{
    "paletzDynamicsMicroconflictsUncertainty2017": "The dynamics of micro - con\ufb02icts and uncertainty in successful and unsuccessful design teams Susannah B . F . Paletz , Center for Advanced Study of Language , University of Maryland , College Park , MD 20742 , USA Joel Chan , Human - Computer Interaction Institute , Carnegie Mellon University , Pittsburgh , PA 15213 , USA Christian D . Schunn , Department of Psychology and Learning , Research , and Development Center , University of Pittsburgh , Pittsburgh , PA 15260 , USA What di\ufb00erentiates successful from unsuccessful design teams ? Building on new research on design innovation that emphasizes interactions between social and cognitive processes , we investigated a potential distinguishing feature : Successful design teams may harness interpersonal con\ufb02icts ( a social design process ) to mitigate uncertainty ( a cognitive design process ) . We analyzed temporal relationships between brief , expressed interpersonal disagreements and subsequent spoken individual uncertainty in 30 h of conversations of 10 successful and 11 unsuccessful engineering product design teams . We discovered that micro - con\ufb02icts were followed by a relative reduction in uncertainty in successful design teams , whereas uncertainty rose after micro - con\ufb02icts in unsuccessful design teams , suggesting that interactions between con\ufb02ict and uncertainty may be a di\ufb00erentiating factor for design team success . (cid:1) 2017 Elsevier Ltd . All rights reserved . Keywords : teamwork , innovation , problem solving , collaborative design , con\ufb02ict W hat di\ufb00erentiates successful from unsuccessful design teams ? After decades of research on this question , design studies researchers have learned much about the nature of successful design teams . For example , successful design teams make heavy use of mental simulations and analogies ( Ball & Christensen , 2009 ; Christensen & Schunn , 2009 ) ; use design tools and media that are appropriate to the phase of the design process ( e . g . , sketching early on , prototypes later on ; Jang & Schunn , 2012 ) ; and work through consensus to build a robust shared understanding of the design prob - lem ( Agogino , Song , & Hey , 2006 ; Dong , 2005 ; Yang , 2010 ) . Yet , much re - mains to be understood about the complex factors that lead to team design success ( Dinar et al . , 2015 ) . For instance , while critical inquiry d which may include con\ufb02ict d is a foundational part of design education and practice ( Dym , Agogino , Eris , Frey , & Leifer , 2005 ; Oh , Ishizaki , Gross , & Yi - Luen Do , 2013 ; Sch \u20ac on , 1983 ) , theoretical and empirical work on team cognition suggests that con\ufb02ict still needs to be appropriately harnessed such that Corresponding author : Susannah Paletz sbfpaletz @ gmail . com www . elsevier . com / locate / destud 0142 - 694X Design Studies 50 ( 2017 ) 39 e 69 http : / / dx . doi . org / 10 . 1016 / j . destud . 2017 . 02 . 002 39 (cid:1) 2017 Elsevier Ltd . All rights reserved . relationship con\ufb02ict is minimized , open - minded discussion is maximized , and the bene\ufb01ts of disagreement can occur ( Jehn , 1997 ; Tjosvold , Wong , & Chen , 2014 ) . Yet , we know little about the preconditions , attendant processes , and mechanisms that make these desirable outcomes possible . An emerging research area in design team innovation emphasizes interactions between social and cognitive processes ( Paletz & Schunn , 2010 ) . The key asser - tion of this perspective is that understanding how social and cognitive pro - cesses are intertwined could help improve our understanding of how design innovation truly occurs , and thus improve interventions designed to improve design team performance ( e . g . , because social dynamics might alter how cogni - tive interventions are perceived , or vice versa ) . This perspective also has the potential to yield fresh insights into pathways to design team success . This social - cognitive perspective is motivated by numerous prior \ufb01ndings of com - plex interactions between social and cognitive processes in teams . For example , simple social phenomena like turn - taking can shape individual mem - ory retrieval dynamics ( Nijstad & Stroebe , 2006 ) . In addition , dissent from a minority opinion holder can trigger a broader information search in other team members , whereas dissent from a majority of team members biases infor - mation search in favor of the dissenting opinion ( Nemeth & Rogers , 1996 ) . In this paper , we investigate how the interplay between disagreement ( a social process ) and individual team members\u2019 uncertainty ( a cognitive process ) could help di\ufb00erentiate successful and unsuccessful design teams . Speci\ufb01cally , we discover that , in successful teams , open expression of disagreements helps to reduce individual uncertainty ( a desirable e\ufb00ect in the design process ) , whereas in unsuccessful teams , similar expressions of disagreement elevate uncertainty levels . Although both disagreement and uncertainty are natural to design teams , harnessing disagreement to resolve uncertainty may be advantageous , if not necessary . 1 Background 1 . 1 Intra - team con\ufb02ict and micro - con\ufb02icts in design Con\ufb02ict has been studied at intra - personal , intrateam , interteam , and national levels , in design teams and between countries ( e . g . , De Dreu & Gelfand , 2008 ; Ozkaramanli , Desmet , & Ozcan , 2016 ) . We focus on intrateam con\ufb02ict as be - tween individuals within the same design team . For this study , we de\ufb01ne con\ufb02ict to be when one team member explicitly opposes or contradicts statements or plans proposed by another team member . Thus , we focus on con\ufb02ict as disagree - ment , which is inherent to problem - solving conversations , regardless of its nega - tive a\ufb00ect , intensity , or directness ( Paletz , Schunn , & Kim , 2011 ; Weingart , Behfar , Bendersky , Todorova , & Jehn , 2015 ) . Such disagreements within a team can arise from di\ufb00erences in values , needs , interests , opinions , goals , or 40 Design Studies Vol 50 No . C May 2017 objectives ( Barki & Hartwick , 2004 ) . This de\ufb01nition draws from the psychology literature , rather than from rhetoric or argumentation theory , which examines di\ufb00erent types of arguments as discourse ( e . g . , Stumpf & McDonnell , 2002 ) . Most importantly , while team con\ufb02ict can be a barrier to design success , con\ufb02ict - ing needs and objectives can also inspire design and promote creativity ( Miron - Spektor , Gino , & Argote , 2011 ; Ozkaramanli et al . , 2016 ) . This paper draws from the organizational behavior , organizational psychol - ogy , and social psychology literature on teams , with the understanding that those research domains have only rarely examined design teams ( see below ) . To better understand the relationship between con\ufb02ict and team performance , social / organizational con\ufb02ict researchers ( De Dreu & Weingart , 2003 ; Jehn , 1995 , 1997 ; Paletz et al . , 2011 ) have found it important to distinguish between task con\ufb02ict ( disagreements about the task ) ; process con\ufb02ict ( disagreements about how to go about doing the task , including scheduling or priorities ) ; and relationship con\ufb02ict ( disagreements about values and personal issues ) . Process and relationship con\ufb02ict are considered problematic , disruptive , and dysfunctional ( e . g . , Jehn , 1997 ; De Dreu & Weingart , 2003 ; De Wit , Greer , & Jehn , 2012 ) . Task con\ufb02ict , in moderate levels and when unrelated to rela - tionship con\ufb02ict , may be positively related to team performance ( De Wit et al . , 2012 ; De Wit , Jehn , & Scheppers , 2013 ; Farh , Lee , & Farh , 2010 ) . Another useful distinction is the length of a con\ufb02ict and how it is measured . Much prior research on con\ufb02ict has examined it via retrospective self - report surveys , and that kind of data has generally found that con\ufb02icts are negatively related to performance ( e . g . , De Dreu & Weingart , 2003 ) . By contrast , obser - vational research on brief con\ufb02icts in - the - moment ( micro - con\ufb02icts ) suggests immediate , sometimes positive relationships with subsequent cognitive pro - cesses ( e . g . , Chiu , 2008a ; Paletz , Schunn , & Kim , 2013 ) . Thus , the often nega - tive portrayal of con\ufb02ict in the organizational behavior literature may be due to how con\ufb02ict is conceptualized and measured , such that brief disagreements are generally healthy , unless they snowball into long , salient con\ufb02icts ( Paletz et al . , 2011 ) . In the aggregate , these micro - process relationships between brief disagreements and cognition may have a positive impact on overall design team performance . The give - and - take of speci\ufb01c , brief disagreements in design and other creative , collaborative settings may not only be normal , but desir - able . Still , understanding the speci\ufb01c social - cognitive processes that are related to these disagreements could not only give insight into why they might be desirable , but also enable teams to negotiate potential tradeo\ufb00s with other as - pects of team performance ( e . g . , team cohesion early in the lifecycle of a team ) . 1 . 2 Uncertainty in design Successful problem solving in real - world , complex domains , such as engi - neering design , relies on e\ufb00ectively detecting and resolving uncertainty Con\ufb02ict and uncertainty in design 41 ( Ball & Christensen , 2009 ; Chan , Paletz , & Schunn , 2012 ; Christensen & Schunn , 2009 ; Schunn & Trafton , 2012 ) . Signi\ufb01cant e\ufb00ort is spent detect - ing , diagnosing , and resolving uncertainty ( Chan et al . , 2012 ; Downey & Slocum , 1975 ; Kahneman & Tversky , 1982 ; Schunn & Trafton , 2012 ) . Psy - chological uncertainty , speci\ufb01cally , is when an individual perceives infor - mation to be incomplete , missing , or vague , regardless of whether or not it is objectively uncertain ( Schunn , 2010 ; Windschitl & Wells , 1996 ) . De - signers can be uncertain not only about how to solve the various problems , but also about what the underlying problems truly are and about what they know ( Ball , Onarheim , & Christensen , 2010 ) . Analogy and mental simula - tion can be used as strategies to resolve uncertainty ( Ball & Christensen , 2009 ; Chan et al . , 2012 ) , and / or co - occur with greater uncertainty ( Christensen & Ball , 2016a ) , and uncertainty can lead to attentive returns to the topic to resolve the uncertain issues later ( Christensen & Ball , 2016b ) . On the other hand , a recent study found that certainty triggered immediate creative reasoning and information elaboration ( Christensen & Ball , 2016b ) . We draw a strong contrast between uncertainty and con\ufb02ict in our conceptu - alizations to avoid circularities : We focus on uncertainty as mental states within individual team members , and con\ufb02ict as interactions between team members . All combinations of uncertainty and con\ufb02ict can occur within the same team ; for example , con\ufb02ict can occur between team members who are currently not uncertain ( e . g . , strongly felt opposing plans ) and uncertainty can occur within team members without con\ufb02ict ( e . g . , individuals are uncer - tain about the same things ) . Although uncertainty is challenging for design teams , having uncertainty is not a sign itself of dysfunction ( Kirschenbaum , Trafton , Schunn , & Trickett , 2014 ) : Indeed , uncertainty is inherent to the ill - structured nature of design problems ( Goel & Pirolli , 1992 ; Tracey & Hutchinson , 2016 ) . Nevertheless , uncertainty that persists over time , unmitigated or unad - dressed , can harm design outcomes . For example , heightened uncertainty can increase a bias against creative ideas ( Mueller , Melwani , & Goncalo , 2012 ) . Also , prolonged uncertainty can increase psychological strain ( Bordia , Hogman , Jones , Gallois , & Callan , 2004 ; Bordia , Hunt , Paulsen , Tourish , & DiFonzo , 2004 ) , potentially reducing team members\u2019 e\ufb00ectiveness in contributing to the team\u2019s problem solving . Thus , devising and deploying appropriate strategies for dealing with uncertainty is central to the design process . Such strategies may include acknowledging it and taking it into account ( Lipshitz & Strauss , 1997 ) , or reducing it by collect - ing additional information , making assumptions and / or analogies , and problem solving ( Berlyne , 1962 ; Chan et al . , 2012 ; Lipshitz & Strauss , 1997 ) . 42 Design Studies Vol 50 No . C May 2017 1 . 3 Research questions about uncertainty and con\ufb02ict in design teams Both team disagreement and individual uncertainty rise and fall naturally over the course of conversations ( Ball & Christensen , 2009 ; Chan et al . , 2012 ; Christensen & Ball , 2016b ; Paletz , Chan , & Schunn , 2016 ; Paletz et al . , 2011 ; McDonnell , 2012 ) . Prior research on interactions between social and cognitive processes has often found important correlations between temporal dynamics of social / cognitive processes ( e . g . , con\ufb02icts sparking analogies dur - ing conversation ; Paletz et al . , 2013 ) . In this work , we hypothesize that tempo - ral dynamics , or patterns of interactions over time , of disagreement and uncertainty can distinguish successful from unsuccessful design teams . The context of design may also in\ufb02uence the e\ufb00ects of con\ufb02ict and uncertainty on each other and on team performance . The broader intrateam con\ufb02ict liter - ature examines a large range of laboratory and natural teams , from executive teams to production teams , including creative teams ( e . g . , De Dreu & Weingart , 2003 ; De Wit et al . , 2012 ) . Design teams , however , have speci\ufb01c fea - tures that may help focus our research questions . First , design teams experi - ence greater uncertainty than many other types of teams , as exhibited by their frequent need to engage in problem \ufb01nding and choose which approach is best to solve their problems ( Mehalik & Schunn , 2006 ) . Second , it is there - fore normal for design teams to tolerate , but also engage with and resolve un - certainty ( Beheshti , 1993 ) . Third , because the goal of design teams is creation , critical thinking and disagreement are valued and taught as part of the social - ization of design students ( Dym et al . , 2005 ) . Even given these norms , however , not all design teams succeed . Successful design teams are likely better at managing and harnessing task con\ufb02ict . Con - straints can be raised in an open manner , thereby increasing the likelihood that disagreement will be productive ( Tjosvold et al . , 2014 ) . One mechanism by which disagreement can lead to success is via the free expression and discussion of di\ufb00erent / opposing views , which encourages information sharing ( De Dreu & West , 2001 ; Tjosvold et al . , 2014 ) . Similarly , dissent arising from minority opinions can lead to greater information search ( Nemeth , 1986 ; Nemeth & Rogers , 1996 ) . Mild disagreement can promote greater information acquisi - tion ( Todorova , Bear , & Weingart , 2014 ) . These dynamics of participation and information sharing have been shown to be vital to team success ( Mesmer - Magnus & DeChurch , 2009 ) . Importantly , information can decrease problem - solving uncertainty ( Lipshitz & Strauss , 1997 ) . Thus , because suc - cessful teams are likely to have e\ufb00ective resources and con\ufb02ict resolution pro - cesses that enable shared understanding ( Kleinsmann & Valkenburg , 2008 ) , con\ufb02ict in those teams should not be dysfunctional , and uncertainty should be productively resolved . We therefore expect that more successful design teams should generally experience a decrease in uncertainty after a Con\ufb02ict and uncertainty in design 43 disagreement . In other words , temporally speaking , we expect disagreements in such teams to be followed by decreases in individual uncertainty . Even as all design teams encounter con\ufb02ict , not all manage it productively . Less successful design teams may have more di\ufb03culty in drawing insights from and resolving task con\ufb02ict . Task con\ufb02ict can mix with relationship con - \ufb02ict , damaging performance ( De Wit et al . , 2012 ) . Minority opinion disagree - ment in teams has been associated with lower con\ufb01dence in team decisions , suggesting that lopsided con\ufb02ict within a team may increase initial uncertainty and the questioning of previously held opinions ( Schulz - Hardt , Jochims , & Frey , 2002 ) d potentially exacerbating and getting in the way of resolving ex - isting levels of uncertainty that are inherent in design teams . Poorly managed disagreement may raise individual uncertainty in the moment that could then remain unresolved and need to be revisited again and again ( Christensen & Ball , 2016b ) ; individuals may use tentative speech to soften a response to a disagreement ( McDonnell , 2012 ) . Thus , because less successful teams are likely to lack e\ufb00ective knowledge resources or con\ufb02ict resolution strategies , con\ufb02ict in such teams should raise uncertainty that is not productively resolved , which may then raise overall levels of uncertainty . From a design perspective , con\ufb02ict could lead to additional uncertainty for less negative reasons . Disagreements reveal di\ufb00erences , ambiguities , or errors that are then noticed by others ( Dama & Dunbar , 1996 ) . Ill - de\ufb01ned problems necessitate the uncovering of design requirements ( Ball et al . , 2010 ; Simon , 1973 ) , such that a team tasked with a poorly - de\ufb01ned problem could disagree about the nature of the problem and its solutions , leading to the revelation of greater uncertainty . Temporally speaking , we expect disagreements in such teams to be followed by increases in individual uncertainty . Motivated by these extant \ufb01ndings , in this paper , we investigate whether there is an interaction between team success and con\ufb02ict , such that micro - con\ufb02icts decrease uncertainty in successful teams , but increase uncertainty in unsuccessful teams . This study explores this question with a dynamic , time - lagged , behav - ioral observation approach using a large dataset of natural problem solving in design teams . Given the prior di\ufb00erences found for task versus process con - \ufb02ict , we examine these relationships separately by these di\ufb00erent types of disagreement . 2 Methods 2 . 1 Research context This study examined engineering student design teams working in a product realization course at a large research university in the Mid - Atlantic United 44 Design Studies Vol 50 No . C May 2017 States . Our overall sample consisted of 57 teams across seven semester - long implementations of the course . In each semester , multidisciplinary student teams , primarily from various dis - ciplines in engineering ( e . g . , mechanical , electrical , industrial , chemical , bioen - gineering ) , took products from concept to functional prototype . Each team was on a di\ufb00erent project from a variety of product domains , such as diaper redesigns , solar - powered water heating systems , and Radio Frequency Identi - \ufb01cation ( RFID ) personnel badge systems . Each project received up to $ 2500 from an external sponsor for design / prototyping e\ufb00orts and had industry , graduate student , and / or faculty mentors . Team members were paid $ 250 to complete background surveys , submit pre - sentations and \ufb01nal documents to the researchers for analysis , and do the bulk of their ( team ) work in a specially prepared room with a table and chairs , a computer with engineering - related software ( e . g . , CAD software ) , and other useful technology . The students were audio e video recorded whenever they used these rooms . Informed consent was obtained from the participants . 2 . 2 Sample Because of the extremely time - consuming nature of observational coding , we sampled 21 teams to be coded for uncertainty and con\ufb02ict , choosing teams that were clearly high vs . low success , focused on hardware projects ( i . e . , excluding unique project types , like software design ) , and regularly used the room set up for video data collection . The 21 teams ranged from three to \ufb01ve members ( M \u00bc 4 , Median \u00bc 4 , SD \u00bc 0 . 63 ) for a total of 84 undergraduate students . Team meeting clips were chosen for transcription based on having no more than 5 min of o\ufb00 - task talk within them and meeting a minimum level of audio and video ( transcribable ) quality . A total of over 30 h of talk was transcribed into utterances ( clauses or thought statements , Chi , 1997 ) and coded for the variables of interest for a total of 38 445 total utterances , 35 148 of which were coded as on - task ( ICC \u00bc 0 . 79 for triple coded sets , average kappa \u00bc 0 . 72 for double coded sets ) . On - task talk included anything relevant to their product or class , including process discussions ( e . g . , shopping trips for materials ) . When a team meeting had a stretch of o\ufb00 - task conversation of over 5 min ( e . g . , about football ) , the video clip was broken up , resulting in 59 video clips ranging from 5 min to 1 h 40 min long ( M \u00bc 31 min , SD \u00bc 23 min ) . Attendees ranged from 2 to 7 individuals with the rare higher numbers because of occasional mentor visits ( M \u00bc 3 . 7 , Median \u00bc 4 . 0 , SD \u00bc 1 . 0 ) . At the team level , 24 % were all male , 29 % of the teams had 50 % females , and the rest had females but in a minority proportion ( proportion female of the team M \u00bc 28 % ) . Con\ufb02ict and uncertainty in design 45 2 . 3 Measures To reveal the patterns of quick , brief social and cognitive process interactions in design teams , this paper draws on recent research that studies interpersonal con\ufb02ict communication behavior within team interactions ( e . g . , Paletz et al . , 2011 ; Poole & Dobosh , 2010 ) . It is within conversations that design team members do the work of joint problem solving and develop group cognition . When studying brief processes , observational methods can tease apart their interplay as they unfold over time ( e . g . , Chiu , 2008a ; Gottman & Notarius , 2000 ; Kau\ufb00eld & Lehmann - Willenbrock , 2012 ; Weingart , 1997 ) . The three primary measures for this study were design team success / outcome , micro - con\ufb02icts and their types , and spoken individual uncertainty . Both un - certainty and micro - con\ufb02icts were 100 % double - coded by two di\ufb00erent sets of paired independent coders who were blind to the research questions of this study , the results of the other variable\u2019s coding , and the team\u2019s success score . Thus , uncertainty and con\ufb02ict could co - occur or not , in any combina - tion , and were not exclusive of each other . All coding disagreements were resolved via consensus . 2 . 3 . 1 Design success Building on professional design practice ( Otto & Wood , 2000 ; Ullman , 2002 ; Ulrich & Eppinger , 2008 ) , we took a requirements - based approach to measuring team design success . Customer needs , given by design brief docu - ments and initial meetings with sponsors , were translated into speci\ufb01c design requirements ( e . g . , safety , cost , ease of use , e\ufb03ciency ) and weighted by impor - tance to the overall product ( using a 1 to 5 importance scale , 5 \u00bc most important ) . Each team\u2019s design requirements were evaluated by a course instructor most familiar with the project and knowledge domain . The instructor rated the de - gree to which the project requirements were met ( e . g . , 0 \u00bc Did not come close to meeting requirement ; 1 \u00bc Fell just short of meeting requirement ; 2 \u00bc Met requirement , but did not exceed signi\ufb01cantly ; 3 \u00bc Signi\ufb01cantly exceeded the requirement ) . While satis\ufb01cing a requirement threshold is important , there is usually a preference for exceeding requirements . For example , while a car manufacturer may obtain an adequate pro\ufb01t if production cost stays below $ 12 000 , an engineering team that produces a car that costs $ 8000 would be even more desirable . All course instructors had extensive experience in product realization , including numerous patents , startup company experience , and in - dustry consulting , as well as relevant content knowledge ( e . g . , materials sci - ence , electrical engineering ) , and were blind to the questions of this study . The \ufb01nal success measure for a given team was calculated by the ratio of the earned success score ( sum of requirement ratings multiplied by their 46 Design Studies Vol 50 No . C May 2017 importance weights ) and the maximum possible success score ( sum of require - ments\u2019 maximum possible ratings multiplied by their importance weights ) , normalized ( multiplied by 100 ) to yield a score on a 0 to 100 scale ( Goncher , Chan , Schunn , & Lovell , 2012 ) . Because of our sampling strategy , our teams fell into a bimodal rather than a normal distribution . We thus used the team\u2019s scores to create a dichotomous variable , 0 \u00bc low success ( success score < 69 , 11 teams ) , and 1 \u00bc high success ( success score > 79 , 10 teams ) . As a validation of our success measure , there was a signi\ufb01cant overall relation - ship between success score and whether or not the teams\u2019 products were sub - mitted for patent or implemented in some fashion by the sponsor , r pb \u00bc 0 . 50 , p < . 001 . In our sample , 7 of the 10 high success teams\u2019 products were submit - ted for patent or implemented , in contrast to only 3 of the 11 low success teams . Information on patenting / implementation was not available for 2 of the teams in our sample ( 1 high and 1 low success ) . We use the ratings as the success measure rather than whether the product was used at the company or patent submissions , because complex factors outside the local project ( e . g . , di\ufb00erences in company intellectual property management approaches , time / budget constraints ) played a large role in determining immediate use at the company or decision to \ufb01le a patent . 2 . 3 . 2 Micro - con\ufb02icts We coded for micro - con\ufb02icts at the utterance level by adapting a pre - existing coding scheme ( i . e . , if this utterance includes con\ufb02ict , the utterance was marked as 1 , and if not , it was marked as 0 ; Paletz et al . , 2011 , 2013 ) . Micro - con\ufb02icts were identi\ufb01ed when a speaker explicitly , whether via tone and / or words , disagreed with something said earlier in the video clip . Simply stating a potentially controversial viewpoint was not su\ufb03cient . Constraints , which limit the search for solutions ( Ball et al . , 2010 ) , can be ( but need not al - ways be ) raised via disagreements ( Paletz , Sumer , & Miron - Spektor , 2016 ) . Coders both read the transcript and listened to / watched the audio e video recording simultaneously to improve reliability ( average con\ufb02ict event kappa of 0 . 70 across coder pairs ) . The micro - con\ufb02icts were coded by utterance for type of con\ufb02ict , using cate - gories typically used in the con\ufb02ict literature ( e . g . , Jehn , 1997 ) : task micro - con\ufb02ict utterances were directly related to the engineering product realization task , including choosing materials , design , and testing ; process micro - con\ufb02ict utterances were about scheduling , communicating their \ufb01ndings ( e . g . , creating presentations , how to dress for presentations ) , how to go about doing work , assigning tasks , prioritization , and who said what to whom ; \ufb01nally , relation - ship micro - con\ufb02icts were about values , being critical of others\u2019 personality and style , and personal likes / dislikes . Because of the potential co - occurrence of these three con\ufb02ict types in real - world settings , their presence / absence Con\ufb02ict and uncertainty in design 47 were coded separately ( task kappa \u00bc 0 . 87 ; process kappa \u00bc 0 . 61 , and relation - ship kappa \u00bc 0 . 65 ) . Micro - con\ufb02ict events were clusters of utterances de\ufb01ned as relating to a very speci\ufb01c topic . For instance , the exchange \u2018It won\u2019t \ufb01t , \u2019 \u2018Well , just make it bigger , \u2019 \u2018Well , I\u2019m dumb , I can\u2019t , \u2019 \u2018You can\u2019t make it bigger ? \u2019 \u2018No , \u2019 \u2018What the hell , [ name ] , you suck\u2019 is actually two micro - con\ufb02icts : one that is predom - inantly process ( the students were discussing whether and how to change the font size on a Powerpoint slide , rather than the content ) , and then a micro - con\ufb02ict event immediately following that is predominantly relationship con - \ufb02ict , where the \ufb01rst speaker\u2019s competence is called into question ( for another example , see Table 1 ) . To test the alternate supposition that successful teams were simply better at resolving con\ufb02icts , micro - con\ufb02ict events were also coded as to whether they were resolved immediately , within 25 utterances of the last micro - con\ufb02ict utter - ance coded ( kappa \u00bc 0 . 79 ) . Coders judged the micro - con\ufb02ict event as resolved if the disagreeing parties came to an agreement , whether by one person acqui - escing to the other\u2019s ( or others\u2019 ) opinion or the parties coming to a compromise . Twenty - \ufb01ve utterances were chosen because it was often not possible to tell if the micro - con\ufb02ict was resolved after that point , as micro - con\ufb02icts blended into each other over time , and some may have been resolved by the participants at untranscribed later times ( e . g . , via email or during other meetings ) . 2 . 3 . 3 Uncertainty Although psychological uncertainty is an internal state , cognitive scientists have frequently operationalized uncertainty via coding individual utterances within design and science team conversations , picking up \u2018hedge words\u2019 ( words that typically accompany and communicate uncertainty ) such as \u2018maybe\u2019 and \u2018possibly\u2019 ( e . g . , Ball & Christensen , 2009 ; Ball et al . , 2010 ; Chan et al . , 2012 ; Trickett , Trafton , Saner , & Schunn , 2007 ) . These kinds of words demonstrate behaviorally , viaverbalexpression , whenanindividualisperceiving uncertainty . This codingschemehad beenpreviously validatedusing convergent and discrim - inantmethods , such ascomparingspeech togesture based uncertainty measures , and distinguishing uncertainty from approximations ( Schunn , 2010 ) . Leveraging the relatively simple and stable predictive nature of hedge words as cues ( in contrast to the much more variable and complex nature of expressed disagreements ) , we employed a two - step semi - automated approach . First , we trained a Support Vector Machine classi\ufb01er ( a supervised machine learning al - gorithm ) on an initial set of human - coded utterances , and used the classi\ufb01er to identify utterances that might contain uncertainty ( for technical details , see Luo , Litman , & Chan , 2013 ) . Second , two trained humans coders assessed for uncertainty those utterances identi\ufb01ed by the algorithm as likely to include 48 Design Studies Vol 50 No . C May 2017 uncertainty . This process saves time and coder e\ufb00ort given the massive dataset , and it also helps to avoid coder drift and decreased validity and reliability due to fatigue . Inter - rater reliability was high ( Cohen\u2019s kappa \u00bc 0 . 80 ) . Table 1 pre - sents an example of both con\ufb02ict and uncertainty from a team that was tasked with creating a product that could be attached to a light bulb socket and indi - cate the power usage , with the end goal of informing consumers of their energy consumption . Table 1 Segmented transcript with uncertainty ( italics ) and micro - conflict ( bold ) codes Speaker Utterance Con\ufb02ict type 4 I thought yeah 4 It would be cool to put some little hydroelectric power generators 3 Yeah 4 You know those little mini ones 4 Right 3 That would be uh 3 And those , these like micro - generators 3 They\u2019d be enough to power maybe a light Task , onset 3 But not a battery Task 3 You know like the overall cost of electricity would not be decreased Task 3 I don\u2019t think Task 3 But then again these are like Task 3 They seem to be high cost solutions Task 3 These generators you know Task , end 4 Mmhmm 1 Yeah 3 Putting in generators 2 Yeah with that liquid water thing 2 Just that some people have their tanks underground too ? 4 Mmhmm right 2 So you wouldn\u2019t be able to do that 3 But perhaps 3 Perhaps that instead of marketing it as an individual solution 3 Maybe it could be a solution for an entire group of people , community , building 2 I think that was one of the things 2 That was just probably the only good thing on this paper 3 Yeah ? 2 Yeah it was 2 It said something about Note . From Team 20070106 . Speakers are numbered based on who spoke \ufb01rst in a transcript . Blocks indicated by section divides . Con\ufb02ict types are the dominant type at that utterance . Con\ufb02ict and uncertainty in design 49 2 . 4 Analyses 2 . 4 . 1 Blocks , video clips , and teams as three - level nested data As the goal was to examine the e\ufb00ects of disagreement on subsequent uncer - tainty , we needed a data structure that supported time - lagged analyses . We created segmented blocks centered on the incidents of the independent vari - able , con\ufb02ict . This event - centering strategy ensures that the predictor block ( i . e . , con\ufb02ict events ) would be relatively homogenous in its content , and the dependent blocks would be standardized in length ( Paletz et al . , 2013 ) . Blocks were created by ( 1 ) identifying the con\ufb02ict event ( or contiguous con\ufb02ict events ) , ( 2 ) segmenting the 10 utterances before and after each con\ufb02ict event as two additional blocks , and ( 3 ) breaking up the rest of the clips into succes - sive blocks of 10 utterances , each ending at the 10th utterance , the end or beginning of the clip , or with the next con\ufb02ict . Ten - utterance long blocks were chosen because initial descriptive visualizations suggested that was the best grain size for observing \ufb02uctuations in uncertainty . Previous analyses of temporal patterns of uncertainty in other datasets suggested that uncertainty in design teams tends to show signi\ufb01cant rises and declines in windows of 15 e 30 utterances ( e . g . , Chan et al . , 2012 ; Christensen & Schunn , 2009 ) . As un - certainty involves some problem solving prior to uncertainty resolution , we conducted analyses at two lags : both 0 e 10 utterances before the dependent variable ( \ufb01rst block , Lag1 ) and 11 e 20 utterances before the dependent vari - able ( second block , Lag2 ) . The number of utterances rather than time was the unit of analysis because the focus of the study was on expressed behavior . That noted , 20 utterances also represented roughly 1 min after the end of a micro - con\ufb02ict , so the \ufb01rst lagged block was at about 0 e 30 s after the con\ufb02ict , and the second at about 30 e 60 s . 2 . 4 . 2 Statistical analyses Most regression analyses assume that ( 1 ) data points are independent , and ( 2 ) that the dependent variable is continuous and normally distributed . Both of these assumptions were violated with these data . First , these data are nested d - blocks within video clips within teams d such that data points were not inde - pendent . Second , the main dependent variable was count data , which is skewed ( mostly zeroes , then ones , etc . ) . Thus , we employed statistical analyses that appropriately accounted for both these features : three - level hierarchical linear modeling , time - lagged , variable exposure overdispersed Poisson models using HLM7 ( 7 . 01 ) software ( Raudenbush , Bryk , & Congdon , 2013 ) . 1 We used three - level hierarchical linear modeling to account for the inherent depen - dence of having blocks ( Level 1 ) within video clips ( Level 2 ) within teams ( Level 3 ) and to handle unequal cell sizes ( Raudenbush & Bryk , 2002 ) . 2 The segmentation method resulted in 3969 blocks at Level 1 , 59 video clips at Level 2 , and 21 teams at Level 3 . Overdispersed Poisson models were necessary 50 Design Studies Vol 50 No . C May 2017 because the main dependent variable was count data ( number of uncertainty utterances in a block ) , and its variance ( 2 . 0 ) was greater than its mean ( 1 . 5 ; s 2 was 1 . 1 instead of 1 ) . Variable exposure Poisson accounts for the number of utterances being di\ufb00erent for di\ufb00erent blocks because the block creation process drew from naturalistic data ( e . g . , endings of clips , beginnings of new con\ufb02ict events ) . 3 We employed time - lagged analyses , such that uncertainty in one 10 - utterance block was predicted by the presence / absence of con\ufb02ict , or a particular con - \ufb02ict , at either Lag1 ( 1 e 10 utterances before ) or Lag2 ( 11 e 20 utterances before ) . In creating the models to test our research questions , the time - lagged con\ufb02ict to uncertainty tests were at Level 1 ( blocks ) . The binary low / high success variable was a main e\ufb00ect at the team level ( Level 3 ) and a moder - ator at the team level on the block level con\ufb02ict variables ( interaction between Level 3 and Level 1 variables ; Figure 1 ) . Before examining our independent variables , we tested for the signi\ufb01cance of several covariates that were plausibly connected to uncertainty or con\ufb02ict and therefore could be potential third - variable confounds . At Level 1 , we tested the percentage of on - task talk that was about the product itself ( rather than , say , presentations or planning ) ; the average number of words per utterance in that block , in order to control for underlying number of words available to be coded ; and the number of speakers in the block , in order to control for uncer - tainty due to more people participating in the conversation . At Level 2 , we tested the number of uncertainty on - task utterances in the overall video clip , the number of people at the meeting , the presence of a non - team member ( e . g . , a client or faculty advisor ) , and two gender composition vectors ( to Figure 1 Three - level data structure of lagged blocks , clips , teams , and team success Con\ufb02ict and uncertainty in design 51 account for the three types of clip gender compositions ) . At Level 3 , we tested the binary success variable , two team - level gender composition vectors , and the size of the team . Gender composition for both the conversational clip and the team was examined because of the known positive relationship be - tween the percent of women in a team and collective intelligence ( Woolley , Chabris , Pentland , Hashmi , & Malone , 2010 ) , which could then be associated with team success or other team processes . We \ufb01rst tested the potential cova - riates at Level 1 , then those signi\ufb01cant at Level 1 with Level 2 , then those sig - ni\ufb01cant at Levels 1 and 2 with Level 3 . Potential covariates had to be statistically signi\ufb01cant to be kept in the \ufb01nal covariate model . Two Level 1 var - iables passed these tests : the average number of words per utterance and the number of speakers in the block , both of which were positively associated with the presence of uncertainty . In keeping with standard practices , Lag2 ef - fects always controlled for Lag1 parallel variables , and interaction e\ufb00ects al - ways controlled for their respective main e\ufb00ects , regardless of statistical signi\ufb01cance . When a signi\ufb01cant interaction e\ufb00ect with team success occurred , we also tested for the main e\ufb00ects of micro - con\ufb02icts on subsequent uncertainty for the data - set broken up into low and high success teams ( i . e . , we conducted simple e\ufb00ects analyses to better understand signi\ufb01cant interactions ) . 3 Results We \ufb01rst describe the frequency of design team uncertainty in the data set . We then present the relationship between di\ufb00erent kinds of design team con\ufb02ict on subsequent uncertainty levels , as well as the moderation tests with team success . 3 . 1 Frequency of uncertainty and micro - con\ufb02icts At the block level , uncertainty occurred relatively frequently , with 70 % of the 3970 blocks having uncertainty , but with a wide variety in amount of uncer - tainty in a block : range 0 e 14 . In contrast , con\ufb02ict was less common : only 6 % of the 3970 blocks had con\ufb02ict . Of the 244 blocks coded as having con\ufb02ict in them , 65 % had at least one utterance of task con\ufb02ict , 57 % had at least one utterance of process con\ufb02ict , and 12 % had at least one utterance of relation - ship con\ufb02ict . The relationship micro - con\ufb02ict frequency ( 1 % of all blocks ) was insu\ufb03cient for use in these analyses . Importantly , as noted in subsequent sections , there were no di\ufb00erences by team success on the simple prevalence of uncertainty or overall con\ufb02ict , task con\ufb02ict , or process con\ufb02ict . Further , more successful teams were not simply better at resolving con\ufb02icts : At the team level , the proportion of quickly resolved con\ufb02ict events out of all on - task con\ufb02icts were the same between the more successful ( M \u00bc 0 . 54 , SD \u00bc 0 . 27 , n \u00bc 10 ) and less successful 52 Design Studies Vol 50 No . C May 2017 ( M \u00bc 0 . 71 , SD \u00bc 0 . 22 , n \u00bc 11 ) teams , t ( 19 ) \u00bc 1 . 57 , p \u00bc 0 . 13 . 4 With uncertainty as the dependent variable using the analysis technique noted previously , and the covariates of number of words per utterance and number of speakers , there was a non - signi\ufb01cant relationship trending in the negative direction between concurrent uncertainty and con\ufb02ict , with a large con\ufb01dence interval . This analysis suggests what the coding procedure implied d that there was no reli - able relationship between uncertainty and con\ufb02ict in the same block . 5 3 . 2 Relationship between con\ufb02ict , team success type , and uncertainty 3 . 2 . 1 Overall con\ufb02ict First , we tested whether con\ufb02ict one block or two blocks earlier ( Lag1 and Lag2 ) had signi\ufb01cant main e\ufb00ects on subsequent uncertainty and then whether they had signi\ufb01cant interactions with the team success variable . Team success alone was not related to uncertainty ( i . e . , both low and high success teams had similar rates of uncertainty ) . The interaction of con\ufb02ict at Lag1 with team suc - cess was not signi\ufb01cant . However , at Lag2 , there was a signi\ufb01cant interaction : high success teams had less uncertainty following micro - con\ufb02icts , compared with low success teams , which had relatively more uncertainty following micro - con\ufb02icts ( see Table 2 , Figure 2 ) . 3 . 2 . 2 Micro - con\ufb02ict by subtype We then repeated these analyses for task and process con\ufb02ict . These micro - con\ufb02ict types were not mutually exclusive , but were negatively correlated with each other : Blocks with task con\ufb02icts were less likely to have process con - \ufb02icts , X 2 ( 1 , 239 ) \u00bc 49 . 82 , p < 0 . 001 , phi \u00bc (cid:2) 0 . 46 ( at Lag2 ) . 7 Thus , testing them Table 2 Full model of micro - conflict and team design success on uncertainty Variable Event rate ratio ( 95 % con\ufb01dence ratio ) Intercept , g 000 a 0 . 17 ( 0 . 13 , 0 . 22 ) * * * Average words per utterance 1 . 07 ( 1 . 05 , 1 . 09 ) * * * Speakers in block 1 . 26 ( 1 . 20 , 1 . 32 ) * * * Con\ufb02ict Lag1 0 . 95 ( 0 . 87 , 1 . 05 ) Con\ufb02ict Lag2 1 . 09 ( 0 . 99 , 1 . 20 ) \u00fe Team Success ( L - 3 ) a 1 . 06 ( 0 . 83 , 1 . 34 ) Con\ufb02ict Lag2 (cid:3) team success 0 . 78 ( 0 . 65 , 0 . 95 ) * Low success teams Con\ufb02ict Lag 2 b 1 . 09 ( 0 . 94 , 1 . 26 ) High success teams Con\ufb02ict Lag 2 c 0 . 85 ( 0 . 72 , 1 . 01 ) \u00fe Note . \u00fe 0 . 10 > p > 0 . 05 ; * p < 0 . 05 ; * * * p < 0 . 001 . Degrees of freedom ( df ) \u00bc 3736 unless otherwise noted . a df \u00bc 23 . b df \u00bc 1951 . c df \u00bc 1782 . Con\ufb02ict and uncertainty in design 53 separately gives a more complete view and helps us determine whether the ef - fect is the same for di\ufb00erent kinds of con\ufb02ict . For task con\ufb02ict , we found that , consistent with the analyses of overall con - \ufb02ict , there were no signi\ufb01cant main e\ufb00ects of team success ( p > 0 . 60 ) , task con - \ufb02ict Lag1 ( p > 0 . 80 ) , and task con\ufb02ict Lag2 ( p > 0 . 40 ) on subsequent uncertainty , controlling for the signi\ufb01cant covariates . However , task con\ufb02ict Lag2 had a signi\ufb01cant interaction with the team success variable ( p < 0 . 05 ) , again showing decreasing uncertainty after task con\ufb02ict in high success teams but increasing uncertainty after task con\ufb02ict in low success teams ( Figure 3A ) . 8 Process con\ufb02ict Lag2 has the same interaction e\ufb00ect with team success ( Table 3 ) with growing uncertainty in low success teams and decreasing uncer - tainty in high success teams following process con\ufb02ict two blocks earlier ( Table 3 , bottom rows , Figure 3B ) . Tables 4 and 5 are illustrative examples of two micro - con\ufb02ict events and the uncertainty that followed them . One team , which did not fare well on its \ufb01nal success score , was tasked with investigating how to put a radio - frequency iden - ti\ufb01cation chip and antenna ( RFID technology ) in drill bits used in manufacturing . As they disagreed about the optimal size they needed , that dis - cussion raised other elements of their task that were uncertain ( Table 4 ) . In the second example , a successful team was working on a presentation and had a minor task disagreement about what the summary should include , with one team member noting it should be a summary of the whole paper . They quickly moved on to discussing in colorful terms how they would do that , with very little uncertainty ( Table 5 ) . While Speaker 1 raised the initial micro - con\ufb02ict , the little uncertainty being communicated seems to be hedges against seeming too domi - nant , rather than uncertainty about any underlying issues . Figure 2 Predicted number of uncertain utterances by team success and time - lagged mi - cro - con\ufb02ict , controlling for covariates 6 54 Design Studies Vol 50 No . C May 2017 Figure 3 Predicted number of uncertain utterances by team success and time - lagged ( A ) task and ( B ) Process micro - con\ufb02ict , controlling for covariates Table 3 Process micro - conflict and team success on uncertainty Variable Event rate ratio ( 95 % con\ufb01dence ratio ) Intercept , g 000 a 0 . 17 ( 0 . 13 , 0 . 22 ) * * * Average words per utterance 1 . 07 ( 1 . 04 , 1 . 09 ) * * * Speakers in block 1 . 26 ( 1 . 20 , 1 . 32 ) * * * Process con\ufb02ict Lag1 0 . 91 ( 0 . 77 , 1 . 08 ) Process con\ufb02ict Lag2 1 . 16 ( 1 . 03 , 1 . 30 ) * Team success ( L - 3 ) a 1 . 06 ( 0 . 83 , 1 . 34 ) Process con\ufb02ict Lag2 (cid:3) team success 0 . 62 ( 0 . 50 , 0 . 78 ) * * * Low success teams Process con\ufb02ict Lag 2 b 1 . 17 ( 0 . 99 , 1 . 38 ) \u00fe High success teams Process con\ufb02ict Lag 2 c 0 . 72 ( 0 . 54 , 0 . 95 ) * Note . \u00fe 0 . 10 > p > 0 . 05 ; * p < 0 . 05 ; * * * p < 0 . 001 . Degrees of freedom ( df ) \u00bc 3736 unless otherwise noted . a df \u00bc 23 . b df \u00bc 1951 . c df \u00bc 1782 . Con\ufb02ict and uncertainty in design 55 Table 4 Example of unsuccessful team ( ID 20070304 ) micro - conflict and uncertainty Speaker Utterance Con\ufb02ict type 2 Let\u2019s \ufb01nd the biggest small drill bit we have , 2 if they\u2019re all the same , 2 I saw some 3 e 3 . 4s 2 3 . 4 might do it , three seven 3 These are 3 This is 3 . 4 here 2 This is too big 1 Well 1 big is better for us Task , onset 1 I think Task 1 Because that will give us more apoxy Task 2 Well 2 This < missing word > 3 Yeah 1 I think we may need < missing word > 2 But I don\u2019t think this is the appropriate size Task 3 Right , Task 3 That looks a little big Task 4 Little bit too big Task , end 3 Would we be allowed to use these drill bits ? 3 Because they are somehow not right 3 < missing word > leave this to join together 2 I could have sworn we had some 3 . 4s 3 We do , 3 They\u2019re right here 2 Okay 3 And there is another set 3 Right here 2 We only got 6 , 2 Better not to mess them up 3 I don\u2019t know 3 What < missing word > 2 Oh 2 These are too small , 2 These are what , 4 . 2 . 2 Yeah 1 Yeah 56 Design Studies Vol 50 No . C May 2017 These examples are merely illustrative and should not replace either an in - depth qualitative analyses or our statistical analysis . Indeed , for some kinds of psychological phenomena , explicit cognitive processes play a strong role , whereas for other psychological phenomena , implicit emotional / memory Table 5 Example of successful team ( ID 20070301 ) micro - conflict and uncertainty Speaker Utterance Con\ufb02ict type 2 So the conclusion should just be a summary of this kind of stuff a 1 I think it\u2019s supposed to be a summary of the whole paper Task , onset 1 Err the whole project Task , end 3 You could do it like that 1 So it\u2019s just a bit more general 1 You don\u2019t have to 1 You should go into detail 2 I know I can bullshit 2 But I want 2 I kinda want 2 Or meant to be 2 Uh , you know 2 Bullshit that everyone else has piled on 1 Well you can kinda , 1 all you have to do for the conclusion is restate the summary 2 Great 1 Just reword the < missing word > 2 Cool 3 So 1 That\u2019s pretty much 1 I guess what you should do tomorrow 1 I don\u2019t know 1 I wasn\u2019t really paying attention too much Note . Bold \u00bc con\ufb02ict ; italic \u00bc uncertainty . Speakers are numbered based on who spoke \ufb01rst in a transcript . Blocks indi - cated by section divides . Con\ufb02ict types are the dominant type at that utterance . a Speaker is referring to cost calculation formulas Table 4 ( continued ) Speaker Utterance Con\ufb02ict type 1 I think that one we have would actually be 2 The 7 ? 1 Yeah Note . Bold \u00bc con\ufb02ict ; italic \u00bc uncertainty . Speakers are numbered based on who spoke \ufb01rst in a transcript . Blocks indi - cated by section divides . Con\ufb02ict types are the dominant type at that utterance . Con\ufb02ict and uncertainty in design 57 processes play a role ( e . g . , priming e\ufb00ects or blocking e\ufb00ects in memory ) . In the case of implicit processes , statistical evidence , as we used , may be more revealing than qualitative analyses of speci\ufb01c speech . These examples show there is no simple explicit mechanism underlying the connection between un - certainty and con\ufb02ict . 3 . 2 . 3 Uncertainty and subsequent micro - con\ufb02icts Because our data are observational , the temporal relationship between con\ufb02ict and uncertainty could be explained by \u2018third variable confounds : \u2019 For example , it is possible that increases in task di\ufb03culty raise both con\ufb02icts and uncertainty ( instead of con\ufb02ict per se leading to decreases / increases in un - certainty ) . To rule out this alternative explanation , we also tested the e\ufb00ects of uncertainty ( Lag1 and Lag2 ) on subsequent task and process con\ufb02icts , as well as the interaction of lagged uncertainty with type of team on con\ufb02ict . If there are third variable confounds , the reverse temporal relationships should be found for the same con\ufb02ict type - to - uncertainty connections obtained above . We used HLM 7 . 01 to conduct three - level hierarchical logistic regression an - alyses , with dichotomous con\ufb02ict variables as the dependent variables ( pres - ence of con\ufb02ict or not ) and percent of uncertainty utterances in the block ( Lag1 and Lag2 ) as the independent variable . For all models , we controlled for three signi\ufb01cant Level - 1 covariates : number of speakers in the block , total number of words , and percent of on - task talk that was speci\ufb01cally about the hardware product itself . The gender variables ( at either the meeting or team level ) and eventual team success were not signi\ufb01cantly related to any subtype of con\ufb02ict . As with the earlier analyses , we controlled for potential main e\ufb00ects and Lag1 e\ufb00ects . Controlling for those three covariates , we found no signi\ufb01cant e\ufb00ects on over - all con\ufb02ict for either uncertainty Lag1 ( p > 0 . 19 ) , uncertainty Lag2 ( p > 0 . 10 ) , team success ( p > 0 . 80 ) , or , most importantly , the interactions between team success and either uncertainty lag variable ( Lag1 , p > 0 . 32 , Lag2 , p > 0 . 15 ) . There was a signi\ufb01cant positive main e\ufb00ect of uncertainty Lag2 on task con - \ufb02ict , odds ratio \u00bc 1 . 69 ( 1 . 01 , 2 . 82 ) , df \u00bc 3717 , p \u00bc 0 . 044 . However , uncertainty Lag1 ( p > 0 . 12 ) , team success ( p > 0 . 34 ) , and , most importantly , the two in - teractions were not signi\ufb01cant ( success (cid:3) uncertainty Lag1 , p > 0 . 64 , X Lag2 , p > 0 . 55 ) . 9 Of particular relevance , there was a marginally signi\ufb01cant interaction between uncertainty Lag2 and the team success variable on process con\ufb02ict , odds ratio \u00bc 0 . 076 ( 0 . 006 , 1 . 035 ) , df \u00bc 3717 , p \u00bc 0 . 053 , when controlling for the three covariates , uncertainty Lag1 ( p > 0 . 25 ) , uncertainty Lag2 ( p > 0 . 17 ) , team success ( p > 0 . 64 ) , and the interaction between success and uncertainty Lag1 ( p > 0 . 22 ) . While this interaction echoed the interaction between process 58 Design Studies Vol 50 No . C May 2017 con\ufb02ict Lag2 and success on subsequent uncertainty , this interaction was much weaker in size and marginal in signi\ufb01cance . Conducting analyses sepa - rately for low and high success teams , there were no signi\ufb01cant relationships for uncertainty ( Lag2 ) on process con\ufb02ict ( p s > 0 . 18 ) . 3 . 2 . 4 Summary of \ufb01ndings The general pattern of our \ufb01ndings was that , for high success teams , some micro - con\ufb02icts ( particularly process con\ufb02icts ) were negatively related to the prevalence of subsequent uncertainty , whereas for low success teams , some micro - con\ufb02icts were positively related to uncertainty in subsequent blocks . This interaction e\ufb00ect was signi\ufb01cant for both task and process micro - con\ufb02ict subtypes , even though these two types of con\ufb02ict were negatively related to each other . Importantly , low and high success teams did not di\ufb00er overall on their levels of uncertainty or any type of con\ufb02ict ; all teams had reg - ular uncertainty and con\ufb02ict . Instead , the di\ufb00erences by team success were spe - ci\ufb01cally in the interrelationship between prior ( Lag2 ) con\ufb02ict and subsequent uncertainty . Further , there was no strong parallel interaction e\ufb00ect between uncertainty and success on subsequent con\ufb02ict . Task con\ufb02ict was signi\ufb01cantly preceded by high uncertainty ( Lag2 ) as a main e\ufb00ect , but this e\ufb00ect occurred for both types of teams . Thus , the observed temporal relationships between con\ufb02ict and uncertainty cannot be explained by general co - occurrence patterns between con\ufb02ict and uncertainty ( e . g . , triggered by a third variable confound ) . 4 Discussion Using novel methods , this study provides evidence that successful and unsuc - cessful design team problem solving can be di\ufb00erentiated in terms of the nature of the temporal relationship between micro - con\ufb02icts between team members and subsequent psychological uncertainty . We observed that , in successful design teams , micro - con\ufb02icts decreased uncertainty , while in unsuccessful teams , micro - con\ufb02icts comparatively increased uncertainty . These correla - tional \ufb01ndings suggest a new hypothesized distinguishing factor between suc - cessful or unsuccessful problem solving : Successful teams may be achieving better problem solving by harnessing their disagreements to reduce individual uncertainty in the moment . One might wonder whether our \ufb01ndings are simply another story about how successful teams are better at resolving con\ufb02icts . If this were true , the lower success teams should have displayed higher levels of con\ufb02ict ( or lower levels of uncertainty ) overall , compared to the higher success teams . However , this was not the case : The lower success teams did not have signi\ufb01cantly more con - \ufb02ict ( of any type ) or more uncertainty overall than the highly successful teams , including the proportion of quickly resolved con\ufb02icts . Thus , our \ufb01ndings cannot be explained by general di\ufb00erences in propensity to con\ufb02ict or uncer - tainty in low vs . high success teams . Con\ufb02ict and uncertainty in design 59 4 . 1 Implications This study makes a contribution to the design studies literature by enriching our understanding of successful design teams . This study deliberately exam - ines brief , minor con\ufb02icts , and \ufb01nds that for successful teams , micro - con\ufb02icts can be potentially bene\ufb01cial by decreasing uncertainty . Uncertainty , particularly brief uncertainty , is not always negative , but it needs to be managed for problem solving to be successful ( Schunn & Trafton , 2012 ) . For example , problem solving in ill - de\ufb01ned domains ( a common situation in real - world engineering design problem solving ) is inherently imbued with un - certainty related to problem detection , \ufb01nding , and structuring ( Goel & Pirolli , 1992 ; Runco , 1994 ) , activities that occur early in the problem solving process ( Mumford , Reiter - Palmon , & Redmond , 1994 ; Runco , 1994 ) . In the course of solving a problem , this uncertainty is assessed , managed , and reduced . This study suggests that teams d speci\ufb01cally , design teams , which encounter a lot of uncertainty relative to other types of teams in organizations d may success - fully manage uncertainty by iteratively raising and dealing with alternative conceptions in a way that helpfully reduces uncertainty in the moment , leaving time to address more issues that arise in problem solving . This study also builds on other team cognition research . Within conversations , team members engage in problem solving , but also develop , challenge , and maintain underlying shared mental models . In particular , design teams are trained to engage in critical inquiry in order to e\ufb00ectively \ufb01nd and solve prob - lems ( Dym et al . , 2005 ; Sch \u20ac on , 1983 ) . Shared mental models occur when indi - vidual mental models are similar with regards to tasks and teamwork ( e . g . , Burke , Stagl , Salas , Peirce , & Kendall , 2006 ; Johnson - Laird , 1980 ; Mathieu , He\ufb00ner , Goodwin , Salas , & Cannon - Bowers , 2000 ) . Both the accuracy and congruence of shared mental models have been identi\ufb01ed as important aspects of group cognition and positively predictive of team performance ( e . g . , Burke et al . , 2006 ; DeChurch & Mesmer - Magnus , 2010 ) . Disagreements may arise due to di\ufb00erences in shared mental models ( Bearman , Paletz , Orasanu , & Thomas , 2010 ; Cronin & Weingart , 2007 ; Paletz & Schunn , 2010 ) . Our \ufb01nd - ings suggest that for more successful teams , disagreements may be used in un - covering and settling these di\ufb00erences . As Dong ( 2005 , p . 458 ) noted , \u201c Team communication re\ufb02ects the formation of mutual expectations and shared under - standings \u201d ( italics in the original ) . However , even with greater tolerance for disagreement , some design teams may get bogged down in irreconcilable is - sues . Di\ufb00erences in shared mental models d for instance , regarding an ill - de\ufb01ned problem d may be too fundamental for some teams to overcome , and / or disagreement may uncover more underlying di\ufb00erences than the team is capable of handling , leaving a state of increased uncertainty . In partic - ular , the examples noted previously illustrate that unsuccessful teams may be grappling with more challenging issues when they are disagreeing . 60 Design Studies Vol 50 No . C May 2017 While our data uncovered temporal relationships between con\ufb02ict and subse - quent uncertainty , they raise further questions about how disagreements might help design teams successfully manage individual uncertainty during complex problem solving . Our analyses of the reverse relationship generally did not support the simplest explanation : that disagreement and uncertainty simply co - occur . Thus , our data spur more theoretically interesting questions about the causal proximal and distal relationships between con\ufb02ict and uncertainty . For example , are successful teams directly better at harnessing con\ufb02ict for resolving uncertainty , such that they resolve uncertainty by increasing the in - formation exchange during con\ufb02icts ( i . e . , information exchange is a mediator of the disagreement - uncertainty relationship ) ? Does con\ufb02ict - driven uncer - tainty resolution better enable teams to achieve a shared understanding by the end of their project ( e . g . , Agogino et al . , 2006 ) ? Were the less successful teams reacting to disagreement with more tentativeness in order to manage and accommodate con\ufb02ict ( McDonnell , 2012 ) ? In addition , the e\ufb00ect for Lag2 rather than Lag1 suggests an incubation period is necessary : It may sim - ply take time , after a disagreement , for individuals to hear and encode the disagreement and for that to a\ufb00ect uncertainty within individuals , before the change in uncertainty is spoken aloud . One possibility is that con\ufb02ict was associated with problem - solving techniques that then impacted uncertainty . This study aligns with emerging research \ufb01nd - ings from the design \ufb01eld and elsewhere suggesting that , despite general nega - tive e\ufb00ects of self - reported con\ufb02ict on team performance ( e . g . , De Dreu & Weingart , 2003 ) , brief or mild disagreements can , in the right circumstances , have positive outcomes on creativity and cognition ( Chiu , 2008a , 2008b ; Goncalo , Polman , & Maslach , 2010 ; Miron - Spektor , Efrat - Treister , Rafaeli , & Schwartz - Cohen , 2011 ; Paletz et al . , 2013 ; Todorova et al . , 2014 ) . For example , in a study on the Mars scientist conversations , process micro - con\ufb02icts were found to increase the likelihood of analogies soon after ( Paletz et al . , 2013 ) . Analogies are useful for team success and creativity ( Dunbar , 1995 , 1997 ) . Indeed , in successful multidisciplinary expert teams , problem - related analogies reduced uncertainty ( Chan et al . , 2012 ) , and analo - gies and mental simulations can be used to reduce uncertainty ( Ball & Christensen , 2009 ) . These studies suggest that analogy and / or mental simula - tion , or other creative cognitive processes , may be possible mediators of the relationship between con\ufb02ict and uncertainty . In addition , in our data , high uncertainty signi\ufb01cantly preceded task con\ufb02ict with no di\ufb00erence across types of teams , suggesting that uncertainty may serve as a prompt for , or in some indirect way lead to , task - related con\ufb02ict and not for the other subtypes . Finally , these \ufb01ndings have potential practical implications for design teams . Design teams , including student design teams like our sample and unlike some teams in other organizational settings , may be encouraged to disagree as part of their divergent search processes ( Dym et al . , 2005 ) , such that brief Con\ufb02ict and uncertainty in design 61 con\ufb02icts may not be a source of anxiety . Indeed , their existence did not di\ufb00er between successful and unsuccessful teams . However , harnessing con\ufb02ict and gaining the best insights can still be a challenge : Team leaders should be aware of and channel what the teams do with that con\ufb02ict in terms of improving cognition and uncovering or resolving uncertainty . In addition , in non - student teams , managers should also be aware of organizational - and project - level forces that can interfere with or enable shared understanding , as well ( Kleinsmann & Valkenburg , 2008 ) . Taking a broader view , this study echoes that teams with greater di\ufb03culties in achieving shared mental models , such as a mental model of the problems themselves , may simply struggle more than other teams , and those other teams have an easier pathway to success . 4 . 2 Limitations and future work Successful and unsuccessful design teams may di\ufb00er on many features , including their initial resources ( material , intellectual , etc . ) ; the di\ufb03culty of their tasks ; their team processes ( e . g . , Kau\ufb00eld & Lehmann - Willenbrock , 2012 ; Post , 2012 ) ; their use of collaboration tools ( e . g . , Jang & Schunn , 2012 ) ; and their ability to leverage external support . This study includes many controls to rule out reverse causality or these possible third variable ex - planations of the obtained relationships . For example , there were no signi\ufb01cant e\ufb00ects for team or meeting gender composition or team size on the prevalence of uncertainty . Importantly , unsuccessful teams had essentially equal levels of un - certainty and disagreement as successful teams , removing the possible alterna - tive explanation that the successful teams simply had less uncertainty or were less likely to disagree d though not what topics they were uncertain or disagreed about . Nevertheless , the issue of causality is not fully resolved . For example , even though successful teams are likely to reduce uncertainty following micro - con\ufb02icts , this relationship may not have in\ufb02uenced their design success . Future research can also unpack the consequences of these patterns between disagreement and uncertainty on other related processes , such as information search , creativity , mental simulation , and analogy ( e . g . , Ball & Christensen , 2009 ; Chiu , 2008a ; Christensen & Ball , 2016a , 2016b ; Paletz et al . , 2013 ) . Given our examples , it is unlikely that any one of these is the only mediator , but that some combination of implicit and explicit processes exist . This study also examined only one design context in one particular culture , given the highly time - consuming data collection and coding methodology . Other domains with di\ufb00erent types of disciplinary and demographic diversity and levels of expertise ( novice / experts ) should be tested , such as professional design teams across di\ufb00erent countries . Comparisons across contexts and cul - tures will be important for identifying the boundary conditions for observed relationships . One interesting comparison is the relationship between the pre - sent \ufb01ndings and a previous investigation of the relationship between con\ufb02ict and uncertainty in conversations of an expert science team working on the 62 Design Studies Vol 50 No . C May 2017 Mars Exploration Rover mission ( Paletz et al . , 2016 ) . Even though that team overall was highly successful , disagreements about the planning and interpre - tation ( of the results ) of the rover\u2019s scienti\ufb01c expeditions led to an increase in expressed uncertainty . Of course , there are large di\ufb00erences in tasks and par - ticipants across the two studies : the previous study examined members of a professional , long - duration team engaged in scienti\ufb01c discovery ; here , we examine more recently - formed , student engineering teams . One possibility is that the present \ufb01ndings do not apply to the success of teams in general , only to teams engaged in design ( vs . scienti\ufb01c discovery , for example ) . Alter - natively , even successful teams may uncover more uncertainty through con\ufb02ict in extremely novel tasks . Future studies that examine more teams across more contexts may be able to identify contextual factors that may moderate the rela - tionship between team success and con\ufb02ict - uncertainty dynamics . 4 . 3 Conclusion In conclusion , we report our discovery of a temporal relationship between brief disagreements ( a social design process ) and subsequent psychological un - certainty ( a cognitive design process ) that di\ufb00ers for design teams with di\ufb00erent problem - solving success and failure . This \ufb01nding was particularly strong for process con\ufb02icts . This discovery sheds light on the nature of success - ful design in teams and spurs additional theoretical questions about the com - plex factors that underlie team design success . Acknowledgments This research was supported by the United States National Science Founda - tion ( NSF ) Science of Science and Innovation Policy Program Grants # SBE - 1064083 and # SBE - 0830210 to the \ufb01rst author when she was at the Uni - versity of Pittsburgh and to the \ufb01rst and third author , respectively , and NSF # SBE - 0738071 to the third author . We are grateful to Roni Reiter - Palmon and Kevin Soo for comments on previous version of this manuscript , Kevin Kim and Mathilda du Toit for statistical advice , and to Carmela Rizzo for research management support . The authors wish to thank Mike Lovell and Kevin Topolski for data collection ; Andrea Goncher and Howard Kuhn for assisting in the team success variable ; Justin Krill , Jake Volpe , Janeen Bost , Jessica Varone , Andrew Bergman , Stephen Burstein , and Shauna Barnes for transcribing ; Carl Fenderson , Abby Pollick , Stephen Burstein , LaNee Jones , Justin Krill , Allison Haley , Sam Rynearson , Courtney Buchanan , Anna Poul - ton , Kevin Gaitonde , Matt Flounders , Claire James , and Jooyang Jang for assistance in coding ; and Kyle Bogue and Megan Loftus for assistance in data management and validation . Notes 1 . Unless otherwise mentioned , we present the test of the unit - speci\ufb01c model with robust standard errors . HLM 7 . 01 uses penalized quasi - likelihood ( PQL ) estimation . Con\ufb02ict and uncertainty in design 63 2 . This choice was statistically justi\ufb01ed . An HLM null model tests the dependent variable without any predictor variables using chi square estimation to determine whether there are signi\ufb01cant higher - level components . The null model of the uncertainty dependent variable showed that there was signi\ufb01cant variance at both Level 2 ( video clip level ) , Tau beta \u00bc 0 . 06 , X 2 ( 38 ) \u00bc 210 . 41 , p < 0 . 001 , and Level 3 ( team level ) , Tau pi \u00bc 0 . 04 , X 2 ( 20 ) \u00bc 58 . 65 , p < 0 . 001 ( ICC \u00bc 4 . 8 % for Level 2 , ICC \u00bc 3 . 1 % for Level 3 ) . In other words , there was a need to use a multilevel model . 3 . As an o\ufb00set variable , we used ( 1 \u00fe ln [ number of utterances ] ) because the o\ufb00set variable cannot equal zero . 4 . Very similar \ufb01ndings were found when controlling for the number of utterances tran - scribed for each group and using a univariate generalized linear model . 5 . Event rate ratio \u00bc 0 . 83 , ( 0 . 69 , 1 . 01 ) , B \u00bc (cid:2) 0 . 18 , SE \u00bc 0 . 10 , df \u00bc 3870 , p \u00bc 0 . 061 . 6 . Graphs created using unstandardized , uncentered variables because the independent var - iables are dummy coded . Covariate e\ufb00ects evaluated at their mean values ( 2 . 51 for num - ber of speakers in block , and 5 . 54 for average words per utterance ) . 7 . Testing only Lag2 con\ufb02ict blocks so as not to arti\ufb01cially in\ufb02ate relationships between types of con\ufb02ict . 8 . Event ratio \u00bc 0 . 80 ( 0 . 63 , 0 . 999 ) , df \u00bc 3736 , p \u00bc 0 . 049 . There was a marginal negative e\ufb00ect for task con\ufb02ict Lag2 on subsequent uncertainty for high success teams , event ratio \u00bc 0 . 84 ( 0 . 70 , 1 . 02 ) , df \u00bc 1782 , p \u00bc 0 . 073 , but no signi\ufb01cant relationship between the variables in low success teams ( p > 0 . 56 ) . 9 . This positive e\ufb00ect of uncertainty Lag2 on task con\ufb02ict was not simply an artifact of con - trolling for non - signi\ufb01cant interactions with success : The e\ufb00ect of uncertainty Lag2 on subsequent task con\ufb02ict remained when the model was only the three signi\ufb01cant Level - 1 covariates and uncertainty Lag1 ( p > 0 . 37 ) and Lag2 , odds ratio \u00bc 1 . 50 ( 1 . 03 , 2 . 19 ) , df \u00bc 3719 , p \u00bc 0 . 034 . References Agogino , A . , Song , S . , & Hey , J . ( 2006 ) . Triangulation of indicators of successful student design teams . International Journal of Engineering Education , 22 , 617 e 625 . Ball , L . J . , & Christensen , B . T . ( 2009 ) . Analogical reasoning and mental simula - tion in design : Two strategies linked to uncertainty resolution . Design Studies , 30 ( 2 ) , 169 e 186 . Ball , L . J . , Onarheim , B . , & Christensen , B . T . ( 2010 ) . Design requirements , epistemic uncertainty and solution development strategies in software design . Design Studies , 13 , 567 e 589 . Barki , H . , & Hartwick , J . ( 2004 ) . Conceptualizing the construct of interpersonal con\ufb02ict . International Journal of Con\ufb02ict Management , 15 , 216 e 244 . Bearman , C . R . , Paletz , S . B . F . , Orasanu , J . , & Thomas , M . J . W . ( 2010 ) . The breakdown of coordinated decision making in distributed systems . Human Factors , 52 , 173 e 188 . Beheshti , R . ( 1993 ) . Design decisions and uncertainty . Design Studies , 14 ( 1 ) , 85 e 95 . Berlyne , D . E . ( 1962 ) . Uncertainty and epistemic curiosity . British Journal of Psy - chology , 53 , 27 e 34 . Bordia , P . , Hobman , E . , Jones , E . , Gallois , C . , & Callan , V . J . ( 2004 ) . Uncertainty during organizational change : Types , consequences , and management strate - gies . Journal of Business and Psychology , 18 , 507 e 532 . Bordia , P . , Hunt , E . , Paulsen , N . , Tourish , D . , & DiFonzo , N . ( 2004 ) . Uncer - tainty during organizational change : Is it all about control ? European Journal of Work and Organizational Psychology , 13 ( 3 ) , 345 e 365 . 64 Design Studies Vol 50 No . C May 2017 Burke , C . S . , Stagl , K . C . , Salas , E . , Peirce , L . , & Kendall , D . ( 2006 ) . Understand - ing team adaptation : A conceptual analysis and model . Journal of Applied Psy - chology , 91 , 1189 e 1207 . Chan , J . , Paletz , S . B . F . , & Schunn , C . D . ( 2012 ) . Analogy as a strategy for sup - porting complex problem solving under uncertainty . Memory and Cognition , 40 , 1352 e 1365 . Chi , M . ( 1997 ) . Quantifying qualitative analyses of verbal data : A practical guide . Journal of the Learning Sciences , 6 , 271 e 315 . Chiu , M . M . ( 2008a ) . E\ufb00ects of argumentation on group micro - creativity ; Statis - tical discourse analyses of algebra students\u2019 collaborative problem - solving . Contemporary Educational Psychology , 33 , 382 e 402 . Chiu , M . M . ( 2008b ) . Flowing toward correct contributions during group prob - lem solving : A statistical discourse analysis . Journal of the Learning Sciences , 17 , 415 e 463 . Christensen , B . T . , & Ball , L . J . ( 2016a ) . Creative analogy use in a heterogeneous design team : The pervasive role of background domain knowledge . Design Studies , 46 , 38 e 58 . Christensen , B . T . , & Ball , L . J . ( 2016b ) . Fluctuating epistemic uncertainty in a design team as a metacognitive driver for creative cognitive processes . In Co - penhagen : Paper presented at Design Thinking Research Symposium 11 . 13 e 15 November . Christensen , B . T . , & Schunn , C . D . ( 2009 ) . The role and impact of mental simu - lation in design . Applied Cognitive Psychology , 23 , 327 e 344 . http : / / dx . doi . org / 10 . 1002 / acp . 1464 . Cronin , M . A . , & Weingart , L . R . ( 2007 ) . Representational gaps , information processing , and con\ufb02ict in functionally diverse teams . Academy of Management Review , 32 , 761 e 773 . http : / / dx . doi . org / 10 . 5465 / AMR . 2007 . 25275511 . Dama , M . , & Dunbar , K . ( 1996 ) . Distributed reasoning : An analysis of where so - cial and cognitive worlds fuse . In Proceedings of the Eighteenth Annual Confer - ence of the Cognitive Science Society ( pp . 166 e 170 ) . DeChurch , L . A . , & Mesmer - Magnus , J . R . ( 2010 ) . The cognitive underpinnings of e\ufb00ective teamwork : A meta - analysis . Journal of Applied Psychology , 95 , 32 e 53 . De Dreu , C . K . W . , & Gelfand , M . ( 2008 ) . Con\ufb02ict in the workplace : Sources , functions , and dynamics across multiple levels of analysis . In C . K . W . De Dreu , & M . Gelfand ( Eds . ) , The Psychology of Con\ufb02ict and Con\ufb02ict Manage - ment in Organizations ( pp . 3 e 54 ) . New York : Erlbaum . De Dreu , C . K . W . , & Weingart , L . R . ( 2003 ) . Task versus relationship con\ufb02ict , team performance , and team member satisfaction : A meta - analysis . Journal of Applied Psychology , 88 , 741 e 749 . De Dreu , C . K . W . , & West , M . A . ( 2001 ) . Minority dissent and team innovation : The importance of participation in decision making . Journal of Applied Psy - chology , 86 , 1191 e 1201 . De Wit , F . R . C . , Greer , L . L . , & Jehn , K . A . ( 2012 ) . The paradox of intragroup con\ufb02ict : A meta - analysis . Journal of Applied Psychology , 97 , 360 e 390 . De Wit , F . R . C . , Jehn , K . A . , & Scheepers , D . ( 2013 ) . Task con\ufb02ict , information processing , and decision - making : The damaging e\ufb00ect of relationship con\ufb02ict . Organizational Behavior and Human Decision Processes , 122 , 177 e 189 . Dinar , M . , Shah , J . J . , Cagan , J . , Leifer , L . , Linsey , J . , Smith , S . M . , et al . ( 2015 ) . Empirical studies of designer thinking : Past , present , and future . Journal of Mechanical Design , 137 . 021101 - 1 - 021101 - 021101 - 13 . Dong , A . ( 2005 ) . The latent semantic approach to studying design team commu - nication . Design Studies , 2 , 445 e 461 . Con\ufb02ict and uncertainty in design 65 Downey , H . K . , & Slocum , J . W . ( 1975 ) . Uncertainty : Measures , research , and sources of variation . Academy of Management Journal , 18 , 562 e 578 . Dunbar , K . ( 1995 ) . How scientists really reason : Scienti\ufb01c reasoning in real - world laboratories . In R . J . Sternberg , & J . E . Davidson ( Eds . ) , The Nature of Insight ( pp . 365 e 395 ) . Cambridge , MA : The MIT Press . Dunbar , K . ( 1997 ) . How scientists think : On - line creativity and conceptual change in science . In T . B . Ward , S . M . Smith , & J . Vaid ( Eds . ) , Creative Thought : An Investigation of Conceptual Structures and Processes ( pp . 461 e 493 ) . Washing - ton : American Psychological Association . Dym , C . L . , Agogino , A . M . , Eris , O . , Frey , D . D . , & Leifer , L . J . ( 2005 ) . Engi - neering design thinking , teaching , and learning . Journal of Engineering Educa - tion , 94 ( 1 ) , 103 e 120 . Farh , J . - L . , Lee , C . , & Farh , C . I . C . ( 2010 ) . Task con\ufb02ict and team creativity : A question of how much and when . Journal of Applied Psychology , 95 , 1173 e 1180 . Goel , V . , & Pirolli , P . ( 1992 ) . The structure of design problem spaces . Cognitive Science , 16 , 395 e 429 . Goncalo , J . A . , Polman , E . , & Maslach , C . ( 2010 ) . Can con\ufb01dence come too soon ? Collective e\ufb03cacy , con\ufb02ict and group performance over time . Organiza - tional Behavior and Human Decision Processes , 113 , 13 e 24 . Goncher , A . , Chan , J . , Schunn , C . , & Lovell , M . ( 2012 ) . A robust and e\ufb03cient function - focused measure of design innovation for design process - outcome studies . Unpublished manuscript . Gottman , J . M . , & Notarius , C . I . ( 2000 ) . Decade review : Observing marital inter - action . Journal of Marriage and the Family , 62 , 927 e 947 . Jang , J . , & Schunn , C . D . ( 2012 ) . Physical design tools support and hinder inno - vative engineering design . Journal of Mechanical Design , 134 ( 4 ) . 041001 - 1 - 9 . Jehn , K . A . ( 1995 ) . A multimethod examination of the bene\ufb01ts and detriments of intragroup con\ufb02ict . Administrative Science Quarterly , 40 , 256 e 282 . Jehn , K . A . ( 1997 ) . A qualitative analysis of con\ufb02ict types and dimensions in orga - nizational groups . Administrative Science Quarterly , 42 , 530 e 557 . Johnson - Laird , P . N . ( 1980 ) . Mental models in cognitive science . Cognitive Sci - ence , 4 , 71 e 115 . Kahneman , D . , & Tversky , A . ( 1982 ) . Variants of uncertainty . Cognition , 11 , 143 e 157 . Kau\ufb00eld , S . , & Lehmann - Willenbrock , N . ( 2012 ) . Meetings matter : E\ufb00ects of team meetings on team and organizational success . Small Group Research , 43 , 130 e 158 . Kirschenbaum , S . S . , Trafton , J . G . , Schunn , C . D . , & Trickett , S . B . ( 2014 ) . Visu - alizing uncertainty : The impact on performance . Human Factors , 56 ( 3 ) , 509 e 520 . http : / / dx . doi . org / 10 . 1177 / 0018720813498093 . Kleinsmann , M . , & Valkenburg , R . ( 2008 ) . Barriers and enablers for creating shared understanding in co - design projects . Design Studies , 29 , 369 e 386 . Lipshitz , R . , & Strauss , O . ( 1997 ) . Coping with uncertainty : A naturalistic decision - making analysis . Organizational Behavior and Human Decision Pro - cesses , 69 , 149 e 163 . Luo , W . , Litman , D . , & Chan , J . ( 2013 ) . Reducing annotation e\ufb00ort on unbal - anced corpus based on cost matrix . In Paper Presented at the 2013 Conference of the North American Chapter of the Association for Computational Linguis - tics : Human Language Technologies ( NAACL HLT 2013 ) Student Research Workshop , Atlanta , GA . 66 Design Studies Vol 50 No . C May 2017 Mathieu , J . E . , He\ufb00ner , T . S . , Goodwin , G . F . , Salas , E . , & Cannon - Bowers , J . A . ( 2000 ) . The in\ufb02uence of shared mental models on team process and perfor - mance . Journal of Applied Psychology , 85 , 273 e 283 . McDonnell , J . ( 2012 ) . Accommodating disagreement : A study of e\ufb00ective design collaboration . Design Studies , 33 , 44 e 63 . Mehalik , M . M . , & Schunn , C . D . ( 2006 ) . What constitutes good design ? A review of empirical studies of the design process . International Journal of Engineering Education , 22 ( 3 ) , 519 e 532 . Mesmer - Magnus , J . R . , & DeChurch , L . A . ( 2009 ) . Information sharing and team performance : A meta - analysis . Journal of Applied Psychology , 94 , 535 e 546 . Miron - Spektor , E . , Efrat - Treister , D . , Rafaeli , A . , & Schwartz - Cohen , O . ( 2011 ) . Others\u2019 anger makes people work harder not smarter : The e\ufb00ect of observing anger and sarcasm on complex thinking . Journal of Applied Psychology , 96 , 1065 e 1075 . Miron - Spektor , E . , Gino , F . , & Argote , L . ( 2011 ) . Paradoxical frames and crea - tive sparks : Enhancing individual creativity through con\ufb02ict and integration . Organizational Behavior and Human Decision Processes , 116 , 229 e 240 . Mueller , J . S . , Melwani , S . , & Goncalo , J . A . ( 2012 ) . The bias against creativity : Why people desire but reject creative ideas . Psychological Science , 23 , 13 e 17 . http : / / dx . doi . org / 10 . 1177 / 0956797611421018 . Mumford , M . D . , Reiter - Palmon , R . , & Redmond , M . R . ( 1994 ) . Problem con - struction and cognition : Applying problem representations in ill - de\ufb01ned do - mains . In M . A . Runco ( Ed . ) , Problem Finding , Problem Solving , and Creativity ( pp . 3 e 39 ) . Norwood , NJ : Ablex . Nemeth , C . J . ( 1986 ) . Di\ufb00erential contributions of majority and minority in\ufb02u - ence . Psychological Review , 93 , 23 e 32 . Nemeth , C . J . , & Rogers , J . ( 1996 ) . Dissent and the search for information . British Journal of Social Psychology , 35 , 67 e 76 . Nijstad , B . A . , & Stroebe , W . ( 2006 ) . How the group a\ufb00ects the mind : A cognitive model of idea generation in groups . Personality and Social Psychological Re - view , 10 , 186 e 213 . http : / / dx . doi . org / 10 . 1207 / s15327957pspr1003 _ 1 . Oh , Y . , Ishizaki , S . , Gross , M . D . , & Yi - Luen Do , E . ( 2013 ) . A theoretical frame - work of design critiquing in architecture studios . Design Studies , 34 ( 3 ) , 302 e 325 . http : / / dx . doi . org / 10 . 1016 / j . destud . 2012 . 08 . 004 . Otto , K . N . , & Wood , K . L . ( 2000 ) . Product Design : Techniques in Reverse Engi - neering and New Product Development . Upper Saddle River , NJ : Prentice Hall . Ozkaramanli , D . , Desmet , P . M . A . , & Ozcan , E . ( 2016 ) . Beyond resolving di - lemmas : Three design directions for addressing intrapersonal concern con\ufb02icts . Design Studies , 32 , 78 e 91 . Paletz , S . B . F . , Chan , J . , & Schunn , C . D . ( 2016 ) . Uncovering uncertainty through disagreement . Applied Cognitive Psychology . http : / / dx . doi . org / 10 . 1002 / acp . 3213 . Published online 17 Feb 2016 . Paletz , S . B . F . , & Schunn , C . D . ( 2010 ) . A social - cognitive framework of multi - disciplinary team innovation . Topics in Cognitive Science , 2 ( 1 ) , 73 e 95 . Paletz , S . B . F . , Schunn , C . D . , & Kim , K . H . ( 2011 ) . Con\ufb02ict under the micro - scope : Micro - con\ufb02icts in naturalistic team discussions . Negotiation and Con\ufb02ict Management Research , 4 , 314 e 351 . Paletz , S . B . F . , Schunn , C . D . , & Kim , K . H . ( 2013 ) . The interplay of con\ufb02ict and analogy in multidisciplinary teams . Cognition , 126 , 1 e 19 . Paletz , S . B . F . , Sumer , A . , & Miron - Spektor , E . ( 2016 ) . Psychological factors sur - rounding disagreement in multicultural design team meetings . In Paper to be Presented at Design Thinking Research Symposium 11 ( DTRS11 ) , Copenhagen , Denmark . Con\ufb02ict and uncertainty in design 67 Poole , M . S . , & Dobosh , M . ( 2010 ) . Exploring con\ufb02ict management processes in jury deliberations through interaction analysis . Small Group Research , 41 , 408 e 426 . http : / / dx . doi . org / 10 . 1177 / 1046496410366310 . Post , C . ( 2012 ) . Deep - level team composition and innovation : The mediating roles of psychological safety and cooperative learning . Group & Organization Man - agement , 37 , 555 e 588 . http : / / dx . doi . org / 10 . 1177 / 1059601112456289 . Raudenbush , S . W . , & Bryk , A . S . ( 2002 ) . Hierarchical Linear Models : Applica - tions and Data Analysis Methods ( 2nd ed . ) . Newbury Park , CA : Sage . Raudenbush , S . W . , Bryk , A . S . , & Congdon , R . ( 2013 ) . HLM 7 . 01 for Windows [ Computer Software ] . Skokie , IL : Scienti\ufb01c Software International , Inc . Runco , M . A . ( 1994 ) . Problem Finding , Problem Solving , and Creativity . Stam - ford , CT : Ablex . Sch \u20ac on , D . A . ( 1983 ) . The Re\ufb02ective Practitioner . How Professionals Think in Ac - tion . New York , NY : Basic Books . Schulz - Hardt , S . , Jochims , M . , & Frey , D . ( 2002 ) . Productive con\ufb02ict in group de - cision making : Genuine and contrived dissent as strategies to counteract biased information seeking . Organizational Behavior and Human Decision Processes , 88 , 563 e 586 . Schunn , C . D . ( 2010 ) . From uncertainty exact to certainly vague : Epistemic uncer - tainty and approximation in science and engineering problem solving . In B . H . Ross ( Ed . ) , The Psychology of Learning and Motivation , Vol . 53 ( pp . 227 e 252 ) . Burlington : Academic Press . Schunn , C . D . , & Trafton , J . G . ( 2012 ) . The psychology of uncertainty in scienti\ufb01c data analysis . In G . Feist , & M . Gorman ( Eds . ) , Handbook of the Psychology of Science ( pp . 461 e 483 ) . New York , NY : Springer Publishing . Simon , H . A . ( 1973 ) . The structure of ill structured problems . Arti\ufb01cial Intelli - gence , 4 ( 3 e 4 ) , 181 e 201 . Stumpf , S . C . , & McDonnell , J . T . ( 2002 ) . Talking about team framing : Using argumentation to analyse and support experiential learning in early design ep - isodes . Design Studies , 23 , 5 e 23 . Tjosvold , D . , Wong , A . S . H . , & Chen , N . Y . F . ( 2014 ) . Constructively managing con\ufb02icts in organizations . Annual Review of Organizational Psychology and Organizational Behavior , 1 , 545 e 568 . http : / / dx . doi . org / 10 . 1146 / annurev - org - psych - 031413 - 091306 . Todorova , G . , Bear , J . B . , & Weingart , L . R . ( 2014 ) . Can con\ufb02ict be energizing ? A study of task con\ufb02ict , positive emotions , and job satisfaction . Journal of Applied Psychology , 99 , 451 e 467 . http : / / dx . doi . org / 10 . 1037 / a0035134 . Tracey , M . W . , & Hutchinson , A . ( 2016 ) . Uncertainty , re\ufb02ection , and designer identity development . Design Studies , 42 , 86 e 109 . http : / / dx . doi . org / 10 . 1016 / j . destud . 2015 . 10 . 004 . Trickett , S . B . , Trafton , J . G . , Saner , L . D . , & Schunn , C . D . ( 2007 ) . \u201cI don\u2019t know what\u2019s going on there\u201d : The use of spatial transformations to deal with and resolve uncertainty in complex visualizations . In M . C . Lovett , & P . Shah ( Eds . ) , Thinking with Data ( pp . 65 e 86 ) . Mahwah , NJ : Erlbaum . Ullman , D . ( 2002 ) . The Mechanical Design Process ( 3 ed . ) . New York , NY . Ulrich , K . T . , & Eppinger , S . D . ( 2008 ) . Product Design and Development ( 4 ed . ) . New York , NY . Weingart , L . R . ( 1997 ) . How did they do that ? The ways and means of studying group processes . Research in Organizational Behavior , 19 , 189 e 239 . Weingart , L . R . , Behfar , K . J . , Bendersky , C . , Todorova , G . , & Jehn , K . A . ( 2015 ) . The directness and oppositional intensity of con\ufb02ict expression . Acad - emy of Management Review , 40 , 235 e 262 . 68 Design Studies Vol 50 No . C May 2017 Windschitl , P . D . , & Wells , G . L . ( 1996 ) . Measuring psychological uncertainty : Verbal versus numeric methods . Journal of Experimental Psychology : Applied , 2 ( 4 ) , 343 e 364 . Woolley , A . W . , Chabris , C . F . , Pentland , A . , Hashmi , N . , & Malone , T . W . ( 2010 ) . Evidence for a collective intelligence factor in the performance of hu - man groups . Science , 330 , 686 e 688 . Yang , M . C . ( 2010 ) . Consensus and single leader decision - making in teams using structured design methods . Design Studies , 31 , 345 e 362 . Con\ufb02ict and uncertainty in design 69",
    "hopeAcceleratingInnovationAnalogy2017": "Accelerating Innovation Through Analogy Mining Tom Hope The Hebrew University of Jerusalem tom . hope @ mail . huji . ac . il Joel Chan Carnegie Mellon University joelchuc @ cs . cmu . edu Aniket Kittur Carnegie Mellon University nkittur @ cs . cmu . edu Dafna Shahaf The Hebrew University of Jerusalem dshahaf @ cs . huji . ac . il ABSTRACT The availability of large idea repositories ( e . g . , the U . S . patent data - base ) could significantly accelerate innovation and discovery by providing people with inspiration from solutions to analogous prob - lems . However , finding useful analogies in these large , messy , real - world repositories remains a persistent challenge for either human or automated methods . Previous approaches include costly hand - created databases that have high relational structure ( e . g . , predicate calculus representations ) but are very sparse . Simpler machine - learning / information - retrieval similarity metrics can scale to large , natural - language datasets , but struggle to account for structural similarity , which is central to analogy . In this paper we explore the viability and value of learning simpler structural representations , specifically , \u201cproblem schemas\u201d , which specify the purpose of a product and the mechanisms by which it achieves that purpose . Our approach combines crowdsourcing and recurrent neural networks to extract purpose and mechanism vector representations from product descriptions . We demonstrate that these learned vectors allow us to find analogies with higher precision and recall than tra - ditional information - retrieval methods . In an ideation experiment , analogies retrieved by our models significantly increased people\u2019s likelihood of generating creative ideas compared to analogies re - trieved by traditional methods . Our results suggest a promising approach to enabling computational analogy at scale is to learn and leverage weaker structural representations . KEYWORDS Computational analogy ; innovation ; creativity ; product dimensions ; text mining ; text embedding 1 INTRODUCTION The ability to find useful analogies is critical to driving innovation in a variety of domains . Many important discoveries in science were driven by analogies : for example , an analogy between bacteria and slot machines helped Salvador Luria advance the theory of bacterial mutation . Analogical reasoning forms the foundation of law , with the effectiveness of an argument often dependent on its Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page . Copyrights for components of this work owned by others than ACM mustbehonored . Abstractingwithcreditispermitted . Tocopyotherwise , orrepublish , to post on servers or to redistribute to lists , requires prior specific permission and / or a fee . Request permissions from permissions @ acm . org . KDD \u201917 , August 13 - 17 , 2017 , Halifax , NS , Canada \u00a9 2017 ACM . 978 - 1 - 4503 - 4887 - 4 / 17 / 08 . . . $ 15 . 00 DOI : 10 . 1145 / 3097983 . 3098038 legal precedents [ 32 ] . Innovation is often spurred by analogy as well : an analogy to a bicycle allowed the Wright brothers to design a steerable aircraft . Whether architecture , design , technology , art , or mathematics , the ability to find and apply patterns from other domains is fundamental to human achievement [ 9 , 17 , 22 , 26 ] . The explosion of available online data represents an unprece - dented opportunity to find new analogies and accelerate human progress across domains . For example , the US Patent database has full text for more than 9 million patents issued from 1976 to the present . InnoCentive 1 contains more than 40 , 000 business , social , policy , scientific , and technical problems and solutions . Quirky 2 , a company that assists inventors in the development process , has had over 2 million product idea submissions . OpenIDEO 3 receives hundreds of solutions for a variety of social problems . Millions of scientific papers and legal cases are searchable on Google Scholar . We believe these data form a treasure trove of analogies that can accelerate problem solving , innovation and discovery . In a striking recent example , a car mechanic invented a simple device to ease difficult childbirths by drawing an analogy to extracting a cork from a wine bottle , which he discovered in a YouTube video . This award - winning device could save millions of lives , particularly in developing countries . We imagine a future in which people could search through data based on deep analogical similarity rather than simple keywords ; lawyers or legal scholars could find legal precedents sharing similar systems of relations to a contemporary case ; and product or service designers could mine myriad potential solutions to their problem . However , sifting through these massive data sources to find relevant and useful analogies poses a serious challenge for both humans and machines . In humans , memory retrieval is highly sensitive to surface similarity , favoring near , within - domain analogs that share object attributes over far , structurally similar analogs that share object relations [ 13 \u2013 15 , 20 ] . Analogical processing also incurs a heavy cognitive load , taxing working memory when even a few relations are required to be processed at once [ 16 ] . Thus searching through datasets with thousands or millions of items for structurally similar ones may be a daunting prospect . Finding analogies is challenging for machines as well , as it is based on having an understanding of the deep relational similarity between two entities that may be very different in terms of sur - face attributes [ 12 ] . For example , Chrysippus\u2019 analogy between sound waves and water waves required ignoring many different surface features between the two [ 20 ] . Recent advances in data 1 innocentive . com 2 quirky . com 3 OpenIDEO . com KDD 2017 Research Paper KDD\u201917 , August 13 \u2013 17 , 2017 , Halifax , NS , Canada 235 mining and information retrieval include a variety of natural lan - guage techniques that use words , parts of speech or other language feature - based vector representations in order to calculate similarity measures ( see [ 31 ] ) . Examples include word embedding models like Word2Vec [ 24 ] , vector - space models like Latent Semantic Index - ing [ 10 ] , and probabilistic topic modeling approaches like Latent Dirichlet Allocation [ 4 ] . These approaches excel at detecting sur - face similarity , but are often unable to detect similarity between documents whose word distributions are disparate . The problem is especially acute when the source and target domains are different ( for example , bacterial mutation and slot machines ) . Another approach to finding analogies has been to use the struc - tural similarity of sentences or texts , such as using coupled clus - tering for detecting structural correspondence of text [ 5 , 35 ] . How - ever , these approaches typically require rich data sets with clear substructures , whereas most descriptions of problems or ideas in existing online databases are short , sparse , or lack consistent struc - ture . Other current methods focus on very narrow analogy tasks , such as four - term analogy problems ( teacher : student = doctor : ? ) , in particular with short strings ( ABC : ABD = KJI : ? ) [ 19 ] . In contrast , we wish to find analogies in real world data , which involve complex representations and a diverse set of analogical relations . In this paper , we are interested in automatically discovering analogies in large , unstructured data sets . In particular , we focus on a corpus of product innovations . There are two in - sights behind our approach that we believe may make this problem tractable despite its longstanding status as a \u201choly grail\u201d in both cognitive science and AI . First , rather than trying to solve the prob - lem of fully structured analogical reasoning , we instead explore the idea that for retrieving practically useful analogies , we can use weaker structural representations that can be learned and reasoned with at scale ( in other words , there is a tradeoff between the ease of extraction of a structure and its expressivity ) . Specifically , we investigate the weaker structural representation of an idea\u2019s pur - pose and mechanism as a way to find useful analogies . The second insight is that advances in crowdsourcing have made it possible to harvest rich signals of analogical structure that can help machine learning models learn in ways that would not be possible with existing datasets alone . This paper combines these two ideas to contribute a technique for computationally finding analogies from unstructured text datasets that go beyond surface features . At a high level , our approach uses the behavioral traces of crowd workers searching for analogies and identifying the purpose and mechanisms of ideas , then developing machine learning models that develop similarity metrics suited for analogy mining . We demonstrate that learning purpose and mechanism representations allows us to find analogies with higher precision and recall than traditional information - retrieval methods based on TF - IDF , LSA , LDA and GloVe , in challenging noisy set - tings . Furthermore , we use our similarity metrics to automatically find far analogies \u2013 products with high purpose similarity , and low mechanism similarity . In a user study , we show that we are able to \u201cinspire\u201d participants to generate more innovative ideas than alternative baselines , increasing the relative proportion of positively - rated ideas by at least 25 % . 2 LEARNING A REPRESENTATION FOR ANALOGIES 2 . 1 Motivation Much work in computation analogy has focused on fully structured data , often with logic - based representations . For example [ 11 ] , CAUSE ( GREATER - THAN [ TEMPERATURE ( coffee ) , TEMPERATURE ( ice - cube ) ] , FLOW ( coffee , ice - cube , heat , bar ) ) These representations , while very expressive , are notoriously difficult to obtain . In this section , we investigate a weaker struc - tural representation . Our goal is to come up with a representation that can be learned , while still being expressive enough to allow analogical mining . Analogies between product ideas are intricately related to their purpose and mechanism . Informally , we think of a product\u2019s purpose as \u201cwhat it does , what it is used for\u201d , and a product\u2019s mechanism is \u201chow it does it , how it works\u201d . The importance of a product\u2019s purpose and mechanism as core components of analogy are the - oretically rooted in early cognitive psychology work on schema induction which define the core components of a schema as a goal and proposed solution to it ( e . g . , [ 15 ] ) . More recently , the practical value of defining a problem schema as a purpose and mechanism has been demonstrated to have empirical benefits in finding and using analogies to augment idea generation ( e . g . , [ 38 \u2013 41 ] ) . Separating an idea into purpose and mechanisms enables core analogical innovation processes such as re - purposing : For a given product ( such as a kitchen - sink cleaner ) and its purpose , finding another way to put it to use ( cleaning windows ) . To that end , as - sume ( for the moment ) that we have for each product i two vectors , p i and m i , representing the product\u2019s purpose and mechanism , respectively . Using this representation , we are able to apply rich queries to our corpus of products , such as : \u2022 Same purpose , different mechanism . Given the corpus of all products P , a product i with ( normalized ) purpose and mech - anism vectors p i , m i , and distance metrics d p ( \u00b7 , \u00b7 ) , d m ( \u00b7 , \u00b7 ) be - tween purpose and mechanism vectors ( respectively ) , solve : argmin \u02dci \u2208P d p ( p i , p \u02dc i ) s . t . d m ( m i , m \u02dc i ) \u2265 threshold , ( 1 ) \u2022 Same Mechanism , different purpose . Solve : argmin \u02dci \u2208P d m ( m i , m \u02dc i ) s . t . d p ( p i , p \u02dc i ) \u2265 threshold , ( 2 ) The decomposition of products into purpose and mechanism also draws inspiration from engineering functional models and ontolo - gies for describing products [ 18 ] . Although there is no set common definition of functions [ 25 ] , much research on functionality has been conducted in areas such as functional representation , engineer - ing design and value engineering . The scope of these ontologies , however , is highly \u201cmechanistic\u201d or engineering - oriented , while in many cases we observe in product data the purpose of a product is more naturally understood in others term \u2013 such as whether it is for entertainment , leisure , or more serious purposes , who is the target user ( adults , children ) , and so forth . KDD 2017 Research Paper KDD\u201917 , August 13 \u2013 17 , 2017 , Halifax , NS , Canada 236 Importantly , our dataset of product descriptions contains noisy texts , often written informally by non - professional people . In these texts product descriptions are often lacking detail or are ill - defined . To automatically describe a product in terms of a formal functional model would require an inordinate amount of meticulous data annotation and collection by professional engineers over a large number of product descriptions . We thus resort to a softer approach , hoping that a compromise on the level of detail will enable data - driven methods to automatically extract useful representations of product purpose and mechanism . Finally , we also make note of the potentially wider applicability of automatically extracting these representations from real - word product descriptions . Identifying the key components and func - tions of products could conceivably improve ( or augment ) search capabilities in internal or external product databases , and perhaps enhance recommender systems by better understanding what a user is looking for in a product and what a product offers . This last idea is connected to a line of work on \u201cproduct dimensions\u201d [ 23 ] , in which it is shown that implicitly identifying the properties of products ( such as that Harry Potter is a book about wizards ) , helps in improving recommendations . The authors propose a method that combines ratings data with textual product reviews , hoping to implicitly recover topics in the text that inform recommendations . We too look at product dimensions , but target only two that are more abstract and broad , and directly learn them in a supervised fashion from annotated data . 2 . 2 Data Innovation Corpus . We test our approach with a corpus of prod - uct descriptions from Quirky . com , an online crowdsourced product innovation website . Quirky is representative of the kinds of datasets we are interested in , because it is large ( at the time of writing , it hosts upwards of 10 , 000 product ideas , of which our corpus included 8500 ) , unstructured ( ideas are described in natural language ) , and covers a variety of domains ( invention categories ) which makes cross - domain analogies possible . The following example illustrates the typical length and \u201cmessiness\u201d of product ideas in this dataset : Thirsty too Pet water bowl / dispenser for your vehicle cup holder . Over spill lip to catch water Has optional sleeve for larger cup holders Optional floor base One way valve so water cant over flow from bottle Small reservoir Reservoir acts as backsplash Water bottle attachment Holds water in your vehicle cupholder for pet Foldable handle to get unit out of holder Dishwasher safe Optional sleeve for larger cup holders Collecting Purpose and Mechanism Data . In addition to the Quirky innovation corpus , we needed to collect analogy - specific data to train our model . Previous approaches to creating structured representations of items for analogical computation , for example predicate logic , are extremely heavyweight and can take tens of person - hours for complex items [ 37 ] . Instead , we aim to develop a lightweight task that avoids complex structure but instead relies on the cognitive expertise and intuitions of people to be able to separate the purpose of a product from its mechanism . By doing so we can Figure 1 : Collectingpurposeandmechanismannotationsfromthe crowd . scale up the collection of purpose and mechanism labels through the use of microtasks and crowdsourcing markets [ 21 ] . Specifically , we show Amazon Mechanical Turk ( AMT ) crowd workers a product description , asking them to annotate the parts of the text they consider to be about the purposes of the product , and the parts related to mechanisms . We frame the problem in simple terms , guiding workers to look for words / phrases / chunks of text talking about \u201cwhat the product does , what it is good for\u201d ( purposes ) , and \u201chow it works , what are its components\u201d ( mechanisms ) . As seen in Figure 1 , we juxtapose two copies of the product text side - by - side , to ease cognitive load and encourage workers not to give purpose and mechanism tags that are too similar or overlapping , thus capturing a potentially richer and more distinct signal . Our corpus consisted of 8500 products . Each product was annotated by four workers . Collecting Analogies . In previous , preliminary work [ 7 ] , we ex - plored the use of crowdsourcing to label analogies , collecting la - beled examples of analogies ( product pairs ) that were fed into a metric - learning deep learning model . While showing promising results , the process of collecting labeled analogies proved expen - sive , requiring considerable cognitive effort from workers and thus more time , limiting the number of labels that can realistically be collected . In addition , in that work , the deep learning model was blind to the rich structures of purpose and mechanism , and had no hope of recovering them automatically due to relative data scarcity . In this paper we take a different approach , focusing our resources on collecting purpose and mechanism annotations from the crowd , while collecting only a small number of curated labeled analogies strictly for the purpose of evaluation ( see Section 3 . 1 for details ) . 2 . 3 Method 2 . 3 . 1 Extracting Purpose and Mechanism vectors . In this section , we describe our approach to learning to extract purpose and mech - anism product representations . We begin with a set of N training product texts X N = { x 1 , x 2 , . . . , x N } , where each x i is a variable - length sequence of tokens ( x 1 i , x 2 i , . . . , x Ti ) . For each document x i , we collect a set of K purpose annotations and K mechanism an - notations , where K is the number of workers who annotate each document . We define the purpose annotation to be a binary vector \u02dcp i k = ( \u02dc p 1 i k , \u02dc p 2 i k , . . . , \u02dc p Ti k ) of the same length as x i , with \u02dc p ji k = 1 if token x ji is annotated as purpose by annotator k , \u02dc p ji k = 0 if not . In the same way , we denote KDD 2017 Research Paper KDD\u201917 , August 13 \u2013 17 , 2017 , Halifax , NS , Canada 237 the mechanism annotation with \u02dcm i k = ( \u02dc m 1 i k , \u02dc m 2 i k , . . . , \u02dc m Ti k ) While on the surface this setting appears to lend itself naturally to sequence - to - sequence learning [ 33 ] , there are a few important differences . A key difference is that in our setting the problem of interest is not to learn to recover the latent \u02dcp ( \u02dcm ) exactly for unseen products , but rather to extract some form of representation that captures the overall purpose and mechanism . That is , what we do care about is the semantic meaning or context our representation captures with respect to product purposes or mechanisms , rather than predicting any individual words . Additionally , sequence - to - sequence models typically involve heavier machinery and work well on large data sets , not suitable for our scenario were we have at most a few thousands of tagged examples . On a more technical note , instead of one sequence in the output , we now have K . A simple solution is to aggregate the annotations , for example by taking the union or intersection of annotations , or considering a token x ji to be positively annotated if it has at least K 2 positive labels . Richer aggregations may also be used . Considering that our focus here is to capture an overall rep - resentation ( not predict a precise sequence ) , however , we resort to a simple and soft aggregation of the K annotations . In simple terms , we look at all words that were annotated , and take a TF - IDF - weighted average of their word vectors . In more formal terms , let w i = ( w 1 i , w 2 i , . . . , w Ti ) be the sequence of GloVe [ 27 ] word vectors ( pre - trained on Common Crawl web data ) , representing ( x 1 i , x 2 i , . . . , x Ti ) . We select all x i word vectors for which \u02dc p ji k = 1 ( \u02dc m ji k = 1 ) for some k , and concatenate them into one sequence . We then compute the TF - IDF scores for tokens in this sequence , find the D tokens with top TF - IDF scores ( D = 5 in our experiments ) , and take the TF - IDF - weighted average of their corresponding D GloVe vectors . We denote the resulting weighted - average vectors as p i \u2208 R 300 and m i \u2208 R 300 for purpose and mechanism annotations , respectively . We consider p i , m i to be target vectors we aim to predict for unseen texts . Embedding ( short ) texts with a weighted - average of word vectors ( such as with TF - IDF ) can lead to surprisingly good results across many tasks [ 2 ] . Furthermore , in our case this simple weighted - average has several advantages . As we will next see , it lends itself to a straightforward machine learning setting , suitable for our modestly - sized data set , and for the objective of finding an overall vector representation that can be used in multiple ways , chiefly the computation of purpose - wise and mechanism - wise distances between products . Additionally , by concatenating all annotations and weighting by TF - IDF , we naturally give more weight in the average vector to words that are more frequently annotated \u2013 thus giving higher impact to words considered important by all annota - tors with respect to purpose / mechanism . 2 . 3 . 2 Learning purpose and mechanism . We now have N train - ing product texts X N = { x 1 , x 2 , . . . , x N } , and N corresponding target tuples Y N = { ( p 1 , m 1 ) , ( p 2 , m 2 ) , . . . , ( p N , m N ) } . We repre - sent each x i with its pre - trained GloVe vectors , w i . Our goal is to learn a function f ( w i ) that predicts ( p i , m i ) . To this end , we model f ( \u00b7 ) with a Recurrent Neural Network as follows . The network takes as input the variable - length sequence w i . This sequence is processed with a bidirectional RNN ( BiRNN ) [ 3 ] with one GRU layer . The BiRNN consists of the forward GRU \u2212\u2212\u2212\u2192 GRU which reads the sequence w i from w 1 i to w Ti , and a backward GRU \u2190\u2212\u2212\u2212 GRU which reads from w Ti to w 1 i , thus in practice capturing the neighborhood of w ji from \u201cboth directions\u201d : \u2212\u2192 h ji = \u2212\u2212\u2212\u2192 GRU ( w ji ) , \u2190\u2212 h ji = \u2190\u2212\u2212\u2212 GRU ( w ji ) , h ji = [ \u2212\u2192 h ji , \u2190\u2212 h ji ] , where we concatenate forward and backward GRU hidden states to obtain h ji , our representation for word j in product i . In our case , we are interested in h Ti which captures the entire product text . Next , let W p and W m be purpose and mechanism weight matrices , respectively . h Ti is a shared representation of the doc - ument , which we now transform into two new vectors , \u02c6p i , \u02c6m i , forming our purpose and mechanism predictions for product i : \u02c6p i = W p h Ti , \u02c6m i = W m h Ti . ( 3 ) Parameters in this network are then tuned to minimize the MSE loss averaged over ( p i , m i ) . In some scenarios , we may care more about predicting either purpose or mechanism , and in that case could incorporate a weight term in the loss function , giving more weight to either p i or m i . 2 . 3 . 3 Purpose and Mechanism vector interpretations . Here , we give intuition about the kinds of representations extracted and the ability to interpret them with very simple tools . We first compute \u02c6p , \u02c6m for held - out product texts . Then , in the first approach to interpreting purpose and mechanism predictions , we find the top 10 GloVe word vectors w most similar to each of \u02c6p , \u02c6m , among all vectors that appear in our vocabulary . In the second approach , we aim to recover a set of 10 word vectors such that their sparse linear combination approximately gives \u02c6p or \u02c6m . More formally , in the spirit of the sparse coding approach in [ 1 ] , consider the collection of all word vectors in our vocabulary V , w 1 , w 2 , . . . , w | V | . We stack them into a matrix W . We aim to solve the following optimization problem : argmin a | | \u02c6p i \u2212 Wa | | 22 s . t . | | a | | 0 \u2264 10 , ( 4 ) where a is a weight vector . Optimization can be done with the Orthogonal Matching Pursuit ( OMP ) [ 6 ] greedy algorithm . In Table 1 , we display some examples of applying these two simple methods , to product texts in test data ( not seen during training ) . The first product is a yogurt maker machine , used for concentrating yogurt under heat , and to reduce time and energy . We observe that words selected as most related to our purpose vector representation include food , produce , concentrate , making , energy , reduce and also words that are typical in the language used in describing advantages of products in our data , such as especially , whole , enough , much . Mechanism words , are indeed overall of a much more \u201cmechanical\u201d nature , including liquid , heat , cooling , pump , steel , machine . In the other examples too , we observe the same pattern : Words selected as most closely - related to purpose or mechanism representations , using simple techniques , empirically appear to reflect corresponding properties in the product text , both in language and in deeper meaning . KDD 2017 Research Paper KDD\u201917 , August 13 \u2013 17 , 2017 , Halifax , NS , Canada 238 Table 1 : Purpose and Mechanism vector interpretation examples . Descriptions shortened . Sparse coding shows only words with | \u03b1 | \u2265 0 . 1 Product Purpose words Mechanism words A small yogurt maker machine for con - centrating yogurt under heat and vac - uum . Has a round base in drum with customized scooper , washable stainless steel drum parts . Reduce time and en - ergy used . Top similar : concentrate , enough , food , even , much , especially , reduce , produce , whole Sparse coding : making , energy , yo - gurt , drum , concentrate , vacuum , heavy , foods , aches , service Top similar : liquid , heat , cooling , pump , steel , machine , water , heating , electric Sparse coding : vacuum , cooled , drum , heavy , ingredients , design , renewable , stainless , vending A cover placed on a car truck to protect from hail . Elastic perimeter to prevent wind from blowing under cover . Snap or velcro slits to open door without re - moving cover . Strong attachment so it wonfit blow away . Inflatable baffles that cover the top , front windshield , side . Top similar : storm , hail , rain , roofs , doors , wind , front , winds , walls Sparse coding : roof , hail , padded , ob - structing , defenses , diesel , windshield , wets Top similar : roof , cover , lining , zipper , bottom , hood , plastic , flap , rubber Sparse coding : front , cover , insulation , hail , buckle , sling , watertight , cutter , blowing A leash accessory with removable com - partments for phone , cards cash , keys , poop bags , treats , bowl . Walk your dog and carry your essentials without pock - ets or a purse bag . Top similar : bags , purse , wallet , carry , leash , backpack , pocket , dog , luggage Sparse coding : bag , leash , compart - ments , pets , phone , eats , practical , hand - ing , pull Top similar : leash , pouch , purse , pocket , bags , pockets , strap , compart - ment , backpack Sparse coding : leash , bag , compart - ments , hand , holders 3 EVALUATION : ANALOGIES As typically done in the context of learning document representa - tions , the key approach to quantitative evaluation is a down - stream task such as document classification . We now evaluate the pre - dicted \u02c6p i , \u02c6m i in the context of their ability to capture distances that reflect analogies , which is the primary focus of this paper . To do so , we first create a dataset of analogies and non - analogies . 3 . 1 Collecting analogies via crowdsourcing We crowdsourced analogy finding within a set of about 8000 Quirky products . AMT crowd workers used our search interface to collect analogies \u2013 pairs of products \u2013 for about 200 seed documents . The search task is powered by a simple word - matching approach . To deal with word variants , we added lemmas for each word to the bag - of - words associated with each product . Each search query was also expanded with lemmas associated with each query term . Search results were ranked in descending order of number of matching terms . Median completion time for each seed was 7 minutes ( work - ers could complete as many seeds as they wanted ) . Further , to deal with potential data quality issues , we recruited 3 workers for each seed ( to allow for majority - vote aggregation ) . Pairs that were tagged as matches became positive examples in our analogy dataset . However , coming up with negative exam - ples was more difficult . Borrowing from information retrieval , we assume that people read the search results sequentially , and treat the implicitly rejected documents ( i . e . , documents that were not matches , despite appearing before matches ) as negatives . It is im - portant to remember that these documents are not necessarily real negatives . To further increase the chance that the document has actually been read , we restrict ourselves to the top - 5 results . Challenges . Getting workers to understand the concept of analo - gies and avoiding tagging products that are superficially similar ( e . g . , \u201cboth smartphone - based\u201d ) as analogies proved a challenge . To address this , we scaffolded the search task by first requiring workers to generate a schema ( or \u201cpattern\u201d ) to describe the core purpose and mechanism of the product , first in concrete terms , and then in more abstract terms ( see an example pattern Figure 2 ) . Workers were then instructed to find other products that matched the abstract schema they created . We found that this scaffolded workflow reduced the number of superficial matches ; yet , a non - negligible portion of the pairs labeled as positive were either superficial matches or near analogies ( i . e . , analogies with many shared surface features ) , likely due to the strong tendency towards surface features in analogical retrieval [ 14 ] . Further , because products were multifaceted , search results may have been implicitly rejected even if they were anal - ogous to the seed if the matching schema was different from the one initially identified by the worker . 3 . 2 Quantitative results In Table 2 , we present precision and recall @ K results . We rank all pairs in the test data ( N = 2500 , with training done on about 5500 products ) based on their distances , according to various metrics , including our own . In summary , across all levels our approach outperformed the baselines , despite a challenging noisy setting . A considerable portion of test product pairs were tagged by workers as analogies despite having only surface similarity , creating mis - labeled positive examples that favor the surface - based baselines . In addition to ranking by purpose - only and mechanism - only , we also concatenate both representations in a vector [ p i , m i ] for prod - uct i , and observe an overall improvement in results , although the \u201cone - dimensional\u201d use of either purpose or mechanism alone still beats the baselines . Using m i only led to considerably better results when looking at precision @ top 1 % , perhaps indicating a tendency by workers to find more mechanism - based analogies . 4 EVALUATION : IDEATION BY ANALOGY Since a major application of the enhanced search and retrieval capabilities of analogy is enhanced creativity , we now evaluate the usefulness of our algorithms . We examine the degree to which our model\u2019s retrieved output improves people\u2019s ability to generate KDD 2017 Research Paper KDD\u201917 , August 13 \u2013 17 , 2017 , Halifax , NS , Canada 239 Figure 2 : Screenshot of analogy search interface Table 2 : Model results . Precision , recall of positive labels @ top - scoring pairs ( ranked by similarity ) . Method Top 1 % Top 5 % Top 10 % Top 15 % Top 20 % Top 25 % Glove + TF - IDF ( top 5 words ) 0 . 565 , 0 . 018 0 . 515 , 0 . 081 0 . 489 , 0 . 153 0 . 468 , 0 . 22 0 . 443 , 0 . 277 0 . 434 , 0 . 339 Glove + TF - IDF 0 . 609 , 0 . 019 0 . 559 , 0 . 087 0 . 487 , 0 . 152 0 . 47 , 0 . 22 0 . 449 , 0 . 281 0 . 426 , 0 . 332 TF - IDF 0 . 63 , 0 . 02 0 . 537 , 0 . 084 0 . 5 0 . 156 0 . 468 , 0 . 22 0 . 464 , 0 . 29 0 . 441 , 0 . 344 LSA 0 . 413 , 0 . 013 0 . 463 , 0 . 072 0 . 446 , 0 . 14 0 . 435 , 0 . 204 0 . 413 , 0 . 258 0 . 399 , 0 . 312 LDA 0 . 435 , 0 . 014 0 . 432 , 0 . 067 0 . 414 , 0 . 129 0 . 398 , 0 . 187 0 . 384 , 0 . 24 0 . 381 , 0 . 298 Purpose only 0 . 674 , 0 . 021 0 . 586 , 0 . 092 0 . 535 , 0 . 167 0 . 505 , 0 . 237 0 . 496 , 0 . 31 0 . 465 , 0 . 363 Mechanism only 0 . 739 , 0 . 023 0 . 586 , 0 . 092 0 . 551 , 0 . 172 0 . 507 , 0 . 237 0 . 482 , 0 . 301 0 . 47 , 0 . 368 concat ( Purpose , Mechanism ) 0 . 696 , 0 . 022 0 . 612 , 0 . 096 0 . 555 , 0 . 173 0 . 507 , 0 . 237 0 . 504 , 0 . 315 0 . 478 , 0 . 373 creative ideas , compared to other methods . To do so we use a standard ideation task in which participants redesign an existing product [ 36 ] , and are given inspirations to help them \u2013 either from our approach , a TF - IDF baseline , or a random baseline . See Figure 3 for an example task given to crowdworkers . Here , the task was to redesign a cell phone case that can charge the phone . The middle part shows the top 3 inspirations per condition . Our assumption is that our approach will provide participants with useful examples that are similar in purpose but provide diverse mechanisms that will help them explore more diverse parts of the design space in generating their ideas . We hypothesize that this approach will lead to better results than the TF - IDF baseline ( highly relevant but non - diverse inspirations , focusing on surface features ) and the random baseline ( highly diverse but low relevance ) . 4 . 1 Generating near - purpose far - mechanism analogies To generate inspirations for the redesign task , we start by using the learned purpose and mechanism representations p i , m i for each document i ( in the test set ) to apply rich queries to our corpus of products . In particular , assuming all vectors are normalized to unit euclidean norm , we can find pairs of products i 1 , i 2 such that d p ( p i 1 , p i 2 ) = p i 1 \u00b7 p i 2 is high ( near purpose ) , while d m ( m i 1 , m i 2 ) = m i 1 \u00b7 m i 2 is low ( far mechanism ) . This type of reasoning , as discussed above , is a core element of analogical reasoning . We take this idea one step forward by clustering by purpose and diversifying by mechanism . In more detail , we take a set of 2500 products not seen during training , and follow a simple and intuitive procedure as follows . Let P T denote our corpus of test - set products . Let S denote the number of seed products we wish to use in our experiment . Let M denote the number of inspirations we wish to produce for each seed { 1 , . . . , P } . Clustering by purpose . First , we find groups of products with similar purpose by clustering by our purpose representation . \u2022 Run K - means ( K = 50 ) , based on vectors p i , \u2200 i \u2208 P T . ( Note that when all vectors are normalized , the euclidean norm on which K - means is based is equivalent to the cosine distance ) . \u2022 For each cluster k \u2208 { 1 , . . . , K } , compute an intra - distance measure ( purpose homogeneity ) d k . We use the MSE . Prune clusters with less than M instances . Rank clusters by d k in de - scending order , pick top P . Call this set of clusters K top - purpose , with corresponding cluster centers \u00af p 1 , . . . , \u00af p S . \u2022 For each cluster k in K top - purpose , select the product i whose vector p i is nearest to the cluster center \u00af p k . This is our k th seed product , denoted by s k . Result diversification by mechanism . We now have a set of seed products , each with a corresponding cluster of products with similar purposes . Next , we need to pick M inspirations per seed . For each seed s k , we have a set of candidate matches U s k , all from the same purpose cluster . We empirically observe that in KDD 2017 Research Paper KDD\u201917 , August 13 \u2013 17 , 2017 , Halifax , NS , Canada 240 the purpose - clusters K top - purpose we generate , some vectors are highly similar to the seed with respect to mechanism , and some less so . In order to generate far - mechanism results for each seed from candidate set U s k , we now turn to diversification of results . The problem of extracting a well - diversified subset of results from a larger set of candidates has seen a lot of work , prominently in the context of information retrieval ( which is closely related to our setting ) . In our case , we assume to have found a set of relevant results U s k according to purpose metric d p ( \u00b7 , \u00b7 ) , and diversify by mechanism metric d m ( \u00b7 , \u00b7 ) . There are many ways to diversify results , mainly differing by objective function and constraints . Two canonical measures are the MAX - MIN and MAX - AVG dispersion problems [ 28 ] . In the former , we aim to find a subset M \u2286 U s k such that | M | = M , and min m i 1 , m i 2 \u2208M d m ( m i 1 , m i 2 ) is maximized . In the latter , we aim to find a subset M \u2286 U s k such that | M | = M , and 2 M ( M \u2212 1 ) (cid:213) m i 1 , m i 2 \u2208M d m ( m i 1 , m i 2 ) is maximized . In other words , in the MAX - MIN problem we find a subset of products M such that the distance between the two nearest products is maximized . In the MAX - AVG problem , we find a subset such the average distance between pairs is maximized . Both problems admit simple greedy algorithms with constant - factor approximations [ 28 ] . We choose the MAX - MIN problem , since we want to avoid displaying too - similar results even once to a user ( who may become frustrated and not proceed to read more inspirations ) . We solve the problem using the GMM algorithm mentioned in [ 28 ] . Each iteration of GMM selects a candidate m \u2208 U s k \u2212M such that the minimum distance from m to an already - selected product in M is the largest among remaining candidates in U s k \u2212M , where we measure distance according to our mechanism metric d m ( \u00b7 , \u00b7 ) . In our experiments , we set P = 12 , M = 12 , for 12 seeds and 12 matches each , respectively . 4 . 2 Experiment design We recruited 38 AMT workers to redesign an existing product , a common creative task in design firms [ 36 ] . To ensure robustness of effects , the experiment included 12 different \u201cseed\u201d products . Participants were paid $ 1 . 5 for their participation . To maximize statistical power , we utilized a within - subjects design with a single manipulated factor , inspiration type : \u2022 ANALOGY : participants receive 12 product inspirations re - trieved by our method detailed above , using near - purpose far - mechanism clustering and diversification . \u2022 BASELINE : SURFACE : participants receive product inspira - tions retrieved using TF - IDF , by finding the top 12 products similar to the seed . This baseline is meant to simulate current search engines . \u2022 BASELINE : RANDOM : participants receive 12 product inspi - rations sampled at random from our product corpus . Since we used a within - subjects design , participants completed the redesign task under each of the 3 inspiration type conditions . Figure 3 : Overview and excerpts of the ideation experiment . Top : Seed product . Workers were asked to solve the same problem in a different way . Middle : Top 3 inspirations for each of the condi - tions . Note that the TF - IDF baseline returns results from the same domain , while our method returns a broader range of products . Bot - tom : Ideas generated by users exposed to the different conditions . The order of conditions was counterbalanced to prevent order ef - fects . To ensure unbiased permutations , we used the Fisher - Yates shuffle to assign seeds to conditions , so that every seed would be seen in all conditions ( by different users ) . Since prior work has shown that people benefit more from analo - gies if they receive them after ideation has begun [ 34 ] , the ideation task proceeded in two phases : 1 ) generating ideas unassisted for one minute , then 2 ) receiving 12 inspirations and generating more ideas for 6 minutes . The inspirations were laid out in four pages , 3 inspirations per page , and the users could freely browse them . Figure 3 provides an overview of the experiment and an excerpt from the data . The original task was to redesign an existing product , in this case a cell phone charger case . The SURFACE baseline retrieves products that are very phone - related ( or case related ) . In contrast , our algorithm retrieves diverse results such as a human pulley - powered electricity generator suit . The bottom of the figure shows ideas generated by users in each condition . Interestingly , KDD 2017 Research Paper KDD\u201917 , August 13 \u2013 17 , 2017 , Halifax , NS , Canada 241 the user exposed to our approach suggested a case that generates power using movement , potentially inspired by the suit . 4 . 3 Results Measures . We are interested in the ability of our approach to en - hance people\u2019s ability to generate creative ideas . Following [ 29 ] , we measured creative output as the rate at which a participant gener - ates good ideas . We recruited five graduate students to judge each idea generated by our participants as good or not . Our definition of \u201cgood\u201d follows the standard definition of creativity in the literature as a combination of novelty , quality , and feasibility [ 30 ] . Each judge was instructed to judge an idea as good if it satisfied all of the follow - ing criteria : 1 ) it uses a different mechanism / technology than the original product ( novelty ) , 2 ) it proposes a mechanism / technology that would achieve the same purpose as the original product ( qual - ity ) , and 3 ) could be implemented using existing technology and does not defy physics ( feasibility ) . Agreement between the judges was substantial , with a Fleiss kappa of 0 . 51 , lending our measure of creativity acceptable inter - rater reliability . The final measure of whether an idea was good or not was computed by thresholding the number of votes , so that good = 1 if at least k judges rated it as good . We report results for both liberal and strict settings k = 2 , 3 . Evaluation . For k = 2 , out of 749 total ideas collected , 249 ideas were judged as good by this measure . As mentioned above , we use the Fisher - Yates shuffle to assign seeds to conditions . To take a conservative approach , as a first step we look only at seeds that appeared across all three conditions ( 9 such seeds ) , to put the con - ditions on par with one another . By this slicing of the data , there were 208 good ideas . The proportion of good ideas in our condition was 46 % ( N = 105 ) . Next was the random baseline with 37 % ( 49 ) , and finally the TF - IDF baseline achieved 30 % ( N = 54 ) . These results are significant by a \u03c7 2 proportion test ( p \u2264 . 01 ) . We thus observe that both in terms of the absolute number of positively - rated ideas and in terms of proportions , our approach was able to generate a considerably large relative positive effect , leading to better ideas . For k = 3 ( the majority vote ) , out of 749 total ideas collected , 184 ideas were judged as good . Again , we start by looking only at seeds that appeared across all three conditions ( 9 such seeds ) . This leaves 154 good ideas . The proportion of good ideas in our condition was 38 % ( N = 118 ) . Next - up was the random baseline with 22 % ( 68 ) , and finally the TF - IDF baseline achieved 21 % ( N = 63 ) , with p < . 01 . By looking at the more conservative majority - vote threshold , the observed effect of our method only increases . Looking only at seeds that appeared across all conditions was a basic way to make sure we cancel out possible confounding fac - tors . A more refined way is attempting to model these effects and condition on them , as follows . We are interested in the likelihood that a given idea is good , or pr ( good ) , as a function of inspiration condition . However , ideas are not independent : each participant generated multiple ideas , and ideas were proposed for different seeds . Failing to account for these dependencies would lead to inaccurate estimates of the effects of the inspirations : some participants may be better at generating ideas than others , while some seeds might be more easy / difficult than others . Therefore , we used a generalized linear mixed model , Figure 4 : Showing proportion estimates by our random - effect lo - gistic regression , for k = 2 ( left ) and k = 3 ( right ) . Participants are significantly more likely to generate good ideas for the redesign ideation task when given inspirations from our analogy approach compared to baseline - surface and baseline - random approaches with a fixed effect of inspiration condition , and random effects of participant and seed ( to model within - participant and within - seed dependencies between ideas ) . For k = 2 , our resulting model ( with fixed effect of inspiration condition ) yields a significant reduction in variance compared to a null model with no fixed effects , Likelihood ratio \u03c7 2 ( 3 ) = 67 . 96 , p < . 01 . The model also yields a reduction in Akaike Information Criterion ( AIC ) , from 682 . 28 in the null model to 620 . 32 , indicating that the improved fit to the data is not due to overfitting . For k = 3 , the model also yields a significant drop in variance compared to a null model , Likelihood ratio \u03c7 2 ( 3 ) = 92 . 38 , p < . 01 , with AIC dropping from 682 . 28 in the null model to 595 . 90 . As Figure 4 shows , our method led to a significantly higher prob - ability for good ideas . For k = 2 , pr ( Good ) = 0 . 71 , 95 % confidence interval = [ 0 . 48 , 0 . 87 ] in our condition . TF - IDF had pr ( Good ) = 0 . 28 [ 0 . 16 , 0 . 44 ] , and random had pr ( Good ) = 0 . 27 [ 0 . 16 , 0 . 41 ] . The advantages of the analogy condition over each baseline are both substantial and statistically significant , B = \u2212 1 . 81 , p < . 01 vs . TF - IDF , and B = \u2212 1 . 88 , p < . 01 vs . random . For k = 3 , we had pr ( Good ) = 0 . 56 , [ 0 . 36 , 0 . 75 ] . TF - IDF had pr ( Good ) = 0 . 16 [ 0 . 08 , 0 . 27 ] , and random had pr ( Good ) = 0 . 14 [ 0 . 08 , 0 . 24 ] , B = \u2212 1 . 94 , p < . 01 vs . TF - IDF , and B = \u2212 2 . 05 , p < . 01 vs . random . Note that confidence intervals for the probability estimates are relatively wide ( more so , unsurprisingly , for k = 2 ) . Replications of this experiment , possibly with more data , could yield results somewhere in between , with more precise estimates on the true size of the effect . The main take - away of this study is that our approach yields a reliable increase in participants\u2019 creative ability . 5 DISCUSSION AND CONCLUSION In this paper , we sought to develop a scalable approach to finding analogies in large , messy , real - world datasets . We explored the po - tential of learning and leveraging a weak structural representation ( i . e . , purpose and mechanism vectors ) for product descriptions . We leverage crowdsourcing techniques to construct a training dataset with purpose / mechanism annotations , and use an RNN to learn purpose and mechanism vectors for each product . We demonstrate that these learned vectors allow us to find analogies with higher precision than traditional information - retrieval similarity metrics like TF - IDF , LSA , GloVe and LDA . Our ideation evaluation experiment further illustrates the effec - tiveness of our approach : participants had a higher likelihood of generating good ideas for the redesign ideation task when they KDD 2017 Research Paper KDD\u201917 , August 13 \u2013 17 , 2017 , Halifax , NS , Canada 242 received inspirations sampled by our analogy approach ( tuned to be similar in purpose , but different in mechanism ) , compared to a traditional ( TF - IDF ) baseline or random sampling approach . From a psychological perspective , the benefits of our inspirations are likely due to our approach\u2019s superior ability to sample diverse yet still structurally similar inspirations , since diversity of examples is a known robust booster for creative ability [ 8 ] . The TF - IDF approach yielded inspirations likely to be relevant , but also likely to be re - dundant and homogeneous , while the random sampling approach yields diversity but not relevance . While moving to a \u201cweak\u201d structural representation based on purpose and mechanism significantly increased the feasibility of analogy - finding , extensions may be necessary to generalize to other domains besides product descriptions . For example , our purpose and mechanism vectors did not distinguish between higher and lower level purposes / mechanisms , or core / peripheral purposes / mechanisms , and also did not encode dependencies between partic - ular purposes / mechanisms . These are potentially fruitful areas for future work and may be especially important when moving from relatively simple product descriptions to more complex data such as scientific papers , in which purposes and mechanisms can exist at multiple hierarchical levels ( e . g . , \u201caccelerate innovation\u201d vs . \u201clearn a vector representation of the purpose of an item\u201d ) . More gener - ally , we believe exploring the tradeoffs between degree of structure , learnability ( including costs of generating training data , accuracy , and generalizability ) and utility for augmenting innovation could lead to interesting points in the design space that could have both theoretical and practical value . Acknowledgments . The authors thank the anonymous reviewers for their helpful comments . This work was supported by NSF grants CHS - 1526665 , IIS - 1149797 , IIS - 1217559 , Carnegie Mellon\u2019s Web2020 initiative , Bosch , Google , ISF grant 1764 / 15 and Alon grant . Dafna Shahaf is a Harry & Abe Sherman assistant professor . REFERENCES [ 1 ] Sanjeev Arora , Yuanzhi Li , Yingyu Liang , Tengyu Ma , and Andrej Risteski . 2016 . Linear algebraic structure of word senses , with applications to polysemy . arXiv preprint arXiv : 1601 . 03764 ( 2016 ) . [ 2 ] Sanjeev Arora , Yingyu Liang , and Tengyu Ma . 2016 . A simple but tough - to - beat baseline for sentence embeddings . ( 2016 ) . [ 3 ] Dzmitry Bahdanau , Kyunghyun Cho , and Yoshua Bengio . 2014 . Neural ma - chine translation by jointly learning to align and translate . arXiv preprint arXiv : 1409 . 0473 ( 2014 ) . [ 4 ] David M . Blei , Andrew Y . Ng , Michael I . Jordan , and John Lafferty . 2003 . Latent Dirichlet Allocation . Journal of Machine Learning Research ( 2003 ) , 993 \u2013 1022 . [ 5 ] Danushka T Bollegala , Yutaka Matsuo , and Mitsuru Ishizuka . 2009 . Measuring the similarity between implicit semantic relations from the web . In Proceedings of the 18th international conference on World wide web . ACM , 651 \u2013 660 . [ 6 ] T Tony Cai and Lie Wang . 2011 . Orthogonal matching pursuit for sparse signal recovery with noise . IEEE Transactions on Information Theory 57 , 7 ( 2011 ) . [ 7 ] JoelChan , TomHope , DafnaShahaf , andAniketKittur . 2016 . ScalingupAnalogy with Crowdsourcing and Machine Learning . In Workshop on Computational Analogy at ICCBR . [ 8 ] Joel Chan and Christian D . Schunn . 2015 . The importance of iteration in creative conceptual combination . Cognition 145 ( Dec . 2015 ) , 104 \u2013 115 . [ 9 ] Darren W Dahl and Page Moreau . 2002 . The influence and value of analogical thinkingduringnewproductideation . JournalofMarketingResearch 39 , 1 ( 2002 ) . [ 10 ] Scott Deerwester , Susan T . Dumais , Geroge W . Furnas , and Thomas K . Landauer . 1990 . Indexing by Latent Semantic Analysis . JASIST 41 , 6 ( 1990 ) , 1990 . [ 11 ] Brian Falkenhainer , Kenneth D . Forbus , and Dedre Gentner . 1989 . The structure - mapping engine : Algorithm and examples . Artificial Intelligence 41 , 1 ( 1989 ) . [ 12 ] DedreGentner . 1983 . Structure - Mapping : ATheoreticalFrameworkforAnalogy * . Cognitive science 7 , 2 ( 1983 ) , 155 \u2013 170 . [ 13 ] Dedre Gentner , Russell Landers , and others . 1985 . Analogical reminding : A good match is hard to find . In Proceedings of the international conference on systems , man and cybernetics . 607 \u2013 613 . [ 14 ] Dedre Gentner , Mary Jo Rattermann , and Kenneth D Forbus . 1993 . The roles of similarity in transfer : Separating retrievability from inferential soundness . Cognitive psychology 25 , 4 ( 1993 ) , 524 \u2013 575 . [ 15 ] M . L . Gick and K . J . Holyoak . 1983 . Schema induction and analogical transfer . Cognitive Psychology 15 , 1 ( 1983 ) , 1 \u2013 38 . [ 16 ] Graeme S Halford , Rosemary Baker , Julie E McCredden , and John D Bain . 2005 . How many variables can humans process ? Psychological science 16 , 1 ( 2005 ) . [ 17 ] Mary B Hesse . 1966 . Models and analogies in science . ( 1966 ) . [ 18 ] Julie Hirtz , Robert B Stone , Daniel A McAdams , Simon Szykman , and Kristin L Wood . 2002 . A functional basis for engineering design : reconciling and evolving previous efforts . Research in engineering Design 13 , 2 ( 2002 ) , 65 \u2013 82 . [ 19 ] Douglas R Hofstadter , Melanie Mitchell , and others . 1994 . The copycat project : A model of mental fluidity and analogy - making . Advances in connectionist and neural computation theory 2 , 31 - 112 ( 1994 ) , 29 \u2013 30 . [ 20 ] Keith James Holyoak and Paul Thagard . 1996 . Mental leaps : Analogy in creative thought . MIT press . [ 21 ] Aniket Kittur , Ed H . Chi , and Bongwon Suh . 2008 . Crowdsourcing User Studies with Mechanical Turk . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201908 ) . ACM , New York , NY , USA , 453 \u2013 456 . [ 22 ] Arthur B Markman and Jeffrey Loewenstein . 2010 . Structural comparison and consumer choice . Journal of Consumer Psychology 20 , 2 ( 2010 ) , 126 \u2013 137 . [ 23 ] Julian McAuley and Jure Leskovec . 2013 . Hidden factors and hidden topics : understanding rating dimensions with review text . In Proceedings of the 7th ACM conference on Recommender systems . ACM , 165 \u2013 172 . [ 24 ] Tomas Mikolov , Kai Chen , Greg Corrado , and Jeffrey Dean . 2013 . Efficient Estimation of Word Representations in Vector Space . arXiv : 1301 . 3781 [ cs ] ( Jan . 2013 ) . http : / / arxiv . org / abs / 1301 . 3781 arXiv : 1301 . 3781 . [ 25 ] Masanori Ookubo , Yusuke Koji , Munehiko Sasajima , Yoshinobu Kitamura , and Riichiro Mizoguchi . 2007 . Towards Interoperability between Functional Tax - onomies using an Ontology - based Mapping . In Proc . of ICED 2007 . [ 26 ] Colin Pask . 2003 . Mathematics and the science of analogies . American Journal of Physics 71 , 6 ( 2003 ) , 526 \u2013 534 . [ 27 ] Jeffrey Pennington , Richard Socher , and Christopher D Manning . 2014 . Glove : Global Vectors for Word Representation . In EMNLP , Vol . 14 . 1532 \u2013 43 . [ 28 ] SekharipuramSRavi , DanielJRosenkrantz , andGiriKumarTayi . 1994 . Heuristic and special case algorithms for dispersion problems . Operations Research 42 , 2 ( 1994 ) , 299 \u2013 310 . [ 29 ] Bruce A . Reinig , Robert O . Briggs , and Jay F . Nunamaker . 2007 . On the Measure - ment of Ideation Quality . Journal of Management Information Systems 23 , 4 ( May 2007 ) , 143 \u2013 161 . [ 30 ] Mark A . Runco and Garrett J . Jaeger . 2012 . The Standard Definition of Creativity . Creativity Research Journal 24 , 1 ( Jan . 2012 ) , 92 \u2013 96 . [ 31 ] Ekaterina Shutova . 2010 . Models of metaphor in NLP . In Proceedings of the 48th annual meeting of the association for computational linguistics . Association for Computational Linguistics , 688 \u2013 697 . [ 32 ] Barbara A Spellman and Frederick Schauer . 2012 . Legal reasoning . Virginia Public Law and Legal Theory Research Paper 2012 - 09 ( 2012 ) . [ 33 ] Ilya Sutskever , Oriol Vinyals , and Quoc V Le . 2014 . Sequence to sequence learning with neural networks . In Advances in neural information processing systems . 3104 \u2013 3112 . [ 34 ] Ian Tseng , Jarrod Moss , Jonathan Cagan , and Kenneth Kotovsky . 2008 . The role oftimingandanalogicalsimilarityinthestimulationofideagenerationindesign . Design Studies 29 , 3 ( 2008 ) , 203 \u2013 221 . [ 35 ] Peter D Turney . 2006 . Similarity of semantic relations . Computational Linguistics 32 , 3 ( 2006 ) , 379 \u2013 416 . [ 36 ] D . Ullman . 2002 . The Mechanical Design Process ( 3 ed . ) . New York , NY . [ 37 ] Swaroop Vattam , Bryan Wiltgen , Michael Helms , Ashok K . Goel , and Jeannette Yen . 2011 . DANE : Fostering Creativity in and through Biologically Inspired Design . In Design Creativity 2010 . [ 38 ] Lixiu Yu , Aniket Kittur , and Robert E Kraut . 2014 . Searching for analogical ideas with crowds . In Proceedings of the 32nd annual ACM conference on Human factors in computing systems . ACM , 1225 \u2013 1234 . [ 39 ] Lixiu Yu , Aniket Kittur , and Robert E Kraut . 2016 . Encouraging fiOutside - the - boxfi Thinking in Crowd Innovation Through Identifying Domains of Expertise . In Proceedings of the 19th ACM Conference on Computer - Supported Cooperative Work & Social Computing . ACM , 1214 \u2013 1222 . [ 40 ] L Yu , B Kraut , and A Kittur . 2014 . Distributed analogical idea generation : inno - vating with crowds . In CHI\u201914 . [ 41 ] Lixiu Yu , Robert E Kraut , and Aniket Kittur . 2016 . Distributed Analogical Idea Generation with Multiple Constraints . In Proceedings of the 19th ACM Conference on Computer - Supported Cooperative Work & Social Computing . ACM , 1236 \u2013 1245 . KDD 2017 Research Paper KDD\u201917 , August 13 \u2013 17 , 2017 , Halifax , NS , Canada 243",
    "zhu2022mem3dg": "ARTICLE Mem3DG : Modeling membrane mechanochemical dynamics in 3D using discrete differential geometry Cuncheng Zhu , 1 Christopher T . Lee , 1 , * and Padmini Rangamani 1 , * 1 Department of Mechanical and Aerospace Engineering , University of California San Diego , La Jolla CA 92093 ABSTRACT Biomembranes adopt varying morphologies that are vital to cellular functions . Many studies use computational modeling to understand how various mechanochemical factors contribute to membrane shape transformations . Compared with approximation - based methods ( e . g . , \ufb01 nite element method [ FEM ] ) , the class of discrete mesh models offers greater \ufb02 ex - ibility to simulate complex physics and shapes in three dimensions ; its formulation produces an ef \ufb01 cient algorithm while main - taining coordinate - free geometric descriptions . However , ambiguities in geometric de \ufb01 nitions in the discrete context have led to a lack of consensus on which discrete mesh model is theoretically and numerically optimal ; a bijective relationship between the terms contributing to both the energy and forces from the discrete and smooth geometric theories remains to be established . We address this and present an extensible framework , Mem3DG , for modeling 3D mechanochemical dynamics of membranes based on discrete differential geometry ( DDG ) on triangulated meshes . The formalism of DDG resolves the inconsistency and provides a unifying perspective on how to relate the smooth and discrete energy and forces . To demonstrate , Mem3DG is used to model a sequence of examples with increasing mechanochemical complexity : recovering classical shape transformations such as 1 ) biconcave disk , dumbbell , and unduloid ; and 2 ) spherical bud on spherical , \ufb02 at - patch membrane ; investigating how the coupling of membrane mechanics with protein mobility jointly affects phase and shape transformation . As high - resolution 3D imaging of membrane ultrastructure becomes more readily available , we envision Mem3DG to be applied as an end - to - end tool to simulate realistic cell geometry under user - speci \ufb01 ed mechanochemical conditions . INTRODUCTION Computational modeling of lipid bilayer mechanics has long been accepted as a way to probe the bio - physical aspects of membrane curvature generation . The ability of lipid bilayers and cellular membranes to bend in response to various applied forces has been studied extensively from the mathematical modeling perspective . However , the nonlinear system of equations that result from such modeling often leads to a computational bottleneck to generate pre - dictions from simulations that can be tested against experimentally observed shapes . In this study , we develop a mesh - based model using discrete differen - tial geometry ( DDG ) to reduce this bottleneck . To justify why our method is necessary and is a computa - tional advance , we \ufb01 rst describe the importance of membrane curvature generation in biology , the cur - rent state of the art in membrane mechanics modeling , and \ufb01 nally explicitly state the goals of our approach . As one of the most essential and conserved structures of cells , cellular membranes perform many functions . First , they form compartments to separate chemical environments . Beyond the passive role of par - titioning space , lipids in the membranes interact with Submitted March 12 , 2022 , and accepted for publication June 8 , 2022 . * Correspondence : ctlee @ ucsd . edu or prangamani @ ucsd . edu Editor Name : J\u00f6rg Enderlein SIGNIFICANCE Cellular membranes have shapes and shape changes that characterize cells / organelles , and support nutrient traf \ufb01 cking among other critical processes . Modeling membrane shape changes using mechanical principles can provide insight into how cells robustly bend membranes to support life . Mathematical and computational strategies to solve the equations describing membrane shape evolution can be complex and challenging without simplifying assumptions . Here , we present a new , general , numerical approach to model arbitrary 3D membrane shapes in response to interaction with curvature - sensing and generating membrane proteins . The accompanying implementation , Mem3DG , is a software tool to make computational membrane mechanics accessible to the general researcher . Biophysical Reports 2 , 100062 , September 14 , 2022 1 https : / / doi . org / 10 . 1016 / j . bpr . 2022 . 100062 (cid:1) 2022 The Author ( s ) . This is an open access article under the CC BY - NC - ND license ( http : / / creativecommons . org / licenses / by - nc - nd / 4 . 0 / ) . proteins and other cellular components in \ufb02 uencing cell signaling ( e . g . , by localizing molecules and acting as an entropic barrier ) ( 1 , 2 ) . Membrane morphology and to - pology changes are critical for traf \ufb01 cking cargo in and out of cells and are very carefully regulated ( 3 \u2013 8 ) . Central to these roles is the ability of the mem - brane to curve and adopt varying morphological con \ufb01 g - urations from spheres to highly - curved and variegated structures . Advances in experimental studies of membrane - pro - tein interactions ( 9 \u2013 20 ) , ultrastructural imaging ( 21 \u2013 30 ) , and image analysis ( 9 \u2013 11 , 31 \u2013 37 ) have re - vealed much about the molecular interactions that regulate membrane curvature . To investigate the me - chanics behind these interactions , many theoretical and computational models in terms of membrane energetics and thermodynamics have been developed ( 7 , 38 \u2013 52 ) . These models , owing to the ease of in silico experimentation , have become an important tool for generating and testing hypotheses ( 53 , 54 ) . These mechanics models and associated simulations have been used to provide intuition on the mechanical requirements for forming and maintaining complex cellular membrane shapes ( 55 \u2013 63 ) . While the utility of this approach has been estab - lished and many models have been developed ( 38 ) , many models are limited by critical assumptions or other technical challenges . For example , the ability to use geometries from membrane ultrastructural imag - ing experiments as a starting condition would improve model realism ( 64 ) . With respect to computational complexity , the solver should be able to model defor - mations and topological changes in three dimensions and be compatible with both energy minimization and time integration for comparing with static and time - se - ries experiments respectively . This is in contrast to the current assumptions of the existence of an axis of sym - metry that is quite commonly made for purposes of ease of simulation ( 65 ) . An additional feature for these solvers should be that their implementation is modular such that the addition of new physics or increasing model complexity should be straightforward . This in - cludes the potential for coupling the membrane model with agent - based and other simulations to propagate other cellular components such as the cytoskeleton ( 66 ) . Thus , new computational tools which are general , easy to use , and without restrictive assumptions are needed to bring modeling closer to experimental obser - vations of membrane shapes in cells . To emphasize the motivations behind our choice of extending and developing a new mesh - based mem - brane model , we provide a summary of the legacy liter - ature in modeling membrane mechanics . The most common theoretical model of membrane bending is the Helfrich - Canham - Evans Hamiltonian ( The Helfrich energy is related to the Willmore energy in the mathe - matics literature ( 67 ) ) , which describes the lipid bilayer as a 2D \ufb02 uid - like structure that exhibits resistance to bending in the out - of - plane direction ( 39 , 40 , 68 \u2013 70 ) . It is a continuum model that describes the bending en - ergy of the membrane as a function of its mean and Gaussian curvatures . The assumptions for the contin - uum are satis \ufb01 ed as long as the deformations are much larger in length scale compared with the individ - ual lipid components . Given the necessary material properties and bound - ary conditions , by minimizing the Helfrich energy , we can obtain the equilibrium shape of the membrane ( 39 , 70 \u2013 72 ) . While straightforward in concept , energy minimization requires the determination of the forces on the membrane , which is a challenging task ( 65 ) . The forces on the membrane are given by the variation of the energy with respect to the embedded coordinate ( i . e . , shape ) of the membrane ( we call this variation the shape derivative , which is distinct from the chemical derivative that will be introduced later in the context of mechanochemical coupling ) . Taking the shape de - rivatives of the Helfrich energy produces the \u201c shape equation , \u201d so termed because solutions of this partial differential equation ( PDE ) , with the prescribed bound - ary conditions , produce con \ufb01 gurations at equilibrium ( i . e . , force balance ) . Solving the shape equation is non - trivial since it is a PDE with fourth - order nonlinear terms . As a result , analytical solutions of the shape equation are known only for a few cases constrained to speci \ufb01 c geometries and boundary conditions ( 42 ) . For most systems , we must resort to the use of numerical methods . The simplest numerical schemes can be formulated by mak - ing restrictive assumptions such as considering only small deformations from a plane ( e . g . , Monge parame - trization ) or assuming that there exists an axis of sym - metry such that the resulting boundary value system can be integrated ( 38 ) . While these methods are suit - able for idealized shapes , these assumptions are not consistent with the membrane shapes found in biology are and thus not general enough for advancing the \ufb01 eld . Solvers of membrane shape in 3D have also been developed and can be categorized into three groups : 1 ) phase \ufb01 eld or level set methods ( 73 \u2013 77 ) , 2 ) \ufb01 nite element method ( FEM ) ( 78 \u2013 85 ) , and 3 ) discrete surface mesh models ( 60 , 86 \u2013 99 ) . These methods and others , reviewed in detail by Guckenberger et al . ( 100 ) , differ in the strategy used to discretize the membrane domain and compute the relevant derivatives . We compare the aforementioned general , 3D models with our estab - lished model criteria in Table 1 and elaborate below . Phase \ufb01 eld and level set methods solve the shape equation by propagating a density \ufb01 eld on an ambient volumetric mesh . The membrane shape is implicit in 2 Biophysical Reports 2 , 100062 , September 14 , 2022 these models and can be found by drawing an isosur - face or level set of the model . While this is ideal for modeling membrane topological changes , the implicit representation of the membrane adds complexity for interfacing with data generated using modern methods of visualizing membrane ultrastructure . The meshes output from ultrastructural studies must be converted into a density or phase \ufb01 eld prior to input to the model . While this conversion is possible , representing the dy - namic and variegated shapes of cellular membranes would require a dense volume mesh , which reduces computational tractability . The implicit surface repre - sentation also complicates the addition of new in - plane physics for end users . FEM and discrete mesh models use an explicit sur - face parametrization ( i . e . , a mesh ) . Thus the meshes output from ultrastructural imaging datasets can be used in these frameworks with minor modi \ufb01 cations ( 32 , 101 ) . FEM relies on elementwise interpolation functions and is commonly derived from the weak formulation of boundary value problems . Comparing FEM methods with our speci \ufb01 cations we identify a few key challenges . First , the numerical evaluation of smooth geometric measurements on arbitrary mani - folds in an FEM framework requires non - intuitive tensor algebra to translate the shape equation in coor - dinate where it is ready to be solved . After this formu - lation , solving the shape equation can require the use of high - order function basis such as the C 1 conform - ing FEM based on subdivision scheme ( 78 , 79 ) or iso - geometric analysis ( IGA ) ( 81 \u2013 83 , 85 ) , which adds code complexity and run - time cost . Extending an FEM framework to incorporate new physics , topologi - cal changes , or interfaces with other models requires advanced mathematical and coding skills . This can restrict the usage to the computational math commu - nity and prevent broad usage by the biophysics community . Finally , evaluating discrete mesh - based methods , which de \ufb01 ne the system energy and / or forces using geometric primitives from a mesh , we \ufb01 nd that they satisfy many of the requirements in Table 1 . Due to the ease of use and implementation , discrete mesh models have gained in popularity and many different schemes can be found in the literature ( 60 , 86 \u2013 99 , 102 , 103 ) . These schemes differ in their approach to de \ufb01 ning and computing geometric measurements necessary for de \ufb01 ning the energy and forces on a discrete object . Discrete geometries have discontinu - ities and limited information that leads to degenerate de \ufb01 nitions for geometric values . For example , there is no canonical de \ufb01 nition for the normal of a vertex of a mesh as opposed to the normal of a smooth geometry ( 89 , 104 , 105 ) . One challenge for selecting the suitable formulation to use is the lack of approximation error metric for which the discrete de \ufb01 nition best matches the smooth theory . Another confounding factor is the step at which the problem is discretized . Some imple - mentations discretize the energy of the system by con - structing standalone discrete energy , which captures the behavior of the Helfrich energy ( 65 ) . From this discrete energy , they take the shape derivatives to obtain an expression for the discrete force . Without careful consideration , the discrete forces derived in this manner are unstructured and there is little resem - blance to expressions of force from smooth theory . A second option is to discretize the smooth force expression directly ( 65 , 100 ) . While this preserves the geometric connection for the forces , there is no longer well - de \ufb01 ned discrete energy . Several discrete mesh methods were benchmarked by Bian et al . ( 89 ) and Guckenberger et al . ( 100 ) who found differences in the accuracy , robustness , and ease of implementation ( 89 , 100 ) . In this work , we outline a discrete mesh framework for modeling membrane mechanics with the following goals in mind : 1 ) we do not make a priori assumptions about axes of symmetry or restrict the coordinates in any way ; 2 ) we resolve the ambiguity in the de \ufb01 nition of geo - metric measurements on the mesh and permit a direct TABLE 1 Comparison of common mathematical frameworks for modeling membrane mechanics with speci \ufb01 cations to advance the mission of computational membrane mechanobiology Phase \ufb01 eld / level set FEM Discrete mesh / Mem3DG General 3D U U U Statics \u00fe dynamics U U U Membrane heterogeneity U U U Incorporation of agents / particles U Incorporation of stochastic dynamics ( e . g . , DPD or MC ) U Explicit surface parametrization U U Coordinate - free evaluation U Ability to support topological changes U requires mesh surgery Error analysis U U A general framework will permit the easy transfer of inputs and results between model and experiments . Models that can be coupled with other modeling schemes representing other cellular components can help address the complexity of cell biology . Discrete mesh models have many desirable traits , with respect to these speci \ufb01 cations , at the cost of forgoing rigorous error analysis . Biophysical Reports 2 , 100062 , September 14 , 2022 3 comparison for both the energy and force expressions in smooth and discrete contexts ; and 3 ) this framework al - lows for use of meshes generated from ultrastructural imaging . We begin by de \ufb01 ning discrete energy that is analogous to the Helfrich energy . Then , using concepts from DDG , we derive discrete shape derivatives analyti - cally and group terms to produce a discrete shape equa - tion . We will show that our discrete shape equation has a clear correspondence between the terms of the smooth shape equations ( 57 , 67 , 70 , 71 ) . Beyond establishing this important connection , we will show that the elegant analytical expressions for discrete variational terms from the DDG also yield improved geometric intuition and numerical accuracy ( 104 , 105 ) . Benchmarking of our expressions was performed with our accompanying software implementation called Membrane Dynamics in 3D using Discrete Differential Geometry ( Mem3DG ) . Mem3DG is written in C \u00fe\u00fe , released under the Mozilla Public License version 2 , and comes with accompanying documentation and tutorials which can be accessed on GitHub ( https : / / github . com / RangamaniLabUCSD / Mem3DG ) . Beyond the computa - tionofdiscrete energiesand forcesona meshofinterest , we also include functionality for performing energy mini - mization and time integration . Using Mem3DG , we vali - date the exactness of the analytical expressions of force terms by numerically examining the convergence of the force terms as a function of system energy perturbation . To illustrate compliance with our tool speci \ufb01 cations , we apply Mem3DG to a sequence of exam - ples with increasing complexity . Finally , we outline the steps to incorporate additional physics such as mem - brane - protein interactions and surface diffusion into Mem3DG . THEORY The lipid bilayer is modeled as a thin elastic shell us - ing the Helfrich - Canham - Evans Hamiltonian or sponta - neous curvature model ( 39 , 69 , 106 ) . The bending energy , E b , of a smooth surface or 2 - manifold , M , can be expressed in terms of the mean H ; Gaussian K , and spontaneous curvature (cid:1) H with ma - terial parameters k the bending and k G the saddle - splay moduli . Additional energy terms E s and E p ac - count for the tension area \u00f0 l - - A \u00de and pressure - volume ( D P - - V ) relationships . The total energy of the bilayer is therefore E \u00bc Z M (cid:2) k \u00f0 H (cid:2) (cid:1) H \u00de 2 \u00fe k G K (cid:3) dA | \ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 } E b \u00fe Z A (cid:1) A l d ~ A | \ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04 } E s (cid:2) Z V (cid:1) V D Pd ~ V | \ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 } E p : ( 1 ) The preferred surface area and volume , (cid:1) A and (cid:1) V , combined with the spontaneous curvature , (cid:1) H , charac - terize the zero - energy state for the system energy . In a nutshell , given the material properties , the system en - ergy is fully determined by its geometric measure - ments such as volume , area , and curvatures . Machinery to express these measurements have been a topic of extensive study in classical differential geometry ( 107 , 108 ) . However , \ufb01 nding the minima of the governing energy , solving the stationary solution to the geometric PDE , can be mathematically and numerically dif \ufb01 cult . While differential geometry pro - vides succinct expressions to describe the measure - ments in a coordinate - free fashion , computational methods often require the introduction of a coordinate basis and subsequent manipulation of expressions us - ing tensor algebra , which can obscure the underlying geometric intuition . As an alternative , forgoing the need for a smooth geometry , one can treat a discrete geometry ( such as a geometric mesh ) as the input . This perspective where the discrete geometry is the actual geometry is that of DDG ( 109 ) . By eliminating the burdens of treating the input mesh as an approximation of a smooth object , DDG capitalizes upon the piecewise nature of meshes to produce ef \ufb01 cient and paralleliz - able \ufb01 nite difference - like formulae which are amenable to algorithmic implementation while main - taining clear geometric meaning . In the following sec - tions , we use concepts from DDG to formulate a discrete analog to the smooth membrane shape prob - lem . Following the derivation of the discrete theory , we describe the development of an accompanying soft - ware implementation called Mem3DG . Notation and preliminaries We assume the following notation conventions and provide a table of important symbols ( Table 2 ) . To aid the reader on how the elements of the mesh are used in the derivation , several fundamental geometric primitives ( i . e . , values on a mesh which are easily measurable ; listed in Table 2A ) are illustrated in Fig . 1 A \u2013 C . We note that , in discrete contexts , the notation , R a , should be considered the discrete ( integrated ) counter - part of a pointwise measurement a in a smooth setting . The rationale and signi \ufb01 cance behind the usage of an integrated quantity in discrete contexts are elaborated in Appendix B and the DDG literature ( 104 , 105 ) . Using this notation , discrete surface integrals are expressed as sums of integrated values over the discrete mesh components listed in Table 2B ( e . g . , P v i R a i is the discrete analog to R M a ) . It is possible to interchange 4 Biophysical Reports 2 , 100062 , September 14 , 2022 between integrated , R a i , and pointwise , a i , quantities by using the dual area ( A i ) , a i \u00bc Z a i = A i : ( 2 ) For simplicity , we will not use separate notations for operators applying in smooth and discrete settings . The context can be inferred from the objects to which the operators are applied . Where it serves to improve our geometric or other intuition , smooth objects will be presented alongside discrete objects for comparison . Obtaining a discrete energy de \ufb01 ned by mesh primitives Following the perspective of DDG , we restrict our input to the family of triangulated manifold meshes , M ( i . e . , discrete 2 - manifolds embedded in R 3 ) ( We will use M for both the smooth and discrete surfaces ) . Paralleling the smooth Helfrich Hamiltonian ( Eq . 1 ) , a functional of geometric measurements of a surface , the discrete Helfrich Hamiltonian is composed of discrete analog of those measurements , E \u00f0 ~ r \u00de \u00bc X v i (cid:5) k i Z \u00f0 H i (cid:2) (cid:1) H i \u00de 2 \u00fe k G Z K i (cid:6) | \ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 } E b \u00fe Z A (cid:1) A l \u00f0 ~ A ; ~ r \u00de d ~ A | \ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 } E s (cid:2) Z V (cid:1) V D P \u00f0 ~ V ; ~ r \u00de d ~ V | \ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 } E p : ( 3 ) In comparison with Eq . 1 , H i and K i are pointwise mean and Gaussian curvature measurements on vertices , R \u00f0 H i (cid:2) (cid:1) H i \u00de 2 is the integrated Willmore mea - sure , and the smooth surface integral is replaced by its discrete analog ( i . e . \ufb01 nite summation ) , P v i ( Table 2B ) . The geometric properties of a given membrane con \ufb01 guration can be connected to the system ' s energy through constitutive relations . In this work , we assume that the surface tension follows a linear stress - strain model ( 110 ) , l \u00f0 A ; ~ r \u00de \u00bc K A A \u00f0 ~ r \u00de (cid:2) (cid:1) A (cid:1) A ; ( 4 ) where (cid:1) A is the preferred surface area of the membrane , and K A is the stretching modulus of the membrane . The osmotic pressure can be de \ufb01 ned based on the van ' t Hoff formula as D P \u00f0 V ; ~ r \u00de \u00bc P in (cid:2) P out \u00bc iRT (cid:7) n V (cid:2) c (cid:8) ; ( 5 ) TABLE 2 Glossary of commonly used symbols and conventions A . Geometric primitives M smooth or discrete 2 - manifold ~ r \u02db R 3 embedded coordinate of M l edge length : corner angle 4 dihedral angle A area of mesh cell , e . g . , face A ijk , edge A ij and vertex A i ~ n surface normal B . Surface integral R a integrated quantity over mesh cell ; e . g . , A i a i or A ijk a ijk P v i sum over all vertices v i of the mesh P e ij sum over all edges e ij of the mesh P f ijk sum over all faces f ijk of the mesh P v j \u02db N \u00f0 a \u00de sum over the vertex v j in the neighborhood of a P e ij \u02db N \u00f0 a \u00de sum over the edges e ij in the neighborhood of a P f ijk \u02db N \u00f0 a \u00de sum over the face f ijk in the neighborhood of a C . Tensors x \u02db R scalar quantity x typeindex sub - and super - script convention ; e . g . , R ~ f b i is the bending force for vertex v i ~ x \u02db R 3 vector quantity x \u00bc f x i g \u00f0 n (cid:3) 1 \u00de indexed scalar quantity ~ x \u00bc f x i g \u00f0 n (cid:3) 3 \u00de indexed vector quantity ~ X matrix or tensor quantity D . Derivatives V ~ r shape derivative V f chemical derivative V ~ q surface gradient _ a time derivative D s Laplace - Beltrami operator E . Physical variables E energy f force density m chemical potential H mean curvature K Gaussian curvature A surface area V enclosed volume $ preferred state ; e . g . , (cid:1) H is the spontaneous curvature f \u02db \u00bd 0 ; 1 (cid:4) protein density parameter l membrane tension D P osmotic pressure across the membrane k bending rigidity k G Gaussian modulus K A stretching modulus K V osmotic strength constant (cid:1) c molar ambient concentration n molar quantity of enclosed solute h Dirichlet energy constant \u03b5 adsorption energy constant x membrane drag constant B protein mobility constant Biophysical Reports 2 , 100062 , September 14 , 2022 5 where i , R , T , c , and n are the van ' t Hoff index , ideal gas constant , temperature , ambient molar concentration , and molar amount of the enclosed solute . Substituting these constitutive relations ( Eqs . 4 and 5 ) into the en - ergy ( Eq . 3 ) , we get explicit expressions for E s and E p , E \u00f0 ~ r \u00de \u00bc E b \u00f0 ~ r \u00de \u00fe 1 2 K A \u00bd A \u00f0 ~ r \u00de (cid:2) (cid:1) A (cid:4) 2 (cid:1) A | \ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 } E s \u00fe iRTn \u00bd r c \u00f0 ~ r \u00de (cid:2) ln r c \u00f0 ~ r \u00de (cid:2) 1 (cid:4) | \ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 } E p ; ( 6 ) where r c \u00bc c = \u00f0 n = V \u00de is the ratio of the concentrations of the ambient and enclosed solutions . Note that the preferred volume , (cid:1) V , which is needed to evaluate the in - tegral in Eq . 3 , is related to to the parameters in Eq . 5 by (cid:1) V \u00bc n = c . If the system is around the isosmotic condi - tion ( e . g . , V / (cid:1) V ) , the leading order of the energy is given as , E p z 1 2 K V \u00f0 V (cid:2) (cid:1) V \u00de 2 (cid:1) V 2 ; ( 7 ) where K V h iRTn groups the phenomenological param - eters . Mathematically , Eqs . 4 and 7 effectively pre - scribe a penalty - based method for area and volume control . An alternative approach is the use of Lagrange multipliers , which have been extensively adopted in the literature ( 40 , 57 , 72 ) . To compute the energy of a system , we must obtain values for several geometric measurements that appear in thediscreteenergy function ( e . g . , H ; K ; A ; V ) . Formea - surements such as the volume and area , there are clear approaches for their evaluation on a triangulated mesh : summing the areas of the triangular faces and summing over the signed volume of tetrahedra ( Fig . 1 E , osmotic pressure and surface tension ) . For other measurements such as the discrete mean and Gaussian curvatures , FIGURE 1 Overview of the DDG framework For a Figure360 author presentation of this \ufb01 gure , see https : / / doi . org / 10 . 1016 / j . bpr . 2022 . 100062 . ( A \u2013 C ) Illustrations of geometric primitives in the neighborhood of ( A ) fan around a vertex , ( B ) diamond around an edge , and C ) triangle on a face . ( D ) Discrete de \ufb01 nition of scalar edge mean curvature , R H ij , scalar vertex Gaussian curvature , R K i , and Laplace - Beltrami operator , R D s \u00f0 $ \u00de . ( E ) Comparative derivation of Helfrich shape equation in both smooth and discrete formulation . 6 Biophysical Reports 2 , 100062 , September 14 , 2022 additional care must be taken . While in smooth contexts these curvatureshaveuniquede \ufb01 nitions , indiscretecon - texts there are multiple approaches for their calculation . For example , the mean curvature can be computed via the application of the cotangent Laplacian , the kernel of the heat equation , or \ufb01 tting polynomials to a local patch ( 65 ) . As mentioned earlier , there are challenges with these approaches that can limit their numerical ac - curacy and obscure the connection to smooth theory . Here using discrete exterior calculus and identi \ufb01 cation of geometric invariants , we produce theoretically and numerically consistent discrete expressions . Similar to the polygonal curve introduced in Appen - dix B , a triangulated mesh has zero curvature on facets and ill - de \ufb01 ned curvature on edges and vertices . Using the Steiner view , the sharp corners formed by vertices and edges are made smooth with portions of spherical and cylindrical shells , which have well - de \ufb01 ned mean curvature ( Fig . 1 D ) . Taking the limit as the radii of the cylinders and spheres decrease , the leading order contribution of total mean curvature is given by the Steiner formula on an edge , Z H ij \u00bc l ij 4 ij 2 ; ( 8 ) referred to as the edge mean curvature , where l ij is the length of edge e ij , and 4 ij is the dihedral angle on e ij ( i . e . , the angle formed by the face normals of the neigh - boring triangles incident to e ij ) ( illustrated in Fig . 1 B ) ( 104 , 105 ) . While not necessary , a triangulated mesh is often realized in R 3 via vertex positions ; thus it is con - ventional to prescribe data on vertices instead of edges . Summation of edgewise quantities over the \u201c fan \u201d neighborhood ( Fig . 1 A ) provides the recipe of converting an edgewise to a vertexwise quantity , \u00f0 $ \u00de i \u00bc 1 2 X e ij \u02db N \u00f0 v i \u00de \u00f0 $ \u00de ij ; ( 9 ) where the prefactor , 1 = 2 , accounts for fact that each edge is shared by two vertices . While we have an integrated mean curvature , the discrete Helfrich Hamiltonian contains a pointwise mean curvature squared term . To de \ufb01 ne a pointwise mean curvature , the size of the domain occupied by the integrated mean curvature needs to be speci \ufb01 ed ( cf . , Appendix B for rationale ) . The area , A i , referred to as the dual area of the vertex v i , can be de \ufb01 ned as one - third of the areal sum of the incident triangles ( fan illustrated in Fig . 1 A ) ( 104 , 105 ) . Applying Eqs . 2 and 9 to Eq . 3 , the pointwise mean curvature is thus , H i \u00bc R H i A i \u00bc X e ij \u02db N \u00f0 v i \u00de l ij 4 ij 4 A i : ( 10 ) Substituting Eq . 10 into the integrated Willmore mea - sure term of Eq . 3 , the integrated Willmore measure can be expressed as a function of the integrated mean and spontaneous curvature , Z \u00f0 H i (cid:2) (cid:1) H i \u00de 2 \u00bc 1 A i (cid:9) Z H i (cid:2) Z (cid:1) H i (cid:10) 2 : ( 11 ) Discrete Gaussian curvature is given by the angle defect formula , Z K i \u00bc 2 p (cid:2) X f ijk \u02db N \u00f0 v i \u00de : kij ; ( 12 ) which is a well - known quantity that preserves many properties parallel to the smooth theory ( e . g . , Gauss - Bonnet , turning number , among other invariants ) . One way to derive the angle defect formula is to compute the area of a spherical n - gon contained by a local Gauss map of the neighboring n faces around a vertex ( 104 , 105 ) . Eq . 12 provides the general geometric de \ufb01 nition to obtain the energetic contributions from the Gaussian curvature terms . In this study , we consider only sys - tems with uniform saddle - splay modulus which do not undergo topological changes . For these systems , the energy can be simpli \ufb01 ed based on the discrete Gauss - Bonnet theorem , which states that X v i Z K i \u00bc 2 pc \u00f0M\u00de (cid:2) X v j \u02db v M Z k gj ; ( 13 ) where c \u00f0M\u00de \u00bc j V j (cid:2) j E j \u00fe j F j is the Euler character - istic of M , a topological invariant where j V j , j E j and j F j represent the number of vertices , edges and faces of the mesh respectively , and R v M k gi \u00bc p (cid:2) P e ij \u02db N \u00f0 v i \u00de : kij is the discrete geodesic curvature , which measures the deviation of the boundary curve from a straight line when the surface is locally \ufb02 attened . In summary , for this work , the Gaussian curvature term is non - constant only when M is not closed , and the en - ergy solely involves the boundary elements . A numerical comparison of the discrete scalar mea - surements with their smooth counterparts is shown in Fig . E . 1 . We note that for all geometric measurements ( i . e . , volume , area , and curvatures ) , unlike in smooth dif - ferential geometry where their numerical evaluation requires the introduction of coordinates , DDG measure - ments are functions of mesh primitives . By substituting these discrete geometric measurements from DDG into Eq . 6 and 3 , we get a numerical recipe for computing the total system energy . Biophysical Reports 2 , 100062 , September 14 , 2022 7 Force from discrete shape derivative of energy We can obtain the force by taking the negative shape derivative of the energy . In continuous settings , the differentiation is an in \ufb01 nite - dimensional problem that requires the use of the calculus of variations and differential geometry to \ufb01 nd analytical expressions ( 39 , 43 , 70 , 71 ) ( Fig . 1 E , smooth ) . Deriving the forces from the discrete energy ( Eq . 3 ) is a much simpler task . Discrete forces can be obtained by taking partial de - rivatives of mesh primitives with respect to the vertex embedded coordinates , ~ r ( Fig . 1 E , discrete ) . Regarding notation , despite the computational differences , the differential operations in both the discrete and smooth contexts are called ( discrete ) shape derivatives and denoted as V ~ r \u00f0 $ \u00de due to the common geometric meaning . We note that the computation of discrete shape derivatives for membrane modeling has been described previously in the literature ( 87 , 89 ) . Also that there are many overlapping de \ufb01 nitions for discrete curvature , energy , and variations thereof in the graphics literature ( 111 \u2013 113 ) . Our work extends upon the prior art that evaluates derivatives algebrai - cally , by introducing simpli \ufb01 cations based upon the grouping of terms and identi \ufb01 cation of geometric ob - jects . These simpli \ufb01 cations have important implica - tions for improving the geometric understanding as well as the run - time and numerical performance of an implementation . At the high level , our goal is to express the forces on each vertex , given a set of physics , using geometric primitives and properties de \ufb01 ned on speci \ufb01 c mesh ele - ments . By grouping terms , we \ufb01 nd that the vertexwise forces arising from the different physics can be ex - pressed as weights that are functions of the input pa - rameters and system con \ufb01 guration , multiplied by basic geometric vectors . We will show that these terms have an exact correspondence to terms in the smooth shape equation ( Fig . 1 E ) . We remark that , in some sense , the force expressions are reminiscent of \ufb01 nite - difference equations , which approximate differentials as a linear combination of values at discrete points . This may have implications for the suitability of modeling smooth biological surfaces with discrete meshes . Force from osmotic pressure For the smooth geometry , the shape derivative of the enclosed volume yields the outward - pointing surface normal with its size equal to the local area element ( 114 ) . For a discrete mesh , the shape derivative of the volume is given by the face normal on triangular faces with its local area element equaling to the face area , which is referred to as the integrated face normal , R ~ n \u00f0 ijk \u00de ( Fig . 1 E , osmotic pressure ) ( 89 , 99 , 104 , 105 ) , where \u00f0 ijk \u00de denotes the symmetry under index permu - tation ( e . g . , a i \u00f0 jk \u00de means a ijk \u00bc a ikj ) . Similar to edge values , the force normal can be converted to vertex normal , Z ~ n i \u00bc V ~ r i V \u00bc 1 3 X f ijk \u02db N \u00f0 v i \u00de Z ~ n \u00f0 ijk \u00de \u00bc 1 3 X f ijk \u02db N \u00f0 v i \u00de A ijk ~ n \u00f0 ijk \u00de ; ( 14 ) where analogous to Eq . 9 , the prefactor 1 = 3 accounts for fact that each face is shared by three vertices . The discrete vertex forces from the derivative of the pressure - volume work , R ~ f p i , is then given by scaling it with the uniform osmotic pressure , Z ~ f p i \u00bc D P Z ~ n i : ( 15 ) Forces from surface tension Next , considering the shape derivative of the surface energy , E s , in smooth contexts , the derivative of the to - tal surface area also points at the surface normal , with its magnitude measuring the size ( dA ) and the curva - ture ( 2 H ) of the local patch ( Fig . 1 E , surface tension ) ( 114 ) . In a discrete case , we can directly compute the derivative of total area on each vertex by summing the area gradient of incident triangles with respect to the vertex position ; the sum is therefore referred to as ( twice of ) the integrated mean curvature vector on vertices , Z 2 ~ H i \u00bc V ~ r i A \u00bc X f ijk \u02db N \u00f0 v i \u00de V ~ r i A ijk \u00bc X f ijk \u02db N \u00f0 v i \u00de Z 2 ~ H i \u00f0 jk \u00de ; ( 16 ) where we de \ufb01 ne R 2 ~ H i \u00f0 jk \u00de h V ~ r i A ijk , and R ~ H i \u00f0 jk \u00de is the mean curvature vector on a triangle face corner . The capillary forces from surface tension , R ~ f s i , are given by scaling the integrated mean curvature vector by the surface tension , Z ~ f s i \u00bc (cid:2) l Z 2 ~ H i : ( 17 ) Evaluating the algebraic sum of area gradients re - veals the \u201c cotangent formula \u201d applied to the vertex posi - tions ( Fig . 1 E , surface tension ) . From independent derivations with unrelated frameworks ( e . g . , discrete exterior calculus and FEM ) , discretizing the smooth Lap - lace - Beltrami operator on a triangulated mesh produces the cotangent formula , which is called the discrete Lap - lace - Beltrami operator , R D s ( 104 , 105 , 111 ) . Inspecting 8 Biophysical Reports 2 , 100062 , September 14 , 2022 the expressions in Fig . 1 E , surface tension , we see that our discrete expression parallels smooth theory , where the mean curvature is related to the coordinates through the application of the smooth Laplace - Beltrami operator , X e ij \u02db N \u00f0 v i \u00de Z 2 ~ H ij \u00bc Z D s ~ r i 4 D s ~ r \u00bc 2 H ~ n : ( 18 ) Forces from bending To evaluate the shape derivative of the discrete bending energy , we consider the terms from the Gaussian and mean curvature separately . Since we do not consider nonuniform saddle - splay modulus and topological changes in this work , the total Gaussian curvature only varies if the surface has boundaries , v M ( cf . , discrete Gauss - Bonnet theorem in section \u201c Obtaining a discrete energy de \ufb01 ned by mesh primitives \u201d ) . The shape derivative of the bending energy requires more algebra and the introduction of halfedges , e ij ( cf . , Appendix C . 1 ) . Here we will focus on the key results and refer the reader to the full der - ivations for each term in Appendix C . 2 . There are four fundamental geometric vectors on halfedges that comprise the bending force at non - boundary vertices : the mean curvature vector ( see Fig . 1 B for indices ) , Z 2 ~ H ij \u00bc 1 2 (cid:9) Z 2 ~ H i \u00f0 jk \u00de \u00fe Z 2 ~ H i \u00f0 jl \u00de (cid:10) ; ( 19 ) the Gaussian curvature vector , Z ~ K ij \u00bc 1 2 4 ij V ~ r i l ij ; ( 20 ) and the two Schla \ufb02 i vectors , Z ~ S ij ; 1 \u00bc 1 2 l ij V ~ r i 4 ij ; Z ~ S ij ; 2 \u00bc 1 2 (cid:7) l jk V ~ r i 4 jk \u00fe l jl V ~ r i 4 jl \u00fe l ji V ~ r i 4 ji (cid:8) ; ( 21 ) which act to smooth the pro \ufb01 le of local dihedral angles . Note that the shape derivatives are taken with respect to different vertices ( i . e . , V ~ r i or V ~ r j ) , such that the mean curvature R ~ H ij , Gaussian curvature R ~ K ij , and Schla \ufb02 i vectors R ~ S ij are asymmetric under index permutation . To account for the orientation , we refer to them as half - edge vector quantities on e ij ( Appendix C . 1 ) . A numeri - cal comparison of the discrete geometric vector with their smooth counterparts is shown in Fig . E . 1 and Fig . E . 2 . The bending force R ~ f b i ( Fig . 1 E , bending ) can be ex - pressed as weights , which are functions of input parameters multiplied by basic geometric measure - ments in scalar and vector form , which parallels the shape derivative of the smooth bending energy , V t ~ r E b \u00bc V t ~ r (cid:5) Z M k \u00f0 H (cid:2) (cid:1) H \u00de 2 dA (cid:6) \u00bc k (cid:2) 2 \u00f0 H (cid:2) (cid:1) H \u00de (cid:11) H 2 (cid:2) K \u00fe (cid:1) HH (cid:12) \u00fe D s \u00f0 H (cid:2) (cid:1) H \u00de (cid:3) dA ; ( 23 ) where V t ~ r \u00bc V ~ r $ ~ n is the shape derivative in the surface normal direction . Comparing the smooth - discrete expressions , we make a few observations : (cid:5) The Schla \ufb02 i vector terms , ~ S , is the discrete analog of the smooth biharmonic term , D s \u00f0 H (cid:2) (cid:1) H \u00de , the high - order local smoothing force . The numerical compar - ison of these two terms , as well as results directly obtained using cotangent formula applied on Z ~ f b i \u00bc X e ij \u02db N \u00f0 v i \u00de (cid:2) (cid:2) k i \u00f0 H i (cid:2) (cid:1) H i \u00de \u00fe k j (cid:11) H j (cid:2) (cid:1) H j (cid:12)(cid:3)Z ~ K ij \u00fe (cid:5) 1 3 k i \u00f0 H i (cid:2) (cid:1) H i \u00de\u00f0 H i \u00fe (cid:1) H i \u00de \u00fe 2 3 k j (cid:11) H j (cid:2) (cid:1) H j (cid:12)(cid:11) H j \u00fe (cid:1) H j (cid:12)(cid:6) Z 2 ~ H ij (cid:2) (cid:5) k i \u00f0 H i (cid:2) (cid:1) H i \u00de Z ~ S ij ; 1 \u00fe k j (cid:11) H j (cid:2) (cid:1) H j (cid:12) Z ~ S ij ; 2 (cid:6) ; ( 22 ) Biophysical Reports 2 , 100062 , September 14 , 2022 9 pointwise scalar mean curvature , are covered in Fig . E . 2 and Fig . E . 1 . (cid:5) Eq . 23 is the normal component of the shape deriva - tive of the bending energy ; an additional tangential component is required if surface heterogeneity ex - ists ( e . g . , k is not spatially uniform ) ( 40 , 65 ) . In contrast , the discrete shape derivative ( Eq . 22 ) is the total derivative in R 3 , which includes both the tangential and normal components ( in the smooth sense since there is no well - de \ufb01 ned vertex normal di - rection in discrete geometry ) . Depending on the extent and symmetry of the heterogeneity , the discrete force can point in any direction in R 3 . (cid:5) The coef \ufb01 cients in Eq . 22 show an intriguing pattern combining contributions from both v i and v j . From a \ufb01 nite - difference approximation standpoint , this re - sults in an approximation scheme for which a rigorous error analysis has not yet been conducted . FIGURE 2 Overview of data \ufb02 ow within Mem3DG . The user provides Mem3DG with an initial condition in the form of a triangulated mesh and vertexwise state and kinematic variables ( green box ) . The main loop ( black loop ) of Mem3DG evaluates the discrete energy and forces and prop - agates the trajectory , among other supporting steps . Modules in dashed lines are optional depending on whether the system mesh has bound - aries and if external forces are speci \ufb01 ed . User - speci \ufb01 ed options and possible extensions of Mem3DG to accommodate various physics are highlighted in yellow boxes . Mem3DG automatically exits the simulation when the system converges or the maximum time step is reached . 10 Biophysical Reports 2 , 100062 , September 14 , 2022 Net force and the bene \ufb01 t of DDG By summing the force terms from each physics , we obtain the net force . Through section \u201c Obtaining a discrete energy de \ufb01 ned by mesh primitives \u201d and sec - tion \u201c Force from discrete shape derivative of energy , \u201d we identify and show a scheme where both 1 ) the force is analytically derived from the discrete energy , and 2 ) both the discrete energy and force mirror the smooth theory . The entire process of de \ufb01 ning energy and con - duction shape derivative do not involve the introduction of coordinate and the use of tensor algebra . Owing to the discontinuities and limited information contained by a discrete geometry , there are ambiguities in geo - metric de \ufb01 nitions that behave otherwise in the smooth context ( cf . , various discrete curvature de \ufb01 nitions for plane curve discussed in Crane and Wardetzky ( 115 ) ) . Intentional choices of certain de \ufb01 nitions of basic discrete geometric measurements can reveal the connection between various de \ufb01 nitions , preserve use - ful geometric invariants , and most naturally re \ufb02 ect the underlying physics . Here many scalar and vector de \ufb01 nitions of geometric measurement are connected through the chain of shape derivatives ( cf . , Fig . A . 2 ) ( 105 ) , which justi \ufb01 es their role in representing either en - ergy or forces . For example , the discrete bending en - ergy is also commonly constructed using the mean curvature vector , R M ~ H $ ~ H dA in literature ( 90 , 99 ) . As shown in section \u201c Forces from surface tension , \u201d the de \ufb01 nition of the mean curvature vector is tightly corre - lated with the surface tension , where directionality is embedded . The inner product used in such discrete en - ergy de \ufb01 nition strips away the directional information . Instead , here we construct energy using the scalar mean curvature , R H , because 1 ) the energy is inher - ently scalar , and 2 ) the discrete curvature exists on the edges . After taking the shape derivative of the en - ergy , the mean curvature vector appears as the effec - tive tension component and the directional information of the vector is used for representing the force . Using the directional information , an arbitrary de \ufb01 nition of a vertex normal is avoided . When weighted homogeneously around a vertex , each fundamental geometric vector that composes the discrete force , R ~ n , R ~ H , R ~ K , R ~ S , can be used to obtain a meaningful de \ufb01 nition of the vertex normal . The heterogeneous weighting of these vectors around the vertex repre - sents the incorporation of functional variation in the tangential direction , which can be used to model het - erogeneities in material and other properties across the membrane ( 40 , 65 ) . Practically , the additional struc - ture provided by the discrete force and energy expres - sions allows the user to inspect term - wise contributions , which can lead to additional insights and analysis . Since the terms of the discrete energy and forces are de \ufb01 ned locally by mesh primitives at vertex neighborhoods , the algorithms are ef \ufb01 cient and straightforward to parallelize . The numerical accuracy of these expressions is benchmarked for several scalar and vector measurements on smooth and discrete sur - faces shown in Fig . E . 1 , Fig . E . 2 , and later discussed in section \u201c Practical considerations for applying Mem3DG to biological problems . \u201d De \ufb01 ning metrics for simulation and error quanti \ufb01 cation For monitoring simulation progress , exactness of force calculations with respect to the discrete energy , and convergence studies of computed quantities upon mesh re \ufb01 nement , we introduce the following norms . L 2 norm From a PDE perspective , the vertex forces are also called the residual of the shape equation , whose solu - tion represents the equilibrium solution . The simulation is terminated when the residual is smaller than a user - speci \ufb01 ed threshold . The rationale for using the L 2 norm is justi \ufb01 ed by perturbing the system con \ufb01 guration and conducting an expansion on the system energy , E \u00f0 ~ r \u00fe e V E \u00f0 ~ r \u00de\u00de \u00bc E \u00f0 ~ r \u00de \u00fe e h V E \u00f0 ~ r \u00de ; V E \u00f0 ~ r \u00dei \u00fe O (cid:11) e 2 (cid:12) \u00bc E \u00f0 ~ r \u00de \u00fe e (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)Z ~ f (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) 2 L 2 \u00fe O (cid:11) e 2 (cid:12) ; ( 24 ) where we refer the inner product of the force matrix as the L 2 norm of the forces . Computationally , this is equivalent to the standard Frobenius matrix L 2 norm , (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)Z ~ f (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) L 2 \u00bc \ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03 trace (cid:9) Z ~ f u Z ~ f (cid:10) s : ( 25 ) Using the L 2 norm and Eq . 24 , we can perform a nu - merical validation of the exactness of the discrete force calculation with respect to the discrete energy . We expect the force to approximate the energy up to second order with respect to the size of a perturbation . This validation will be further elaborated in section \u201c Membrane dynamics with full mechanochemical feedback . \u201d L 1 norm A scale - invariant L 1 norm is well suited to quantify the magnitude of the error on varying domain size and mesh resolution . Given a vertexwise local scalar mea - surement , R a , or a vector measurement , R ~ a , and their reference values , R (cid:1) a , and R (cid:1) ~ a , Biophysical Reports 2 , 100062 , September 14 , 2022 11 (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)Z a (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) L 1 \u00bc X v i (cid:13)(cid:13)(cid:13) Z a i (cid:2) Z (cid:1) a i (cid:13)(cid:13)(cid:13) A (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)Z ~ a (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) L 1 \u00bc X v i (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) Z ~ a i (cid:2) Z ~ a i (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) L 2 A ; ( 26 ) where the normalizing factor , the total surface area A , is used to obtain a pointwise estimate of the error . The L 1 norm is applied in the local comparison of discrete and smooth measurements , which we further elaborate in section \u201c Practical considerations for applying Mem3DG to biological problems . \u201d SOFTWARE IMPLEMENTATION : Mem3DG Along with the theoretical developments , we have developed an accompanying software implementation written in C \u00fe\u00fe called Mem3DG . Our goal in developing this software is to enable the easy use and application of the corresponding theory developed above to biolog - ical problems of interest . Mem3DG is a library that contains several compo - nents to support this goal . Fig . 2 provides a synopsis of Mem3DG . The input to Mem3DG includes a triangu - lated mesh with its coordinate ~ r embedded in R 3 . Users can choose to use Mem3DG to construct idealized meshes ( e . g . , icosphere , cylinder , or \ufb02 at hexagonal patch ) as an input or to read in meshes from several common mesh formats . Meshes are stored and manipulated in Mem3DG using the halfedge data struc - ture provided by Geometry Central ( 116 ) . The supported input \ufb01 le formats are those which are read - able by hapPLY and Geometry Central ( 116 , 117 ) . Once a mesh and parameters are loaded , Mem3DG can evaluate the discrete energy and forces of the sys - tem . Mem3DG adopts a modular design that facilitates the use of different energy and force components and has utilities which help the user to specify the phys - ics and governing parameters . Mem3DG also supports local system simulations where the input mesh has boundaries . Additional details about the supported boundary conditions are given in section \u201c Prescribing boundary conditions with force masking . \u201d To perform energy minimization and time integration of the system , various schemes have been imple - mented . These schemes are described in section \u201c Time integration and energy minimization . \u201d As dis - cussed further in section \u201c Practical considerations for applying Mem3DG to biological problems , \u201d when a user wishes to use Mem3DG to represent complex biological membrane geometries , additional care regarding the quality of the mesh is necessary . Mem3DG includes algorithms for basic mesh regulariza - tion and remeshing , which can be toggled by the user to support their applications . The simulation terminates when it reaches the time limit or the system reaches equilibrium , whose criteria is determined using the en - ergy L 2 norm introduced in section \u201c De \ufb01 ning metrics for simulation and error quanti \ufb01 cation . \u201d A user can choose between several formats to output a trajectory over time or the con \ufb01 guration of the local minima from Mem3DG . In addition to the mesh outputs supported by Geometry Central , we have also developed a scheme for outputting mesh trajectories in NetCDF format ( 118 ) . Mem3DG can read and visualize the output trajectories and mesh con \ufb01 gurations using Ge - ometry Central and Polyscope ( 116 , 119 ) . For rapid prototyping and enumeration of simulation conditions , we have also developed a Python API called PyMem3DG . The functionality in C \u00fe\u00fe is exposed in Py - thon using bindings from pybind11 ( 120 ) . Illustrative examples of using both Mem3DG and PyMem3DG are provided in the online tutorials . For the experiments dis - cussed in this work , all of the simulations were per - formed using PyMem3DG and the accompanying code and initial con \ufb01 gurations are on GitHub : https : / / github . com / RangamaniLabUCSD / Mem3DG . De \ufb01 ning properties of a membrane reservoir for systems with open boundaries To facilitate correspondence with wet experiments and to support the reduction of computational cost , it is possible to construct systems using meshes with open boundaries in Mem3DG . For example , when modeling the formation of a small endocytic bud from a large cell , the deformation is small compared with the broader system . If we assume that the bulk of the cell is invariant with respect to bud formation , the computational burden can be reduced by modeling only the local deformation ; we can assume that the modeled patch is attached to an implicit membrane reservoir . To de \ufb01 ne this coupled system , the constant area ( A r ) and volume ( V r ) of the reservoir must also be provided . The total area and volume of the broader system is given by A \u00bc A patch \u00fe A r , and V \u00bc V patch \u00fe V r , where A patch and V patch are area and \u201c enclosed volume \u201d of the mesh patch respectively . In our models , we enforce that all elements of a boundary loop are on the same plane ; this way V patch can be unambiguously de \ufb01 ned as the enclosed volume when each boundary loop is closed by a planar sheet . The capability to model systems attached to a reservoir re - duces the modeled degrees of freedom while enabling intuitive physics to simplify the process of mimicking experimental conditions using Mem3DG . 12 Biophysical Reports 2 , 100062 , September 14 , 2022 Prescribing boundary conditions with force masking Mem3DG supports modeling membranes with and without boundaries : a sphere ( with no boundaries ) , a disk ( with one boundary ) , and an open cylinder ( with two boundaries ) . For systems without boundaries , the discrete forces conserve angular and translational mo - mentum of system ( as was also noted by Bian et al . ( 89 ) ) . Because the ( discrete ) potential energy is invariant under rigid body motions ( i . e . , the energy of the membrane is given only by the geometry ) , and the discrete forces are analytically derived from the energy , the discrete forces will not contribute to any rigid body motions since these components do not change sys - tem energy . In other words , the forces that lead to rigid body motion are , by construction , orthogonal to the shape derivative of the potential energy . To study sys - tems with boundaries , Mem3DG currently supports three types of boundary conditions : (cid:5) Roller , where the movement of boundary vertices is restricted along a given direction or plane . (cid:5) Pinned , where the position of boundary vertices are pinned while the curvature is allowed to vary . (cid:5) Fixed , where both the position and the boundary cur - vature are \ufb01 xed for vertices on the boundary . The different boundary conditions are achieved by masking the elements of the force matrix correspond - ing to the boundary vertices and their neighborhood . For example , to apply roller boundary conditions , we mask the Z - component of the force on the boundary vertices , therefore constraining their movement to the X - Y plane ; pinned boundary conditions mask all force components for the boundary vertices to \ufb01 x their posi - tion ; \ufb01 xed boundary conditions mask all force compo - nents for the outermost three layers to \ufb01 x both their position and curvature . Time integration and energy minimization In this work , we use the forward Euler algorithm to inte - grate the system dynamics and the nonlinear conjugate gradient method to solve for equilibrium conditions . Both solvers are complemented by a backtracking line search algorithm , which satis \ufb01 es Wolfe conditions to support adaptive time - stepping and robust minimi - zation ( 121 ) . The forward Euler scheme was chosen as the simplest dynamical propagator ; physically it represents over - damped conditions where the environment of the membrane is too viscous for the system to carry any inertia . Mathematically , the physics is described by , _ ~ r \u00bc 1 x Z ~ f \u00bc 1 x Z (cid:7) ~ f b \u00fe ~ f s \u00fe ~ f p (cid:8) ; ( 27 ) where x is the drag coef \ufb01 cient . From an optimization perspective , forward Euler is equivalent to the gradient descent method for minimizing an objective function , which is the discrete energy in our case . A second propagator is the nonlinear conjugate gradient method for locally minimizing the discrete en - ergy to yield the equilibrium shape of the membrane . Since the system is nonlinear , we periodically perform forward Euler ( gradient descent ) steps after several conjugate gradient steps . This approach of iterating be - tween conjugate gradient and gradient descent steps is commonplace in the literature for solving nonlinear systems ( 121 ) . We note that other time integrators and energy min - imizers are also compatible with Mem3DG . Included in the software are reference implementations of velocity Verlet integration ( for symplectic time integration ) , and limited - memory Broyden - Fletcher - Goldfarb - Shanno al - gorithm ( L - BFGS , a quasi - Newton method to the equi - librium shape for large - scale problems where fast computation is needed ) . We do not discuss these addi - tional solvers in this work . Practical considerations for applying Mem3DG to biological problems As we have noted above , in the DDG perspective , the mesh is the geometry and thus the formulation of the discrete forces and energies is exact . There are there - fore very few restrictions on the resolution and quality of the input mesh . However , in biophysics , we often consider biological membranes as smooth systems . We expect that many users of Mem3DG may wish to approximate a smooth system using our discrete model . In doing so , they make an implicit assumption that such an approximation is reasonable . Although the relationships between geometric objects and their shapes are preserved between the smooth and discrete contexts , our ability to approximate a smooth problem with a discrete mesh is not guaranteed . Similar to \ufb01 nite differences and FEM , additional con - straints on mesh quality and resolution must be imposed . To verify and understand the limitations of the assumption that the discrete mesh is the geometry and includes all of the geometric information , we numerically test the convergence of the discrete quantities under variation of resolution on an oblate spheroid mesh . The additional details regarding these numerical experiments are presented in Appendix D . Setting the characteristic length scale of the \ufb01 nest mesh to be h \u00bc 1 , as the mesh coarsens ( i . e . , mesh size increases ) h increases . Fig . 3 shows the scaling relationship of the deviation in magnitude between the smooth and discrete quantities . Fig . 3 A shows the convergence property of global measurements Biophysical Reports 2 , 100062 , September 14 , 2022 13 that determines the energy ( Eqs . 1 and 3 ) , including the total area , A , enclosed volume , V , and total Gaussian curvature and mean curvature ( squared ) , R M KdA , R M HdA and R M H 2 dA , respectively . Except for the total Gaussian curvature being an exact topological invariant , all integrated quantities exhibit approximately second - order convergence rate . We acknowledge that convergence of global mea - surements does not imply that local measurements will also converge . To validate the convergence of local measurements , which determines the conver - gence of local forces on the membrane ( e . g . , Eqs . 15 , 17 , and 22 ) , we utilize the L 1 norm ( Eq . 26 ) to eval - uate the deviation of local quantities from their smooth counterparts . Fig . 3 B shows the local conver - gence plot . Similar to their global counterparts , local scalar mean and Gaussian curvature , R H , and R K , converge at O\u00f0 h 2 \u00de . Fig . 3 B also shows the conver - gence of vector quantities , which not only contribute to the magnitude of the force but also set the direc - tion of the force . The test shows that most vector quantities converge slightly slower than their scalar counterparts . Two terms exhibit poor convergence , the Schla \ufb02 i vector term in Eq . 22 , H R S , and a scalar counterpart , R D s H . The latter term corresponds to the direct application of the cotangent Laplacian ( Eq . 17 ) to the pointwise scalar mean curvature \ufb01 eld ; this approach is not used in our framework but is common in the literature ( 65 ) . Both non - convergent expressions are discrete representations of the bihar - monic term , D s H , which have been noted to be sensi - tive to noises of vertex coordinates in the prior literature ( 100 ) . Recall that the biharmonic term is the fourth - order derivative of the embedded coordi - nates . Although the traditional approximation theories suggest that higher - order derivatives often exhibit slower rates of convergence ( 122 ) , to the best of our knowledge , there is not yet a rigorous study that connects DDG with an approximation theory . Never - theless , we anticipate that similar principles hold . Two spatial plots comparing local measurements be - tween smooth and discrete contexts are provided in the appendix ( Fig . E . 1 and Fig . E . 2 ) ; each test is con - ducted using the \ufb01 nest mesh size ( h \u00bc 1 ) . Based on this numerical validation , we conclude that the computation of energy converges with a second - order rate ( Fig . 3 A ) . While most components of the forces converge , the biharmonic term remains a limiting factor . One other practical consideration for our models is that the Helfrich Hamiltonian , matching the in - plane \ufb02 uidity of biological membranes , has no resistance to shearing . Without additional constraints , the mesh is susceptible to shearing motions , which can deteriorate mesh quality in some conditions ( 83 ) . This can degrade the implicit assumption that the discrete mesh represents a smooth geometry . To ensure that such an approximation can remain valid throughout a trajectory , we have incorporated algorithmic solutions to adaptively maintain an iso - tropically well - resolved discrete geometry . This is achieved by two operations : 1 ) mesh regularization using local force constraints , which are common in FEM ( 78 , 82 , 83 , 85 ) ( Appendix E . 2 ) ; and 2 ) mesh muta - tions such as decimating , \ufb02 ipping , and collapsing edges . Beyond regularization , these local force con - straints can also be used to model underlying physics of a problem of interest . For example , similar restoring forces between vertices ( Eq . E . 1 ) have been adopted to model actin - spectrin cortex in red blood cell ( 95 ) . Mesh mutations are also a common practice to cope with deterioration and a means to perturb system con \ufb01 guration in other mesh simula - tions that use a Monte Carlo integration ( 60 , 89 \u2013 95 ) . The algorithms for mesh regularization and mutation are further described in Appendix E . FIGURE 3 Comparison of discrete quantities with their smooth counterparts on spheroid shape . ( A ) Convergence plot of global quanti - ties , including total area , volume , mean curva - ture ( squared ) , and Gaussian curvature ; and ( B ) Convergence plot of L 1 norm of scalar and vector local quantities , including the mean curvature , Gaussian curvature , and the biharmonic term . 14 Biophysical Reports 2 , 100062 , September 14 , 2022 RESULTS AND DISCUSSION To further validate the method and to provide a sense of how Mem3DG can be used and extended to solve more complex physics , we apply Mem3DG to a sequence of ex - amples with increasing complexity . First , we model well - studied systems with homogeneous membrane conditions . We show that Mem3DG is capable of repro - ducing the classical solutions without imposing the axisymmetric constraint commonly adopted by other solvers . The later examples set a blueprint for extending and modifying Mem3DG for particular systems of inter - est . We introduce new energy and corresponding force terms to expand the formulation for complex systems of interest . We emphasize that the goal of these exam - ples is to illustrate the generality of the theory and soft - ware and to outline speci \ufb01 c steps for future extensions ; we do not perform rigorous experimental comparisons , nor do we extract scienti \ufb01 c insights . Additional care must be taken to mimic speci \ufb01 c biological experiments for model validity , which is left for future work . Each of the following sections considers a different class of membrane biophysics problem of increasing complexity in the coupling of the in - plane protein den - sity parameter , f \u02db \u00bd 0 ; 1 (cid:4) . To mimic the various in \ufb02 u - ences protein - lipid interactions can have on the bilayer , the protein density can be set to in \ufb02 uence membrane properties such as the spontaneous curva - ture , (cid:1) H \u00f0 f \u00de , and bending rigidity , k \u00f0 f \u00de . More complex phenomena such as the production of in - plane interfa - cial forces from membrane - protein phase separation ( 55 , 59 , 123 ) can also be modeled . In our \ufb01 nal proof of concept , we extend Mem3DG to support full mechano - chemical dynamics , where proteins can mobilize in and out of plane through adsorption and lateral diffu - sion , based on its coupling with membrane material properties and shape transformation . These scenarios highlight the relative ease of extending Mem3DG with additional physics and the potential utility to simulate realistic experimental scenarios . Note that , for all of the examples , unless otherwise speci \ufb01 ed , the bending rigidity of membrane , k , is assumed to be the rigidity of a bare membrane , k b \u00bc 8 : 22 (cid:3) 10 (cid:2) 5 m m $ nN . Despite the superior performance of the nonlinear con - jugate gradient method in \ufb01 nding an energy minimizing con \ufb01 guration , to maintain both static and dynamic interpretability , we perform all simulations using a for - ward Euler integrator unless otherwise noted . All simu - lations presented in this work were conducted on a standard workstation with Intel Xeon processors . Although the numerical algorithms are amenable to parallelization , Mem3DG is currently a single - threaded program . Using a single core , the simulations here complete in minutes and up to 2 hours . Modeling spherical and cylindrical membranes with homogeneous physical properties We begin our examples by using Mem3DG to \ufb01 nd the equilibrium shapes of membranes with homogeneous protein density , f . We ask , given an initial membrane FIGURE 4 Recovertypical equilibriumshapes of membranes with homogeneous material properties . ( A and B ) Equilibrium solutions un - der different osmolarity ( (cid:1) c ) and spontaneous curvature ( (cid:1) H ) conditions , with initial condition of ( A ) Oblate spheroid and ( B ) Prolate spheroid . Wevarytheosmolaritybyadjustingtheconcen - tration of the ambient solution , (cid:1) c , while holding the enclosed amount of solute , n , constant . ( C ) Equilibrium solutions of a tubular membrane structureundervariationsinosmolarityandsur - face tension . Biophysical Reports 2 , 100062 , September 14 , 2022 15 con \ufb01 guration with uniform bending modulus and spon - taneous curvature , what are the minimizers of the sys - tem energy ? The answers are the classical equilibrium solutions to the shape equation obtained analytically ( 42 ) , and numerically using many methods with different assumptions ( 39 , 124 ) . We will show solutions obtained using Mem3DG with topologies of sphere and tube ( Fig . 4 ) . These geometries were selected not only because of their potential for comparison with the leg - acy literature but also because they are reminiscent of biological membranous structures such as red blood cell ( 97 , 98 , 125 , 126 ) , cell - cell tunneling and tethering ( 127 \u2013 129 ) , and neuron beading ( 130 , 131 ) , among other biological processes . Starting with closed topological spheres , Fig . 4 A and B shows the typical equilibrium shapes under osmotic stress while the surface area is conserved . The preferred area of the vesicle , (cid:1) A \u00bc 4 p m m 2 , represents a sphere of radius 1 m m . This constraint is achieved by prescribing a large stretching modulus , K A , such that the areal strain , \u00f0 A (cid:2) (cid:1) A \u00de = A , is less than 1 % . The strength constant of osmotic pressure , K V is set to be 0 . 1 m m $ nN . Initializing the simulations from an oblate spheroid , as the osmolarity increases ( e . g . , the normalized ambient solution , (cid:1) c = n ) , we recover the well - known biconcave red blood cell shape ( 97 , 98 , 106 , 124 ) ( Fig . 4 A ) . The vesicle adopts a more convex con \ufb01 guration as we increase the spontaneous curvature , indicating an overall increase in its mean cur - vature with the concomitant decrease of areas with negative mean curvature ( the dimple regions ) . In contrast , starting from a prolate spheroid , as the spon - taneous curvature increases , the vesicle adopts a dumbbell con \ufb01 guration as the energetically preferred state ( Fig . 4 B ) . The size of the beads on the dumbbell is governed by the osmolarity , (cid:1) c = n . These trends with respect to the variations of the spontaneous curvature and osmolarity are consistent with the analytical and numerical results in the broader literature ( 42 , 89 ) . Qual - itatively the predicted geometries of closed vesicles with homogeneous spontaneous curvature match the predictions of a detailed benchmark of mesh - based methods performed by Bian et al . ( 89 ) . We also modeled the shapes of membranes starting from an open cylinder con \ufb01 guration under different os - motic and surface tension conditions ( Fig . 4 C ) . This problem is related to a well - studied phenomenon called the Plateau - Rayleigh instability ( 132 , 133 ) . The Plateau - Rayleigh instability describes how surface tension breaks up a falling stream of \ufb02 uid into liquid droplets . Compared with a liquid stream , a lipid membrane pro - vides additional resistance against instability due to its rigidity . Zhongcan and Helfrich ( 134 ) obtain stability regimes as a function of membrane bending rigidity and spontaneous curvature using the spectral stability analysis ( 134 ) . Although osmotic pressure is often re - ported as an important cause of morphological insta - bility ( 131 , 135 \u2013 137 ) , the effect of osmotic pressure is dif \ufb01 cult to isolate in wet experiments because the change to osmolarity affects the surface tension , which is a key driver of the instability . In our simula - tions , the two effects are decoupled , making the inves - tigation of individual contributions to the morphology possible . All shapes in Fig . 4 C evolve from the initial tubular mesh with radius of 1 m m and axial length of 19 . 9 m m , under a constant spontaneous curvature of 1 m m (cid:2) 1 . These simulations are set up as local models ( cf . , section \u201c De \ufb01 ning properties of a membrane reser - voir for systems with open boundaries \u201d ) where the explicit mesh is assumed to be coupled to a membrane reservoir . Additional geometric information de \ufb01 ning the membrane reservoir and boundary conditions are required to initialize the local model . The tubular struc - ture is considered to be a cylinder that connects two otherwise detached domains ( e . g . , membrane reser - voirs ) , which combined have a total reservoir volume , V r \u00bc 4 : 19 m m 3 . The strength of osmotic pressure , K V , is set to be 0 . 01 m m $ nN . To isolate the effect of os - motic pressure and surface tension on the morphology , we prescribe a speci \ufb01 c surface tension that we as - sume to be invariant with respect to changes to the sur - face area . On the two boundary loops of the mesh , we apply roller boundary conditions , which restrict the movement of boundary vertices in the axial direction . The length of the tube is thus constrained to be 19 . 9 m m , while the radius of the tube including the boundaries is free to shrink or expand . As the osmolarity increases from the reference con - dition ( (cid:1) c = n \u00bc 0 : 022 m m (cid:2) 3 ) ( Video S1 ) , we obtain near constant - mean - curvature surfaces such as unduloid pearl structure at (cid:1) c = n \u00bc 0 : 030 m m (cid:2) 3 ( Video S2 ) , and cylindrical tube at (cid:1) c = n \u00bc 0 : 051 m m (cid:2) 3 , which follow the trends from both analytical ( 42 , 138 ) and experi - mental observations ( 19 , 131 , 135 ) . As we increase the surface tension from the reference condition ( l \u00bc 1 (cid:3) 10 (cid:2) 7 nN $ m m (cid:2) 1 ) to a tension - dominated regime ( l \u00bc 1 (cid:3) 10 (cid:2) 4 nN $ m m (cid:2) 1 ) , we obtain the beads - on - a - string structure that minimizes the sur - face - to - volume ratio ( Video S3 ) . The formation of beads - on - a - string is an interesting con \ufb01 guration that has been identi \ufb01 ed in biological membranes and other systems ( 130 , 131 ) . Note that our simulations revealed a symmetric metastable state where two large beads form at either end ( Appendix A ) , connected by a thin tube , prior to adopting the asymmetric conformation shown in Fig . 4 C . We believe that discretization arti - facts such as mesh mutations act as a perturbation to break the symmetry of the metastable intermediate and transition the membrane to a single bead con \ufb01 gu - ration ( see Fig . A . 1 ) . 16 Biophysical Reports 2 , 100062 , September 14 , 2022 These examples with uniform spontaneous curvature pro \ufb01 le prove the ability of Mem3DG to reproduce the ex - pected classical solutions for spherical and tubular membrane geometries . Note that no axisymmetric constraint is imposed in these simulations . Mem3DG sol - ves the system in full three dimensions and the symmet - rical con \ufb01 gurations are due to the problem physics . The ability to adapt to changing and complex curvatures of the membrane using discrete mesh is achieved using mesh mutation and other manipulations within solver steps . For example , the pinched neck regions of the tubes are automatically decimated with \ufb01 ner triangles than other regions of the mesh . For a global closed membrane simulation such as in Fig . 4 A , B , we do not remove any rigid body motions from the system ; since the forces from DDG are exact and we used the forward Euler integrator , no arti \ufb01 cial rigid body motions are intro - duced throughout the simulation . These examples show that that the derivation of the discrete energy and forces along with the software implementation are accurate and proceed to test Mem3DG with more complex examples . Modeling endocytic budding mechanisms Our goal is to highlight the potential of Mem3DG and its associated framework for investigating mechanical phenomena relevant to cellular biology . Endocytosis is a cellular process in which cells uptake cargo from the extracellular environment ; the transported material is engulfed by the cell membrane , which then buds off to form a vesicle ( 13 ) . Endocytosis occurs through various mechanisms , including clathrin - mediated endocytosis ( 13 , 139 ) . It has been shown that clathrin aggregates on the plasma membrane , helping to deform the membrane and form a spherical bud ( 9 , 13 , 59 ) . However , the speci \ufb01 c mechanisms of how membrane - clathrin interactions facilitate membrane curvature generation remain unresolved . While there is signi \ufb01 cant literature investigating the many pro - posed mechanisms , here we develop models to demonstrate the bud formation via spatially localized spontaneous curvature , combined with a line tension term arising from phase separations on the membrane ( 140 ) . We model endocytic budding on a circular patch with radius 1 m m ( a disc with one boundary loop ) . We as - sume that the patch is a local system which is coupled to a large vesicle ( section \u201c De \ufb01 ning properties of a membrane reservoir for systems with open bound - aries \u201d ) . A heterogeneous protein density , f \u02db \u00bd 0 ; 1 (cid:4) , is applied to mimic the distribution of clathrin and other scaffolding proteins . Shown in Fig . 5 A , the protein den - sity is high ( f \u00bc 1 ) toward the center of a geodesic disk with radius 0 . 5 m m ) and decreases toward the boundaries ( f \u00bc 0 ) . During simulation , the geodesic distance to the center of the patch is periodically computed using the heat method ( 141 ) . Vertexwise f is assigned based on the stair - step pro \ufb01 le smoothed by the hyperbolic tangent function applied to the geodesic distance . Each experiment is initialized as a \ufb02 at patch and the displacement of boundary vertices is restricted using a \ufb01 xed boundary condition . Since the patch is viewed as a small piece within a larger closed vesicle reservoir , we assume that the surface tension is constant . A common model to account for the preferential bending owing to protein - membrane interactions is through the spontaneous curvature ; we assume (cid:1) H \u00f0 f \u00de \u00bc (cid:1) H c f , where (cid:1) H c \u00bc 6 m m (cid:2) 1 is the spontaneous curvature imposed by the membrane protein coat . Pro - teins such as clathrin are known to form stiff scaffolds on the membrane . Similar to the spontaneous curva - ture , we can assume a linear relationship between bending rigidity and protein density , k \u00f0 f \u00de \u00bc k b \u00fe k c f , where constant k b is the rigidity of the bare membrane , and k c is additional rigidity of the protein scaffold . Shown in Fig . 5 A \u2013 C and Video S4 , is the control simulation where we set the contribution to the rigidity from protein to be the same as that of the raw mem - brane , k c \u00bc k b . Fig . 5 A shows the initial \ufb02 at con \ufb01 gura - tion of the control experiment ; the color bar shows the heterogeneous spontaneous curvature resulting from the prescribed protein density pro \ufb01 le . In the control experiment , the bending force is resisted by the surface tension ( Fig . 5 C ) until , at the \ufb01 nal frame in Fig . 5 B ( t \u00bc 5 ) , the membrane reaches the equilibrium con \ufb01 guration where the surface tension cancels with the bending force . In a second model , we assume that the scaffolding proteins are much more rigid than the bare membrane , k c \u00bc 3 k b . Fig . 5 D \u2013 F and Video S5 show the bud formation due to this increased protein scaffolding effect . The greater rigidity results in an increase of initial bending energy , which outcom - petes the resistance from the surface tension ( Fig . 5 F ) . Fig . 5 E shows the shape evolution from a \ufb02 at patch to a successful bud with a pinched neck . Fig . 5 D shows the signed projection of the bending force onto the ver - tex normal , R f bi \u00bc R ~ f b i $ ~ n i , at T \u00bc 15 . ( Outward - point - ing angle - weighted normal ; the same applies to the interfacial line tension . ) We can see an \u201c effective line tension \u201d driven by the heterogeneous spontaneous curvature that constricts the neck . This phenomenon is theoretically explored in detail by Alimohamadi et al . ( 58 ) . For our third model , based on the prior observations that protein phase separation on surfaces can lead to a line tension ( 140 ) , we incorporate a Ginzburg - Landau interfacial energy into the system , Biophysical Reports 2 , 100062 , September 14 , 2022 17 E d \u00bc 1 2 X f ijk h Z (cid:13)(cid:13)(cid:13)(cid:13) V ~ q f (cid:13)(cid:13)(cid:13)(cid:13) 2 ijk / 1 2 Z M h (cid:13)(cid:13)(cid:13)(cid:13) V ~ q f (cid:13)(cid:13)(cid:13)(cid:13) 2 dA ( 28 ) where h , referred to as the Dirichlet energy constant , governs the strength of the energy penalty , and V ~ q f is the discrete surface gradient of the protein density pro \ufb01 le . The term is similar to previous modeling efforts by Elliott and Stinner ( 80 ) and Ma and Klug ( 79 ) using FEM ; because we use the protein phase sep - aration as a prior , we exclude the double - well term , which models the thermodynamics of phase separa - tion , and incorporate only the Dirichlet energy compo - nent that penalizes the heterogeneity of membrane composition . De \ufb01 ned as the slope of the linearly interpolation of f on faces of the mesh , f ijk , the discrete surface gradient of the protein density is , V ~ q f i \u00bc 1 2 A ijk X e i \u02db N \u00f0 f ijk \u00de f i ~ e t i ; ( 29 ) where following illustration in Fig . 1 C , ~ e i is the vector aligned with the halfedge e i , with its length of l i , and \u00f0 $ \u00de t represents a 90 (cid:6) counterclockwise rotation in the plane of f ijk . The resulting line tension force R ~ f d is then the shape derivative of the Dirichlet energy , V ~ r E d , which acts to minimize the region with sharp het - erogeneity . The detailed derivation of the shape deriva - tive is elaborated in Appendix C . 3 , where we follow the formulaic approach by taking geometric derivatives of basic mesh primitives shown in Eq . C . 13 . Note that despite bearing the same name , this line tension force differs from from those resulting from line energy , which prescribes the line tension energy based on the interfacial edge length ( 55 , 142 ) . The line tension force from Dirichlet energy is used to model the out - of - plane component resulting from either entropic or enthalpic repulsion at the interface between heterogeneous membrane protein aggregates . The Dirichlet energy is based on a 2D \ufb01 eld variable and the line tension is only effective when phase separation of the \ufb01 eld vari - able occurs , where the interfacial line , or , more pre - cisely , thin areal band , between phases exists . With the introduction of protein evolution later in sections \u201c Protein aggregation on the realistic mesh of a den - dritic spine \u201d and \u201c Membrane dynamics with full mech - anochemical feedback , \u201d thickness depends on the competition between the Dirichlet energy and other competing aggregational potential . FIGURE 5 Budding dynamics by robust mechanisms of protein scaffolding and interfacial line tension constriction . ( A \u2013 C ) Control group , ( D \u2013 F ) bending - driven scaffolding mechanism , and ( G \u2013 I ) Interfacial line tension assisted budding . ( A ) Spontaneous curvature distribution , (cid:1) H , on initially \ufb02 at patch . ( D ) Normal projection of the bending force at T \u00bc 15 . ( G ) Normal projection of the line tension force at T \u00bc 7 . ( B , E , H ) Shape evo - lution through time - series snapshots of the Y - Z cross sections of the membrane , corresponding to the vertical dash lines in ( C ) , ( F ) , ( I ) trajectory plots of system energy and its competing components . 18 Biophysical Reports 2 , 100062 , September 14 , 2022 Fig . 5 G \u2013 I and Video S6 show the trajectory where we used control bending rigidity , k c \u00bc k b , and the addi - tional interfacial line tension , h \u00bc 5 (cid:3) 10 (cid:2) 4 m m $ nN . We \ufb01 nd that the interfacial line tension , jointly with the bending force , lowers the system energy and helps the formation of a spherical bud ( Fig . 5 I and H ) . Fig . 5 G shows the snapshot ( t \u00bc 7 ) with the color map representing the signed normal projection of the interfacial line tension that acts to constrict the neck . These examples of endocytic bud formation help to illustrate the utility of Mem3DG and the accompanying theoretical framework . Since physical parameters are assigned on a per - vertex basis , it is straightforward to incorporate heterogeneity such as the nonuniform bending rigidity and spontaneous curvature . In smooth theory and its derived discrete mesh models , when the membrane is heterogeneous , it is required to decom - pose the force separately in normal and tangential di - rection ( 40 , 65 ) . In contrast , the general derivation of the discrete bending force following the formalism of DDG permits modeling membrane with heterogeneous material properties without any modi \ufb01 cation to its formulation ( section \u201c Forces from bending \u201d ) . The intro - duction of Dirichlet energy and line tension force serves to highlight the relative ease to extend the modeled physics . Protein aggregation on the realistic mesh of a dendritic spine While the prior examples have focused on the mechan - ical response of the membrane given a bound protein distribution , we can also model the inverse problem . Given the membrane shape , how do curvature - sensing proteins diffuse in the plane of the membrane and distribute over the domain ? And how does the resultant protein distribution in \ufb02 uence the stresses of the sys - tem ? To model the protein dynamics , we use three terms corresponding to protein binding , curvature sensitivity , and lateral diffusion . To model the binding of proteins to the membrane , we assume that the energy of adsorption , \u03b5 , is constant and uniform across the surface such that the discrete adsorption energy is , E a \u00bc \u03b5 X i Z f i ; ( 30 ) where f i is an order parameter representing the area density of protein at each vertex . Taking the derivative with respect to f , referred to as the chemical derivative , m ai \u00bc (cid:2) V f E a \u00bc (cid:2) Z \u03b5 ; ( 31 ) we obtain the adsorption component of the chemical potential . To account for protein curvature sensitivity , we \ufb01 nd the chemical potential of the bending energy , m bi \u00bc (cid:2) V f E b \u00bc Z (cid:2) 2 k i \u00f0 H i (cid:2) (cid:1) H i \u00de V f (cid:1) H i (cid:2) \u00f0 H i (cid:2) (cid:1) H i \u00de 2 V f k i (cid:3) ; ( 32 ) where we assume that V f k i \u00bc k c , and V f (cid:1) H i \u00bc (cid:1) H c where k c and (cid:1) H c are constant parameters de \ufb01 ned in Section \u201c Modeling endocytic budding mechanisms . \u201d The \ufb01 rst term of Eq . 32 endows the protein with curva - ture - sensitive binding . The second term of Eq . 32 is the shape mismatch penalty ; considering the binding of a rigid protein that induces a signi \ufb01 cant spontaneous curvature change , if the curvature of membrane is far from this new spontaneous curvature , then the shape mismatch between the membrane and proteins will pre - vent binding . Alternatively , if the protein is more \ufb02 exible , a shape mismatch results in a small energetic penalty . The in - plane diffusion of the protein is accounted for by the chemical derivative of the smoothing Dirichlet energy , E d , m di \u00bc (cid:2) V f E d \u00bc (cid:2) Z h D s f i ; ( 33 ) where h is the same Dirichlet energy constant intro - duced in Eq . 28 that governs the strength of interfacial line tension , R ~ f d . The total chemical potential captures the bending , adsorption and diffusion components . A mobility rate constant , B , determines the time scale of the chemical response , _ f \u00bc B m \u00bc B \u00f0 m b \u00fe m a \u00fe m d \u00de : ( 34 ) We investigate the in \ufb02 uence of curvature - dependent binding to a realistic dendritic spine geometry , which was reconstructed from electron micrographs and curated using GAMer 2 ( Fig . 6 A ) ( 32 ) . A summary of the parameters used in the simulation is shown in Table 3 . The mean curvature of the spine geometry is shown Fig . 6 C . We isolate the effect of curvature - dependent binding by assuming that the shape of the spine is \ufb01 xed and impose Dirichlet boundary conditions at the base on the spine to \ufb01 x the protein concentration , f \u00bc 0 : 1 ( Fig . 6 A ) . Starting from a homogeneous protein distribution , f 0 \u00bc 0 : 1 , Fig . 6 B and Video S7 show the evolution of the protein distribution and a trajectory of the sys - tem energy . Note that , for simplicity , we have turned off the adsorption energy term since it only shifts the basal protein - membrane interactions , which will also be set by the Dirichlet boundary condition . Mem3DG constrains the range of f \u02db \u00f0 0 ; 1 \u00de using the interior point Biophysical Reports 2 , 100062 , September 14 , 2022 19 method ( 121 ) . Due to the curvature sensitivity of the protein , illustrated by the snapshots ( Fig . 6 B , T \u00bc 350 ) representing the \ufb01 nal protein distribution , the protein aggregates toward regions of high curva - ture ( e . g . , on the spine head ) . Although the proteins can reduce the bending energy by modulating the local bending modulus and sponta - neous curvature , the protein distribution at equilibrium does not cancel out the bending energy . We expect that the Dirichlet energy term , which limits f to be smooth across the geometry , restricts the protein from further aggregating to the extent required to cancel out the bending energy . The components of forces on the initial and \ufb01 nal con \ufb01 gurations of the spine are compared in Fig . 6 D \u2013 F . The initial homoge - neous protein distribution has no line tension forces and a bending force shown in Fig . 6 D . After the protein distribution reaches the steady state , line tension appears in response to membrane heterogeneity Fig . 6 E . We hypothesize that , similar to section \u201c Modeling endocytic budding mechanisms , \u201d the line tension constricts the neck of the spine and helps to support the cup - like structures in the spine head . While , in most regions , the distribution of proteins reduces the force , several regions experience increased stress Fig . 6 F . Note that the magnitude of the forces gener - ated by proteins in this model is orders of magnitude smaller than the initial bending force . Thus , this example demonstrates that Mem3DG can be used on meshes imported from realistic geometries of cellular substructures . Membrane dynamics with full mechanochemical feedback In this section , we will demonstrate the use of Mem3DG to model the complete mechanochemical feedback of a protein - membrane system . For the following simula - tions , not only can proteins bind in a curvature - depen - dent manner but the membrane can also deform , leading to a feedback loop . We have introduced all of the force terms in previous sections except the shape derivative of the adsorption energy , FIGURE 6 Protein aggregation on a realistic dendritic spine geometry . ( A ) Mesh of the dendritic spine and its boundary elements . ( B ) Trajectory of protein evolution and components of system energy . ( C ) Mean curvature of the geometry . The normal component of ( D ) the bending force at t \u00bc 0 , ( E ) the line tension force produced by the equilibrium protein distribution , and ( F ) the difference in the bending force pro \ufb01 le produced by \ufb01 nal protein distribution as opposed to the initial distribution . TABLE 3 Parameters used in section \u201c Protein aggregation on the realistic mesh of a dendritic spine \u201d Parameters Values f 0 0 . 1 k c 0 nN $ m m (cid:1) H c 10 m m (cid:2) 1 B 3 nN (cid:2) 1 $ m m (cid:2) 1 $ s (cid:2) 1 h 0 . 01 m m $ nN 20 Biophysical Reports 2 , 100062 , September 14 , 2022 Z ~ f a i \u00bc (cid:2) V ~ r E a \u00bc (cid:2) X e ij \u02db N \u00f0 v i \u00de \u03b5 (cid:9) f i 3 \u00fe 2 f j 3 (cid:10)Z 2 ~ H i ; ( 35 ) which accounts for the change in the area of protein coverage ( i . e . , expanded coverage if \u03b5 < 0 ) . Revisiting the claim that all discrete forcing is exact with respect to the discrete energy , we validate by examining the convergence of the forcing terms with respect to the size of perturbation to the system con \ufb01 guration , e ( Fig . 7 A ) . This is based on the leading order expansion done in Eq . 24 , which concludes that the forcing terms are exact if their rate of convergence is at least second order . Shown in Fig . 7 A , this is true for all forcing terms ; note that , since the adsorption energy , E a , is a linear function with respect to f , m a can be determined to the machine precision for all perturbation sizes . A meaningful discrete - smooth comparison of all terms in the energy and forcing similar to section \u201c Practical considerations for applying Mem3DG to biological problems \u201d requires the analytical solutions of the bending force and inter - facial line tension arising from the spatially heteroge - neous membrane properties , which , to the best of our knowledge , are not available . In section \u201c Modeling endocytic budding mechanisms , \u201d we introduced a het - erogeneous membrane composition as a static prop - erty . By prescribing the protein density pro \ufb01 le , we can get hints about how to form membrane buds from a patch . Here we lift this assumption and simu - late the dynamics of osmotic pressure - driven budding from a vesicle . The dynamics couples the protein - membrane mechanochemical feedback and includes protein binding and diffusion introduced in section \u201c Protein aggregation on the realistic mesh of a FIGURE 7 Modeling budding from a vesicle driven by the full mechanochemical feedback of membrane - protein interactions . ( A ) Validation of the exactness of the discrete forcing with respect to the discrete energy . The terms correspond to forces from bending f b , tension area f s , pres - sure - volume f p , Dirichlet f d , and protein binding f a . m d , m b , and m a are the chemical potential of diffusion , bending , and binding respectively . ( B ) The time trajectory of budding dynamics in hypertonic , isotonic , and hypotonic osmotic conditions . ( C ) The \ufb01 nal snapshot of the system con \ufb01 gura - tion under hypertonic , isotonic , and hypotonic conditions . ( D ) Similar geometries to those shown in ( C ) have been observed in experiments by Saleem et al . ( 59 ) . Biophysical Reports 2 , 100062 , September 14 , 2022 21 dendritic spine . \u201d The expressions of discrete free en - ergy and forcings do not change but we allow the membrane con \ufb01 guration and protein density to interact and evolve simultaneously . We start each simulation from a sphere with a uniform protein concentration , f \u00bc f 0 \u00bc 0 : 1 . We consider the evolution of the system in different osmotic conditions : hyper - , iso - , and hypotonic , (cid:1) V \u00bc 2 : 91 ; 3 : 95 and 4 : 99 m m 3 , respectively . Additional param - eters for these simulations are given in Table 4 . Fig . 7 B shows plots of the mechanical , k ~ f k 2 L 2 , and chemical response , k m k 2 L 2 , along with the protein density , \u00f0 f max \u00fe f min \u00de = 2 , over the trajectory for each osmotic condition . Note that under hypo - and isotonic condi - tions , the system reaches the ( quasi ) steady state where both the shape and protein distribution stabilize , while , in a hypertonic solution , the system continues to experience strong mechanical force and protein mobility , which we expect to drive further morpholog - ical changes of the membrane beyond our simulation stopping point . Fig . 7 C shows the \ufb01 nal snapshot of each simulation across the osmotic conditions with the protein density represented by the color map . In hy - pertonic conditions , the osmotic pressure provides suf - \ufb01 cient perturbations to membrane morphology , which initializes the positive feedback loop between mem - brane curvature generation and protein aggregation . This mechanochemical feedback jointly promotes the outward bending of the membrane and results in bud formation ( Fig . 7 C , hypertonic ; Video S8 ) . Under isotonic and hypotonic conditions , the osmolarity does not permit the large change in the volume required to form spherical buds with a thin neck . In hy - potonic condition , the pressure - tension balance pro - vides substantial stability to the initial spherical con \ufb01 guration . In comparison , in the isotonic condition , the comparable competition between the chemical and mechanical response leads to a patterned protein dis - tribution and an undulating morphology ( Fig . 7 C , hypo - tonic ; Video S9 ) . This example illustrates the possibility to utilize Mem3DG to model a full mechanochemical feedback between membrane and protein . Although we do not intend to replicate the exact experimental conditions and assumptions , the geometries obtained from these simulations resemble the shapes obtained by Saleem et al . ( 59 ) who investigated budding from spherical vesicles under differing osmotic conditions ( Fig . 7 D ) ( 59 ) . SUMMARY In this work , we introduce a new perspective for con - structing a 3D membrane mechanics model on discrete meshes . The goal of our approach is to close the gap between existing discrete mesh - based models ( 60 , 86 \u2013 96 , 99 , 102 , 103 ) and the smooth theory . Specif - ically , we seek to advance the discussion behind the choice of algorithmic approaches for computing geo - metric values required for the discrete energy and force ( 65 , 89 , 99 , 100 ) . We start by writing a discrete energy , Eq . 3 , mirroring the spontaneous curvature model . Then using the perspective of DDG , we show that there is a formulaic approach for deriving the corresponding discrete force terms based on fundamental geometric vectors . By identifying geometric invariants and grouping terms , the resulting discrete forces have exact correspondence to the traditional smooth theory . This helps us to facilitate the comparison between smooth and discrete contexts enabling new geometric perspectives and understanding of numerical accu - racy . Moreover , the discrete force terms are functions of readily accessible geometric primitives , and the formulation is amenable for ef \ufb01 cient implementation and extension . We have produced a reference software implementa - tion called Mem3DG . Using Mem3DG , we validate our the - ory by reproducing the solutions to the classical shape transformations of a spherical and tubular vesicle . We further demonstrate the derivation and incorporation of additional physics terms to model protein - membrane interactions . Following our formulaic approach using DDG , we derived the discrete analog of various physics , such as the interfacial line tension , surface - bulk adsorption , protein lateral diffusion , and curvature - dependent protein aggregation . To exemplify all the introduced physics , the full mechanochemical coupling between the membrane shape and protein density evo - lution results in protein localization , pattern formation , and budding . These examples serve to highlight the extensibility of the framework , which does not require the introduction of coordinates and advanced tensor calculus commonly needed to solve problems on arbi - trary manifolds . The software implementation Mem3DG was designed to facilitate coordination between TABLE 4 Parameters used in section \u201c Membrane dynamics with full mechanochemical feedback \u201d for models with full mechanochemical feedback Parameters Values f 0 0 . 1 k c 8 . 22 (cid:3) 10 (cid:2) 5 m m $ nN (cid:1) H c 10 m m (cid:2) 1 K V 0 . 5 nN $ m m K A 1 nN $ m m (cid:2) 1 B 3 nN (cid:2) 1 $ m m (cid:2) 1 $ s (cid:2) 1 x 1 nN $ s $ m m (cid:2) 1 \u03b5 (cid:2) 1 (cid:3) 10 (cid:2) 3 nN $ m m h 0 . 1 m m $ nN (cid:1) V 2 . 91 , 3 . 95 , and 4 . 99 m m 3 22 Biophysical Reports 2 , 100062 , September 14 , 2022 theoretical modeling and wet experiments ; we hope to support the modeling of scenes with well - resolved pro - tein - membrane interactions such as in the electron tomograms ( 143 ) . We expect that as the advances in biophysical modeling and membrane ultrastructure im - aging progresses , Mem3DG will emerge as a useful tool to test new hypotheses and understand cellular mechanobiology . APPENDICES A . Supplemental \ufb01 gures B . Rationale for integrated measurements in discrete contexts The rationale for why an integrated measurement in discrete con - texts is the natural counterpart to pointwise measurements in smooth contexts can be demonstrated by considering the curvature of a discrete polygonal curve . If we attempt to de \ufb01 ne the curvature , C , of the discrete polygonal curve in a na\u00efve pointwise manner , following the procedure in smooth settings , we will \ufb01 nd zero curva - ture along edges and in \ufb01 nite curvature ( owing to the discontinuity ) on vertices . Thus the traditional view of curvature from smooth manifolds reveals no useful information about the geometry of the discrete curve . We must \ufb01 nd another geometric relationship that can translate between smooth and discrete contexts to maintain the geometric connection . One relationship from smooth differential geometry is the equiva - lence of the integrated curvature and the turning angle j ( i . e . , the total angle by which the tangent vector of the curve turns over some domain l ) . Returning to the discrete context , we can seek to preserve this relationship between the integrated curvature and turning angle by \ufb01 nding a compatible de \ufb01 nition . Since the discrete turning angle , j i , between two connected edges of the discrete polygonal curve is well de \ufb01 ned , we can set the discrete curvature , R C , of a vertex , v i , to be (cid:9)Z C (cid:10) i h j i : ( B . 1 ) We note that the notation for the discrete curvature , \u00f0 R C \u00de i is used only in this illustrative example ; in the remainder of the text , we will omit the parenthesis and use the simpli \ufb01 ed notation , R C i . To make sense of the integral over a discrete object , additional care must be taken to represent the curvature from a distributional sense ( 104 ) . This is related to traditional approximation methods , such as the point allocation method , which bridges a smooth and FIGURE A . 2 Steiner ' s formula in continuous and discrete geometry : chain of smooth and discrete shape derivatives of integrated geo - metric measurements ( 144 ) . FIGURE A . 1 A symmetric metastable state with two beads instead of a single larger bead is observed , prior to collapsing to the solution shown in Fig . 4 C , high tension . Biophysical Reports 2 , 100062 , September 14 , 2022 23 discrete problem by convoluting the smooth problem with impulse functions ( e . g . , the Dirac delta function ) at a \ufb01 nite number of loca - tions ( 122 ) . As we have shown , integrated geometric measurements enable us to preserve geometric relationships ( from smooth contexts ) for discrete objects , and are thus preferred over pointwise de \ufb01 nitions . Nevertheless , we often require a pointwise discrete measurement for use in algorithms and visualization . An integrated measurement can be converted to a meaningful pointwise discrete measurement by normalizing the value over a domain . For the discrete polygonal curve , thedomaincanbethedualvertexlength , l i ( i . e . , thediscreteanalogof l ) . l i is given by half of the sum lengths of the two incident edges . A point - wise curvature on the vertex v i is then given by , C i \u00bc Z C i = l i \u00bc j i = l i : ( B . 2 ) Another rationale for using an integrated value for a discrete geo - metric measurement is that we can arrive at the same de \ufb01 nition from multiple perspectives . Returning to the de \ufb01 nition of the curva - ture of a polygonal curve , without introducing the turning angle , we can arrive at the same result by adopting the Steiner view ( 104 , 145 ) ( we use the Steiner view to de \ufb01 ne the discrete curvature of a surface in section \u201c Obtaining a discrete energy de \ufb01 ned by mesh primitives \u201d ) . In the Steiner view , we replace the sharp vertices with a smooth circular arc with radius e such that the discrete geometry is made smooth such that the curvature is well de \ufb01 ned everywhere . As the only curved section , every circular arc has a discrete ( integrated ) curvature , Z C \u00bc Z arc C ds \u00bc C arc l arc \u00bc 1 e \u00f0 ej \u00de \u00bc j ; ( B . 3 ) where C arc \u00bc 1 = e is the curvature of the circular arc , and l arc \u00bc ej is the arc length . We see that , in the Steiner view , the integrated curva - ture is still equivalent to the turning angle . Following similar logic , other discrete de \ufb01 nitions are described in section \u201c Obtaining a discrete energy de \ufb01 ned by mesh primitives \u201d and the DDG literature ( 104 , 105 ) . C . Discrete shape and chemical derivatives of discrete energy C . 1 . Halfedge on a triangulated mesh A scalar quantity on an edge is symmetric with respect to index permutation . For example , the scalar mean curvature ( Eq . 8 ) , Z H ij \u00bc Z H ji \u00bc l ij 4 ij 2 : ( C . 1 ) However , as we will show in detail in the following sections , this symmetry does not apply to vector quantities , which compose the discrete shape derivative of the energy , force . For example , the corre - sponding mean curvature vector , Z ~ H ij s Z ~ H ji : ( C . 2 ) To highlight the directionality of vector quantities and disambig - uate the notation , here we review the concept of a halfedge on a triangulated mesh . Given any non - boundary edge , e ij , on a manifold mesh , there exits two associated halfedges , e ij and e ji ( Fig . C . 1 ) . This convention leads to an oriented ( counterclockwise ) halfedge loop on each triangle face and subsequently a well - de \ufb01 ned 1 ) 90 (cid:6) counterclockwise rotation of the halfedge in the plane of the face ( e . g . , e lj / e t lj ) , and 2 ) face normal ( outward ) based on the right hand rule ( Fig . C . 1 ) . Beside being used to differentiate vector / scalar quantities , the concept of halfedge is widely adopted data structure for managing connected graphs , or meshes , for which we refer the reader to the broader literature ( 116 , 146 ) . C . 2 . Deriving the bending force as the shape derivative of bending energy The geometric derivatives of mesh primitives , including edge length , l , dihedral angle , 4 , and vertex dual area , A , are given as V ~ r i l ij \u00bc ~ e ji l ij ; ( C . 3a ) V ~ r i 4 ij \u00bc 1 l ij (cid:9) cot : ijk ~ n ijk \u00fe cot : ijl ~ n ilj (cid:10) ; ( C . 3b ) V ~ r i 4 jk \u00bc (cid:2) 1 l jk (cid:11) cot : ijk \u00fe cot : ikj (cid:12) ~ n ijk \u00bc (cid:2) l jk 2 A ijk ~ n ijk ; ( C . 3c ) V ~ r i A i \u00bc 1 3 X f ijk \u02db N \u00f0 v i \u00de V ~ r i A ijk \u00bc 1 6 X e ij \u02db N \u00f0 v i \u00de (cid:11) cot : ikj \u00fe cot : ilj (cid:12) ~ e ji ; ( C . 3d ) V ~ r i A j \u00bc 1 3 X f ijk \u02db N \u00f0 e ij \u00de V ~ r i A ijk \u00bc 1 6 (cid:7) ~ e t jk \u00fe ~ e t lj (cid:8) ; ( C . 3e ) where ~ n ijk is the unit normal vector of the face f ijk , and ~ e ji is the vector aligned with the halfedge , e ji , with its length of l ij ( 147 ) . The indices and nomenclature in Eqs . C . 3b , C . 3c and C . 3e are illustrated in the diamond neighborhood ( Fig . C . 1 ) and those of Eq . C . 3d are illustrated in the fan neighborhood ( Fig . 1 A ) . FIGURE C . 1 Schematics for halfedges on a triangulated mesh . 24 Biophysical Reports 2 , 100062 , September 14 , 2022 To simplify the expression and provide more structure for the sub - sequent discrete variation , it is convenient to de \ufb01 ne some funda - mental curvature vectors , Z 2 ~ H ij \u00bc 1 2 (cid:7) V ~ r i A ijk \u00fe V ~ r i A ijl (cid:8) \u00bc 1 4 ~ e t jk \u00fe ~ e t lj ! ( C . 4a ) Z ~ K ij \u00bc 1 2 4 ij V ~ r i l ij ( C . 4b ) Z ~ S ij ; 1 \u00bc 1 2 l ij V ~ r i 4 ij \u00bc 1 2 (cid:9) cot : ijk ~ n ijk \u00fe cot : ijl ~ n ilj (cid:10) ( C . 4c ) Z ~ S ij ; 2 \u00bc 1 2 (cid:7) l jk V ~ r i 4 jk \u00fe l jl V ~ r i 4 jl \u00fe l ji V ~ r i 4 ji (cid:8) \u00bc (cid:2) 1 2 (cid:9) cot : jki ~ n ijk \u00fe cot : ilj ~ n ilj (cid:10) ; ( C . 4d ) where the mean curvature vector , R ~ H , results from area gradient ; Gaussian curvature vector , R ~ K , and the Schla \ufb02 i vector , R ~ S , consist of the two components of the variation of total mean curvature , 12 P e ij l ij 4 ij . The asymmetry of vector quantities in Eq . C . 4 under index permutation ( Eq . C . 2 ) arises from the vertex we take the shape deriv - ative with respect to ( i . e . , v i , or v j ) ; because of the asymmetry , we can associate each Schla \ufb02 i vector with a unique halfedge . Similar to the translation from edge values to vertex value ( Eq . 9 ) , we can also translate the halfedge value to vertex value by summing all halfedge values over the fan neighborhood , Z \u00f0 $ \u00de i \u00bc X e ij \u02db N \u00f0 v i \u00de Z \u00f0 $ \u00de ij : ( C . 5 ) Note that , unlike translating edge values , there is no prefactor 1 = 2 for translating halfedge values because each halfedge is uniquely associated with one vertex . The translated curvature vectors on a ver - tex cane compared against vertexwise smooth analytical solutions as benchmarked in section \u201c Practical considerations for applying Mem3DG to biological problems . \u201d Now we have all of the elements needed to derive the derivatives of the discrete Willmore bending en - ergy . Because the discrete energy is locally supported by the vertex , v i , and its 1 - ring neighbors , v j \u02db N \u00f0 v i \u00de , we can separate them into the \u201c diagonal \u201d term , and \u201c off - diagonal \u201d term , Using the derivatives of geometric primitives in Eq . C . 3 , we can assemble the derivatives of local pointwise mean curvature for both the diagonal term , V ~ r i H i \u00bc 1 4 X e ij \u02db N \u00f0 v i \u00de V ~ r i l ij 4 ij A i \u00bc 1 4 A i X e ij \u02db N \u00f0 v i \u00de (cid:7) 4 ij V ~ r i l ij \u00fe l ij V ~ r i 4 ij (cid:8) (cid:2) H i A i V ~ r i A i \u00bc 1 A i X e ij \u02db N \u00f0 v i \u00de 1 2 (cid:9) Z ~ K ij \u00fe Z ~ S ij ; 1 (cid:10) (cid:2) 2 3 H i Z ~ H ij ; ( C . 7 ) and for the off - diagonal term , When written in the halfedge form , factoring out the fundamental curvature vectors introduced in Eq . C . 4 , we obtain the discrete bending force as Z ~ f b i \u00bc (cid:2) V ~ r i E b \u00bc (cid:2) V ~ r i X i k i \u00f0 H i \u00f0 ~ r \u00de (cid:2) (cid:1) H i \u00de 2 A i \u00f0 ~ r \u00de ! \u00bc (cid:2) V ~ r i (cid:2) k i \u00f0 H i (cid:2) (cid:1) H i \u00de 2 A i (cid:3) | \ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 } diagonal (cid:2) X v j \u02db N \u00f0 v i \u00de V ~ r i h k j (cid:11) H j (cid:2) (cid:1) H j (cid:12) 2 A j i | \ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 { z\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04\ufb04 } off - diagonal : ( C . 6 ) V ~ r i H j \u00bc 1 4 X e jk \u02db N \u00f0 v j \u00de V ~ r i l jk 4 jk A j \u00bc 1 4 A j (cid:7) l jk V ~ r i 4 jk \u00fe l jl V ~ r i 4 jl \u00fe 4 ji V ~ r i l ji \u00fe l ji V ~ r i 4 ji (cid:8) (cid:2) H j A j V ~ r i A j \u00bc 1 2 A j (cid:9) Z ~ K ij \u00fe Z ~ S ij ; 2 (cid:10) (cid:2) 4 3 H j Z ~ H ij : ( C . 8 ) Biophysical Reports 2 , 100062 , September 14 , 2022 25 When the surface is not closed , boundary vertices , v i \u02db v M , experience an additional force from the Gaussian curvature component , Z ~ f b i \u00bc RHS of Eq : \u00f0 C : 9 \u00de \u00fe X f ijk \u02db N \u00f0 v i \u00de \u00f0(cid:2) 1 \u00de p V ~ r i : f ijk ; v i \u02db v M ; ( C . 10 ) where p is the number of boundary vertices in face f ijk , and : f ijk \u00bc 8 < : : ijk if v j ; v k ; v M ; : ikj if v j \u02db v M ; : ijk if v k \u02db v M : ( C . 11 ) C . 3 . Deriving the line tension and diffusion as the shape and chemical derivatives of the Dirichlet energy Since the discrete Dirichlet energy is constructed on the triangular face and therefore does not involve any neighborhood , we simplify the notation by adopting the convention illustrated in Fig . 1 C . The gradient of protein density is given by the slope of the \ufb01 tted plane over the vertexwise protein density , which is piecewise constant for each face , V ~ q f i \u00bc 1 2 A ijk X ~ e k \u02db N \u00f0 f ijk \u00de f k ~ e t k ; ( C . 12 ) where we adopt the counterclockwise convention ( e . g . , ~ e k \u00bc ~ e ji ) and \u00f0 $ \u00de t represents a 90 (cid:6) counterclockwise rotation in plane of the face , f ijk . C . 3 . 1 . Line tension from the shape derivative of the Dirichlet energy . Substituting the de \ufb01 nition of the discrete gradient into the Dirichlet energy ( Eq . 28 ) , we expand the energy in terms of mesh primitives , whose geometric derivatives are given in Eq . C . 3 . Additional formulae are needed to compute the geometric derivatives of the outer angles of the triangle ( Fig . 1 C ) V ~ r i : k \u00bc ~ n (cid:3) ~ e j k ~ e j k 2 ( C . 13a ) V ~ r i : j \u00bc ~ n (cid:3) ~ e k k ~ e k k 2 ( C . 13b ) V ~ r i : i \u00bc (cid:2) (cid:7) V ~ r i : k \u00fe V ~ r i : j (cid:8) ; ( C . 13c ) which arise from the calculation of the L 2 norm of the gradients as the result of vector inner product . When combined , the geometric deriva - tives for the quadratic gradient term is Then we can get the \ufb01 nal shape derivative by combining the area gradient , or the mean curvature vector ( Eq . C . 4 ) . C . 3 . 2 . Surface diffusion from the chemical derivative of the Dirichlet energy . In the case where we are evolving the protein dis - tribution , we need the chemical derivative of the Dirichlet energy . Before we look into the discrete case , we can \ufb01 rst tackle the problem Z ~ f b i \u00bc X e ij \u02db N \u00f0 v i \u00de (cid:2) (cid:2) k i \u00f0 H i (cid:2) (cid:1) H i \u00de \u00fe k j (cid:11) H j (cid:2) (cid:1) H j (cid:12)(cid:3)Z ~ K ij \u00fe (cid:5) 1 3 k i \u00f0 H i (cid:2) (cid:1) H i \u00de\u00f0 H i \u00fe (cid:1) H i \u00de \u00fe 2 3 k j (cid:11) H j (cid:2) (cid:1) H j (cid:12)(cid:11) H j \u00fe (cid:1) H j (cid:12)(cid:6) Z 2 ~ H ij (cid:2) (cid:5) k i \u00f0 H i (cid:2) (cid:1) H i \u00de Z ~ S ij ; 1 \u00fe k j (cid:11) H j (cid:2) (cid:1) H j (cid:12) Z ~ S ij ; 2 (cid:6) : ( C . 9 ) V ~ r i (cid:15) X f k ~ e t k ; X f k ~ e t k (cid:16) \u00bc \u00fe f k f k ~ e k (cid:2) 2 f j f j ~ e j \u00fe 2 f j f i k ~ e i k (cid:9) (cid:2) b e j cos : k \u00fe k ~ e j k V ~ r i \u00f0 cos : k \u00de (cid:10) \u00fe 2 f i f k k ~ e i k (cid:9) b e k cos : j \u00fe k ~ e k k V ~ r i (cid:11) cos : j (cid:12)(cid:10) \u00fe 2 f j f k (cid:9) (cid:2) b e j k ~ e k k cos : i \u00fe k ~ e j k b e k cos : i \u00fe k ~ e j kk ~ e k k V ~ r i \u00f0 cos : i \u00de (cid:10) ( C . 14 ) 26 Biophysical Reports 2 , 100062 , September 14 , 2022 in the smooth setting , which is a classic textbook example . Using the Green ' s \ufb01 rst identity , or integration by parts on a 2 - manifold , Z M (cid:11) j D s 4 \u00fe V ~ q j $ V ~ q 4 (cid:12) dA \u00bc I v M j V ~ q 4 $ ~ ndS ; ( C . 15 ) and ignoring the boundary term at the right hand side , we arrive at an alternative expression for the Dirichlet energy , E d \u00bc 1 2 Z M h k V ~ q f k 2 dA \u00bc (cid:2) 1 2 Z M hf D s f dA : ( C . 16 ) The same procedure can be followed in the discrete case . The discrete Dirichlet energy ( Eq . 28 ) can be written in matrix form , E d \u00bc 1 2 h f u ~ G u ~ T ~ G f ( C . 17 ) where ~ G is the gradient tensor , which maps scalar value on vertices to vector values on faces , and ~ T \u00bc diag \u00f0 A face \u00de is the j f j(cid:3) j f j diagonal matrix with entries corresponding to the area of each mesh triangle face . Through integration by parts on discrete geometry , the discrete Dirichlet energy can be equivalently expressed as E d \u00bc 1 2 h f u ~ L f ; ( C . 18 ) which is a quadratic form with respect to the cotangent Laplacian matrix , ~ L ( 104 , 105 ) . The chemical derivative of the Dirichlet energy , or the diffusion potential , is m d \u00bc (cid:2) V f E d \u00bc (cid:2) h Z D s f \u00bc (cid:2) h ~ L f : ( C . 19 ) In other words , the chemical gradient \ufb02 ow of the Dirichlet energy is the diffusion equation . Note that ~ L \u00bc ~ G u ~ T ~ G , ~ G u is referred to as the discrete divergence operator , which maps face vectors to scalars on vertices ( 146 ) . D . Discrete - smooth comparison on spheroid The smooth - discrete comparison is done on the spheroid with the parametrization , \u00f0 x ; y ; z \u00de \u00bc \u00f0 a cos b cos q ; a cos b sin q ; c sin b \u00de ; ( D . 1 ) where a \u00bc 1 , b \u00bc 0 : 5 , b is the parametric latitude and q is the azimuth coordinate . All geometric measurements of the smooth geometry used for benchmarking were obtained using the symbolic algebra software Sympy . The corresponding discrete measurements are computed using Mem3DG , whose input spheroid mesh is mapped from a subdivided icosphere . The subsequent error norms for local measurements are computed based on de \ufb01 nitions used in section \u201c De \ufb01 ning metrics for simulation and error quanti \ufb01 cation . \u201d E . Mesh regularization and mesh mutation E . 1 . Mesh mutation Mesh mutation and re \ufb01 nement in combination with vertex shifting are the default methods to ensure that the mesh remains well condi - tioned and well resolved during simulation . Mesh mutations include edge \ufb02 ipping , collapsing , and splitting , which change the connectivity of the mesh . Vertex shifting moves the vertex to the barycenter of the fan neighborhood without changing the mesh topology ( Fig . 1 A ) . Mem3DG has a suite of possible criteria to initiate mesh mutation . Here we list the most important ones : 1 ) \ufb02 ip the edge of the non - De - launay diamond neighborhood ( Fig . 1 B ) , 2 ) collapse the shortest FIGURE E . 1 Pointwise magnitude comparison of continuous and discrete measurements : ( A ) scalar mean curvature , ( B ) scalar Gaussian cur - vature , ( C ) ( scalar ) bi - Laplacian term V H based on the cotan formula , ( D ) vector mean curvature , ( E ) vector Gaussian curvature , and ( F ) ( vector ) bi - Laplacian term based on Schla \ufb02 i vector . Note that the result of the cotangent Laplacian approach in ( C ) produces a scalar result while our approach using the Schla \ufb02 i vector in ( F ) is a vector result , thus their direct comparison is not meaningful . Biophysical Reports 2 , 100062 , September 14 , 2022 27 edge in a skinny triangle face , and 3 ) split the edge with high ( geodesic ) curvature . For additional details , please refer to the soft - ware documentation . For practical use , although mesh mutation introduces additional complexity in data write - out and computational costs associated with varying ( usually growing ) mesh size , it nevertheless provides a robust algorithm to ensure the good mesh quality needed for valid discrete - smooth comparisons ( section \u201c Practical consider - ations for applying Mem3DG to biological problems \u201d ) in static frames . For dynamical simulation , mesh mutations introduce an arbi - trary interpolation of state variables , such as the position , velocity , and protein density . Rigorous study on how to interpolate these quan - tities to ensure the conservation of energy , momentum , and mass re - mains to be done . Similarly , the interpolation used in this study introduces discontinuities of curvature and can create jumps in forces ; this is particularly severe for terms with higher - order deriva - tives such as the biharmonic term in bending force ( Eq . 22 ) . E . 2 . Mesh regularization Mesh regularization can be used when mesh mutations are not desired . The regularization force consists of three weakly enforced constraining forces : the edge ( length ) , ~ f e , face ( area ) , ~ f f , and confor - mality ( angle ) , ~ f c , regularization forces , ~ f e i \u00bc (cid:2) K e X e ij \u02db N \u00f0 v i \u00de (cid:11) l ij (cid:2) (cid:1) l ij (cid:12) (cid:1) l ij V ~ ri l ij ; ( E . 1a ) ~ f f i \u00bc (cid:2) K f X f ijk \u02db N \u00f0 v i \u00de (cid:11) A ijk (cid:2) (cid:1) A ijk (cid:12) (cid:1) A ijk V ~ r i A ijk ; ( E . 1b ) ~ f c i \u00bc (cid:2) K c X e ij \u02db N \u00f0 v i \u00de (cid:11) l ij (cid:2) (cid:1) l ij (cid:12) (cid:1) l ij V ~ r i l ij ; ( E . 1c ) which are in the order of strongest to weakest . The length - cross - ratio , l ij \u00bc l il l jk = l ki l jl is a metric of discrete conformality on triangulated mesh , where the indices is illustrated in Fig . 1 A and B ( 148 ) . Regula - rization forces require the input of a reference value for geometric measurements , (cid:1) l , (cid:1) A , and (cid:1) l , which can be derived from a well - condi - tioned reference mesh ( usually the initial input mesh for the simula - tion ) . The intensity of each regularization force is controlled with parameters K e , K f , and K c . For practical use , regularization constraints should be minimally imposed because of their impact on system dynamics . In the worst case , regularization constraints can prevent the optimizer from reach - ing an energy minima . Thus a good practice is to start a simulation with no : conformality , face area , and \ufb01 nally edge length regulariza - tion , and subsequently raise the intensity / type of constraints based on the mesh quality desired . We do not recommend imposing con - straints stronger than the face areal constraints , ~ f f . In addition to be - ing numerical regularizers for the triangulated mesh , they can serve as model for certain additional physics . For example , the edge length regularization has been adopted to model the additional local rigidity from actin - spectrin cortex in red blood cells ( 95 ) . The areal regulari - zation can be used to model local incompressibility of the membrane . The conformality regularization can be used to isolate shearing resistance . SUPPORTING MATERIAL Supplemental information can be found online at https : / / doi . org / 10 . 1016 / j . bpr . 2022 . 100062 . AUTHOR CONTRIBUTIONS P . R . conceived the research . P . R . , C . Z . , and C . T . L . designed the research . C . Z . developed the theory and analyzed the results . C . Z . and C . T . L . developed the software . C . Z . , C . T . L . , and P . R . wrote and edited the article . All authors read and approved the \ufb01 nal article . ACKNOWLEDGMENTS The authors would like to acknowledge Dr . Ali Behzadan , Prof . Ravi Ramamoorthi , and Prof . Albert Chern for helpful discussions and crit - ical feedback . This work was supported in part by the National Insti - tutes of Health R01GM132106 , National Science Foundation DMS 1934411 , Of \ufb01 ce of Naval Research N00014 - 20 - 1 - 2469 , and an Air Force Of \ufb01 ce of Scienti \ufb01 c Research DURIP FA9550 - 19 - 1 - 0181 to P . R . C . T . L . also acknowledges support from a Hartwell Foundation Postdoctoral Fellowship . DECLARATION OF INTERESTS The authors declare no competing interests . REFERENCES 1 . Groves , J . T . , and J . Kuriyan . 1989 . Molecular mechanisms in signal transduction at the membrane . Nat . Struct . Mol . Biol . 17 : 659 \u2013 665 . https : / / doi . org / 10 . 1038 / nsmb . 1844 . 2 . Cheng , X . , and J . C . Smith . 2019 . Biological membrane organiza - tion and cellular signaling . Chem . Rev . 119 : 5849 \u2013 5880 . FIGURE E . 2 Pointwise directional comparison of continuous and discrete measurements : discrete vertex normal based on ( A ) volume gradient , V ~ r V , mean curvature vector , ~ H , and Gaussian curvature vector , ~ K , and ( B ) Schla \ufb02 i vector , H ~ S . 28 Biophysical Reports 2 , 100062 , September 14 , 2022 3 . Anitei , M . , and B . Ho \ufb02 ack . 2012 . Bridging membrane and cyto - skeleton dynamics in the secretory and endocytic pathways . Nat . Cell Biol . 14 : 11 \u2013 19 . 4 . Bonifacino , J . S . , and B . S . Glick . 2004 . The mechanisms of vesicle budding and fusion . Cell . 116 : 153 \u2013 166 . 5 . Herrmann , J . M . , and A . Spang . 2015 . In Membrane Traf \ufb01 cking , Second Edition , B . L . Tang , ed Springer , New York , NY , pp . 1 \u2013 12 . https : / / doi . org / 10 . 1007 / 978 - 1 - 4939 - 2309 - 0 _ 1 . 6 . McMahon , H . T . , and E . Boucrot . 2011 . Molecular mechanism and physiological functions of clathrinmediated endocytosis . Nat . Rev . Mol . Cell Biol . 12 : 517 \u2013 533 . 7 . Zimmerberg , J . , and M . M . Kozlov . 2006 . How proteins produce cellular membrane curvature . Nat . Rev . Mol . Cell Biol . 7 : 9 \u2013 19 . 8 . Farsad , K . , and P . D . Camilli . 2003 . Mechanisms of membrane deformation . Curr . Opin . Cell Biol . 15 : 372 \u2013 381 . 9 . Avinoam , O . , M . Schorb , . , M . Kaksonen . 2015 . Endocytic sites mature by continuous bending and remodeling of the clathrin coat . Science . 348 : 1369 \u2013 1372 . https : / / doi . org / 10 . 1126 / sci - ence . aaa9555 . 10 . Terasaki , M . , T . Shemesh , . , T . A . Rapoport . 2013 . Stacked endoplasmic reticulum sheets are connected by helicoidal membrane motifs . Cell . 154 : 285 \u2013 296 . https : / / doi . org / 10 . 1016 / j . cell . 2013 . 06 . 031 . 11 . Shibata , Y . , T . Hemesh , . , T . A . Rapoport . 2010 . Mechanisms determining the morphology of the peripheral ER . Cell . 143 : 774 \u2013 788 . https : / / doi . org / 10 . 1016 / j . cell . 2010 . 11 . 007 . 12 . McMahon , H . T . , and J . L . Gallop . 2005 . Membrane curvature and mechanisms of dynamic cell membrane remodelling . Na - ture . 438 : 590 \u2013 596 . https : / / doi . org / 10 . 1038 / nature04396 . 13 . McMahon , H . T . , and E . Boucrot . 2015 . Membrane curvature at a glance . J . Cell Sci . 128 : 1065 \u2013 1070 . 14 . Kozlov , M . M . , F . Campelo , . , H . T . McMahon . 2014 . Mecha - nisms shaping cell membranes . Curr . Opin . Cell Biol . 29 : 53 \u2013 60 . 15 . Scheve , C . S . , P . A . Gonzales , . , J . C . Stachowiak . 2013 . Steric pressure between membranebound proteins opposes lipid phase separation . J . Am . Chem . Soc . 135 : 1185 \u2013 1188 . https : / / doi . org / 10 . 1021 / ja3099867 . 16 . Stachowiak , J . C . , C . C . Hayden , and D . Y . Sasaki . 2010 . Steric con \ufb01 nement of proteins on lipid membranes can drive curvature and tubulation . Proc . Natl . Acad . Sci . USA . 107 : 7781 \u2013 7786 . https : / / doi . org / 10 . 1073 / pnas . 0913306107 . 17 . Stachowiak , J . C . , E . M . Schmid , . , C . C . Hayden . 2012 . Mem - brane bending by protein \u2013 protein crowding . Nat . Cell Biol . 14 : 944 \u2013 949 . https : / / doi . org / 10 . 1038 / ncb2561 . 18 . Stachowiak , J . C . , F . M . Brodsky , and E . A . Miller . 2013 . A cost \u2013 bene \ufb01 t analysis of the physical mechanisms of membrane cur - vature . Nat . Cell Biol . 15 : 1019 \u2013 1027 . https : / / doi . org / 10 . 1038 / ncb2832 . 19 . Yuan , F . , H . Alimohamadi , . , J . C . Stachowiak . 2021 . Mem - brane bending by protein phase separation . Proc . Natl . Acad . Sci . USA . 118 . e2017435118 . https : / / doi . org / 10 . 1073 / pnas . 2017435118 . 20 . Chabanon , M . , J . C . Stachowiak , and P . Rangamani . 2017 . Sys - tems biology of cellular membranes : a convergence with biophysics : systems biology of cellular membranes . Wiley Inter - discip . Rev . Sys . Biol . Med . 9 : e1386 . https : / / doi . org / 10 . 1002 / wsbm . 1386 . 21 . Knott , G . , H . Marchman , . , B . Lich . 2008 . Serial section scan - ning electron microscopy of adult brain tissue using focused ion beam milling . J . Neurosci . 28 : 2959 \u2013 2964 . https : / / doi . org / 10 . 1523 / JNEUROSCI . 3189 - 07 . 2008 . 22 . Peddie , C . J . , and L . M . Collinson . 2014 . Exploring the third dimension : volume electron microscopy comes of age . Micron . 61 : 9 \u2013 19 . https : / / doi . org / 10 . 1016 / J . MICRON . 2014 . 01 . 009 . 23 . Titze , B . , and C . Genoud . 2016 . Volume scanning electron micro - scopy for imaging biological ultrastructure . Biol . Cell . 108 : 307 \u2013 323 . https : / / doi . org / 10 . 1111 / boc . 201600024 . 24 . Knott , G . , and C . Genoud . 2013 . Is EM Dead ? J . Cell Sci . 126 : 4545 \u2013 4552 . https : / / doi . org / 10 . 1242 / jcs . 124123 . 25 . Micheva , K . D . , and S . J . Smith . 2007 . Array tomography : a new tool for imaging the molecular architecture and ultrastructure of neural circuits . Neuron . 55 : 25 \u2013 36 . https : / / doi . org / 10 . 1016 / j . neuron . 2007 . 06 . 014 . 26 . Villa , E . , M . Schaffer , . , W . Baumeister . 2013 . Opening windows into the cell : focusedionbeam milling for cryoelectron tomogra - phy . Curr . Opin . Struct . Biol . 23 : 771 \u2013 777 . https : / / doi . org / 10 . 1016 / j . sbi . 2013 . 08 . 006 . 27 . NixonAbell , J . , C . J . Obara , . , C . Blackstone . 2016 . Increased Spatiotemporal resolution reveals highly dynamic dense tubular matrices in the peripheral ER . Science . 354 : aaf3928 . https : / / doi . org / 10 . 1126 / science . aaf3928 . 28 . Hoffman , D . P . , G . Shtengel , . , D . R . Stabley . 2020 . Correlative threedimensional superresolution and blockface electron mi - croscopy of whole vitreously frozen cells . Science . 367 : eaaz5357 . https : / / doi . org / 10 . 1126 / science . aaz5357 . 29 . Xu , C . S . , K . J . Hayworth , . , H . F . Hess . 2017 . Enhanced FIBSEM Systems for LargeVolume 3D Imaging . Elife . 6 : e25916 . https : / / doi . org / 10 . 7554 / eLife . 25916 . 30 . Frey , T . G . , and C . A . Mannella . 2000 . The internal structure of mitochondria . Trends Biochem . Sci . 25 : 319 \u2013 324 . 31 . Wu , Y . , C . Whiteus , . , P . De Camilli . 2017 . Contacts between the Endoplasmic Reticulum and Other membranes in neurons . Proc . Natl . Acad . Sci . USA . 114 : E4859 \u2013 E4867 . https : / / doi . org / 10 . 1073 / pnas . 1701078114 . 32 . Lee , C . T . , J . G . Laughlin , . , P . Rangamani . 2020 . 3D mesh pro - cessing using GAMer 2 to enable reactiondiffusion simulations in realistic cellular geometries . PLoS Comput . Biol . 16 : e1007756 . 33 . Mendelsohn , R . , G . C . Garcia , . , G . Perkins . 2021 . Morpholog - ical principles of neuronal mitochondria . J . Comp . Neurol . 530 : 886 \u2013 902 . 34 . Salfer , M . , J . F . Collado , . , A . Mart\u00ednez - S\u00e1nchez . 2020 . Reliable Estimation of Membrane Curvature for CryoElectron Tomogra - phy . PLoS Comput . Biol . 16 : e1007962 . 35 . Tamada , H . , J . Blanc , . , G . W . Knott . 2020 . Ultrastructural com - parison of dendritic spine morphology preserved with cryo and chemical \ufb01 xation . Elife . 9 : e56384 . 36 . Davies , K . M . , M . Strauss , . , W . K \u20ac uhlbrandt . 2011 . Macromolec - ular organization of atp synthase and complex i in whole mito - chondria . Proc . Natl . Acad . Sci . USA . 108 : 14121 \u2013 14126 . https : / / doi . org / 10 . 1073 / pnas . 1103621108 . 37 . Davies , K . M . , C . Anselmi , . , W . K \u20ac uhlbrandt . 2012 . Structure of the Yeast F1FoATP Synthase Dimer and Its Role in Shaping the Mitochondrial Cristae . Proc . Natl . Acad . Sci . USA . 109 : 13602 \u2013 13607 . https : / / doi . org / 10 . 1073 / pnas . 1204593109 . 38 . Ramakrishnan , N . , P . B . Sunil Kumar , and R . Radhakrishnan . 2014 . Mesoscale computational studies of membrane bilayer remodeling by curvatureinducing proteins . Phys . Rep . 543 : 1 \u2013 60 . https : / / doi . org / 10 . 1016 / j . physrep . 2014 . 05 . 001 . 39 . Helfrich , W . 1973 . Elastic properties of lipid bilayers : theory and possible experiments . Z . Naturforsch . C . 28 : 693 \u2013 703 . 40 . Steigmann , D . J . 1999 . Fluid \ufb01 lms with curvature elasticity . Arch . Ration . Mech . Anal . 150 : 127 \u2013 152 . 41 . Campelo , F . , C . Arnarez , . , M . M . Kozlov . 2014 . Helfrich model of membrane bending : from gibbs theory of liquid interfaces to membranes as thick anisotropic elastic layers . Adv . Colloid Interface Sci . 208 : 25 \u2013 33 . 42 . Naito , H . , M . Okuda , and O . Y . Zhongcan . 1995 . New Solutions to the Helfrich Variation Problem for the Shapes of Lipid Bilayer Biophysical Reports 2 , 100062 , September 14 , 2022 29 Vesicles : Beyond Delaunay ' s Surfaces . Phys . Rev . Lett . 74 : 4345 \u2013 4348 . https : / / doi . org / 10 . 1103 / PhysRevLett . 74 . 4345 . 43 . Deserno , M . 2015 . Fluid Lipid Membranes : From Differential Ge - ometry to Curvature Stresses . Chem . Phys . Lipids . 185 : 11 \u2013 45 . https : / / doi . org / 10 . 1016 / j . chemphyslip . 2014 . 05 . 001 . 44 . Argudo , D . , N . P . Bethel , . , M . Grabe . 2016 . Continuum Descrip - tions of Membranes and Their Interaction with Proteins : To - wards Chemically Accurate Models . Biochim . Biophys . Acta . 1858 : 1619 \u2013 1634 . https : / / doi . org / 10 . 1016 / j . bbamem . 2016 . 02 . 003 . 45 . Argudo , D . , N . P . Bethel , . , M . Grabe . 2017 . New Continuum Ap - proaches for Determining ProteinInduced Membrane Deforma - tions . Biophys . J . 112 : 2159 \u2013 2172 . https : / / doi . org / 10 . 1016 / j . bpj . 2017 . 03 . 040 . 46 . Hamm , M . , and M . Kozlov . 2000 . Elastic energy of tilt and bending of \ufb02 uid membranes . Eur . Phys . J . E . 3 : 323 \u2013 335 . 47 . Agrawal , A . , and D . J . Steigmann . 2009 . BoundaryValue prob - lems in the theory of lipid membranes . Continuum Mech . Therm . 21 : 57 \u2013 82 . https : / / doi . org / 10 . 1007 / s00161 - 009 - 0102 - 8 . 48 . Rangamani , P . , A . Benjamini , . , G . Oster . 2014 . Small scale membrane mechanics . Biomech . Model . Mechanobiol . 13 : 697 \u2013 711 . https : / / doi . org / 10 . 1007 / s10237 - 013 - 0528 - 6 . 49 . Liese , S . , E . M . Wenzel , . , A . Carlson . 2020 . Protein crowding mediates membrane remodeling in upstream escrtinduced for - mation of intraluminal vesicles . Proc . Natl . Acad . Sci . USA . 117 : 28614 \u2013 28624 . 50 . Liese , S . , and A . Carlson . 2021 . Membrane shape remodeling by protein crowding . Biophys . J . 120 : 2482 \u2013 2489 . 51 . Bassereau , P . , R . Jin , . , T . R . Weikl . 2018 . The 2018 bio - membrane curvature and remodeling roadmap . J . Phys . D Appl . Phys . 51 : aacb98 . https : / / doi . org / 10 . 1088 / 1361 - 6463 / aacb98 . 52 . Baumgart , T . , B . R . Capraro , . , S . L . Das . 2011 . Thermody - namics and mechanics of membrane curvature generation and sensing by proteins and lipids . Annu . Rev . Phys . Chem . 62 : 483 \u2013 506 . 53 . Lee , C . T . , M . Akamatsu , and P . Rangamani . 2021 . Value of models for membrane budding . Curr . Opin . Cell Biol . 71 : 38 \u2013 45 . https : / / doi . org / 10 . 1016 / j . ceb . 2021 . 01 . 011 . 54 . Carlsson , A . E . 2018 . Membrane bending by actin polymeriza - tion . Curr . Opin . Cell Biol . 50 : 1 \u2013 7 . 55 . Liu , J . , M . Kaksonen , . , G . Oster . 2006 . Endocytic vesicle scis - sion by lipid phase boundary forces . Proc . Natl . Acad . Sci . USA . 103 : 10277 \u2013 10282 . 56 . Liu , J . , Y . Sun , . , G . F . Oster . 2009 . The mechanochemistry of endocytosis . PLoS Biol . 7 : e1000204 . 57 . Hassinger , J . E . , G . Oster , . , P . Rangamani . 2017 . Design prin - ciples for robust vesiculation in clathrinmediated endocytosis . Proc . Natl . Acad . Sci . USA . 114 : E1118 \u2013 E1127 . 58 . Alimohamadi , H . , R . Vasan , J . E . Hassinger , J . C . Stachowiak , and P . Rangamani . 2018 . The role of traction in membrane cur - vature generation . Mol . Biol . Cell . 29 : 2024 \u2013 2035 . 59 . Saleem , M . , S . Morlot , . , A . Roux . 2015 . A balance between membrane elasticity and polymerization energy sets the shape of spherical clathrin coats . Nat . Commun . 6 : 6249 . https : / / doi . org / 10 . 1038 / ncomms7249 . 60 . Tachikawa , M . , and A . Mochizuki . 2017 . Golgi apparatus selfor - ganizes into the characteristic shape via postmitotic reassem - bly dynamics . Proc . Natl . Acad . Sci . USA . 114 : 5177 \u2013 5182 . https : / / doi . org / 10 . 1073 / pnas . 1619264114 . 61 . Natesan , R . , and R . Radhakrishnan . 2015 . Phenomenology Based Multiscale Models as Tools to Understand Cell Mem - brane and Organelle Morphologies . . Advances in Planar Lipid Bilayers and Liposomes , 22 . Elsevier , ISBN 9780128028780 , pp . 129 \u2013 175 . https : / / doi . org / 10 . 1016 / bs . adplan . 2015 . 06 . 004 . 62 . Agrawal , N . J . , and R . Radhakrishnan . 2009 . Calculation of Free Energies in Fluid Membranes Subject to Heterogeneous Curva - ture Fields . Phys . Rev . 80 : 011925 . 63 . Agrawal , N . J . , J . Nukpezah , and R . Radhakrishnan . 2010 . Mini - mal Mesoscale Model for ProteinMediated Vesiculation in ClathrinDependent Endocytosis . C . V . Rao . PLoS Comput . Biol . 6 : e1000926 . 64 . Ma , R . , and J . Berro . 2021 . Endocytosis against High Turgor Pressure Is Made Easier by Partial Coating and Freely Rotating Base . Biophys . J . 120 : 1625 \u2013 1640 . 65 . Guckenberger , A . , and S . Gekle . 2017 . Theory and Algorithms to Compute Helfrich Bending Forces : A Review . J . Phys . Con - dens . Matter . 29 : 203001 . https : / / doi . org / 10 . 1088 / 1361 - 648X / aa6313 . 66 . Akamatsu , M . , R . Vasan , . , D . G . Drubin . 2020 . Principles of selforganization and load adaptation by the actin cytoskeleton during clathrinmediated endocytosis . Elife . 9 : e49840 . 67 . Willmore , T . 1996 . Riemannian Geometry . Oxford University Press . https : / / books . google . com / books ? id \u00bc L6qtDAEACAAJ . 68 . Evans , E . A . , and R . Skalak . 2018 . Mechanics and thermody - namics of biomembranes . CRC press . 69 . Canham , P . B . 1970 . The minimum energy of bending as a possible explanation of the biconcave shape of the human red blood cell . J . Theor . Biol . 26 : 61 \u2013 81 . 70 . Jenkins , J . T . 1977 . The equations of mechanical equilibrium of a model membrane . SIAM J . Appl . Math . 32 : 755 \u2013 764 . 71 . Jenkins , J . T . 1977 . Static equilibrium con \ufb01 gurations of a model red blood cell . J . Math . Biol . 4 : 149 \u2013 169 . 72 . Seifert , T . , O . Zsch\u00f6rnig , . , K . Arnold . 1997 . Beta - blockers inhibit the modi \ufb01 cation of low - density lipoproteins by sodium hypochlorite in vitro . Chem . Phys . Lipids . 85 : 13 \u2013 21 . 73 . Du , Q . , C . Liu , and X . Wang . 2004 . A Phase Field Approach in the Numerical Study of the Elastic Bending Energy for Vesicle Mem - branes . J . Comput . Phys . 198 : 450 \u2013 468 . https : / / doi . org / 10 . 1016 / j . jcp . 2004 . 01 . 029 . 74 . Du , Q . , and X . Wang . 2007 . Convergence of numerical approxi - mations to a phase \ufb01 eld bending elasticity model of membrane deformations . Int . J . Numer . Anal . Model . 4 : 441 \u2013 459 . 75 . Salac , D . , and M . Miksis . 2011 . A level set projection model of lipid vesicles in general \ufb02 ows . J . Comput . Phys . 230 : 8192 \u2013 8215 . 76 . Biben , T . , and C . Misbah . 2003 . Tumbling of vesicles under shear \ufb02 ow within an advected \ufb01 eld approach . Phys . Rev . E - Stat . Nonlinear Soft Matter Phys . 67 : 031908 . https : / / doi . org / 10 . 1103 / PhysRevE . 67 . 031908 . 77 . Biben , T . , K . Kassner , and C . Misbah . 2005 . Phase \ufb01 eld approach to threedimensional vesicle dynamics . Phys . Rev . E - Stat . Nonlinear Soft Matter Phys . 72 : 041921 . https : / / doi . org / 10 . 1103 / PhysRevE . 72 . 041921 . 78 . Feng , F . , and W . S . Klug . 2006 . Finite Element Modeling of Lipid Bilayer Membranes . J . Comput . Phys . 220 : 394 \u2013 408 . https : / / doi . org / 10 . 1016 / j . jcp . 2006 . 05 . 023 . 79 . Ma , L . , and W . S . Klug . 2008 . Viscous regularization and radap - tive remeshing for \ufb01 nite element analysisof lipid membrane me - chanics . J . Comput . Phys . 227 : 5816 \u2013 5835 . 80 . Elliott , C . M . , and B . Stinner . 2010 . Modeling and Computation of Two Phase Geometric Biomembranes Using Surface Finite Ele - ments . J . Comput . Phys . 229 : 6585 \u2013 6612 . https : / / doi . org / 10 . 1016 / j . jcp . 2010 . 05 . 014 . 81 . Rangarajan , R . , and H . Gao . 2015 . A Finite Element Method to Compute ThreeDimensional Equilibrium Con \ufb01 gurations of Fluid Membranes : Optimal Parameterization , Variational Formulation and Applications . J . Comput . Phys . 297 : 266 \u2013 294 . https : / / doi . org / 10 . 1016 / j . jcp . 2015 . 05 . 001 . 82 . Sauer , R . A . , T . X . Duong , . , D . J . Steigmann . 2017 . A Stabilized Finite Element Formulation for Liquid Shells and Its Application 30 Biophysical Reports 2 , 100062 , September 14 , 2022 to Lipid Bilayers . J . Comput . Phys . 330 : 436 \u2013 466 . https : / / doi . org / 10 . 1016 / j . jcp . 2016 . 11 . 004 . 83 . Vasan , R . , S . Rudraraju , . , P . Rangamani . 2020 . A Mechanical Model Reveals That NonAxisymmetric Buckling Lowers the En - ergy Barrier Associated with Membrane Neck Constriction . Soft Matter . 16 : 784 \u2013 797 . https : / / doi . org / 10 . 1039 / C9SM01494B . 84 . TorresS\u00e1nchez , A . , D . Mill\u00e1n , and M . Arroyo . 2019 . Modelling \ufb02 uid deformable surfaces with an emphasis on biological inter - faces . J . Fluid Mech . 872 : 218 \u2013 271 . https : / / doi . org / 10 . 1017 / jfm . 2019 . 341 . 85 . Auddya , D . , X . Zhang , . , S . Rudraraju . 2021 . Biomembranes un - dergo complex , nonaxisymmetric deformations governed by kirchhof \ufb02 ove kinematics and revealed by a three dimensional computational framework . Proc . Math . Phys . Eng . Sci . 477 : 20210246 . https : / / doi . org / 10 . 1098 / rspa . 2021 . 0246 . 86 . Gompper , G . , and D . M . Kroll . 1996 . Random surface discretiza - tions and the renormalization of the bending rigidity . J . Phys . I France . 6 : 1305 \u2013 1320 . 87 . J \u20ac ulicher , F . 1996 . The morphology of vesicles of higher topolog - ical genus : conformal degeneracy and conformal modes . J . Phys . II France . 6 : 1797 \u2013 1824 . 88 . Kantor , Y . , and D . R . Nelson . 1987 . Phase transitions in \ufb02 exible polymeric surfaces . Phys . Rev . A Gen . Phys . 36 : 4020 \u2013 4032 . 89 . Bian , X . , S . Litvinov , and P . Koumoutsakos . 2020 . Bending Models of Lipid Bilayer Membranes : Spontaneous Curvature and AreaDifference Elasticity . Comput . Methods Appl . Mech . Eng . 359 : 112758 . https : / / doi . org / 10 . 1016 / j . cma . 2019 . 112758 . 90 . Brakke , K . A . 1992 . The Surface Evolver . Exp . Math . 1 : 141 \u2013 165 . https : / / doi . org / 10 . 1080 / 10586458 . 1992 . 10504253 . 91 . Jie , Y . , L . Quanhui , . , O . Y . ZhongCan . 1998 . Numerical Obser - vation of Nonaxisymmetric Vesicles in Fluid Membranes . Phys . Rev . E . 58 : 4730 \u2013 4736 . https : / / doi . org / 10 . 1103 / PhysRevE . 58 . 4730 . 92 . Kroll , D . M . , and G . Gompper . 1992 . The Conformation of Fluid Membranes : Monte Carlo Simulations . Science . 255 : 968 \u2013 971 . https : / / doi . org / 10 . 1126 / science . 1546294 . 93 . Atilgan , E . , and S . X . Sun . 2007 . Shape Transitions in Lipid Mem - branes and Protein Mediated Vesicle Fusion and Fission . J . Chem . Phys . 126 : 095102 . https : / / doi . org / 10 . 1063 / 1 . 2483862 . 94 . Bahrami , A . H . , and G . Hummer . 2017 . Formation and Stability of Lipid Membrane Nanotubes . ACS Nano . 11 : 9558 \u2013 9565 . https : / / doi . org / 10 . 1021 / acsnano . 7b05542 . 95 . Noguchi , H . , and G . Gompper . 2005 . Shape Transitions of Fluid Vesicles and Red Blood Cells in Capillary Flows . Proc . Natl . Acad . Sci . USA . 102 : 14159 \u2013 14164 . https : / / doi . org / 10 . 1073 / pnas . 0504243102 . 96 . Tsai , K . , S . Britton , . , M . Alber . 2020 . Role of Combined Cell Membrane and Wall Mechanical Properties Regulated by Polar - ity Signals in Cell Budding . Phys . Biol . 17 : 065011 . https : / / doi . org / 10 . 1088 / 1478 - 3975 / abb208 . 97 . Lim H W , G . , M . Wortis , and R . Mukhopadhyay . 2008 . Soft Mat - ter , Section : 2 . John Wiley & Sons , Ltd , pp . 83 \u2013 139 . https : / / doi . org / 10 . 1002 / 9783527623372 . ch2a . https : / / onlinelibrary . wiley . com / doi / pdf / . 98 . Lim H W , G . , M . Wortis , and R . Mukhopadhyay . 2021 . Stomato - cytediscocyteechinocyte sequence of the human red blood cell : Evidence for the bilayercouple hypothesis from membrane mechanics . en . Proc . Natl . Acad . Sci . USA . 99 : 16766 \u2013 16769 . https : / / doi . org / 10 . 1073 / pnas . 202617299 . 99 . Fai , T . G . , B . E . Grif \ufb01 th , . , C . S . Peskin . 2022 . Immersed Bound - ary Method for Variable Viscosity and Variable Density Prob - lems Using Fast Constant - Coef \ufb01 cient Linear Solvers I : Numerical Method and Results . SIAM J . Sci . Comput . 35 : B1132 \u2013 B1161 . https : / / doi . org / 10 . 1137 / 120903038 . 100 . Guckenberger , A . , M . P . Schraml , . , S . Gekle . 2016 . On the Bending Algorithms for Soft Objects in Flows . Comput . Phys . Commun . 207 : 1 \u2013 23 . https : / / doi . org / 10 . 1016 / j . cpc . 2016 . 04 . 018 . 101 . Shewchuk , J . R . 2002 . What Is a Good Linear Finite Element ? Interpolation , Conditioning , Anisotropy , and Quality Measures . Proc . 11th Int . Meshing Roundtable . 94720 : 115 \u2013 126 . 102 . Pezeshkian , W . , M . K\u00f6nig , . , J . H . Ipsen . 2019 . A multiscale approach to membrane remodeling processes . Front . Mol . Bio - sci . 6 : 59 . 103 . Sadeghi , M . , T . R . Weikl , and F . No (cid:3) e . 2018 . ParticleBased Mem - brane Model for Mesoscopic Simulation of Cellular Dynamics . J . Chem . Phys . 148 : 044901 . https : / / doi . org / 10 . 1063 / 1 . 5009107 . 104 . A . Chern , Discrete Differential Geometry , 2020 . 105 . K . Crane , Discrete Differential Geometry : An Applied Introduc - tion , 2020 . 106 . Evans , E . A . 1974 . Bending Resistance and Chemically Induced Moments in Membrane Bilayers . Biophys . J . 14 : 923 \u2013 931 . 107 . Do Carmo , M . P . 2016 . Differential geometry of curves and sur - faces : revised and updated second edition . Courier Dover Pub - lications . 108 . Lee , J . M . 2012 . Introduction to Smooth Manifolds218 . Springer New York , New York , NY . https : / / doi . org / 10 . 1007 / 978 - 1 - 4419 - 9982 - 5 . 109 . A . I . Bobenko , ed 2008 . Discrete Differential Geometry Birkh \u20ac auser , Basel \u037e Boston . 110 . Phillips , R . , J . Kondev , . , N . Orme . 2012 . Physical Biology of the Cell . Garland Science , Taylor & Francis Group , New York , Nov . 111 . Meyer , M . , M . Desbrun , . , A . H . Barr . 2003 . In Visualization and Mathematics III , G . Farin . , eds Springer Berlin Heidelberg , Ber - lin , Heidelberg , pp . 35 \u2013 57 . https : / / doi . org / 10 . 1007 / 978 - 3 - 662 - 05105 - 4 _ 2 . 112 . M . Wardetzky , S . Mathur , . , E . Grinspun , presented at the ACM SIGGRAPH ASIA 2008 Courses on SIGGRAPH Asia ' 08 , ( 2022 ) pp . 1 \u2013 5 , DOI 10 . 1145 / 1508044 . 1508063 . 113 . Wardetzky , M . 2008 . Convergence of the Cotangent Formula : An Overview . In Discrete Differential Geometry , 38 A . I . Bobenko , J . M . Sullivan , P . Schr\u00f6der , and G . M . Ziegler , eds . , pp . 275 \u2013 286 . https : / / doi . org / 10 . 1007 / 978 - 3 - 7643 - 8621 - 4 _ 15 . 114 . M . Deserno , Notes on Differential Geometry . 115 . Crane , K . , and M . Wardetzky . 2022 . A Glimpse into Discrete Dif - ferential Geometry . Not . Am . Math . Soc . 64 : 1153 \u2013 1159 . https : / / doi . org / 10 . 1090 / noti1578 . 116 . Sharp , N . , K . Crane . , 2019 . geometrycentral . www . geometrycentral . net . 117 . Sharp , N . , K . Crane . , 2018 . hapPLY . https : / / github . com / nmwsharp / happly . 118 . Rew , R . , and G . Davis . 1990 . NetCDF : an interface for scienti \ufb01 c data access . IEEE Comput . Graph . Appl . 10 : 76 \u2013 82 . 119 . Sharp , N . . , 2019 . Polyscope . www . polyscope . run . 120 . Jakob , W . , J . Rhinelander , and D . Moldovan . 2017 . pybind11 \u2013 Seamless operability between C \u00fe\u00fe 11 and Python . https : / / github . com / pybind / pybind11 . 121 . Nocedal , J . , and S . J . Wright . 2006 . Numerical Optimization , 2nd ed . Springer , New York . 122 . Hughes , T . J . 2012 . The \ufb01 nite element method : linear static and dynamic \ufb01 nite element analysis . Courier Corporation . 123 . J \u20ac ulicher , F . , and R . Lipowsky . 1996 . Shape transformations of vesicles with intramembrane domains . Phys . Rev . E Stat . Phys . Plasmas Fluids Relat . Interdiscip . Topics . 53 : 2670 \u2013 2683 . 124 . Deuling , H . J . , and W . Helfrich . 1976 . Red blood cell shapes as explained on the basis of curvature elasticity . Biophys . J . 16 : 861 \u2013 868 . 125 . Evans , E . , and Y . C . Fung . 1972 . Improved Measurements of the Erythrocyte Geometry . Microvasc . Res . 4 : 335 \u2013 347 . https : / / doi . org / 10 . 1016 / 0026 - 2862 ( 72 ) 90069 - 6 . Biophysical Reports 2 , 100062 , September 14 , 2022 31 126 . Alimohamadi , H . , A . S . Smith , . , P . Rangamani . 2020 . Non - uni - form distribution of myosin - mediated forces governs red blood cell membrane curvature through tension modulation . PLoS Comput . Biol . 16 : e1007890 . https : / / doi . org / 10 . 1371 / journal . pcbi . 1007890 . 127 . Gerdes , H . H . , and R . N . Carvalho . 2008 . Intercellular transfer mediated by tunneling nanotubes . Curr . Opin . Cell Biol . 20 : 470 \u2013 475 . 128 . Pearce , K . M . , M . Bell , . , S . Scarlata . 2020 . G a qMediated Cal - cium Dynamics and Membrane Tension Modulate Neurite Plas - ticity . Mol . Biol . Cell . 31 : 683 \u2013 694 . https : / / doi . org / 10 . 1091 / mbc . E19 - 09 - 0536 . 129 . Alimohamadi , H . , B . Ovryn , and P . Rangamani . 2020 . Modeling membrane nanotube morphology : the role of heterogeneity in composition and material properties . Sci . Rep . 10 : 2527 . 130 . Datar , A . , J . Ameeramja , . , P . A . Pullarkat . 2019 . The Roles of Microtubules and Membrane Tension in Axonal Beading , Retraction , and Atrophy . Biophys . J . 117 : 880 \u2013 891 . https : / / doi . org / 10 . 1016 / j . bpj . 2019 . 07 . 046 . 131 . Pullarkat , P . A . , P . Dommersnes , . , A . Ott . 2006 . Osmotically Driven Shape Transformations in Axons . Phys . Rev . Lett . 96 : 048104 . https : / / doi . org / 10 . 1103 / PhysRevLett . 96 . 048104 . 132 . J . Plateau , Experimental and Theoretical Statics of Liquids Sub - ject to Molecular Forces Only , 1873 . 133 . L . Rayleigh , On the Instability of Jets , 1878 . 134 . O . Y . Zhongcan , W . Helfrich , Bending energy of vesicle mem - branes : general expressions for the \ufb01 rst , second , and third vari - ation of the shape energy and applications to spheres and cylinders . Physical Review A , 39 . 5280 135 . BarZiv , R . , and E . Moses . 1994 . Instability and \u201c pearling \u201d states produced in tubular membranes by competition of curvature and tension . Phys . Rev . Lett . 73 : 1392 \u2013 1395 . 136 . Sanborn , J . , K . Ogl \u0119 cka , . , A . N . Parikh . 2013 . Transient Pearl - ing and Vesiculation of Membrane Tubes under Osmotic Gradi - ents . Faraday Discuss . 161 : 167 \u2013 176 . https : / / doi . org / 10 . 1039 / C2FD20116J . 137 . Nelson , P . , T . Powers , and U . Seifert . 2021 . Dynamical Theory of the Pearling Instability in Cylindrical Vesicles . en . Phys . Rev . Lett . 74 : 3384 \u2013 3387 . https : / / doi . org / 10 . 1103 / PhysRevLett . 74 . 3384 . 138 . Mladenov , I . 2002 . New solutions of the shape equation . Eur . Phys . J . B Condens . Matter . 29 : 327 \u2013 330 . https : / / doi . org / 10 . 1140 / epjb / e2002 - 00310 - y . 139 . Watanabe , S . , B . R . Rost , . , E . M . Jorgensen . 2013 . Ultrafast Endocytosis at Mouse Hippocampal Synapses . Nature . 504 : 242 \u2013 247 . https : / / doi . org / 10 . 1038 / nature12809 . 140 . Baumgart , T . , S . T . Hess , and W . W . Webb . 2003 . Imaging Coex - isting Fluid Domains in Biomembrane Models Coupling Curva - ture and Line Tension . Nature . 425 : 821 \u2013 824 . https : / / doi . org / 10 . 1038 / nature02013 . 141 . Crane , K . , C . Weischedel , and M . Wardetzky . 2017 . The Heat Method for Distance Computation . Commun . ACM . 60 : 90 \u2013 99 . https : / / doi . org / 10 . 1145 / 3131280 . 142 . Frey , F . , F . Ziebert , and U . S . Schwarz . 2019 . Dynamics of parti - cle uptake at cell membranes . Phys . Rev . E . 100 : 052403 . https : / / doi . org / 10 . 1103 / PhysRevE . 100 . 052403 . 143 . Serwas , D . , M . Akamatsu , . , D . G . Drubin . 2022 . Mechanistic in - sights into actin force generation during vesicle formation from cryo - electron tomography . Dev . Cell . 57 : 1132 \u2013 1145 . https : / / doi . org / 10 . 1016 / j . devcel . 2022 . 04 . 012 . 144 . K . Crane , Lecture slide in discrete differential geometry , 15 , 2021 . 145 . Steiner , J . 2013 . Ueber parallele Fl \u20ac achen . In Jacob Steiner ' s Ge - sammelte Werke : Herausgegeben auf Veranlassung der k\u00f6ni - glich preussischen Akademie der Wissenschaften ( Cambridge Library Collection - Mathematics ) . K . Weierstrass , ed . Cam - bridge University Press , Cambridge , pp . 171 \u2013 176 . https : / / doi . org / 10 . 1017 / CBO9781139567947 . 016 . 146 . Jacobson , A . , D . Panozzo . , 2018 . libigl : A simple C \u00fe\u00fe geom - etry processing library . https : / / libigl . github . io / . 147 . E . Grinspun , A . N . Hirani , . , P . Schr\u00f6der , presented at the Pro - ceedings of the 2003 ACM SIGGRAPH / Eurographics sympo - sium on Computer animation , ( 2003 ) pp . 62 \u2013 67 . 148 . Soliman , Y . , A . Chern , . , P . Schr\u00f6der . 2021 . Constrained Will - more Surfaces . ACM Trans . Graph . 40 : 07300301 . https : / / doi . org / 10 . 1145 / 3450626 . 3459759 . 32 Biophysical Reports 2 , 100062 , September 14 , 2022",
    "smaldinoNaturalSelectionBad2016": "rsos . royalsocietypublishing . org Research Cite this article : Smaldino PE , McElreath R . 2016 The natural selection of bad science . R . Soc . opensci . 3 : 160384 . http : / / dx . doi . org / 10 . 1098 / rsos . 160384 Received : 1 June 2016 Accepted : 17 August 2016 Subject Category : Psychology and cognitive neuroscience Subject Areas : theoretical biology / computer modelling and simulation / statistics Keywords : metascience , cultural evolution , statistical power , replication , incentives , Campbell\u2019s Law Author for correspondence : Paul E . Smaldino e - mail : paul . smaldino @ gmail . com Electronic supplementary material is available at http : / / dx . doi . org / 10 . 1098 / rsos . 160384 or via http : / / rsos . royalsocietypublishing . org . The natural selection of bad science Paul E . Smaldino 1 and Richard McElreath 2 1 CognitiveandInformationSciences , UniversityofCalifornia , Merced , CA95343 , USA 2 DepartmentofHumanBehavior , Ecology , andCulture , MaxPlanckInstitutefor EvolutionaryAnthropology , Leipzig , Germany PES , 0000 - 0002 - 7133 - 5620 ; RME , 0000 - 0002 - 0387 - 5377 Poor research design and data analysis encourage false - positive \ufb01ndings . Such poor methods persist despite perennial calls for improvement , suggesting that they result from something more than just misunderstanding . The persistence of poor methods results partly from incentives that favour them , leading to the natural selection of bad science . This dynamic requires no conscious strategizing\u2014no deliberate cheating nor loa\ufb01ng\u2014 by scientists , only that publication is a principal factor for career advancement . Some normative methods of analysis have almost certainly been selected to further publication instead of discovery . In order to improve the culture of science , a shift must be made away from correcting misunderstandings and towards rewarding understanding . We support this argument with empirical evidence and computational modelling . We \ufb01rst present a 60 - year meta - analysis of statistical power in the behavioural sciences and show that power has not improved despite repeated demonstrations of the necessity of increasing power . To demonstrate the logical consequences of structural incentives , we then present a dynamic model of scienti\ufb01c communities in which competing laboratories investigate novel or previously published hypotheses using culturally transmitted research methods . As in the real world , successful labs produce more \u2018progeny , \u2019 such that their methods are more often copied and their students are more likely to start labs of their own . Selection for high output leads to poorer methods and increasingly high false discovery rates . We additionally show that replication slows but does not stop the process of methodological deterioration . Improving the quality of research requires change at the institutional level . 2016 The Authors . Published by the Royal Society under the terms of the Creative Commons Attribution License http : / / creativecommons . org / licenses / by / 4 . 0 / , which permits unrestricted use , providedtheoriginalauthorandsourcearecredited . 2 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The more any quantitative social indicator is used for social decision - making , the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor . Donald T . Campbell ( 1976 , p . 49 ) [ 1 ] I\u2019ve been on a number of search committees . I don\u2019t remember anybody looking at anybody\u2019s papers . Number and IF [ impact factor ] of pubs are what counts . Terry McGlynn ( realscientists ) ( 21 October 2015 , 4 : 12 p . m . Tweet . ) 1 . Introduction In March 2016 , the American Statistical Association published a set of corrective guidelines about the use and misuse of p - values [ 2 ] . Statisticians have been publishing guidelines of this kind for decades [ 3 , 4 ] . Beyond mere signi\ufb01cance testing , research design in general has a history of shortcomings and repeated corrective guidelines . Yet misuse of statistical procedures and poor methods has persisted and possibly grown . In \ufb01elds such as psychology , neuroscience and medicine , practices that increase false discoveries remain not only common , but normative [ 5 \u2013 11 ] . Why have attempts to correct such errors so far failed ? In April 2015 , members of the UK\u2019s science establishment attended a closed - door symposium on the reliability of biomedical research [ 12 ] . The symposium focused on the contemporary crisis of faith in research . Many prominent researchers believe that as much as half of the scienti\ufb01c literature\u2014not only in medicine , by also in psychology and other \ufb01elds\u2014may be wrong [ 11 , 13 \u2013 15 ] . Fatal errors and retractions , especially of prominent publications , are increasing [ 16 \u2013 18 ] . The report that emerged from this symposium echoes the slogan of one anonymous attendee : \u2018Poor methods get results . \u2019 Persistent problems with scienti\ufb01c conduct have more to do with incentives than with pure misunderstandings . So \ufb01xing them has more to do with removing incentives that reward poor research methods than with issuing more guidelines . As Richard Horton , editor of The Lancet , put it : \u2018Part of the problem is that no one is incentivised to be right\u2019 [ 12 ] . This paper argues that some of the most powerful incentives in contemporary science actively encourage , reward and propagate poor research methods and abuse of statistical procedures . We term this process the natural selection of bad science to indicate that it requires no conscious strategizing nor cheating on the part of researchers . Instead , it arises from the positive selection of methods and habits that lead to publication . How can natural selection operate on research methodology ? There are no research \u2018genes\u2019 . But science is a cultural activity , and such activities change through evolutionary processes [ 19 \u2013 25 ] . Philosophers of science such as Campbell [ 19 ] , Popper [ 26 ] and Hull [ 27 ] have discussed how scienti\ufb01c theories evolve by variation and selection retention . But scienti\ufb01c methods also develop in this way . Laboratory methods can propagate either directly , through the production of graduate students who go on to start their own labs , or indirectly , through prestige - biased adoption by researchers in other labs . Methods which are associated with greater success in academic careers will , other things being equal , tend to spread . The requirements for natural selection to produce design are easy to satisfy . Darwin outlined the logic of natural selection as requiring three conditions : ( i ) There must be variation . ( ii ) That variation must have consequences for survival or reproduction . ( iii ) Variation must be heritable . In this case , there are no biological traits being passed from scienti\ufb01c mentors to apprentices . However , research practices do vary . That variation has consequences\u2014habits that lead to publication lead to obtaining highly competitive research positions . And variation in practice is partly heritable , in the sense that apprentices acquire research habits and statistical procedures from mentors and peers . Researchers also acquire research practice from successful role models in their \ufb01elds , even if they do not personally know them . Therefore , when researchers are rewarded primarily for publishing , then habits which promote publication are naturally selected . Unfortunately , such habits can directly undermine scienti\ufb01c progress . This is not a new argument . But we attempt to substantially strengthen it . We support the argument both empirically and analytically . We \ufb01rst review evidence that institutional incentives are likely to increase the rate of false discoveries . Then we present evidence from a literature review of repeated calls for improved methodology , focusing on the commonplace and easily understood issue of statistical 3 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . power . We show that despite over 50 years of reviews of low statistical power and its consequences , there has been no detectable increase . While the empirical evidence is persuasive , it is not conclusive . It is equally important to demonstrate that our argument is logically sound . Therefore , we also analyse a formal model of our argument . Inspecting the logic of the selection - for - bad - science argument serves two purposes . First , if the argument cannot be made to work in theory , then it cannot be the correct explanation , whatever the status of the evidence . Second , formalizing the argument produces additional clarity and the opportunity to analyse and engineer interventions . To represent the argument , we de\ufb01ne a dynamical model of research behaviour in a population of competing agents . We assume that all agents have the utmost integrity . They never cheat . Instead , research methodology varies and evolves due to its consequences on hiring and retention , primarily through successful publication . As a result our argument applies even when researchers do not directly respond to incentives for poor methods . We show that the persistence of poor research practice can be explained as the result of the natural selection of bad science . 2 . Institutional incentives for scientific researchers The rate at which new papers are added to the scienti\ufb01c literature has steadily increased in recent decades [ 28 , 29 ] . This is partly due to more opportunities for collaboration , resulting in more multi - author papers [ 30 , 31 ] . However , the increases in publication rate may also be driven by changing incentives . Recently , Brischoux & Angelier [ 32 ] looked at the career statistics of junior researchers hired by the French CNRS in evolutionary biology between 2005 and 2013 . They found persistent increases in the average number of publications at the time of hiring : newly hired biologists now have almost twice as many publications as they did 10 years ago ( 22 in 2013 versus 12 . 5 in 2005 ) . These numbers re\ufb02ect intense competition for academic research positions . The world\u2019s universities produce many more PhDs than there are permanent academic positions for them to \ufb01ll [ 33 \u2013 35 ] , and while this problem has escalated in recent years , it has been present for at least two decades [ 36 ] . Such competition is all the more challenging for researchers who graduate from any but the most prestigious universities , who face additional discrimination on the job market [ 37 ] . Although there may be jobs available outside of academia\u2014 indeed , often better - paying jobs than university professorships\u2014tenure - track faculty positions at major research universities come with considerable prestige , \ufb02exibility and creative freedom , and remain desirable . Among those who manage to get hired , there is continued competition for grants , promotions , prestige and placement of graduate students . Given this competition , there are incentives for scientists to stand out among their peers . Only the top graduate students can become tenure - track professors , and only the top assistant professors will receive tenure and high pro\ufb01le grants . Recently , the Nobel laureate physicist Peter Higgs , who pioneered theoretical work in the search for fundamental particles in the 1960s , lamented \u2018Today I wouldn\u2019t get an academic job . It\u2019s as simple as that . I don\u2019t think I would be regarded as productive enough\u2019 ( The Guardian , 6 Dec 2013 ) . Of course , such a statement is speculative ; a young Peter Higgs might well get a job today . But he might not be particularly successful . Evidence suggests that only a very small proportion of scientists produce the bulk of published research and generate the lion\u2019s share of citation impact [ 38 ] , and it is these researchers who are likely to be the source of new labs and PhDs . Supporting this argument is evidence that most new tenure - track positions are \ufb01lled by graduates from a small number of elite universities\u2014typically those with very high publication rates [ 37 ] . One method of distinguishing oneself might be to portray one\u2019s work as groundbreaking . And indeed , it appears that the innovation rate has been skyrocketing . Or claims at innovation , at any rate . In the years between 1974 and 2014 , the frequency of the words \u2018innovative\u2019 , \u2018groundbreaking\u2019 and \u2018novel\u2019 in PubMed abstracts increased by 2500 % or more [ 39 ] . As it is unlikely that individual scientists have really become 25 times more innovative in the past 40 years , one can only conclude that this language evolution re\ufb02ects a response to increasing pressures for novelty , and more generally to stand out from the crowd . Another way to distinguish oneself is through sheer volume of output . Substantial anecdotal evidence suggests that number of publications is an overwhelmingly important factor in search committee decision - making . Output may be combined with impact\u2014some researchers place emphasis on metrics such the h - index , de\ufb01ned as the largest number h such that an individual has h publications with at least h citations each [ 40 ] . Yet volume alone is often perceived as a measure of researcher quality , particularly for early - career researchers who have not yet had the time to accrue many citations . Although the degree to which publication output is used for evaluation\u2014as well as what , exactly , constitutes acceptable productivity\u2014varies by discipline , our argument applies to all \ufb01elds in which the number of published papers , however scaled , is used as a benchmark of success . 4 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Whenever a quantitative metric is used as a proxy to assess a social behaviour , it becomes open to exploitation and corruption [ 1 , 41 , 42 ] . This is often summarized more pithily as \u2018when a measure becomes a target , it ceases to be a good measure\u2019 . For example , since the adoption of the h - index , researchers have been observed to arti\ufb01cially in\ufb02ate their indices through self - citation [ 43 ] and even a clever type of quasi - fraud . With the goal of illustrating how the h - index system might be gamed , researchers created six fake papers under a fake author that cited their papers extensively [ 44 ] . They posted these papers to a university server . When the papers were indexed by Google , their h - indices on Google Scholar increased dramatically . Strategies that target evaluative metrics may be invented by cheaters , but they may propagate through their consequences . 1 Our argument is that incentives to generate a lengthy CV have particular consequences on the ecology of scienti\ufb01c communities . The problem stems , in part , from the fact that positive results in support of some novel hypothesis are more likely to be published than negative results , particularly in high - impact journals . For example , until recently , the Journal of Personality and Social Psychology refused to publish failed replications of novel studies it had previously published [ 45 ] . Many researchers who fail to garner support for their hypotheses do not even bother to submit their results for publication [ 46 ] . The response to these incentives for positive results is likely to increase false discoveries . If researchers are rewarded for publications and positive results are generally both easier to publish and more prestigious than negative results , then researchers who can obtain more positive results\u2014 whatever their truth value\u2014will have an advantage . Indeed , researchers sometimes fail to report those hypotheses that fail to generate positive results ( lest reporting such failures hinders publication ) [ 47 ] , even though such practices make the publication of false positives more likely [ 11 ] . One way to better ensure that a positive result corresponds to a true effect is to make sure one\u2019s hypotheses have \ufb01rm theoretical grounding and that one\u2019s experimental design is suf\ufb01ciently well powered [ 15 ] . However , this route takes effort and is likely to slow down the rate of production . An alternative way to obtain positive results is to employ techniques , purposefully or not , that drive up the rate of false positives . Such methods have the dual advantage of generating output at higher rates than more rigorous work , while simultaneously being more likely to generate publishable results . Although sometimes replication efforts can reveal poorly designed studies and irreproducible results , this is more the exception than the rule [ 48 ] . For example , it has been estimated that less than 1 % of all psychological research is ever replicated [ 49 ] and failed replications are often disputed [ 50 \u2013 52 ] . Moreover , even \ufb01rmly discredited research is often cited by scholars unaware of the discreditation [ 53 ] . Thus , once a false discovery is published , it can permanently contribute to the metrics used to assess the researchers who produced it . False discoveries in the literature can obviously result from fraud or p - hacking [ 54 ] , but there are many ways that false discoveries can be generated by perfectly well - intentioned researchers . These are easy to spot when the results are absurd ; for example , following standard methods for their \ufb01elds , researchers have observed a dead Atlantic salmon exhibiting neural responses to emotional stimuli [ 55 ] and university students apparently demonstrating \u2018pre - cognitive\u2019 abilities to predict the outcome of a random number generator [ 56 ] . However , false discoveries are usually not identi\ufb01able at a glance , which is why they are problematic . In some cases , poor or absent theory amounts to hypotheses being generated almost at random , which substantially lowers the probability that a positive result represents a real effect [ 13 , 15 , 57 ] . Interpretation of results after data is collected can also generate false positives through biased selection of statistical analyses [ 58 ] and post hoc hypothesis formation [ 8 ] . Campbell\u2019s Law , stated in this paper\u2019s epigraph , implies that if researchers are incentivized to increase the number of papers published , they will modify their methods to produce the largest possible number of publishable results rather than the most rigorous investigations . We propose that this incentivization can create selection pressures for the cultural evolution of poor methods that produce publishable \ufb01ndings at the expense of rigour . It is important to recognize that this process does not require any direct strategizing on the part of individuals . To draw an analogy from biological evolution , giraffe necks increased over time , not because individual animals stretched their necks , but because those with longer necks could more effectively monopolize food resources and thereby produce more offspring . In the same way , common methodologies in scienti\ufb01c communities can change over time not only because established researchers are strategically changing their methods , but also because certain researchers are more successful in transmitting their methods to younger generations . Incentives in\ufb02uence both the patterns of innovation and the nature of selection . Importantly , it is not necessary that strategic innovation be common in order for \ufb01tter strategies to spread rapidly in the research population when they appear . 1 Incentives to increase one\u2019s h - index may also encourage researchers to engage in high - risk hypothesizing , particularly on \u2018hot\u2019 research topics , because they can increase their citation count by being corrected . 5 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 . Case study : statistical power has not improved As a case study , let us consider the need for increased statistical power . Statistical power refers to the probability that a statistical test will correctly reject the null hypothesis when it is false , given information about sample size , effect size and likely rates of false positives 2 [ 59 ] . Because many effects in the biomedical , behavioural and social sciences are small [ 60 \u2013 62 ] , it is important for studies to be suf\ufb01ciently high - powered . On the other hand , low - powered experiments are substantially easier to perform when studying humans or other mammals , particularly in cases where the total subject pool is small , the experiment requires expensive equipment or data must be collected longitudinally . It is clear that low - powered studies are more likely to generate false negatives . Less clear , perhaps , is that low power can also increase the false discovery rate and the likelihood that reported effect sizes are in\ufb02ated , due to their reduced ability to mute stochastic noise [ 13 , 63 , 64 ] . The nature of statistical power sets up a contrast . In an imaginary academic environment with purely cooperative incentives to reveal true causal models of nature , increasing power is often a good idea . Perfect power is impossible . But very low power brings no bene\ufb01ts . However , in a more realistic academic environment that only publishes positive \ufb01ndings and rewards publication , an ef\ufb01cient way to succeed is to conduct low power studies . Why ? Such studies are cheap and can be farmed for signi\ufb01cant results , especially when hypotheses only predict differences from the null , rather than precise quantitative differences and trends [ 3 ] . We support this prediction in more detail with the model in a later section\u2014it is possible for researchers to publish more by running low - power studies , but at the cost of \ufb01lling the scienti\ufb01c literature with false discoveries . For the moment , we assert the common intuition that there is a con\ufb02ict of interest between the population and the individual scientist over statistical power . The population bene\ufb01ts from high power more than individual scientists do . Science does not possess an \u2018invisible hand\u2019 mechanism through which the naked self - interest of individuals necessarily brings about a collectively optimal result . Scientists have long recognized this con\ufb02ict of interest . The \ufb01rst highly cited exhortation to increase statistical power was published by Cohen in 1962 , as a reaction to the alarmingly low power of most psychology studies at the time [ 65 ] . The response , or lack of response , to such highly cited exhortations serves as a \ufb01rst - order test of which side of the con\ufb02ict of interest is winning . A little over two decades after Cohen\u2019s original paper , two meta - analyses by Sedlmeier & Gigerenzer [ 66 ] and Rossi [ 67 ] examined a total of 25 reviews of statistical power in the psychological and social science literature between 1960 and 1984 . These studies found that not only was statistical power quite low , but that in the intervening years since Cohen [ 65 ] , no improvement could be discerned . Recently , Vankov et al . [ 68 ] observed that statistical power in psychological science appears to have remained low to the present day . We expanded this analysis by performing a search on Google Scholar among papers that had cited Sedlmeier & Gigerenzer [ 66 ] ( the more highly cited of the two previous meta - analyses ) using the search terms \u2018statistical power\u2019 and \u2018review\u2019 . We collected all papers that contained reviews of statistical power from published papers in the social , behavioural and biological sciences , and found 19 studies from 16 papers published between 1992 and 2014 . Details about our methods and data sources can be found in the appendix . We focus on the statistical power to detect small effects of the order d = 0 . 2 , the kind most commonly found in social science research . These data , along with the data from Sedlmeier & Gigerenzer [ 66 ] and Rossi [ 67 ] , are plotted in \ufb01gure 1 . Statistical power is quite low , with a mean of only 0 . 24 , meaning that tests will fail to detect small effects when present three times out of four . More importantly , statistical power shows no sign of increase over six decades ( R 2 = 0 . 00097 ) . The data are far from a complete picture of any given \ufb01eld or of the social and behavioural sciences more generally , but they help explain why false discoveries appear to be common . Indeed , our methods may overestimate statistical power because we draw only on published results , which were by necessity suf\ufb01ciently powered to pass through peer review , usually by detecting a non - null effect . 3 Why does low power , a conspicuous and widely appreciated case of poor research design , persist ? There are two classes of explanations . First , researchers may respond directly to incentives and strategically reason that these poor methods help them maximize career success . In considering the 2 We differentiate statistical power from power more generally ; the latter is the probability that one\u2019s methods will return a positive result given a true effect , and is a Gestalt property of one\u2019s methods , not only of one\u2019s statistical tools . 3 It is possible that the average statistical power of research studies does , in fact , sometimes increase for a period , but is then quelled as a result of publication bias . Publication in many disciplines is overwhelmingly biased towards non - null results . Therefore , average powercouldincrease ( atleastintheshortrun ) , but , especiallyifhypothesisselectiondidnotimprovedatacorrespondingrate , itmight simply lead to the scenario in which higher - powered studies merely generated more null results , which are less likely to be published [ 69 ] . Labs employing lower - powered studies would therefore have an advantage , and lower - powered methods would continue to propagate . 6 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1955 1965 1975 1985 1995 2005 2015 s t a ti s ti ca l po w e r Figure 1 . Averagestatisticalpowerfrom44reviewsofpaperspublishedinjournalsinthesocialandbehaviouralsciencesbetween1960 and 2011 . Data are power to detect small effect sizes ( d = 0 . 2 ) , assuming a false - positive rate of \u03b1 = 0 . 05 , and indicate both very low power ( mean = 0 . 24 ) but also no increase over time ( R 2 = 0 . 00097 ) . persistence of low statistical power , Vankov et al . [ 68 ] suggest the following : \u2018Scientists are human and will therefore respond ( consciously or unconsciously ) to incentives ; when personal success ( e . g . promotion ) is associated with the quality and ( critically ) the quantity of publications produced , it makes more sense to use \ufb01nite resources to generate as many publications as possible\u2019 ( p . 1037 , emphasis in original ) . Second , researchers may be trying to do their best , but selection processes reward misunderstandings and poor methods . Traditions in which people believe these methods achieve broader scienti\ufb01c goals will have a competitive edge , by crowding out alternative traditions in the job market and limited journal space . Vankov et al . provide some evidence for this among psychologists : widespread misunderstandings of power and other statistical issues . What these misunderstandings have in common is that they all seem to share the design feature of making positive results\u2014true or false\u2014more likely . Misunderstandings that hurt careers are much less commonplace . Reality is probably a mix of these explanations , with some individuals and groups exhibiting more of one than the other . Our working assumption is that most researchers have internalized scienti\ufb01c norms of honest conduct and are trying their best to reveal true explanations of important phenomena . However , the evidence available is really insuf\ufb01cient . Analyses of data in evolutionary and historical investigations are limited in their ability to infer dynamical processes [ 70 ] , particularly when those data are sparse , as with investigations of scienti\ufb01c practices . To really investigate such a population dynamic hypothesis , we need a more rigorous demonstration of its logic . 4 . An evolutionary model of science To validate the logic of the natural selection of bad science , we develop and analyse a dynamical population model . Such a model simultaneously veri\ufb01es the logic of a hypothesis and helps to re\ufb01ne its predictions , so that it can be more easily falsi\ufb01ed [ 71 , 72 ] . Our model is evolutionary : researchers compete for prestige and jobs in which the currency of \ufb01tness is number of publications , and more successful labs will have more progeny that inherit their method . The foundation is a previously published mathematical model [ 15 ] in which a population of scientists investigate both novel and previously tested hypotheses and attempt to communicate their results to produce a body of literature . Variation in research quality , replication rates and publication biases are all present in the dynamics . That model was de\ufb01ned analytically and solved exactly for the probability that a given hypothesis is true , conditional on any observed publication record . Here we extend this model to focus on a \ufb01nite , heterogeneous population of N labs . We assume the following : \u2014 Each lab has a characteristic power , the ability to positively identify a true association . This power is not only the formal power of a statistical procedure . Rather it arises from the entire chain of inference . \u2014 Increasing power also increases the rate of false positives , unless effort is exerted . \u2014 Increasing effort decreases the productivity of a lab , because it takes longer to perform rigorous research . 7 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . It is important to understand why increasing power tends to also increase false positives without the application of effort . It is quite easy to produce a method with very high power : simply declare support for every proposed association , and you are guaranteed to never mistakenly mark a true hypothesis as false . Of course , this method will also yield many false positives . One can decrease the rate of false positives by requiring stronger evidence to posit the existence of an effect . However , doing so will also decrease the power\u2014because even true effects will sometimes generate weak or noisy signals\u2014unless effort is exerted to increase the size and quality of one\u2019s dataset . This follows readily from the logic of signal detection theory [ 73 ] . There are alternative ways to formalize this trade - off . For example , we might instead assume signal threshold and signal noise to be characteristics of a lab . That would hew closer to a signal detection model . What we have done here instead is focus on a behavioural hypothesis , that researchers tend to reason as if a research hypothesis were true and select methods that make it easier to \ufb01nd true effects . This maintains or increases effective power , even if nominal statistical power is low . But it simultaneously exaggerates false - positive rates , because many hypotheses are in fact not true . Formally , either system could be translated to the other , but each represents a different hypothesis about the behavioural dimensions along which inferential methods change . The trade - off between effective power and research effort has also been invoked in less formal arguments [ 64 , p . 288 ] . Given their inferential characteristics , labs perform experiments and attempt to publish their results . But positive results are easier to publish than negative results . Publications are rewarded ( through prestige , grant funding and more opportunities for graduate students ) , and more productive labs are in turn more likely to propagate their methodologies onto new labs ( such as those founded by their successful graduate students ) . New labs resemble , but are not identical to , their parent labs . The model has two main stages : Science and Evolution . In the Science stage , each lab has the opportunity to select a hypothesis , investigate it experimentally and attempt to communicate their results through a peer - reviewed publication . Hypotheses are assumed to be strictly true or false , though their essential epistemological states cannot be known with certainty but can only be estimated using experiments . In the Evolution stage , an existing lab may \u2018die\u2019 ( cease to produce new research ) , making room in the population for a new lab that adopts the methods of a progenitor lab . More successful labs are more likely to produce progeny . 4 . 1 . Science The Science stage consists of three phases : hypothesis selection , investigation and communication . Every time step , each lab i , in random order , begins a new investigation with probability h ( e i ) , where e i is the characteristic effort that the lab exerts towards using high quality experimental and statistical methods . Higher effort results in better methods ( more speci\ufb01cally , it allows higher power for a given false - positive rate ) , and also results in a lengthier process of performing and analysing experiments . 4 The probability of tackling a new hypothesis on a given time step is h ( e i ) = 1 \u2212 \u03b7 log 10 e i , ( 4 . 1 ) where \u03b7 is a constant re\ufb02ecting the extent to which increased effort lowers the lab\u2019s rate of producing new research . For simplicity , e i is bounded between 1 and 100 for all labs . In all our simulations \u03b7 = 0 . 2 , which ensured that h stayed non - negative . This value is fairly conservative ; in most of our simulations , labs were initialized with a fairly high rate of investigation , h = 0 . 63 , at their highest level of effort . However , our results are robust for any monotonically decreasing , non - negative function h ( e i ) . If an experimental investigation is undertaken , the lab selects a hypothesis to investigate . With probability r i , this hypothesis will be one that has been supported at least once in the published literature\u2014i . e . it will be an attempt to replicate prior research . Otherwise , the lab selects a novel , untested hypothesis ( at least as represented in the literature ) for investigation . Novel hypotheses are true with probability b , the base rate for any particular \ufb01eld . It is currently impossible to accurately calculate the base rate ; it may be as high as 0 . 1 for some \ufb01elds but it is likely to be much lower in many others [ 13 \u2013 15 ] . 4 We acknowledge that even methods with low power and / or high rates of false positives may require considerable time and energy to apply , and might therefore be considered effortful in their own right . For readers with such an objection , we propose substituting the word rigour to de\ufb01ne the variable e i . 8 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 power f a l s e - po s iti v e r a t e e = 1 e = 10 e = 75 Figure 2 . The relationship between power and false - positive rate , modified by effort , e . Runs analysed in this paper were initialized with e 0 = 75 ( shown in orange ) , such that \u03b1 = 0 . 05 when power is 0 . 8 . Labs vary in their investigatory approach . Each lab i has a characteristic power , W i , associated with its methodology , which de\ufb01nes its probability of correctly detecting a true hypothesis , Pr ( + | T ) . Note again that power here is a characteristic of the entire investigatory process , not just the statistical procedure . It can be very high , even when sample size is low . The false - positive rate , \u03b1 i , must be a convex function of power . We assume it to be a function of both power and the associated rigour and effort of the lab\u2019s methodology : \u03b1 i = W i 1 + ( 1 \u2212 W i ) e i . ( 4 . 2 ) This relationship is depicted in \ufb01gure 2 . What this functional relationship re\ufb02ects is the necessary signal - detection trade - off : \ufb01nding all the true hypotheses necessitates labelling all hypotheses as true . Likewise , in order to never label a false hypothesis as true , one must label all hypotheses as false . Note that the false discovery rate \u2014the proportion of positive results that are in fact false positives\u2014is determined not only by the false - positive rate , but also by the base rate , b . When true hypotheses are rarer , false discoveries will occur more frequently [ 13 , 15 ] . A feature of our functional approach is that increases in effort do not also increase power ; the two variables are independent in our model . This is unexpected from a pure signal detection perspective . Reducing signal noise , by increasing experimental rigour , will tend to in\ufb02uence both true and false - positive rates . However , our approach makes sense when power is maintained by contingent procedures that are invoked conditional on the nature of the evidence [ 58 ] . If we instead view researchers\u2019 inferential procedures as \ufb01xed , prior to seeing the data , the independence of effort and power is best seen as a narrative convenience : effort is the sum of all methodological behaviours that allow researchers to increase their power without also increasing their rate of false positives . All investigations yield either positive or negative results : a true hypothesis yields a positive result with probability W i , and a false hypothesis yields a positive result with probability \u03b1 i . Upon obtaining these results , the lab attempts to communicate them to a journal for publication . We assume that positive novel results are always publishable , while negative novel results never are . Both con\ufb01rmatory and discon\ufb01rmatory replications are published , possibly at lower rates . We adopt this framework in part because it approximates widespread publication practices . Positive and negative replications are communicated with probabilities c R + and c R \u2212 , respectively . Communicated results enter the published literature . Labs receive pay - offs for publishing their results , and these pay - offs\u2014which may be thought of in terms of factors such as prestige , in\ufb02uence or funding\u2014make their methodologies more likely to propagate in the scienti\ufb01c community at large . Labs accumulate pay - offs throughout their lifespans . Pay - offs differ for novel and replicated results , with the former being larger . Pay - offs can also accrue when other research groups attempt to replicate a lab\u2019s original hypothesis . These pay - offs can be positive , in cases of con\ufb01rmatory replication , or punitive , in cases of failure to replicate . Values for pay - offs and all other parameter values are given in table 1 . 9 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Table 1 . Global model parameters . parameter definition values tested N number of labs 100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . b base rate of true hypotheses 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . r 0 initial replication rate for all labs { 0 , 0 . 01 , 0 . 2 , 0 . 5 } . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . e 0 initial effort for all labs 75 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . w 0 initial power for all labs 0 . 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . \u03b7 influence of effort on productivity 0 . 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . c R + probability of publishing positive replication 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . c R \u2212 probability of publishing negative replication 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V N pay - off for publishing novel result 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V R + pay - off for publishing positive replication 0 . 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V R \u2212 pay - off for publishing negative replication 0 . 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V O + pay - off for having novel result replicated 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V O \u2212 pay - off for having novel result fail to replicate \u2212 100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . d number of labs sampled for death and birth events 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . \u03bc r probability of r mutation { 0 , 0 . 01 } . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . \u03bc e probability of e mutation { 0 , 0 . 01 } . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . \u03bc w probability of w mutation { 0 , 0 . 01 } . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . \u03c3 r standard deviation of r mutation magnitude 0 . 01 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . \u03c3 e standard deviation of e mutation magnitude 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . \u03c3 w standard deviation of w mutation magnitude 0 . 01 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 . 2 . Evolution At the end of each time step , once all labs have had the opportunity to perform and communicate research , there follows a stage of selection and replication . First , a lab is chosen to die . A random sample of d labs is obtained , and the oldest lab of these is selected to die , so that age correlates coarsely but not perfectly with fragility . If multiple labs in the sample are equally old , one of these is selected at random . The dying lab is then removed from the population . Next , a lab is chosen to reproduce . A random sample of d labs is obtained , and from among these the lab with the highest accumulated pay - off is chosen to reproduce . This skews reproduction towards older labs as well as towards more successful labs , which agrees with the observation that more established labs and scientists are more in\ufb02uential . However , because age does not correlate with the rate at which pay - offs are accumulated , selection will favour those strategies which can increase pay - offs most quickly . A new lab with an age of zero is then created , imperfectly inheriting the attributes of its parent lab . Power , effort and replication rate independently \u2018mutate\u2019 with probabilities \u03bc w , \u03bc e and \u03bc r , respectively . If a mutation occurs , the parameter value inherited from the parent lab is increased or decreased by an amount drawn from a Gaussian distribution with a mean of zero and a standard deviation of \u03c3 w , \u03c3 e or \u03c3 r for power , effort or replication rate . If a mutation modi\ufb01es a parameter\u2019s value below or above its prescribed range , it is truncated to the minimum or maximum value . It is , of course , unrealistic to assume that all researchers have the same expected \u2018lifespan\u2019 . Many researchers disappear from academia quite early in their careers , shortly after receiving their degree or completing a postdoc . Nevertheless , the simplifying assumption of equal expected lifespan is , if anything , a conservative one for our argument . If the factors that lead a researcher to drop out of the \ufb01eld early in his or her career are unrelated to publication , then this is irrelevant to the model\u2014it is simply noise , and incorporated in the stochastic nature of the death algorithm . On the other hand , if the ability to progress in one\u2019s career is directly in\ufb02uenced by publications , then our model is , if anything , a muted demonstration of the strength of selection on publication quantity . 10 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . After death and reproduction , the published literature is truncated to a manageable size . Because few \ufb01elds have more than a million relevant publications ( assessed through searches for broad key words in Google Scholar\u2014most \ufb01elds probably have far fewer relevant papers ) , because most replications target recent work ( e . g . 90 % of replications in psychology target work less than 10 years old [ 14 ] ) and because decreased data availability for older papers makes replication more dif\ufb01cult [ 74 ] , we restrict the size of the literature available for replication to the 1 million most recently published results . This assumption was made in part to keep the computational load at manageable levels and to allow for long evolutionary trajectories . At the end of every evolutionary stage , the oldest papers in the published literature are removed until the total number is less than or equal to 1 million . 4 . 3 . Simulation runs Table 1 describes all the model parameters and their default and range values . The data are averaged from 50 runs for each set of parameter values tested . Each simulation was run for 1 million time steps , with data collected every 2000 time steps . In addition to the evolvable traits of individual research groups , we also collected data on the false discovery rate ( FDR ) for all positive results published each time step . Our goal here is illustrative rather than exploratory , and so our analysis focuses on a few illuminative examples . For the interested reader , we provide the Java code for the full model in the electronic supplementary material . 5 . Simulation results While our model is thankfully much simpler than any real scienti\ufb01c community , it is still quite complex . So we introduce bits of the dynamics in several sections , so that readers can learn from the isolated forces . 5 . 1 . The natural selection of bad science First , ignore replication and focus instead on the production of novel \ufb01ndings ( r 0 = \u03bc r = 0 ) . Consider the evolution of power in the face of constant effort ( \u03bc e = 0 , \u03bc w = 0 . 01 ) . As power increases so too does the rate of false positives . Higher power therefore increases the total number of positive results a lab obtains , and hence the number of papers communicated . As such , both W and \u03b1 go to unity in our simulations , such that all results are positive and thereby publishable ( \ufb01gure 3 ) . It is easy to prove that unimpeded power ( and false - positive rate ) will increase to unity . Fitness is directly tied to the number of publications a lab produces , so anything that increases number of publications also increases \ufb01tness . The probability that a novel hypothesis is true is the base rate , b . For a lab with power W and effort e , the probability that a test of a novel hypothesis will lead to a positive ( and therefore publishable ) result is therefore Pr ( + ) = Pr ( + | T ) Pr ( T ) + Pr ( + | F ) Pr ( F ) = bW + ( 1 \u2212 b ) W 1 + ( 1 \u2212 W ) e . ( 5 . 1 ) By differentiation , it can be shown that this probability is strictly increasing as a function of W . This implies that if selection favours ever higher discovery rates , power will continue to increase to the point at which it is matched by a countervailing force , that is , by whatever factors limit the extent to which power and false - positive rate can change . This \ufb01rst case is unrealistic . Research groups would never be able to get away with methods for which all hypotheses are supported , not to mention that increasing power without bound is not pragmatically feasible . However , one can imagine institutions to keep power relatively high , insofar as at least some aspects of power are directly measurable . At least some aspects of experimental power , such as statistical power , are directly measurable . False positives , on the other hand , are notoriously dif\ufb01cult for peer reviewers to assess [ 75 , 76 ] . If power is measurable and kept high through institutional enforcement , publication rates can still be increased by reducing the effort needed to avoid false positives . We ran simulations in which power was held constant but in which effort could evolve ( \u03bc w = 0 , \u03bc e = 0 . 01 ) . Here selection favoured labs who put in less effort towards ensuring quality work , which increased publication rates at the cost of more false discoveries ( \ufb01gure 4 ) . When the focus is on the production of novel results and negative \ufb01ndings are dif\ufb01cult to publish , institutional incentives for publication quantity select for the continued degradation of scienti\ufb01c practices . 11 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 0 . 2 0 0 . 4 0 . 6 0 . 8 1 . 0 20 000 40 000 60 000 80 000 100 000 time power a FDR Figure 3 . Power evolves . The evolution of mean power ( W ) , false - positive rate ( \u03b1 ) and false discovery rate ( FDR ) . 0 20 40 60 80 0 0 . 2 0 0 . 4 0 . 6 0 . 8 1 . 0 200000 400000 600000 800000 1000000 e ff o r t a , F D R time a FDR effort Figure 4 . Effort evolves . The evolution of low mean effort corresponds to evolution of high false - positive and false discovery rates . 5 . 2 . The ineffectuality of replication Novel results are not and should not be the sole focus of scienti\ufb01c research . False discoveries and ambiguous results are inevitable with even the most rigorous methods . The only way to effectively separate true from false hypotheses is through repeated investigation , including both direct and conceptual replication [ 14 , 15 ] . Can replication impede the evolution of bad science ? If replications are dif\ufb01cult to publish , this is unlikely . On the other hand , consider a scenario in which replication efforts are easy to publish and the failure to replicate a lab\u2019s novel result causes substantial losses to the prestige of that lab . Might the introduction of replication then counteract selection for low - effort methodologies ? We repeated the previous runs , but this time initialized each lab so that 1 % of all investigations would be replications of ( randomly selected ) hypotheses from the published literature . This is consistent with empirical estimations of replication rates in psychology [ 49 ] . We then allowed the replication rate to evolve through mutation and selection ( r 0 = 0 . 01 , \u03bc w = 0 , \u03bc r = \u03bc e = 0 . 01 ) . Conditions in these runs were highly favourable to replication . We assumed that all replication efforts would be publishable , and worth half as much as a novel result in terms of evolutionary \ufb01tness ( e . g . in terms of associated prestige ) . Additionally , having one\u2019s original novel result successfully replicated by another lab increased the value of that \ufb01nding by 10 % , but having one\u2019s original result fail to replicate was catastrophic , carrying a penalty equal to 100 times the value of the initial \ufb01nding ( i . e . V O + = 0 . 1 , V O \u2212 = \u2212 100 ) . This last assumption may appear unrealistically harsh , but research indicates that retractions can lead to a substantial decrease in citations to researchers\u2019 prior work [ 77 ] . In addition , some have suggested that institutions should incentivize reproducibility by providing some sort of \u2018money - back guarantee\u2019 if results fail to replicate [ 78 ] , which could end up being highly punitive to individual researchers . More generally , these assumptions are extremely favourable to the idea that replication might deter poor methods , since false positives carry the highest risk of failing to replicate . 12 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 20 40 60 80 0 0 . 2 0 0 . 4 0 . 6 0 . 8 1 . 0 200000 400000 600000 800000 1000000 e ff o r t a , F D R , r e p li ca ti on time a FDR replication effort Figure 5 . The coevolution of effort and replication . 0 20 40 60 80 0 200 000 400 000 600 000 800 000 1 000 000 e ff o r t time no replication 25 % replication 50 % replication Figure 6 . The evolution of effort when zero , 25 % or 50 % of all studies performed are replications . We found that the mean rate of replication evolved slowly but steadily to around 0 . 08 . Replication was weakly selected for , because although publication of a replication was worth only half as much as publication of a novel result , it was also guaranteed to be published . On the other hand , allowing replication to evolve could not stave off the evolution of low effort , because low effort increased the false - positive rate to such high levels that novel hypotheses became more likely than not to yield positive results ( \ufb01gure 5 ) . As such , increasing one\u2019s replication rate became less lucrative than reducing effort and pursuing novel hypotheses . Because effort decreased much more rapidly than replication rates increased , we considered the possibility that substantially higher replication rates might be effective at keeping effort rates high , if only they could be enforced . To test this hypothesis , we ran simulations in which the replication rate could not mutate ( \u03bc r = 0 ) but was initialized to very high levels . High rates of replication did slow the decline of effort , but even extremely high replication rates ( as high as 50 % ) did not stop effort from eventually bottoming out ( \ufb01gure 6 ) . We note emphatically that we are not suggesting that highly punitive outcomes for failures to replicate are desirable , since even high - quality research will occasionally fail to replicate . Rather , we are pointing out that even in the presence of such punitive outcomes , institutional incentives for publication quality will still select for increasingly low - quality methods . 5 . 3 . Why is replication not sufficient ? Replication is not suf\ufb01cient to curb the natural selection of bad science because the top performing labs will always be those who are able to cut corners . Replication allows those labs with poor methods to be penalized , but unless all published studies are replicated several times ( an ideal but implausible scenario ) , some labs will avoid being caught . In a system such as modern science , with \ufb01nite career 13 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . r = 0 r = 0 . 01 r = 0 . 05 300 200 100 0 400 300 200 100 0 400 400 300 200 100 0 \u2013 200 \u2013 150 \u2013 100 \u2013 50 0 50 300 200 100 0 c oun t total pay - off total pay - off total pay - off 0 10 20 30 low effort high effort low effort high effort low effort high effort 40 50 \u2013 500 \u2013 400 \u2013 300 \u2013 200 \u2013 100 0 100 \u2013 1500 \u2013 1200 \u2013 900 \u2013 600 \u2013 300 0 ( b ) ( a ) ( c ) Figure 7 . Lab pay - offs from the non - evolutionary model . Each graph shows count distributions for high and low effort labs\u2019 total pay - offs after 110 time steps , 100 of which included replication . ( a \u2013 c ) Total count for each pay - off is totalled from 50 runs for each condition . Panel ( c ) includes an inset that displays the same data as the larger graph , but for a narrower range of pay - offs . The punishment for having one\u2019s novel result fail to replicate is orders of magnitude greater than the benefit of publishing , reflected in the discrete peaks in ( b ) and ( c ) . opportunities and high network connectivity , the marginal return for being in the top tier of publications may be orders of magnitude higher than an otherwise respectable publication record [ 37 , 38 , 79 ] . Within our evolutionary model it was dif\ufb01cult to lay bare the precise relationship between replication , effort and reproductive success , because all these parameters are entangled . We wanted to understand exactly why the most successful labs appear to be those with low effort even when failure to replicate was highly punitive , making low effort potentially quite costly . To do this , we created a simpli\ufb01ed version of our model that omitted any evolution . Power was \ufb01xed at 0 . 8 and all labs were either high effort ( H ) or low effort ( L ) . High effort labs had effort e H = 0 . 75 , corresponding to a false - positive rate of \u03b1 = 0 . 05 and an investigation rate of h = 0 . 625 . Low effort labs had effort e L = 0 . 15 , corresponding to a false - positive rate of \u03b1 = 0 . 2 and an investigation rate of h = 0 . 765 . The population was initialized with 50 % high effort labs . Labs conducted research and attempted to communicate their \ufb01ndings as in the original model . We allowed ten time steps without any replication to establish a baseline body of literature , and then ran the simulation for 100 additional time steps , equivalent to the expected lifespan of a lab in the evolutionary model . During this time , each hypothesis investigated was a replication with probability r . This simpli\ufb01ed model allowed us to examine the distribution of the pay - offs ( resulting from both the bene\ufb01ts of successful publications and punishment from failed replications ) and to directly compare high and low effort labs . Figure 7 shows the distributions of pay - offs from three replication rates . Without replication , low effort labs have an unambiguous advantage . As the rate of replication increases , the mean pay - off for high effort labs can surpass the mean pay - off for low effort labs , as the former are less likely to be punished . However , the laws of probability dictate that some labs will escape either producing false positives or being caught doing so , and among these the very highest performers will be those who exert low effort . When the top labs have disproportionate in\ufb02uence on funding and graduate student success , this type of small advantage can cascade to continuously select for lower effort and increasing numbers of false discoveries , as seen in our evolutionary model . 6 . Discussion Incentives drive cultural evolution . In the scienti\ufb01c community , incentives for publication quantity can drive the evolution of poor methodological practices . We have provided some empirical evidence that this occurred , as well as a general model of the process . If we want to improve how our scienti\ufb01c culture functions , we must consider not only the individual behaviours we wish to change , but also the social forces that provide affordances and incentives for those behaviours [ 1 , 80 ] . We are hardly the \ufb01rst to consider a need to alter the incentives for career success in science [ 68 , 69 , 81 \u2013 89 ] . However , we are the \ufb01rst to illustrate the evolutionary logic of how , in the absence of change , the existing incentives will necessarily lead to the degradation of scienti\ufb01c practices . An incentive structure that rewards publication quantity will , in the absence of countervailing forces , select for methods that produce the greatest number of publishable results . This , in turn , will lead to the natural selection of poor methods and increasingly high false discovery rates . Although we have focused on false discoveries , there are additional negative repercussions of this kind of incentive structure . Scrupulous research on dif\ufb01cult problems may require years of intense work before yielding coherent , 14 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . publishable results . If shallower work generating more publications is favoured , then researchers interested in pursuing complex questions may \ufb01nd themselves without jobs , perhaps to the detriment of the scienti\ufb01c community more broadly . Good science is in some sense a public good , and as such may be characterized by the con\ufb02ict between cooperation and free riding . We can think of cooperation here as the opportunity to create group - bene\ufb01cial outcomes ( i . e . quality research ) at a personal cost ( i . e . diminished \u2018\ufb01tness\u2019 in terms of academic success ) . To those familiar with the game theory of cooperative dilemmas , it might therefore appear that continued contributions to the public good\u2014cooperation rather than free riding\u2014could be maintained through the same mechanisms known to promote cooperation more generally , including reciprocity , monitoring and punishment [ 90 ] . However , the logic of cooperation requires that the bene\ufb01t received by cooperators can be measured in the same units as the pay - off to free riders : i . e . units of evolutionary \ufb01tness . It is possible that coalitions of rigorous scientists working together will generate greater output than less rigorous individuals working in isolation . And indeed , there has been an increase in highly collaborative work in many \ufb01elds [ 30 , 31 ] . Nevertheless , such collaboration may also be a direct response to incentives for publication quantity , as contributing a small amount to many projects generates more publications than does contributing a large amount to few projects . Cooperation in the sense of higher quality research provides a public good in the sense of knowledge , but not in the sense of \ufb01tness for the cultural evolution of methodology . Purely bottom - up solutions are therefore unlikely to be suf\ufb01cient . That said , changing attitudes about the assessment of scientists is vital to making progress , and is a driving motivation for this presentation . Institutional change is dif\ufb01cult to accomplish , because it requires coordination on a large scale , which is often costly to early adopters [ 91 , 92 ] . Yet such change is needed to ensure the integrity of science . It is therefore worth considering the types of institutions we might want . It might appear that journals and peer reviewers need only to adopt increasingly strict standards as bars to publication . However , although this may place some limits on the extent to which individuals can game the system , it still affords the opportunity to do so [ 93 ] and incentives to do so will exist as long as success is tied to publication . Punishing individuals for failure to replicate their original results is unlikely to be effective at stopping the evolution of bad science . Many important discoveries initially appear unlikely , and designing clear experiments is dif\ufb01cult . Eliminating false positives entirely is likely to be impossible [ 86 ] . In addition , failed replications may themselves be false negatives . Moreover , replication is dif\ufb01cult or impossible for some \ufb01elds , such as those involving large clinical samples or historical data . An overemphasis on replication as the saviour of all science is biased in favour of certain \ufb01elds over others . It is therefore inadvisable to be overly harsh on researchers when their results fail to replicate . A more drastic suggestion is to alter the \ufb01tness landscape entirely by changing the selection pressures : the incentives for success . This is likely to be quite dif\ufb01cult . Consider that the stakeholders who fund and support science may have incentives that are not always aligned with those of active researchers [ 82 ] . For example , funders may expect \u2018deliverables\u2019 in the form of published papers , which in turn may pressure scientists to conduct research in such a manner to maximize those deliverables , even if incentives for promotion or hiring are changed . Another impediment is the constraint on time and cognitive resources on the part of evaluators . The quality of a researcher is dif\ufb01cult to assess , particularly since there are many ways to be a good scientist , making assessment a high - dimensionality optimization problem . Quantitative metrics such as publication rates and impact factor are used partly because they are simple and provide clear , unambiguous comparisons between researchers . Yet these are precisely the qualities that allow such metrics to be exploited . In reality , the true \ufb01tness landscape of scienti\ufb01c career success is multidimensional . Although some publications are probably necessary , there are other routes to success beside the accrual of a lengthy CV . We should support institutions that facilitate these routes , particularly when they encourage high quality research , and resist the temptation or social pressure to paper count . We believe that our interpretation of the model is quite robust , provided that at least one of the following two criteria are met : ( i ) negative results are harder to publish or ( ii ) lowering effort increases the rate of a lab\u2019s output . We chose parameter values that we deemed reasonable and strove to err in a direction unfavourable to our hypotheses , as a guard against con\ufb01rmation bias . For example , the replication rate was set extremely high , and failure to replicate was made extremely punitive . However , we have not presented a full sensitivity analysis . So we note again that we provide our model source code in the electronic supplementary material , so that interested readers can verify and extend our analysis . Our model treats each publication of a novel result as equivalent , but of course they are not . Instead , the most prestigious journals receive the most attention and are the most heavily cited [ 94 ] . Anecdotally , we have heard many academics\u2014ranging from graduate students to full professors\u2014discussing job candidates largely in terms of the prestigious journals they either published in or failed to publish in . 15 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Consideration of journal prestige would change our model somewhat , but the implications would be similar , as the existence of such journals creates pressure to publish in those journals at all costs . The major change to our model would be additional selection for highly surprising results , which are more likely to be false . Investigations have found that the statistical power of papers published in prestigious ( high impact factor ) journals is no different from those with lower impact factors [ 84 ] , while the rate of retractions for journals is positively correlated with impact factor [ 95 ] . Although this is likely to be at least partly due to the increased attention paid to those papers , it is well known that high impact journals often reject competent papers deemed insuf\ufb01ciently novel . Our model presents a somewhat pessimistic picture of the scienti\ufb01c community . Let us be clear : many scientists are well aware of the dif\ufb01culties surrounding evaluation , and many hiring and grant committees take efforts to evaluate researchers on the quality of their work rather than the quantity of their output or where it is published . Moreover , we believe that many if not most scientists are truly driven to discover truth about the world . It would be overly cynical to believe that scientists are driven only by extrinsic incentives , fashioning researchers to conform to the expectations of Homo economicus . However , a key feature of our evolutionary model is that it assumes no ill intent on the part of the actors , and does not assume that anyone actively alters their methods in response to incentives . Rather , selection pressures at the institutional level favoured those research groups that , for whatever reason , used methods that generated a higher volume of published work . Any active social learning , such as success - biased copying , will serve only to accelerate the pace of evolution for such methods . Despite incentives for productivity , many scientists employ rigorous methods and learn new things about the world all the time that are validated by replication or by being effectively put into practice . In other words , there is still plenty of good science out there . One reason is that publication volume is rarely the only determinant of the success or failure of a scientist\u2019s career . Other important factors include the importance of one\u2019s research topic , the quality of one\u2019s work , and the esteem of one\u2019s peers . The weight of each factor varies among disciplines , and in some \ufb01elds such factors may work positively to promote behaviours leading to high - quality research , particularly when selection for those behaviours is enculturated into institutions or disciplinary norms . In such cases , this may be suf\ufb01cient to counteract the negative effects of incentives for publication volume , and so maintain high levels of research quality . If , on the other hand , success is largely determined by publication output or related quantitative metrics , then those who care about quality research should be on high alert . In which direction the scale tips in one\u2019s own \ufb01eld is a critical question for anyone interested in the future of science . Whenever quantitative metrics are used as proxies to evaluate and reward scientists , those metrics become open to exploitation if it is easier to do so than to directly improve the quality of research . Institutional guidelines for evaluation at least partly determine how researchers devote their energies , and thereby shape the kind of science that gets done . A real solution is likely to be patchwork , in part because accurately rewarding quality is dif\ufb01cult . 5 Real merit takes time to manifest , and scrutinizing the quality of another\u2019s work takes time from already busy schedules . Competition for jobs and funding is stiff , and reviewers require some means to assess researchers . Moreover , individuals differ on their criteria for excellence . Boiling down an individual\u2019s output to simple , objective metrics , such as number of publications or journal impacts , entails considerable savings in terms of time , energy and ambiguity . Unfortunately , the long - term costs of using simple quantitative metrics to assess researcher merit are likely to be quite great . If we are serious about ensuring that our science is both meaningful and reproducible , we must ensure that our institutions incentivize that kind of science . Ethics . Our meta - analysis used only previously published data . All simulated scientists were humanely euthanized . Data accessibility . All data for our meta - analysis were drawn from previously published material . Our methods and sources are documented in the electronic supplementary material . The Java source code for our agent - based model is likewise available in the electronic supplementary material . Authors\u2019 contributions . P . E . S . and R . M . conceived and designed the research and wrote the paper . P . E . S . carried out the meta - analysis and wrote the model code and analysed its output . All authors gave \ufb01nal approval for publication . Competinginterests . We have no competing interests . Funding . This research received no speci\ufb01c grant from any funding agency in the public , commercial or not - for - pro\ufb01t sectors . Acknowledgements . For critical feedback on earlier drafts of this manuscript , we thank Clark Barrett , William Baum , Bret Beheim , Monique Borgerhoff Mulder , John Bunce , Emily Newton , Karthik Panchanathan , Joel Steele , Peter Trimmer and Bruce Winterhalder . 5 Not to mention that any measure of quality is partly dependent on ever - changing disciplinary , institutional and departmental needs . 16 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References 1 . CampbellDT . 1976 Assessingtheimpactofplanned socialchange . Hanover , NH : ThePublicAffairs Center . 2 . WassersteinRL , LazarNA . 2016TheASA\u2019sstatement on p - values : context , process , andpurpose . Am . Stat . 70 , 129 \u2013 133 . ( doi : 10 . 1080 / 00031305 . 2016 . 1154108 ) 3 . MeehlPE . 1967Theory - testinginpsychologyand physics : amethodologicalparadox . Phil . Sci . 34 , 103 \u2013 115 . ( doi : 10 . 1086 / 288135 ) 4 . CohenJ . 1994Theearthisround ( p < . 05 ) . Am . Psychol . 49 , 997 \u2013 1003 . ( doi : 10 . 1037 / 0003 - 066X . 49 . 12 . 997 ) 5 . EnserinkM . 2012FinalreportonStapelblamesfield asawhole . Science 338 , 1270 \u2013 1271 . ( doi : 10 . 1126 / science . 338 . 6112 . 1270 ) 6 . VulE , HarrisC , WinkielmanP , PashlerH . 2009 PuzzlinglyhighcorrelationsinfMRIstudiesof emotion , personality , andsocialcognition . Perspect . Psychol . Sci . 4 , 274 \u2013 290 . ( doi : 10 . 1111 / j . 1745 - 6924 . 2009 . 01125 . x ) 7 . PetersonD . 2016Thebabyfactory : difficultresearch objects , disciplinarystandards , andtheproduction ofstatisticalsignificance . Socius 2 , 1 \u2013 10 . ( doi : 10 . 1177 / 2378023115625071 ) 8 . KerrNL . 1998HARKing : hypothesizingafterthe resultsareknown . Pers . Soc . Psychol . Rev . 2 , 196 \u2013 217 . ( doi : 10 . 1207 / s15327957pspr0203 _ 4 ) 9 . JohnLK , LoewensteinG , PrelecD . 2012Measuring theprevalenceofquestionableresearchpractices withincentivesfortruthtelling . Psychol . Sci . 23 , 524 \u2013 532 . ( doi : 10 . 1177 / 0956797611430953 ) 10 . IoannidisJPA . 2014Researchaccomplishmentsthat aretoogoodtobetrue . IntensiveCareMed . 40 , 99 \u2013 101 . ( doi : 10 . 1007 / s00134 - 013 - 3100 - z ) 11 . SimmonsJP , NelsonLD , SimonsohnU . 2011 False - positivepsychology : undisclosedflexibilityin datacollectionandanalysisallowspresenting anythingassignificant . Psychol . Sci . 22 , 1359 \u2013 1366 . ( doi : 10 . 1177 / 0956797611417632 ) 12 . HortonR . 2015Offline : whatismedicine\u2019s5sigma ? Lancet 385 , 1380 . ( doi : 10 . 1016 / S0140 - 6736 ( 15 ) 60696 - 1 ) 13 . IoannidisJPA . 2005Whymostpublishedresearch findingsarefalse . PLoSMed . 2 , e124 . ( doi : 10 . 1371 / journal . pmed . 0020124 ) 14 . PashlerH , HarrisCR . 2012Isthereplicabilitycrisis overblown ? Threeargumentsexamined . Perspect . Psychol . Sci . 7 , 531 \u2013 536 . ( doi : 10 . 1177 / 1745691612 463401 ) 15 . McElreathR , SmaldinoPE . 2015Replication , communication , andthepopulationdynamicsof scientificdiscovery . PLoSONE 10 , e0136088 . ( doi : 10 . 1371 / journal . pone . 0136088 ) 16 . IoannidisJPA . 2005Contradictedandinitially strongereffectsinhighlycitedclinicalresearch . J . Am . Med . Assoc . 294 , 218 \u2013 228 . ( doi : 10 . 1001 / jama . 294 . 2 . 218 ) 17 . IoannidisJPA . 2012Whyscienceisnotnecessarily self - correcting . Perspect . Psychol . Sci . 7 , 645 \u2013 654 . ( doi : 10 . 1177 / 1745691612464056 ) 18 . SteenRG , CasadevallA , FangFC . 2013Whyhasthe numberofscientificretractionsincreased ? PLoSONE 8 , e68397 . ( doi : 10 . 1371 / journal . pone . 0068397 ) 19 . CampbellDT . 1965Variationandselectiveretention insocio - culturalevolution . In Socialchangein developingareas : areinterpretationofevolutionary theory ( edsHBarringer , GBlanksten , RMack ) , pp . 19 \u2013 49 . Cambridge , MA : Schenkman . 20 . SkinnerBF . 1981Selectionbyconsequences . Science 213 , 501 \u2013 504 . ( doi : 10 . 1126 / science . 7244649 ) 21 . BoydR , RichersonPJ1985 Cultureandthe evolutionaryprocess . Chicago , IL : Universityof ChicagoPress . 22 . MesoudiA2011 Culturalevolution : howDarwinian theorycanexplainhumancultureandsynthesizethe socialsciences . Chicago , IL : UniversityofChicago Press . 23 . WhitenA , HindeRA , LalandKN , StringerCB . 2011 Cultureevolves . Phil . Trans . R . Soc . B 366 , 938 \u2013 948 . ( doi : 10 . 1098 / rstb . 2010 . 0372 ) 24 . SmaldinoPE . 2014Theculturalevolutionof emergentgroup - leveltraits . Behav . BrainSci . 37 , 243 \u2013 295 . ( doi : 10 . 1017 / S0140525X13001544 ) 25 . AcerbiA , MesoudiA . 2015Ifweareallcultural Darwinianswhat\u2019sthefussabout ? Clarifyingrecent disagreementsinthefieldofculturalevolution . Biol . Philos . 30 , 481 \u2013 503 . ( doi : 10 . 1007 / s10539 - 015 - 9490 - 2 ) 26 . PopperK1979 Objectiveknowledge : anevolutionary approach . Oxford , UK : OxfordUniversityPress . 27 . HullDL . 1988 Scienceasaprocess : anevolutionary accountofthesocialandconceptualdevelopmentof science . Chicago , IL : UniversityofChicagoPress . 28 . LarsenPO , vonInsM . 2010Therateofgrowthin scientificpublicationandthedeclineincoverage providedbyScienceCitationIndex . Scientometrics 84 , 575 \u2013 603 . ( doi : 10 . 1007 / s11192 - 010 - 0202 - z ) 29 . BornmannL , MutzR . 2015Growthrateofmodern science : abibliometricanalysisbasedonthe numberofpublicationsandcitedreferences . J . Assoc . Inf . Sci . Technol . 66 , 2215 \u2013 2222 . ( doi : 10 . 1002 / asi . 23329 ) 30 . NaboutJC , ParreiraMR , TeresaFB , CarneiroFM , da CunhaHF , deSouzaOndei L , CaramoriSS , Soares TN . 2015Publish ( inagroup ) orperish ( alone ) : the trendfromsingle - tomulti - authorshipinbiological papers . Scientometrics 102 , 357 \u2013 364 . ( doi : 10 . 1007 / s11192 - 014 - 1385 - 5 ) 31 . WardilL , HauertC . 2015Cooperationand coauthorshipinscientificpublication . Phys . Rev . E 91 , 012825 . ( doi : 10 . 1103 / PhysRevE . 91 . 012825 ) 32 . BrischouxF , AngelierF . 2015Academia\u2019s never - endingselectionforproductivity . Scientometrics 103 , 333 \u2013 336 . ( doi : 10 . 1007 / s11192 - 015 - 1534 - 5 ) 33 . CyranoskiD , GilbertN , LedfordH , NayarA , YahiaM . 2011ThePhDfactory . Nature 472 , 276 \u2013 279 . ( doi : 10 . 1038 / 472276a ) 34 . SchillebeeckxM , MaricqueB , LewisC . 2013The missingpiecetochangingtheuniversityculture . Nat . Biotechnol . 31 , 938 \u2013 941 . ( doi : 10 . 1038 / nbt . 2706 ) 35 . PowellK . 2015Thefutureofthepostdoc . Nature 520 , 144 \u2013 147 . ( doi : 10 . 1038 / 520144a ) 36 . KerrE . 1995Toomanyscientists , toofewacademic jobs . Nat . Med . 1 , 14 . ( doi : 10 . 1038 / nm0195 - 14 ) 37 . ClausetA , ArbesmanS , LarremoreDB . 2015 Systematicinequalityandhierarchyinfaculty hiringnetworks . Sci . Adv . 1 , e1400005 . ( doi : 10 . 1126 / sciadv . 1400005 ) 38 . IoannidisJPA , BoyackKW , KlavansR . 2014Estimates ofthecontinuouslypublishingcoreinthescientific workforce . PLoSONE 9 , e101698 . ( doi : 10 . 1371 / journal . pone . 0101698 ) 39 . VinkersV , TijdinkJ , OtteW . 2015Useofpositiveand negativewordsinscientificPubMedabstracts between1974and2014 : retrospectiveanalysis . Br . Med . J . 351 , h6467 . ( doi : 10 . 1136 / bmj . h6467 ) 40 . HirschJE . 2005Anindextoquantifyanindividual\u2019s scientificresearchoutput . Proc . NatlAcad . Sci . USA 102 , 16569 \u2013 16572 . ( doi : 10 . 1073 / pnas . 0507655102 ) 41 . GoodhartC . 1975Problemsofmonetary management : theUKexperience . In Papersin monetaryeconomics , volumeI . Sydney , Australia : ReserveBankofAustralia . 42 . LucasRE . 1976Econometricpolicyevaluation : a critique . In Carnegie - Rochesterconferenceserieson publicpolicy , vol . 1 , pp . 19 \u2013 46 . Amsterdam , The Netherlands : Elsevier . 43 . BartneckC , KokkelmansS . 2011Detecting h - index manipulationthroughself - citationanalysis . Scientometrics 87 , 85 \u2013 98 . ( doi : 10 . 1007 / s11192 - 010 - 0306 - 5 ) 44 . L\u00f3pez - C\u00f3zarED , Robinson - Garc\u00edaN , Torres - Salinas D . 2014TheGoogleScholarexperiment : howto indexfalsepapersandmanipulatebibliometric indicators . J . Assoc . Inf . Sci . Technol . 65 , 446 \u2013 454 . ( doi : 10 . 1002 / asi . 23056 ) 45 . AldhousP . 2011Journalrejectsstudiescontradicting precognition . NewSci . Seehttps : / / www . newscientist . com / article / dn20447 - journal - rejects - studies - contradicting - precognition . 46 . FrancoA , MalhotraN , SimonovitsG . 2014 Publicationbiasinthesocialsciences : unlockingthe filedrawer . Science 345 , 1502 \u2013 1505 . ( doi : 10 . 1126 / science . 1255484 ) 47 . FrancoA , SimonovitsG , MalhotraN . 2015 Underreportinginpoliticalsciencesurvey experiments : comparingquestionnairesto publishedresults . Polit . Anal . 23 , 306 \u2013 312 . ( doi : 10 . 1093 / pan / mpv006 ) 48 . SchmidtS . 2009Shallwereallydoitagain ? The powerfulconceptofreplicationisneglectedinthe socialsciences . Rev . Gen . Psychol . 13 , 90 \u2013 100 . ( doi : 10 . 1037 / a0015108 ) 49 . MakelMC , PluckerJA , HegartyB . 2012Replications inpsychologyresearch : howoftendotheyreally occur ? Perspect . Psychol . Sci . 7 , 537 \u2013 542 . ( doi : 10 . 1177 / 1745691612460688 ) 50 . BissellM . 2013Reproducibility : therisksofthe replicationdrive . Nature 503 , 333 \u2013 334 . ( doi : 10 . 1038 / 503333a ) 51 . SchnallS . 2014Cleandata : statisticalartefactswash outreplicationefforts . Soc . Psychol . 45 , 315 \u2013 320 . ( doi : 10 . 1027 / 1864 - 9335 / a000204 ) 52 . KahnemanD . 2014Anewetiquetteforreplication . Soc . Psychol . 45 , 310 \u2013 311 . 53 . CampanarioJM . 2000Fraud : retractedarticlesare stillbeingcited . Nature 408 , 288 . ( doi : 10 . 1038 / 35042753 ) 54 . AschwandenC . 2015Scienceisn\u2019tbroken . FiveThirtyEight . Seehttp : / / fivethirtyeight . com / features / science - isnt - broken / . 55 . BennettCM , MillerM , WolfordG . 2009Neural correlatesofinterspeciesperspectivetakinginthe post - mortemAtlanticsalmon : anargumentfor multiplecomparisonscorrection . Neuroimage 47 , S125 . ( doi : 10 . 1016 / S1053 - 8119 ( 09 ) 71202 - 9 ) 17 r s o s . r o y a l s o c i e t y p u b l i s h i n g . o r g R . S o c . o p e n s c i . 3 : 1 60 3 8 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 . WagenmakersE - J , WetzelsR , BorsboomD , vander MaasHL . 2011Whypsychologistsmustchangethe waytheyanalyzetheirdata : thecaseofpsi : commentonBem ( 2011 ) . J . Pers . Soc . Psychol . 100 , 426 \u2013 432 . ( doi : 10 . 1037 / a0022790 ) 57 . GigerenzerG . 1998Surrogatesfortheories . Theory Psychol . 8 , 195 \u2013 204 . ( doi : 10 . 1177 / 095935439808 2006 ) 58 . GelmanA , LokenE . 2014Thestatisticalcrisisin science . Am . Sci . 102 , 460 \u2013 465 . ( doi : 10 . 1511 / 2014 . 111 . 460 ) 59 . CohenJ . 1992Statisticalpoweranalysis . Curr . Dir . Psychol . Sci . 1 , 98 \u2013 101 . ( doi : 10 . 1111 / 1467 - 8721 . ep10768783 ) 60 . RichardFD , BondCF , Stokes - ZootaJJ . 2003One hundredyearsofsocialpsychologyquantitatively described . Rev . Gen . Psychol . 7 , 331 \u2013 363 . ( doi : 10 . 1037 / 1089 - 2680 . 7 . 4 . 331 ) 61 . AguinisH , BeatyJC , BoikRJ , PierceCA . 2005Effect sizeandpowerinassessingmoderatingeffectsof categoricalvariablesusingmultipleregression : a 30 - yearreview . J . Appl . Psychol . 90 , 94 \u2013 107 . ( doi : 10 . 1037 / 0021 - 9010 . 90 . 1 . 94 ) 62 . K\u00fchbergerA , FritzA , ScherndlT . 2014Publication biasinpsychology : adiagnosisbasedonthe correlationbetweeneffectsizeandsamplesize . PLoSONE 9 , e105825 . ( doi : 10 . 1371 / journal . pone . 0105825 ) 63 . ButtonKS , IoannidisJPA , MokryszC , NosekBA , Flint J , RobinsonESJ , Munaf\u00f2MR . 2013Powerfailure : whysmallsamplesizeunderminesthereliabilityof neuroscience . Nat . Rev . Neurosci . 14 , 365 \u2013 376 . ( doi : 10 . 1038 / nrn3475 ) 64 . LakensD , EversERK . 2014Sailingfromtheseasof chaosintothecorridorofstability : practical recommendationstoincreasetheinformational valueofstudies . Perspect . Psychol . Sci . 9 , 278 \u2013 292 . ( doi : 10 . 1177 / 1745691614528520 ) 65 . CohenJ . 1962Thestatisticalpowerof abnormal - socialpsychologicalresearch : areview . J . Abnorm . Soc . Psychol . 65 , 145 \u2013 153 . ( doi : 10 . 1037 / h0045186 ) 66 . SedlmeierP , GigerenzerG . 1989Dostudiesof statisticalpowerhaveaneffectonthepowerof studies ? Psychol . Bull . 105 , 309 \u2013 316 . ( doi : 10 . 1037 / 0033 - 2909 . 105 . 2 . 309 ) 67 . RossiJS . 1990Statisticalpowerofpsychological research : whathavewegainedin20years ? J . Consult . Clin . Psychol . 58 , 646 \u2013 656 . ( doi : 10 . 1037 / 0022 - 006X . 58 . 5 . 646 ) 68 . VankovI , BowersJ , Munaf\u00f2MR . 2014Onthe persistenceoflowpowerinpsychologicalscience . Q . J . Exp . Psychol . 67 , 1037 \u2013 1040 . ( doi : 10 . 1080 / 17470218 . 2014 . 885986 ) 69 . NosekBA , SpiesJR , MotylM . 2012Scientificutopia II . Restructuringincentivesandpracticesto promotetruthoverpublishability . Perspect . Psychol . Sci . 7 , 615 \u2013 631 . ( doi : 10 . 1177 / 1745691612459058 ) 70 . SmaldinoPE , CalanchiniJ , PickettCL . 2015Theory developmentwithagent - basedmodels . Organ . Psychol . Rev . 5 , 300 \u2013 317 . ( doi : 10 . 1177 / 204138 6614546944 ) 71 . WimsattWC . 1987Falsemodelsasmeanstotruer theories . In Neutralmodelsinbiology ( edsMNitecki , AHoffman ) , pp . 23 \u2013 55 . Oxford , UK : Oxford UniversityPress . 72 . EpsteinJM . 2008Whymodel ? J . Artif . Soc . Soc . Simul . 11 , 12 . 73 . MacMillanNA , CreelmanCD1991 Detectiontheory : a user\u2019sguide . Cambridge , UK : CambridgeUniversity Press . 74 . VinesTH etal . 2014Theavailabilityofresearchdata declinesrapidlywitharticleage . Curr . Biol . 24 , 94 \u2013 97 . ( doi : 10 . 1016 / j . cub . 2013 . 11 . 014 ) 75 . MacArthurD . 2012Faceuptofalsepositives . Nature 487 , 427 \u2013 428 . ( doi : 10 . 1038 / 487427a ) 76 . Eyre - WalkerA , StoletzkiN . 2013Theassessmentof science : therelativemeritsofpost - publication review , theimpactfactor , andthenumberof citations . PLoSBiol . 11 , e1001675 . ( doi : 10 . 1371 / journal . pbio . 1001675 ) 77 . LuSF , JinGZ , UzziB , JonesB . 2013Theretraction penalty : evidencefromthewebofscience . Sci . Rep . 3 , 3146 . ( doi : 10 . 1038 / srep03146 ) 78 . RosenblattM . 2016Anincentive - basedapproachfor improvingdatareproducibility . Sci . Transl . Med . 8 , 336ed5 . ( doi : 10 . 1126 / scitranslmed . aaf5003 ) 79 . FrankRH . 2011 TheDarwineconomy : liberty , competition , andthecommongood . Princeton , NJ : PrincetonUniversityPress . 80 . WilsonDS , HayesSC , BiglanA , EmbryDD . 2014 Evolvingthefuture : towardascienceofintentional change . Behav . BrainSci . 37 , 395 \u2013 460 . ( doi : 10 . 1017 / S0140525X13001593 ) 81 . NosekBA , LakensD . 2014Registeredreports : a methodtoincreasethecredibilityofpublished results . Soc . Psychol . 45 , 137 \u2013 141 . ( doi : 10 . 1027 / 1864 - 9335 / a000192 ) 82 . IoannidisJPA . 2014Howtomakemorepublished researchtrue . PLoSMed . 11 , e1001747 . ( doi : 10 . 1371 / journal . pmed . 1001747 ) 83 . FischerJ , RitchieEG , HanspachJ . 2012Anacademia beyondquantity : areplytoLoyola etal . andHalme etal . TrendsEcol . Evol . 27 , 587 \u2013 588 . ( doi : 10 . 1016 / j . tree . 2012 . 08 . 009 ) 84 . BrembsB , ButtonK , Munaf\u00f2M . 2013Deepimpact : unintendedconsequencesofjournalrank . Front . Hum . Neurosci . 7 , 291 . ( doi : 10 . 3389 / fnhum . 2013 . 00291 ) 85 . BegleyCG , BuchanAM , DirnaglU . 2015Institutions mustdotheirpartforreproducibility . Nature 525 , 25 \u2013 27 . ( doi : 10 . 1038 / 525025a ) 86 . MacCounR , PerlmutterS . 2015Hideresultstoseek thetruth . Nature 526 , 187 \u2013 189 . ( doi : 10 . 1038 / 526187a ) 87 . GigerenzerG , MarewskiJN . 2015Surrogatescience : theidolofauniversalmethodforscientific inference . J . Manag . 41 , 421 \u2013 440 . ( doi : 10 . 1177 / 0149206314547522 ) 88 . SillsJ . 2016Measuresofsuccess . Science 352 , 28 \u2013 30 . ( doi : 10 . 1126 / science . 352 . 6281 . 28 ) 89 . SarewitzD . 2016Thepressuretopublishpushes downquality . Nature 533 , 147 . ( doi : 10 . 1038 / 533147a ) 90 . RandDG , NowakMA . 2013Humancooperation . TrendsCogn . Sci . 17 , 413 \u2013 425 . ( doi : 10 . 1016 / j . tics . 2013 . 06 . 003 ) 91 . NorthDC . 1990 Institutions , institutionalchangeand economicperformance . Cambridge , UK : Cambridge UniversityPress . 92 . AokiM . 2007Endogenizinginstitutionsand institutionalchanges . J . Inst . Econ . 3 , 1 \u2013 31 . ( doi : 10 . 1017 / S1744137406000531 ) 93 . MartinBR . 2016Editors\u2019JIF - boostingstrategems : whichareappropriateandwhichnot ? Res . Policy 45 , 1 \u2013 7 . ( doi : 10 . 1016 / j . respol . 2015 . 09 . 001 ) 94 . IoannidisJPA . 2006Concentrationofthemost - cited papersinthescientificliterature : analysisofjournal ecosystems . PLoSONE 1 , e5 . ( doi : 10 . 1371 / journal . pone . 0000005 ) 95 . FangFC , CasadevallA . 2011Retractedscienceand theretractionindex . Infect . Immun . 79 , 3855 \u2013 3859 . ( doi : 10 . 1128 / IAI . 05661 - 11 )",
    "jingHouseholdSecondaryAttack2020": "www . thelancet . com / infection Vol 20 October 2020 1141 Articles Household secondary attack rate of COVID - 19 and associated determinants in Guangzhou , China : a retrospective cohort study Qin - Long Jing * , Ming - Jin Liu * , Zhou - Bin Zhang * , Li - Qun Fang * , Jun Yuan * , An - Ran Zhang , Natalie E Dean , Lei Luo , Meng - Meng Ma , Ira Longini , Eben Kenah , Ying Lu , Yu Ma , Neda Jalali , Zhi - Cong Yang , Yang Yang Summary Background As of June 8 , 2020 , the global reported number of COVID - 19 cases had reached more than 7 million with over 400 000 deaths . The household transmissibility of the causative pathogen , severe acute respiratory syndrome coronavirus 2 ( SARS - CoV - 2 ) , remains unclear . We aimed to estimate the secondary attack rate of SARS - CoV - 2 among household and non - household close contacts in Guangzhou , China , using a statistical transmission model . Methods In this retrospective cohort study , we used a comprehensive contact tracing dataset from the Guangzhou Center for Disease Control and Prevention to estimate the secondary attack rate of COVID - 19 ( defined as the probability that an infected individual will transmit the disease to a susceptible individual ) among household and non - household contacts , using a statistical transmission model . We considered two alternative definitions of household contacts in the analysis : individuals who were either family members or close relatives , such as parents and parents - in - law , regardless of residential address , and individuals living at the same address regardless of relationship . We assessed the demographic determinants of transmissibility and the infectivity of COVID - 19 cases during their incubation period . Findings Between Jan 7 , 2020 , and Feb 18 , 2020 , we traced 195 unrelated close contact groups ( 215 primary cases , 134 secondary or tertiary cases , and 1964 uninfected close contacts ) . By identifying households from these groups , assuming a mean incubation period of 5 days , a maximum infectious period of 13 days , and no case isolation , the estimated secondary attack rate among household contacts was 12\u22194 % ( 95 % CI 9\u22198 \u2013 15\u22194 ) when household contacts were defined on the basis of close relatives and 17\u22191 % ( 13\u22193 \u2013 21\u22198 ) when household contacts were defined on the basis of residential address . Compared with the oldest age group ( \u226560 years ) , the risk of household infection was lower in the youngest age group ( < 20 years ; odds ratio [ OR ] 0\u221923 [ 95 % CI 0\u221911 \u2013 0\u221946 ] ) and among adults aged 20 \u2013 59 years ( OR 0\u221964 [ 95 % CI 0\u221943 \u2013 0\u221997 ] ) . Our results suggest greater infectivity during the incubation period than during the symptomatic period , although differences were not statistically significant ( OR 0\u221961 [ 95 % CI 0\u221927 \u2013 1\u221938 ] ) . The estimated local reproductive number ( R ) based on observed contact frequencies of primary cases was 0\u22195 ( 95 % CI 0\u221941 \u2013 0\u221962 ) in Guangzhou . The projected local R , had there been no isolation of cases or quarantine of their contacts , was 0\u22196 ( 95 % CI 0\u221949 \u2013 0\u221974 ) when household was defined on the basis of close relatives . Interpretation SARS - CoV - 2 is more transmissible in households than SARS - CoV and Middle East respiratory syndrome coronavirus . Older individuals ( aged \u226560 years ) are the most susceptible to household transmission of SARS - CoV - 2 . In addition to case finding and isolation , timely tracing and quarantine of close contacts should be implemented to prevent onward transmission during the viral incubation period . Funding US National Institutes of Health , Science and Technology Plan Project of Guangzhou , Project for Key Medicine Discipline Construction of Guangzhou Municipality , Key Research and Development Program of China . Copyright \u00a9 2020 Elsevier Ltd . All rights reserved . Introduction The ongoing COVID - 19 pandemic , caused by the novel severe acute respiratory syndrome corona virus 2 ( SARS - CoV - 2 ) , has now affected 188 countries worldwide . As of June 8 , 2020 , more than 7 million reported cases and over 400 000 deaths had been reported . 1 Older individuals ( aged \u226570 years ) and individuals with chronic conditions such as diabetes and cardiopulmonary disease are most susceptible to severe disease and death . 2 Efficient viral transmission via droplets and fomites is potentially supplemented by other transmission routes such as aerosol and faecal contamination . 3 , 4 Accumulating evidence suggests that presymptomatic or asymp tomatic carriers can transmit the virus . 5 \u2013 7 Within - household transmission is suspected to have contributed sub stantially to the con - tinued increase in cases in China following the introduction of nationally enforced restric tions on human movement . 8 , 9 Isolation of cases and quarantine of their close contacts at home are frequently recommended as a disease control measure in countries with COVID - 19 outbreaks , but such Lancet Infect Dis 2020 ; 20 : 1141 \u2013 50 Published Online June 17 , 2020 https : / / doi . org / 10 . 1016 / S1473 - 3099 ( 20 ) 30471 - 0 See Comment page 1103 For the Chinese translation of the abstract see Online for appendix 1 * Contributed equally Guangzhou Center for Disease Control and Prevention , Guangzhou , China ( Q - L Jing PhD , Z - B Zhang MPH , J Yuan MD , L Luo PhD , M - M Ma MD , Y Lu MD , Y Ma MD , Z - C Yang MD ) ; Department of Biostatistics , College of Public Health and Health Professions , Emerging Pathogens Institute , University of Florida , Gainesville , FL , USA ( M - J Liu BS , A - R Zhang BS , N E Dean BS , I Longini PhD , N Jalali MS , Y Yang PhD ) ; State Key Laboratory of Pathogen and Biosecurity , Beijing Institute of Microbiology and Epidemiology , Beijing , China ( A - R Zhang , L - Q Fang PhD ) ; Department of Epidemiology , School of Public Health , Cheeloo College of Medicine , Shandong University , Jinan , China ( A - R Zhang ) ; and Division of Biostatistics , College of Public Health , Ohio State University , Columbus , OH , USA ( E Kenah ScD ) Correspondence to : Dr Yang Yang , Department of Biostatistics , College of Public Health and Health Professions , Emerging Pathogens Institute , University of Florida , Gainesville , FL 32611 , USA yangyang @ ufl . edu or Dr Zhi - Cong Yang , Guangzhou Center for Disease Control and Prevention , Guangzhou 510440 , China yangzc @ gzcdc . org . cn Articles 1142 www . thelancet . com / infection Vol 20 October 2020 restrictions are likely to have little or no effect on trans - mission within households . To date , transmissibility of the disease has primarily been assessed at the population level , using mathematical models , or at the individual level in synthetic popula - tions using agent - based models coupled with statistical methods . 10 \u2013 12 Transmissibility within households or through other types of close contact remains under investigated , despite the importance of these social interactions in shaping the overall dynamics of disease spread and in determining the effectiveness of mitigation strategies . 13 Data obtained from contract tracing provide the most accurate information about human - to - human trans - missibility of any infectious pathogen , because trans - missibility can be assessed more accurately by accounting for individual - level exposure history . Available estimates for the secondary attack rate ( defined as the probability that an infected individual will transmit the disease to a susceptible individual [ eg , household or close contacts ] ) of SARS - CoV - 2 are based on contact tracing data for hundreds of cases in Shenzhen and Guangzhou in China , and ten cases in the USA . 14 \u2013 16 These estimates represent the proportion of confirmed infections among all traced contacts , which does not fully account for heterogeneity in individual exposure history , the possi bility of transmission among contacts themselves , or the infection risks from untraced contacts or fomites . We aimed to estimate the secondary attack rate of SARS - CoV - 2 among household and non - household close contacts in Guangzhou , using a statistical transmission model . This model accounts for individual - level exposure history and the potential existence of tertiary cases . We also aimed to assess the effects of age and sex on the infectivity of COVID - 19 cases and susceptibility of their close contacts , and the relative infectivity of COVID - 19 cases during the incubation period compared with the period of illness . Methods Case definition A suspected COVID - 19 case was defined as an individual who met one or more epidemiological criteria ( had travelled to , or resided in , Wuhan or nearby cities in the 14 days before symptom onset ; had contact history with an individual with COVID - 19 [ confirmed by RT - PCR ] in the 14 days before symptom onset ; had contact history with patients who had fever or respiratory symptoms and came from Wuhan or communities with reported COVID - 19 cases in the 14 days before symptom onset ; or was related to a cluster of COVID - 19 cases ) and two or more clinical criteria ( had fever or respiratory symptoms ; had radio - graphical characteristics of pneumonia ; or normal or reduc ed leucocyte counts , or reduced lymphocyte counts during the acute phase of the disease ; appendix 2 p 1 ) . A confirmed case was defined as a suspected case with positive detection of SARS - CoV - 2 nucleic acid by real - time RT - PCR or viral genes that are highly homologous to SARS - CoV - 2 by sequencing using respiratory speci mens . An individual with laboratory confirmation , but without clinical signs , mainly found by community screening and contact tracing , was considered asymptomatic ( appendix 2 pp 2 , 3 ) . In this study , asymptomatic infections were analysed as confirmed cases . Epidemiological investigation and contact tracing Epidemiological investigations were done by county - level Centers for Disease Control and Prevention ( CDC ) Research in context Evidence before this study The transmissibility of severe acute respiratory syndrome coronavirus 2 ( SARS - CoV - 2 ) in household and community settings remains underinvestigated . On April 1 , 2020 , we searched PubMed and medRxiv using the search terms ( \u201cnovel coronavirus\u201d OR \u201cCOVID - 19\u201d OR \u201cSARS - CoV - 2\u201d OR ) AND ( \u201chousehold\u201d OR \u201cfamily\u201d ) AND ( \u201ctransmissibility\u201d OR \u201cattack rate\u201d ) . Our search yielded three articles that investigated household secondary attack rates of COVID - 19 in multiple family clusters . In studies of contact tracing data in Shenzhen and Guangzhou , two cities in southern China , the secondary attack rates were 14\u00b79 % and 10\u00b72 % , respectively , but these estimates represented the proportions of infections among household contacts . In a study of the close contacts of ten US patients with COVID - 19 the estimated household secondary attack rate was 10\u00b75 % , however , the sample size was too small for reliable interpretation and only symptom onset of primary cases was examined . Transmission events from COVID - 19 cases during their incubation period or from asymptomatic carriers have been reported , but infectiousness before symptom onset has not been quantified . Added value of this study Using contact tracing data from patients with COVID - 19 in Guangzhou , China , we implemented a statistical transmission model to estimate secondary attack rate for household and non - household close contacts . To our knowledge , our model is the first to account for heterogeneous individual - level exposure history , tertiary transmission , potential exposure to untraced infection sources , and asymptomatic infections . Additionally , we assessed the effects of age , sex , epidemic phase , and household size on transmissibility of the virus and relative infectivity before and after symptom onset . Implications of all the available evidence SARS - CoV - 2 can be transmitted efficiently within households and during the incubation period of COVID - 19 cases . Because presymptomatic and asymptomatic transmission has been observed , case - isolation alone is inadequate for mitigating the pandemic . Comprehensive tracing and timely quarantine of close contacts of COVID - 19 cases should be implemented to prevent onward transmission during their incubation periods . See Online for appendix 2 Articles www . thelancet . com / infection Vol 20 October 2020 1143 offices in China within 24 h after a suspected or confirmed case was reported ( appendix 2 pp 2 , 3 ) . For each suspected or confirmed case , we recorded demographic , clinical , diagnostic , and occupational data , baseline health conditions , clinical samples and laboratory test results , and exposure history in the 14 days before symptom onset using a standardised investigation form . Typically , a suspected case would be changed to a confirmed case or removed from the national surveillance system once laboratory test results were available . A close contact was defined as an individual who had unprotected close contact ( within 1 m ) with a confirmed case within 2 days before their symptom onset or sample collection , including but not limited to household members , care givers , and individuals in the same workplace , classroom , hospital ward , or transportation vehicle . Close contacts were quarantined at designated places ( eg , hotel rooms ) or at home and followed up for 14 days , and nasal swabs were collected at day 1 and day 14 and tested by RT - PCR ( appendix 2 pp 3 , 4 ) . These data were collected as part of a continuing public health response required by the National Health Commission of China , thus the require - ment for written informed consent was waived . This study was approved by the ethics committee of the Guangzhou CDC ( Guangzhou , China ) . Analyses involving personally identifiable data were done at Guangzhou CDC . Anonymised data were used for all other analyses . In our analyses , individuals who were linked by contact tracing were considered a close contact group . Cases in the same close contact group formed a case cluster . An imported case was defined as an individual who resided in or had travelled to Hubei province ( of which Wuhan is the capital ) in the 2 weeks before symptom onset ; otherwise , individuals were considered a local case . In a close contact group , a local case with symptom onset 1 day or less ( 3 days or less for an imported case ) from the earliest onset day in the close contact group was considered a primary case ; otherwise , this case was considered a secondary case . For asymptomatic infections , the primary or secondary case status was determined by the collection dates of the earliest SARS - CoV - 2 positive specimens . We used two definitions of house hold contacts : individuals who were either family members or close relatives , such as parents and parents - in - law , regardless of residential address , and individuals living at the same address regardless of relationship . Statistical analysis We used standard non - parametric tests ( Fisher\u2019s exact test , the Kruskal - Wallis test , and the Mann - Whitney U test ) to compare characteristics between demographic groups . Similar to the published literature , 14 \u2013 16 we calcu - lated the proportions of confirmed infections among all traced contacts for subgroups and herein , we refer to these proportions as data - based secondary attack rate estimates , which do not account for the fact that infections among contacts are not necessarily secon - dary , and could be tertiary . All summary analyses were done using R statistical software ( version 3 . 6 . 1 ) . The spatial distribution of case clusters was mapped at the com munity level using ArcGIS ( Environmental Systems Research Institute , Redlands , CA , USA ) , with a directed graph indicating potential transmission chains . We estimated effective reproductive numbers ( R t ) based on the contact tracing data ( appendix 2 pp 4 \u2013 6 ) . Since the transmission relationship remains unclear , we investi gated three scenarios defined by the following assump tions : scenario 1 , all imported cases were primary cases , and all secondary cases were infected by primary cases in the same case cluster ; scenario 2 , which was identical to scenario 1 , with the additional assumption that local primary cases might have been infected by earlier cases in other case clusters ; and scenario 3 , which was identical to scenario 2 , with the additional assumption that imported secondary cases were considered secondary cases rather than primary cases . Scenarios 1 and 3 served as the lower and upper bounds of the R t . A chain - binomial statistical model was used to estimate secondary attack rate and local reproductive number , with an expectation - maximisation algorithm to account for uncertainty in the infection time of asymptomatic infections ( appendix 2 pp 6 \u2013 10 ) . 17 Possible distributions of the incubation period and the infectious period were derived from the literature and our previous research on a separate contact tracing database ( appendix 2 pp 10 \u2013 12 ) . 5 , 18 \u2013 20 On the basis of the available literature ( appendix 2 p 10 \u2013 11 ) , we reported the estimates associated with a mean incubation period of 5 days and a maximum infectious period of 13 days as the primary results . Briefly , this model estimated the probabilities , p 1 and p 2 , of viral transmission from an infectious household contact and from an infectious non - household contact ( eg , friends , coworkers , passengers ) respectively , to a susceptible person per daily contact . Additionally , each susceptible person was subject to a constant daily probability , b , of being infected by an unspecified external source , which accounted for untraced contacts and fomites . We assumed that the infectivity of a COVID - 19 case differs between the incubation period and the period after symptom onset ( ie , illness period ) . We used ( D min , D max ) to represent the whole infectious period with the symptom onset day set as day 0 , and modelled the effective daily transmission probability as on days D min \u2264 l < 0 ( incubation period ) and p k ( l ) * \u2013 1 \u2013 p k ( l ) = p k \u2013 1 \u2013 p k \u00d7 OR * p k ( l ) = p k * Articles 1144 www . thelancet . com / infection Vol 20 October 2020 on days 0 \u2264 l < D max ( illness period ) , where the odds ratio ( OR ) measures the relative infectivity of the illness period versus the incubation period . The secondary attack rate was defined as : where \u03c6 * ( l ) was the prespecified relative infectivity level on the l th day of the infectious period based on previous studies , 19 peaking near the time of symptom onset . Herein , we refer to the derived secondary attack rate estimates as model - based . The local reproductive number ( ie , the mean number of infections a symptomatic case could generate via household and non - household contacts ) was defined as : where n 1 ( l ) and n 2 ( l ) were the mean numbers of household and non - household contacts per primary case on day l ( appendix 2 pp 12 , 13 ) . We assessed the effect of age , sex , house hold size , and epidemic phase on susceptibility and infectivity by regressing transmission probabilities on these characteristics associated with either the suscep tible person or the infectious person in each potential trans - mission \u2013 exposure pair ( appendix 2 pp 13 , 14 ) . Additi - onally , we assessed the goodness - of - fit of the model under different settings of the natural history of disease ( appendix 2 pp 14 , 15 ) . Role of the funding source The funder of the study had no role in study design , data collection , data analysis , data interpretation , or the writing of the report . The corresponding authors had full access to all the data in the study , and had final responsibility for the decision to submit for publication . Results Between Jan 7 , 2020 , and Feb 18 , 2020 , 349 laboratory - confirmed SARS - CoV - 2 infections were reported to Guangzhou CDC , among whom 19 ( 5 % ) indi viduals were asymptomatic . Contact tracing identified 195 unrelated close contact groups ( 215 primary cases , 134 secondary or tertiary cases , and 1964 uninfected close contacts ) . The median size of close contact groups was six ( IQR 4 \u2013 10 ; range 1 \u2013 274 ) . For 138 ( 71 % ) of these 195 close contact groups , no secondary cases were identified . Time from symptom onset to hospital admission and to laboratory confirmation was longer for primary cases than for secondary cases and was longer in January than in February ( appendix 2 p 19 ) . Among the 349 cases , the most common symptoms were fever ( 258 [ 74 % ] cases ) , cough ( 207 [ 59 % ] ) , fatigue ( 74 [ 21 % ] ) , sore throat ( 60 [ 17 % ] ) , chills ( 55 [ 16 % ] ) , and myalgia ( 51 [ 15 % ] ) . Radio - graphic abnormality was observed in 238 ( 80 % ) of 349 cases ( appendix 2 p 20 ) . Most patients with COVID - 19 were adults ( aged 20 \u2013 59 years ; table 1 ) . The majority of primary cases ( 158 [ 73 % ] of 215 cases ) and nearly half of all secondary cases ( 66 [ 46 % ] of 134 cases ) had recently travelled to or resided in Hubei province ( referred to as imported cases hereafter ) . The overall data - based secon dary attack rates were 13\u22192 % ( 95 % CI 10\u22199 \u2013 15\u22197 ) among household contacts and 2\u22194 % ( 1\u22196 \u2013 3\u22193 ) among non - household contacts , when household was defined on the basis of close relatives . Within house holds , the data - based secondary attack rates were lower in the youngest age group ( age < 20 years ; 5\u22192 % [ 95 % CI 2\u22194 \u2013 9\u22197 ] ) than the 20 \u2013 59 years age group ( 14\u22198 % [ 95 % CI 11\u22197 \u2013 18\u22194 ] ; p = 0\u00b70009 ) and the oldest age group ( age \u226560 years ; 18\u22194 % [ 12\u22195 \u2013 25\u22196 % ] ; p = 0\u00b70003 ) . Similar findings were observed among non - household contacts , but the differences between age groups were not statistically significant . We identified no significant differences in secondary attack rate between sexes . Secondary attack rate estimates decreased between January and February , 2020 , within and outside households ( both p < 0\u00b70001 ) . When household was defined by residential address , the data - based secondary attack rate among household contacts increased to 17\u22192 % ( 95 % CI 14\u22191 \u2013 20\u22196 ) and was 2\u22196 % ( 1\u22199 \u2013 3\u22196 ) among non - household contacts . Most COVID - 19 cases were reported in the densely populated districts ( where 56 % of the total population of Guangzhou reside ) including Yuexiu , Liwan , Haizhu , Tianhe , and Baiyun ( figure 1 ) . Four clusters with five or more secondary cases ( excluding tertiary cases or further generations ) were identified , with one cluster identified in Yuexiu , Haizhu , Baiyun and Panyu districts , and all primary cases in these clusters were imported ( figure 1B \u2013 D ) . The longest transmission chain had three generations subsequent to the primary case , which occurred in the Panyu district ( figure 1D ) . Five other clusters had two subsequent generations ( figure 1B , figure 1C ) . Most of the reported residential locations of primary and secondary cases within the same clusters were identical , but some non - household secondary cases resided in areas that were distant from the primary cases . Most transmissions occurred between household members ( appendix 2 p 28 ) . The first imported primary case had symptom onset on Jan 7 , 2020 , and arrived in Guangzhou on Jan 13 , 2020 . The earliest local primary case had symptom onset on Jan 16 , 2020 , when at least two imported cases had already been reported in Guangzhou . The number of imported cases peaked at the same time as the epidemic in Guangzhou around Jan 27 , 2020 , 4 days after the lockdown was implemented in Wuhan ( figure 2 ) . After Jan 27 , 2020 , the number of imported cases decreased and the epidemic waned quickly , with only sporadic cases reported by the middle of February . In the early phase of the epidemic , the R t reached 1\u00b74 in scenario 3 , around 1\u00b72 in scenario 2 , and 0\u00b77 in scenario 1 . For all scenarios , by Jan 27 , 2020 , the R t had declined to less than 0\u00b75 , 1 \u2013 \u220f l = D D maxmin [ 1 \u2212 p k ( l ) \u03c6 * ( l ) ] , k = 1 , 2 , * \u2211 k = 1 \u2211 l = DD maxmin { n k ( l ) p k ( l ) \u03c6 ( l ) \u03a0 m = D 2 min l \u2013 1 [ 1 \u2013 p k ( l ) \u03c6 ( l ) ] } * * * * Articles www . thelancet . com / infection Vol 20 October 2020 1145 which is likely to reflect the tightening of control measures in Guangzhou . We excluded 12 close contact groups with primary cases but no recorded contact , one close contact group in which all members ( two asymptomatic primary cases and two uninfected individuals ) stayed in Guangzhou for only 1 day , and 25 contacts with missing data on age or sex . Thus , 182 close contact groups with a total of 332 cases ( 317 symptomatic and 15 asymptomatic ) and 1937 uninfected contacts were included in our transmission analysis . We estimated the secondary attack rates among household and non - household contacts for the combinations of mean incubation period of 4 \u2013 7 days and maximum infectious period of 13 , 16 , and 22 days ( appendix 2 p 17 ) . Assessment of the goodness - of - fit for these settings indicated that all models fit the data satisfactorily , because the model - predicted numbers of infections were consistent with the observed values with only small differences before and after the peak of the epidemic ( appendix 2 p 30 ) . Assuming a mean incubation period of 5 days and maximum infectious period of 13 days , when household contact was defined on the basis of close relatives , assuming there was no case isolation , the estimated Primary cases Secondary cases Uninfected close contacts Overall Data - based secondary attack rate * Household Non - household Household Non - household Household Non - household Close relatives Age , years < 20 10 / 215 ( 5 % ) 9 / 103 ( 9 % ) 1 / 31 ( 3 % ) 163 / 681 ( 24 % ) 70 / 1283 ( 5 % ) 253 / 2313 ( 11 % ) 5\u00b72 % ( 2\u00b74 \u2013 9\u00b77 ) 1\u00b74 % ( 0\u00b704 \u2013 7\u00b76 ) 20 \u2013 59 145 / 215 ( 67 % ) 67 / 103 ( 65 % ) 22 / 31 ( 71 % ) 385 / 681 ( 57 % ) 961 / 1283 ( 75 % ) 1580 / 2313 ( 68 % ) 14\u00b78 % ( 11\u00b77 \u2013 18\u00b74 ) 2\u00b72 % ( 1\u00b74 \u2013 3\u00b74 ) \u226560 60 / 215 ( 28 % ) 27 / 103 ( 26 % ) 8 / 31 ( 26 % ) 120 / 681 ( 18 % ) 247 / 1283 ( 19 % ) 462 / 2313 ( 20 % ) 18\u00b74 % ( 12\u00b75 \u2013 25\u00b76 ) 3\u00b71 % ( 1\u00b74 \u2013 6\u00b71 ) Sex Female 107 / 215 ( 50 % ) 57 / 103 ( 55 % ) 17 / 31 ( 55 % ) 341 / 681 ( 50 % ) 627 / 1283 ( 49 % ) 1149 / 2313 ( 50 % ) 14\u00b73 % ( 11\u00b70 \u2013 18\u00b72 ) 2\u00b76 % ( 1\u00b75 \u2013 4\u00b72 ) Male 108 / 215 ( 50 % ) 46 / 103 ( 45 % ) 14 / 31 ( 45 % ) 335 / 681 ( 49 % ) 651 / 1283 ( 51 % ) 1154 / 2313 ( 50 % ) 12\u00b71 % ( 9\u00b70 \u2013 15\u00b78 ) 2\u00b71 % ( 1\u00b72 \u2013 3\u00b75 ) Month\u2020 January 193 / 215 ( 90 % ) 98 / 103 ( 95 % ) 29 / 31 ( 94 % ) 545 / 681 ( 80 % ) 681 / 1283 ( 53 % ) 1546 / 2313 ( 67 % ) 15\u00b72 % ( 12\u00b76 \u2013 18\u00b73 ) 4\u00b71 % ( 2\u00b78 \u2013 5\u00b78 ) February 22 / 215 ( 10 % ) 5 / 103 ( 5 % ) 2 / 31 ( 6 % ) 136 / 681 ( 20 % ) 602 / 1283 ( 47 % ) 767 / 2313 ( 33 % ) 3\u00b75 % ( 1\u00b72 \u2013 8\u00b71 ) 0\u00b733 % ( 0\u00b704 \u2013 1\u00b72 ) Household size \u22646 people 160 / 215 ( 74 % ) 69 / 103 ( 67 % ) 23 / 31 ( 74 % ) 302 / 681 ( 44 % ) 874 / 1283 ( 68 % ) 1428 / 2313 ( 62 % ) 18\u00b76 % ( 14\u00b78 \u2013 22\u00b79 ) 2\u00b76 % ( 1\u00b76 \u2013 3\u00b78 ) > 6 people 55 / 215 ( 26 % ) 34 / 103 ( 33 % ) 8 / 31 ( 26 % ) 379 / 681 ( 56 % ) 409 / 1283 ( 32 % ) 885 / 2313 ( 38 % ) 8\u00b72 % ( 5\u00b78 \u2013 11\u00b73 ) 1\u00b79 % ( 0\u00b78 \u2013 3\u00b77 ) Origin Imported 158 / 215 ( 73 % ) 59 / 103 ( 57 % ) 3 / 31 ( 10 % ) NA NA NA NA NA Local 57 / 215 ( 27 % ) 44 / 103 ( 43 % ) 28 / 31 ( 90 % ) NA NA NA NA NA Residential address Age , years < 20 10 / 215 ( 5 % ) 8 / 93 ( 9 % ) 2 / 41 ( 5 % ) 117 / 449 ( 26 % ) 116 / 1515 ( 8 % ) 253 / 2313 ( 11 % ) 6\u00b74 % ( 2\u00b78 \u2013 12\u00b72 ) 1\u00b77 % ( 0\u00b72 \u2013 5\u00b70 ) 20 \u2013 59 145 / 215 ( 67 % ) 59 / 93 ( 63 % ) 30 / 41 ( 73 % ) 260 / 449 ( 58 % ) 1086 / 1515 ( 72 % ) 1580 / 2313 ( 68 % ) 18\u00b75 % ( 14\u00b74 \u2013 23\u00b72 ) 2\u00b77 % ( 1\u00b78 \u2013 3\u00b78 ) \u226560 60 / 215 ( 28 % ) 26 / 93 ( 28 % ) 9 / 41 ( 22 % ) 67 / 449 ( 15 % ) 300 / 1515 ( 20 % ) 462 / 2313 ( 20 % ) 28\u00b70 % ( 19\u00b71 \u2013 38\u00b72 ) 2\u00b79 % ( 1\u00b73 \u2013 5\u00b75 ) Sex Female 107 / 215 ( 50 % ) 53 / 93 ( 57 % ) 21 / 41 ( 51 % ) 227 / 449 ( 51 % ) 741 / 1515 ( 49 % ) 1149 / 2313 ( 50 % ) 18\u00b79 % ( 14\u00b75 \u2013 24\u00b70 ) 2\u00b78 % ( 1\u00b77 \u2013 4\u00b72 ) Male 108 / 215 ( 50 % ) 40 / 93 ( 43 % ) 20 / 41 ( 49 % ) 218 / 449 ( 49 % ) 768 / 1515 ( 51 % ) 1154 / 2313 ( 50 % ) 15\u00b75 % ( 11\u00b73 \u2013 20\u00b75 ) 2\u00b75 % ( 1\u00b76 \u2013 3\u00b79 ) Month\u2020 January 193 / 215 ( 90 % ) 88 / 93 ( 95 % ) 39 / 41 ( 95 % ) 362 / 449 ( 81 % ) 864 / 1515 ( 57 % ) 1546 / 2313 ( 67 % ) 19\u00b76 % ( 16\u00b70 \u2013 23\u00b75 ) 4\u00b73 % ( 3\u00b71 \u2013 5\u00b79 ) February 22 / 215 ( 10 % ) 5 / 93 ( 5 % ) 2 / 41 ( 5 % ) 87 / 449 ( 19 % ) 651 / 1515 ( 43 % ) 767 / 2313 ( 33 % ) 5\u00b74 % ( 1\u00b78 \u2013 12\u00b72 ) 0\u00b731 % ( 0\u00b704 \u2013 1\u00b71 ) Household size \u22646 people 188 / 215 ( 87 % ) 79 / 93 ( 85 % ) 32 / 41 ( 78 % ) 309 / 449 ( 69 % ) 1191 / 1515 ( 79 % ) 1799 / 2313 ( 78 % ) 20\u00b74 % ( 16\u00b75 \u2013 24\u00b77 ) 2\u00b76 % ( 1\u00b78 \u2013 3\u00b77 ) > 6 people 27 / 215 ( 13 % ) 14 / 93 ( 15 % ) 9 / 41 ( 22 % ) 140 / 449 ( 31 % ) 324 / 1515 ( 21 % ) 514 / 2313 ( 22 % ) 9\u00b71 % ( 5\u00b71 \u2013 14\u00b78 ) 2\u00b77 % ( 1\u00b72 \u2013 5\u00b71 ) Origin Imported 158 / 215 ( 73 % ) 56 / 93 ( 60 % ) 6 / 41 ( 15 % ) NA NA NA NA NA Local 57 / 215 ( 27 % ) 37 / 93 ( 40 % ) 35 / 41 ( 85 % ) NA NA NA NA NA Data are n / N ( % ) or secondary attack rate ( 95 % CI ) . When household was defined on the basis of close relatives , the overall data - based secondary attack rates were 13\u00b72 % ( 10\u00b79 \u2013 15\u00b77 ) among household contacts and 2\u00b74 % ( 1\u00b76 \u2013 3\u00b73 ) among non - household contacts . When household was defined on the basis of residential address , the overall data - based secondary attack rates were 17\u00b72 % ( 14\u00b71 \u2013 20\u00b76 ) among household contacts and 2\u00b76 % ( 1\u00b79 \u2013 3\u00b76 ) among non - household contacts . Contact type was determined by an individual\u2019s relationship with the primary cases of each close contact group . NA = not applicable . * Calculated as the number of secondary cases divided by the sum of secondary cases and non - cases . \u2020Secondary cases and non - cases in each close contact group were allocated to January or February , 2020 , on the basis of the number of days in the infectious period of the primary case that occurred in January compared with that in February . Table 1 : Demographic composition of the study population stratified by case type and contact type Articles 1146 www . thelancet . com / infection Vol 20 October 2020 secondary attack rates were 12\u22194 % ( 95 % CI 9\u22198 \u2013 15\u22194 % ) among household contacts and 7\u22199 % ( 95 % CI 5\u22193 \u2013 11\u22198 % ) among non - household contacts . A longer incubation period was associated with a slightly lower secondary attack rate estimate than a shorter incubation period , and a longer infectious period was associated with a slightly higher secondary attack rate estimate than a shorter infectious period ( table 2 ; appendix 2 p 21 ) . When different mean incubation periods and maximum infectious periods were considered , the secondary attack rate varied from 11\u22194 % ( 95 % CI 9\u00b70 \u2013 14\u00b72 ) to 18\u00b70 % ( 13\u00b79 \u2013 23\u00b70 ) among household contacts , and from 7\u00b75 % ( 5\u00b70 \u2013 11\u00b72 ) to 12\u00b72 % ( 8\u00b70 \u2013 18\u00b71 ) among non - household contacts ( appendix 2 p 21 ) . The estimated local R based on observed contact freque ncies of primary cases was 0\u22195 ( 95 % CI 0\u221941 \u2013 0\u221962 ) , which was insensitive to the assumed incubation and infectious periods . Thus , a typical case infected 0\u00b75 indi vi duals on average in Figure 1 : Spatial distribution of COVID - 19 case clusters on the basis of contact tracing data from Guangzhou , China , from Jan 7 , 2020 , to Feb 18 , 2020 Overall distribution of COVID - 19 case clusters in Guangzhou ( A ) , and distribution in the subregions defined in panel A ( B \u2013 G ) . Individuals were considered as primary cases if their symptom onset dates were the earliest or 1 day ( \u2264 3 days for an imported case ) after the earliest in the cluster and as secondary cases otherwise . Non - infected contacts are not shown . The displayed location of each case is randomly perturbed away from the actual residential address . D E F G Baiyun B 0\u00b70 1\u00b725 2\u00b75 5\u00b70km Foshan Liwan Haizhu Huangpu Panyu Tianhe Baiyun Foshan Panyu Zengcheng Huizhou Zengcheng Nansha Dongguan Huangpu Huadu 0\u00b70 1\u00b725 2\u00b75 5\u00b70 km 0 1 2 4km 0 0\u00b75 1 2km 0 0\u00b75 1 2km 0 1 2 4km N N N N N N Contact relationship Primary case Household secondary case Non - household secondary case Direction of transmission Coprimary cases Onset date Imported cases with symptom onset on or before Jan 23 , 2020 Imported cases with symptom onset after Jan 23 , 2020 Local cases with symptom onset on or before Jan 23 , 2020 Local cases with symptom onset after Jan 23 , 2020 Boundary Population density 0 114 000 ( km\u00b2 ) A N 0\u00b70 12\u00b75 25\u00b70 50\u00b70km Qingyuan Huizhou Conghua Jiangmen Foshan Zhongshan Shenzhen Dongguan Zengcheng Huadu Huangpu Shaoguan D B E C F Nansha Baiyun Panyu C Articles www . thelancet . com / infection Vol 20 October 2020 1147 Guangzhou , implying inefficient transmission of the disease under the control measures . The projected local R , had there been no isolation of cases or quarantine of their contacts , was 0\u22196 ( 95 % CI 0\u221949 \u2013 0\u221974 ) when household was defined on the basis of close relatives . Higher estimates of projected local R were associated with shorter incubation periods and longer infectious periods ( appendix 2 p 21 ) . When household was defined on the basis of close relatives , the daily transmission probability during the incubation period was similar to that during the illness period ( esti mated OR 0\u221961 [ 95 % CI 0\u221927 \u2013 1\u221938 ] , table 3 ) ; however , the difference was much larger when longer incubation periods were assumed ( table 3 ; appendix 2 p 22 \u2013 23 ) . Individuals aged 60 years and older were the group most susceptible to SARS - CoV - 2 infection ( table 4 ; appendix 2 p 24 ) . Assuming a mean incubation period of 5 days and maximum infectious period of 13 days , in comparison with the oldest age group ( \u226560 years ) , the risk of infection was lower in the youngest age group ( < 20 years ; OR 0\u221923 [ 95 % CI 0\u221911 \u2013 0\u221946 ] ) and the 20 \u2013 59 year age group ( 0\u221964 [ 0\u221943 \u2013 0\u221997 ] ) . The person - to - person transmissibility of the virus declined over time to some extent ( February vs January OR 0\u221942 [ 95 % CI 0\u00b717 \u2013 1\u00b707 ] ) . These estimated ORs were insensitive to the assumptions about the natural history of disease ( appendix 2 p 24 ) . The estimated probability of daily transmission was two times higher in households of six people or less than in larger households ( more than six people ; appendix 2 p 25 ) . We found no association between age and infectivity and no associations between sex and susceptibility or infectivity . Restricting household contacts to those who were living at the same address as the primary case regardless of relationship resulted in higher secondary attack rate estimates among household contacts , ranging from 16\u00b71 % ( 95 % CI 12\u00b75 \u2013 20\u00b74 ) to 24\u00b73 % ( 18\u00b75 \u2013 31\u00b72 ) , but lower secondary attack rate estimates among non - household contacts ( ranging from 6\u00b78 % [ 5\u00b70 \u2013 9\u00b72 ] to 9\u00b73 % [ 6\u00b75 \u2013 13\u00b71 ] ) under the various settings ( ie , mean incubation and maximum infectious periods ) of the natural history of disease ( appendix 2 p 21 ) . Assuming an incubation time of 5 days and an infectious period of 13 days , the estimated secondary attack rates were 17\u22191 % ( 95 % CI 13\u22193 \u2013 21\u22198 ) among household contacts and 7\u22193 % ( 5\u22194 \u2013 9\u22199 ) among non - household contacts ( table 2 ) . The effect of age and the relative infectivity of virus in the illness period versus the incubation period remained similar ( table 4 ; appendix 2 p 24 ) . Discussion We retrospectively characterised the spatiotemporal epidemiology and transmissibility of SARS - CoV - 2 in Guangzhou , the most populated city in southern China , from early January up to mid - February , 2020 . The rapid decline in the R t indicates the effectiveness of the control policy implemented in the city . Social distancing or other potential personal behavioural changes might have also shifted the contact pattern between household members , as shown by the two - fold reduction in the probability of household transmission observed between January and February . Additionally , we assessed the effects of host features and disease stage on susceptibility and Figure 2 : Epidemic curve based on symptom onset dates of COVID - 19 cases in Guangzhou from Jan 6 , 2020 , to Feb 18 , 2020 Estimated R t for three scenarios : scenario 1 , all imported cases ( who travelled to or resided in Hubei province 14 days before symptom onset ) considered as primary cases , and all secondary cases were infected by primary cases in the same case cluster ; scenario 2 , which is identical to scenario 1 , with the additional assumption that local primary cases might have been infected by earlier cases in other clusters ; and scenario 3 , which is identical to scenario 2 , with the additional assumption that imported secondary cases were considered as infected by primary cases in the same cluster . R t = effective reproductive number . Jan 6 , 2020 Jan 13 , 2020 Jan 20 , 2020 Jan 27 , 2020 Feb 3 , 2020 Feb 18 , 2020 Feb 10 , 2020 0 C a s e s ( n ) Date 20 30 40 10 R t 0\u00b75 0 1\u00b70 1\u00b75 2\u00b70 Imported primary cases Imported secondary cases Local primary cases Local secondary cases Scenario 1 Scenario 2 Scenario 3 95 % CI 95 % CI 95 % CI Mean incubation period of 5 days Mean incubation period of 7 days 13 - day infectious period 22 - day infectious period 13 - day infectious period 22 - day infectious period Close relatives Secondary attack rate , % ( 95 % CI ) Household 12\u00b74 % ( 9\u00b78 \u2013 15\u00b74 ) 15\u00b75 % ( 11\u00b77 \u2013 20\u00b72 ) 11\u00b74 % ( 9\u00b70 \u2013 14\u00b72 ) 13\u00b71 % ( 9\u00b79 \u2013 17\u00b71 ) Non - household 7\u00b79 % ( 5\u00b73 \u2013 11\u00b78 ) 10\u00b74 % ( 6\u00b77 \u2013 15\u00b78 ) 7\u00b75 % ( 5\u00b70 \u2013 11\u00b72 ) 8\u00b79 % ( 5\u00b77 \u2013 13\u00b76 ) Local R ( 95 % CI ) With quarantine 0\u00b750 ( 0\u00b741 \u2013 0\u00b762 ) 0\u00b751 ( 0\u00b739 \u2013 0\u00b766 ) 0\u00b751 ( 0\u00b741 \u2013 0\u00b763 ) 0\u00b751 ( 0\u00b739 \u2013 0\u00b767 ) No quarantine 0\u00b760 ( 0\u00b749 \u2013 0\u00b774 ) 0\u00b776 ( 0\u00b759 \u2013 1\u00b700 ) 0\u00b756 ( 0\u00b745 \u2013 0\u00b769 ) 0\u00b765 ( 0\u00b749 \u2013 0\u00b785 ) Residential address Secondary attack rate , % ( 95 % CI ) Household 17\u00b71 % ( 13\u00b73 \u2013 21\u00b78 ) 21\u00b72 % ( 15\u00b78 \u2013 27\u00b78 ) 16\u00b71 % ( 12\u00b75 \u2013 20\u00b74 ) 18\u00b73 % ( 13\u00b76 \u2013 24\u00b71 ) Non - household 7\u00b73 % ( 5\u00b74 \u2013 9\u00b79 ) 9\u00b73 % ( 6\u00b75 \u2013 13\u00b71 ) 6\u00b78 % ( 5\u00b70 \u2013 9\u00b72 ) 7\u00b78 % ( 5\u00b75 \u2013 11\u00b70 ) Local R ( 95 % CI ) With quarantine 0\u00b750 ( 0\u00b740 \u2013 0\u00b761 ) 0\u00b750 ( 0\u00b738 \u2013 0\u00b765 ) 0\u00b750 ( 0\u00b741 \u2013 0\u00b762 ) 0\u00b751 ( 0\u00b739 \u2013 0\u00b766 ) No quarantine 0\u00b759 ( 0\u00b748 \u2013 0\u00b772 ) 0\u00b774 ( 0\u00b757 \u2013 0\u00b796 ) 0\u00b755 ( 0\u00b745 \u2013 0\u00b767 ) 0\u00b763 ( 0\u00b748 \u2013 0\u00b782 ) Estimates were reported using two different definitions of household contact ( close relatives or individuals sharing the same residential address ) and for selected settings of the natural history of disease . This model was not adjusted for age group , epidemic phase , or household size . R = reproductive number . Table 2 : Model - based estimates of secondary attack rates among household and non - household contacts , and local R with and without quarantine Articles 1148 www . thelancet . com / infection Vol 20 October 2020 infectivity . We found that patients with COVID - 19 were at least as infectious in the incubation periods as during their illness periods , and that older people ( aged > 60 years ) are most susceptible to household infection of SARS - CoV - 2 . When household contact was defined by residential address , our model - based secondary attack rate estimate for the Guangzhou contact tracing data was 17\u22191 % , which is higher than data - based secondary attack rate estimates of 14\u22199 % for Shenzhen and 10\u22192 % for Guangzhou under the same household definition . 14 , 15 Generally , a data - based secondary attack rate estimate would be higher than a model - based secondary attack rate estimate since data - based estimates do not exclude tertiary transmission and untraced exposure . However , these data - based secondary attack rate estimates reflect transmissibility under control measures such as case isolation , whereas our model - based secondary attack rate estimates assumed exposure of a susceptible individual during the whole infectious period of an infector , which is more epidemiologically relevant and generalisable . Model - based household secondary attack rate estimates for SARS - CoV or the Middle East respiratory syndrome coronavirus ( MERS - CoV ) are not available ; however , a small number of studies have reported data - based secondary attack rate estimates in the household or comparable settings . For SARS - CoV , the secondary attack rate was estimated to be 4\u22196 \u2013 8\u00b70 % in Beijing , Hong Kong , and Singapore . 21 , 22 The daily transmis - sion probabilities during the illness period , however , are comparable between SARS - CoV ( 0\u00b7013 [ 95 % CI 0\u00b7011 \u2013 0\u00b7016 ] ) and SARS - CoV - 2 ( 0\u00b7016 [ 0\u00b7008 \u2013 0\u00b7029 ] ; appendix 2 p 22 ) . 23 Information about household transmissibility of MERS - CoV is less clear . In a multicity household study in Saudi Arabia , the household secondary attack rate was 4 % ( 95 % CI 2 \u2013 7 ) . 24 In an outbreak among 828 female workers who lived in an Mean incubation period of 5 days Mean incubation period of 7 days 13 - day infectious period 22 - day infectious period 13 - day infectious period 22 - day infectious period Close relatives Transmission probabilities for household contacts ( \u00d710\u207b\u00b2 ) Incubation 1\u00b784 ( 1\u00b736 \u2013 2\u00b749 ) 1\u00b791 ( 1\u00b744 \u2013 2\u00b754 ) 2\u00b709 ( 1\u00b763 \u2013 2\u00b769 ) 2\u00b711 ( 1\u00b766 \u2013 2\u00b768 ) Illness 1\u00b713 ( 0\u00b761 \u2013 2\u00b708 ) 0\u00b780 ( 0\u00b744 \u2013 1\u00b746 ) 0\u00b754 ( 0\u00b719 \u2013 1\u00b757 ) 0\u00b741 ( 0\u00b716 \u2013 1\u00b705 ) Transmission probabilities for non - household contacts ( \u00d710\u207b\u00b2 ) Incubation 1\u00b716 ( 0\u00b773 \u2013 1\u00b783 ) 1\u00b725 ( 0\u00b781 \u2013 1\u00b792 ) 1\u00b737 ( 0\u00b79 \u2013 2\u00b707 ) 1\u00b74 ( 0\u00b793 \u2013 2\u00b71 ) Illness 0\u00b771 ( 0\u00b735 \u2013 1\u00b743 ) 0\u00b752 ( 0\u00b726 \u2013 1\u00b705 ) 0\u00b735 ( 0\u00b711 \u2013 1\u00b709 ) 0\u00b727 ( 0\u00b71 \u2013 0\u00b775 ) Transmission probability from an external source ( \u00d710\u207b\u2074 ) 1\u00b771 ( 0\u00b778 \u2013 3\u00b778 ) 1\u00b749 ( 0\u00b765 \u2013 3\u00b744 ) 1\u00b754 ( 0\u00b761 \u2013 3\u00b786 ) 1\u00b738 ( 0\u00b754 \u2013 3\u00b756 ) OR 0\u00b761 ( 0\u00b727 \u2013 1\u00b738 ) 0\u00b741 ( 0\u00b719 \u2013 0\u00b789 ) 0\u00b726 ( 0\u00b708 \u2013 0\u00b786 ) 0\u00b719 ( 0\u00b707 \u2013 0\u00b755 ) Residential address Transmission probabilities for household contacts ( \u00d710\u207b\u00b2 ) Incubation 2\u00b764 ( 1\u00b79 \u2013 3\u00b766 ) 2\u00b777 ( 2\u00b703 \u2013 3\u00b776 ) 3\u00b703 ( 2\u00b729 \u2013 4\u00b700 ) 3\u00b707 ( 2\u00b735 \u2013 4\u00b702 ) Illness 1\u00b758 ( 0\u00b784 \u2013 2\u00b795 ) 1\u00b71 ( 0\u00b759 \u2013 2\u00b705 ) 0\u00b779 ( 0\u00b728 \u2013 2\u00b721 ) 0\u00b757 ( 0\u00b722 \u2013 1\u00b746 ) Transmission probabilities for non - household contacts ( \u00d710\u207b\u00b2 ) Incubation 1\u00b708 ( 0\u00b775 \u2013 1\u00b755 ) 1\u00b714 ( 0\u00b781 \u2013 1\u00b761 ) 1\u00b723 ( 0\u00b789 \u2013 1\u00b769 ) 1\u00b725 ( 0\u00b792 \u2013 1\u00b770 ) Illness 0\u00b764 ( 0\u00b733 \u2013 1\u00b724 ) 0\u00b745 ( 0\u00b723 \u2013 0\u00b787 ) 0\u00b732 ( 0\u00b711 \u2013 0\u00b791 ) 0\u00b723 ( 0\u00b709 \u2013 0\u00b761 ) Transmission probability from an external source ( \u00d710\u207b\u2074 ) 1\u00b774 ( 0\u00b779 \u2013 3\u00b784 ) 1\u00b754 ( 0\u00b767 \u2013 3\u00b753 ) 1\u00b753 ( 0\u00b76 \u2013 3\u00b787 ) 1\u00b74 ( 0\u00b754 \u2013 3\u00b762 ) OR 0\u00b759 ( 0\u00b726 \u2013 1\u00b735 ) 0\u00b739 ( 0\u00b718 \u2013 0\u00b786 ) 0\u00b726 ( 0\u00b708 \u2013 0\u00b782 ) 0\u00b718 ( 0\u00b706 \u2013 0\u00b752 ) Data are estimates ( 95 % CI ) . Estimates of the daily probability of infection from an external source and the ORs for the relative infectivity during the illness versus incubation period are also provided . Estimates are reported using two different definitions of household contact ( close relatives , or only individuals sharing the same residential address ) and for selected settings of the natural history of disease ( ie , mean incubation and maximum infectious periods ) . This model was not adjusted for age group , epidemic phase , or household size . OR = odds ratio . Table 3 : Model - based estimates of daily transmission probabilities for household contacts and non - household contacts during the incubation and illness periods Mean incubation period of 5 days Mean incubation period of 7 days 13 - day infectious period 22 - day infectious period 13 - day infectious period 22 - day infectious period Close relatives Susceptibility Age < 20 years vs \u226560 years 0\u00b723 ( 0\u00b711 \u2013 0\u00b746 ) 0\u00b722 ( 0\u00b711 \u2013 0\u00b746 ) 0\u00b722 ( 0\u00b711 \u2013 0\u00b745 ) 0\u00b722 ( 0\u00b711 \u2013 0\u00b745 ) Age 20 \u2013 59 years vs \u226560 years 0\u00b764 ( 0\u00b743 \u2013 0\u00b797 ) 0\u00b764 ( 0\u00b742 \u2013 0\u00b796 ) 0\u00b763 ( 0\u00b742 \u2013 0\u00b795 ) 0\u00b763 ( 0\u00b742 \u2013 0\u00b794 ) February vs January 0\u00b742 ( 0\u00b717 \u2013 1\u00b707 ) 0\u00b746 ( 0\u00b719 \u2013 1\u00b710 ) 0\u00b736 ( 0\u00b712 \u2013 1\u00b705 ) 0\u00b738 ( 0\u00b713 \u2013 1\u00b709 ) Infectivity Illness vs incubation period 0\u00b760 ( 0\u00b727 \u2013 1\u00b736 ) 0\u00b742 ( 0\u00b719 \u2013 0\u00b791 ) 0\u00b729 ( 0\u00b710 \u2013 0\u00b788 ) 0\u00b721 ( 0\u00b707 \u2013 0\u00b758 ) Residential address Susceptibility Age < 20 years vs \u226560 years 0\u00b722 ( 0\u00b711 \u2013 0\u00b746 ) 0\u00b722 ( 0\u00b711 \u2013 0\u00b745 ) 0\u00b722 ( 0\u00b711 \u2013 0\u00b744 ) 0\u00b722 ( 0\u00b711 \u2013 0\u00b744 ) Age 20 \u2013 59 years vs \u226560 years 0\u00b767 ( 0\u00b745 \u2013 1\u00b700 ) 0\u00b767 ( 0\u00b745 \u2013 1\u00b700 ) 0\u00b766 ( 0\u00b744 \u2013 0\u00b799 ) 0\u00b766 ( 0\u00b744 \u2013 0\u00b799 ) February vs January 0\u00b757 ( 0\u00b723 \u2013 1\u00b739 ) 0\u00b762 ( 0\u00b727 \u2013 1\u00b744 ) 0\u00b750 ( 0\u00b718 \u2013 1\u00b736 ) 0\u00b753 ( 0\u00b720 \u2013 1\u00b742 ) Infectivity Illness vs incubation period 0\u00b754 ( 0\u00b723 \u2013 1\u00b726 ) 0\u00b738 ( 0\u00b717 \u2013 0\u00b784 ) 0\u00b724 ( 0\u00b707 \u2013 0\u00b779 ) 0\u00b718 ( 0\u00b706 \u2013 0\u00b752 ) Data are OR ( 95 % CI ) . Estimates were reported for selected settings of the natural history of disease ( ie , mean incubation and maximum infectious periods ) . This model was adjusted for age group , epidemic phase , and household size . OR = odds ratio . Table 4 : Model - based effects of age group and epidemic phase on susceptibility and relative infectivity during the illness period compared with the incubation period Articles www . thelancet . com / infection Vol 20 October 2020 1149 expatriate dormitory consisting of 24 villas in Riyadh , Saudi Arabia , 19 workers in seven villas were infected . If each affected villa was considered a household with a mean of 34\u00b75 residents including a single primary case , we estimated a secondary attack rate of 5\u00b71 % ( 95 % CI 2\u00b78 \u2013 9\u00b70 ) . 25 We conclude that SARS - CoV - 2 is more transmissible than SARS - CoV and MERS - CoV in households . We quantified the infectivity of patients with COVID - 19 during their incubation period using household data , but epidemiological evidence has been published previously . Transmission to secondary cases during the incubation period of the primary case has been reported in Germany and China . 7 , 26 An analysis of 77 transmission pairs within and outside of China estimated that nearly half of all transmission events could have occurred during the incubation period . 5 By contrast , shedding of SARS - CoV peaks 6 \u2013 11 days after illness onset , and asymptomatic and mild MERS - CoV cases were hypothesised to transmit inefficiently , indicating the importance of symptoms in the transmission of these two coronaviruses . 27 \u2013 29 This finding indicates the importance of testing close contacts of COVID - 19 cases to identify and isolate infections in the incubation period . Alternatively , the lower estimated transmission probability of SARS - CoV - 2 during the illness period than during the incubation period could be partially attributed to self - distancing within house holds when the primary cases developed symptoms . The infectivity measured by the transmission model accounts for both the biological process of viral shedding and the social contact process , and our data cannot separate the two processes . We estimated that the local R of SARS - CoV - 2 was relatively low ( around 0\u22195 ) , which is consistent with the mean R t . We estimated that without isolation of cases or quarantine of their contacts and assuming a mean incubation period of 5 days , the local reproductive number would have been about 20 \u2013 50 % higher , increasing to 0\u22196 \u2013 0\u00b776 ( appendix 2 p 21 ) . The subcritical local reproductive number ( R < 1 ) in the absence of isolation is due to the small average number of contacts per person per day ( appendix 2 , p 18 ) , which is likely to be a result of the stringent control measures that were implemented in the city . Although the effect of case isolation seems moderate , considering the high infectivity of the virus during the incubation period , quarantine of asymptomatic contacts could have prevented more onward transmissions . Our analysis has several limitations . The model suggests that a longer assumed incubation period was associated with higher estimated infectivity in this period , which could be due to the rapid quarantine of close contacts after symptom onset of cases . When few transmissions occurred during the illness period , bias could occur due to the paucity of data . Additionally , we were unable to reliably quantify the infectivity of asymptomatic infections , since only two of the 15 asymptomatic infections included in the household transmission analyses were considered primary cases . Some asymptomatic infections might have been missed since close contacts were tested only twice and the tests were done 14 days apart . Moreover , our assumption that asymptomatic infections have the same infectivity as symptomatic cases during their incubation period might not be realistic . Furthermore , it is likely that some imported primary cases might have been infected locally and that some asymptomatic infections or cases might have been missed by contact tracing or by false negative tests , which could lead to underestimation of the R t and the secondary attack rate . The infectiousness of patients with COVID - 19 during their incubation periods is high and could substantially increase the difficulty of curbing the ongoing pandemic . Active case finding and isolation in conjunction with comprehensive contact tracing and quarantine are useful for preventing infected contacts from spreading the virus during their incubation periods , which could be crucial when societal restrictions on human movement and mixing are lifted . The provision of comfortable facilities for exposed contacts to quarantine or for mild cases to isolate away from their families could be a valuable strategy to limit onward transmission within households . Contributors QLJ , JY , ZBZ , LL , MMM , YL , YM , and ZCY collected the data . YY , ZCY , and LQF conceived the statistical analysis plan . QLJ , MJL , ARZ , and NJ cleaned the data . MJL , QLJ , ARZ , and NJ did statistical analyses under the supervision of YY , ZCY , and LQF . YY drafted the manuscript . ND , IL , and EK contributed to interpretation of results and writing of the manuscript . Declaration of interests We declare no competing interests . Data sharing Data sharing requests should be directed to Guangzhou Center for Disease Control and Prevention . Acknowledgments This study was supported by grants from the US National Institutes of Health ( R01 AI139761 , R01 AI116770 and R37 AI32042 ) , the Science and Technology Plan Project of Guangzhou ( 201804010121 ) , the Project for Key Medicine Discipline Construction of Guangzhou Municipality ( 2017 \u2013 2019 \u2013 04 ) , and the Key Research and Development Program of China ( 2019YFC1200604 ) . We thank M Elizabeth Halloran ( Fred Hutchinson Cancer Research Center , Seattle , WA , USA ) for helpful discussions . We also thank the staff members of all district - level Center for Disease Control and Prevention and community health service centres in Guangzhou for their assistance in field investigation and data collection . References 1 Dong E , Du H , Gardner L . An interactive web - based dashboard to track COVID - 19 in real time . Lancet Infect Dis 2020 ; 20 : 533 \u2013 34 . 2 Wu Z , McGoogan JM . Characteristics of and important lessons from the coronavirus disease 2019 ( COVID - 19 ) outbreak in China : summary of a report of 72 314 cases from the Chinese Center for Disease Control and Prevention . JAMA 2020 ; 323 : 1239 . 3 van Doremalen N , Bushmaker T , Morris DH , et al . Aerosol and surface stability of SARS - CoV - 2 as compared with SARS - CoV - 1 . N Engl J Med 2020 ; 382 : 1564 \u2013 67 . 4 Xu Y , Li X , Zhu B , et al . Characteristics of pediatric SARS - CoV - 2 infection and potential evidence for persistent fecal viral shedding . Nat Med 2020 ; 26 : 502 \u2013 05 . Articles 1150 www . thelancet . com / infection Vol 20 October 2020 5 He X , Lau EHY , Wu P , et al . Temporal dynamics in viral shedding and transmissibility of COVID - 19 . Nat Med 2020 ; 26 : 672 \u2013 75 . 6 Bai Y , Yao L , Wei T , et al . Presumed asymptomatic carrier transmission of COVID - 19 . JAMA 2020 ; 323 : 1406 . 7 Rothe C , Schunk M , Sothmann P , et al . Transmission of 2019 - nCoV Infection from an asymptomatic contact in Germany . N Engl J Med 2020 ; 382 : 970 \u2013 71 . 8 Liu J , Liao X , Qian S , et al . Community transmission of severe acute respiratory syndrome coronavirus 2 , Shenzhen , China , 2020 . Emerg Infect Dis 2020 ; 26 : 1320 \u2013 23 . 9 She J , Jiang J , Ye L , Hu L , Bai C , Song Y . 2019 novel coronavirus of pneumonia in Wuhan , China : emerging attack and management strategies . Clin Transl Med 2020 ; 9 : 19 . 10 Wu JT , Leung K , Leung GM . Nowcasting and forecasting the potential domestic and international spread of the 2019 - nCoV outbreak originating in Wuhan , China : a modelling study . Lancet 2020 ; 395 : 689 \u2013 97 . 11 Chinazzi M , Davis JT , Ajelli M , et al . The effect of travel restrictions on the spread of the 2019 novel coronavirus ( COVID - 19 ) outbreak . Science 2020 ; 368 : 395 \u2013 400 . 12 Tian H , Liu Y , Li Y , et al . An investigation of transmission control measures during the first 50 days of the COVID - 19 epidemic in China . Science 2020 ; 368 : 638 \u2013 42 . 13 Liu Y , Eggo RM , Kucharski AJ . Secondary attack rate and superspreading events for SARS - CoV - 2 . Lancet 2020 ; 395 : e47 . 14 Bi Q , Wu Y , Mei S , et al . Epidemiology and transmission of COVID - 19 in 391 cases and 1286 of their close contacts in Shenzhen , China : a retrospective cohort study . Lancet Infect Dis 2020 ; published online April 27 . DOI : 10 . 1016 / S1473 - 3099 ( 20 ) 30287 - 5 . 15 Luo L , Liu D , Liao X - L , et al . Modes of contact and risk of transmission in COVID - 19 among close contacts . medRxiv 2020 ; published online March 26 . DOI : 10 . 1101 / 2020 . 03 . 24 . 20042606 ( preprint ) . 16 Burke RM , Midgley CM , Dratch A , et al . Active monitoring of persons exposed to patients with confirmed COVID - 19\u2014United States , January \u2013 February 2020 . MMWR Morb Mortal Wkly Rep 2020 ; 69 : 245 \u2013 46 . 17 Yang Y , Longini IM Jr , Halloran ME , Obenchain V . A hybrid EM and Monte Carlo EM algorithm and its application to analysis of transmission of infectious diseases . Biometrics 2012 ; 68 : 1238 \u2013 49 . 18 Li Q , Guan X , Wu P , et al . Early transmission dynamics in Wuhan , China , of novel coronavirus - infected pneumonia . N Engl J Med 2020 ; 382 : 1199 \u2013 207 . 19 Lauer SA , Grantz KH , Bi Q , et al . The incubation period of coronavirus disease 2019 ( COVID - 19 ) from publicly reported confirmed cases : estimation and application . Ann Intern Med 2020 ; 172 : 577 \u2013 82 . 20 W\u00f6lfel R , Corman VM , Guggemos W , et al . Virological assessment of hospitalized patients with COVID - 2019 . Nature 2020 ; 581 : 465 \u2013 69 . 21 Goh DL , Lee BW , Chia KS , et al . Secondary household transmission of SARS , Singapore . Emerg Infect Dis 2004 ; 10 : 232 \u2013 34 . 22 Lau JTF , Lau M , Kim JH , Tsui HY , Tsang T , Wong TW . Probable secondary infections in households of SARS patients in Hong Kong . Emerg Infect Dis 2004 ; 10 : 235 \u2013 43 . 23 Pitzer VE , Leung GM , Lipsitch M . Estimating variability in the transmission of severe acute respiratory syndrome to household contacts in Hong Kong , China . Am J Epidemiol 2007 ; 166 : 355 \u2013 63 . 24 Drosten C , Meyer B , M\u00fcller MA , et al . Transmission of MERS - coronavirus in household contacts . N Engl J Med 2014 ; 371 : 828 \u2013 35 . 25 Van Kerkhove MD , Alaswad S , Assiri A , et al . Transmissibility of MERS - CoV infection in closed setting , Riyadh , Saudi Arabia , 2015 . Emerg Infect Dis 2019 ; 25 : 1802 \u2013 09 . 26 Cai J , Sun W , Huang J , Gamber M , Wu J , He G . Indirect virus transmission in cluster of COVID - 19 cases , Wenzhou , China , 2020 . Emerg Infect Dis 2020 ; 26 : 1343 \u2013 45 . 27 Cheng PK , Wong DA , Tong LK , et al . Viral shedding patterns of coronavirus in patients with probable severe acute respiratory syndrome . Lancet 2004 ; 363 : 1699 \u2013 700 . 28 Memish ZA , Assiri AM , Al - Tawfiq JA . Middle East respiratory syndrome coronavirus ( MERS - CoV ) viral shedding in the respiratory tract : an observational analysis with infection control implications . Int J Infect Dis 2014 ; 29 : 307 \u2013 08 . 29 Al Hosani FI , Kim L , Khudhair A , et al . Serologic follow - up of Middle East Respiratory Syndrome Coronavirus cases and contacts \u2013 Abu Dhabi , United Arab Emirates . Clin Infect Dis 2019 ; 68 : 409 \u2013 18 .",
    "hopeScalingCreativeInspiration2022": "Scaling Creative Inspiration with Fine - Grained Functional Aspects of Ideas Tom Hope tomh @ allenai . org Allen Institute for AI The University of Washington Ronen Tamari Hebrew University of Jerusalem Hyeonsu Kang Carnegie Mellon University Daniel Hershcovich University of Copenhagen , Denmark Joel Chan University of Maryland Aniket Kittur Carnegie Mellon University Dafna Shahaf dshahaf @ cs . huji . ac . il Hebrew University of Jerusalem ABSTRACT Large repositories of products , patents and scientific papers offer an opportunity for building systems that scour millions of ideas and help users discover inspirations . However , idea descriptions are typi - cally in the form of unstructured text , lacking key structure that is required for supporting creative innovation interactions . Prior work has explored idea representations that were either limited in expres - sivity , required significant manual effort from users , or dependent on curated knowledge bases with poor coverage . We explore a novel rep - resentation that automatically breaks up products into fine - grained functional aspects capturing the purposes and mechanisms of ideas , and use it to support important creative innovation interactions : func - tional search for ideas , and exploration of the design space around a focal problem by viewing related problem perspectives pooled from across many products . In user studies , our approach boosts the qual - ity of creative search and inspirations , substantially outperforming strong baselines by 50 - 60 % . 1 INTRODUCTION Human creativity often relies on detecting structural matches across distant ideas and adapting them by transferring mechanisms from one domain to another [ 16 , 17 , 37 , 38 ] . For example , microwave ovens were discovered by repurposing radar technology developed during World War II . Teflon , today chiefly used in non - stick cook - ware , was first used in armament development . Recognizing the potential of this innovation process , major organizations such as NASA and Procter & Gamble actively engage in searching for op - portunities to adapt existing technologies for new markets [ 27 ] . Online repositories of millions of products , scientific papers , and patents present an opportunity to augment and scale this core process of innovation . The large scale and diversity of these repositories is Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page . Copyrights for components of this work owned by others than ACM must be honored . Abstracting with credit is permitted . To copy otherwise , or republish , to post on servers or to redistribute to lists , requires prior specific permission and / or a fee . Request permissions from permissions @ acm . org . CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA \u00a9 2022 Association for Computing Machinery . important , because inspiration can be found in unexpected places \u2013 for example , a car mechanic recently invented a simple device to ease childbirths by adapting a trick for extracting a cork stuck in a wine bottle , which he discovered in a YouTube video [ 1 ] . However , the predominant way human problem - solvers currently interact with these repositories \u2014 via standard search engines \u2014 does not tap into their potential for augmenting and scaling human ingenuity . Core to this limitation is the representation of ideas in the form of unstructured textual descriptions . This representation hinders creative innovation interactions that require traversing multiple levels of granularity and abstraction around a focal problem , to \u201cbreak out\u201d of fixation on the details of a specific problem by exploring the design space and viewing novel perspectives on problems and solutions [ 15 , 23 , 60 , 65 ] . Toward addressing this challenge , our vision in this paper is to develop a novel representation of ideas that can support exploration and abstraction of fine - grained functional aspects in large - scale idea repositories \u2013 aspects such as the purposes and mechanisms of products . More specifically , our goal is to obtain a representation having two key capabilities : ( I ) The representation would be able to automatically disentangle raw descriptions into fine - grained func - tional \u201cunits\u201d that support search and discovery of products that match on certain functions but not others . ( II ) This representation should also allow navigating the landscape of ideas at different resolutions \u2014 enabling users to \u201czoom\u201d in and out at desired levels of abstraction of a given problem and connect to inspirations in seemingly distant areas . As an example , consider an inventor looking for a way to wash clothes without water ( e . g . , in space , or where water is scarce ) . Fig - ure 1 illustrates our vision . Breaking down product descriptions into fine - grained functions ( capability I above ) could allow an automated system to find ideas that match on certain purpose aspects ( washing clothes ) but not certain mechanism aspects ( usage of water ) . This could lead to solutions like cleaning mechanisms based on dry ice or chemical coating . Zooming out and abstracting the problem to a more general fram - ing ( capability II ) might lead to broader ideas for the problem of cleaning such as techniques for removing dirt or odor \u2013 each result - ing in novel problem perspectives and inspirations . In Figure 1 , each node represents a cluster of documents with a similar purpose and a r X i v : 2102 . 09761v3 [ c s . H C ] 17 F e b 2022 CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Hope et al . the user can explore neighboring clusters to find inspirations ( e . g . , dry shampoo ) . This can also expand the innovator\u2019s conception of the problem space itself , such as the assumption that clothes should be cleaned and reworn ( vs . biodegradable ) . In this paper , we develop a scalable computational model of ideas that brings us closer to this vision . We train a neural network to extract spans of text describing purposes and mechanisms in product texts , and use them to build a span embedding representation that allows aspect - based matching between ideas . We then use this representation of individual ideas to automatically mine connections between problems and solutions across entire repositories and build a \u201cfunctional network\u201d that resembles functional ontologies used in engineering and design ideation [ 39 , 51 ] , which are typically hand - crafted and limited in scale . Our approach could facilitate many applications in creative inno - vation due to its ability to decompose idea texts into fine - grained functional aspects , and to surface related problem perspectives at multiple levels of abstraction \u2014 two fundamental drivers of creativ - ity support . In this paper we instantiate the approach in two prototype systems , probing its value regarding each of these capabilities : \u2022 Functional aspect - based search for alternative uses . One important use of our novel representation is to enhance the expressivity of search engines over idea repositories . This way , our representation could support expressive search for alternative , atypical uses of products to identify potentially high - value adaptation opportunities . \u2022 Exploring alternative levels of problem perspectives . Re - cent work [ 42 ] showed that problem - solvers are often inter - ested in exploring different reformulations of the problem when searching for inspiration . Our fine - grained span repre - sentations facilitate mining of recurring functional relations , such as purposes that are often mentioned together or mecha - nisms associated with purposes . This level of detail enables us to map the landscape of ideas with a network of purposes and mechanisms , allowing us to automatically traverse neigh - boring problems and solutions around a focal problem and surface novel inspirations to users . Previous work highlighted the importance of functional represen - tations for supporting ideation [ 12 , 42 , 52 , 60 , 65 , 100 ] , but these methods require significant manual effort from the user , rely on resources with limited coverage , or have limited expressivity ( we discuss this work in more detail in \u00a7 2 ) . We seek to advance the state of the art by developing a novel representation that is both ex - pressive and scalable , and exploring the applications it unlocks . We believe our representation may serve as a useful building block for novel creativity support tools that can help users find and recombine the inspirations latent in unstructured idea repositories at a scale previously impossible . In summary , in this work we contribute : \u2022 A novel computational representation of ideas with fine - grained functional aspects for purposes and mechanisms . \u2022 Empirical demonstrations of the flexibility and utility of the representation for computational support of core creative tasks : ( 1 ) searching for alternative , atypical product uses for potential adaptation opportunities ; and ( 2 ) creating a func - tional concept graph that enables innovators to explore the design space around a focal problem . Through two empirical user studies we demonstrate that our representation signifi - cantly outperforms both previous work and state - of - the art embedding baselines on these tasks . We achieve Mean Aver - age Precision ( MAP ) of 87 % in the alternative product uses search , and 62 % of our inspirations for design space explo - ration are found to be useful and novel \u2013 a relative boost of 50 - 60 % over the best - performing baselines . 2 RELATED WORK : IDEATION SYSTEMS AND COGNITIVE MECHANISMS The contributions in this paper relate broadly to previous work on systems that use structured representations for supporting ideation , and studies that seek to understand the cognitive process of creativity and its implications . We provide a brief discussion of these two themes . Cognitive Theories of Creative Thinking with Prior Ideas . A core aspect of creative thinking that distinguishes it from regular problem solving is the need for divergent thinking [ 21 , 85 , 87 ] - to construct and explore diverse ideas that are quite different from the obvious path of ideation . This process of divergence is shaped by prior knowledge [ 97 ] \u2014 such as past ideas , external stimulation , and examples \u2014 in ways that sometimes hinders creativity through mechanisms like fixation [ 54 ] , and sometimes leads to creative break - throughs [ 92 ] . Our design goals for this research are guided by past research on cognitive mechanisms that enable helpful interactions with past ideas . For example , research on insight problem - solving has uncov - ered the role of re - representation of past ideas by decomposing them into conceptual chunks and then recombining and / or repurposing them into new solutions [ 61 , 73 ] . Similar patterns in terms of core helpful interactions with prior ideas have been observed in in situ studies of expert designers\u2019 dissection of past ideas into component aspects and features for repurposing and recombination into new ideas [ 26 , 48 , 49 ] . Another core process is analogical abstraction , where innovators think about and retrieve past ideas not in terms of their surface features , but in terms of deeper structural features or schemata , such as their underlying purposes ( goals ) and mechanisms [ 40 , 41 ] . These abstracted \u201cschemas \" can then facilitate analogical transfer of ideas across domains that can lead to groundbreaking discoveries [ 19 , 50 , 60 , 72 , 79 ] . For example , the ancient Greeks studied the properties of sound waves by analogy to ripples in water ; Nobel laureate Salvador Luria used abstract structural similarity between a slot machine and bacteria mutations to understand bac - terial replication [ 77 ] ; and in computer science and optimization , analogies to processes in nature inspired algorithms such as simu - lated annealing [ 59 ] , genetic algorithms [ 35 ] , and momentum - based gradient descent [ 83 ] . This process of abstraction over past ideas is also an important contributor to the ability to reformulate problems [ 20 , 23 , 24 , 31 , 57 ] . For example , innovators tasked with a problem ( find more room to store e - waste ) might consider a related , more general goal ( reducing environmental pollution ) to inspire new solu - tions , or consider \u201csibling\u201d formulations ( create alternative materials that are biodegradable ) . This ability to reframe a problem using other problems that bear some abstract relation to it is known to be a pow - erful way to combat fixation and boost creativity [ 16 , 31 , 36 \u2013 38 ] . Scaling Creative Inspiration with Fine - Grained Functional Aspects of Ideas CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Purpose : wash clothes Mechanism : NOT water Waterless washing machine scrubs clothes clean with dry ice . Dry shampoo uses starch - based active ingredients to soak up the oils and sweat from your hair , making it appear cleaner . Fragrance makes your hair smell fresh . A chemical coating causing cotton materials to clean themselves of stains and remove odors when exposed to sunlight . Biodegradable yarn from cellulose fiber from citrus fruit peels . Can be spun into high - quality fabric that feels like silk . clean , remove dirt wash clothes , clean clothes remove bad odor get rid of dirty objects 1 2 3 4 1 1 , 2 2 , 3 4 Example Documents from Clusters Exploring the Purpose Graph Figure 1 : Extracting fine - grained purposes and mechanisms at scale enables mapping the landscape of ideas . Suppose an inventor is looking for a way to wash clothes without water . On the left we see a snippet from the graph of purposes . Each node in the graph represents a cluster of similar purpose spans extracted from documents ( labels are manually generated for illustration purposes ) . Edges reflect abstract structural similarity , capturing co - occurrence patterns of spans in the corpus ( see Section 5 . 1 and Figure 9 ) . On the right we see example documents containing purposes from the four clusters ( corresponding cluster numbers appear in boxes ) . Purpose / mechanism spans in documents are shown in pink / green , respectively . One could find direct matches to the query \u2013 i . e . , documents with purpose from cluster 1 and mechanism not \u201cwater\u201d ( e . g . , waterless washing machine using dry ice ) , or explore neighboring purpose clusters , reformulating the problem as removing odor , removing dirt , or getting rid of the dirty clothes , each resulting in a different set of inspirations . This abstraction and reframing process has also been observed in studies of example curation [ 49 , 56 ] . A core unifying thread across these mechanisms is the need for particular structured , yet flexible , representations of ideas in terms of their component aspects , such as analogical schemas , or conceptual chunks . In this paper , we focus on developing representations with these properties , starting with the decomposition of ideas into their component purposes and mechanisms . As we discuss in the next section , developing computational representations with these proper - ties that can operate over large scale repositories of ideas remains a formidable open challenge . Utilizing Structured Representations for Ideation Systems . A main focus of creativity techniques and prototypes has been building computational systems for exploring the space of possible solutions to problems and alternative problem perspectives [ 31 ] . To do so , such systems often leverage structured knowledge representations for mapping the design space and linking across different prob - lems . For example , the WordTree method [ 65 ] \u2013 a prominent design method in creative engineering design \u2013 directs designers to break their problem into subfunctions , and then use the WordNet database [ 75 ] to explore abstractions and related functional aspects . Like - wise , a recent study [ 42 ] asked designers to select product aspects to abstract using WordNet and the Cyc ontology [ 63 ] , which aimed to serve as a general - purpose repository of commonsense knowl - edge in structured form . These and other general - purpose knowledge bases ( e . g . , NELL [ 76 ] and DBpedia [ 29 ] ) largely encode categorical knowledge ( e . g . , is - a , has - a ) and rarely functional knowledge ( e . g . , used - for ) , and often suffer from poor coverage of concepts in real - world products [ 42 ] . Knowledge bases and ontologies that do focus on functions , behaviors , and structures [ 6 , 51 , 96 ] have primarily been hand - crafted and are therefore even more limited in coverage . Work attempting to scale up has shown promise but is limited in expressivity or interpretability , such as modeling patents in terms of verbs and nouns [ 32 ] , using principal component analysis [ 25 ] , or learning coarse aggregate vectors that capture only one overall product purpose and mechanism [ 52 ] that cannot disentangle the dif - ferent aspects of a product , unlike our work presented in this paper . While full abstraction of ideas currently remains a holy grail , here we investigate whether learning nuanced functional aspects might enable a limited form of abstraction useful for augmenting creativity and providing a first step towards true automated abstraction . 3 LEARNING A FINE - GRAINED FUNCTIONAL REPRESENTATION Our goal in this section is to construct a representation that can support the creative innovation tasks and interactions discussed in the Introduction . CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Hope et al . Figure 2 : Crowdsourcing interface for fine - grained purposes and mechanisms . We propose to use span representations [ 62 ] . Given a product text description , we extract tagged spans of text corresponding to purposes and mechanisms ( see Figure 2 ) , and represent the product as a set of span embeddings . More technically , we use a standard sequence tagging formulation , with X \ud835\udc41 = { x 1 , x 2 , . . . , x \ud835\udc41 } a training set of \ud835\udc41 texts , each a sequence of tokens x \ud835\udc56 = ( \ud835\udc65 1 \ud835\udc56 , \ud835\udc65 2 \ud835\udc56 , . . . , \ud835\udc65 \ud835\udc47\ud835\udc56 ) , and Y \ud835\udc41 a corresponding set of label sequences , Y \ud835\udc41 = { y 1 , y 2 , . . . , y \ud835\udc41 } , y \ud835\udc56 = { \ud835\udc66 1 \ud835\udc56 , \ud835\udc66 2 \ud835\udc56 , . . . , \ud835\udc66 \ud835\udc47\ud835\udc56 } , where each \ud835\udc66 \ud835\udc57 indicates token \ud835\udc57 \u2019s label ( purpose / mechanism / other ) . In later sections , we represent each product \ud835\udc56 as a set of purpose span embedding vectors and a set of mechanism span embedding vectors . 3 . 1 Data We use real - world product idea descriptions taken from crowd - sourced innovation website Quirky . com and used in [ 52 ] , including 8500 user - generated texts describing inventions across diverse do - mains ( e . g . , kitchen products , health and fitness , clean energy ) . Texts typically include multiple purposes and mechanisms . Texts in Quirky use very nonstandard language , including grammatical and spelling errors ( e . g . , \u201cFolds Up Perfect For Carrying . you can walk - on , put your mouth on and or hands on . numbers in any configuration 4 learning to De / Composing Numbers . \u201d ) . Annotation . To create a dataset annotated with purposes and mecha - nisms , we collect crowdsourced annotations on Amazon Mechanical Turk ( AMT ) . In the similar annotation task of [ 52 ] workers were reported to annotate long , often irrelevant spans . We thus guided workers to focus on shorter spans . To further improve quality and encourage more fine - grained annotations , we limited maximal span length that could be annotated , and disabled the annotation of stop - words . Fig . 2 shows our tagging interface ; rectangles are taggable chunks . For quality control , we required US - based workers with approval rate over 95 % and at least 1000 approved tasks , and filtered unreasonably fast users . In total , we had 400 annotating workers . Workers were paid $ 0 . 1 per task . This rate was computed aiming for an hourly rate of $ 7 , where completion time was estimated via a small - scale pilot study . However , in the full study we were sur - prised to find the median completion time was much higher , reaching 100 seconds . We note that this figure could be skewed ( e . g . , due to workers queuing of tasks or the ability to take breaks ) . While a manual inspection of the annotations revealed they are mostly satisfactory , we observe two main issues : First , there are of - ten multiple correct annotations . Second , workers provide partial tagging \u2013 in particular , if similar spans appear in different sentences , very few workers bother tagging more than one instance ( despite instructions ) . These issues would have made computing evaluation metrics problematic . We thus decided to use the crowdsourced an - notations as a bronze - standard for training and development sets only . For a reliable evaluation , we collected gold - standard test sets annotated by two CS graduate students . Annotators were instructed to mark all the relevant chunks , resulting in high inter - annotator agreement of 0 . 71 . We collect 22316 annotated training sentences and 512 gold sentences , for a total of 238 , 399 \ud835\udc61\ud835\udc5c\ud835\udc58\ud835\udc52\ud835\udc5b\ud835\udc60 ( tag proportions : 14 . 5 % mechanism , 15 . 9 % purpose , 69 . 6 % other ) . A note on related annotated data . There has been recent work on the related topic of information extraction from scientific papers by classifying sentences , citations , or phrases . Recent supervised approaches [ 9 , 55 , 66 ] use annotations which are often provided by either paper authors themselves , NLP experts , domain experts , or involve elaborate ( multi - round ) annotation protocols . Sequence tagging models are often trained and evaluated on ( relatively ) clean , succinct sentences [ 71 , 101 ] . When trained on noisy texts , results typically suffer drastically [ 2 ] . Our corpus of product descriptions is significantly noisier than scientific papers , and our training annota - tions were collected in a scalable , low - cost manner by non - experts . Using noisy crowdsourced annotation for training and development only is consistent with our quest for a lightweight annotation ap - proach that would still enable training useful models . 3 . 2 Extracting Spans After collecting annotations , we can now train models to extract the spans . We explore several models likely to have sufficient power to learn our proposed novel representation , with the goal of selecting the best performing one . In particular , we chose two approaches that are common for related sequence - tagging problems , such as named entity recognition ( NER ) and part - of - speech ( POS ) tagging : a common baseline and a recent state - of - the - art model . We also tried a model - enrichment approach with syntactic relational inputs . We briefly describe the models we used below , with full technical descriptions and implementation details , data and code appearing in the supplementary material ( Appendix A . 1 ) . We note that our goal in this section is to find a reasonable model whose output could support creative downstream tasks ; many other architectures are possible and could be considered in future work . \u2022 BiLSTM - CRF . A BiLSTM - CRF [ 53 ] neural network , a common baseline approach for NER tasks , enriched with semantic and syn - tactic input embeddings known to often boost performance [ 101 ] . We adopt the \u201cmulti - channel\u201d strategy as in [ 101 ] , concatenating input word embeddings ( pretrained GloVe vectors [ 81 ] ) with part - of - speech ( POS ) and NER embeddings . A conditional random field ( CRF ) model over the BiLSTM outputs maximizes the tag sequence log likelihood under a pairwise transition model between adjacent tags [ 5 ] . \u2022 Pre - trained Language Model ( Pooled Flair ) . A pre - trained lan - guage model [ 4 ] based on contextualized string embeddings , re - cently shown to outperform other powerful models such as BERT [ 22 ] in NER and POS tagging tasks and achieve state - of - art results . \u2022 GCN . We also explore a model - enrichment approach with syn - tactic relational inputs . We employ a graph convolutional network ( GCN ) [ 58 ] over dependency - parse edges [ 101 ] . GCNs are known Scaling Creative Inspiration with Fine - Grained Functional Aspects of Ideas CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA 0 10 20 30 40 50 60 70 80 90100 Top K ( % ) 40 45 50 55 60 65 70 75 P r e c i s i on mechanism purpose 40 45 50 55 60 65 70 75 Configuration P R F 1 Enriched BiLSTM 45 . 24 39 . 01 41 . 90 Pooled - Flair 53 . 30 39 . 80 45 . 50 GCN 47 . 85 47 . 93 47 . 89 GCN self - train 49 . 00 52 . 00 50 . 50 Figure 3 : Left : Precision @ K results for the best performing model ( GCN + self - training ) . Right : Raw extraction accuracy evaluation . All approaches use CRF loss . GCN with syntactic edges outperforms baselines . Self - training further improves re - sults . Random - label achieves only 16 . 01 F 1 . to be useful for propagating relational information and utilizing syntactic cues [ 71 , 101 ] . The linguistic cues are of special relevance and interest to us , as they are known to exist for purpose / mechanism mentions in texts [ 32 ] . 3 . 3 Evaluation of Extraction Accuracy In this section we assess extraction accuracy ( whether we are able to extract purpose and mechanism spans of text ) . In the next sections , we evaluate the utility of the extracted spans for enabling creative innovation tasks . To evaluate raw accuracy of the model\u2019s predictions , we use the standard IOB label markup to encode the purpose and mechanism spans ( 5 possible labels per token , { Beginning , Inside } x { Purpose , Mechanism } plus an \" Outside \" label ) . We conduct experiments using a train / development / test split of 18702 / 3614 / 512 . Due to our challenging setting , we train models on bronze - standard annotations with noisy and partial tagging done by non - experts ; for evaluation we use a curated gold - standard test set ( Section 3 ) . See Figure 3 ( right ) for results : GCN reaches an F 1 score of \u223c 48 % , out - performing the BiLSTM - CRF model ( enriched with multi - channel GloVe , POS , NER and dependency relation embeddings ) by 6 % . GCN also surpasses the strong Pooled - Flair pre - trained language model by nearly 2 . 5 % . A random baseline guessing each token by label frequencies ( Section 3 ) achieves 16 . 01 F 1 . We interpret these results as possibly attesting to the utility of graph representations and features capturing syntactic and semantic information when labels are noisy . As a sanity check , we also computed precision @ K ( Figure 3 , left ) . As expected , precision is higher with low values of \ud835\udc3e , and gradually degrades . Precision for mechanisms is higher Figure 4 : Comparing our GCN model predictions ( right ) to human annotations ( left ) . Interestingly , our model managed to correct some annotator errors ( \u201cit\u2019s\u201d , \u201cheated\u201d , \u201ccoffee warm\u201d , \u201cbeverages\u201d ) . Purposes in pink , mechanisms in green . than for purposes . Interestingly , a manual inspection revealed many cases where despite the noisy training setting , our models managed to correct mistaken or partial annotations ( see Figure 4 ) . Self - Training . According to the results , we chose GCN as our best - performing model . We experimented adding self - training [ 86 ] to GCN . Self - training is a common approach in semi - supervised learn - ing where we iteratively re - label \u201cO\u201d tags in training data with model predictions . A large portion of our training sentences are ( erroneously ) un - annotated by workers , perhaps due to annotation fatigue , introducing bias towards the \u201cO\u201d label . Self - training with GCN shows an improvement in F 1 by an ad - ditional 2 . 6 % , substantially increasing recall ( more than 12 % over Flair ) , see Figure 3 , right . Self - training stopped after 2 iterations , following no gain in F 1 on the development set . In the following two sections we demonstrate that our extraction model\u2019s accuracy , while far from perfect , is sufficient for achieving good performance on the downstream tasks which are at the focus of this paper . One main reason for this gap is that our downstream tasks involve aggregation of multiple extracted spans : Product descriptions will typically mention salient mechanisms / purposes several times in the text , such that the effect of local false positives / negatives is mitigated if overall the key aspects are captured somewhere in the text . Further , as we discuss in \u00a7 5 . 1 , our approach also aggregates purposes and mechanisms across the entire corpus , not just single texts , learning from patterns observed sufficiently many times across multiple texts and thus removing noise introduced by extraction errors . As future information extraction technologies advance , our task could benefit from improved extraction accuracy to further reduce the rate of false positives and negatives . 4 FINE - GRAINED FUNCTIONAL SEARCH FOR ALTERNATIVE USES In the previous section we suggested a model for extracting purpose and mechanism spans and assessed extraction accuracy . Our focus in this paper is to study the utility of the extracted purposes and mechanisms , in terms of the user interactions they enable . In the following sections we explore two tasks demonstrating the value of our novel representation for supporting creative innovation . We start with a task involving search for alternative uses . CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Hope et al . Motivation . Our task is inspired by one of the most well - known divergent thinking tests [ 46 ] for measuring creative ability \u2013 the alternative uses test [ 47 ] , where participants are asked to think of as many uses as possible for some object . Aside from serving as a measure of creativity , the ability to find alternative uses for technolo - gies has important applications in engineering , science and industry . Technologies developed at NASA , the US space agency , have led to over 2 , 000 spinoffs , finding new uses in computer technology , agri - culture , health , transportation , and even consumer products 1 . Procter & Gamble , the multinational consumer goods company , has invested in systematic search for ideas to re - purpose and adapt from other industries , such as using a compound that speeds up wound healing to treat wrinkles - an idea that led to a new line of anti - wrinkle products [ 27 ] . And very recently , the COVID - 19 pandemic provided a stark example of human innovation , with many companies seeking to pivot and re - purpose existing products to fit the new climate [ 28 ] . One teaching story is that of John Osher , creator of the \u201cSpin Pop\u201d \u2014 a lollipop with a mechanism for twirling in your mouth . After sell - ing his invention , Osher\u2019s team systematically searched for new ideas \u2014 \u201crather than having an idea come to us\u201d 2 . The group eventually landed on the \u201dSpin Brush\u201d \u2013 a cheap electric toothbrush adapted from the same twirling mechanisms . This case of repurposing an existing technology involved a systematic search process rather than pure serendipity . Introducing automation could help accelerate the search process by scouring many relevant problems available online , but the task is challenging for existing search systems , requiring a fine - grained , multi - aspect understanding of products . Illustrative Example . Consider a company that manufactures light bulbs . The company is familiar with straightforward usages of their product ( lamps , flashlights ) , and wants to identify non - standard uses and expand to new markets . Finding uses for a lightbulb that are not about the standard purpose of illuminating a space would be difficult to do with a standard search query over an idea repository , as the term \u201clights\u201d or \u201clighting\u201d will bring back lots of results close to \u201clamps , \u201d \u201cflashlight , \u201d and the like . In contrast , with our representation each idea is associated with mechanism and purpose aspects , and one could form a query such as mechanism = \u201clight bulb\u201d , purpose = NOT \u201clight\u201d . Using our system , the searcher adds \u201clight\u201d as a mechanism and also adds \u201clight\u201d as a negative purpose ( i . e . , results should not include \u201clight\u201d as a purpose ) . Our prototype returns examples such as billiard laser instructor devices ( Table 1 ) , warning signs on food packages to get attention of kids with allergies , and lights attached to furniture to protect your pinky toes at night ( Figure 5 ) . 4 . 1 Study Design We have built a search engine prototype supporting our representa - tion . Figure 5 shows the top two results for the light bulb scenario : warning lights on food for kids with allergies , and lights attached to furniture to protect your pinky toe at night . These are non - standard recombinations [ 30 ] ( light + allergies , light + furniture guard ) that could lead the company to new markets . 1 https : / / spinoff . nasa . gov / 2 https : / / www . allbusiness . com / the - man - the - legend - john - osher - inventor - of - the - spin - brush - part - i - 2 - 7665547 - 1 . html Figure 5 : Applications for light where light is not in the purpose . Left : Interface . Right : Two of the results and their automatic annotations ( purposes in pink , mechanisms in green ) . We conduct an experiment simulating scenarios where users wish to find novel / uncommon uses of mechanisms . Table 1 shows the sce - narios and examples . To choose these scenarios for the experiment , we find popular / common mechanisms in the dataset and their most typical uses . For example , one frequent mechanism is RFID , which is typically used for purposes such as \u201clocating\u201d and \u201ctracking\u201d . We then create queries searching for different uses \u2013 purposes that do not include concepts related to the typical uses of a given mechanism . To automate scenario selection , we cluster mechanisms ( see Section 5 . 1 ) , select frequent mechanisms from the top 5 largest mechanism clusters , and identify purposes strongly co - occurring with them ( e . g . , \u201cRFID\u201d co - occurs with \u201clocating\u201d , \u201ctracking\u201d ) to avoid . Our Approach . We represent each product \ud835\udc56 as a set of purpose vectors P \ud835\udc56 (cid:66) { p 1 \ud835\udc56 , p 2 \ud835\udc56 , . . . , p \ud835\udc43 \ud835\udc56 \ud835\udc56 } , and a set of mechanism vectors M \ud835\udc56 (cid:66) { p 1 \ud835\udc56 , p 2 \ud835\udc56 , . . . , p \ud835\udc40 \ud835\udc56 \ud835\udc56 } extracted with our GCN model . Simi - larly , we define a set of query vectors q \ud835\udc5d (cid:66) q 1 , q 2 , . . . q \ud835\udc44 \ud835\udc5d and q \ud835\udc5a (cid:66) q 1 , q 2 , . . . q \ud835\udc44 \ud835\udc5a . Each query chunk can be negated , meaning it should not appear . Finally , we define distance metrics \ud835\udc51 \ud835\udc5d ( \u00b7 , \u00b7 ) , \ud835\udc51 \ud835\udc5a ( \u00b7 , \u00b7 ) between sets of purposes and mechanisms . For example , to locate a dog using RFID but not GPS : argmin \ud835\udc56 \ud835\udc51 \ud835\udc5d ( { q \u201clocate dog\u201d } , P \u02dc \ud835\udc56 ) \ud835\udc60 . \ud835\udc61 . \ud835\udc51 \ud835\udc5a ( { q \u201cGPS\u201d } , M \u02dc \ud835\udc56 ) \u2265 threshold \ud835\udc51 \ud835\udc5a ( { q \u201cRFID\u201d } , M \u02dc \ud835\udc56 ) \u2264 threshold ( 1 ) Scaling Creative Inspiration with Fine - Grained Functional Aspects of Ideas CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA We explore two alternatives for computing distance metrics \ud835\udc51 \ud835\udc5a , \ud835\udc51 \ud835\udc5d : \u2022 FineGrained - AVG . \ud835\udc51 \ud835\udc5d ( q \ud835\udc5d , P \ud835\udc56 ) is 1 minus the dot product be - tween average query and purpose vectors ( normalized to unit norm ) . We define \ud835\udc51 \ud835\udc5a similarly . \u2022 FineGrained - MAXMIN . We match each element in q \ud835\udc5d with its nearest neighbor in P \ud835\udc56 , and then find the minimum over the distances between matches . \ud835\udc51 \ud835\udc5d is defined as 1 minus the minimum . All vectors are normalized . We define \ud835\udc51 \ud835\udc5a similarly . This captures cases where queries match only a small subset of product chunks , erring on the side of caution with a max - min approach . Baselines . We test our model against : \u2022 AvgGloVe . A weighted average of GloVe vectors of the entire text ( excluding stopwords ) , similar to standard NLP approaches for retrieval and textual similarity . We average query terms and normalize to unit norm . Distance is computed via the dot product . \u2022 Aggregate purpose / mechanism . Representing each document with the model in [ 52 ] . This model takes raw text as input , applies a BiLSTM neural network and produces two vectors correspond - ing to aggregate purpose and mechanism of the document . We average and normalize query vectors , and use the dot product . For all four methods , we handle negative ( purpose ) queries by filtering out all products whose similarity is lower than \ud835\udf06 , where lambda is a threshold selected to be the 90 th percentile of similarities ( 1 minus the distances ) . This corresponds to the threshold seen in the example in Eq . 1 . 4 . 2 Results We recruited five engineering graduate students ( three female , two male ) to judge the retrieved product ideas . Each participant provided binary relevance feedback [ 88 ] ( yes / no ) to the top 20 results from each of the four methods , shuffled randomly so that judges are blind to the condition . 3 See Figure 6 for results . We report Non Cummulative Discounted Gain ( NDCG ) and Mean Average Precision ( MAP ) , two common metrics in information retrieval [ 88 ] . Our FineGrained - AVG wins for both metrics , followed by FineGrained - MAXMIN . The baselines perform much worse , with the aggregate - vectors approach in [ 52 ] outperforming standard embedding - based retrieval with GloVe . Im - portantly , our approach achieves high MAP ( 85 % - 87 % ) in absolute terms , in addition to a large relative improvement over the baselines ( MAP of 40 % - 60 % ) . Qualitative Analysis . Table 1 shows example results of FineGrained - AVG . For instance , a query for using light not for lighting results in laser - based billiard instructions . A query for using RFID not for locating or tracking results in an idea for an RFID - based lock , or RFIDs used at supermarket checkouts . To give an intuition for what might be driving our quantitative findings , we examine examples of retrieved results . For instance , with the query for using light for the non - standard purpose of cleaning , the top ranked result retrieved by FineGrained - AVG is a UV Light Sterilizer , with extracted purposes including Sterilizes bacteria , Keep public and people healthy and Cleaner 3 Inter - rater agreement measured across all scenarios was at 50 % by both Fleiss kappa and Krippendorff\u2019s alpha tests . Figure 6 : Results for search evaluation test case . Mean average precision ( MAP ) and Normalized Discounted Cumulative Gain ( NDCG ) by method , averaged across queries . Methods in bold use our model . fresher air , and the top result from FineGrained - MAXMIN is sim - ilarly a Standalone bug zapper bulb that uses uv light / black light . Conversely , the top result for both baselines ( standard search and aggregate - vectors ) is a Toilet / Bathroom Light , with \u201ca sensor light that glows around your toilet\u201d and \u201cextra batteries if you lose elec - tricity in the bathroom\u201d . It appears that both baselines were not able to accurately capture and disentangle purposes and mechanisms , despite the aggregate - vector being explicitly designed for that . More generally , it appears that the aggregate - vector approach squashes multiple purposes together by design into one soft , ag - gregate vector , which in this case includes concepts like toilet and bathroom that are somewhat topically related to cleaning . The aggre - gate approach had similar issues in the next product ideas it retrieved ( e . g . , Switch that glows in the dark , a Dash Light to illuminate ash trays ) . Overall , our results demonstrate that fine - grained purposes and mechanisms lead to better functional search expressivity than ap - proaches based on distributional representations or coarse purpose - mechanism vectors . 5 EXPLORING THE DESIGN SPACE WITH A FUNCTIONAL CONCEPT GRAPH In this section we test the value of our novel representation for supporting users in exploring the design space for solving a given problem . We use our span - based representation to construct a corpus - wide graph of purpose / mechanism concepts . We demonstrate the utility of this approach in an ideation task , helping users identify useful inspirations in the form of problems that are related to their own . Our goal is to help users \u201cbreak out\u201d of fixation on a certain domain , a well - known hindrance to innovation [ 15 , 60 ] . Doing so is challenging because it requires some level of abstraction : being able to go beyond the details of a concrete problem to connect to a part CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Hope et al . Query Example results Mechanism : light . Purpose : NOT light Billiard laser instructor ( projector ) Mechanism : solar energy . Purpose : NOT generating power Light bulbs with built - in solar chips . Mechanism : water . Purpose : NOT cleaning , NOT drinking A lighter that burns hydrogen generated from water and sunlight . Mechanism : RFID . Purpose : NOT locating , NOT tracking A digital lock for your luggage with RFID access . Mechanism : light . Purpose : cleaning A UV box to clean and sanitize barbells at the gym . Table 1 : Scenarios and example results retrieved by our FineGrained - AVG method . Queries reflect non - trivial uses of mechanisms ( e . g . , using water not for drinking / cleaning , retrieving a lighter running on hydrogen from water and sunlight ) . of the design space that may look dissimilar on the surface , but has abstract similarity . Numerous studies in engineering and cognitive psychology have shown the benefits of problem abstractions for ideation [ 32 , 34 , 45 , 60 , 64 , 98 , 99 ] . However , these studies either involve non - scalable methods ( relying on highly - structured annota - tions , or on crowd - sourcing ) or simple , syntactical pattern - matching heuristics incapable of capturing deeper abstract relations . In [ 52 ] ( aggregate - vectors baseline from the previous section ) , crowdwork - ers were given a product description from the Quirky database , and asked to come up with ideas for products that solve the same prob - lem in a different way . Aggregate vectors representing purpose and mechanism were used to find near - purpose , far - mechanism analo - gies . Thus , finding analogies relied on having a given mechanism to control for structural distance . Unlike [ 52 ] , in our setup we assume a more realistic scenario where we are given only a short problem description \u2013 e . g . , generat - ing power for a phone , reminding someone to take medicine , folding laundry \u2013 and aim to find inspirational stimuli [ 45 ] in the \u201csweet spot\u201d for creative ideation \u2013 structurally related to the given problem , not too near yet also not too far [ 33 ] . Functional Concept Graph . To address this challenge , we build a tool inspired by functional modeling [ 51 ] , which we call a Func - tional Concept Graph . A functional model is , roughly put , a hierar - chical ontology of functions and ways to achieve them , and is a key concept in engineering design . Such models are especially useful for innovation , allowing problem - solvers to break out of a fixed overly - concrete purpose or mechanism and move up and down the hierarchy . Despite their great potential , today\u2019s functional models are constructed manually , and thus do not scale . While automati - cally inducing full abstraction hierarchies / ontologies of functional properties of real - world products remains a daunting challenge , in our approach we construct a rough approximation \u2014 simple enough to extract automatically from noisy product texts , while still being useful for exploring the design space and suggesting inspirations to users . Specifically , in our approach Functional Concept Graphs consist of nodes corresponding to purposes or mechanisms , and edges reflect semantic relatedness that is not guaranteed to directly encode abstraction . We build this graph by observing fine - grained co - occurrences of concepts appearing together in products , using rule - mining to infer which concept is likely to be more general to ( roughly ) capture different levels of abstraction . For example , Figure 7 shows an actual subgraph from our auto - matically constructed functional concept graph related to electricity , power and charging . Products that mention certain purposes ( e . g . , \u201ccharge your phone \" ) will often mention other , structurally related problems that could be more general / abstract ( e . g . , \u201cgenerate power \" ) Figure 7 : An example of our learned functional concept graph extracted from texts . Mechanism in green , purpose in pink . Ti - tles are tags nearest to cluster centroids ( redacted to fit ) . or more specific ( \u201cwireless phone charging \" ) , resulting in edges in our graph ( only high - confidence edges are shown ) . A designer could go from the problem of charging batteries to the more general prob - lem of generating power , and from there to another branch ( e . g . , solar power and mechanical stored energy ) , to get inspired by structurally related ideas . 5 . 1 Building a Functional Concept Graph We develop a method to infer this representation from co - occurrence patterns of the fine - grained spans of text . Naively looking for co - occurrences of problems may yield inspirations too near to the orig - inal \ud835\udc5d \ud835\udc56 , as many frequently co - occurring purposes tend to be very similar , while we are interested discovering the more abstract re - lations . In addition , raw chunks of text extracted from our tagging model have countless variants that are not sufficiently abstract and are thus sparsely co - occurring . We thus design our approach to en - courage abstract inspirations . As an overview of our approach before presenting the technical details , we take the following two steps : I . Concept discretization . Intuitively , nodes in our graph should correspond to groups of related spans ( \u201ccharging\u201d , \u201ccharging the battery\u201d , \u201ccharging a laptop\u201d ) . To achieve this , we take all purpose and mechanism spans \u02c6 P , \u02c6 M in the corpus , extracted using our GCN model , and cluster them ( separately ) , using pre - trained vector representations . We refer to the clusters C \ud835\udc5d , C \ud835\udc5a as concepts . II . Relations . We employ rule - mining [ 80 ] to discover a set of re - lations R between concepts ( see \u00a7 5 . 2 for implementation details ) . Scaling Creative Inspiration with Fine - Grained Functional Aspects of Ideas CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Relations are Antecedent = \u21d2 Consequent , with weights correspond - ing to rule confidence . To illustrate our intuition , suppose that when \u201cprevent head injury\u201d appears in a product description , the condi - tional probability of \u201csafety\u201d appearing too is large ( but not the other way around ) . In this case , we can ( weakly ) infer that preventing head injuries is a sub - purpose of \u201csafety\u201d . Indeed , manually observing the purpose - purpose edges , the one - directional relations captured are often sub - purpose , and the bi - directional ones often encode abstract similarity . Similarly , for mechanism concepts the one - directional relations are often part of ( \u201ccell phone\u201d and \u201cbattery\u201d ) , and bi - directional are mechanisms that co - occur often . For pairs of purpose and mechanism concepts , the relation is often functionality ( \u201ccharger\u201d , \u201ccharge\u201d ) . Exploring more relations is left for future work . 5 . 2 Study Design & Implementation Next , we set out to test the utility of the functional concept graph in an ideation task . In our setup participants are given problems ( e . g . , reminding people to take their medication in the morning ) and are asked to think of creative solutions . Participants were also given a list of potential inspirations , grouped into boxes , and were instructed to mark whether each was novel and helpful . They were also encouraged to explain the solution it inspired . Figure 8 shows our interface . In this example , seeing inspirations about health monitors caused one user to suggest monitoring the person to find the best time to remind them to take medicine ; see - ing inspirations about coffee caused them to suggest integrating medicine reminders into coffee machines . To create a set of seed problems , a graduate student mapped between problems from WikiHow . com ( a website of how - to guides ) to purposes in our data . Using this source allowed us to collect real - world problems that are broadly familiar , with succinct and self - explanatory titles that do not require further reading to understand . The student was tasked with confirming that our Quirky dataset contains idea descriptions that mention these problems . For a given problem in WikiHow ( how to remember to take medication ) , they performed keyword search over 17 \ud835\udc3e purpose spans gleaned by our model from Quirky , and found matching spans ( morning medicine reminder ) . We use those matching spans as our seed problem descrip - tion given to users ( purple text in Figure 8 ) . We collect 25 problems this way . Table 2 shows more examples , such as Tracking distance walked , folding laundry or sensing dryness level . Inspirations are other purpose spans from our dataset ( see Table 2 ) , selected automatically using our approach or baseline approaches . Our Method . For our approach , we construct a functional concept graph as in Section 5 . 1 . To cluster related spans into concept nodes , we explore two common and powerful vector representations of spans to capture semantic similarity : \u2022 GloVe [ 81 ] pre - trained word embeddings , averaged across tokens . \u2022 BERT - based [ 84 ] contextualized vectors that have been fine - tuned for semantic similarity tasks 4 . 4 We use RoBERTa - large - STS - SNLI , available at github . com / UKPLab / sentence - transformers . Figure 8 : A snippet from our ideation interface for \u201cmorn - ing medicine reminder\u201d . Users see inspirations grouped into boxes . Each box is supposed to represent a concept \u2013 a clus - ter of related spans as found by our method or by the baselines ( see \u00a7 5 . 1 ) . Users indicate which inspirations were useful , and what ideas they inspired . For example , seeing \u201creal time health checker\u201d inspired one user to suggest a monitoring device for finding the best time for reminding to take the medicine . We cluster the spans using K - Means + + 5 [ 8 ] . We then apply the Apriori algorithm 6 to automatically mine association rules between clusters , [ 80 ] and use the confidence metric to select the top rules 7 . To use the mined rules between purpose nodes ( clusters ) for select - ing inspirations shown to users , we start from the purpose node corresponding to the given problem and take its consequents ; as explained earlier , this captures a weak signal of abstract similarity . Some of these nodes contain tens of spans in them . Thus , we also explore two approaches to \u201csummarize\u201d each concept cluster with representative spans displayed to users \u2013 one that attempts to summarize the cluster independently of the seed problem , and one that takes the seed problem into account : \u2022 TextRank [ 74 ] . We construct a graph where nodes are the spans in a cluster and edges represent textual similarity . We run PageRank [ 78 ] on this graph , selecting the top \ud835\udc3e spans to present . \u2022 Nearest spans . Following the findings in [ 33 ] , select the top \ud835\udc3e spans in C \ud835\udc5d that are nearest to the query \ud835\udc5d \ud835\udc56 . ( For both approaches , we use \ud835\udc3e = 5 ) . Baselines . 8 \u2022 Purpose span similarity . Given a problem \ud835\udc5d \ud835\udc56 , we find the \ud835\udc3e = 5 nearest purpose spans of text in our corpus ( out of 17 \ud835\udc3e purposes ) . We experiment with the same two vector representations used by our approach : GloVe and BERT . This method is similar to applying the methodology in [ 52 ] to our setting , where in our setting we are given only a problem \ud835\udc5d \ud835\udc56 and no mechanism \ud835\udc5a \ud835\udc56 is available to control for structural distance . While this approach relies on our model for extracting purpose spans , we consider it a baseline to study the added value of our hierarchy . 5 \ud835\udc3e = 250 selected automatically with elbow - based criteria on silhouette scores . 6 http : / / www . borgelt . net / pyfim . html . 7 We use the top 3 rules in our experiment . 8 We note that all methods and baselines include both single and multi - word spans of text as inspirations , ensuring users are blind to the condition . CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Hope et al . Problem Inspirations Rater explanation Track distance walked Protect children Get ideas from devices that keep track of children Folding laundry Store toilet paper Roll laundry around a tube instead of folding Dispense medicine Pet bowl that keeps ants away Based on pet bowls that can dispense food during the day Sense dryness level Voltage reading Use electric current to measure water level ( safely ) Waterproof Ideas from sensors in waterproof devices Temperature reading Morning medicine reminder Schedule coffee , coffee alarm Alarm clock with coffee and medicine reminders Send vital data , real - time health checker Health trackers to tell if medicine not taken , alert accordingly Heart rate monitoring , con - tinuously monitor glucose Find the best time to take medicine Table 2 : Example inspirations and explanations given by human evaluators . \u2022 Linguistic abstraction . We use the WordNet [ 75 ] lexical data - base to extract hypernyms ( for each token in \ud835\udc5d \ud835\udc56 ) , in order to capture potential abstractions . WordNet is often used in similar fashion for design - by - analogy studies [ 42 , 45 , 64 ] . \u2022 Random concepts . Random inspirations are often considered as a baseline in ideation studies since diversity of examples is a known booster for creative ability [ 52 ] . For each task , we select a random cluster from C \ud835\udc5d and display its TextRank summary . Study Participants . We recruit 10 raters to evaluate the inspirations , via university mailing lists . 8 raters were engineering graduate stu - dents , and the remaining two raters included a senior engineering professor and an architect . This cohort is intended to reflect a target user base of people interested in innovation and involved in creative inventive thinking as part of their work . Rating Collection . In our study , each method generated \ud835\udc3e = 5 spans ( concept summaries ) , which are grouped and displayed together in a box ( see Figure 8 ) . For each problem a rater views 8 boxes in randomized order , to avoid bias . Raters were instructed to mark inspirations they consider useful and relevant for solving a given problem , while being not about the same problem . Raters were also encouraged to write comments , especially for non - trivial cases which they found of interest ( see Table 2 ) . In total , raters viewed 2584 boxes , or 12920 purpose descriptors . 5 . 3 Results Analyzing inspirations . Table 2 and Figure 8 show examples of problems , inspirations and user explanations from our study . For in - stance , users facing the \u201cmorning medicine reminder\u201d problem were presented with nearby concepts in the Functional Concept Graph that included health monitoring and coffee machines . To explore why these concepts are connected in our graph and why they are potentially useful as inspirations , we make use of the direct inter - pretability of our approach . We examine the purpose co - occurrences from which the Functional Concept Graph was constructed . Figure 9 shows the subgraph with concept nodes of Making hot drinks , alerting / reminding , health monitoring , medicine delivery , and edges representing products in which two adjacent purposes were co - mentioned ( e . g . , a \u201ccoffee machine alarm\u201d product that men - tioned the purposes of making hot drinks and alerting / reminding , or Alert / remind Making hot drinks Medicine delivery Medical monitoring Coffee machine alarm Smart medicine injector Smart medicine injector pill reminder Smart medicine injector Figure 9 : Excerpt from our Functional Concept Graph . Nodes represent concepts ( clusters of purposes ) . To give intuition for how edges were created , they are annotated with example prod - ucts containing spans from both concepts . All nodes and edges in this figure were automatically constructed and used to create the user - facing inspirations shown in Figure 8 . This figure pro - vides a graphic illustration ( not shown to users ) explaining how the boxes in Figure 8 are generated , with node names provided here by us for readability . The problem of \u201cmedicine morning reminder\u201d is mapped ( via embedding ) to the Alert / remind con - cept cluster ( as named by us ) , which is linked to the concepts of medical monitoring and making hot drinks through products such as \u201csmart medicine injector\u201d and \u201ccoffee machine alarm\u201d ( among others , not displayed in the figure ) . These links serve as the source for inspirations in our study , as seen in Figure 8 . a \u201csmart medicine injector \" that mentioned both alerting / reminding and medicine delivery ) . This explains why the concepts are nearby in the graph , as there are multiple products in our dataset that share purposes from both concepts . For example , a \u201cpill reminder\u201d product refers to the problem of forgetting to take medicine at prescribed times ( Sends notification if you forgot to take your AM or PM meds ) , while a \u201csmart injec - tor\u201d device administers medicine on set time intervals . At the same time , both of these products mention purposes of medicine delivery . Scaling Creative Inspiration with Fine - Grained Functional Aspects of Ideas CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA When our graph construction algorithm observes enough similar co - occurrence patterns between the concepts of alerting and medicine delivery , across multiple products , an edge is added between the two in the graph . Similarly , an \u201cAlarm coffee maker\u201d product mentions the purposes of time management and making coffee at a set time as well as alerting when the coffee is ready , explaining how it emerges as a potential inspiration in our graph . This type of linkage or overlap between an original problem space and inspiration problems helps get at a sweet - spot of innovation [ 16 ] by finding ideas that are not too near and not too far from the original problem , helping users break out of fixation as discussed earlier in this section . Users in our study used these inspirations to come up with a tracker that alerts the user at the best time to take a medicine and a coffee machine reminding the user to take their medication with their morning coffee . Those creative directions demonstrate the utility of the Functional Concept Graph for exploring the design space . Quantitative results . Figure 10 shows the results of the user study . On the left , we show the proportion of inspirations ( individual spans ) selected by at least two raters , for each method . Our approach sig - nificantly outperforms all the baselines . The effect is particularly pronounced for the BERT - based approach , with 51 % of inspirations found useful , while the best baseline reaches less than 30 % . Interest - ingly , for both BERT and GloVe representations , the Nearest - span summarization approach fares better , potentially due to striking a balance between being too far / near the initial problem \ud835\udc5d \ud835\udc56 . Figure 10 ( right ) shows the proportion of inspiration boxes that got at least 2 individual inspirations marked ( by at least 2 raters ) . This metric measures the effect of a box as one unit , as each box is meant to represent a coherent cluster . Our method is able to reach 62 % , while the best baseline ( GloVe search on purpose spans ) yields only 39 % . Again , the nearest - span summarization is prefered to TextRank . Importantly , for both individual inspiration spans and inspirations boxes , 51 % - 62 % are rated as useful \u2013 high figures considering the challenging nature of the task . 6 DISCUSSION AND CONCLUSION In this paper we introduced a novel span - based representation of ideas in terms of their fine - grained purposes and mechanisms and used it to develop new tools for creative ideation . We trained a model to extract spans from a noisy , real - world corpus of products . We used this representation to support human creativity in two applications : expressive search for alternative , uncommon uses of products , and generating a graph to help problem - solvers explore the design space around their problem . In both ideation studies , we were able to achieve high accuracies , significantly outperform baselines and help boost user creativity . 6 . 1 Limitations While our results showed the promise of a functional aspect - based representation , and demonstrated potentially feasible technical ap - proaches for extracting this representation from unstructured text , the approach has several limitations . Challenging Annotation Task for Crowds . First , the annotation task proved to be somewhat difficult for crowdworkers , and the out - puts were noisy . One direction for future work would be to explore weak supervision approaches to augment annotation . One issue that might exacerbate the problem is that sometimes the boundary be - tween purpose and mechanism is fuzzy , and it is genuinely difficult to tell how to annotate the span . Limited Functional Schema . In a similar vein , it might be interest - ing to explore more expressive schemas , containing elements other than just purpose and mechanism ( similar to [ 12 ] ) . One particularly useful element to add might be context / constraints ( e . g . , nanoscale ) , to restrict the search space to feasible solutions . Surface Form Abstraction . Another limitation of our approach is that our functional aspects ( and resulting embeddings ) remain quite closely anchored to the original texts . This limits their ability to be used to match across domains , to make connections such as inspiring new optimization approaches by analogy to \" heating / cooling \" sched - ules in metallurgy . Achieving abstraction to match across distant domains without burdening the user with a combinatorial explosion of noisy matches remains an open problem . We wonder if abstract - ing key objects or entities in a purpose functional aspect \u2014 such as a more automated approach to replacing objects with their \" com - monsense \" properties \u2014 might be more feasible than attempting to abstract from an entire product description or abstract , given that the chunk is already a rich signal of the product\u2019s functional meaning . 6 . 2 Future Work and Broader Implications Moving to future directions , we are excited about the potential of functional aspects to lead to advances in the interpretability of content - based recommender systems in these complex domains . Keywords are inherently interpretable , but are limited in their ca - pacity to support crossing knowledge boundaries ; and until now , embedding - based approaches ( e . g . , [ 52 ] ) have not always led to interpretable justifications for matches . Functional aspects could provide the basis of not just more powerful search operators , but also more interpretable results and feedback loops . Deeper Functional Graph Exploration . A key component for the above might be expanding on our use of functional graphs , built from the extracted functional aspects . In our experiments , we used our functional concept graph to retrieve inspirations from \" around \" the problem . But what would it take to be able to explore this graph ? Could we identify and optimize for latent coordinates in the func - tional space , moving \" up \" and \" down abstraction levels , or \" across \" sibling nodes in a functional graph ? Taking inspiration from network perspectives on ideation [ 11 , 44 , 94 ] , could we retrieve interesting \" lineages \" of ideas , or compute the potential inspiration value of functional aspects based on network connectivity metrics ? Could we combine these content - based functional aspects with measures of use ( e . g . , citations ) , to enrich approaches that combine content - and social - based signals , such as literature - based discovery [ 93 , 95 ] ? Identifying Overlap and Gaps Across Fields . These approaches rely on identifying interesting overlaps in concepts that simultane - ously coincide with disjunctions in citations , as signals of potentially impactful \" undiscovered public knowledge \" : a persistent challenge is how to define \" concepts \" - keyword or unstructured text approaches can lead to combinatorial explosions of noise to sift through , and con - trolled vocabulary ( e . g . , MEDLine ) can help increase the signal to noise ratio , but are only available in specialized circumstances [ 89 ] . CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Hope et al . Figure 10 : Inspiration user study results . Left : Proportion of inspirations selected by at least 2 raters , per condition . Right : Proportion of boxes ( clusters ) with at least 2 spans marked by \u2265 2 raters . Functional aspects might be a useful bridge between unstructured text and controlled vocabularies for identifying points of overlap and disjunction between different fields , accelerating the discovery of gems hidden in plain sight . Functional aspects for Collaborative Ideation . Future work could also explore new interactions in collaborative and crowd innovation that might be enabled by the ability to quickly extract functional aspects in idea corpora . The source of our primary dataset here , Quirky , was actually a crowd innovation platform . HCI research on these platforms have begun to emphasize moving away from mere \" selection \" of best ideas from large samples of ideas , towards sup - porting generative collaboration over ideas . Open problems include synthesizing major themes in large - scale corpora of user - generated ideas and identifying gaps in the idea space [ 13 , 68 , 90 ] , as well as supporting intelligent matching and structuring ways for crowd innovators to collaborate and build on each others\u2019 ideas [ 14 , 69 ] between crowd innovators . We are excited about the potential for functional aspects to assist with these functions , as a complement to other approaches like crowd - powered synthesis [ 7 , 18 , 43 , 91 ] . Here , too , the potential for functional aspects to be highly inter - pretable could power novel explorations of mixed - initiative systems for augmenting collaborative ideation at scale [ 67 , 91 ] . Mapping of Design Spaces . Beyond supporting richer search for creative inspiration , a data - driven approach to extracting functional aspects and learning relationships between the aspects could power much more expansive approaches to mapping out design spaces for entire domains or problem areas , identifying key subproblems and constraints and novel paths through the design space . Mapping approaches like this , such as technological roadmapping [ 10 ] , have already shown significant promise for reinvigorating research and development in real - world applications such as neural recording [ 70 ] . However , these mapping exercises are still highly manual and labor - intensive processes ; computational support for such tasks could have transformative impacts on innovation . REFERENCES [ 1 ] The car mechanic who uncorked a childbirth revolution . BBC News , 2013 . [ 2 ] G . Aguilar , S . Maharjan , A . P . L . Monroy , and T . Solorio . A multi - task approach for named entity recognition in social media data . In 3rd Workshop on Noisy User - generated Text , 2017 . [ 3 ] A . Akbik , T . Bergmann , D . Blythe , K . Rasul , S . Schweter , and R . Vollgraf . Flair : An easy - to - use framework for state - of - the - art nlp . In NAACL - HLT , 2019 . [ 4 ] A . Akbik , T . Bergmann , and R . Vollgraf . Pooled contextualized embeddings for named entity recognition . In NAACL , 2019 . [ 5 ] A . Akbik , D . Blythe , and R . Vollgraf . Contextual string embeddings for sequence labeling . In International Conference on Computational Linguistics , 2018 . [ 6 ] G . Altshuller . 40 principles : TRIZ keys to innovation . 2002 . [ 7 ] P . Andr\u00e9 , A . Kittur , and S . P . Dow . Crowd Synthesis : Extracting Categories and Clusters from Complex Data . In Proceedings of the 17th ACM Conference on Computer Supported Cooperative Work & Social Computing , CSCW \u201914 , pages 989 \u2013 998 , New York , NY , USA , 2014 . ACM . [ 8 ] D . Arthur and S . Vassilvitskii . k - means + + : The advantages of careful seeding . In ACM - SIAM symposium on Discrete algorithms , 2007 . [ 9 ] I . Augenstein , M . Das , S . Riedel , L . Vikraman , and A . McCallum . Semeval 2017 task 10 : Scienceie - extracting keyphrases and relations from scientific publica - tions . arXiv preprint arXiv : 1704 . 02853 , 2017 . [ 10 ] E . S . Boyden and A . H . Marblestone . Architecting Discovery : A Model for How Engineers Can Help Invent Tools for Neuroscience . Neuron , 102 ( 3 ) : 523 \u2013 525 , May 2019 . Publisher : Elsevier . [ 11 ] H . Cai , E . Y . Do , and C . M . Zimring . Extended linkography and distance graph in design evaluation : An empirical study of the dual effects of inspiration sources in creative design . Design Studies , 31 ( 2 ) : 146 \u2013 168 , 2010 . [ 12 ] J . Chan , J . Chang , T . Hope , D . Shahaf , and A . Kittur . Solvent : A mixed initiative system for finding analogies between research papers . CSCW , 2018 . [ 13 ] J . Chan , S . Dang , and S . P . Dow . Comparing Different Sensemaking Approaches for Large - Scale Ideation . In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems , pages 2717 \u2013 2728 . ACM , 2016 . [ 14 ] J . Chan , S . Dang , and S . P . Dow . Improving Crowd Innovation with Expert Facilitation . In Proceedingsofthe19thACMConferenceonComputer - Supported CooperativeWork & SocialComputing , CSCW\u201916 , pages1223 \u2013 1235 , NewYork , NY , USA , 2016 . ACM . [ 15 ] J . Chan , S . P . Dow , and C . D . Schunn . Do The Best Design Ideas ( Really ) Come From Conceptually Distant Sources Of Inspiration ? Design Studies , 2015 . [ 16 ] J . Chan , K . Fu , C . Schunn , J . Cagan , K . Wood , and K . Kotovsky . On the benefits and pitfalls of analogies for innovative design : Ideation performance based on analogical distance , commonness , and modality of examples . Journal of mechanical design , 2011 . [ 17 ] J . Chan , T . Hope , D . Shahaf , and A . Kittur . Scaling up analogy with crowdsourc - ing and machine learning . In ICCBR - 16 . [ 18 ] L . B . Chilton , G . Little , D . Edge , D . S . Weld , and J . A . Landay . Cascade : Crowdsourcing taxonomy creation . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems , pages 1999 \u2013 2008 . ACM , 2013 . Scaling Creative Inspiration with Fine - Grained Functional Aspects of Ideas CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA [ 19 ] D . W . Dahl and P . Moreau . The influence and value of analogical thinking during new product ideation . Journal of Marketing Research , 2002 . [ 20 ] E . De Bono . Lateral thinking . [ 21 ] E . De Bono . Lateral thinking : a textbook of creativity . Penguin UK , 2010 . 01064 . [ 22 ] J . Devlin , M . - W . Chang , K . Lee , and K . Toutanova . Bert : Pre - training of deep bidirectional transformers for language understanding . In NAACL - HLT , 2019 . [ 23 ] K . Dorst . The core of \u201cdesign thinking\u201d and its application . Design Studies , 2011 . [ 24 ] H . Dubberly and S . Evenson . On Modeling : The Analysis - synthesis Bridge Model . interactions , 2008 . [ 25 ] J . R . Duflou and P . - A . Verhaegen . Systematic innovation through patent based product aspect analysis . CIRP Annals - Manufacturing Technology , 2011 . [ 26 ] C . Eckert and M . Stacey . Adaptation of Sources of Inspiration in Knitwear Design . Creativity Research Journal , 15 ( 4 ) : 355 \u2013 384 , 2003 . 00000 . [ 27 ] K . Essick . Technologyscouts : hopingtofindthenextbigthing . ScienceBusiness , Feb . 2006 . [ 28 ] K . Essick . Innovation and creativity in a time of crisis . Science Business , 2020 . [ 29 ] M . F\u00e4rber , F . Bartscherer , C . Menne , and A . Rettinger . Linked data quality of dbpedia , freebase , opencyc , wikidata , and yago . Semantic Web , 2018 . [ 30 ] L . Fleming . Recombinant uncertainty in technological search . Management science , 47 ( 1 ) : 117 \u2013 132 , 2001 . [ 31 ] J . Frich , L . MacDonald Vermeulen , C . Remy , M . M . Biskjaer , and P . Dalsgaard . Mapping the landscape of creativity support tools in hci . In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems , pages 1 \u2013 18 , 2019 . [ 32 ] K . Fu , J . Cagan , K . Kotovsky , and K . L . Wood . Discovering Structure In Design Databases Through Functional And Surface Based Mapping . JMD , 2013 . [ 33 ] K . Fu , J . Chan , J . Cagan , K . Kotovsky , C . Schunn , and K . Wood . The Meaning of Near and Far : The Impact of Structuring Design Databases and the Effect of Distance of Analogy on Design Output . JMD , 2013 . [ 34 ] K . Fu , J . Chan , C . Schunn , J . Cagan , and K . Kotovsky . Expert representation of design repository space : A comparison to and validation of algorithmic output . Design Studies , 2013 . [ 35 ] M . GenandL . Lin . Geneticalgorithms . WileyEncyclopediaofComputerScience and Engineering , 2007 . [ 36 ] D . Gentner , S . Brem , R . Ferguson , P . Wolff , A . B . Markman , and K . Forbus . Analogy and creativity in the works of johannes kepler . Creative thought : An investigation of conceptual structures and processes , 1997 . [ 37 ] D . Gentner and K . J . Kurtz . Relational Categories . In Categorization inside and outside the laboratory : Essays in honor of Douglas L . Medin , APA decade of behavior series . American Psychological Association , Washington , DC , US , 2005 . [ 38 ] D . Gentner and A . B . Markman . Structure mapping in analogy and similarity . American psychologist , 1997 . [ 39 ] K . Gericke and B . Eisenbart . The integrated function modeling framework and its relation to function structures . AI EDAM , 2017 . [ 40 ] M . L . GickandK . J . Holyoak . Analogicalproblemsolving . Cognitivepsychology , 12 ( 3 ) : 306 \u2013 355 , 1980 . [ 41 ] M . L . Gick and K . J . Holyoak . Schema induction and analogical transfer . Cogni - tive Psychology , 15 ( 1 ) , 1983 . [ 42 ] K . Gilon , J . Chan , F . Y . Ng , H . Lifshitz - Assaf , A . Kittur , and D . Shahaf . Analogy mining for specific design needs . In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems , CHI \u201918 , pages 121 : 1 \u2013 121 : 11 . ACM , 2018 . [ 43 ] V . Girotto , E . Walker , and W . Burleson . The Effect of Peripheral Micro - tasks on Crowd Ideation . In Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems , CHI \u201917 , pages 1843 \u2013 1854 , New York , NY , USA , 2017 . ACM . [ 44 ] M . Gon\u00e7alves and P . Cash . The life cycle of creative ideas : Towards a dual - process theory of ideation . Design Studies , 72 : 100988 , Jan . 2021 . 00000 . [ 45 ] K . Goucher - Lambert and J . Cagan . Crowdsourcing inspiration : Using crowd generated inspirational stimuli to support designer ideation . Design Studies , 2019 . [ 46 ] J . P . Guilford . Three faces of intellect . American psychologist , 1959 . [ 47 ] J . P . Guilford . The nature of human intelligence . 1967 . [ 48 ] A . Hargadon and R . I . Sutton . Technology Brokering and Innovation in a Product Development Firm . Administrative Science Quarterly , 42 ( 4 ) : 716 \u2013 749 , 1997 . 03534 Publisher : [ Sage Publications , Inc . , Johnson Graduate School of Management , Cornell University ] . [ 49 ] S . R . Herring , C . - C . Chang , J . Krantzler , and B . P . Bailey . Getting inspired ! : understanding how and why examples are used in creative design practice . In Proceedings of the 27th international conference on Human factors in computing systems , pages 87 \u2013 96 . ACM , 2009 . [ 50 ] M . B . Hesse . Models and analogies in science . 1966 . [ 51 ] J . Hirtz , R . Stone , D . A . McAdams , S . Szykman , and K . Wood . A functional basis for engineering design : reconciling and evolving previous efforts . Research in engineering Design , 2002 . [ 52 ] T . Hope , J . Chan , A . Kittur , and D . Shahaf . Accelerating innovation through analogy mining . In KDD , 2017 . [ 53 ] Z . Huang , W . Xu , and K . Yu . Bidirectional lstm - crf models for sequence tagging . arXiv preprint arXiv : 1508 . 01991 , 2015 . [ 54 ] D . G . Jansson and S . M . Smith . Design fixation . Design Studies , 12 ( 1 ) : 3 \u2013 11 , 1991 . [ 55 ] D . Jin and P . Szolovits . Hierarchical neural networks for sequential sentence classification in medical scientific abstracts . arXiv preprint arXiv : 1808 . 06161 , 2018 . [ 56 ] A . Kerne , N . Lupfer , R . Linder , Y . Qu , A . Valdez , A . Jain , K . Keith , M . Carrasco , J . Vanegas , and A . Billingsley . Strategies of Free - Form Web Curation : Processes of Creative Engagement with Prior Work . In Proceedings of the 2017 ACM SIGCHI Conference on Creativity and Cognition , C & C \u201917 , pages 380 \u2013 392 , New York , NY , USA , 2017 . ACM . [ 57 ] A . Kerne , A . M . Webb , S . M . Smith , R . Linder , N . Lupfer , Y . Qu , J . Moeller , and S . Damaraju . Using metrics of curation to evaluate information - based ideation . ACM Transactions on Computer - Human Interaction ( ToCHI ) , 21 ( 3 ) : 1 \u2013 48 , 2014 . [ 58 ] T . N . Kipf and M . Welling . Semi - Supervised Classification with Graph Convolu - tional Networks . sep 2016 . [ 59 ] S . Kirkpatrick , C . D . Gelatt , and M . P . Vecchi . Optimization by simulated annealing . science , 1983 . [ 60 ] A . Kittur , L . Yu , T . Hope , J . Chan , H . Lifshitz - Assaf , K . Gilon , F . Ng , R . E . Kraut , and D . Shahaf . Scaling up analogical innovation with crowds and ai . PNAS , 2019 . [ 61 ] G . Knoblich , S . Ohlsson , H . Haider , and D . Rhenius . Constraint relaxation and chunk decomposition in insight problem solving . Journal of Experimental Psychology : Learning , Memory , and Cognition , 25 ( 6 ) : 1534 \u2013 1555 , 1999 . 00691 . [ 62 ] T . Kuribayashi , H . Ouchi , N . Inoue , P . Reisert , T . Miyoshi , J . Suzuki , and K . Inui . An empirical study of span representations in argumentation structure parsing . In Proceedings of the 57th Annual Meeting of the Association for Computa - tional Linguistics , pages 4691 \u2013 4698 , Florence , Italy , July 2019 . Association for Computational Linguistics . [ 63 ] D . B . Lenat . Cyc : a large - scale investment in knowledge infrastructure . In Communications of the ACM , 1995 . [ 64 ] J . Linsey , A . Markman , and K . Wood . Design by analogy : a study of the wordtree method for problem re - representation . JMD , 2012 . [ 65 ] J . Linsey , A . B . Markman , and K . L . Wood . WordTrees : A method for design - by - analogy . In Proceedings of the 2008 ASEE Annual Conference , 2008 . [ 66 ] Y . Luan , L . He , M . Ostendorf , and H . Hajishirzi . Multi - task identification of entities , relations , and coreferencefor scientific knowledge graph construction . In Proc . Conf . Empirical Methods Natural Language Process . ( EMNLP ) , 2018 . [ 67 ] M . Mackeprang , C . M\u00fcller - Birn , and M . T . Stauss . Discovering the Sweet Spot of Human - Computer Configurations : A Case Study in Information Extraction . Proc . ACM Hum . - Comput . Interact . , 3 ( CSCW ) : 195 : 1 \u2013 195 : 30 , Nov . 2019 . [ 68 ] N . Mahyar , D . V . Nguyen , M . Chan , J . Zheng , and S . P . Dow . The Civic Data Deluge : Understanding the Challenges of Analyzing Large - Scale Community Input . In Proceedings of the 2019 on Designing Interactive Systems Conference , DIS \u201919 , pages 1171 \u2013 1181 , New York , NY , USA , June 2019 . Association for Computing Machinery . 00012 . [ 69 ] T . W . Malone , J . V . Nickerson , R . J . Laubacher , L . H . Fisher , P . de Boer , Y . Han , and W . B . Towne . Putting the Pieces Back Together Again : Contest Webs for Large - Scale Problem Solving . In Proceedings of the 2017 ACM Conference on Computer Supported Cooperative Work and Social Computing , CSCW \u201917 , pages 1661 \u2013 1674 , New York , NY , USA , 2017 . ACM . [ 70 ] A . H . Marblestone , B . M . Zamft , Y . G . Maguire , M . G . Shapiro , T . R . Cybulski , J . I . Glaser , D . Amodei , P . B . Stranges , R . Kalhor , D . A . Dalrymple , D . Seo , E . Alon , M . M . Maharbiz , J . M . Carmena , J . M . Rabaey , E . S . Boyden , G . M . Church , and K . P . Kording . Physical Principles for Scalable Neural Recording . Frontiers in Computational Neuroscience , 7 , 2013 . arXiv : 1306 . 5709 . [ 71 ] D . Marcheggiani and I . Titov . Encoding Sentences with Graph Convolutional Networks for Semantic Role Labeling . 1 , 2017 . [ 72 ] A . B . Markman and J . Loewenstein . Structural comparison and consumer choice . Journal of Consumer Psychology , 2010 . [ 73 ] T . McCaffrey . Innovation Relies on the Obscure : A Key to Overcoming the Classic Problem of Functional Fixedness . Psychological Science , 23 ( 3 ) : 215 \u2013 218 , 2012 . 00117 . [ 74 ] R . Mihalcea and P . Tarau . Textrank : Bringing order into text . In EMNLP , 2004 . [ 75 ] G . A . Miller . Wordnet : a lexical database for english . Communications of the ACM , 1995 . [ 76 ] T . Mitchell , W . Cohen , E . Hruschka , P . Talukdar , B . Yang , J . Betteridge , A . Carl - son , B . Dalvi , M . Gardner , B . Kisiel , et al . Never - ending learning . Communica - tions of the ACM , 2018 . [ 77 ] A . Murray . Salvador luria and max delbr\u00fcck on random mutation and fluctuation tests . Genetics , 2016 . [ 78 ] L . Page , S . Brin , R . Motwani , and T . Winograd . The pagerank citation ranking : Bringing order to the web . Technical report , Stanford InfoLab , 1999 . [ 79 ] C . Pask . Mathematics and the science of analogies . American Journal of Physics , 2003 . CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Hope et al . [ 80 ] N . Pasquier , Y . Bastide , R . Taouil , and L . Lakhal . Discovering frequent closed itemsets for association rules . In Database Theory\u2014ICDT\u201999 . 1999 . [ 81 ] J . Pennington , R . Socher , and C . D . Manning . Glove : Global vectors for word representation . In EMNLP , 2014 . [ 82 ] M . E . Peters , S . Ruder , and N . A . Smith . To tune or not to tune ? adapting pretrained representations to diverse tasks . In RepL4NLP @ ACL , 2019 . [ 83 ] N . Qian . On the momentum term in gradient descent learning algorithms . Neural networks , 1999 . [ 84 ] N . Reimers and I . Gurevych . Sentence - bert : Sentence embeddings using siamese bert - networks . In EMNLP , 2019 . [ 85 ] M . A . Runco and S . Acar . Divergent thinking . In The Cambridge handbook of creativity , 2nd ed , pages 224 \u2013 254 . Cambridge University Press , New York , NY , US , 2019 . 00878 . [ 86 ] M . Sachan and E . Xing . Self - training for jointly learning to ask and answer questions . In NAACL - HLT , 2018 . [ 87 ] R . K . Sawyer . Explaining creativity : the science of human innovation . Oxford University Press , New York , 2nd edition , 2012 . [ 88 ] H . Sch\u00fctze , C . D . Manning , and P . Raghavan . Introduction to information retrieval , volume 39 . Cambridge University Press Cambridge , 2008 . [ 89 ] Y . Sebastian , E . - G . Siew , and S . O . Orimaye . Emerging approaches in literature - based discovery : techniques and performance review . The Knowledge Engineer - ing Review , 32 , Jan . 2017 . [ 90 ] P . Siangliulue , K . C . Arnold , K . Z . Gajos , and S . P . Dow . Toward Collaborative Ideation at Scale : Leveraging Ideas from Others to Generate More Creative and Diverse Ideas . In Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing , CSCW \u201915 , pages 937 \u2013 945 , New York , NY , USA , 2015 . ACM . [ 91 ] P . Siangliulue , J . Chan , S . P . Dow , and K . Z . Gajos . IdeaHound : Improving Large - scale Collaborative Ideation with Crowd - powered Real - time Semantic Modeling . In UIST\u201916 , 2016 . [ 92 ] U . N . Sio , K . Kotovsky , and J . Cagan . Fixation or inspiration ? A meta - analytic review of the role of examples on design processes . Design Studies , 39 : 70 \u2013 99 , July 2015 . [ 93 ] N . R . Smalheiser . Rediscovering Don Swanson : The Past , Present and Future of Literature - based Discovery . Journal of Data and Information Science , 2 ( 4 ) : 43 \u2013 64 , Dec . 2017 . [ 94 ] R . Sosa . Accretion theory of ideation : evaluation regimes for ideation stages . Design Science , 5 : e23 , 2019 . [ 95 ] D . R . Swanson and N . R . Smalheiser . Undiscovered Public Knowledge : a Ten - Year Update . In Proceedings of KDD \u201996 , page 4 , 1996 . [ 96 ] S . Vattam , B . Wiltgen , M . Helms , A . K . Goel , and J . Yen . DANE : Fostering Creativity in and through Biologically Inspired Design . In Design Creativity 2010 . 2011 . [ 97 ] T . B . Ward . What\u2019s old about new ideas . In The creative cognition approach , pages 157 \u2013 178 . 1995 . [ 98 ] L . Yu , A . Kittur , and R . E . Kraut . Searching for analogical ideas with crowds . In CHI , 2014 . [ 99 ] L . Yu , B . Kraut , and A . Kittur . Distributed analogical idea generation : innovating with crowds . In CHI\u201914 , 2014 . [ 100 ] L . Yu , R . E . Kraut , and A . Kittur . Distributed analogical idea generation with multiple constraints . In Proceedings of the 19th ACM Conference on Computer - Supported Cooperative Work & Social Computing . ACM , 2016 . [ 101 ] Y . Zhang , P . Qi , and C . D . Manning . Graph convolution over pruned dependency trees improves relation extraction . In EMNLP , 2018 . A TECHNICAL APPENDIX A . 1 Model Details \u2022 BiLSTM - CRF . A BiLSTM - CRF [ 53 ] neural network , a common baseline approach for NER tasks , enriched with semantic and syn - tactic input embeddings known to often boost performance [ 101 ] . We first pass the input sentence x = ( \ud835\udc65 1 , \ud835\udc65 2 , . . . , \ud835\udc65 \ud835\udc47 ) through an embedding module resulting in v 1 : \ud835\udc47 , v \ud835\udc56 \u2208 R \ud835\udc51 \ud835\udc52 , where \ud835\udc51 \ud835\udc52 is the embedded space dimension . We adopt the \u201cmulti - channel\u201d strat - egy as in [ 101 ] , concatenating input word embeddings ( pretrained GloVe vectors [ 81 ] ) with part - of - speech ( POS ) and NER embed - dings . We additionally add an embedding corresponding to the incoming dependency relation . The sequence of token embeddings is then processed with a BiLSTM layer to obtain contextualized word representations h ( 0 ) 1 : \ud835\udc47 , h \ud835\udc56 \u2208 R \ud835\udc51 \u210e , where \ud835\udc51 \u210e is the hidden state dimension . The outputs are fed into a linear layer \ud835\udc53 to obtain per - word tag scores \ud835\udc53 (cid:16) h ( \ud835\udc3f ) 1 (cid:17) , \ud835\udc53 (cid:16) h ( \ud835\udc3f ) 2 (cid:17) , . . . , \ud835\udc53 (cid:16) h ( \ud835\udc3f ) \ud835\udc47 (cid:17) . These are used as inputs to a conditional random field ( CRF ) model which maximizes the tag sequence log likelihood under a pairwise transition model between adjacent tags [ 5 ] . \u2022 Pooled Flair . A pre - trained language model [ 4 ] based on contex - tualized string embeddings , recently shown to outperform pow - erful approaches such as BERT [ 22 ] in NER and POS tagging tasks and achieve state - of - art results . Flair 9 uses a character - based language model pre - trained over large corpora , combined with a memory mechanism that dynamically aggregates embeddings of each unique string encountered during training and a pooling op - eration to distill a global word representation . We follow [ 4 ] and concatenate pre - trained GloVe vectors to token embeddings , add a CRF decoder , and freeze the language - model weights rather than fine - tune them [ 22 , 82 ] . \u2022 GCN . We also explore a model - enrichment approach with syn - tactic relational inputs . We employ a graph convolutional network ( GCN ) [ 58 ] over dependency - parse edges [ 101 ] . GCNs are known to be useful for propagating relational information and utilizing syntactic cues [ 71 , 101 ] . The linguistic cues are of special relevance and interest to us , as they are known to exist for purpose / mechanism mentions in texts [ 32 ] . We used a GCN with same token embeddings as in the BiLSTM - CRF baseline , with a BiLSTM layer for sequential context and a CRF decoder . For the graph fed into the GCN , we use a pre - computed syntactic edges with dependency parsing : For sentence x 1 : \ud835\udc47 , we convert its dependency tree to A \ud835\udc60\ud835\udc66\ud835\udc5b where A \ud835\udc60\ud835\udc66\ud835\udc5b\ud835\udc56\ud835\udc57 = 1 for any two tokens \ud835\udc65 \ud835\udc56 , \ud835\udc65 \ud835\udc57 connected by a dependency edge . We also add self - loops A \ud835\udc60\ud835\udc52\ud835\udc59\ud835\udc53 = \ud835\udc3c ( to propagate from h ( \ud835\udc59 \u2212 1 ) \ud835\udc56 to h ( \ud835\udc59 ) \ud835\udc56 [ 101 ] ) . Following [ 101 ] , we normalize activations to reduce bias toward high - degree nodes . For an \ud835\udc3f - layer GCN , denoting h ( \ud835\udc59 ) \ud835\udc56 \u2208 R \ud835\udc51 \u210e to be the \ud835\udc59 - th layer output node , the GCN operation can be written as \u210e ( \ud835\udc59 ) \ud835\udc56 = \ud835\udf0e (cid:169) (cid:173) (cid:171) \u2211\ufe01 \ud835\udc5f \u2208R \uf8ee\uf8ef\uf8ef \uf8ef\uf8ef \uf8f0 \ud835\udc5b \u2211\ufe01 \ud835\udc57 = 1 A \ud835\udc5f\ud835\udc56\ud835\udc57 W ( \ud835\udc59 ) \ud835\udc5f \u210e ( \ud835\udc59 \u2212 1 ) \ud835\udc57 / \ud835\udc51 \ud835\udc5f\ud835\udc56 + b ( \ud835\udc59 ) \ud835\udc5f \uf8f9\uf8fa\uf8fa \uf8fa\uf8fa \uf8fb (cid:170) (cid:174) (cid:172) 9 https : / / github . com / flairNLP / flair Scaling Creative Inspiration with Fine - Grained Functional Aspects of Ideas CHI \u201922 , April 29 - May 5 , 2022 , New Orleans , LA , USA Figure 11 : Schema of our GCN model . where R = { syn , self } , \ud835\udf0e is the ReLU activation function , W ( \ud835\udc59 ) \ud835\udc5f is a linear transformation , b ( \ud835\udc59 ) \ud835\udc5f is a bias term and \ud835\udc51 \ud835\udc5f\ud835\udc56 = (cid:205) \ud835\udc47\ud835\udc57 = 1 A \ud835\udc5f\ud835\udc56\ud835\udc57 is the degree of token \ud835\udc56 w . r . t \ud835\udc5f . In the GCN architecture , \ud835\udc3f layers corre - spond to propagating information across \ud835\udc3f - order neighborhoods . We set the contextualized word vectors h ( 0 ) 1 : \ud835\udc47 to be the input to the GCN , and use h ( \ud835\udc3f ) 1 : \ud835\udc47 as the output word representations . Figure 11 illustrates the GCN model architecture . Similarly to [ 71 ] , we do not model edge directions or dependency types in the GCN layers , to avoid over - parameterization in our data - scarce setting . We also attempted edge - wise gating [ 71 ] to mitigate noise propagation but did not see improvements , similarly to [ 101 ] . In our experiments , we followed standard GCN training proce - dures . Specifically , we base our model on the experimental setup detailed in [ 101 ] ( see also the authors\u2019 code which we adapt for our architecture , at https : / / github . com / qipeng / gcn - over - pruned - trees ) . We pre - process the data using the spaCy ( https : / / spacy . io ) package for tokenization , dependency parsing , and POS / NER - tagging . We use pretrained GloVE embeddings of dimension 300 , and NER , POS and dependency relation embeddings of size 30 each , giving a total embedding dimension \ud835\udc51 \ud835\udc52 = 390 . The bi - directional LSTM and GCN layers\u2019 hidden dimension is \ud835\udc51 \u210e = 200 , with 1 hidden layer for the LSTM . We find that the setting of 2 hidden layers works best for the GCNs . We also tried training with edge label information based on syntactic relations , but found this hurts performance . The training itself was carried out using SGD with gradient clipping ( cutoff 5 ) for 100 epochs , selecting the best model on the development set . For the Pooled - Flair approach [ 4 ] , we use the FLAIR framework [ 3 ] , with the settings obtaining SOTA results for CONLL - 2003 as in [ 4 ] ( see https : / / github . com / flairNLP / flair / blob / master / resources / docs / EXPERIMENTS . md ) . We also experiment with non - pooled embeddings and obtain similar results . We experiment with initial learning rate and batch size settings described in [ 4 ] , finding 0 . 1 and 32 to work best , respectively .",
    "carman2023structures": "STRUCTURAL BIOLOGY Structures of the free and capped ends of the actin filament Peter J . Carman 1 , 2 \u2020 , Kyle R . Barrie 1 , 2 \u2020 , Grzegorz Rebowski 1 , Roberto Dominguez 1 , 2 * The barbed and pointed ends of the actin filament ( F - actin ) are the sites of growth and shrinkage and the targets of capping proteins that block subunit exchange , including CapZ at the barbed end and tropomodulin at the pointed end . We describe cryo - electron microscopy structures of the free and capped ends of F - actin . Terminal subunits at the free barbed end adopt a \u201c flat \u201d F - actin conformation . CapZ binds with minor changes to the barbed end but with major changes to itself . By contrast , subunits at the free pointed end adopt a \u201c twisted \u201d monomeric actin ( G - actin ) conformation . Tropomodulin binding forces the second subunit into an F - actin conformation . The structures reveal how the ends differ from the middle in F - actin and how these differences control subunit addition , dissociation , capping , and interactions with end - binding proteins . A ctin is the most abundant cytosolic pro - tein in eukaryotes . The cellular pool of actin is approximately evenly divided between monomeric ( G - actin ) and fil - amentous ( F - actin ) forms ( 1 ) . G - actin consists of four subdomains which , based on their disposition in F - actin , form the outer ( subdomains 1 and 2 ) and inner ( subdomains 3 and 4 ) domains ( 2 ) ( Fig . 1A ) . Two clefts separate the outer and inner domains : the nucleotide - binding cleft and , opposite of it , a hydrophobic cleft that mediates interactions with numerous actin - binding proteins and withactin itselfinthe filament ( 3 ) . Adenosine triphosphate ( ATP ) hydrolysis by actin is one of several factors that control the G - to F - actin transition . Although G - actin is a slow adeno - sine triphosphatase ( ATPase ) , ATP hydrolysis is accelerated manyfold in F - actin as a result of a rotation of the outer domain relative to the inner domain , which produces a flatter subunit conformation and reorients side chains and water molecules in the catalytic site for hydro - lysis ( 4 , 5 ) . In this way , F - actin acquires polarity , i . e . , structural and kinetic asymmetry : ATP - actin binds to the barbed ( or + ) end of F - actin , followed by fast ATP hydrolysis , slow g - phosphate release , and dissociation of adeno - sinediphosphate ( ADP ) \u2013 actinfromthepointed ( or \u2212 ) end ( 1 , 6 ) . Unbound monomers undergo rapid ADP - to - ATP exchange and become ready to rejoin the barbed end . This process , known as filament treadmilling , is regulated by pro - teins that either accelerate subunit addition at the barbed end ( formins , Ena / VASP ) ( 7 \u2013 9 ) , accelerate subunit dissociation from the pointed end ( cyclase - associated protein and cofilin ) ( 10 , 11 ) , or cap the ends to stop subunit exchange . In muscle sarcomeres andnonmuscle cells , the main capping proteins are CapZ ( also known as capping protein or CP ) ( 12 ) at the barbed end and tropomodulin ( Tmod ) at the pointed end ( 13 ) . Recent cryo - electron microscopy ( cryo - EM ) helical reconstructions of F - actin in different nucleotide states [ ADP , ADP - Pi ( inorganic phosphate ) , and ATP analogs ] provide high - resolution insightsinto thestructure ofsubunits in the middle of F - actin and the conformational changesthatleadtoATPhydrolysisuponpolym - erization ( 5 , 14 , 15 ) . However , the structure of F - actin changes little as a function of the bound nucleotide or divalent cation ( Ca 2 + or Mg 2 + ) . Thus , although the rate constants for the re - actions of ATP - and ADP - actin addition and dissociation at the ends of F - actin were deter - mined ~ 40 years ago ( 1 , 6 ) , the structural bases of kinetic asymmetry at the ends of F - actin re - main poorly understood . On the basis of single - particle cryo - EM structures , we reveal how the barbed and pointed ends are conformationally different and primed for subunit addition and dissociation , respectively , and how CapZ and Tmod block subunit exchange . Results Cryo - EM structures of the filament ends We used single - particle cryo - EM to determine structures of the ends of F - actin . Actin alone tends to form long filaments in cryo - EM mi - crographs , which is ideal for helical recon - struction but produces few filament ends per micrograph . To increase the number of ends , we used an optimized mixture of CapZ , Tmod , tropomyosin , and actin that produced shorter filaments and thus , more ends per micrograph ( materials and methods ; fig . S1 , A to C ) . Filament end particles were harvested with the program Topaz , using a particle - picking pipeline based on custom - trained neural networks ( 16 ) . Starting from a subset of manually picked ends , iterations of Topaz training and particle picking resulted in a dataset of 1 , 102 , 793 particles . During data processing , this dataset diverged into three subclasses that produced three - dimensional reconstructionsofthefreepointedend , Tmod - capped pointed end , and CapZ - capped barbed end at resolutions of 2 . 84 , 3 . 26 , and 2 . 79 \u00c5 , respectively ( fig . S1D ) . Free barbed ends were not observed , consistent with CapZ \u2019 s subna - nomolar affinity for the barbed end and long half - time for dissociation ( 12 ) . Datasets were collected without CapZ to obtain the structure of the free barbed end . Attempts to shorten the filaments by shearing mostly failed because filaments reannealed fast before vitrification with or without Tmod . A total of 413 , 076 par - ticles were collected from two datasets , which , combined with a different data processing strategy , yielded a 3 . 30 - \u00c5 resolution structure of the free barbed end ( fig . S2 ) . A 2 . 26 - \u00c5 resolution structure of the middle of F - actin was obtained by helical reconstruction and used as reference for comparisons . This struc - ture issimilar to other structuresofthemiddle of F - actin , with a C a root mean square devia - tion of 0 . 35 \u00c5 with the highest - resolution structure ( 2 . 24 \u00c5 ) reported thus far ( 5 ) . Poly - merization started from Mg 2 + - ATP - actin , but subunits in all the structures described here contain bound ADP , including subunits at the barbed and pointed ends , indicating that hydrolysis and g - phosphate release occurred in the ~ 3 hours that elapsed before vitrifica - tion . The helical rise and twist of subunits is also similar in all the structures . Data collec - tion , refinement , and structure validations are shown in figs . S3 to S7 and table S1 . Tropomyosin density was only observed in a subset ( ~ 13 % ) of the Tmod - capped pointed end subclass , resulting in a 4 . 8 - \u00c5 resolution reconstruction that shows tropomyosin inter - acting with Tmod on both sides of F - actin ( fig . S8A ) . The median length of the filaments in micrographswas48 nm , which corresponds to ~ 8 . 7 actin protomers and ~ 1 . 25 tropomyosin molecules along the long - pitch helix ( fig . S1B ) . The low occupancy of tropomyosin is therefore consistent with the extremely low ( millimolar ) affinity of single tropomyosin molecules for F - actin ( 17 ) . That tropomyosin was only ob - served in conjunction with Tmod probably re - sults from the interaction of these two proteins at the pointed end , which mutually enhances their affinities for F - actin ( 13 , 18 ) . The low res - olution of the Tmod - tropomyosin map pre - cludes a detailed analysis of this interaction . The barbed end With one exception , the conformation of the two terminal subunits at the barbed end cor - responds to the classical flat conformation of subunits in the middle of the filament ( 4 , 5 , 14 , 15 ) ( Fig . 1 , A to C , movie S1 , and table S2 ) . Three main features characterize this conformation : ( i ) a ~ 20\u00b0 scissorlike rota - tion of the outer domain relative to the inner RESEARCH Carman et al . , Science 380 , 1287 \u2013 1292 ( 2023 ) 23 June 2023 1 of 6 1 Department of Physiology , Perelman School of Medicine , University of Pennsylvania , Philadelphia , PA , USA . 2 Biochemistry and Molecular Biophysics Graduate Group , Perelman School of Medicine , University of Pennsylvania , Philadelphia , PA , USA . * Corresponding author . Email : droberto @ pennmedicine . upenn . edu \u2020 These authors contributed equally to this work . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on O c t ob e r 26 , 2023 domain that flattens the actin protomer and changes the overall conformation of both the nucleotide and hydrophobic clefts ( Fig . 1A ) ; ( ii ) the D loop ( residues R39 to K50 ) adopts an extended conformation to insert into the hydrophobic cleft of the subunit immediately above ; and ( iii ) residues in the hydrophobic cleft change conformation to host the D loop of the subunit immediately below . Specifi - cally , the W loop ( I165 to P172 ) and the C terminus of actin ( P367 to F375 ) form a pincer - like structure that grasps the D loop of the subunit below . The latter change does not occur in barbed - end subunits because they do not interact with the D loop ( Fig . 1 , D and E ) . A cryo - EM structure of the CapZ hetero - dimer ( Fig . 2 , A and B ) at the barbed end was recently reported ( 19 ) , but CapZ was only partially visualized and at low resolution ( 6 to 7 \u00c5 ) . CapZ was also visualized at the barbed end of the actin - related protein 1 ( Arp1 ) minifilament of dynactin ( 20 ) . Yet , Arp1 shares only 51 % sequence identity with actin and its interaction with CapZ differs somewhat . Com - pared with these structures , the 2 . 79 - \u00c5 resolu - tion structure described here ( Fig . 2C ) displays large differences in the overall disposition and specific interactions of CapZ at the barbed end , and the relative orientation of actin ( or Arp1 ) subunits in the filament ( fig . S9A ) . As in the structure of the free barbed end , the terminal and penultimate subunits in the CapZ - capped structure have an F - actin conformation with a G - actin \u2013 like W loop ( table S2 ) . However , the short - pitch pair formed by these two subunits is splayed apart by ~ 1 . 5 \u00c5 compared with a short - pitch pair in the middle of F - actin ( Fig . 2D ) . Subunit splaying results from a ~ 2\u00b0 rotation of the terminal subunit away from the longitudi - nal axis of F - actin , which may be induced by CapZ binding but is likely favored by weakened contacts between the two long - pitch helices at the barbed end . Indeed , the interstrand plug ( Q263 to G273 ) provides the main contact be - tween the long - pitch helices ( or strands ) of F - actin ( 2 ) . In the middle of the filament , the plug from one subunit binds at the interface between two subunits of the opposite strand , a pattern that is broken at the barbed end where the plug of the terminal subunit contacts only one subunit of the opposite strand ( Fig . 2D ) . CapZ is a heterodimer of related a and b subunits ( Fig . 2A ) ( 12 ) . The heterodimer is described as having a mushroom - like shape , where the mushroom head binds the barbed end ( Fig . 2B ) . CapZ can adopt distinct filament - bound and unbound conformations . Alloste - ric regulators of CapZ , including CARMIL , CD2AP , and CKIP , have a capping - protein interaction ( CPI ) motif that binds around the Carman et al . , Science 380 , 1287 \u2013 1292 ( 2023 ) 23 June 2023 2 of 6 A B C D E Fig . 1 . Free barbed end . ( A ) Schematic representation of the G - to F - actin transition . A scissorlike ~ 20\u00b0 rotation of the outer domain ( subdomains 1 and 2 ) relative to the inner domain ( subdomains 3 and 4 ) produces a flatter conformation of subunits in F - actin . ( B ) Cryo - EM map of the free barbed end at 3 . 30 - \u00c5 resolution . The terminal and penultimate subunits are shown in two different shades of blue . ( C ) The terminal and penultimate subunits ( blue ) adopt the classical F - actin conformation of subunits in the middle of F - actin ( gray ) . Subunits were superimposed based on the inner domain ( surface representation ) to highlight differences in orientation of the outer domain ( ribbon representation ) . ( D ) Fit to the cryo - EM map of the W loop of the terminal subunit ( blue ) versus a subunit from the middle of the filament ( gray ) . ( E ) In the middle of F - actin , the W loop and the C terminus of actin form a pincer - like structure that grasps the D loop of the subunit below ( gray and cyan , respectively ) . This interaction is absent for subunits at the barbed end ( blue ) , resulting in the W loop adopting a G - actin conformation and the C terminus moving by 2 . 0 \u00c5 ( red arrows ) . Single - letter abbreviations for amino acids referenced throughout are as follows : R , Arg ; K , Lys ; I , Ile ; P , Pro ; F , Phe ; Q , Gln ; G , Gly ; E , Glu ; H , His ; L , Leu ; N , Asn ; A , Ala ; T , Thr RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on O c t ob e r 26 , 2023 mushroom stalk and stabilizes the unbound conformation to promote uncapping ( 21 , 22 ) . The filament - bound state is characterized by a flatter conformation of the mushroom head , resultingfromthedistalendsoftheheadmovingupwardstowardthebarbedend ( 19 , 20 ) . This conformation has also been observed in CapZ complexes with twinfilin - actin ( 23 ) and V - 1 ( 22 ) , a protein that sterically competes with CapZ binding to the barbed end ( fig . S9B ) . In the current structure , CapZ also adopts a flat con - formation , bending by ~ 15\u00b0 from its filament - unboundconformation ( Fig . 2EandmovieS1 ) . Flattening of the mushroom head leads to changes in two long , pseudo - symmetric helices of CapZ a ( E221 to R259 ) and CapZ b ( H209 to L243 ) that run along the barbed - end interface ( Fig . 2F ) . Both helices contain a p bulge , a deformation of the a - helical fold that often plays a functionalrole ( 24 ) . In the barbed end \u2013 bound structure , the bulge in the CapZ b helix travels along the helical axis toward the N terminus by about half a helical turn , which reorients the side chains of N229 , E230 , and I231 at the barbed - end interface . In CapZ a , the p bulge coincides with a kink in the helix , which becomes more pronounced in the filament - bound structure to reorient the side chains of D252 and K256 at the barbed - end interface . Other important elements of the interaction are the a tentacle ( CapZ a residues R260 to A286 ) and the b tentacle ( CapZ b residues R244 to N277 ) ( 25 ) , immediately C - terminal to the p - bulge helices ( Fig . 2G ) . The tentacles are well resolved in our struc - ture and their conformation differs from other filament - bound structures ( 19 , 20 ) , as well as unbound structures in which the tentacles are disordered ( 21 \u2013 23 ) . Both tentacles comprise amphipathic helices that project upwards and present their hydrophobic sides to the hydro - phobic clefts of the terminal ( b tentacle ) and penultimate ( a tentacle ) subunitsofthefilament . The pointed end Unexpectedly , the firsttwo subunits atthe free pointed end adopt aG - actinconformation and their D loops are disordered , as observed in Carman et al . , Science 380 , 1287 \u2013 1292 ( 2023 ) 23 June 2023 3 of 6 A C B F D E G Fig . 2 . CapZ - capped barbed end . ( A ) Domain diagram of CapZ subunits a and b . ( B ) Mushroom - like structure of CapZ and transition between filament - bound and unbound states . In filament - bound CapZ , the mushroom head flattens and the tentacles engage the hydrophobic clefts of barbed - end subunits . ( C ) Cryo - EM map of the CapZ - capped barbed end at 2 . 79 - \u00c5 resolution , showing the terminal and penultimate actin subunits in two shades of blue and CapZ a and CapZ b in pink and magenta , respectively . ( D ) Splaying of the short - pitch pair formed by the terminal and penultimate subunits ( blue ) as compared to a short - pitch pair from the middle of F - actin ( gray ) . The short - pitch pairs were superimposed based on the penultimate subunit to highlight the splaying of the terminal subunit , showing a maximum displacement of 1 . 5 \u00c5 resulting from a rotation of ~ 2\u00b0 ( red arrow ) . Splaying is likely favored by missing contacts of the interstrand plug at the barbed end compared to the middle of F - actin ( dashed red curves ) . ( E ) Comparison of the filament - bound ( pink , magenta ) and unbound ( 22 ) ( gray , PDB code 3AA7 ) structures of CapZ . The superimposition , based on CapZ a , highlights a ~ 15\u00b0 rotation of CapZ b ( red arrow ) that flattens the mushroom head . ( F ) The actin - binding surface of CapZ consists mostly of two antiparallel helices and the a and b tentacles . Comparison of the filament - bound ( pink , magenta ) and unbound ( gray ) structures shows that the helices contain p bulges that change conformation between these two states ( red arrows ) . The tentacles , which are disordered in the unbound structure , project out in the filament - bound structure to engage the two barbed - end subunits . ( G ) Close - up view of CapZ \u2019 s interaction with the barbed end , with the binding interface colored pink ( CapZ a ) and magenta ( CapZ b ) . CapZ residues participating in the interaction are shown and labeled in fig . S9C . RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on O c t ob e r 26 , 2023 most G - actin structures ( Fig . 3 , A and B , movie S1 , and table S2 ) . However , their hydrophobic clefts host the D loop of subunits immediately below , which imposes changes in the W loop and C terminus characteristic of the F - actin conformation . The opposite was observed at the barbed end , where the last two subunits have an F - actin conformation with G - actin - like features in their hydrophobic cleft ( Fig . 1 ) . We draw three conclusions from these observa - tions . First , theG - toF - actintransitioniscaused by interactions of subdomains 2 and 4 with the subunit immediately above along the long - pitch helix . These interactions impose an F - actin conformation to subunits at the barbed end , whereas their absence at the pointed end allows subunits to revert to the G - actin confor - mation . Second , subunit additionatthe pointed end is conformationally unfavorable , analo - gous to the way G - actin to G - actin interactions are rate limiting during nucleation ( 1 , 26 ) . Third , subunits at the pointed end appear con - formationally primed for dissociation . Contrary to CapZ , which capped all barbed ends , Tmod capped only ~ 20 % of pointed ends , resulting in a 3 . 26 - \u00c5 resolution structure ( Fig . 4 , A and B , and movie S1 ) . Tmod \u2019 s low pointed - end occupancy may be in part attri - buted to inefficient pointed - end capping in the absence of tropomyosin ( 13 , 18 ) , which binds weakly to short filaments ( 17 ) . Tmod comprises alternating tropomyosin and actin - binding sites ( Fig . 4A ) . Likely owing to thelack of interactions with tropomyosin , the two tropo - myosin binding sites are disordered in the Tmod - capped structure ( Fig . 4B ) , but appear ordered in the low - resolution Tmod - tropomyosin map ( fig . S8A ) . In the Tmod - capped structure , actin binding site 1 ( ABS1 , P58 to K99 ) has an extended conformation , interrupted by a single a helix , and interacts from N to C terminus with subdomains 4 , 2 , and 1 of the first actin subunit ( Fig . 4B and fig . S9D ) . Actin binding site 2 ( ABS2 , L161 to T348 ) consists mostly of a leucine - rich repeat ( LRR ) domain , which forms a wedge - like structure that inserts into a cleft at the interface between the first three Carman et al . , Science 380 , 1287 \u2013 1292 ( 2023 ) 23 June 2023 4 of 6 A B Fig . 3 . Free pointed end . ( A ) Cryo - EM map of the free pointed end at 2 . 84 - \u00c5 resolution , showing the first and second subunits in two different shades of green . The D loop of subunits at the pointed end is disordered . ( B ) The first ( left ) and second ( right ) subunits at the pointed end adopt a G - actin conformation , where the outer domain is rotated ~ 20\u00b0 ( red arrow ) compared with subunits in the middle of F - actin ( gray ) . Subunits were superimposed based on their inner domains ( surface representation ) to highlight differences in the orientation of their outer domains ( ribbon representation ) . C A B Fig . 4 . Tmod - capped pointed end . ( A ) Domain diagram of Tmod , comprising alternating tropomyosin ( TMBS1 and TMBS2 , disordered in the structure ) and actin ( ABS1 and ABS2 ) binding sites . ( B ) Cryo - EM map of the Tmod - capped pointed end at 3 . 26 - \u00c5 resolution , showing the first and second subunits in two shades of green and Tmod in orange . The view on the right is approximately down the longitudinal axis of F - actin and shows a ribbon representation of Tmod . ABS1 caps the first actin subunit , whereas ABS2 wedges into a cleft formed by the first three actin subunits . Most of the interactions with actin are mediated by the b - strand to a - helix loops of the LRR domain of ABS2 . Tmod residues involved in the interaction are shown and labeled in fig . S9D . ( C ) The first actin subunit ( left ) adopts a G - actin conformation with the outer domain rotated by ~ 20\u00b0 ( red arrow ) compared with subunits in the middle of F - actin ( gray ) . The second actin subunit ( right ) adopts an F - actin conformation . Subunits were superimposed based on their inner domains ( surface represen - tation ) to highlight differences in orientation of their outer domains ( ribbon representation ) . RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on O c t ob e r 26 , 2023 pointed - end subunits . The loops that connect the b strands to the a helices of the LRR do - main account for most of the interactions with actin subunits ( Fig . 4B ) . We had previ - ouslyproposedamodelofTmodatthepointedendthatwasbasedoncrystalstructuresofABS1andABS2boundtoG - actin ( 18 ) . This model accurately predicted Tmod \u2019 s binding mode , but not the conformation of actin sub - units at the pointed end , which is crucial to understand the details of this interaction . Thus , another reason for Tmod \u2019 s inefficient pointed - end binding appears to be that it re - quires a conformational change that forces the second subunit into an F - actin conforma - tion , whereas the first subunit remains in the G - actin conformation ( Fig . 4C and table S2 ) . The interaction of ABS2 with the three pointed - end subunits is likely responsible for the conformational change in the second subunit , because binding of ABS2 would be sterically hindered for the G - actin conformation . In a recently reported structure of the spectrin - actin complex ( 27 ) , the pointed end is capped by Tmod together with a previously unknown protein ( SH3BGRL2 ) . This complex also con - tains tropomyosin , whose position is con - strained by interactions with several other proteins . Likely because of these added inter - actions , the first actin subunitinthespectrin - actin complex has an F - actin conformation and tropomyosin is shifted azimuthally by ~ 2 . 0 \u00c5 compared with its apparent position in our Tmod - tropomyosin map ( fig . S8 ) . Conclusions We found that the ends of F - actin display conformational differences that correlate with kinetic differences in subunit addition and dis - sociation ( Fig . 5 and movie S1 ) . Notably , these structuraldifferences dependon theposition of subunits in the filament and not on the nucleotide state , which is responsible only for minor differences both in the monomer ( 28 , 29 ) and the filament ( 5 , 14 , 15 ) . Consistent with early studies of filament elongation and de - polymerization kinetics ( 1 , 6 ) , subunits at the barbed end have an F - actin conformation and are thus structurally primed for subunit addition . Yet , their exposed hydrophobic cleft have a G - actin \u2013 like conformation , which may explain why many proteins that bind G - actin can also stably or transiently bind the barbed end , including profilin ( 1 , 30 ) , gelsolin ( 31 ) , formins ( 7 ) , twinfilin ( 23 , 32 ) , andWH2domain \u2013 containing proteins ( 33 ) . CapZ undergoes major structural changes upon binding to the barbed end , which remains mostly unaffected by this interaction . An unexpected finding was that subunits at the pointed end adopt a G - actin conformation andthusappearprimedfordissociationandlesslikelytoacceptincomingsubunits . According to this model , slow pointed - end association can be viewed as conceptually analogous to the rate - limiting step during nucleation , i . e . , the formation of a polymerization seed from G - actin subunits ( 26 ) . Tmod forms a knot - like structure around the first three subunits , forc - ing the second subunit into an F - actin confor - mation and inhibiting subunit addition and dissociationbystericallyblockingthepointedend . Thenucleotidestatecontrolstheassociation and dissociation kinetics at the ends of F - actin ( 1 , 6 ) . The ADP - bound structures described here are directly relevant to the situation at the pointed end , which accumulates ADP - actin . At the barbed end , ATP hydrolysis likely occurs immediately upon subunit addition , followed by slow g - phosphate release . Yet , as in the fila - ment ( 5 , 14 , 15 ) , the structure of the barbed end is unlikely to change much as a function of the bound nucleotide . Therefore , we suggest that the structural differences observed in this work explain the asymmetric association of ATP - actin monomers to the barbed and pointed ends of F - actin , with ATP - bound monomers being more likely to undergo the G - to F - actin transition required for preferential binding to the barbed end than ADP - bound monomers ( Fig . 5 ) . In this way , this work provides a mech - anistic understanding of years of biochemical evidence and supports fundamental conclusions about the structure , kinetics , and capping in - teractions at the ends of the actin filament . REFERENCES AND NOTES 1 . T . D . Pollard , Handb . Exp . Pharmacol . 235 , 331 \u2013 347 ( 2017 ) . 2 . R . Dominguez , K . C . Holmes , Annu . Rev . Biophys . 40 , 169 \u2013 186 ( 2011 ) . 3 . R . Dominguez , Trends Biochem . Sci . 29 , 572 \u2013 578 ( 2004 ) . 4 . T . Oda , M . Iwasa , T . Aihara , Y . Ma\u00e9da , A . Narita , Nature 457 , 441 \u2013 445 ( 2009 ) . 5 . W . Oosterheert , B . U . Klink , A . Belyy , S . Pospich , S . Raunser , Nature 611 , 374 \u2013 379 ( 2022 ) . 6 . T . D . Pollard , J . Cell Biol . 103 , 2747 \u2013 2754 ( 1986 ) . 7 . D . R . Kovar , E . S . Harris , R . Mahaffy , H . N . Higgs , T . D . Pollard , Cell 124 , 423 \u2013 435 ( 2006 ) . 8 . S . D . Hansen , R . D . Mullins , J . Cell Biol . 191 , 571 \u2013 584 ( 2010 ) . 9 . D . Breitsprecher et al . , EMBO J . 30 , 456 \u2013 467 ( 2011 ) . 10 . T . Kotila et al . , Nat . Commun . 10 , 5320 ( 2019 ) . 11 . S . Shekhar , J . Chung , J . Kondev , J . Gelles , B . L . Goode , Nat . Commun . 10 , 5319 ( 2019 ) . 12 . M . Edwards et al . , Nat . Rev . Mol . Cell Biol . 15 , 677 \u2013 689 ( 2014 ) . 13 . V . M . Fowler , R . Dominguez , Biophys . J . 112 , 1742 \u2013 1760 ( 2017 ) . 14 . F . Merino et al . , Nat . Struct . Mol . Biol . 25 , 528 \u2013 537 ( 2018 ) . 15 . S . Z . Chou , T . D . Pollard , Proc . Natl . Acad . Sci . U . S . A . 116 , 4265 \u2013 4274 ( 2019 ) . 16 . T . Bepler et al . , Nat . Methods 16 , 1153 \u2013 1160 ( 2019 ) . 17 . W . M . Schmidt , W . Lehman , J . R . Moore , Cytoskeleton 72 , 292 \u2013 303 ( 2015 ) . 18 . J . N . Rao , Y . Madasu , R . Dominguez , Science 345 , 463 \u2013 467 ( 2014 ) . 19 . J . Funk et al . , Nat . Commun . 12 , 5329 ( 2021 ) . Carman et al . , Science 380 , 1287 \u2013 1292 ( 2023 ) 23 June 2023 5 of 6 Fig . 5 . Model of subunit association and dissociation at the free and capped ends of F - actin . Structures described here show that subunits in F - actin have different conformations depending on whether they are in the middle or at the ends of F - actin , and are different from those of G - actin . Notably , these are not nucleotide - dependent conformational differences , which are relatively minor in both G - and F - actin ( see text ) . The conformational differences at the ends of F - actin correlate with the association and dissociation constants of subunits at the ends of F - actin . Only the main pathway at equilibrium is depicted , with ATP - actin preferentially adding to the barbed end and ADP - actin dissociating from the pointed end [ see ( 1 ) for other possible reactions ] . Structural differences explain the asymmetric association of ATP - actin monomers to the barbed and pointed ends of F - actin , with ATP - bound monomers more likely to undergo the G - to F - actin transition required for preferential binding to the barbed end than ADP - bound monomers . CapZ and Tmod inhibit subunit exchange at the barbed and pointed ends , respectively , by structrual mechanisms revealed in this study . RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on O c t ob e r 26 , 2023 20 . L . Urnavicius et al . , Nature 554 , 202 \u2013 206 ( 2018 ) . 21 . M . Hernandez - Valladares et al . , Nat . Struct . Mol . Biol . 17 , 497 \u2013 503 ( 2010 ) . 22 . S . Takeda et al . , PLOS Biol . 8 , e1000416 ( 2010 ) . 23 . D . M . Mwangangi , E . Manser , R . C . Robinson , Sci . Adv . 7 , eabd5271 ( 2021 ) . 24 . J . P . Cartailler , H . Luecke , Structure 12 , 133 \u2013 144 ( 2004 ) . 25 . A . Narita , S . Takeda , A . Yamashita , Y . Ma\u00e9da , EMBO J . 25 , 5626 \u2013 5633 ( 2006 ) . 26 . D . Sept , J . A . McCammon , Biophys . J . 81 , 667 \u2013 674 ( 2001 ) . 27 . N . Li et al . , Cell 186 , 1912 \u2013 1929 . e18 ( 2023 ) . 28 . L . R . Otterbein , P . Graceffa , R . Dominguez , Science 293 , 708 \u2013 711 ( 2001 ) . 29 . P . Graceffa , R . Dominguez , J . Biol . Chem . 278 , 34172 \u2013 34180 ( 2003 ) . 30 . N . Courtemanche , T . D . Pollard , Biochemistry 52 , 6456 \u2013 6466 ( 2013 ) . 31 . S . Nag , M . Larsson , R . C . Robinson , L . D . Burtnick , Cytoskeleton 70 , 360 \u2013 384 ( 2013 ) . 32 . M . Hakala et al . , Nat . Cell Biol . 23 , 147 \u2013 159 ( 2021 ) . 33 . M . F . Carlier , C . Husson , L . Renault , D . Didry , Int . Rev . Cell Mol . Biol . 290 , 55 \u2013 85 ( 2011 ) . ACKNOWLEDGMENTS Data collection was performed at the Electron Microscopy Resource Lab ( EMRL ) and The Beckman Center for Cryo - Electron Microscopy , University of Pennsylvania ( Research Resource Identifier SCR _ 022375 ) . We thank M . Boczkowska for assistance with sample preparation and S . Steimle for assistance with data collection . Funding : This study was supported by National Institutes of Health ( NIH ) grant R01 GM073791 ( R . D . ) and NIH grant F31 HL156431 ( P . J . C . ) . Computational resources were supported by NIH grant S10 OD023592 . Author contributions : Conceptualization : P . J . C . , K . R . B . , and R . D . ; Methodology : P . J . C . , K . R . B . , and R . D . ; Investigation : P . J . C . , K . R . B . , and R . D . ; Resources : P . J . C . , K . R . B . , and G . R . ; Visualization : P . J . C . , K . R . B . , and R . D . ; Funding acquisition : P . J . C . and R . D . ; Project administration : R . D . ; Supervision : R . D . ; Writing \u2013 original draft : R . D . ; Writing \u2013 review and editing : P . J . C . , K . R . B . , and R . D . Competing interests : The authors declare no competing interests . Data and materials availability : Molecular models and cryo - EM density maps have been deposited with the following accession codes : CapZ - capped barbed end ( PDB ID 8F8Q , EMD - 28933 ) , free barbed end ( PDB ID 8F8R , EMD - 28934 ) , free pointed end ( PDB ID 8F8S , EMD - 28935 ) , Tmod - capped pointed end ( PDB ID 8F8T , EMD - 28936 ) , and F - actin in the ADP state ( PDB ID 8F8P , EMD - 28932 ) . All other data necessary to evaluate the claims in this paper are present in the text or supplementary materials . License information : Copyright \u00a9 2023 the authors , some rights reserved ; exclusive licensee American Association for the Advancement of Science . No claim to original US government works . https : / / www . science . org / about / science - licenses - journal - article - reuse SUPPLEMENTARY MATERIALS science . org / doi / 10 . 1126 / science . adg6812 Materials and Methods Figs . S1 to S9 Tables S1 and S2 References ( 34 \u2013 41 ) MDAR Reproducibility Checklist Movie S1 View / request a protocol for this paper from Bio - protocol . Submitted 13 January 2023 ; accepted 17 May 2023 Published online 25 May 2023 10 . 1126 / science . adg6812 Carman et al . , Science 380 , 1287 \u2013 1292 ( 2023 ) 23 June 2023 6 of 6 RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on O c t ob e r 26 , 2023 Use of this article is subject to the Terms of service Science ( ISSN 1095 - 9203 ) is published by the American Association for the Advancement of Science . 1200 New York Avenue NW , Washington , DC 20005 . The title Science is a registered trademark of AAAS . Copyright \u00a9 2023 The Authors , some rights reserved ; exclusive licensee American Association for the Advancement of Science . No claim to original U . S . Government Works Structures of the free and capped ends of the actin filament Peter J . Carman , Kyle R . Barrie , Grzegorz Rebowski , and Roberto Dominguez Science 380 ( 6651 ) , . DOI : 10 . 1126 / science . adg6812 Editor\u2019s summary Actin is a key protein in cellular motility , forming filaments that give dynamic structure to cellular appendages and serve as scaffold for motor proteins . The ends of these filaments are constantly extending and dissociating , and their structure has been difficult to capture . Carman et al . optimized a mixture of binding proteins so that they could visualize the ends in different forms . The structures revealed the internal conformations of the actin subunits at the ends , and the authors discuss how differences from the main filament structure facilitate differential addition or loss of subunits at the ends . They also found that a binding protein called tropomodulin stabilizes the pointed end and a protein called cap - Z binds at the barbed end . \u2014Michael A . Funk View the article online https : / / www . science . org / doi / 10 . 1126 / science . adg6812 Permissions https : / / www . science . org / help / reprints - and - permissions D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on O c t ob e r 26 , 2023",
    "jeeRelationalScaffoldingEnhances2019": "https : / / doi . org / 10 . 1177 / 0956797619864601 Psychological Science 2019 , Vol . 30 ( 9 ) 1287 \u2013 1302 \u00a9 The Author ( s ) 2019 Article reuse guidelines : sagepub . com / journals - permissions DOI : 10 . 1177 / 0956797619864601 www . psychologicalscience . org / PS ASSOCIATION FOR PSYCHOLOGICAL SCIENCE Research Article Science portrays a world governed by invisible entities and processes . The orbiting of electrons around the nucleus of an atom or of planets around the sun in our solar system cannot be directly observed . One is imper - ceptibly small , the other incredibly vast . Belief in the existence of these invisible systems is hardly a matter of common sense . Indeed , revolutionary scientific breakthroughs\u2014such as the heliocentric model of the solar system , evolution by natural selection , and the germ theory of disease\u2014have drastically restructured our understanding of the world ( Kuhn , 1962 ) . The advancement of these and other fundamental scientific ideas has involved creating , testing , and refining models that represent the hidden nature of reality ( Nersessian , 2010 ) . Science education aims to support model - based learning from a young age ( American Association for the Advancement of Science , 2009 ; National Research Council , 2012 ) . Yet students often emerge from school with incomplete and incorrect ideas about the way the world works ( Chi , Roscoe , Slotta , Roy , & Chase , 2012 ; McCloskey , 1983 ; Shtulman , 2017 ) . To realize the explanatory power of a scientific model , a student must grasp its relation to observable phenomena . This is straightforward when models depict familiar objects in idealized form or events happening more slowly or quickly than usual . It is far more challenging when models portray invisible entities and processes , whose relationship to observable phenomena is nonobvious and often counterintuitive . Like interpreting an analogy , making sense of a sci - entific model can involve relating seemingly disparate sets of information . Hence , also like analogy , the map - ping between observations and model may depend on 864601 PSSXXX10 . 1177 / 0956797619864601 Jee , Anggoro Relational Scaffolding and Scientific Models research - article 2019 Corresponding Author : Benjamin D . Jee , Worcester State University , Department of Psychology , Sullivan Academic Center , Suite S - 241 , 486 Chandler St . , Worcester , MA 01602 E - mail : bjee @ worcester . edu Relational Scaffolding Enhances Children\u2019s Understanding of Scientific Models Benjamin D . Jee 1 and Florencia K . Anggoro 2 1 Department of Psychology , Worcester State University , and 2 Department of Psychology , College of the Holy Cross Abstract Models are central to the practice and teaching of science . Yet people often fail to grasp how scientific models explain their observations of the world . Realizing the explanatory power of a model may require aligning its relational structure to that of the observable phenomena . In the present study , we tested whether relational scaffolding \u2014guided comparisons between observable and modeled events\u2014enhances children\u2019s understanding of scientific models . We tested relational scaffolding during instruction of third graders about the day / night cycle , a topic that involves relating Earth - based observations to a space - based model of Earth\u2019s rotation . Experiment 1 found that participants ( N = 108 ) learned more from instruction that incorporated relational scaffolding . Experiment 2 ( N = 99 ) found that guided comparison\u2014not merely viewing observable and modeled events\u2014is a critical component of relational scaffolding , especially for children with low initial knowledge . Relational scaffolding could be applied broadly to assist the many students who struggle with science . Keywords relational learning , structural alignment , comparison , relational scaffolding , science education , model - based instruction , open data , open materials Received 11 / 7 / 18 ; Revision accepted 4 / 26 / 19 1288 Jee , Anggoro a process of structural alignment , in which correspon - dences are established in accordance with a deeper , shared system of relations ( Gentner & Markman , 1997 ) . With scientific models , causal attribution is also crucial ; the model explains what is observed . Given these paral - lels , theory and research on analogical thinking could provide a basis for new approaches to model - based science instruction . Our aim was to test a support for model - based learning\u2014 relational scaffolding \u2014that involves guiding a student through systematic comparisons between observable phenomena and corresponding modeled events . Com - paring analogous cases brings common relational struc - ture into focus ( Gentner & Markman , 1997 ) . When cases lack surface similarity , explicit comparison supports analogical retrieval and mapping ( Goldwater & Gentner , 2015 ; Holyoak & Koh , 1987 ; Kurtz , Miao , & Gentner , 2001 ) . Thus , relational scaffolding should be most effec - tive when models involve entities and processes that bear little resemblance to observable phenomena . Applying relational scaffolding throughout a com - plex system of relations could illuminate the system\u2019s structural coherence\u2014its systematicity \u2014and discourage students from fragmented , incoherent explanations ( Gentner & Toupin , 1986 ) . Comprehensive visual com - parisons also reduce the burden of mentally represent - ing and aligning observed and modeled events during instruction ( Mayer & Moreno , 2003 ; Richland , Zur , & Holyoak , 2007 ) . Thus , relational scaffolding should especially benefit students with little prior knowledge , who often misinterpret scientific explanations and experi - ence cognitive overload during instruction ( Kalyuga , 2007 ; McNamara , Kintsch , Songer , & Kintsch , 1996 ; Vosniadou & Skopeliti , 2017 ) . The present research tested relational scaffolding with a fundamental and notoriously challenging topic : the day / night cycle . Children in the United States are expected to understand this topic between third grade and fifth grade ( National Research Council , 2012 ) . How - ever , many third and fifth graders are confused about the connection between Earth\u2019s motion and the day / night cycle , stating , for example , that the Earth orbits the sun in a single day ( Vosniadou & Brewer , 1994 ) . Even seventh - and eighth - grade U . S . students frequently endorse incorrect explanations ( Sadler et al . , 2010 ) . Grasping the scientific model of the day / night cycle involves reconciling observations from an Earth - based perspective\u2014including sunrise , midday , sunset , and midnight\u2014with models that adopt a large - scale space - based perspective of the Earth \u2013 sun system ( see Fig . 1 ) . In relational scaffolding , instructors support structural alignment by showing video footage of these two per - spectives simultaneously ( in a split - screen display ) and indicating the relevant temporal , spatial , and causal co r res pondences . Relational scaffolding is intended to supplement rather than replace model - based science instruction . We therefore tested it in a sequence of instruction that progressed from a familiar frame of reference\u2014an embodied simulation in which the student enacted Earth\u2019s rotation\u2014to a space - based simulation using a physical 3 - D model . Embodied simulation was intended to provide an intuitive , bridging analog for the external model ( Clement , 1993 ) and to clarify the alignment between apparent and actual motion through direct physical experience ( Kontra , Lyons , Fischer , & Beilock , 2015 ) . Relational scaffolding incorporated video footage of each simulation activity , recorded from a third - person / space - based perspective and a first - person / Earth - based perspective , as shown in Figure 2 . We tested the effects of relational scaffolding in two experiments . In Experiment 1 , third - grade participants Fig . 1 . Earth\u2019s rotation from a space - based perspective ( left ) and the apparent motion of the sun from an Earth - based perspective ( right ) . The images were obtained from nasa . gov ( left ) and stellarium . org ( right ) ; arrows have been added here to indicate the direc - tion of motion . Relational Scaffolding and Scientific Models 1289 were randomly assigned to four conditions . One group completed the sequence of instruction outlined above , progressing from an embodied to a 3 - D model - simulation activity with supplementary relational scaffolding ( relational - scaffolding condition ) . A second group com - pleted the same two simulations but without relational scaffolding ( no - relational - scaffolding condition ) . A third group repeated the 3 - D model simulation several times ( models - only condition ) . A fourth group received no instruction ( control ) . We interviewed students about the day / night cycle before and after instruction . We rea - soned that if relational scaffolding helps students relate their observations to a scientific model , participants in the relational - scaffolding condition should acquire the most knowledge . Moreover , if relational scaffolding is especially effective for students with little prior knowl - edge , any advantage of the relational - scaffolding condi - tion should be greater for low - knowledge participants . To assess changes in broader domain knowledge , we included items from a space - science - concept inventory ( Sadler et al . , 2010 ) at pretest and posttest . Given the role of language in conceptual learning ( e . g . , Gentner , Anggoro , & Klibanoff , 2011 ) and of spatial thinking in space - science understanding ( e . g . , Plummer , Kocareli , & Slagle , 2014 ) , we assessed verbal and spatial abilities prior to instruction and included these factors in the analyses of student learning outcomes . Spatial tests were also administered at a delayed posttest to explore potential changes resulting from the spatially intense instruction . Embodied Simulation 3 - D Model Simulation Third - Person / Space - Based Perspective First - Person / Earth - Based Perspective Fig . 2 . Screen captures of relational - scaffolding video footage from an embodied simulation ( top row ) and a 3 - D model simula - tion ( bottom row ) . In the embodied simulation , the student enacted Earth\u2019s rotation in relation to the yellow ball representing the sun ( top left ) . In the 3 - D model simulation , the student watched from a third - person perspective as the model of Earth rotated on its axis ( bottom left ) . Images on the right were captured by a GoPro camera mounted on the participant\u2019s head ( top right ) and on model Earth ( bottom right ) . Arrows indicate paths of motion or apparent motion . 1290 Jee , Anggoro Experiment 1 Method Participants . A total of 147 third - grade children were recruited from public elementary schools in Worcester , Massachusetts . Thirty - six ( 24 % ) left the study before the Session 6 posttest ( attrition is discussed further in the Supplemental Material available online ) . Three partici - pants were removed from the data set for scoring more than 2 standard deviations below age level on a vocabu - lary assessment ( the Peabody Picture Vocabulary Test ; see below ) . The remaining sample consisted of 108 third graders ( 62 female , 46 male ; age : M = 8 . 6 years , SD = 0 . 5 ) . This sample size provided power of greater than . 90 to detect a medium - to - large effect size ( Cohen\u2019s \u0192 2 \u2265 . 25 ) for our planned linear multiple regression analysis ( G * Power ; Faul , Erdfelder , Buchner , & Lang , 2009 ) . Each participant was randomly assigned to one of the four con - ditions ( 27 per condition ) . Knowledge measures . Day / night - cycle interview . The day / night - cycle inter - view was our primary measure of understanding . Interview - ers followed a script of questions and follow - up questions ( see Pretest Interview Script at osf . io / kszge ) . Several ques - tions required a verbal response . Others involved the use of 3 - D objects ( rubber balls representing Earth and the sun ) to model specific events in the day / night cycle . Items from the Astronomy and Space Science Concept Inventory ( ASSCI ) . The ASSCI measures students\u2019 mas - tery of astronomical concepts found in national standards ( Sadler et al . , 2010 ) . We selected nine items ( seven for kindergarten \u2013 Grade 4 and two for Grades 5 \u2013 8 ) broadly related to , but not covered in , our instruction . Further details on the instrument and selected items can be found in the Supplemental Material . Cognitive - ability measures . Verbal ability . The Peabody Picture Vocabulary Test , fourth edition ( Dunn & Dunn , 2007 ) , measures receptive vocabulary . Four pictures are presented on each trial . The test administrator says a word , and the participant must point to the picture to which it corresponds . Mental rotation . The spatial - relations subtest of the Primary Mental Abilities Test ( Thurstone & Thurstone , 1962 ) measures mental - rotation ability . On each trial , the participant must identify a rotated shape that forms a complete square with a second part . Perspective taking . The Perspective Taking Test for Children ( Frick , M\u00f6hring , & Newcombe , 2014 ) involves identifying from a set of four options the picture taken by a toy photographer within a simple scene . The objects in the scene and the photographer\u2019s position vary across trials . Instructional activities . Embodied simulation . The embodied - simulation activity followed a script based on the lesson plan for Kinesthetic Astronomy ( Morrow , 2000 ) . After orienting the partici - pant to his or her role as Earth ( see Fig . 2 , top left ) , the researcher guided the participant through a simulated 24 - hr day . The participant\u2019s first - person observations were recorded using a head - mounted GoPro camera ( see Fig . 2 , top right ) . A video recorder on a tripod captured the session from a third - person perspective . At the end of the activity , the participant performed one slow , careful rota - tion that supplied video footage for relational scaffolding . 3 - D model simulation . This activity involved a physi - cal 3 - D model with a sun and rotating Earth ( see Fig . 2 , bottom left ) . The simulation followed a detailed script similar in content and structure to the embodied simu - lation . The participant was prompted to look out from behind model Earth to adopt an Earth - based perspective for key events . Relational scaffolding . Relational Scaffolding 1 used video of the embodied simulation . The participant\u2019s third - and first - person videos were edited to create an approxi - mately 20 - s to 30 - s split - screen video of Earth\u2019s rotation as seen from each perspective . A trained researcher guided the participant through footage of midday , sunset , midnight , and sunrise , shown on a 13 - in . MacBook Pro computer ( see the Relational Scaffolding 1 script at osf . io / kszge ) . The researcher pointed at and between the videos to convey the correspondences , repeating key events several times ( for an excerpt , see Fig . 3 ) . Relational Scaffolding 2 used video from both the embodied and 3 - D model simulations . The footage was displayed on a 13 - in . MacBook Pro computer in a 2 \u00d7 2 matrix , as shown in Figure 4 . A trained researcher followed a script ( see the Relational Scaffolding 2 script at osf . io / kszge ) that highlighted corresponding per - spectives ( see Figs . 4a and 4b ) and the higher - order relations between the simulations ( see Figs . 4c and 4d ) . Procedure . Participants completed one study session per day , twice a week for 3 weeks at their school during after - school hours . The delayed posttest occurred about 6 to 7 weeks later . Participants met with the same researcher in a quiet room ( usually a classroom ) each day . Each ses - sion lasted about 20 to 30 min . The sessions were video - taped ( with parental permission ) . In Session 1 , participants completed the verbal - ability test and mental - rotation test . Session 2 consisted of the 1291 \u201c N o w l e t \u2019 s t ake a l ook a t t h e t h i ng s y o u d i d l a s t t i m e . Le t \u2019 s c o m p a r e h o w i t l ooke d f r o m t h e t w o c a m e r a s . O n e s h o w s t h e s un a nd h o w y o u s pun o r r o t a t e d a r o und w h e n y o u w e r e E a r t h . T h i s v i d e o w a s t ake n f r o m a c a m e r a o n t h e s i d e a nd y o u c a n s ee b o t h t h e s un a nd y o u a s E a r t h . \u201d \u201c N o w t h i s i s r ea ll y c oo l ! T h i s o t h e r v i d e o w e h a v e h e r e w a s t ake n f r o m t h e c a m e r a o n y o u r h ea d a nd s h o w s w h a t i t l ooke d li ke t o y o u a s E a r t h . R e m e m b e r , t h e r e d d o t o n y o u r c h e s t i s w h e r e W o r c e s t e r i s . W h a t t h e c a m e r a r e c o r d e d i s p r e tt y s i m il a r t o w h a t a p e r s o n w o u l d s ee i f t h e y w e r e l ook i ng up f r o m W o r c e s t e r . \u201d \u201c W h e n Ea r t h h a s r o t a t e d f a r e n o ugh a r o und , w e s t a r t t o s ee t h e s un a g a i n i n W o r c e s t e r . A p e r s o n i n W o r c e s t e r w o u l d s ee t h e s un a pp ea r i n t h e ea s t . W e c a ll t h a t s un r i s e . D o y o u s ee h o w t h e r o t a t i o n o f E a r t h m ake s i t l ook li ke t h e s un i s r i s i ng i n t h e ea s t ? \u201d [ R e p l a y e v e n t a t l ea s t o n c e ] \u201c W h e r e w e r e y o u l oo k i ng t o s ee t h e s un ? [ P a r t i c i p a n t a n s w e r s + f ee db a c k ] . S o , y o u w e r e l ook i ng d o w n y o u r ea s t w a r d a r m . Th e s un i s s t a r t i ng t o a pp ea r i n t h e ea s t . W h a t i s i t c a ll e d ? [ A n s w e r + f ee db a c k ] . W h y d oe s i t l ook li ke t h e s un i s a pp ea r i ng ? [ A n s w e r + f ee db a c k ] . S o , a s y o u c a n s ee h e r e , w h e n y o u s p i n f a r e n o ugh t o t h e ea s t , y o u c a n b e g i n t o s ee t h e s un a g a i n d o w n y o u r ea s t w a r d a r m . \u201d a b c d F i g . 3 . S c h e m a t i c d i a g r a m o f R e l a t i on a l S c a ff o l d i n g 1 . T h i s e x c e r p t i s a bou t s un r i s e . E a c h p a n e l s ho w s a s c r ee n s ho t f r o m v i d e o o f t h e t h i r d - p e r s on / s p a c e - b a s e d p e r s p e c t i ve ( on t h e l e f t ) a nd t h e f i r s t - p e r s on / E a r t h - b a s e d p e r s p e c t i v e ( on t h e r i g h t ) . T e x t f r o m t h e s e ss i on s c r i p t i s quo t e d a bo ve t h e i m a g e s . B o l d t e x t s i g n i f i e s w h e n t h e r e s e a r c h e r p o i n t e d t o on e o f t h e p e r s p e c t i ve s . I n t h e bo tt o m r o w , a rr o w s i nd i c a t e t h e p a r t i c i p a n t \u2019 s m o t i on ( t h i r d p e r s on ) a nd t h e m od e l s un \u2019 s a pp a r e n t m o t i on ( f i r s t p e r s on ) a s t h ey a pp e a r e d i n t h e v i d e o . A s i n g l e h a nd i nd i c a t e s a p o i n t i n g g e s t u r e f r o m t h e r e s e a r c h e r . T h e do tt e d a rr o w i n ( c ) i nd i c a t e s a s e qu e n t i a l g e s t u r e b e t w ee n t h e t w o v i d e o s . T h e t w o do tt e d a rr o w s i n ( d ) i nd i c a t e b a c k - a nd - f o r t h g e s t u r i n g b e t w ee n t h e v i d e o s . T h e f o r m a t f o r t h i s d i a g r a m i s b a s e d on t h e w o r k o f Y u a n , U tt a l , a nd G e n t n e r ( 2017 ) . 1292 \u201c T h e s e t w o v i d eo s a r e s i m il a r b e c a u s e t h e y s h o w w h a t t h i ng s m i gh t l ook li ke f r o m E a r t h . \u201d \u201c A nd t h e s e t w o v i d eo s a r e s i m il a r b e c a u s e t h e y s h o w w h a t t h i ng s m i gh t l o ok li ke f r o m o u t e r s p a c e . \u201d \u201c W h e n y o u w e r e Ea r t h , f r o m w h i c h d i r e c t i o n d i d i t l ook li ke t h e s un w a s a pp ea r i ng ? [ P a r t i c i p a n t a n s w e r s + f ee db a c k ] . F r o m w h i c h d i r e c t i o n d oe s i t l ook li ke t h e s un i s a pp ea r i ng f o r t h e m o d e l Ea r t h ? [ A n s w e r + f ee db a c k ] \u201d \u201c W h e n y o u w e r e t h e E a r t h , y o u r o t a t e d e a s t w a r d un t il y o u s t a r t e d t o s ee t h e s un . I n t h e m o d e l , t h e Ea r t h a l s o r o t a t e s e a s t w a r d un t il t h e s un l oo k s li k e i t a pp ea r s i n t h e ea s t f r o m W o r c e s t e r o n t h e m o d e l E a r t h . W h a t i s i t c a ll e d w h e n y o u r o t a t e d f a r e n o ugh ea s t t h a t y o u c o u l d s ee t h e s un ? [ P a r t i c i p a n t a n s w e r s + f ee db a c k ] \u201d a b c d F i g . 4 . S c h e m a t i c d i a g r a m o f R e l a t i on a l S c a ff o l d i n g 2 . T h i s e x c e r p t i s a bou t s un r i s e . E a c h p a n e l s ho w s a s c r ee n s ho t o f t h e 2 \u00d7 2 d i s p l a y f r o m v i d e o o f t h e e m bod i e d s i m u l a - t i on a nd 3 - D m od e l s i m u l a t i on o f t h e t h i r d - p e r s o n / s p a c e - b a s e d p e r s p e c t i ve ( on t h e l e f t ) a nd t h e f i r s t - p e r s on / E a r t h - b a s e d p e r s p e c t i ve ( on t h e r i g h t ) . T e x t f r o m t h e s e ss i on s c r i p t i s quo t e d a bo ve t h e i m a g e s . B o l d t e x t s i g n i f i e s w h e n t h e r e s e a r c h e r p o i n t e d t o a n eve n t f r o m on e o f t h e p e r s p e c t i ve s . S o li d a rr o w s i nd i c a t e t h e p a r t i c i p a n t \u2019 s m o t i on ( t h i r d p e r s on ) a nd t h e m od e l s un \u2019 s a pp a r e n t m o t i on ( f i r s t p e r s on ) a s t h ey a pp e a r e d i n t h e v i d e o . A s i n g l e h a nd i nd i c a t e s a p o i n t i n g g e s t u r e f r o m t h e r e s e a r c h e r . T h e do tt e d a rr o w s i n ( c ) a nd ( d ) i nd i c a t e s e qu e n t i a l g e s t u r e s b e t w ee n t h e t w o v i d e o s . T h e t w o do tt e d a rr o w s i n ( a ) a nd ( b ) i nd i c a t e b a c k - a nd - f o r t h g e s t u r i n g b e t w ee n t h e v i d e o s . T h e f o r m a t f o r t h i s d i a g r a m i s b a s e d on t h e w o r k o f Y u a n , U tt a l , a nd G e n t n e r ( 2017 ) . Relational Scaffolding and Scientific Models 1293 day / night - cycle - understanding interview ( pretest ) , ASSCI items , and the perspective - taking test . Sessions 3 to 5 varied among conditions . Participants in the relational - scaffolding condition completed the embod - ied simulation in Session 3 , Relational Scaffolding 1 in Session 4 , and both the 3 - D model simulation and Rela - tional Scaffolding 2 in Session 5 . Participants in the no - relational - scaffolding condition completed the embodied simulation in Session 3 and again in Session 4 , followed by the 3 - D model simulation in Session 5 . Participants in the models - only condition completed the 3 - D model simulation in Sessions 3 , 4 , and 5 . Partici - pants in the control condition did not complete instruc - tional sessions . They remained in a supervised classroom during this time . In all conditions , Session 6 consisted of the day / night - cycle - understanding interview ( posttest ) and ASSCI items . Session 7 included the day / night - cycle - understanding interview ( delayed posttest ) , the mental - rotation test , and the perspective - taking test . Equivalent forms of the mental - rotation test , perspective - taking test , and ASSCI items were administered in counterbalanced order across participants in each condition . Figure 5 provides an overview of the procedure . Coding the day / night - cycle interviews . We created a 27 - component coding rubric , building on prior measures of space - science understanding ( e . g . , Plummer et al . , 2014 ; Vosniadou & Brewer , 1994 ) . The components represent scientifically consistent ideas scored as correct / present versus incorrect / absent on the basis of the participant\u2019s interview responses . The internal consistency ( Cronbach\u2019s \u03b1 ) of the rubric was . 80 at pretest , . 88 at posttest , and . 86 Experiment 1 Condition Pretest Instruction Posttest Delayed posttest Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Control Verbal - Ability Test + Mental - Rotation Test Day / Night Cycle Interview + Concept - Inventory Items + Perspective - Taking Test No Instruction No Instruction No Instruction Day / Night Cycle Interview + Concept - Inventory Items Day / Night Cycle Interview + Mental - Rotation Test + Perspective - Taking Test Models Only 3 - D Model Simulation 3 - D Model Simulation 3 - D Model Simulation No Relational Scaffolding Embodied Simulation Embodied Simulation 3 - D Model Simulation Relational Scaffolding Embodied Simulation Relational Scaffolding 1 3 - D Model Simulation + Relational Scaffolding 2 Experiment 2 Condition Pretest Instruction Posttest Delayed posttest Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Sequential Scaffolding Verbal - Ability Test + Mental - Rotation Test Day / Night Cycle Interview + Concept - Inventory Items + Perspective - Taking Test Embodied Simulation Sequential Scaffolding 1 3 - D Model Simulation + Sequential Scaffolding 2 Day / Night Cycle Interview + Concept - Inventory Items Day / Night Cycle Interview + Mental - Rotation Test + Perspective - Taking Test Relational Scaffolding Embodied Simulation Relational Scaffolding 1 3 - D Model Simulation + Relational Scaffolding 2 Fig . 5 . Overview of procedure for Experiments 1 and 2 . 1294 Jee , Anggoro at delayed posttest . The 27 components and information about the coding process are provided in the Supplemen - tal Material . Results Table 1 shows the results for the demographic , cognitive - ability , and knowledge measures . There were no prein - struction differences between conditions on any variable\u2014gender : \u03c7 2 = 2 . 58 , p = . 46 ; all other measures : F s < 1 . 40 , p s > . 25 . We conducted a multiple regression analysis to predict posttest day / night - cycle understanding ( score on the 27 - component rubric ) from the participant\u2019s age ; gender ( male = 0 , female = 1 ) ; pretest verbal - ability , mental - rotation , and perspective - taking scores ; and pretest day / night understanding . Table 2 shows the zero - order cor - relations between variables . We also included a set of orthogonal contrasts in the regression model ( Davis , 2010 ) to test ( a ) the effect of receiving instruction versus none at all ( instructional conditions vs . control ) , ( b ) the effect of receiving embodied and 3 - D model simulation versus 3 - D model simulation alone ( no relational scaffolding and relational scaffolding vs . models only ) , and ( c ) the effect of receiving relational scaffolding specifically ( no rela - tional scaffolding vs . relational scaffolding ) . Because the no - relational - scaffolding - versus - relational - scaffolding con - trast directly tested the relational - scaffolding effect , we crossed this factor with pretest understanding to test the hypothesized Relational Scaffolding \u00d7 Prior Knowledge interaction ( West , Aiken , & Krull , 1996 ) . We included this interaction in the regression along with two others that crossed no relational scaffolding versus relational scaffold - ing with pretest mental - rotation and perspective - taking scores . All predictors were mean centered for the analysis . The regression model accounted for 53 % of the vari - ance in posttest understanding , F ( 12 , 95 ) = 9 . 02 , standard error of the estimate ( SEE ) = 3 . 98 , p < . 0001 . Figure 6 conveys the main findings of the orthogonal contrasts . Participants in the instructional conditions learned more than control participants ( \u03b2 = 0 . 55 ; 95 % confidence inter - val , or CI = [ 0 . 40 , 0 . 69 ] , p < . 0001 , partial r 2 = . 38 ) . Par - ticipants who received embodied simulation ( no relational scaffolding and relational scaffolding ) learned about as much as models - only participants ( \u03b2 = 0 . 09 , 95 % CI = [ \u2013 0 . 05 , 0 . 23 ] , p = . 20 , partial r 2 = . 02 ) . Impor - tantly , participants in the relational - scaffolding condition acquired the greatest understanding , significantly more than no - relational - scaffolding participants ( \u03b2 = 0 . 24 , 95 % CI = [ 0 . 09 , 0 . 38 ] , p = . 002 , partial r 2 = . 10 ) . We confirmed in a post hoc analysis\u2014restructuring the orthogonal contrasts\u2014that relational - scaffolding participants also Table 1 . Means for Demographic and Cognitive Variables for Experiment 1 Variable type and variable Condition Overall Control Models only No relational scaffolding Relational scaffolding Demographics Gender ( female : male ) 14 : 13 18 : 9 13 : 14 17 : 10 62 : 46 Age 8 . 7 ( 0 . 6 ) 8 . 6 ( 0 . 5 ) 8 . 6 ( 0 . 4 ) 8 . 7 ( 0 . 5 ) 8 . 6 ( 0 . 5 ) Pretest and posttest n 27 27 27 27 108 Delayed posttest n 19 19 24 21 83 Pretest Verbal ability ( PPVT standard score ) 100 . 4 ( 13 . 8 ) 100 . 8 ( 16 . 3 ) 99 . 3 ( 13 . 5 ) 99 . 5 ( 13 . 9 ) 100 . 0 ( 14 . 2 ) Mental rotation ( PMA - SR score ) 9 . 8 ( 2 . 4 ) 9 . 2 ( 3 . 0 ) 8 . 8 ( 2 . 7 ) 8 . 3 ( 3 . 3 ) 9 . 0 ( 2 . 9 ) Perspective taking ( PTT - C score ) 10 . 2 ( 4 . 0 ) 8 . 8 ( 3 . 8 ) 9 . 1 ( 3 . 5 ) 8 . 2 ( 3 . 1 ) 9 . 1 ( 3 . 6 ) Space - science concept ( ASSCI score ) 2 . 7 ( 1 . 5 ) 2 . 7 ( 1 . 1 ) 2 . 9 ( 1 . 5 ) 3 . 0 ( 1 . 3 ) 2 . 8 ( 1 . 3 ) Day / night understanding score 8 . 5 ( 4 . 1 ) 7 . 7 ( 4 . 4 ) 7 . 3 ( 3 . 8 ) 7 . 2 ( 3 . 6 ) 7 . 7 ( 4 . 0 ) Posttest Space - science concept ( ASSCI score ; n = 98 ) 3 . 0 ( 1 . 7 ) 3 . 0 ( 1 . 2 ) 2 . 9 ( 1 . 2 ) 2 . 5 ( 1 . 3 ) 2 . 8 ( 1 . 4 ) Day / night understanding score 7 . 8 ( 4 . 0 ) 13 . 3 ( 5 . 2 ) 12 . 3 ( 4 . 7 ) 15 . 8 ( 4 . 9 ) 12 . 3 ( 5 . 5 ) Delayed posttest Mental rotation ( PMA - SR score ; n = 77 ) 9 . 8 ( 2 . 8 ) 8 . 9 ( 2 . 4 ) 9 . 7 ( 2 . 5 ) 10 . 0 ( 3 . 3 ) 9 . 6 ( 2 . 8 ) Perspective taking ( PTT - C score ; n = 81 ) 12 . 0 ( 4 . 0 ) 10 . 0 ( 3 . 3 ) 11 . 3 ( 4 . 7 ) 9 . 9 ( 4 . 0 ) 10 . 8 ( 4 . 1 ) Day / night understanding score 9 . 1 ( 4 . 8 ) 12 . 4 ( 4 . 9 ) 11 . 1 ( 5 . 0 ) 13 . 5 ( 4 . 6 ) 11 . 5 ( 5 . 0 ) Delay interval ( weeks ) 7 . 1 ( 2 . 8 ) 6 . 3 ( 2 . 2 ) 6 . 8 ( 2 . 5 ) 6 . 6 ( 2 . 4 ) 6 . 7 ( 2 . 5 ) Note : Standard deviations are given in parentheses . PPVT = Peabody Picture Vocabulary Test ; PMA - SR = spatial - relations subtest of the Primary Mental Abilities Test ; PTT - C = Perspective Taking Test for Children ; ASSCI = Astronomy and Space Science Concept Inventory . Relational Scaffolding and Scientific Models 1295 achieved higher understanding than models - only par - ticipants ( \u03b2 = 0 . 20 , 95 % CI = [ 0 . 06 , 0 . 34 ] , p = . 009 , partial r 2 = . 07 ) . Pretest day / night understanding was also a significant predictor in the model ( \u03b2 = 0 . 39 , 95 % CI = [ 0 . 22 , 0 . 55 ] , p < . 0001 , partial r 2 = . 19 ) ; however , no other factor or interaction was a significant predictor , including Relational Scaffolding \u00d7 Prior Knowledge ( \u03b2 s = \u22120 . 03 to 0 . 11 , p s > . 20 ) . Participants\u2019 broader space - science knowledge changed little ( see ASSCI means in Table 1 ) . A 2 ( test session ) \u00d7 4 ( condition ) mixed analysis of variance ( ANOVA ) found no significant difference in mean scores on the ASSCI items from pretest to posttest , F ( 1 , 94 ) < 1 , p = . 87 , \u03b7 p \u00b2 < . 01 , nor was there a difference between conditions , F ( 3 , 94 ) < 1 , p = . 97 , \u03b7 p \u00b2 < . 01 , or a Test Session \u00d7 Condition interaction , F ( 1 , 94 ) = 1 . 0 , p = . 32 , \u03b7 p \u00b2 = . 03 . Eighty - three participants ( 77 % ) completed the delayed posttest . Day / night - cycle - understanding scores were analyzed in a multiple regression analysis . To pre - serve statistical power , we included as predictors only those variables that were significant in the posttest analysis : pretest understanding , the instructional - conditions - versus - control contrast , and the no - relational - scaffolding - versus - relational - scaffolding contrast . ( An analysis that included all of the posttest predictors yielded the same pattern of results . ) Delay interval ( weeks between posttest and delayed posttest ) was added as a predictor . All variables were mean centered for the analysis . The regression model accounted for 35 % of the vari - ance in delayed - posttest day / night - cycle - understanding scores , F ( 4 , 78 ) = 5 . 89 , SEE = 10 . 64 , p < . 0001 . The results mirrored those at posttest . Participants in the instructional conditions maintained higher understand - ing than those in the control condition ( \u03b2 = 0 . 30 , 95 % CI = [ 0 . 12 , 0 . 49 ] , p < . 01 , partial r 2 = . 12 ) . Participants in the relational - scaffolding condition had higher understanding than no - relational - scaffolding partici - pants ( see Table 1 ) ; however , this difference diminished and was not statistically significant ( \u03b2 = 0 . 18 , 95 % CI = [ \u2013 0 . 01 , 0 . 36 ] , p = . 06 , partial r 2 = . 05 ) . Pretest under - standing was a significant predictor ( \u03b2 = 0 . 52 , 95 % CI = [ 0 . 33 , 0 . 71 ] , p < . 0001 , partial r 2 = . 28 ) . Delay inter - val was not significant ( \u03b2 = \u22120 . 13 , 95 % CI = [ \u22120 . 32 , 0 . 06 ] , p = . 18 , partial r 2 = . 02 ) , although understanding scores decreased with longer intervals . We also analyzed participants\u2019 spatial - test perfor - mance before and after instruction for both Experiments 1 and 2 . The details are provided in the Supplemental Material . In short , we found no change in mental - rotation performance in either experiment . Perspective - taking scores increased significantly in both experiments ; however , the control and instructional conditions showed about the same level of improvement . Table 2 . Pearson Correlations Between Experiment 1 Variables Variable 2 3 4 5 6 7 8 9 10 11 12 1 . Gender \u2212 . 01 . 10 \u2212 . 06 . 10 \u2212 . 02 . 20 * \u2212 . 07 . 18 . 05 \u2212 . 05 . 05 2 . Age \u2212 . 43 * * . 02 \u2212 . 08 \u2212 . 05 \u2212 . 16 \u2212 . 07 \u2212 . 13 . 04 \u2212 . 03 \u2212 . 11 3 . Verbal ability . 23 * . 24 * . 11 . 33 * * . 29 * * . 27 * * . 21 . 19 . 27 * 4 . Pretest mental rotation . 30 * * . 13 . 14 . 17 . 06 . 30 * * . 24 * . 13 5 . Pretest perspective taking . 17 . 38 * * . 12 . 18 . 21 . 71 * * . 13 6 . Pretest space - science concept . 07 . 32 * * . 21 * . 11 . 22 * . 07 7 . Pretest day / night - understanding . 18 . 40 * * . 09 . 32 * * . 46 * * 8 . Posttest space - science concept ( n = 98 ) . 12 \u2212 . 16 . 23 * . 05 9 . Posttest day / night - understanding . 21 . 21 . 64 * * 10 . Delayed posttest mental rotation ( n = 77 ) . 24 * . 16 11 . Delayed perspective taking ( n = 81 ) . 17 12 . Delayed posttest day / night - understanding ( n = 83 ) Note : Gender was coded 0 for male and 1 for female . The fourth edition of the Peabody Picture Vocabulary Test ( PPVT ) measured verbal ability , the spatial - relations subtest of the Primary Mental Abilities Test ( PMA - SR ) measured mental rotation , the Perspective Taking Test for Children ( PTT - C ) measured perspective taking , and the Astronomy and Space Science Concept Inventory ( ASSCI ) measured children\u2019s space - science concept . * p < . 05 . * * p < . 01 . 1296 Jee , Anggoro Discussion Participants in the instructional conditions greatly increased their understanding of the day / night cycle ( although not of space - science concepts more broadly ) . Participants who received relational scaffolding gained the most knowledge . These effects were somewhat attenuated by the delayed posttest ( 6 \u2013 7 weeks after instruction ) , but the general pattern remained . The results are consistent with our hypothesis that systematic comparison of observable and modeled events facilitates understanding of scientific models . Yet further experimentation is required to separate the effects of comparison from other aspects of the relational - scaffolding condition that may enhance learning . Nota - bly , only the relational - scaffolding condition included videos of observable and modeled events . Viewing this footage could reduce extraneous cognitive load ( Sweller , 1994 ) , enabling a student to devote limited mental resources to sense - making processes ( Mayer & Moreno , 2003 ) . It is precisely for topics that require attention to relations within a system ( high element interactivity ) that a reduction in extraneous load should be especially beneficial ( Sweller , 1994 ) . Experiment 2 Experiment 2 teased apart guided comparison from the viewing of video footage . We compared the relational - scaffolding condition with a condition in which videos of observable and modeled events were presented sequen - tially , without explicit comparison . If guided comparison is integral to the relational - scaffolding effect , then the relational - scaffolding condition should produce greater understanding than the sequential - scaffolding condition . Zeroing in on the role of guided comparison also permits a more direct test of the hypothesis that comprehensive , guided comparisons are especially helpful for students with little prior understanding . If so , any advantage of the relational - scaffolding condition should be most pro - nounced among lower - knowledge students . In Experiment 2 , a new group of third - grade partici - pants received day / night - cycle instruction , supple - mented with either relational scaffolding or sequential scaffolding , in which videos of observable ( Earth - based ) and modeled ( space - based ) events were presented one after the other . In Experiment 2 , we also varied whether the scaffolding involved video of the participant\u2019s own embodied simulation ( self footage ) or a research assis - tant\u2019s simulation ( stock footage ) . If relational scaffolding requires personalized footage , it would be challenging to implement in formal educational settings . The footage variable was crossed with the scaffolding manipulation to create four experimental conditions . Method Participants . A total of 128 third - grade children were recruited at elementary schools in Worcester , Massachu - setts . Twenty - eight children ( 22 % ) dropped out before 0 3 6 9 12 15 18 21 24 27 D a y / N i gh t - C y c l e - U nd e r s t a nd i ng S c o r e Control Models Only Relational Scaffolding No Relational Scaffolding Pretest Posttest Pretest Posttest Pretest Posttest Pretest Posttest Fig . 6 . Pretest and posttest understanding scores for the day / night - cycle interview , separately for each of the four experimental conditions . Individual participant scores are shown in gray ; condition means are in orange . Error bars show standard errors . Individual scores were jittered to produce separation on the x - axis . Relational Scaffolding and Scientific Models 1297 the posttest . One child was removed from the data set because the child\u2019s Peabody Picture Vocabulary Test score was more than 2 standard deviations below age level . Our remaining sample included 99 third graders ( 59 female , 40 male ; age : M = 8 . 6 years , SD = 0 . 4 ) . This sample size provided power of greater than . 90 to detect a medium to large effect ( Cohen\u2019s \u0192 2 \u2265 . 25 ) for our planned analysis ( Faul et al . , 2009 ) . Each participant was randomly assigned to one of the four conditions ( 25 per condition ; the sequential - scaffolding / self - footage condi - tion had 24 ) . Materials . Experiment 2 used the measures and instruc - tional activities from Experiment 1 and two new sequential - scaffolding activities . Sequential Scaffolding 1 used footage of the embodied simulation , either self or stock . The first - person perspective was shown first . The participant was prompted to attribute the sun\u2019s apparent motion to Earth\u2019s rotation ( third - person perspective ) , although the sun and Earth were not shown simultaneously . The third - person perspective was shown second and likewise referenced the first - person perspective . Sequential Scaffolding 2 pre - sented video of the 3 - D model simulation , first from an Earth - based perspective and then from a space - based perspective . The activity used a similar script to that for Sequential Scaffolding 1 . When stock footage was shown , the script adopted a third - person perspective ( e . g . , \u201cRemember what they recorded from the camera on their head\u201d ) . Scripts for Sequential Scaffolding 1 and Sequen - tial Scaffolding 2 can be accessed at osf . io / kszge . Procedure . Participants completed the procedure at their school during after - school hours . They completed one session per day , twice a week for 3 weeks . The delayed posttest was about 4 weeks later ( sooner than in Experiment 1 to accommodate a school break ) . Sessions 1 and 2 ( pretest ) , 6 ( posttest ) , and 7 ( delayed posttest ) were the same as in Experiment 1 . The instructional ses - sions ( Sessions 3 \u2013 5 ) for the relational - scaffolding condi - tion were equivalent to those for the relational - scaffolding condition from Experiment 1 . Participants in the sequential - scaffolding condition completed the embodied simulation in Session 3 , Sequential Scaffolding 1 in Session 4 , and both the 3 - D model simulation and Sequential Scaffold - ing 2 in Session 5 . Figure 5 provides an overview of the procedure . To code the day / night - cycle interviews , we used the 27 - component rubric from Experiment 1 . The internal consistency ( Cronbach\u2019s \u03b1 ) of the instrument was . 81 at pretest , . 85 at posttest , and . 83 at delayed posttest . Results Table 3 shows the results for the demographic and other variables . There were no preinstruction differences between conditions in gender composition , \u03c7 2 = 0 . 89 , p = . 35 , or any other variable , t s < 1 . 25 , p s > . 20 . We conducted a multiple regression analysis to pre - dict posttest understanding from age ; gender ( male = 0 , female = 1 ) ; pretest verbal - ability , mental - rotation , and perspective - taking scores ; pretest understanding ; video condition ( stock = 0 , self = 1 ) ; and comparison condition ( sequential scaffolding = 0 , relational scaf - folding = 1 ) . In this model , comparison condition directly tested the effect of guided comparison . We crossed this factor with pretest understanding to test the hypothesized Guided Comparison \u00d7 Prior Knowl - edge interaction . We also crossed comparison condition with pretest mental - rotation and perspective - taking scores to explore possible interactions with spatial abil - ity . Predictors were mean centered for the analysis . Table 4 shows the zero - order correlations between variables . The regression model accounted for 31 % of the vari - ance in posttest day / night - cycle understanding , F ( 11 , 87 ) = 3 . 52 , SEE = 4 . 79 , p < . 001 . The main finding was a significant Comparison Condition \u00d7 Pretest Under - standing interaction ( \u03b2 = \u22120 . 41 , 95 % CI = [ \u22120 . 73 , \u2013 0 . 09 ] , p = . 01 , partial r 2 = . 07 ) . Figure 7 ( left panel ) shows mean posttest day / night - cycle understanding scores for relational - scaffolding and sequential - scaffolding partici - pants who scored below and above the median on the pretest ( i . e . , low initial knowledge vs . high initial knowledge ) . For low - knowledge ( but not high - knowl - edge ) participants , there was a clear advantage of rela - tional scaffolding over sequential scaffolding . Neither manipulated variable\u2014video condition ( \u03b2 = \u22120 . 07 ) or comparison condition ( \u03b2 = 0 . 07 ) \u2014was itself a signifi - cant predictor ( p s > . 45 ) . Pretest understanding was significant ( \u03b2 = 0 . 72 , 95 % CI = [ 0 . 40 , 1 . 00 ] , p < . 0001 , partial r 2 = . 19 ) ; however , no other variable was signifi - cant ( \u03b2 s = 0 . 01 \u2013 0 . 12 , p s > . 25 ) , nor was there another interaction ( \u03b2 s = 0 . 01 , p s = . 96 ) . Participants\u2019 scores on the ASSCI increased slightly , by about 0 . 5 items ( see Table 2 ) . A 2 ( test session ) \u00d7 2 ( comparison condition ) \u00d7 2 ( video condition ) mixed ANOVA revealed that this increase was significant , F ( 1 , 90 ) = 11 . 96 , p < . 001 , \u03b7 p \u00b2 = . 12 , but there was no effect of comparison condition or video condition ( F s < 1 , p s > . 60 , \u03b7 p \u00b2s < . 01 ) and no interactions ( F s < 1 . 95 , p s > . 15 , \u03b7 p \u00b2s < . 03 ) . Seventy participants ( 71 % ) completed the delayed posttest . An analysis of spatial - test performance before and after instruction can be found in the Supplemental Material . We analyzed delayed - posttest day / night under - standing in a multiple regression analysis . To preserve power , we included as predictors only those variables involved in the significant Pretest Understanding \u00d7 Comparison Condition interaction from posttest : pretest 1298 Jee , Anggoro understanding , comparison condition , and the interac - tion term . We also added delay interval ( in weeks ) as a predictor . All variables were mean centered for the analysis . The regression model accounted for 23 % of the vari - ance in delayed - posttest day / night - cycle understanding , F ( 4 , 65 ) = 4 . 94 , SEE = 4 . 61 , p < . 01 . The Pretest Under - standing \u00d7 Comparison Condition interaction was not significant ( \u03b2 = \u22120 . 32 , 95 % CI = [ \u22120 . 67 , 0 . 03 ] , p = . 07 , partial r 2 = . 05 ) . Instead , comparison condition emerged as a significant predictor ( \u03b2 = 0 . 27 , 95 % CI = [ 0 . 06 , 0 . 49 ] , p < . 05 , partial r 2 = . 09 ) . Figure 7 ( right panel ) shows that mean delayed - posttest understanding was higher overall in the relational - scaffolding condition , although this relational - scaffolding advantage was especially pro - nounced among low - knowledge participants . Pretest understanding was also a significant predictor ( \u03b2 = 0 . 60 , 95 % CI = [ 0 . 24 , 0 . 95 ] , p < . 01 , partial r 2 = . 15 ) , but delay interval was not ( \u03b2 = \u22120 . 13 , 95 % CI = [ \u22120 . 35 , 0 . 09 ] , p = . 25 ) . Discussion Experiment 2 found that participants with relatively low initial knowledge benefited from explicit , guided com - parisons between videos of Earth - and space - based perspectives\u2014the relational - scaffolding condition . Viewing the same videos without comparison\u2014the sequential - scaffolding condition\u2014was inferior . This relational - scaffolding advantage emerged across levels of prior knowledge by the delayed posttest , although low - knowledge participants remained the primary beneficiaries . The second manipulated variable\u2014whether scaffold - ing involved the participant\u2019s personal observations ( self footage ) versus another person\u2019s ( stock footage ) \u2014 was unrelated to learning outcomes . We note that the stock footage showed the same materials , setup , and sequence as the participant\u2019s own embodied simulation . It is possible that stock footage would be less effective if these properties were altered . That said , the results are encouraging for the prospect of implementing rela - tional scaffolding in educational contexts , such as class - rooms , where individualized footage is unfeasible . General Discussion Our findings demonstrate that relational scaffolding\u2014 systematic , guided comparison of observable and cor - responding modeled events\u2014supports students\u2019 understanding of scientific models . We tested relational scaffolding in the context of instruction about the day / night cycle , a fundamental science topic that involves linking Earth - based observations to a space - based Table 3 . Means for Demographic and Cognitive Variables for Experiment 2 Variable type and variable Sequential scaffolding Relational scaffolding Overall Stock footage Self footage Stock footage Self footage Demographics Gender ( female : male ) 12 : 13 20 : 4 12 : 12 14 : 11 59 : 40 Age 8 . 6 ( 0 . 4 ) 8 . 5 ( 0 . 4 ) 8 . 6 ( 0 . 5 ) 8 . 7 ( 0 . 4 ) 8 . 6 ( 0 . 4 ) Pretest and posttest n 25 24 25 25 99 Delayed posttest N 19 16 22 13 70 Pretest Verbal ability ( PPVT standard score ) 103 . 0 ( 13 . 7 ) 103 . 9 ( 15 . 3 ) 100 . 0 ( 12 . 9 ) 100 . 0 ( 14 . 5 ) 101 . 7 ( 14 . 0 ) Mental rotation ( PMA - SR score ) 9 . 0 ( 2 . 2 ) 9 . 3 ( 1 . 9 ) 9 . 4 ( 2 . 2 ) 9 . 5 ( 2 . 4 ) 9 . 3 ( 2 . 2 ) Perspective taking ( PTT - C score ) 9 . 7 ( 2 . 8 ) 10 . 0 ( 3 . 6 ) 10 . 8 ( 4 . 1 ) 10 . 5 ( 3 . 7 ) 10 . 3 ( 3 . 6 ) Space - science concept ( ASSCI score ) 2 . 7 ( 1 . 4 ) 2 . 5 ( 1 . 4 ) 2 . 4 ( 1 . 7 ) 2 . 8 ( 1 . 2 ) 2 . 6 ( 1 . 4 ) Day / night understanding score 6 . 8 ( 3 . 6 ) 8 . 1 ( 4 . 3 ) 7 . 4 ( 5 . 2 ) 6 . 6 ( 4 . 2 ) 7 . 2 ( 4 . 3 ) Posttest Space - science concept ( ASSCI score ; n = 94 ) 3 . 3 ( 1 . 5 ) 2 . 9 ( 1 . 5 ) 3 . 3 ( 1 . 4 ) 3 . 0 ( 1 . 7 ) 3 . 1 ( 1 . 5 ) Day / night understanding score 13 . 6 ( 5 . 9 ) 14 . 5 ( 6 . 0 ) 15 . 2 ( 4 . 7 ) 13 . 8 ( 5 . 2 ) 14 . 3 ( 5 . 4 ) Delayed posttest Mental rotation ( PMA - SR score ; n = 51 ) 9 . 3 ( 2 . 6 ) 9 . 9 ( 2 . 8 ) 9 . 5 ( 3 . 1 ) 9 . 4 ( 2 . 1 ) 9 . 5 ( 2 . 6 ) Perspective taking ( PTT - C score ; n = 68 ) 11 . 6 ( 4 . 1 ) 11 . 5 ( 4 . 9 ) 13 . 0 ( 3 . 8 ) 13 . 0 ( 4 . 2 ) 12 . 3 ( 4 . 2 ) Day / night understanding score 12 . 7 ( 5 . 5 ) 13 . 6 ( 4 . 6 ) 15 . 3 ( 4 . 6 ) 16 . 2 ( 5 . 6 ) 14 . 3 ( 5 . 1 ) Delay interval ( weeks ) 3 . 9 ( 1 . 6 ) 3 . 9 ( 1 . 8 ) 3 . 7 ( 1 . 3 ) 4 . 5 ( 1 . 9 ) 4 . 0 ( 1 . 6 ) Note : Standard deviations are given in parentheses . PPVT = Peabody Picture Vocabulary Test ; PMA - SR = spatial - relations subtest of the Primary Mental Abilities Test ; PTT - C = Perspective Taking Test for Children ; ASSCI = Astronomy and Space Science Concept Inventory . Relational Scaffolding and Scientific Models 1299 model of planetary motion . Across two experiments , relational scaffolding was found to enhance third grad - ers\u2019 understanding of the day / night cycle , especially among students with little prior knowledge . This research speaks to the potential of extending theories of analogical thinking to fundamental issues in science education ( see also Goldwater & Schalk , 2016 ; Jee et al . , 2010 ) . When a scientific model has no obvious connection to observable phenomena , struc - tural alignment may be crucial for comprehension . Rela - tional scaffolding facilitates the alignment process through several coordinated supports . Explicit compari - sons illuminated shared relational structure ( Goldwater & Gentner , 2015 ) . A trained researcher pointed at and Table 4 . Pearson Correlations Between Experiment 2 Variables Variable 2 3 4 5 6 7 8 9 10 11 12 1 . Gender . 08 \u2212 . 05 . 11 < . 01 . 06 . 09 . 11 . 02 . 03 \u2212 . 04 \u2212 . 06 2 . Age \u2212 . 38 * * \u2212 . 01 \u2212 . 09 \u2212 . 05 . 01 \u2212 . 10 < . 01 . 19 . 01 . 12 3 . Verbal ability . 27 * * . 25 * . 19 . 23 * . 26 * . 22 * . 01 . 09 . 01 4 . Pretest mental rotation . 05 . 11 . 37 * * . 27 * * . 23 * . 32 * . 16 . 01 5 . Pretest perspective taking . 03 . 29 * * . 24 * . 21 * . 02 . 53 * * . 13 6 . Pretest space - science concept . 13 . 46 * * . 11 . 18 . 06 . 22 7 . Pretest day / night understanding . 36 * * . 46 * * . 21 . 25 * . 33 * * 8 . Posttest space - science concept ( n = 94 ) . 42 * * . 13 . 23 . 37 * * 9 . Posttest day / night understanding . 12 . 34 * * . 63 * * 10 . Delayed posttest mental rotation ( n = 51 ) . 26 . 02 11 . Delayed posttest perspective taking ( n = 68 ) . 19 12 . Delayed posttest day / night understanding ( n = 70 ) Note : Gender was coded 0 for male and 1 for female . The fourth edition of the Peabody Picture Vocabulary Test ( PPVT ) measured verbal ability , the spatial - relations subtest of the Primary Mental Abilities Test ( PMA - SR ) measured mental rotation , the Perspective Taking Test for Children ( PTT - C ) measured perspective taking , and the Astronomy and Space Science Concept Inventory ( ASSCI ) measured children\u2019s space - science concept . * p < . 05 . * * p < . 01 . 0 3 6 9 12 15 18 21 24 27 SS Individual Score RS Individual Score SS Group Mean RS Group Mean 0 3 6 9 12 15 18 21 24 27 Posttest Delayed Posttest D a y / N i gh t - C y c l e - U nd e r s t a nd i n g S c o r e Below Median Above Median Below Median Above Median Fig . 7 . Posttest and delayed - posttest day / night - cycle - understanding scores for participants who scored below and above the median on the pretest . Results are shown separately for sequential scaffolding ( SS ) and relational scaffolding ( RS ) . Individual participant scores are shown in gray ; means are in orange . Error bars show standard errors . Individual scores were jittered to produce separation on the x - axis . SS = sequential scaffolding ; RS = relational scaffolding . 1300 Jee , Anggoro between videos of observed and modeled events to clarify key correspondences ( Richland et al . , 2007 ; Yuan , Uttal , & Gentner , 2017 ) . The scaffolding was com - prehensive , underscoring the coherence of the scientific model and discouraging inaccurate , piecemeal explana - tions ( Au & Romo , 1996 ; Chi et al . , 2012 ) . This system - atic approach may be most effective when models are complex , unfamiliar , or counterintuitive . Future applica - tions of relational scaffolding should therefore consider the subject matter , students\u2019 background knowledge , and common conceptions that might impede ( or pro - mote ) science learning ( Shtulman , 2017 ; Vosniadou & Skopeliti , 2014 ) . Relational scaffolding is well suited to novices , who lack conceptual knowledge for inferring relational structure ( McNamara et al . , 1996 ) and are susceptible to cognitive overload during instruction ( Sweller , 1994 ) . An intriguing possibility is that relational scaffolding can help level the playing field for children who are poorly prepared for science education . Science - achievement gaps emerge in the early school years and often persist , largely because of disparities in students\u2019 basic science knowledge ( Morgan , Farkas , Hillemeier , & Maczuga , 2016 ) . Methods that assist underprepared students would profoundly impact many children and , ultimately , increase the pool of qualified candidates for careers in science . Important questions remain about the model - based instruction that accompanies relational scaffolding . Our Experiment 1 found that embodied simulation did not increase student learning beyond models - only instruc - tion . This seems at odds with evidence that physical experience improves learning of science concepts ( e . g . , Kontra et al . , 2015 ) . However , rather than passively observing , participants repeatedly adopted an Earth - based perspective ( looking out from model Earth ) dur - ing the 3 - D model activity . A fully enacted simulation may be unnecessary when physical ( or virtual ) model - ing is immersive , providing sensorimotor feedback link - ing model - based observations to embodied actions ( DeSutter & Stieff , 2017 ; Lindgren , Tscholl , Wang , & Johnson , 2016 ) . Nonetheless , our findings leave open the possibility that embodied simulation helps students understand an external model when the two are explic - itly aligned\u2014that is , through relational scaffolding . Fur - ther experimentation is required to explore this and other questions about how model - based instruction contributes to the relational - scaffolding effect . The day / night cycle was a prime candidate for testing relational scaffolding . Yet many other funda - mental scientific models portray invisible entities and processes with no obvious connection to observable phenomena . Seasonal change relates to Earth\u2019s tilt and orbit , energy transfer and state changes involve invisible molecular activity , and illness and immunity relate to microscopic viruses , vaccines , and immune cells . Nonscientific conceptions are widespread for each of these topics ( Shtulman , 2017 ) . Each is thus a promising subject for future tests of the relational - scaffolding approach . Action Editor Erika E . Forbes served as action editor for this article . Author Contributions Both authors created the study concept , developed the study design and materials , and supervised the training of research assistants . B . D . Jee developed the data coding and analyzed and interpreted the data . F . K . Anggoro recruited the partici - pants and managed the data collection . B . D . Jee drafted the manuscript , and F . K . Anggoro provided critical revisions . Both authors approved the final manuscript for submission . Acknowledgments We thank David Uttal , Cherilynn Morrow , and Dedre Gentner for providing valuable feedback throughout the project . We thank the project coordinators and research assistants from College of the Holy Cross and Worcester State University who assisted with data collection and coding . We are grateful to the schools , teachers , parents , and students who took part in the research . Declaration of Conflicting Interests The author ( s ) declared that there were no conflicts of interest with respect to the authorship or the publication of this article . Funding This research was supported by the Institute of Education Sciences , U . S . Department of Education , through Grant R305A150228 to the College of the Holy Cross . The opinions expressed are those of the authors and do not represent the views of the Institute of Education Sciences or the U . S . Department of Education . Supplemental Material Additional supporting information can be found at http : / / journals . sagepub . com / doi / suppl / 10 . 1177 / 0956797619864601 Open Practices All data and materials have been made publicly available via the Open Science Framework and can be accessed at osf . io / kszge . The design and analysis plans for this study were not preregistered . The complete Open Practices Disclosure for this article can be found at http : / / journals . sagepub . com / doi / suppl / 10 . 1177 / 0956797619864601 . This article has received the Relational Scaffolding and Scientific Models 1301 badges for Open Data and Open Materials . More information about the Open Practices badges can be found at http : / / www . psychologicalscience . org / publications / badges . References American Association for the Advancement of Science . ( 2009 ) . Benchmarks for science literacy . New York , NY : Oxford University Press . Au , T . K . , & Romo , L . F . ( 1996 ) . Building a coherent concep - tion of HIV transmission : A new approach to AIDS edu - cation . In D . L . Medin ( Ed . ) , The Psychology of Learning and Motivation ( Vol . 35 , pp . 193 \u2013 241 ) . San Diego , CA : Academic Press . Chi , M . T . H . , Roscoe , R . D . , Slotta , J . D . , Roy , M . , & Chase , C . C . ( 2012 ) . Misconceived causal explanations for emergent processes . Cognitive Science , 36 , 1 \u2013 61 . Clement , J . ( 1993 ) . Using bridging analogies and anchoring intuitions to deal with students\u2019 preconceptions in physics . Journal of Research in Science Teaching , 30 , 1241 \u2013 1257 . Davis , M . J . ( 2010 ) . Contrast coding in multiple regression analysis : Strengths , weaknesses , and utility of popular coding structures . Journal of Data Science , 8 ( 1 ) , 61 \u2013 73 . DeSutter , D . , & Stieff , M . ( 2017 ) . Teaching students to think spatially through embodied actions : Design principles for learning environments in science , technology , engineer - ing , and mathematics . Cognitive Research : Principles and Implications , 2 ( 1 ) , Article 22 . doi : 10 . 1186 / s41235 - 016 - 0039 - y Dunn , L . M . , & Dunn , D . M . ( 2007 ) . PPVT - 4 : Peabody Picture Vocabulary Test . San Antonio , TX : Pearson Assessments . Faul , F . , Erdfelder , E . , Buchner , A . , & Lang , A . - G . ( 2009 ) . Statistical power analyses using G * Power 3 . 1 : Tests for correlation and regression analyses . Behavior Research Methods , 41 , 1149 \u2013 1160 . Frick , A . , M\u00f6hring , W . , & Newcombe , N . S . ( 2014 ) . Picturing perspectives : Development of perspective - taking abilities in 4 - to 8 - year - olds . Frontiers in Psychology , 5 , Article 386 . doi : 10 . 3389 / fpsyg . 2014 . 00386 Gentner , D . , Anggoro , F . K . , & Klibanoff , R . S . ( 2011 ) . Structure mapping and relational language support children\u2019s learn - ing of relational categories . Child Development , 82 , 1173 \u2013 1188 . Gentner , D . , & Markman , A . B . ( 1997 ) . Structure mapping in analogy and similarity . American Psychologist , 52 , 45 \u2013 56 . Gentner , D . , & Toupin , C . ( 1986 ) . Systematicity and sur - face similarity in the development of analogy . Cognitive Science , 10 , 277 \u2013 300 . Goldwater , M . B . , & Gentner , D . ( 2015 ) . On the acquisition of abstract knowledge : Structural alignment and expli - cation in learning causal system categories . Cognition , 137 , 137 \u2013 153 . Goldwater , M . B . , & Schalk , L . ( 2016 ) . Relational categories as a bridge between cognitive and educational research . Psychological Bulletin , 142 , 729 \u2013 757 . Holyoak , K . J . , & Koh , K . ( 1987 ) . Surface and structural sim - ilarity in analogical transfer . Memory & Cognition , 15 , 332 \u2013 340 . Jee , B . D . , Uttal , D . H . , Gentner , D . , Manduca , C . , Shipley , T . F . , Tikoff , B . , . . . Sageman , B . ( 2010 ) . Commentary : Analogical thinking in geoscience education . Journal of Geoscience Education , 58 , 2 \u2013 13 . Kalyuga , S . ( 2007 ) . Expertise reversal effect and its impli - cations for learner - tailored instruction . Educational Psychology Review , 19 , 509 \u2013 539 . Kontra , C . , Lyons , D . J . , Fischer , S . M . , & Beilock , S . L . ( 2015 ) . Physical experience enhances science learning . Psycho - logical Science , 26 , 737 \u2013 749 . Kuhn , T . S . ( 1962 ) . The structure of scientific revolutions . Chicago , IL : University of Chicago Press . Kurtz , K . J . , Miao , C . H . , & Gentner , D . ( 2001 ) . Learning by analogical bootstrapping . The Journal of the Learning Sciences , 10 , 417 \u2013 446 . Lindgren , R . , Tscholl , M . , Wang , S . , & Johnson , E . ( 2016 ) . Enhancing learning and engagement through embodied interaction within a mixed reality simulation . Computers & Education , 95 , 174 \u2013 187 . Mayer , R . E . , & Moreno , R . ( 2003 ) . Nine ways to reduce cognitive load in multimedia learning . Educational Psychologist , 38 , 43 \u2013 52 . McCloskey , M . ( 1983 ) . Naive theories of motion . In D . Gentner & A . L . Stevens ( Eds . ) , Mental models ( pp . 299 \u2013 324 ) . Hillsdale , NJ : Erlbaum . McNamara , D . S . , Kintsch , E . , Songer , N . B . , & Kintsch , W . ( 1996 ) . Are good texts always better ? Interactions of text coherence , background knowledge , and levels of under - standing in learning from text . Cognition and Instruction , 14 , 1 \u2013 43 . Morgan , P . L . , Farkas , G . , Hillemeier , M . M . , & Maczuga , S . ( 2016 ) . Science achievement gaps begin very early , per - sist , and are largely explained by modifiable factors . Educational Researcher , 45 , 18 \u2013 35 . Morrow , C . A . ( 2000 ) . Kinesthetic astronomy : The sky time les - son . The Physics Teacher , 38 ( 4 ) , Article 252 . doi : 10 . 1119 / 1 . 880520 National Research Council . ( 2012 ) . A framework for K - 12 science education : Practices , crosscutting concepts , and core ideas . Washington , DC : The National Academies Press . Nersessian , N . J . ( 2010 ) . Creating scientific concepts . Cambridge , MA : MIT Press . Plummer , J . D . , Kocareli , A . , & Slagle , C . ( 2014 ) . Learning to explain astronomy across moving frames of reference : Exploring the role of classroom and planetarium - based instructional contexts . International Journal of Science Education , 36 , 1083 \u2013 1106 . Richland , L . E . , Zur , O . , & Holyoak , K . J . ( 2007 ) . Cognitive supports for analogies in the mathematics classroom . Science , 316 , 1128 \u2013 1129 . Sadler , P . M . , Coyle , H . , Miller , J . L . , Cook - Smith , N . , Dussault , M . , & Gould , R . R . ( 2010 ) . The Astronomy and Space Science Concept Inventory : Development and validation of assessment instruments aligned with the K \u2013 12 national science standards . Astronomy Education Review , 8 ( 1 ) , Article 010111 . doi : 10 . 3847 / AER2009024 1302 Jee , Anggoro Shtulman , A . ( 2017 ) . Scienceblind : Why our intuitive theories about the world are so often wrong . New York , NY : Basic Books . Sweller , J . ( 1994 ) . Cognitive load theory , learning difficulty , and instructional design . Learning and Instruction , 4 , 295 \u2013 312 . Thurstone , T . G . , & Thurstone , L . L . ( 1962 ) . Primary mental abilities tests . Chicago , IL : Science Research Associates . Vosniadou , S . , & Brewer , W . F . ( 1994 ) . Mental models of the day / night cycle . Cognitive Science , 18 , 123 \u2013 183 . Vosniadou , S . , & Skopeliti , I . ( 2014 ) . Conceptual change from the framework theory side of the fence . Science & Education , 23 , 1427 \u2013 1445 . Vosniadou , S . , & Skopeliti , I . ( 2017 ) . Is it the Earth that turns or the Sun that goes behind the mountains ? Students\u2019 misconceptions about the day / night cycle after reading a science text . International Journal of Science Education , 39 , 2027 \u2013 2051 . West , S . G . , Aiken , L . S . , & Krull , J . L . ( 1996 ) . Experimental personality designs : Analyzing categorical by continuous variable interactions . Journal of Personality , 64 , 1 \u2013 48 . Yuan , L . , Uttal , D . , & Gentner , D . ( 2017 ) . Analogical processes in children\u2019s understanding of spatial representations . Developmental Psychology , 53 , 1098 \u2013 1114 .",
    "chanSOLVENTMixedInitiative2018": "31 SOLVENT : A Mixed Initiative System for Finding Analogies between Research Papers JOEL CHAN , University of Maryland JOSEPH CHEE CHANG , Carnegie Mellon University TOM HOPE and DAFNA SHAHAF , Hebrew University of Jerusalem ANIKET KITTUR , Carnegie Mellon University Scientific discoveries are often driven by finding analogies in distant domains , but the growing number of papers makes it difficult to find relevant ideas in a single discipline , let alone distant analogies in other domains . To provide computational support for finding analogies across domains , we introduce Solvent , a mixed - initiative system where humans annotate aspects of research papers that denote their background ( the high - level problems being addressed ) , purpose ( the specific problems being addressed ) , mechanism ( how they achieved their purpose ) , and findings ( what they learned / achieved ) , and a computational model constructs a semantic representation from these annotations that can be used to find analogies among the research papers . We demonstrate that this system finds more analogies than baseline information - retrieval approaches ; that annotators and annotations can generalize beyond domain ; and that the resulting analogies found are useful to experts . These results demonstrate a novel path towards computationally supported knowledge sharing in research communities . 1 CCS Concepts : \u2022 Human - centered computing \u2192 Collaborative and social computing systems and tools ; Additional Key Words and Phrases : Scientific discovery ; computer - supported cooperative work ; analogy ; crowdsourcing ACM Reference format : Joel Chan , Joseph Chee Chang , Tom Hope , Dafna Shahaf , and Aniket Kittur . 2018 . SOLVENT : A Mixed Initiative System for Finding Analogies between Research Papers . Proc . ACM Hum . - Comput . Interact . 2 , CSCW , Article 31 ( November 2018 ) , 21 pages . https : / / doi . org / 10 . 1145 / 3274300 1 INTRODUCTION Analogies are an important driver of scientific progress . For example , Darwin\u2019s theory of evolution was conceived by analogy to human population dynamics [ 21 ] ; Salvador Luria\u2019s Nobel - Prize - winning work on bacterial mutation was inspired by analogy to a slot machine [ 31 ] ; the simulated annealing optimization algorithm was inspired by the annealing process for removing imperfections in metals [ 26 ] ; and information foraging theory was inspired by analogy to how animals forage for food [ 36 , 41 ] . 1 This abstract demonstrates the annotation scheme we use in our mixed - initiative system : background ( yellow ) , purpose ( red ) , mechanism ( blue ) , and findings ( gray ) . Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page . Copyrights for components of this work owned by others than the author ( s ) must be honored . Abstracting with credit is permitted . To copy otherwise , or republish , to post on servers or to redistribute to lists , requires prior specific permission and / or a fee . Request permissions from permissions @ acm . org . \u00a9 2018 Copyright held by the owner / author ( s ) . Publication rights licensed to Association for Computing Machinery . 2573 - 0142 / 2018 / 11 - ART31 $ 15 . 00 https : / / doi . org / 10 . 1145 / 3274300 Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 2 G . Zhou et al . However , as the number of papers grows , individual researchers increasingly struggle to discover relevant analogies from work within their own discipline , let alone analogies across different domains or disciplines ( e . g . , from metallurgy to optimization , for simulated annealing ) . The problem is especially acute as many fields of study now require recombining knowledge across multiple disciplines [ 24 , 45 ] . For example , cognitive science relies on insights from linguistics , artificial intelligence , neuroscience , cognitive psychology , and education ; creativity and innovation research must integrate insights from the study of individual cognitive factors , motivational / emotional factors , team dynamics , and societal / economic incentive structures [ 38 ] . Individual scientists lack the time and resources to keep up with all the possible conferences , journals , and talks that might hold analogical insights for problems they are working on . For these reasons , computational systems that can mine analogies between research papers could significantly accelerate scientific progress by reducing the cost of knowledge discovery . A core challenge to building such systems is creating appropriate semantic representations of documents that can support analogical matching . The essence of analogy is matching a target knowledge representation ( e . g . , document , abstract , research problem description ) with other source knowledge representations that share its core relational structure [ 15 ] ; when the analogous sources also have many other details that are very different from the target , they are known as far or domain - distant analogies . For example , the relational structure of the annealing process can be described as \u201cMINIMIZING imperfections in a substance BY HEATING and gradually COOLING the substance\u201d ( relations in ALL - CAPS ) . This relational structure can then be analogically matched to other sources that share all or part of the structure , such as optimization problems , which share a similar relation of \u201cMINIMIZING error in a model\u201d , despite vast differences in their domain features ( e . g . , physical temperature and metals VS . algorithms and digital data ) . We build on promising recent work by Hope et al . [ 22 ] that takes a mixed - initiative approach to creating such representations for consumer product descriptions : 1 ) crowds annotate aspects of the documents that denote their purpose ( what they are trying to achieve ) and mechanism ( how they achieve that purpose ) , and 2 ) these annotations are used to construct semantic vector representations that capture the overall purpose and mechanism of each document . The intuition behind this approach is that the overall representations of purpose and mechanism constitute a \" soft \" relational schema [ 16 ] because of the in - built causal relation between purpose and mechanism . This soft schema can then be used for analogical queries ( e . g . , finding other products with similar purpose but different mechanisms \u2013 a classic analogical problem solving query [ 16 ] ) . Hope et al . demonstrated that crowd workers could produce purpose and mechanism annotations reliably and quickly ( usually in under a minute per document ) for several thousand consumer product descriptions , and that the semantic vector representations could be used to find analogies between products , even when they were from different domains ( e . g . , a \u201cdigital collar\u201d that alerts a pet owner when the pet is too far away , and a bluetooth wristband that alerts a parent if the child wearing it is too far away ) , and at a higher rate than traditional information retrieval approaches ( e . g . , TF - IDF , GloVe word embeddings [ 35 ] ) . We are interested in extending Hope et al\u2019s [ 22 ] approach to help researchers find analogies from the research literature for their own work . For example , could researchers use this approach to find mechanisms ( e . g . , HEATING and gradually COOLING a substance ) from other research papers with similar purposes ( e . g . , MINIMIZING imperfections in a substance ) that might inspire new ideas for a target purpose ( e . g . , INCREASE and gradually DECREASE the degree of stochasticity in the optimization process to MINIMIZE error in the model , which is the essence of simulated annealing [ 26 ] ) ? Or find other papers with the same purpose and mechanism ( e . g . , for competitive analysis ) , even if those papers are from different domains ? Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 3 However , key features of research papers may prevent a straightforward application of Hope et al\u2019s [ 22 ] approach . First , research papers are often written in complex , domain - specific language not easily understood by laypeople ( e . g . , \u201cHMM\u201d , \u201cregularization\u201d , \u201ccross - validation\u201d , \u201chuman computation\u201d ) . The complexity of the concepts and language might greatly increase the cost of human annotation , rendering the method unsuitable for mining analogies across large , complex , multidisciplinary corpora of research papers . Research papers also often include multiple purposes that might have hierarchical causal relationships with each other ( e . g . , \u201cexpose model parameters and error patterns\u201d TO \u201csupport algorithmic accountability\u201d ) . Additionally , many research papers aim to understand some phenomenon , instead of contribute novel mechanisms to add value for some user : for these papers , the purpose - mechanism schema might not capture the core relational structure of these papers , and thus fail to find useful analogies for these papers . In this paper , we explore approaches to address these challenges for extracting and using purpose - mechanism representations to discover analogical relationships in complex scientific literature . Our investigation yields the following contributions : ( 1 ) Annotating the purpose and mechanism aspects of research papers is scalable in terms of cost ( < 1 minute per paper ) , not critically dependent on annotator expertise , and generalizable across domains . We explore this by deploying the method with expert annotators on papers in their domain ( Study 1 ) and outside their domain ( Study 2 ) , and with novice crowd annotators ( Study 3 ) . ( 2 ) Purpose - mechanism representations of research papers can be used to discover analogical relationships that are systematically missed by traditional state - of - the - art semantic models ( Study 1 ) , including distant analogies found valuable to a domain expert ( Study 2 ) . ( 3 ) Extending the annotation scheme to incorporate other aspects of research papers ( e . g . , their higher - level problem \u201cbackground\u201d , and the key findings of the paper ) yields measurable gains in analogy - finding performance ( Study 1 ) . These results suggest that Hope et al\u2019s [ 22 ] mixed - initiative system ( with key modifications from our investigation ) is a promising approach for finding analogies amongst research papers . We call this modifed mixed - initiative system solvent , to capture its core intution : using a mixture of crowdsourcing and machine learning to \" dissolve \" research papers into their constituent soft relational schemas ( e . g . , purpose - mechanism schemas ) , which can then be used to find analogical combinations of research papers that yield novel discoveries and innovations . We offer solvent as a novel path towards computationally supporting knowledge sharing in research communities . 2 RELATED WORK 2 . 1 Citation Recommendation The problem of finding analogous research papers can be thought of as a special case of the problem of citation recommendation : given some target manuscript that one is writing , find other papers that are suitable citations [ 20 , 42 ] . Our work here on finding suitable semantic representations of documents is similar to content - based approaches that leverage semantic representations of documents to search for recommendations [ 5 ] , and complementary to graph - based approaches ( e . g . , based on properties of citation / co - authorship networks [ 29 , 37 ] ) to this problem . 2 . 2 Computational Analogy Computational analogy is a well - studied problem in cognitive science and artificial intelligence . A major branch of research has devised algorithms for reasoning over rich , relational representa - tions , such as Gentner and colleagues\u2019 structure - mapping engine [ 14 , 15 ] , Hummel and Holyoak\u2019s LISA analogy engine [ 23 ] , and Vattam and colleagues\u2019 [ 44 ] Design Analogy to Nature Engine Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 4 G . Zhou et al . ( DANE ) . These systems can achieve impressive , human - like performance on analogy tasks , but require very detailed and rich relational representations : for example , to reason about the sim - ulated annealing analogy , the structure - mapping engine [ 14 , 15 ] would require a representa - tion of the concept of annealing that looks something like \u201cCAUSE ( AND ( HEAT ( metalworker , metal ) , CONTROL ( metalworker , COOL ( metalworker , metal ) ) ) , MINIMIZE ( metalworker , IMPER - FECTIONS _ IN ( metal ) ) \u201d . These representations are very costly to obtain : for example , Vattam and colleagues [ 44 ] estimated that converting a single ( complex ) biological system into a formal rep - resentation requires between forty and one hundred person - hours of work ; further , automatic extraction of relational representations across domains remains a difficult open problem [ 19 ] . On the other hand , many computational approaches have been developed that ignore relational structure , opting instead to learn semantic representations from the distributional statistics of words across documents in a corpus\u2014e . g . , word embedding models like Word2Vec [ 32 ] , vector - space models like Latent Semantic Indexing [ 13 ] , and probabilistic topic modeling approaches like Latent Dirichlet Allocation [ 6 ] \u2013 scale well to very large datasets , and perform well at matching based on overall / surface similarity , but struggle to detect analogies between documents when surface similarity is low . Our work extends a new thread of research [ 22 ] that explores how to extract and exploit \u201csoft\u201d relational structure : not going all the way to fully specified relational representations , but still attempting to model some kind of relational structure ( e . g . , semantic representations of the overall purpose and mechanism of a product , which can be used to define a \" soft \" schema for the product ) . 2 . 3 Crowd - Powered and Mixed - Initiative Knowledge Modeling Many researchers have explored the general problem of applying crowdsourcing techniques to synthesize a knowledge model of some domain , from taxonomies of items within a domain [ 10 , 46 ] , to summarizing answers from Web sources for a given question [ 8 , 18 ] , to planning conference talk schedules based in part on topical similarity [ 1 , 9 ] , to organizing ideas for a given problem by topical similarity during collaborative idea generation [ 40 ] . Many of these efforts are mixed - initiative sys - tems , where crowds either generate semantic data to support the creation of a computational seman - tic model [ 40 ] , or use computational methods\u2014such as information - theoretic methods\u2014to support more efficient crowdsourcing strategies [ 7 , 46 ] , or effectively aggregate crowds\u2019 semantic judgments [ 10 ] . These applications tend to be about general topicality or similarity of documents / items within some domain , rather than about analogical relationships between documents / items that might be from different domains . Closer to the problem of supporting analogy , other researchers have explored using crowdsourc - ing to perform or assist with extracting relations [ 30 ] . Most work in this area aims to extract single commonsense facts ( e . g . , \u201cPortland IS _ IN Oregon\u201d , Obama \u2018WORKED _ AT\u2018 the White House , similar to open information extraction [ 3 ] , instead of systems of relations ( e . g . , minimize imperfections in a metal BY controlled heating and cooling of the metal ) : thus , while these relations can be quite useful for improving artificial commonsense reasoning , they are not as useful for reasoning about analogies between scientific papers , especially across multiple disciplines . However , some recent work has examined how to use crowdsourcing to assist with extraction of more specialized relations , for example mechanical engineering - related functions and mechanisms in biological texts [ 2 ] , or experiment details for biomedical research ( e . g . , population studied , treatment given ) [ 43 ] . We are inspired by these efforts ( particularly the approach of using mixed - initiative systems ) , but differ in that we aim to crowdsource knowledge models for complex corpora ( like scientific papers ) across many different knowledge domains . Additionally , instead of attempting to model specific rela - tions and entities in detail , we aim to extract general semantic categories ( e . g . , purpose / mechanism ) that are relevant for analogical reasoning . Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 5 2 . 4 Ontologies of scientific and scholarly discourse The general semantic categories we aim to crowdsource are closely related to work on ontologies of scholarly and scientific discourse [ 11 , 12 ] , with an especially close relation to efforts like the Core Information about Scientific Papers ( CISP ) ontology [ 28 ] , which breaks papers out into \" Goal of investigation \" , \" Motivation \" , \" Object of investigation \" , \" Research method , \" Experiment \" , \" Observation , \" Result \" , and \" Conclusion \" . While our goal differs from this work in aiming to extract elements that support analogy in particular ( vs . construct a complete / useful model of scientific discourse ) , we extend this body of work by systematically exploring how these discourse - like ontology elements might be efficiently crowdsourced ( vs . relying entirely on automation or expert researcher annotations [ 27 ] ) , and what particular combinations of these elements might be useful for supporting analogy mining . 3 ADAPTING THE METHOD FOR RESEARCH PAPERS Applying the purpose - mechanism annotation analogical search method [ 22 ] to the domain of research papers may seem straightforward at first glance : most research projects have research goals ( e . g . , research problems , research questions ) , and methods for achieving those goals . However , there are several major challenges involving in moving from the domain of simple consumer product descriptions to more complex structures and terminology , such as found in scientific literature . 3 . 0 . 1 Complex language . First , research papers tend to be significantly more complex than the typical consumer product idea . For example , most people know that a utility belt is for holding / storing things , so words like \" store / keep \" are purpose words ; in contrast , without con - tent / domain knowledge , it may be difficult to discern which terms denote purpose / mechanism in research papers . For example , a hidden Markov model ( HMM ) is a common mechanism for analyzing sequences of events . However , without specialized domain knowledge it is possible that an annotator might not be able to label it as a mechanism . Fortunately , as we will show , people are able to leverage rich syntactic cues to determine which portions ( not specific words ) denote the purpose / mechanism / etc of the paper , even in the absence of understanding the details of those purposes or mechanisms . 3 . 0 . 2 Hierarchy of problems . Research papers also often address multiple distinct purposes that are hierarchically dependent on each other . For example , a research paper might explore how to improve creativity in teams ( high - level problem ) by improving the ability of team members\u2019 to build on each others\u2019 ideas ( mid - level subproblem ) , and specifically aim to investigate ways to surface the most diverse and best ideas ( low - level subproblem ) . In most cases , the mechanisms contributed by a given paper are most directly causally related to the lower level subproblems ( e . g . , crowdsourcing techniques for summarizing ideas ) ; consequently , the purpose - mechanism schema of a particular paper is often best described by its low - level subproblems , along with the mechanisms it contributes for solving those subproblems . To address this challenge , it may be useful to distinguish between the research background ( higher - level problems ) and the specific problem ( s ) being addressed in the paper to find more directly analogous matches . The background context can also be \" partialed out \" ( e . g . by ignoring it ) to enable discovery of analogies between different areas / fields , e . g . , matching papers that are about crowdsourcing techniques for content analysis in qualitative analysis vs . online classrooms vs . brainstorming teams . 3 . 0 . 3 Mechanisms vs . findings . Finally , many research papers do not fit the purpose - mechanism model exactly . Specifically , many papers cannot really be understood in terms of answering \u201chow\u201d questions ( e . g . , how might we improve work quality for novice crowd workers ) ; instead , their Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 6 G . Zhou et al . primary goal is to understand some phenomenon ( e . g . , when and why do people disclose personal information on social media ? ) . For these \u201cunderstanding - oriented\u201d papers , the purpose remains informative ( what question they want to understand ) , but the mechanism ( how they found the answer ) may no longer capture the core contribution of the paper towards that purpose ; instead , the findings of the paper are the \u201canswer\u201d to the purpose of the paper . For example , a research paper whose purpose is to investigate people\u2019s motivations for disclosing personal information on social media may have ( according to our schema ) particular research methods ( e . g . , interviews with users , content analysis ) as mechanisms , but the answer to the purpose might be the findings it reports ( e . g . , users avoid disclosing if they want to manage their online identity across multiple social contexts ) . Thus , the relevant schema to match on might be a purpose - findings schema , or even just a findings schema , as opposed to a purpose - mechanism one . 3 . 1 Modified annotation scheme To address these challenges , we adapted Hope et al\u2019s [ 22 ] purpose - mechanism annotation scheme to incorporate two new elements ( background and findings ) . This yields the following modified annotation scheme with 4 elements : ( 1 ) Background : What is the intellectual context of this work ? What other ( higher - level ) goals / questions can be furthered by this work ? How might this help other research ( ers ) ? ( 2 ) Purpose : What specific thing ( s ) do the paper\u2019s authors want to do or know ? ( 3 ) Mechanism : How did the paper\u2019s authors do it or find out ? ( 4 ) Findings : Did it work ? What did the paper\u2019s authors find out ? Table 1 ( top ) shows examples of these annotations for a system - oriented vs . a more \u201cunderstanding - oriented\u201d paper . This modified annotation scheme defines the target aspects that solvent trans - forms into soft relational schemas , which are used for analogical matching . 3 . 2 Creating semantic vector representations from the annotations Each of the four annotations ( background , purpose , mechanism , and findings ) yields a set of words . For example , in Table 1 , the set of words highlighted in red are annotated as \" purpose \" . We are interested in taking the four sets of annotations W = { w 1 , w 2 , w 3 , w 4 } ( where each w i is a set of words ) , and creating representations that capture the respective aspects of the paper . We follow a common approach in natural language processing and information retrieval , and build semantic vector representations based on each annotated set of words w i . These vector representations are solvent \u2019s \" soft \" relational schemas that can be used for analogical matching . In particular , we represent words with their word vectors ( e . g . , trained with word2vec [ 33 ] or GloVe [ 35 ] ) , and then for each set w i we compute a weighted average of the words in w i to obtain a representation for aspect i . The average is weighted by the TF - IDF score of each word vector 2 . Note that the TF - IDF score is calculated on an annotation - specific basis : for example , when computing TF - IDF scores for purpose vectors , we use only words that were tagged as purpose when computing term and document frequencies 3 . This is a conceptually similar approach to that taken by Hope et al [ 22 ] and helps us tease apart words that are important for denoting the purpose / mechanism / etc . of documents , as opposed to generalized importance . The end result is that each document is represented by 4 annotation - specific semantic vectors ( background , purpose , mechanism , and findings ) , which we denote as b i \u2208 R D , p i \u2208 R D , m i \u2208 R D , and f i \u2208 R D , respectively ( where D is the dimensionality of the base word vectors we use to represent the words ) . 2 This step is implemented in get - vectors . py in the code provided in the supplementary material 3 This step is implemented in compute - tf - idfs . py in the code provided in the supplementary material Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 7 IdeaHound : Self - sustainable Idea Generation in Creative Online Communities A Comparison of Social , Learning , and Finan - cial Strategies on Crowd Engagement and Out - put Quality One main challenge in large creative online communities is helping their members find inspirational ideas from a large pool of ideas . A high - level approach to address this chal - lenge is to create a synthesis of emerging solution space that can be used to provide participants with creative and diverse inspirational ideas of others . Existing approaches to gener - ate the synthesis of solution space either require community members to engage in tasks that detract from the main activ - ity of generating ideas or depend on external crowd workers to help organize the ideas . We built IDEAHOUND a collab - orative idea generation system that demonstrates an alter - native \u201dorganic\u201d human computation approach , where com - munity members ( rather than external crowds ) contribute feedback about ideas as a byproduct of an activity that nat - urally integrates into the ideation process . This feedback in turn helps the community identify diverse inspirational ideas that can prompt community members to generate more high - quality and diverse ideas . A significant challenge for crowdsourcing has been increas - ing worker engagement and output quality . We explore the effects of social , learning , and financial strategies , and their combinations , on increasing worker retention across tasks and change in the quality of worker output . Through three experiments , we show that 1 ) using these strategies together increased workers\u2019 engagement and the quality of theirwork ; 2 ) asocialstrategywasmosteffectiveforincreas - ing engagement ; 3 ) a learning strategy was most effective in improving quality . The findings of this paper provide strate - gies for harnessing the crowd to perform complex tasks , as well as insight into crowd workers\u2019 motivation . Top 10 nearest words for Abstract : ideas idea analogies proposals productively sub - strates fictions discoveries brainstorm schemas Abstract : efficient expressive retention responsive suscep - tibility differentially enjoyment disengagement alertness Background : ideas creative innovative inspirational analo - gies productively fictions brainstorm substrates wishing Background : crowdworkersworkershepherdingsourcing harnessing intrinsically legitimacy rubrics extrinsic Purpose : curate interleave scaffold organise devise stimu - late formulate ideas consolidating mobilize Purpose : retention differentially worker soundness media - tors maximize susceptibility redefining greatly decreasing Mechanism : idea ideas ideation endeavor grassroots real - ization productively scaffold generative prospector Mechanism : nine eleven twelve fourteen fieldwork com - prehensively separately uncompensated sixteen thirteen Findings : ideas community cultivate cohesive heteroge - neous divergent productively repertoire mobilize socialize Findings : tactic efficient disengagement responsive playful attentiveness partnerships paradoxically expressive Table 1 . Examples of modified annotation scheme applied to a system - oriented paper ( top left ) and an \u201cunderstanding\u201d - oriented paper ( top right ) , and top 10 nearest words for all abstract words VS . each annotation vector for each paper ( bottom ) To give an intuition of what the vectors mean , Table 1 ( bottom ) shows the top 10 nearest neighbor words for each annotation type for two example papers , as well as for a vector representing the full abstract ( i . e . , TF - IDF - weighted average of all non - stopwords in the abstract ) . The nearest neighbors are quite different across the different annotation types and from the full abstract vector , and also map well to the meaning of those annotation types . For example , while the overall abstract for the first example is about improving creativity in online communities ( as seen in the top words for the abstract vector , { ideas , idea , analogies , proposals . . . } ) , the core purpose of enabling efficient synthesis of the ideas generated by the community is well captured by the words { curate , interleave , scaffold , organise , devise . . . } , and the mechanism of \u201cpiggy - backing\u201d on ideation is captured by the mechanism top words , { idea , ideas , ideation , endeavor , grassroots . . . } . The term \u201cgrassroots\u201d in Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 8 G . Zhou et al . particular is interesting because it captures the idea of sustainability , which is completely missed in the top words for the abstract and background vectors . In contrast , for the second , understanding - oriented paper , the mechanism top words are very research - methods oriented , { nine , eleven , twelve , fourteen , fieldwork . . . } . 3 . 3 Analogical similarity metrics Using these specialized vectors , solvent can support a number of interesting analogical queries : ( 1 ) purpose + mechanism : find other research papers with a similar purpose AND mechanism , by computing the average of the cosine between their purpose vectors and the cosine betwen their mechanism vectors , i . e . , \u27e8 cos ( p 1 , p 2 ) , cos ( m 1 , m 2 ) \u27e9 for any pair of paper 1 and paper 2 . By ignoring background / implications , we aim to achieve some level of abstraction away from the domain context . ( 2 ) purpose : find other research papers with a similar purpose , but not necessarily a similar mechanism , with cos ( m 1 , m 2 ) This should find alternative mechanisms for solving some target problem . ( 3 ) mechanism : find other research papers with a similar mechanism , but not necessarily a similar purpose , with cos ( p 1 , p 2 ) . This should find alternative applications for some mechanism . ( 4 ) findings : find other research papers with similar findings , with cos ( f 1 , f 2 ) . This should help find analogies between \u201cunderstanding\u201d papers ( where the core schema is not necessarily a purpose - mechanism pairing ) . 4 STUDY 1 : SOURCING ANNOTATIONS FROM DOMAIN - EXPERT RESEARCHERS We begin our investigation of solvent \u2019s quality and feasibility by sourcing and using annotations from domain expert researchers . This deployment should provide an upper bound on solvent \u2019s quality and cost considerations . 4 . 1 Dataset As an initial test , we deployed our modified mixed - initative system on 50 papers published at the CSCW conference 4 from 2013 - 2017 . This modest number of papers is small enough for experts to manually find analogies ( with reasonable confidence that we have covered most analogies ) , yet large enough to produce interpretable quantitative results ( N = 1 , 225 possible pairs between papers ) . Further , the match to our content / domain expertise ( social computing ) allows us to better define what counts as analogies , and judge the performance of the different metrics accordingly . 4 . 2 Annotation and Vectors Two members of the research team served as annotators . Each paper abstract took a median of 1 minute to annotate . In Studies 2 and 3 , we explore how the method extends to settings where annotators lack content / domain expertise . We originally used pre - trained GloVe [ 35 ] vectors trained on the Common Crawl dataset 5 . However , baseline performance was very poor . We therefore opted for a doc2vec model [ 33 ] , with 600 dimensions , trained on a smaller but more focused and relevant corpus ( 4 , 402 papers from CHI and CSCW from 2010 to 2017 ) . 4 . 3 Baselines We compare the performance of our analogical similarity metrics to two baselines : 4 http : / / cscw . acm . org / 5 https : / / nlp . stanford . edu / projects / glove / Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 9 Match Type Title & Explanation Near analogy : Background + Purpose + Mechanism Toward Collaborative Ideation at Scale : Leveraging Ideas from Others to Generate More Creative and Diverse Ideas : Use crowd - sourcing to identify diverse idea sets to improve creative idea generation Far analogy : Purpose + Mecha - nism Crowd Guilds : Worker - led Reputation and Feedback on Crowd - sourcing Platforms : Make worker certification self - sustainable by using crowdsourced peer review to certify good workers Non - analogy Making Decisions From a Distance : The Impact of Technological Mediation on Riskiness and Dehumanization : No similarity Table 2 . Example expert - found analogies for the IdeaHound paper in the CSCW dataset , with notes on each analogy ( in italics ) . One non - analogy also shown for comparison . Analogies are mapped to the following : core schema from the IdeaHound paper : Make creative idea generation self - sustainable by using crowdsourced peer review to identify diverse idea sets . ( 1 ) all words baseline : create TF - IDF - weighted vectors for each paper from all the words in each paper\u2019s abstract ( denoted as a i ) , and compute cos ( a 1 , a 2 ) , for every pair of paper 1 and paper 2 . This baseline is meant to emulate current content - based recommendation practices , which often operate on some combination of the abstract / full - text . ( 2 ) bkground / purpose + mechanism baseline : create a new semantic vector bp i that captures both background and purpose aspects , by concatenating the weighted vectors for w b ( background tokens ) and w p ( purpose tokens ; each weighted by their respective annotation - specific tf - idf weights ) , and computing the average of this concatenation . We then look for matches with similar background / purpose and mechanism , i . e . , \u27e8 cos ( bp 1 , bp 2 ) , cos ( m 1 , m 2 ) \u27e9 . Comparing our new metrics ( especially the purpose + mechanism and purpose metrics , which ignore background ) to this baseline tests whether it is necessary to separate background and purpose , or whether it is sufficient to simply group all the purposes of a given paper as a single annotation type ( inspired by the purpose - mechanism method from Hope et al . [ 22 ] ) . 4 . 4 Performance Measures To evaluate the performance of the similarity metrics , we use Precision @ K , defined as the number of known analogies in the dataset found in the top K % of matches . Exploring different levels of K allows us to explore how the performance of a given metric changes depending on how conservative it is ( lower K = more conservative , which accepts a higher risk of false negatives in favor of a higher precision of matches returned to the searcher ) . These analogies were manually found by a member of the research team through exhaustive examination of the papers in two phases . First , the research team member created affinity maps of the research papers , and identified shared schemas and relations . Second , these schemas were used to find papers that fit the schema ( thereby defining analogy pairs ) . Finally , we used an initial prototype GloVe model ( with Common Crawl ) to suggest new matches we might have missed . We ended up with 259 analogy pairs ( approximately 21 % of 1225 total possible pairs across 50 papers ; see Table 2 for some example analogies found for the IdeaHound paper from Table 1 6 ) . The research team member who found the analogies in the dataset has 8 years of PhD - level expertise in researching social computing and analogy , providing both domain and process ( for finding analogies ) expertise . The team member considered papers to be analogically related if they share a relational mapping ( e . g . , similar purposes and mechanisms , or similar backgrounds and 6 The full set of known analogy pairs is available in the supplementary material Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 10 G . Zhou et al . K = 1 % 2 % 5 % 10 % 15 % 20 % 25 % all words baseline . 67 ( . 03 ) . 67 ( . 06 ) . 46 ( 11 ) . 37 ( . 18 ) . 35 ( . 25 ) . 31 ( . 29 ) . 29 ( . 34 ) bkground / purpose + mechanism . 84 ( . 05 ) . 84 ( . 09 ) . 64 ( . 16 ) . 50 ( . 24 ) . 43 ( . 30 ) . 37 ( . 35 ) . 33 ( . 39 ) purpose + mechanism . 92 ( . 05 ) . 92 ( . 09 ) . 73 ( . 18 ) . 50 ( . 24 ) . 40 ( . 29 ) . 36 ( . 34 ) . 34 ( . 40 ) purpose . 50 ( . 03 ) . 50 ( . 05 ) . 38 ( . 10 ) . 38 ( . 18 ) . 33 ( . 23 ) . 31 ( . 29 ) . 30 ( . 35 ) mechanism . 92 ( . 05 ) . 80 ( . 08 ) . 66 ( . 16 ) . 49 ( . 23 ) . 45 ( . 32 ) . 39 ( . 37 ) . 35 ( . 42 ) findings . 75 ( . 04 ) . 50 ( . 05 ) . 37 ( . 09 ) . 33 ( . 16 ) . 31 ( . 22 ) . 29 ( . 28 ) . 27 ( . 32 ) Table 3 . Proportion of known analogy pairs found in Study 1 for each metric , varying by K ( lower K = more conservative ; recall scores in parentheses ; pairs sorted by similarity ) . Best - performing scores at each K are bold - underlined . Our bkground / purpose + mechanism , purpose + mechanism , and mechanism metrics consistently find analogy pairs at substantially higher precision than the all words baseline at each level of K . purposes ) . For example , motivating contributions in citizen science relationally maps to motivating scientists to share their code ( e . g . , motivating ( x to contribute to y ) ) . Papers with similar backgrounds ( e . g . , vision science ) but different purposes and mechanisms would not be considered analogies . This is consistent with the literature on analogical mapping ( e . g . , [ 15 ] ) . Importantly , two papers with similar background would still be considered analogies if their purposes or mechanisms mapped relationally ; these are known in the analogy literature as near analogies . Note that because these are all social computing papers , many of the analogies are from the same / similar domains / areas ; therefore , we expect the all words metric to be able to find many of these near analogies . Separating background from purpose also may not be as important for finding near analogies . For this reason , our performance measure in this dataset is a conservative test of the efficacy of our mixed - initiative system . Yet , social computing is broad enough as a field that it allows for some cross - domain analogies ( e . g . , crowdsourcing grades for assignments in MOOCs , vs . crowdsourcing discovery of creative ideas in online innovation platforms ) . We examine precision for K \u2208 ( 1 , 2 , 5 , 10 , 15 , 20 , 25 ) . For each similarity metric , and for each K , we compute similarities between all 1 , 225 possible pairs in the dataset , rank based on those similarities , take the top K % of matches , and then compute how many of those matches were known analogies . 4 . 5 Results 4 . 5 . 1 We find unique analogies missed by the all words baseline . Table 3 shows the quantitative results of our experiment . Three of our similarity metrics ( bkground / purpose + mechanism , pur - pose + mechanism and mechanism ) consistently return a higher proportion of analogies at multiple settings of K , yielding on average gains of 26 % , 31 % , and 30 % in precision compared to all words . The advantage of our metrics is especially substantial at lower levels of K : for example , in the top 1 % and 2 % of matches , our best metric ( purpose + mechanism ) has both a large increase over the all words baseline ( a 37 % increase from . 67 to . 92 ) , and a high precision value in absolute terms . This is significant because we did not distinguish between near and far analogies , and there were many near analogies in the dataset ; these near analogies are likely to dominate at lower levels of K for all words ( which is tuned largely for surface similarity ) . Additionally , there was low overlap between the matches found by our metrics and those found by all words : for example , only 42 % of the purpose + mechanism , and 50 % of the mechanism matches were shared with the all words baseline at K = 5 . Referring back to the matches in Table 2 , the all words baseline was only able to find the near analogy ( \u201cTowards Collaborative Ideation at Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 11 Source : IdeaHound : Self - sustainable Idea Generation in Creative Online Communities Analogy : Crowd Guilds : Worker - led Reputation and Feedback on Crowdsourcing Platforms Top 10 nearest words for Background : ideas creative innovative inspirational analo - gies productively fictions brainstorm substrates wishing Background : workers decentralized crowd crowds dis - persed sourcing equitably collaborators volunteers harness Purpose : curate interleave scaffold organise devise stim - ulate formulate ideas consolidating mobilize Purpose : reputation scores score ratings accountabil - ity legitimacy fairness responsiveness scoring relevancy Mechanism : idea ideas ideation endeavor grassroots re - alization productively scaffold generative prospector Mechanism : crowd workers crowds guilds worker peer sourcing instructors equitably freelancing Table 4 . Illustrative analogy ( right ) found for the IdeaHound paper ( left ) in the CSCW dataset by our pur - pose + mechanism and purpose metrics , but missed by the all words and bkground / purpose + mechanism baseline metrics . Bold - underlined words denote conceptual overlap across matching vectors . Note how different the background vectors are , while there is overlap in concepts for both purpose and mechanism vectors . Both papers leverage community mechanisms [ grassroots , peersourcing ] for the purpose of curating things [ ideas , workers ] , despite key differences in background ( creativity / ideation vs . crowd labor markets ) . This example illustrates the value of separating higher - level ( background ) problems from lower - level ( purpose ) problems during annotation . Scale\u201d ) , while the other metrics were able to find both the near and far analogy ( \u201cCrowd Guilds : Worker - led Reputation and Feedback on Crowdsourcing Platforms\u201d ) . 4 . 5 . 2 Modifications to the annotation scheme are helpful . While bkground / purpose + mecha - nism finds more analogies than all words across all levels of K , it is consistently outperformed by purpose + mechanism ( which separates background and purpose ) across all levels of K . Inspection of the matches suggests that separating background and purpose allows us to find analogies between papers that have different higher - level goals ( and might be thought of as in different domains ) . Table 4 shows one example of this , where purpose + mechanism found an analogy between a paper on devising scalable methods to organise many ideas from an online creative community and distribute them to inspire community members ( \u201cIdeaHound : Self - Sustainable Idea Generation in Creative Online Communities\u201d ) , and a paper on using peer ratings to rank workers by quality in online labor markets ( \u201cCrowd Guilds : Worker - led Reputation and Feedback on Crowdsourcing Platforms\u201d ) . This match was missed by both all words and bkground / pur - pose + mechanism , in part because the background vectors are quite different from each other ( one heavily emphasizing creativity / ideation , and the other heavily emphasizing crowd work ) . 4 . 5 . 3 Findings annotations help find near analogies . While the findings metric does return significantly more matches in the top 1 % of pairs , it does not outperform the all words baseline at K > = 2 . This is partially explained by the high amount of overlap between the matches found by findings and those found by the all words baseline metric ( 64 % are shared , in contrast to the low levels of overlap for the purpose + mechanism and mechanism metrics ) . We will return to this point in the discussion . 5 STUDY 2 : USING SOLVENT TO FIND ANALOGIES WITH REAL - WORLD VALUE Study 1 demonstrated that applying solvent to a corpus of social computing papers enabled us to retrieve known analogical relationships , as defined by a domain expert . In Study 2 , we explore Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 12 G . Zhou et al . whether solvent can provide real - world value for researchers actually looking for analogical inspirations for their work . 5 . 1 Scenario We evaluated our approach with a local mechanical engineering research group at a highly research active , private research university in the Midwestern United States . The group is working on cutting - edge interdisciplinary work at the intersection of bioengineering and mechanical engineering : specifically , they are exploring how to create interesting new 2D and 3D structures at many different scales by stretching / folding polymers . The research group has been struggling to find examples of this work to build on and compare against : it is distributed across such diverse domains as mechanical engineering , civil engineering , aerospace engineering , materials science , mathematics , and design , among others , many of which are outside of the specific domain expertise of the research group . Some examples of their analogy information needs include : \u2022 Competitive analysis : find work from other research groups that are attacking the same problems with similar techniques \u2022 Inspiration : find relevant research that can suggest new properties of polymers to leverage , or new techniques for stretching / folding or exploring 2D / 3D structure designs \u2022 Application : find novel application domains that could benefit from the unique advantages of their fabrication method ( e . g . , non - invasively constructing temporary medical implants in situ ) The team has spent the last year or so conducting literature searches using various standard approaches ( e . g . , Google Scholar , library databases , citations in relevant [ review ] papers , conver - sations with colleagues ) , and have been frustrated by how slow and error - prone the process has been . They are also concerned that they are missing things . The difficulty stems in part from the relative newness of the field , and the degree of fragmentation of knowledge across many different disciplines . In this evaluation , we explore whether solvent can help them find analogies they were not able to find through keyword / database and citation tree search . 5 . 2 Dataset Based on conversations with the group , we identified three out - of - domain sources of research papers to mine from : materials science , civil engineering , and aerospace engineering . The papers were sampled from the 1000 most highly cited papers of 2016 - 2017 ( indexed on Web of Science ) for each of the 3 domains . Two members of our research team ( neither of whom have any training or formal knowledge in any of the domains involved in this scenario ) annotated 90 papers sampled from this larger corpus . As before , rather than use a semantic model trained on a larger but more generalized corpus ( e . g . , Common Crawl ) , we used a model trained on a smaller but more relevant corpus of documents to create aggregate vectors . Specifically , we used word vectors from a continuous - bag - of - words ( CBOW ) word2vec model [ 32 ] trained on 3 , 000 papers in the dataset . We then used the same method as before to create annotation - specific semantic vectors for each of the 90 annotated papers . 5 . 3 Evaluation We used an abstract for one of the research group\u2019s recent manuscripts as a query document . We first highlighted and constructed semantic vectors for the abstract , and then sampled the top 20 matches for the abstract from each of our similarity metrics from before . This corresponds to a Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 13 setting of K at approximately 20 % ( top 20 most similar out of 90 possible matches ) ; we chose this less conservative setting of K to reflect our prior expectation that the number of cross - domain matches that would be useful to the team is likely to be relatively few . We collapsed duplicates and blinded them by removing information about which metric produced the match , resulting in a set of 53 unique possible matches . We showed these matches to a member of their research group ( the lead PhD student ) . We asked the researcher look through the matches as s / he might look through a Google Search result list , and evaluate 1 ) whether each match is in fact useful for their literature review , 2 ) why ( not ) , and 3 ) whether they had previously encountered that match . 5 . 4 Results 5 . 4 . 1 Our metrics find more analogies than all words baseline . Overall , the researcher identified 7 out of the 53 matches as relevant and new . This reflects the extremely challenging setting of finding useful and novel analogies among 90 randomly sampled papers from outside the core domain of the work . 5 out of those 7 analogies were found by one of / both the purpose + mechanism and mechanism similarity metrics ( an aggregate precision of approximately . 25 ; comparable to precision at K = . 2 in Study 1 ) , while 2 were found by the all words metric ( precision = . 1 ) . 5 . 4 . 2 Our metrics find different analogies from all words baseline . In addition to this numerical difference , there was also a qualitative difference in the types of matches returned . To illustrate , the most relevant match found by the all words metric was almost a direct replication of their work ( manipulating the structure of a membrane in a controlled way by applying tension to the membrane ) from an aerospace engineering paper . The researcher was uncertain if she had seen this paper before , and found it useful for competitive analysis . In contrast , the most relevant match found by the purpose + mechanism metric was a paper on using multi - agent systems to generate a variety of geometric structures . The mechanism metric also found an interesting potential analogy of a civil engineering paper analyzing web crippling phenomena in steel beams under tension . Both of these matches were judged to be both potentially useful ( the first for inspiring alternative approaches to systematically explore 2D / 3D structures by applying different levels and types of tension , the second for potentially inspiring new mathematical models to analyze their polymers , which could reveal new properties to leverage for their designs ) and novel ( neither had been seen by the researcher ) . 6 STUDY 3 : SCALING UP SOLVENT THROUGH CROWDSOURCING Thus far we have shown that solvent increases our ability to find known analogies , as well as analogies that can provide real - world value to a domain expert . However , in both Study 1 and 2 , our annotations were sourced from expert researchers , who make up a highly limited resource pool . In this final section , we explore how we might scale up solvent by crowdsourcing annotations from a wider range of non - experts who might provide a much larger pool of possible human judgments . Study 2 ( where researchers annotated papers outside of their domain / discipline ) suggests that annotation does not critically depend on domain knowledge ; still , non - experts with little to no experience writing or reading research papers as a \u201cgenre\u201d may produce annotations that are too noisy to support effective analogy - mining ( or require expensive quality control mechanisms that would make the cost of crowdsourcing prohibitively high ) . Therefore , in Study 3 , we explore whether crowdsourced annotations for papers in the set of 50 CSCW papers can replicate or approximate the analogy - finding gains shown in Study 1 with researcher annotations . We crowdsource annotations from workers on Upwork ( where we recruited workers with general writing expertise but no Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 14 G . Zhou et al . Fig . 1 . A screenshot of the crowd - facing interface for annotating the abstracts . Workers annotate aspects by first selecting the annotation type ( an in - line description is provided as a reminder ) and then using an intuitive click - and - drag highlighting interaction to annotate one or more words . domain or research experience ) and workers on Amazon Mechanical Turk ( who represent a larger population of workers with little or no expertise requirements ) . 6 . 1 Crowdsourcing setup Workers used an intuitive drag - and - highlight interaction to annotate each abstract ( see Figure 1 ) . We included a training step before the main task : 2 gold standard examples to illustrate the annotations in context , and a single training example ( where workers would first annotate and then compare with the gold standard ) . 6 . 1 . 1 Deployment on Upwork . We recruited two workers with general copywriting expertise from Upwork . The first worker held a bachelor\u2019s degree in English Literature , and a master\u2019s degree in business administration , had more than nine years experience writing and editing academic , business , literary and technical documents , and was paid at a rate of $ 30 / hr . The second worker held a master\u2019s degree in English Literature and a certificate in Publishing , had experience with editing book - length projects for major publishers and creating research dossiers for news publications , and was paid at a rate of $ 20 / hr . Annotation times were similar to the MTurk and Study 1 deployments , with a median completion time per document of 1 minute . Each worker provided annotations for half ( 25 ) of the papers in the dataset . Thus , each paper\u2019s annotations was based on one worker , similar to the deployment with domain - expert researchers in Study 1 . 6 . 1 . 2 Deployment on Amazon Mechanical Turk . We screened for workers with at least 95 % approval rate for at least 5 , 000 tasks . The overall task ( including login , training , and the actual task ) took a median of 4 . 0 minutes ; actual time annotating the main document took a median of 1 . 3 minutes . We paid workers $ 0 . 70 for each task completion , for an effective average hourly rate of $ 10 / hr . We obtained annotations from an average of 3 workers for each document . We aggregated annotations across workers for each word by majority vote ( weighted by each worker\u2019s performance on the training example ) . 6 . 2 Results 6 . 2 . 1 Crowd annotations had substantial agreement with researcher annotations . Table 5 shows the accuracies ( i . e . , agreement with researcher annotations ) by annotation type . Overall , Upwork Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 15 Overall Bkgd Find Mech Purp Upwork 0 . 78 0 . 75 0 . 94 0 . 72 0 . 68 MTurk 0 . 59 0 . 66 0 . 71 0 . 53 0 . 42 Table 5 . Crowd annotation agreement with researcher annotations K = 1 % 2 % 5 % 10 % 15 % 20 % 25 % all words . 67 ( . 03 ) . 67 ( . 06 ) . 46 ( 11 ) . 37 ( . 18 ) . 35 ( . 25 ) . 31 ( . 29 ) . 29 ( . 34 ) bkground / purpose + mechanism Upwork . 75 ( . 04 ) . 67 ( . 06 ) . 64 ( . 15 ) . 43 ( . 20 ) . 39 ( . 28 ) . 33 ( . 31 ) . 31 ( . 37 ) MTurk . 75 ( . 04 ) . 63 ( . 06 ) . 57 ( . 14 ) . 43 ( . 20 ) . 34 ( . 24 ) . 32 ( . 30 ) . 29 ( . 34 ) purpose + mechanism Upwork . 83 ( . 04 ) . 75 ( . 07 ) . 54 ( . 13 ) . 42 ( . 20 ) . 34 ( . 24 ) . 33 ( . 31 ) . 30 ( . 35 ) MTurk . 83 ( . 04 ) . 71 ( . 07 ) . 52 ( . 13 ) . 42 ( . 20 ) . 37 ( . 26 ) . 32 ( . 30 ) . 31 ( . 36 ) mechanism Upwork . 75 ( . 04 ) . 58 ( . 06 ) . 51 ( . 12 ) . 39 ( . 19 ) . 33 ( . 23 ) . 31 ( . 30 ) . 30 ( . 36 ) MTurk . 75 ( . 04 ) . 54 ( . 05 ) . 44 ( . 11 ) . 34 ( . 16 ) . 32 ( . 23 ) . 29 ( . 27 ) . 28 ( . 33 ) Table 6 . Proportion of known analogies found at each level of K for each similarity metric , with annotations by Upwork and MTurk crowd workers ( recall in parentheses ; best - performing scores bold - underlined . Our bkground / purpose + mechanism and purpose + mechanism metrics continue to outperform the all words baseline ( with substantial reductions from researcher - based annotations ) , while the mechanism metric\u2019s advantage is almost entirely eliminated for both Upwork and MTurk annotations . workers\u2019 annotations matched the researcher annotations 78 % of the time , and MTurk workers\u2019 an - notations matched 59 % of the time . There was considerable variation across papers , with accuracies as high as 96 % agreement for some papers , and as low as 4 % for others . 6 . 2 . 2 Crowds struggled with purpose and mechanism annotations . In general , both groups of workers struggled most with purpose and mechanism annotations ( 68 % and 72 % for the Upwork workers , and 42 % and 53 % agreement for the MTurk workers ) . First , workers often confused background and purpose . The most frequent error for background annotations was incorrectly annotating it as purpose ( 13 % of background annotations for Upwork ; 21 % for MTurk ) . Purpose was also frequently confused for background ( 10 % of purpose annotations for Upwork ; 16 % for MTurk ) . Mechanism was also frequently confused for purpose and findings : the most frequent error for purpose was incorrectly annotating it as mechanism ( 17 % of purpose annotations for Upwork ; 28 % for MTurk ) , while mechanism was confused for purpose and findings quite frequently as well ( 8 % and 9 % of mechanism annotations for Upwork ; 10 % and 31 % for MTurk ) . One possible explanation is that researchers frequently mix multiple aspects in their sentences : for example , they often described the purpose and mechanism of their paper in the same sentence , or described the mechanism in more detail by claiming that it worked ( findings ) . We return to this point in the discussion to consider how our scheme might be modified to account for variations in writing style . 6 . 2 . 3 Crowd annotations still improve analogy - mining . As Table 6 shows , despite the modest levels of agreement , specialized vectors constructed from crowd workers\u2019 aggregated annotations were still able to support bkground / purpose + mechanism and purpose + mechanism metrics that Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 16 G . Zhou et al . outperformed the all words baseline , albeit with substantial reductions in the size of the advantage . In particular , the purpose + mechanism metric outperforms all words throughout the range of K , for both Upwork and MTurk deployments . Note that the advantage of the specialized metrics is reduced the least for bkground / pur - pose + mechanism ( retaining substantial advantages over all words , and even outperforming purpose + mechanism , past K = 2 ) , while the advantage of the mechanism metric disappears past K = 1 . The generally low accuracy for purpose and mechanism annotations might explain these patterns , since the purpose + mechanism and mechanism metrics depend critically on these annotations . 7 DISCUSSION 7 . 1 Summary and Implications of Findings Across three studies , we show that applying solvent to research papers yields a substantial improvement ( approximately 25 - 30 % overall improvement in Study 1 ) over a state - of - the - art infor - mation retrieval ( IR ) technique ( TF - IDF - weighted average semantic vectors ) for finding analogies between those research papers . The advantage holds for near analogies ( where standard IR is expected to perform well ) , and across domains ( seen especially in Study 2 with the mechanical engineering research group ) . We further show that these annotations do not require extensive con - tent / domain expertise to be useful : researchers can produce useful annotations for papers outside their discipline ( Study 2 ) , and naive crowd workers can produce useful annotations for research papers ( Study 3 ; albeit with noticable reductions in effectiveness compared to researcher - sourced annotations ) . Across these deployments , we find that the time cost of annotation is approximately a minute per paper . Our modified annotation scheme ( and the usefulness of partialing out background , or leveraging findings ) points to solvent \u2019s core insight ( which extends Hope et al\u2019s [ 22 ] approach ) : \" dissolving \" documents into relational elements ( e . g . , purpose , mechanism , findings ) that have in - built causal relations with each other enables us to create soft relational schemas that are useful for analogical matching . This insight might prove useful in other domains than research papers : for example , annotating the precedents , facts , and decisions in legal cases to support case law reasoning , annotating student learning goals and associated exercises and / or examples in lesson plans to support innovation in teaching , or annotating plot trajectory and associated tropes in writing . While we suspect that solvent might be useful in many of these domains , we think that domains where these dimensions are hidden / obscured by surrounding texts that serve many other functions ( e . g . , preambles in legal opinions ) might especially benefit from solvent \u2019s soft schema extraction techniques . This insight opens up a design space that could help resolve the cost - accuracy tradeoff for computational analogy : rather than spending prohibitively high amounts of resources to specify rich relational structures manually ( thereby maximizing accuracy for analogical matching ) , or completely automating knowledge modeling ( but ignoring structure , thereby losing accuracy for analogical matching ) , we can explore the middle ground of extracting soft relational schemas in a cost - effective manner . This design space mirrors current Semantic Web efforts that seek to relax formality requirements for shared ontologies , and explore how machine - and crowdsourced semantics might complement or even replace more formal and precise ( but infeasible to obtain for the broader Web ) ontologies [ 4 ] . 7 . 2 Limitations and Future Work 7 . 2 . 1 Extending to larger datasets . While our results show significant promise , we acknowledge that our datasets are relatively small . This is partially mitigated by the diversity of domains represented : our data includes papers from diverse domains and contribution types / framings ( e . g . , Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 17 civil engineering papers on numerical analysis of beam strength used a different structure than empirical CHI paper abstracts ) . However , we do aim to test this approach across larger and more diverse corpora . Part of the bottleneck is having a good set of gold standard matches that we can use to evaluate our approach : conservatively , many datasets might contain approximately 5x as many analogy pairs as there are papers ( similar to the 259 known analogy pairs in our set of 50 papers in Study 1 ) ; finding analogies by hand remains a costly process , and having relatively complete coverage of known analogies is critical for meaningful precision and recall performance measures . However , alternative evaluations ( e . g . , experiments with searchers , real - world prototype deployments ) are possible . Extending the method to larger datasets may require increases in the scalability of the method . However , our findings with the crowdsourcing study suggest that obtaining annotations for more realistic - sized datasets may not be cost - prohibitive : for example , our deployments suggest that annotating a corpus of 10 , 000 papers would require on the order of 10 \u2013 30 , 000 person - minutes ( or 166 - 500 person - hours ) of work . This suggests that such a corpus could be annotated for as little as under $ 4 , 000 ( assuming a single trained full - time employee working at $ 20 / hr for 1 - 2 months , similar to one of our Upwork workers ) , quite possibly a significant bargain given the high potential upside of transformative cross - disciplinary breakthroughs and reductions in redundant work and dead ends . This work could also be crowdsourced for a comparable cost ( and significantly quicker turnaround time ) , although more sophisticated aggregation and quality control mechanisms than majority vote ( see , e . g . , [ 39 ] ) , or worker - centric screening / training strategies [ 34 ] might be necessary for crowdsourcing efforts similar to our MTurk deployment . Further , with a few thousand papers , ( semi ) supervised machine learning models like recurrent neural networks ( e . g . , as in Hope et al [ 22 ] ) or conditional random - field models might be trained to replace or augment human judgments . As noted earlier , we are also intrigued by the possibility of more \u201ccommunity - sourced\u201d models of crowdsourcing , where community members contribute semantic annotations as a seamless part of the primary task they are motivated to do ( e . g . , bookmarking important steps in a how - to video [ 25 ] , or arranging ideas on a virtual whiteboard in the natural course of brainstorming [ 40 ] ) . In this spirit , could we design a future where the many hours that scientists spend reading , annotating , organizing , and summarizing research papers ( e . g . , for literature reviews , or peer review ) could contribute not just to their primary task of doing research , but also towards supporting a computational infrastructure for knowledge sharing ? These approaches could pave the way towards more self - sustainable computational infrastructures for knowledge sharing . Note also that in some fields ( e . g . , human factors , biomedical research ) , structured abstracts are commonly produced by researchers , suggesting that the publication process could be modified slightly to author - source these annotations at a tolerable / minimal cost to authors , particularly if the value of these annotations was made apparent to the authors . 7 . 2 . 2 Ensuring unbiased coverage . Recall that we modified the annotation scheme ( e . g . , adding a findings annotation ) to reflect the different epistemological goals that exist among research papers ( e . g . , understanding vs . system - building ) . We had mixed success with the findings annotations : at the highest level of matching ( top 1 % of matches ) , the metric outperformed the all words baseline , but not at higher levels of K . One possible explanation is that our modest - sized dataset didn\u2019t contain enough \" same - findings \" analogies to have high precision at higher levels of K . Another possible explanation is that our core insight of getting a schema \" for free \" by identifying separate sets of annotations that are joined by a causal relation glosses over the relational structure within an annotation type - in the case of understanding - oriented papers , the relevant schema may consist of relations between objects as hypothesized in the purpose statement , or in the findings ( e . g . , people Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 18 G . Zhou et al . decide when to self - disclose on social media depending on how the disclosure might reflect on their online identity ) . Relatedly , our conversations with the mechanical engineering researchers revealed that the properties of the objects ( e . g . , the particular ways that polymers respond to physical loads ) are a key aspect of identifying relevant analogies to work in other domains . Unfortunately , these properties are often left implicit in the text . This challenge could be addressed by augmenting the vector representations of the documents with key properties ( e . g . , from knowledge bases like ConceptNet 7 and CYC 8 ) . How to select which among the numerous possible properties of an entity might be useful for analogical matching would be an interesting and challenging problem for future work ; some recent work is already beginning to explore this for consumer product descriptions [ 17 ] More generally , we want to ensure that relying on this annotation schema does not cause us to systematically miss or misrepresent the contributions of papers written in different genres ( e . g . , that don\u2019t emphasize explaining mechanisms , or emphasize reporting findings instead of clearly stating the problem being addressed ) , or with different kinds of contributions ( e . g . , conceptual / theory or review papers ) , or levels of writing / communication quality . We did notice some limitations in our deployments , particularly in trying to apply a purpose / mechanism to review / survey papers , when mechanisms were dropped in favor of motivating the problem in shorter abstracts ( possibly with strict word limits ) , or when the research problem was not clearly / explicitly stated ; anecdotally , we noticed that this last writing style ( missing problem statements , but rich descriptions of findings and methods ) seemed to be especially common in our materials science and engineering papers ; this might partially explain the lower performance of our metrics in Study 2 ( in addition to the difficulty of explicitly searching in randomly sampled papers from other domains ) . Future work that builds on this should explore how we might expand the annotation schemes to fit different genres / contributions ( while retaining the central insight of identifying soft schemas ) , or supplement the knowledge models constructed by our methods with alternative methods that are less dependent on the way a paper is written . 7 . 2 . 3 Usefulness in Real - World Settings . Finally , while our metrics outperformed our instantiation of a traditional state - of - the - art content - based information retrieval approach ( i . e . , leveraging word embedding representations of the whole abstracts ) , our absolute precision was high only for very conservative ( low ) levels of K ( e . g . , recommending only the top 1 % most similar pairs as analogies ) . In smaller search spaces or for niche topics with fewer \" true \" matches , potential users of our approach might have to work harder to find good matches ; in many cases , however , we believe that even returning the top 1 % most similar matches might return more potential matches than the average user would want to sift through . Our approach also differs significantly from other production systems in that it ignores other important signals available from the citation graph ( e . g . , quality , relatedness ) . In practice , we think it would be useful to combine our content - based approach with graph - based approaches [ 42 ] , e . g . , re - ranking with graph - based signals after retrieving initial candidates via our approach , or re - ranking using our approach after retrieving initial candidates via graph - based approaches . 8 CONCLUSION In this paper , we introduced solvent , a mixed - initiative system for re - representing and finding analogies between research papers across different domains , in which humans produce lightweight annotations of key aspects of research paper abstracts , and a computational model constructs semantic vectors from the annotations and uses them to finds analogies between the papers . We 7 http : / / conceptnet . io / 8 http : / / www . opencyc . org / Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 19 show that solvent holds promise for efficiently finding analogies between research papers across domains , opening up novel pathways towards computationally augmenting knowledge sharing for scientific progress . 9 ACKNOWLEDGEMENTS This work was funded by NSF grants CHS - 1526665 and CHS - 1816242 , ISF grant 1764 / 15 , the HUJI Cyber Security Research Center in conjunction with the Israel National Cyber Bureau in the Prime Minister\u2019s Office , Carnegie Mellon University\u2019s Web2020 Initiative , and Bosch Research Institute . We thank Yla Tausczik and Ping Wang for comments on earlier drafts of this manuscript . REFERENCES [ 1 ] Paul Andr\u00c3\u013e , Haoqi Zhang , Juho Kim , Lydia Chilton , Steven P . Dow , and Robert C . Miller . 2013 . Community clustering : Leveraging an academic crowd to form coherent conference sessions . In First AAAI Conference on Human Computation and Crowdsourcing . [ 2 ] Ryan Arlitt , Friederich Berthelsdorf , Sebastian Immel , and Robert B . Stone . 2014 . The Biology Phenomenon Categorizer : A Human Computation Framework in Support of Biologically Inspired Design . Journal of Mechanical Design ( 2014 ) . https : / / doi . org / 10 . 1115 / 1 . 4028348 [ 3 ] Michele Banko , Michael J Cafarella , Stephen Soderland , Matthew Broadhead , and Oren Etzioni . 2007 . Open Information Extraction from the Web . . In IJCAI , Vol . 7 . 2670 \u2013 2676 . [ 4 ] Abraham Bernstein , James Hendler , and Natalya Noy . 2016 . A New Look at the Semantic Web . Commun . ACM 59 , 9 ( Aug . 2016 ) , 35 \u2013 37 . https : / / doi . org / 10 . 1145 / 2890489 [ 5 ] Chandra Bhagavatula , Sergey Feldman , Russell Power , and Waleed Ammar . 2018 . Content - based citation recommenda - tion . arXiv preprint arXiv : 1802 . 08301 ( 2018 ) . [ 6 ] David M . Blei , Andrew Y . Ng , Michael I . Jordan , and John Lafferty . 2003 . Latent Dirichlet Allocation . Journal of Machine Learning Research ( 2003 ) , 993 \u2013 1022 . [ 7 ] Jonathan Bragg and Daniel S . Weld . 2013 . Crowdsourcing Multi - Label Classification for Taxonomy Creation . In First AAAI Conference on Human Computation and Crowdsourcing . [ 8 ] Joseph C . Chang , Aniket Kittur , and Nathan Hahn . 2016 . Alloy : Clustering with crowds and computation . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . ACM . [ 9 ] Lydia B . Chilton , Juho Kim , Paul Andr\u00c3\u013e , Felicia Cordeiro , James A . Landay , Daniel S . Weld , Steven P . Dow , Robert C . Miller , andHaoqiZhang . 2014 . Frenzy : CollaborativeDataOrganizationforCreatingConferenceSessions . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201914 ) . ACM , New York , NY , USA , 1255 \u2013 1264 . https : / / doi . org / 10 . 1145 / 2556288 . 2557375 [ 10 ] Lydia B . Chilton , Greg Little , Darren Edge , Daniel S . Weld , and James A . Landay . 2013 . Cascade : Crowdsourcing taxonomy creation . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . ACM , 1999 \u2013 2008 . https : / / doi . org / 10 . 1145 / 2470654 . 2466265 [ 11 ] Paolo Ciccarese , Elizabeth Wu , Gwen Wong , Marco Ocana , June Kinoshita , Alan Ruttenberg , and Tim Clark . 2008 . The SWAN biomedical discourse ontology . Journal of Biomedical Informatics 41 , 5 ( Oct . 2008 ) , 739 \u2013 751 . https : / / doi . org / 10 . 1016 / j . jbi . 2008 . 04 . 010 [ 12 ] Tim Clark , Paolo N . Ciccarese , and Carole A . Goble . 2014 . Micropublications : a semantic model for claims , evidence , arguments and annotations in biomedical communications . Journal of Biomedical Semantics 5 ( July 2014 ) , 28 . https : / / doi . org / 10 . 1186 / 2041 - 1480 - 5 - 28 [ 13 ] Scott Deerwester , Susan T . Dumais , Geroge W . Furnas , and Thomas K . Landauer . 1990 . Indexing by Latent Semantic Analysis . JASIST 41 , 6 ( 1990 ) , 1990 . [ 14 ] Brian Falkenhainer , Kenneth D Forbus , and Dedre Gentner . 1989 . The structure - mapping engine : Algorithm and examples . Artificial intelligence 41 , 1 ( 1989 ) , 1 \u2013 63 . [ 15 ] Dedre Gentner . 1983 . Structure - Mapping : A Theoretical Framework for Analogy * . Cognitive science 7 , 2 ( 1983 ) , 155 \u2013 170 . [ 16 ] M . L . Gick and K . J . Holyoak . 1983 . Schema induction and analogical transfer . Cognitive Psychology 15 , 1 ( 1983 ) , 1 \u2013 38 . [ 17 ] Karni Gilon , Joel Chan , Felicia Y Ng , Hila Lifshitz Assaf , Aniket Kittur , and Dafna Shahaf . 2018 . Analogy Mining for Specific Design Needs . In Proceedings of the 2018 ACM SIGCHI Conference on Human Factors in Computing . [ 18 ] Nathan Hahn , Joseph Chang , Ji Eun Kim , and Aniket Kittur . 2016 . The Knowledge Accelerator : Big Picture Thinking in Small Pieces . In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems ( CHI \u201916 ) . ACM , New York , NY , USA , 2258 \u2013 2270 . https : / / doi . org / 10 . 1145 / 2858036 . 2858364 Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . 31 : 20 G . Zhou et al . [ 19 ] Silvana Hartmann , Ilia Kuznetsov , Teresa Martin , and Iryna Gurevych . 2017 . Out - of - domain FrameNet Semantic Role Labeling . In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics : Volume 1 , Long Papers , Vol . 1 . 471 \u2013 482 . [ 20 ] Qi He , Jian Pei , Daniel Kifer , Prasenjit Mitra , and Lee Giles . 2010 . Context - aware Citation Recommendation . In Proceedings of the 19th International Conference on World Wide Web ( WWW \u201910 ) . ACM , New York , NY , USA , 421 \u2013 430 . https : / / doi . org / 10 . 1145 / 1772690 . 1772734 [ 21 ] K . J . Holyoak and P . Thagard . 1996 . The analogical scientist . In Mental Leaps : Analogy in Creative Thought , K . J . Holyoak and P . Thagard ( Eds . ) . Cambridge , MA , 185 \u2013 209 . [ 22 ] Tom Hope , Joel Chan , Aniket Kittur , and Dafna Shahaf . 2017 . Accelerating Innovation Through Analogy Mining . In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . ACM , 235 \u2013 243 . [ 23 ] John E Hummel and Keith J Holyoak . 2003 . A symbolic - connectionist theory of relational inference and generalization . Psychological review 110 , 2 ( 2003 ) , 220 . [ 24 ] Benjamin F . Jones . 2009 . The Burden of Knowledge and the Death of the Renaissance Man : Is Innovation Getting Harder ? Review of Economic Studies 76 , 1 ( 2009 ) , 283 \u2013 317 . https : / / doi . org / 10 . 1111 / j . 1467 - 937X . 2008 . 00531 . x [ 25 ] Juho Kim , Phu Tran Nguyen , Sarah Weir , Philip J . Guo , Robert C . Miller , and Krzysztof Z . Gajos . 2014 . Crowdsourcing Step - by - step Information Extraction to Enhance Existing How - to Videos . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201914 ) . ACM , New York , NY , USA , 4017 \u2013 4026 . https : / / doi . org / 10 . 1145 / 2556288 . 2556986 [ 26 ] Scott Kirkpatrick , C Daniel Gelatt , Mario P Vecchi , et al . 1983 . Optimization by simulated annealing . science 220 , 4598 ( 1983 ) , 671 \u2013 680 . [ 27 ] Maria Liakata , Shyamasree Saha , Simon Dobnik , Colin Batchelor , and Dietrich Rebholz - Schuhmann . 2012 . Automatic recognition of conceptualization zones in scientific articles and two life science applications . Bioinformatics 28 , 7 ( April 2012 ) , 991 \u2013 1000 . https : / / doi . org / 10 . 1093 / bioinformatics / bts071 [ 28 ] Maria Liakata , Simone Teufel , Advaith Siddharthan , Colin R Batchelor , and others . 2010 . Corpora for the Conceptuali - sation and Zoning of Scientific Papers . . In LREC . Citeseer . [ 29 ] Yicong Liang , Qing Li , and Tieyun Qian . 2011 . Finding Relevant Papers Based on Citation Relations . In Web - Age Information Management ( Lecture Notes in Computer Science ) . Springer , Berlin , Heidelberg , 403 \u2013 414 . https : / / doi . org / 10 . 1007 / 978 - 3 - 642 - 23535 - 1 _ 35 [ 30 ] Angli Liu , Stephen Soderland , Jonathan Bragg , Christopher H Lin , Xiao Ling , and Daniel S Weld . 2016 . Effective Crowd Annotation for Relation Extraction . . In HLT - NAACL . 897 \u2013 906 . [ 31 ] Salvador E Luria and Max Delbr\u00fcck . 1943 . Mutations of bacteria from virus sensitivity to virus resistance . Genetics 28 , 6 ( 1943 ) , 491 . [ 32 ] Tomas Mikolov , Kai Chen , Greg Corrado , and Jeffrey Dean . 2013 . Efficient Estimation of Word Representations in Vector Space . arXiv : 1301 . 3781 [ cs ] ( Jan . 2013 ) . http : / / arxiv . org / abs / 1301 . 3781 arXiv : 1301 . 3781 . [ 33 ] Tomas Mikolov , Ilya Sutskever , Kai Chen , Greg S Corrado , and Jeff Dean . 2013 . Distributed Representations of Words and Phrases and their Compositionality . In Advances in Neural Information Processing Systems 26 , C . J . C . Burges , L . Bottou , M . Welling , Z . Ghahramani , and K . Q . Weinberger ( Eds . ) . Curran Associates , Inc . , 3111 \u2013 3119 . [ 34 ] Tanushree Mitra , C . J . Hutto , and Eric Gilbert . 2015 . Comparing Person - and Process - centric Strategies for Obtaining Quality Data on Amazon Mechanical Turk . In Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems ( CHI \u201915 ) . ACM , New York , NY , USA , 1345 \u2013 1354 . https : / / doi . org / 10 . 1145 / 2702123 . 2702553 [ 35 ] Jeffrey Pennington , Richard Socher , and Christopher D Manning . 2014 . Glove : Global vectors for word representation . Proceedings of the Empiricial Methods in Natural Language Processing ( EMNLP 2014 ) 12 ( 2014 ) , 1532 \u2013 1543 . [ 36 ] Peter Pirolli and Stuart Card . 1999 . Information foraging . Psychological review 106 , 4 ( 1999 ) , 643 . [ 37 ] Xiang Ren , Jialu Liu , Xiao Yu , Urvashi Khandelwal , Quanquan Gu , Lidan Wang , and Jiawei Han . 2014 . ClusCite : effective citation recommendation by information network - based clustering . In Knowledge Discovery and Data Mining . 821 \u2013 830 . https : / / doi . org / 10 . 1145 / 2623330 . 2623630 [ 38 ] R . Keith Sawyer . 2012 . Explaining creativity : the science of human innovation ( 2nd ed . ) . Oxford University Press , New York . [ 39 ] Aashish Sheshadri and Matthew Lease . 2013 . SQUARE : A Benchmark for Research on Computing Crowd Consensus . In Proceedings of the 1st AAAI Conference on Human Computation ( HCOMP ) . 156 \u2013 164 . http : / / ir . ischool . utexas . edu / square / documents / sheshadri . pdf [ 40 ] Pao Siangliulue , Joel Chan , Bernd Huber , Steven P . Dow , and Krzysztof Z . Gajos . 2016 . IdeaHound : Self - sustainable Idea Generation in Creative Online Communities . In Proceedings of the 19th ACM Conference on Computer Supported Cooperative Work and Social Computing Companion ( CSCW \u201916 Companion ) . ACM , New York , NY , USA , 98 \u2013 101 . https : / / doi . org / 10 . 1145 / 2818052 . 2874335 [ 41 ] David W Stephens and John R Krebs . 1986 . Foraging theory . Princeton University Press . Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 . Mixed Initiative System for Finding Analogies 31 : 21 [ 42 ] TrevorStrohman , W . BruceCroft , andDavidJensen . 2007 . RecommendingCitationsforAcademicPapers . In Proceedings of the 30th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval ( SIGIR \u201907 ) . ACM , New York , NY , USA , 705 \u2013 706 . https : / / doi . org / 10 . 1145 / 1277741 . 1277868 [ 43 ] Yalin Sun , Pengxiang Cheng , Shengwei Wang , Hao Lyu , Matthew Lease , Iain Marshall , and Byron C . Wallace . 2016 . Crowdsourcing Information Extraction for Biomedical Systematic Reviews . In 4th AAAI Conference on Human Compu - tation and Crowdsourcing ( HCOMP ) : Works - in - Progress Track . http : / / arxiv . org / abs / 1609 . 01017 3 pages . arXiv : 1609 . 01017 . [ 44 ] Swaroop Vattam , Bryan Wiltgen , Michael Helms , Ashok K . Goel , and Jeannette Yen . 2011 . DANE : Fostering Creativity in and through Biologically Inspired Design . In Design Creativity 2010 . http : / / link . springer . com / chapter / 10 . 1007 / 978 - 0 - 85729 - 224 - 7 _ 16 [ 45 ] S . Wuchty , B . F . Jones , and B . Uzzi . 2007 . The increasing dominance of teams in production of knowledge . Science 316 , 5827 ( 2007 ) , 1036 \u2013 1039 . [ 46 ] James Zou , Kamalika Chaudhuri , and Adam Kalai . 2015 . Crowdsourcing Feature Discovery via Adaptively Chosen Comparisons . In Third AAAI Conference on Human Computation and Crowdsourcing . Received April 2018 ; revised July 2018 ; accepted August 2018 Proc . ACM Hum . - Comput . Interact . , Vol . 2 , No . CSCW , Article 31 . Publication date : November 2018 .",
    "reinhardt2023angstromresolution": "Nature | Vol 617 | 25 May 2023 | 711 Article \u00c5ngstr\u00f6m - resolution fluorescence microscopy Susanne C . M . Reinhardt 1 , 2 , 5 , Luciano A . Masullo 1 , 5 , Isabelle Baudrexel 1 , 3 , 5 , Philipp R . Steen 1 , 2 , 5 , Rafal Kowalewski 1 , 2 , Alexandra S . Eklund 1 , 3 , Sebastian Strauss 1 , 2 , Eduard M . Unterauer 1 , 2 , Thomas Schlichthaerle 1 , 2 , Maximilian T . Strauss 1 , 2 , Christian Klein 3 , 4 & Ralf Jungmann 1 , 2 \u2709 Fluorescence microscopy , with its molecular specificity , is one of the major characterization methods used in the life sciences to understand complex biological systems . Super - resolution approaches 1 \u2013 6 can achieve resolution in cells in the range of 15 to 20 nm , but interactions between individual biomolecules occur at length scales below 10 nm and characterization of intramolecular structure requires \u00c5ngstr\u00f6m resolution . State - of - the - art super - resolution implementations 7 \u2013 14 have demonstrated spatial resolutions down to 5 nm and localization precisions of 1 nm under certain in vitro conditions . However , such resolutions do not directly translate to experiments in cells , and \u00c5ngstr\u00f6m resolution has not been demonstrated to date . Here we introdue a DNA - barcoding method , resolution enhancement by sequential imaging ( RESI ) , that improves the resolution of fluorescence microscopy down to the \u00c5ngstr\u00f6m scale using off - the - shelf fluorescence microscopy hardware and reagents . By sequentially imaging sparse target subsets at moderate spatial resolutions of > 15 nm , we demonstrate that single - protein resolution can be achieved for biomolecules in whole intact cells . Furthermore , we experimentally resolve the DNA backbone distance of single bases in DNA origami with \u00c5ngstr\u00f6m resolution . We use our method in a proof - of - principle demonstration to map the molecular arrangement of the immunotherapy target CD20 in situ in untreated and drug - treated cells , which opens possibilities for assessing the molecular mechanisms of targeted immunotherapy . These observations demonstrate that , by enabling intramolecular imaging under ambient conditions in whole intact cells , RESI closes the gap between super - resolution microscopy and structural biology studies and thus delivers information key to understanding complex biological systems . The localization precision ( \u03c3 SMLM ) of a target molecule in widefield single - molecule localization microscopy ( SMLM ) 15 is ultimately and fundamentally limited by the number of photons ( N ) collected per blinking event : \u03c3 \u2248 \u03c3 N SMLM DIFF ( \u03c3 DIFF is the s . d . of the point spread function ( PSF ) of the optical imaging system 16 ; Fig . 1a ) . Multiple localizations of the same target ( Fig . 1b , top ) are distributed around the true position due to their finite precision . Two or more points not resolvable by SMLM produce overlapping distributions of localizations , thus pre - cluding unique assignment of localizations to respective targets ( Fig . 1b , bottom ) . However , if each localization could be assigned to a specific target by colour , barcode or any other molecular identity , they could be unambiguously grouped per target 2 . The centre of each group of localizations can be calculated with a precision far better than \u03c3 SMLM . In essence , applying the principle of localization microscopy to distinguishable groups of K super - resolved localizations , precision is increased from the s . d . ( \u03c3 SMLM ) to the s . e . m . ( \u03c3 K SMLM ) . Collecting an arbitrarily large number of localizations yields an arbitrary increase in precision . Notably , this increase in precision occurs regardless of the precision achieved in individual localizations ( \u03c3 SMLM ) . We introduce a straightforward implementation of this concept using Exchange - PAINT 17 , a variant of DNA - PAINT 18 , for identical target molecules ( Fig . 1c ) . DNA - PAINT uses the programmable , repetitive but transient binding of dye - labelled \u2018imager\u2019 strands to their com - plementary \u2018docking\u2019 strands on target molecules of interest 9 , 18 . The transient nature of the binding leads to an apparent \u2018blinking\u2019 of the target , necessary to perform SMLM . Exchange - PAINT uses orthogonal DNA barcodes combined with imaging and washing cycles to allow for sequential target multiplexing . In our implementation we \u2018multiplex\u2019 a single target species by separating it into multiple , sparser subsets . By imaging them sequentially , sufficiently spaced and isolated groups of localizations are measured . Determining the centre of each group of localizations yields a resolution enhancement ( Fig . 1d ) . We call this implementation resolution enhancement by sequential imaging ( RESI ) , and the resulting localizations RESI localizations . https : / / doi . org / 10 . 1038 / s41586 - 023 - 05925 - 9 Received : 21 July 2022 Accepted : 7 March 2023 Published online : 24 May 2023 Open access Check for updates 1 Max Planck Institute of Biochemistry , Planegg , Germany . 2 Faculty of Physics and Center for NanoScience , Ludwig Maximilian University , Munich , Germany . 3 Department of Chemistry and Biochemistry , Ludwig Maximilian University , Munich , Germany . 4 Roche Innovation Center Zurich , Roche Pharma Research and Early Development , Schlieren , Switzerland . 5 These authors contributed equally : Susanne C . M . Reinhardt , Luciano A . Masullo , Isabelle Baudrexel , Philipp R . Steen . \u2709 e - mail : jungmann @ biochem . mpg . de 712 | Nature | Vol 617 | 25 May 2023 Article By application of RESI in silico ( Methods ) , we demonstrated a reso - lution improvement ( Extended Data Fig . 1 ) over super - resolution akin to the improvement of super - resolved over diffraction - limited meas - urements ( Fig . 1e ) . For routinely obtainable DNA - PAINT localization precision ( approximately 3 nm ) and number of localizations per target ( in the order of hundreds ) , RESI could achieve precision well below one nanometre , thus entering the \u00c5ngstr\u00f6m scale ( Fig . 1f ) according to \u03c3 = \u03c3 K RESI SMLM . For an experimental proof of principle of RESI we used self - assembled DNA origami structures to precisely position orthogonal DNA strands 9 , 19 . We first designed DNA origami featuring two docking strands spaced 5 nm apart , a distance previously resolved with DNA - PAINT 7 , 9 , to verify the accuracy and precision of RESI . Using two sequential imag - ing rounds and an alignment procedure ( Methods ) to conduct RESI , we were able to accurately recapitulate the 5 nm point - to - point distance with precision improved by a factor of K = 381 \u224820 average ( Extended Data Figs . 2 and 3 ) . We next performed RESI in three dimensions ( 3D ) using recently developed 3D DNA origami disk structures 20 and measured distances between docking strands of 2 . 5 \u00b1 0 . 4 nm in xy and 11 . 3 \u00b1 0 . 8 nm in z . This demonstrates that RESI resolution enhancement applies in all three dimensions , surpassing current state - of - the - art 3D super - resolution capabilities ( Extended Data Figs . 4 and 5 ; for imaging parameters see Extended Data Table 1 ) . RESI resolves single nuclear pore complex proteins To demonstrate the applicability of RESI in a cellular context , we next imaged structural proteins of the nuclear pore complex ( NPC ) . As the major gatekeeper of nucleocytoplasmic transport , the NPC is a key target for structural biology research 21 . We furthermore chose the NPC as a model system because it has been well studied using a variety of imaging approaches , including cryo - electron microscopy ( cryo - EM ) 22 , fluorescence microscopy and super - resolution techniques 23 , 24 . Figure 2a presents a typical diffraction - limited and DNA - PAINT image of Nup96 molecules ( tagged with monomeric enhanced green fluo - rescent protein ( mEGFP ) ) labelled with DNA - conjugated anti - GFP nanobodies . Nup96 is a structural NPC protein ( part of the so - called Y - complex ) present in eight pairs exhibiting an eight - fold symmetry on both cytoplasmic and nuclear rings , totalling 32 copies per NPC ( Fig . 2b ) . Individual pairs of Nup96 proteins , spaced approximately 10 nm laterally and 3 nm axially , cannot be routinely resolved with current state - of - the - art super - resolution implementations 25 \u2013 28 . To enable RESI , neighbouring copies of Nup96 proteins must be labelled with orthogonal DNA sequences . To this end , we opted for a stochas - tic labelling approach by incubating the sample with anti - GFP nano - bodies , each conjugated with one out of four orthogonal sequences ( Fig . 2c ) . We note that , with an increasing number of expected targets below the classical DNA - PAINT resolution limit , a larger number of orthogonal labelling sequences 29 \u2014and thus imaging rounds\u2014is nec - essary to guarantee sufficiently spaced groups of localizations ( for details on this requirement see Methods ) . Sequential 3D image acqui - sition in four rounds led to sufficiently spaced localization groups representing single Nup96 target molecules ( Fig . 2d ) . Subsequent RESI super - localization of these groups allowed us to routinely visu - alize individual copies of Nup96 proteins ( Fig . 2e ) . We note that this was achieved across the whole field of view ( roughly 67 \u00d7 67 \u00b5m 2 ) totalling over 1 , 000 NPCs during a total image acquisition time of Single - molecule localization RESI a c b d e f (cid:86) RESI \u2248 K (cid:86) SMLM K (cid:86) R E S I ( n m ) Diffraction limited DNA - PAINT RESI 10 40 0 2 1 \u00c5 scale K = 10 K = 40 K = 100 100 300 nm 10 nm 10 nm 2 nm 2 nm 2 nm d 2 < 2 . 35 (cid:86) SMLM d 1 > > (cid:86) SMLM ~ DNA - PAINT d 1 \u2248 20 nm d 2 \u2248 2 nm (cid:86) SMLM (cid:86) SMLM \u2248 N (cid:86) DIFF (cid:86) DIFF 10 nm 1 \u03bc m 200 nm Sequential super - resolution K localizations I m a g i n g r o un d s (cid:86) SMLM (cid:86) RESI 10 nm 10 nm Round 1 Round 2 Sequence 1 Sequence 2 Round n Round 2 Round 1 Fig . 1 | RESI concept . a , In SMLM , \u03c3 SMLM of a single dye scales with \u03c3 N DIFF , ultimately limiting the achievable spatial resolution . b , SMLM approaches such as DNA - PAINT feature approximately 10 nm spatial resolution ( resolution approximated as full - width at half - maximum \u2248 2 . 35 \u03c3 SMLM ) . Whereas targets separated by 20 nm ( d 1 ) can thus be routinely resolved , objects spaced 2 nm apart ( d 2 ) are unresolvable because the resulting distributions of localizations overlap . c , Using orthogonal DNA sequences ( blue and green ) and sequential acquisition as in Exchange - PAINT , localizations from targets spaced more closely than the SMLM resolution limit can be unambiguously assigned for each target . d , Combining all localizations per target ( K ) for each imaging round improves localization precision from s . d . ( \u03c3 SMLM ) to s . e . m . ( \u03c3 RESI ) . e , As super - resolution revolutionized fluorescence microscopy , RESI results in another paradigm shift by reapplying the concept of localization to super - resolution data . f , Localization precision in RESI scales with K 1 , and thus resolution improvement in RESI is independent of \u03c3 SMLM , reaching localization precision on the \u00c5ngstr\u00f6m scale . Nature | Vol 617 | 25 May 2023 | 713 100 min ( see Extended Data Fig . 6 for representative data ) . The recon - structed RESI image features an average lateral localization precision of approximately 1 nm , representing a sixfold improvement over the individual DNA - PAINT acquisition rounds . We therefore achieved label - size - limited resolution , allowing us to resolve individual Nup96 molecules ( Fig . 2f ) . Generally , label size not only limits spatial reso - lution but furthermore could lead to inaccuracies such as biased observed distances due to linkage errors . We then performed unbiased 3D averaging of 1 , 217 NPCs using a recently developed model - free approach for SMLM data 30 . The result - ing 3D average ( Fig . 2g ) not only allows recapitulation of the eight - fold symmetry of Nup96 in both cytoplasmic and nuclear rings ( which has previously been achieved with super - resolution 23 \u2013 28 ) , but enables reso - lution of individual Nup96 proteins in a structural average ( Fig . 2h ) . Enabled by RESI\u2019s unprecedented spatial resolution , we were able to recapitulate distances of Nup96 proteins of 11 . 9 \u00b1 1 . 2 nm laterally and 5 . 4 \u00b1 0 . 4 nm axially from the structural average image ( Fig . 2i ) . Both lateral and axial orientation , as well as tilt , of Nup96 pairs are consist - ent with cryo - EM data 22 . We resolved this spatial arrangement for most Nup96 protein pairs ( Extended Data Fig . 7 ) , which was previously out of reach for optical microscopy . Imaging DNA bases at \u00c5ngstr\u00f6m resolution To assay the ultimately achievable spatial resolution by RESI , we designed a flat , rectangular DNA origami structure featuring six pairs ( spaced 20 nm apart ) of directly adjacent orthogonal docking strands at a distance of only one DNA base pair ( red and blue strands in Fig . 3a ) . This yielded a designed in - plane distance of around 7 \u00c5 along the backbone of one strand of the DNA double helix 31 . The structures also contain DNA - PAINT docking strands for precise alignment between sequential imaging rounds ( green strands in Fig . 3a ) . State - of - the - art DNA - PAINT image acquisition 32 at approximately 5 nm spatial reso - lution yielded six localization clouds in a 20 nm grid arrangement but failed to resolve the individual docking strands at subnanometre single - base - pair distances ( Fig . 3b ) . Remarkably , RESI resolves the individual docking strand positions ( Fig . 3c ) in all DNA origami structures . We note that RESI achieved this D i ff r ac t i o n li m i t e d D N A - PA I N T NR CR NR 0 80 Side view 0 40 Position ( nm ) 0 30 P o s i t i o n ( n m ) (cid:86) \u2248 5 \u00c5 (cid:86) \u2248 10 \u00c5 (cid:86) \u2248 10 \u00c5 3D RESI 3 D D N A - P A I N T 3 D R E S I Round 1 Round 2 Round 3 Round 4 Nup96 - GFP nanobody CR NR 11 . 9 nm 5 . 4 nm 20 NPC NPC structure Stochastic labelling Z p o s i t i o n ( n m ) 5 \u03bc m 500 nm a b c d Sequential 3D DNA - PAINT imaging 20 nm 5 nm Side NR CR 20 nm 20 \u00c5 20 nm e f g h i CR 20 nm x y z z y x x y z 3 . 2 nm 1 0 . 3 n m Fig . 2 | NPC proteins in whole cells resolved with \u00c5ngstr\u00f6m precision by RESI . a , Diffraction - limited and DNA - PAINT overview image of Nup96 - mEGFP labelled with DNA - conjugated anti - GFP nanobodies . Zoomed - in view ( bottom right ) shows high labelling efficiency and image quality for standard DNA - PAINT conditions , recapitulating the eight - fold symmetry of the NPC . b , Cryo - EM structure representation of the location of Nup96 proteins ( red ; C - terminal mEGFP position marked in blue ) as part of the Y - complex in nuclear and cytoplasmic rings ( NR and CR , respectively ) . Adapted from PDB 7PEQ . Nup96 is present in 32 copies per NPC . c , To enable RESI , Nup96 - mEGFP proteins were stochastically labelled with orthogonal DNA sequences by incubation of the sample with anti - GFP nanobodies , each conjugated with one of four orthogonal sequences ( represented by blue , yellow , magenta and green dots ) . d , Sequential 3D imaging ( colour represents z position ) of the four labels yielded sufficiently spaced localization distributions . The average number of localizations per target is K average = 38 ( background represents cryo - EM structure from b for context ) . e , Comparison of 3D DNA - PAINT ( top left ) and 3D RESI ( bottom right ) for the same NPC illustrating improvement in spatial resolution by RESI . Localizations are rendered as gaussians with \u03c3 DNA - PAINT and \u03c3 RESI , respectively . f , Localization precision ( \u03c3 RESI ) as good as 5 \u00c5 was achieved by combining K localizations for each target , unambiguously resolving single Nup96 proteins . g , The 3D NPC cryo - EM structure was recapitulated using optical microscopy by applying a model - free average 30 of 1 , 217 NPCs from a single nucleus . h , RESI resolved adjacent Nup96 in a structural average by optical microscopy . i , Consistent with the cryo - EM structure ( taking into account linkage error arising from label size ) , adjacent Nup96 proteins were spaced 11 . 9 \u00b1 1 . 2 nm apart laterally ( top ) and 5 . 4 \u00b1 0 . 4 nm axially ( bottom ) . 714 | Nature | Vol 617 | 25 May 2023 Article in an image acquisition time of 100 min featuring an approximately 67 \u00d7 67 \u00b5m 2 field of view containing more than 2 , 000 DNA origami structures ( see Extended Data Fig . 8 for representative DNA origami structures ) . RESI allows us to routinely resolve strands spaced apart by only one DNA base pair . Strikingly , we measured a distance of 8 . 5 \u00b1 1 . 7 \u00c5 between two single docking strands in an individual DNA origami struc - ture ( Fig . 3d ) . This demonstrates an unprecedented resolution in optical microscopy by distinguishing structures closer than one nanometre . We note that our resolution claim is based on the most fundamental and strict definition : the ability to spatially distinguish point objects . We measured a distance of 9 . 5 \u00b1 2 . 6 \u00c5 between adjacent docking strands in an average of 42 DNA origami ( Extended Data Fig . 9 ) , which is within 1 s . d . of the expected backbone distance 31 of around 7 \u00c5 . To quantify resolution gain , we calculated RESI localizations for different values of K underlying DNA - PAINT localizations ( Methods ) . We demonstrate that the effective localization precision scales as \u03c3 = \u03c3 K RESI SMLM , yielding an average localization precision of 1 . 3 \u00c5 for an average K = 254 ( Fig . 3e ) , experimentally confirming the in silico results ( Fig . 1f ) . RESI not only yields virtually unlimited numbers of localiza - tions per target , but also avoids detrimental photophysical effects caused by spatial proximity of fixed - dye labels because , in DNA - PAINT imaging , two adjacent dyes are never present simultaneously . It has recently been reported 33 that , at sub - 10 - nm distances , photophysical near - field interactions play a major role in modulation of photoswitch - ing kinetics , thus effectively preventing fixed - dye SMLM techniques from accessing this resolution scale . This ultimately limits the achiev - able resolution of even the most photon - efficient techniques available for single - molecule localization , such as MINFLUX or MINSTED , despite their subnanometre precision , unless combined with DNA - PAINT . The experimentally demonstrated subnanometre resolution illustrates the capacity of RESI to enable structural biology studies using DNA - based imaging at hitherto elusive scales . CD20 receptor organization Finally , we applied RESI to address and resolve a cell - biological ques - tion currently under debate that has so far been beyond reach for both cryo - EM in a native cellular context and incumbent super - resolution techniques . Specifically we studied the organization of CD20 mem - brane receptors , which are prime targets for therapeutic antibody treatment of B cell - derived blood cancers and autoimmune diseases 34 . In the case of the most frequently used therapeutic anti - CD20 antibody , rituximab ( RTX ) , the spatial rearrangement of CD20 in the cell membrane is thought to play an important role in its efficacy 35 , 36 . Recent cryo - EM studies detected CD20 as a dimer in complex with two individual RTX - fragment antigen - binding regions 37 , 38 , suggest - ing a linear chain - like assembly of CD20 in the presence of the full antibody 38 . On the other hand , when incubated with the full RTX anti - body , a trimeric ring of alternating RTX molecules and CD20 dimers was detected in EM images 37 . The fact that cryo - EM experiments are performed in detergent solution raises the question about which molecular arrangements are actually present in the cell . Currently CD20 organization when bound to full RTX antibodies in intact cells cannot be assessed , thus precluding the investigation of whether CD20 clusters are of linear or circular nature . Moreover , even though in vitro studies showed that CD20 dimers can form without antibody binding , the quantitative assessment of CD20 dimerization in untreated cells is currently limited . Here we applied RESI to study the molecular arrangement of CD20 in Chinese hamster ovary ( CHO ) cells transiently transfected with mEGFP - CD20 , using four rounds of probe exchange in a total imag - ing time of 4 . 4 h . In the diffraction - limited overview and DNA - PAINT super - resolution image of untreated cells , CD20 appeared homogene - ously distributed ( Fig . 4a , b ( top ) and Extended Data Fig . 10a ) whereas RTX - treated cells exhibited apparent CD20 clusters ( Fig . 4b ( bottom ) and Extended Data Figs . 11a and 12a ) . Comparison of DNA - PAINT and RESI for both untreated and RTX - treated cells shows sub - 10 - nm - spaced CD20 pairs in the RESI images ( Fig . 4c , right ) that were unresolvable with DNA - PAINT ( Fig . 4c , left ) . RESI images suggest that CD20 is present in dimers and chain - like , higher - order structures in untreated and RTX - treated cells , respectively ( Fig . 4d ) . To quantitatively assess the existence of dimers in untreated cells , we performed first nearest - neighbour distance ( NND ) analysis for both DNA - PAINT and RESI data , demonstrating nonrandom distribu - tions in both cases ( Fig . 4e ) . RESI at 1 nm localization precision shows a substantial fraction of sub - 10 - nm distances in the NND histogram , which enables quantitative assessment of the degree of CD20 dimeri - zation . We performed numerical simulations and a least - squares fit ( Methods ) that yielded a composition of 53 \u00b1 1 % monomers and 47 \u00b1 1 % dimers with average intradimer distance of 13 . 5 \u00b1 0 . 3 nm ( Fig . 4f , solid line ) . For comparison , we simulated NND distributions corre - sponding to a population of 100 % monomers ( Fig . 4f , dotted line ) , 7 \u00c5 a Round 1 Round 2 b c DNA - PAINT RESI d K (cid:86) R E S I ( \u00c5 ) e 20 15 10 5 K = 5 K = 50 K = 100 1 50 100 9 0 n m 6 0 n m 8 . 5 \u00c5 (cid:86) \u2248 1 . 2 \u00c5 5 \u00c5 DNA origami structure 300 250 K = 254 20 nm 1 bp 5 nm 5 nm 5 nm 3 . 4 \u00c5 0 Fig . 3 | RESI resolves the distance of single DNA base pairs at \u00c5ngstr\u00f6m resolution . a , DNA origami with docking strands spaced by a single base pair ( bp ; red and blue strands , with alignment markers in green ) provided a platform to demonstrate the highest resolution achievable by RESI . b , DNA - PAINT resolved 20 nm spacing but the resolution was insufficient to distinguish individual docking sites , spaced one base apart . c , RESI resolves the adjacent docking strands . d , A Euclidean distance of 8 . 5 \u00b1 1 . 7 \u00c5 was calculated from individual localizations with an average precision of 1 . 2 \u00c5 ( left ) for the single - base - pair backbone distance , which is within 1 s . d . of the expected value of roughly 7 \u00c5 ( right ) . e , Experimental localization precision in RESI is in good agreement with \u03c3 K SMLM ( blue line , K ) , yielding an average localization precision of 1 . 3 \u00c5 for the experimental data from all n = 42 DNA origami ( insets correspond to exemplary point pair in d ) . Error bars represent mean \u00b1 1 s . d . Nature | Vol 617 | 25 May 2023 | 715 further demonstrating that CD20 molecules are not present solely as monomers . Because all NND distributions except for the first order are consistent with a complete spatial random ( CSR ) distribution at the experimentally measured density , we exclude the presence of higher - order assemblies for untreated CD20 ( Fig . 4g ) . Our findings present quantitative experimental evidence that CD20 exists as dimers in an intact cell membrane . By contrast , RESI analysis of CD20 in RTX - treated cells yielded first to fourth NND distributions inconsistent with a CSR model ( Fig . 4h and Extended Data Fig . 12d , e ) . This suggests a higher - order arrange - ment of CD20 molecules after RTX treatment and confirms recent cryo - EM - derived models 37 , 38 . Finally we probed the existence of hexameric , ring - like arrangements by comparison with numerical simulations ( Extended Data Fig . 13 ) . The characteristics of the experimentally detected CD20 clusters sug - gest the absence of isolated hexamers and support the hypothesis of predominantly linear , chain - like structures ( Extended Data Fig . 13h ) . Discussion We introduce RESI , a conceptually new approach in SMLM to improve the spatial resolution of optical microscopy to the \u00c5ngstr\u00f6m scale . RESI achieves this by combining multiple localizations from single targets to obtain a \u2018super - super - resolution\u2019 image after separating their localiza - tions by sequential imaging ( for example , using DNA - barcoded probes ) . In this way RESI precision\u2014and thus resolution\u2014scales not only with the number of photons ( N ) detected per localization but also with the number of localizations ( K ) acquired per target . RESI thus provides a new precision scaling law : \u03c3 = \u2248 \u03c3 K \u03c3 K N RESI SMLM DIFF . This applies if a suffi - ciently large number of orthogonal labelling sequences and thus imag - ing rounds guarantee adequately spaced groups of localizations ( Extended Data Fig . 14 ) . Importantly , resolution enhancement is isotropic in three dimensions . For our current experimental implementation , RESI approaches structural biology resolution with an all - optical approach in intact cells using off - the - shelf labelling reagents and a simple inverted fluorescence microscope operated under ambient conditions . We were able to experimentally demonstrate \u00c5ngstr\u00f6m spatial resolution below the physical size of a dye . This was achieved due to three specific advan - tages of DNA - PAINT leading to unbiased target sampling : ( 1 ) the rota - tional flexibility of the target - bound docking strand ( even in the case of longer repetitive - sequence motifs 32 ) ; ( 2 ) the freely rotating dipole of the dye attached to the imager sequence ; and ( 3 ) the fact that two adjacent imagers are never present simultaneously . Furthermore , because RESI images are not obtained from single localizations but from groups of localizations per target , the method presents a uniquely robust feature compared with other SMLM tech - niques : it shifts the focus from enhancement of only optical precision ( \u03c3 OPT ) to improvement in overall precision ( \u03c3 \u03c3 \u03c3 \u2248 + SMLM OPT2 MEC2 ) by averaging out the uncertainty effects of mechanical instability ( \u03c3 MEC ) , provided the latter is normally distributed . e h f g RTX treated 2 . 35 (cid:86) PAINT N o r m a li z e d f r e q u e n c y 0 20 First NND ( nm ) 40 60 80 100 Untreated Untreated Untreated RESI Untreated RTX treated d dimer d n d CD20 GFP (cid:86) \u2248 5 nm (cid:86) \u2248 0 . 5 nm RESI d dimer DNA - PAINT d CSR b DNA - PAINT DNA - PAINT 200 nm c U n t r ea t e d R T X t r ea t e d DNA - PAINT 4 nm 9 nm 5 nm 8 nm 0 20 First NND ( nm ) 40 60 80 100 0 . 16 0 . 10 0 . 10 0 50 n th NND ( nm ) 100 150 200 0 . 16 0 . 10 0 . 16 0 20 n th NND ( nm ) 40 60 80 100 0 . 04 0 . 04 0 . 04 0 . 05 0 . 03 0 . 01 N o r m a li z e d f r e q u e n c y N o r m a li z e d f r e q u e n c y N o r m a li z e d f r e q u e n c y d n d CSR d dimer DNA - PAINT RESI d dimer Dimers and monomers Monomers d CSR d dimer d CSR First NND SecondNNDThird NND Fourth NND First NND SecondNNDThird NND Fourth NND D i ff r a c t i o n li m i t e d a CD20 receptors D N A - P A I N T 5 \u03bc m 200 nm 50 nm 50 nm 50 nm 50 nm Anti - GFP nanobody 2 nm Fig . 4 | RESI shows CD20 receptor ( re ) organization at subnanometre precision following drug treatment . a , Diffraction - limited and DNA - PAINT overview image of CHO cells expressing mEGFP - CD20 labelled with anti - GFP nanobodies . b , Zoomed - in DNA - PAINT image showing apparently randomly distributed CD20 receptors for untreated cells ( top ) and clustered receptor arrangement for RTX - treated cells ( bottom ) . c , Comparison of DNA - PAINT and RESI for both untreated and RTX - treated cells showing sub - 10 - nm - spaced receptor pairs in the RESI images , which are unresolvable with DNA - PAINT . d , RESI data suggest that CD20 proteins occur in dimers ( spaced at d dimer ) , which are in turn distributed according to complete spatial randomness ( CSR ; distances between dimers , d CSR ) in untreated cells . Chains of dimers were observed following administration of RTX . e , Whole - cell analysis of first NNDs of CD20 receptors ( histograms of distances and kernel density estimation are shown ) . Only RESI , but not DNA - PAINT , allows the routine detection of sub - 10 - nm distances between proteins . Whereas DNA - PAINT overestimates dimer distance , RESI shows a label - limited distance of 13 . 5 nm ( see main text for discussion ) . f , Fitting RESI NND data from e to a numerical model reveals CD20 dimers and monomers . g , CD20 receptors in untreated cells showed second to fourth NNDs consistent with CSR , thus excluding the presence of higher - order protein complexes . h , CD20 receptors in RTX - treated cells , however , showed first to fourth NNDs , inconsistent with complete spatial randomness . 716 | Nature | Vol 617 | 25 May 2023 Article With RESI we measured areas of 67 \u00d7 67 \u00b5m 2 in 100 min , making it applicable as a sufficiently high - throughput tool for cell biology . Resolving receptor patterns at single - protein resolution could enable \u2018spatial diagnostics\u2019 as a prescreening method for personalized treat - ments , and serve as a tool for biomedical discovery of patterned thera - peutics\u2014for example , by guiding drug design principles . RESI performance and accuracy could be further improved by advances in intramolecular labelling approaches such as orthogonal unnatural amino acids 39 . RESI is thus poised to close the gap between 3D fluorescence super - resolution microscopy in whole intact cells and cryo - EM structural studies of individual supramolecular complexes , introducing a paradigm shift in fluorescence imaging by pushing opti - cal microscopy to \u00c5ngstr\u00f6m resolutions . Online content Any methods , additional references , Nature Portfolio reporting summa - ries , source data , extended data , supplementary information , acknowl - edgements , peer review information ; details of author contributions and competing interests ; and statements of data and code availability are available at https : / / doi . org / 10 . 1038 / s41586 - 023 - 05925 - 9 . 1 . Hell , S . W . & Wichmann , J . Breaking the diffraction resolution limit by stimulated emission : stimulated - emission - depletion fluorescence microscopy . Opt . Lett . 19 , 780 \u2013 782 ( 1994 ) . 2 . Betzig , E . Proposed method for molecular optical imaging . Opt . Lett . 20 , 237 \u2013 239 ( 1995 ) . 3 . Klar , T . A . , Jakobs , S . , Dyba , M . , Egner , A . & Hell , S . W . Fluorescence microscopy with diffraction resolution barrier broken by stimulated emission . Proc . Natl Acad . Sci . USA 97 , 8206 \u2013 8210 ( 2000 ) . 4 . Betzig , E . et al . Imaging intracellular fluorescent proteins at nanometer resolution . Science 313 , 1642 \u2013 1645 ( 2006 ) . 5 . Rust , M . J . , Bates , M . & Zhuang , X . Sub - diffraction - limit imaging by stochastic optical reconstruction microscopy ( STORM ) . Nat . Methods 3 , 793 \u2013 795 ( 2006 ) . 6 . Sharonov , A . & Hochstrasser , R . M . Wide - field subdiffraction imaging by accumulated binding of diffusing probes . Proc . Natl Acad . Sci . USA 103 , 18911 \u2013 18916 ( 2006 ) . 7 . Dai , M . , Jungmann , R . & Yin , P . Optical imaging of individual biomolecules in densely packed clusters . Nat . Nanotechnol . 11 , 798 \u2013 807 ( 2016 ) . 8 . Balzarotti , F . et al . Nanometer resolution imaging and tracking of fluorescent molecules with minimal photon fluxes . Science 355 , 606 \u2013 612 ( 2017 ) . 9 . Schnitzbauer , J . , Strauss , M . T . , Schlichthaerle , T . , Schueder , F . & Jungmann , R . Super - resolution microscopy with DNA - PAINT . Nat . Protoc . 12 , 1198 \u2013 1228 ( 2017 ) . 10 . Reymond , L . et al . SIMPLE : structured illumination based point localization estimator with enhanced precision . Opt . Express 27 , 24578 \u2013 24590 ( 2019 ) . 11 . Gu , L . et al . Molecular resolution imaging by repetitive optical selective exposure . Nat . Methods 16 , 1114 \u2013 1118 ( 2019 ) . 12 . Cnossen , J . et al . Localization microscopy at doubled precision with patterned illumination . Nat . Methods 17 , 59 \u2013 63 ( 2020 ) . 13 . Weber , M . et al . MINSTED fluorescence localization and nanoscopy . Nat . Photonics 15 , 361 \u2013 366 ( 2021 ) . 14 . Masullo , L . A . et al . An alternative to MINFLUX that enables nanometer resolution in a confocal microscope . Light Sci . Appl . 11 , 199 ( 2022 ) . 15 . Lelek , M . et al . Single - molecule localization microscopy . Nat . Rev . Methods Primers https : / / doi . org / 10 . 1038 / s43586 - 021 - 00038 - x ( 2021 ) . 16 . Mortensen , K . I . , Churchman , L . S . , Spudich , J . A . & Flyvbjerg , H . Optimized localization analysis for single - molecule tracking and super - resolution microscopy . Nat . Methods 7 , 377 \u2013 381 ( 2010 ) . 17 . Jungmann , R . et al . Multiplexed 3D cellular super - resolution imaging with DNA - PAINT and Exchange - PAINT . Nat . Methods 11 , 313 \u2013 318 ( 2014 ) . 18 . Jungmann , R . et al . Single - molecule kinetics and super - resolution microscopy by fluorescence imaging of transient binding on DNA origami . Nano Lett . 10 , 4756 \u2013 4761 ( 2010 ) . 19 . Rothemund , P . W . Folding DNA to create nanoscale shapes and patterns . Nature 440 , 297 \u2013 302 ( 2006 ) . 20 . Eklund , A . S . , Comberlato , A . , Parish , I . A . , Jungmann , R . & Bastings , M . M . C . Quantification of strand accessibility in biostable DNA origami with single - staple resolution . ACS Nano 15 , 17668 \u2013 17677 ( 2021 ) . 21 . Beck , M . & Hurt , E . The nuclear pore complex : understanding its function through structural insight . Nat . Rev . Mol . Cell Biol . 18 , 73 \u2013 89 ( 2017 ) . 22 . Schuller , A . P . et al . The cellular environment shapes the nuclear pore complex architecture . Nature 598 , 667 \u2013 671 ( 2021 ) . 23 . Loschberger , A . et al . Super - resolution imaging visualizes the eightfold symmetry of gp210 proteins around the nuclear pore complex and resolves the central channel with nanometer resolution . J . Cell Sci . 125 , 570 \u2013 575 ( 2012 ) . 24 . Szymborska , A . et al . Nuclear pore scaffold structure analyzed by super - resolution microscopy and particle averaging . Science 341 , 655 \u2013 658 ( 2013 ) . 25 . Heydarian , H . et al . 3D particle averaging and detection of macromolecular symmetry in localization microscopy . Nat . Commun . 12 , 2847 ( 2021 ) . 26 . Wang , W . , Heydarian , H . , Huijben , T . , Stallinga , S . & Rieger , B . Joint registration of multiple point clouds for fast particle fusion in localization microscopy . Bioinformatics 38 , 3281 \u2013 3287 ( 2022 ) . 27 . Gwosch , K . C . et al . Reply to : Assessment of 3D MINFLUX data for quantitative structural biology in cells . Nat . Methods 20 , 52 \u2013 54 ( 2023 ) . 28 . Weber , M . et al . MINSTED nanoscopy enters the Angstrom localization range . Nat . Biotechnol . https : / / doi . org / 10 . 1038 / s41587 - 022 - 01519 - 4 ( 2022 ) . 29 . Agasti , S . S . et al . DNA - barcoded labeling probes for highly multiplexed Exchange - PAINT imaging . Chem . Sci . 8 , 3080 \u2013 3091 ( 2017 ) . 30 . Wu , Y . L . et al . Maximum - likelihood model fitting for quantitative analysis of SMLM data . Nat . Methods 20 , 139 \u2013 148 ( 2023 ) . 31 . Drew , H . R . et al . Structure of a B - DNA dodecamer : conformation and dynamics . Proc . Natl Acad . Sci . USA 78 , 2179 \u2013 2183 ( 1981 ) . 32 . Strauss , S . & Jungmann , R . Up to 100 - fold speed - up and multiplexing in optimized DNA - PAINT . Nat . Methods 17 , 789 \u2013 791 ( 2020 ) . 33 . Helmerich , D . A . et al . Photoswitching fingerprint analysis bypasses the 10 - nm resolution barrier . Nat . Methods 19 , 986 \u2013 994 ( 2022 ) . 34 . Pavlasova , G . & Mraz , M . The regulation and function of CD20 : an \u201cenigma\u201d of B - cell biology and targeted therapy . Haematologica 105 , 1494 \u2013 1506 ( 2020 ) . 35 . Glennie , M . J . , French , R . R . , Cragg , M . S . & Taylor , R . P . Mechanisms of killing by anti - CD20 monoclonal antibodies . Mol . Immunol . 44 , 3823 \u2013 3837 ( 2007 ) . 36 . Pierpont , T . M . , Limper , C . B . & Richards , K . L . Past , present , and future of rituximab\u2014the world\u2019s first oncology monoclonal antibody therapy . Front . Oncol . 8 , 163 ( 2018 ) . 37 . Rouge , L . et al . Structure of CD20 in complex with the therapeutic monoclonal antibody rituximab . Science 367 , 1224 \u2013 1230 ( 2020 ) . 38 . Kumar , A . , Planchais , C . , Fronzes , R . , Mouquet , H . & Reyes , N . Binding mechanisms of therapeutic antibodies to human CD20 . Science 369 , 793 \u2013 799 ( 2020 ) . 39 . Beliu , G . et al . Bioorthogonal labeling with tetrazine - dyes for super - resolution microscopy . Commun . Biol . 2 , 261 ( 2019 ) . Publisher\u2019s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations . Open Access This article is licensed under a Creative Commons Attribution 4 . 0 International License , which permits use , sharing , adaptation , distribution and reproduction in any medium or format , as long as you give appropriate credit to the original author ( s ) and the source , provide a link to the Creative Commons licence , and indicate if changes were made . The images or other third party material in this article are included in the article\u2019s Creative Commons licence , unless indicated otherwise in a credit line to the material . If material is not included in the article\u2019s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use , you will need to obtain permission directly from the copyright holder . To view a copy of this licence , visit http : / / creativecommons . org / licenses / by / 4 . 0 / . \u00a9 The Author ( s ) 2023 Methods Materials Unmodified DNA oligonucleotides , as well as DNA oligonucleotides modified with C3 - azide and Cy3B , were purchased from MWG Eurofins and Metabion . The M13mp18 and p7560 scaffold was obtained from Tilibit . Magnesium chloride ( 1 M , no . AM9530G ) , sodium chloride ( 5 M , no . AM9759 ) , ultrapure water ( no . 10977 - 035 ) , Tris ( 1 M , pH 8 . 0 , no . AM9855G ) , EDTA ( 0 . 5 M , pH 8 . 0 , no . AM9260G ) and 10\u00d7 PBS ( no . 70011051 ) were purchased from Thermo Fisher Scientific . BSA ( no . A4503 - 10G ) was ordered from Sigma - Aldrich . Triton X - 100 ( no . 6683 . 1 ) was purchased from Carl Roth . Sodium hydroxide ( no . 31627 . 290 ) was purchased from VWR . Paraformaldehyde ( no . 15710 ) and glutaraldehyde ( no . 16220 ) were obtained from Electron Microscopy Sciences . Tween - 20 ( no . P9416 - 50ML ) , glycerol ( no . 65516 - 500 ml ) , methanol ( no . 32213 - 2 . 5L ) , proto - catechuate 3 , 4 - dioxygenase pseudomonas ( PCD , no . P8279 ) , 3 , 4 - dihydroxybenzoic acid ( PCA , no . 37580 - 25G - F ) and ( \u00b1 ) - 6 - hydroxy - 2 , 5 , 7 , 8 - tetra - methylchromane - 2 - carboxylic acid ( trolox , no . 238813 - 5G ) were ordered from Sigma - Aldrich . Neutravidin ( no . 31000 ) was pur - chased from Thermo Fisher Scientific . Biotin - labelled BSA ( no . A8549 ) and sodium azide ( no . 769320 ) were obtained from Sigma - Aldrich . Cover - slips ( no . 0107032 ) and glass slides ( no . 10756991 ) were purchased from Marienfeld and Thermo Fisher Scientific , respectively . Fetal bovine serum ( FBS , no . 10500 - 064 ) , 1\u00d7 PBS ( pH 7 . 2 , no . 20012 - 019 ) , 0 . 05 % trypsin - EDTA ( no . 25300 - 054 ) , salmon sperm DNA ( no . 15632011 ) , OptiMEM ( no . 31985062 ) and Lipofectamine LTX ( no . A12621 ) were purchased from Thermo Fisher Scientific . Gold nanoparticles ( 90 nm , no . G - 90 - 100 ) were ordered from Cytodiagnostics . Nanobodies against GFP ( clone 1H1 ) with a single ectopic cysteine at the C terminus for site - specific conjugation were purchased from Nanotag Biotechnologies . DBCO - PEG4 - Maleimide ( no . CLK - A108P ) was purchased from Jena Bioscience . Buffers The following buffers were used for sample preparation and imaging . \u2022 Buffer A : 10 mM Tris pH 8 . 0 , 100 mM NaCl and 0 . 05 % Tween - 20 \u2022 Buffer B : 10 mM MgCl 2 , 5 mM Tris - HCl pH 8 . 0 , 1 mM EDTA and 0 . 05 % Tween - 20 pH 8 . 0 \u2022 Buffer C : 1\u00d7 PBS , 1 mM EDTA , 500 mM NaCl pH 7 . 4 , 0 . 02 % Tween , optionally supplemented with 1\u00d7 trolox , 1\u00d7 PCA and 1\u00d7 PCD \u2022 Blocking buffer : 1\u00d7 PBS , 1 mM EDTA , 0 . 02 % Tween - 20 , 0 . 05 % NaN 3 , 2 % BSA , 0 . 05 mg ml \u2013 1 sheared salmon sperm DNA \u2022 Two - dimensional ( 2D ) DNA origami folding buffer : 10 mM Tris , 1 mM EDTA , 12 . 5 mM MgCl 2 pH 8 . 0 \u2022 3D DNA origami folding buffer : 5 mM Tris , 1 mM EDTA , 5 mM NaCl , 20 mM MgCl2 pH 8 . 0 \u2022 1\u00d7 TA buffer : 40 mM Tris pH 8 . 0 , 20 mM acetic acid PCA , PCD and trolox Trolox ( 100\u00d7 ) was made by the addition of 100 mg of trolox to 430 \u00b5l of 100 % methanol and 345 \u00b5l of 1 M NaOH in 3 . 2 ml of water . PCA ( 40\u00d7 ) was made by mixing 154 mg of PCA in 10 ml of water and NaOH and adjustment of pH to 9 . 0 . PCD ( 100\u00d7 ) was made by the addition of 9 . 3 mg of PCD to 13 . 3 ml of buffer ( 100 mM Tris - HCl pH 8 . 0 , 50 mM KCl , 1 mM EDTA , 50 % glycerol ) . DNA - PAINT docking and imager sequences Four orthogonal DNA sequence motifs were used to label targets in four RESI rounds . The docking strands were 5xR1 ( TCCTCCT CCTCCTCCTCCT ) , 5xR2 ( ACCACCACCACCACCACCA ) , 7xR3 ( CTCTCTCTCTCTCTCTCTC ) and 7xR4 ( ACACACACACACACACACA ) . The respective imagers were R1 ( AGGAGGA - Cy3B ) , R2 ( TGGTGGT - Cy3B ) , R3 ( GAGAGAG - Cy3B ) and R4 ( TGTGTGT - Cy3B ) . The design of 2D RESI origami required extension of the R1 site at the 5\u2032 end such that the adjacent R1 and R3 docking strands could be spaced apart by a single base pair . Thus , the docking strand 5\u2032 5xR1 ( TCCTCCTCCTCCTCCTCCT ) and the 5\u2032 R1 imager ( Cy3B - AGGAGGA ) were used rather than the 3\u2032 versions for both 2D DNA origamis . DNA origami self - assembly ( 2D ) All 2D DNA origami structures were designed in caDNAno 40 . Self - assembly of DNA origami was accomplished in a one - pot reaction mix with a total volume of 40 \u00b5l , consisting of 10 nM scaffold strand ( for sequence , see Supplementary Data 2 ) , 100 nM folding staples ( Supple - mentary Data 1 ) , 500 nM biotinylated staples ( Supplementary Data 1 ) and 1 \u00b5M staple strands with docking site extensions ( Supplementary Data 1 ) in 2D DNA origami folding buffer . The reaction mix was then subjected to a thermal annealing ramp using a thermocycler . First , it was incubated at 80 \u00b0C for 5 min , cooled using a temperature gradient from 60 to 4 \u00b0C in steps of 1 \u00b0C per 3 . 21 min and finally held at 4 \u00b0C . DNA origami self - assembly ( 3D ) The 3D DNA origami disk structure was designed in caDNAno 40 . Self - assembly of the DNA origami disk was accomplished in a one - pot reaction mix of 50 \u00b5l total volume , consisting of 20 nM scaffold strand p7560 ( for sequence , see Supplementary Data 3 ) , 200 nM core folding staples ( Supplementary Data 1 ) , 200 nM staple sequences without handle extension ( Supplementary Data 1 ) , 500 nM biotinylated sta - ples ( Supplementary Data 1 ) , 2 \u00b5M staple strands with R4 docking site extensions and 4 \u00b5M staple strands with R1 or R3 docking site exten - sions ( Supplementary Data 1 ) in 3D DNA origami folding buffer . The reaction mix was then subjected to a thermal annealing ramp using a thermocycler . It was first incubated at 80 \u00b0C for 5 min then cooled using a temperature gradient from 60 \u00b0C to 20 \u00b0C in steps of 1 \u00b0C h \u2013 1 and finally held at 20 \u00b0C . DNA origami purification After self - assembly , structures were purified by agarose gel electropho - resis ( 1 . 5 % agarose , 1\u00d7 TA , 10 mM MgCl 2 , 0 . 5\u00d7 SybrSafe ) at 4 . 5 V cm \u2013 1 for 1 . 5 h . Gel bands were cut , crushed and the origami stored in low - binding Eppendorf tubes at \u221220 \u00b0C . DNA origami sample preparation and imaging For sample preparation , a bottomless six - channel slide ( ibidi , no . 80608 ) was attached to a coverslip . First , 80 \u00b5l of biotin - labelled BSA ( 1 mg ml \u2013 1 , dissolved in buffer A ) was flushed into the chamber and incu - bated for 5 min . The chamber was then washed with 360 \u00b5l of buffer A . A volume of 100 \u00b5l of neutravidin ( 0 . 1 mg ml \u2013 1 , dissolved in buffer A ) was then flushed into the chamber and allowed to bind for 5 min . After washing with 180 \u00b5l of buffer A and subsequently with 360 \u00b5l of buffer B , 80 \u00b5l of biotin - labelled DNA structures ( approximately 200 pM ) in buffer B was flushed into the chamber and incubated for 5 min . For measurement of the DNA origami disk , additional 2D DNA origami structures with 12 target sites 9 spaced 20 nm apart were incubated together , with the 3D disk origami serving as fiducials for drift cor - rection . After DNA origami incubation the chamber was washed with 540 \u00b5l of buffer B . For DNA origami disk structures , 150 \u00b5l of gold nano - particles ( diluted 1 : 10 in buffer B ) was flushed through and incubated for 5 min before washing with 540 \u00b5l of buffer B . Finally , 180 \u00b5l of the imager solution in buffer B was flushed into the chamber . The chamber remained filled with imager solution and imaging was then performed . Between imaging rounds , the sample was washed three times with 1 ml of buffer B until no residual signal from the previous imager solution was detected . Then , the next imager solution was introduced . For RESI , two imaging rounds were performed with imagers R1 and R4 present in round 1 and the imagers R3 and R4 in round 2 ( R1 and R3 probe the sites of interest for RESI and R4 serves alignment purposes ) . Nanobody \u2013 DNA conjugation Nanobodies were conjugated as described previously 32 . Unconju - gated nanobodies were thawed on ice , then 20 - fold molar excess of Article bifunctional DBCO - PEG4 - Maleimide linker was added and reacted for 2 h on ice . Unreacted linker was removed by buffer exchange to PBS using Amicon centrifugal filters ( 10 , 000 MWCO ) . The DBCO - modified nanobodies were reacted with 5\u00d7 molar excess of azide - functionalized DNA ( R1 , R2 , R3 and R4 ) overnight at 4 \u00b0C . Unconjugated protein and free DNA were removed by anion exchange chromatography using an \u00c4KTA pure system equipped with a Resource Q 1 ml column . Cell culture CHO cells ( CCL - 61 , ATCC ) were cultured in Gibco Ham\u2019s F - 12K ( Kaighn\u2019s ) medium supplemented with 10 % FBS ( no . 11573397 , Gibco ) . U2OS - CRISPR - Nup96 - mEGFP cells ( a gift from the Ries and Ellenberg laboratories ) were cultured in McCoy\u2019s 5A medium ( Thermo Fisher Scientific , no . 16600082 ) supplemented with 10 % FBS . Cells were pas - saged every 2 \u2013 3 days using trypsin - EDTA . Nup96 EGFP imaging U2OS - CRISPR - Nup96 - mEGFP cells were seeded on ibidi eight - well high glass - bottom chambers ( no . 80807 ) at a density of 30 , 000 cm \u2013 2 . Cells were fixed with 2 . 4 % paraformaldehyde in PBS for 30 min at room temperature . After fixation , cells were washed three times with PBS . Gold nanoparticles ( 200 \u00b5l ) were incubated for 5 min and washed three times with PBS . Blocking and permeabilization were performed with 0 . 25 % Triton X - 100 in blocking buffer for 90 min . After washing with PBS , cells were incubated with 100 nM anti - GFP nanobodies in blocking buffer for 60 min at room temperature . To enable RESI , the nanobody solution consisted of 25 nM R1 , R2 , R3 and R4 docking - strand - coupled anti - GFP nanobodies with a total nanobody concentration of 100 nM . Unbound nanobodies were removed by washing three times with PBS , followed by washing once with buffer C for 10 min . Postfixation was performed with 2 . 4 % paraformaldehyde in PBS for 15 min . After wash - ing 3\u00d7 with PBS , the imager solution in buffer C was flushed into the chamber . Between imaging rounds the sample was washed with 1 \u2013 2 ml of PBS until no residual signal from the previous imager solution was detected . Then , the next imager solution was introduced . First , imag - ers R1 , R2 , R3 and R4 were added simultaneously to the sample to perform a standard DNA - PAINT measurement ; then , RESI imaging was conducted via four subsequent imaging rounds with only one of the imagers . Cloning mEGFP - Alfa - CD20 was cloned by insertion of Alfa - CD20 into the mEGFP - C1 plasmid ( no . 54759 , Addgene ) . An Alfa - CD20 gblock ( obtained from Integrated DNA Technologies ) was amplified with primers cggcatggacgagct and gtacaagtccgga and , after cutting with restriction enzymes BsrGI and BamHI , Gibson assembly was performed ( 2\u00d7 mix , NEB ) . mEGFP - CD20 imaging CHO cells were seeded on ibidi eight - well high glass - bottom chambers ( no . 80807 ) the day before transfection at a density of 15 , 000 cm \u2013 2 . Transfection with mEGFP - CD20 was carried out with Lipofectamine LTX as specified by the manufacturer . CHO cells were allowed to express mEGFP - CD20 for 16 \u2013 24 h . Then , the medium was replaced with fresh F - 12K medium + 10 % FBS ( in the untreated case ) or with F - 12K medium + 10 % FBS + 10 ug ml \u2013 1 RTX - Alexa 647 ( a gift from Roche Glycart ) ( in the RTX - treated case ) , followed by incubation for 30 min . After washing two times with fresh medium for 5 min , cells were fixed with 250 \u00b5l of prewarmed 4 % PFA + 0 . 1 % glutaraldehyde in PBS for 15 min . CHO cells were washed three times with PBS and quenched with 100 mM Tris pH 8 . 0 for 5 min . Permeabilization was carried out for 5 min with 0 . 2 % Triton X - 100 in PBS , followed by three washes with PBS . Cells were blocked in blocking buffer for 1 h at room temperature ( RT ) . Anti - GFP nanobodies were incubated at a total concentration of 25 nM overnight at 4 \u00b0C ; for RESI with four rounds this yielded 6 . 25 nM each of GFP - Nb - R1 / 2 / 3 / 4 . After washing three times with PBS at RT for 15 min , cells were postfixed with 4 % PFA at RT for 10 min followed by washing and postfixation as described above . Gold nanoparticles ( 90 nm ) were diluted 1 : 3 in PBS and incubated for 10 min at RT and the sample was washed two times with PBS to remove unbound gold . The imager solution in buffer C for the first round was incubated for 5 min and then replaced with fresh imager , after which the first acquisition round was started . Between imaging rounds the sample was washed with at least 2 ml of PBS until no residual signal from the previous imager solution was detected . Then , the next imager solution was introduced . RESI imaging was conducted via four subsequent imaging rounds with only one of the imagers . In the final imaging round , imagers R1 , R2 , R3 and R4 were added simultaneously to the sample to perform a standard DNA - PAINT measurement . Microscopy setup Fluorescence imaging was carried out using an inverted microscope ( Nikon Instruments , Eclipse Ti2 ) with the Perfect Focus System , applying an objective - type TIRF configuration equipped with an oil - immersion objective ( Nikon Instruments , Apo SR TIRF \u00d7100 / numeri - cal aperture 1 . 49 , oil ) . A 560 nm laser ( MPB Communications , 1 W ) was used for excitation . The laser beam was passed through a cleanup filter ( Chroma Technology , no . ZET561 / 10 ) and coupled to the microscope objective using a beam splitter ( Chroma Technology , no . ZT561rdc ) . Fluorescence was spectrally filtered with an emission filter ( Chroma Technology , nos . ET600 / 50m and ET575lp ) and imaged on an sCMOS camera ( Andor , Zyla 4 . 2 Plus ) without further magnification , resulting in an effective pixel size of 130 nm ( after 2 \u00d7 2 binning ) . The readout rate was set to 200 MHz . Images were acquired by choosing a region of interest of size 512 \u00d7 512 pixels . 3D imaging was performed using a cylindrical lens ( Nikon Instruments , N - STORM ) in the detection path . Raw microscopy data were acquired using \u00b5Manager 41 ( v . 2 . 0 . 1 ) . Total internal reflection illumination was used for 2D and 3D DNA origami data , as well as for CD20 acquisition . Highly inclined and laminated optical sheet ( HILO ) illumination was employed for the acquisition of NPC data . Detailed imaging conditions for the respective experiments are shown in Extended Data Table 1 . Imaging parameters and duration Due to target and sample heterogeneity the optimal imager concentra - tion , c , used to achieve sparse blinking varies . Here we used concentra - tions from 100 pM ( Nup96 ) to 800 pM ( DNA origami ) . Optimal imager concentrations were determined visually for each sample . Concentra - tions were altered until blinking was frequent but sufficiently sparse to achieve good DNA - PAINT resolution . The average number of expected binding events per binding site during a DNA - PAINT measurement is given by the duration of the measurement t measurement and the mean dark time \u03c4 dark ( defined as \u03c4 = k c dark 1 \u00d7 on , with k on being the on - rate of a given imager strand ) as : \uf8eb \uf8ed\uf8ec \uf8f6 \uf8f8\uf8f7 n t \u03c4 t k c = = \u00d7 \u00d7 . bindingevents measurement dark measurement on The average number of localizations per binding event is given by the mean bright time \u03c4 bright and camera exposure time t exposure as n \u03c4 t = . locsperbindingevent bright exposure \uf8eb \uf8ed\uf8ec \uf8ec \uf8f6 \uf8f8\uf8f7 \uf8f7 Therefore , the average number of localizations expected per binding site over the course of the measurement is \uf8eb \uf8ed\uf8ec \uf8f6 \uf8f8\uf8f7 \uf8eb \uf8ed \uf8ec \uf8ec \uf8f6 \uf8f8 \uf8f7 \uf8f7 \uf8eb \uf8ed \uf8ec \uf8ec \uf8f6 \uf8f8 \uf8f7 \uf8f7 n t \u03c4 \u03c4 t t k c \u03c4 t = \u00d7 = \u00d7 \u00d7 \u00d7 . loc measurement dark bright exposure measurement on bright exposure It follows that the total acquisition time necessary to collect n loc localizations is , on average , t t n \u03c4 k c = \u00d7 \u00d7 \u00d7 . measurement exposure loc bright on The necessary number of localizations , n loc , is calculated using \u03c3 = \u03c3 n RESI DNA PAINT loc \u2010 , and thus ( ) n = \u03c3 \u03c3 loc 2 DNA PAINT RESI \u2010 with the DNA - PAINT locali - zation precision \u03c3 DNA - PAINT . For expected imager concentrations between 50 and 800 pM , expo - sure times between 100 and 200 ms and kinetics reported previously 32 , the times required to collect 16 localizations ( 1 nm RESI precision given \u03c3 DNA - PAINT = 4 nm ) vary between 42 s ( R2 , 800 pM , 100 ms exposure time ) and 314 min ( R5 , 50 pM , 200 ms exposure time ) . DNA - PAINT analysis Raw fluorescence data were subjected to super - resolution reconstruc - tion using the Picasso software package 9 ( latest version available at https : / / github . com / jungmannlab / picasso ) . Drift correction was per - formed with a redundant cross - correlation and gold particles as fidu - cials for cellular experiments , or with single DNA - PAINT docking sites as fiducials for origami experiments . Channel alignment Alignment of subsequent imaging rounds was performed iteratively in Picasso 9 , starting with a redundant cross - correlation and followed by gold fiducial alignment for cellular experiments . Every DNA ori - gami was equipped with additional DNA - PAINT docking sites that were imaged simultaneously with the sites of interest in all imaging rounds , thus enabling their use as fiducials . First , redundant cross - correlation ( 2D and 3D origami measurements ) and gold alignment ( 3D measure - ments ) were performed in Picasso Render . To correct for nanoscopic movement of individual DNA origami during buffer exchange , channel alignment was not only performed on the full field of view but , addition - ally , small regions of interest containing only one DNA origami were selected . Within each region of interest , alignment was then conducted via the fiducial docking sites of the DNA origami . This was performed outside of Picasso in a custom Python script , not only to find the optimal translation between channels but also to correct for possible rotations of the DNA origami . Clustering and RESI Clustering of DNA - PAINT localizations . After channel alignment , DNA - PAINT data were analysed using a custom clustering algorithm for each imaging round . This algorithm is based on the fact that , in DNA - PAINT , localizations are independent measurements of the posi - tion of a target molecule and are observed to be Gaussian distributed . To assign localizations to a specific target molecule , we first used a gradient ascent method to find the centre of a localization cloud for each target . We then assigned all localizations circularly distributed around the centre point to the same target molecule . This is a valid approx imation because , due to the reduction of effective target density by RESI\u2019s sequential imaging approach , the majority of localization clouds from single targets are spaced sufficiently apart . The clustering algorithm uses two input parameters : radius r , which sets the final size of the clusters and defines a circular environment around each localization , and the minimal number of localizations , n min , representing a lower threshold for the number of DNA - PAINT localizations in any cluster . First , the number of neighbouring localizations within distance r from each localization is calculated . If a given localization has more neighbours within its r radius than all neighbouring localizations , it is considered a local maximum . If there are more than n min localizations within a circle of radius r around such a local maximum , these locali - zations are assigned to the same cluster ; the remainder are not con - sidered to be part of a cluster and are omitted from further analysis . Further filtering of clusters is performed to exclude clusters that originate from unspecific sticking of imagers to the sample . Firstly , the mean frame ( mean value of the frame numbers in which localizations occurred ) of all localizations assigned to the same cluster is calcu - lated . In the case of repetitive blinking the mean frame is expected to be around half the total number of frames 42 . The algorithm therefore excludes all clusters with a mean frame in the first or last 20 % of frames . Secondly , sticking events in the middle of the acquisition time can be identified by dividing the acquisition time into 20 time windows each containing 5 % of frames . If any of these time windows contains more than 80 % of localizations in the cluster , it is excluded as a sticking event . The choice of the clustering radius r and the threshold n min depend on the respective experimental conditions . A suitable value for n min can be estimated by picking localization clouds originating from single target molecules ( that is , well separated ) in Picasso Render , exporting pick properties and plotting a histogram of the number of localizations in each pick . n min is chosen to differentiate between populations corre - sponding to single targets and to background localizations . The radius r scales with the size of the localization clouds and thus the localization precision . If too large a value is chosen , adjacent clus - ters might not be separated ; if r is too small , \u2018subclustering\u2019 within one localization can occur . The latter also translates to a peak in NND at twice the clustering radius . A good a priori starting value for r is represented by approximately twofold the localization precision of the underlying DNA - PAINT measurement . Picasso Render offers a tool ( Test Clusterer ) in which the effect of different clustering parameters can be tested for a small region of interest . For 3D clustering , an additional radius for the z direction is intro - duced because the spread of localizations in z is approximately twofold greater compared with x and y . Calculation and rendering of RESI localization . Following cluster analysis , the centres of the DNA - PAINT localization groups were calcu - lated as weighted ( wtd ) means by employing the squared inverse localization precisions ( ) lp 1 2 as weights . For x and y coordinates : x wx w w lp \u00af = \u2211 \u2211 , = 1 . i N i i i N i i wtd = 1 = 1 2 For z coordinates a standard mean without weights is used to calcu - late z positions . The precision of the resulting RESI localization is the weighted s . e . m . of the underlying grouped localizations : \u2211 \u2211 s s N x N x N N w x x w ( ) = ( ) = Var ( ) , where Var ( ) = \u2212 1 ( \u2212 ) . x x i N i i i N i wtd wtd wtd wtd = 1 wtd 2 = 1 The choice for lp 1 / 2 as weights is based on the following argument : under the hypothesis that localizations are independent and normally distributed with the same mean , the weighted mean based on inverse variances as weights is the maximum likelihood estimator of the mean of the whole set of localizations . Therefore , the variance of the weighted mean is minimal ( the estimator is optimal ) when the inverse variances of individual measurements lp 1 / 2 are chosen as weights . Finally , we take the average of the resulting x and y s . e . m . as the final precision of each RESI localization . For z coordinates the precision is estimated to be two times xy precision . Saving RESI localizations in a Picasso hdf5 file allowed us to render them as Gaussians with s . d . corresponding to their respective precision . Article RESI resolution estimation Evaluation of in silico RESI precision with numerical simulations . To evaluate the performance of RESI , in silico numerical simulations were performed . The algorithm consists of the following steps . ( 1 ) A grid of defined positions of the binding sites ( ground truth ) is generated . Typically , a grid of positions was generated ( Extended Data Fig . 1a , top left ) . ( 2 ) SMLM ( DNA - PAINT ) localizations are simulated as samples from a 2D Gaussian distribution with \u03c3 = \u03c3 SMLM . A large number ( M ) of localizations is generated per binding site ( Extended Data Fig . 1a , top right ) . ( 3 ) For each binding site , subsets of K localizations are randomly selected ( K < < M ) . This results in n = M K subsets of SMLM localiza - tions ( Extended Data Fig . 1a , bottom left ) that are then averaged to generate n RESI localizations ( Extended Data Fig . 1a , bottom right ) . ( 4 ) The resulting RESI localizations are then shown in a histogram ( Extended Data Fig . 1b ) and the trace ( tr ) of the covariance matrix is calculated . RESI precision is estimated as \u03c3 tr x y = ( cov ( , ) ) RESI 1 2 ( Extended Data Fig . 1c ) . This definition has been used before in the field as a scalar metric for 2D variance 8 . ( 5 ) Steps 3 and 4 are repeated for different values of K to numerically study \u03c3 = \u03c3 RESI ( K ) . Evaluation of experimental RESI precision by resampling of local - izations . To evaluate the precision of RESI in experimental data , an analogous method was used . Briefly , the M total of DNA - PAINT localizations of each group corresponding to a single binding site was randomly resampled into subsets of K localizations , then steps 4 and 5 above were performed to evaluate \u03c3 RESI . The plotted \u03c3 RESI in Fig . 3d is the average value of all single binding sites in the dataset . Error bars represent the s . d . of the different \u03c3 RESI values calculated for different binding sites . Note that this analysis can be performed only for K < < M to have suf - ficient n = M K RESI localizations for a statistically significant estimation . Because final RESI localization takes into account all M DNA - PAINT localizations , final precision is extrapolated as \u03c3 = \u03c3 M RESI SMLM . Stochastic labelling : simulations and user guidelines In RESI , the sparsity of binding sites in the sample is achieved by labell - ing a single species of biomolecules with different orthogonal DNA sequences . The labelling process is performed in a stochastic manner : n different labels ( for example , DNA - conjugated nanobodies ) targeting the same protein species are simultaneously incubated in the sample and thus the probability of each single protein being labelled with a certain sequence i ( i = 1 , \u2026 , n ) is p = i n 1 , given that the same concentration of each label is used . Subsequently , n imaging rounds are performed to record all groups of localizations required to obtain the final RESI image . The minimum number of labels ( n ) and rounds necessary to achieve sufficient sparsity of binding sites in each imaging round will depend mainly on three factors : SMLM localization precision and density and the molecular arrangement of the protein of interest . Here we describe how these parameters affect the final RESI results using a few practical examples . Case 1 : protein structure with oligomers not resolvable with DNA - PAINT . A typical study case is that of single proteins arranged in dimers , which in turn present another specific spatial organization in space . This is the case , for example , of the Nup96 in the NPC . In this case stochastic labelling has to be such that the probability of labelling two proteins forming a dimer with different sequences is sufficiently high . For n rounds of labelling / imaging , the probability is P p n ( diff . seq . ) = 1\u2212 = 1\u2212 1 i for n = 4 labelling / imaging rounds P ( diff . seq . ) \u2248 75 % . We chose n = 4 to demonstrate that it provides a relatively high P ( diff . seq . ) with only a few imaging rounds . We note , however , that n > 4 could be used to increase P ( diff . seq . ) and hence to maximize the sparsity of labelled binding sites in each round . To resolve a set of an arbitrary number of molecules , m , spaced more closely than the resolution of DNA - PAINT , they must be labelled with n orthogonal sequences . In general , the proportion of m molecules labelled with n orthogonal sequences , and thus the proportion of resolvable sets of molecules , follows the equation P m n n n m n ( , ) = ! ( \u2212 ) ! . m Case 2 : proteins distributed similarly to CSR at a certain density . This is a common case\u2014for example , for membrane receptors . If pro - teins are distributed in a CSR fashion ( Extended Data Fig . 14a ) at a given density , DNA - PAINT can already resolve single proteins that are suf - ficiently spaced from their NNs . We will consider that proteins at a distance d = 4 \u00d7 \u03c3 DNA - PAINT are reliably resolved ( note that this criterion is significantly stricter than 2 . 35 \u00d7 \u03c3 DNA - PAINT ) . Then , for a given density , the NND histogram can be computed and the fraction of distances below d calculated ( Extended Data Fig . 14b ) . This represents the frac - tion of single proteins , F , that will not be resolved by DNA - PAINT . Here we plot F as a function of both density and resolution ( Extended Data Fig . 14c ) . Such a map already provides a tool to understand the level of SMLM resolution needed to resolve single proteins at a given density . RESI can be interpreted here as a way to reduce the effective density by splitting targets into different stochastically labelled subsets . Hence , the effective density of each round will be reduced according to the formula \u03c1 = n density . Extended Data Fig . 14d shows one - dimensional cuts of the 2D map to provide guidelines to choosing the number of ortho - gonal sequences ( and hence imaging rounds ) needed to be able to perform RESI efficiently . For example , for an initial resolution of 20 nm ( \u03c3 = 5 nm ) , which is typical for DNA - PAINT in a cellular context , and a density of density = 200 \u00b5 molecules m 2 ( relatively high ) , n = 4 different sequences are sufficient to provide P ( diff . seq . ) \u2248 90 % for proteins below d ( Extended Data Fig . 14d ) . These proteins will then be resolv - able by RESI . Model - free averaging Model - free averaging of Nup96 data was performed for both DNA - PAINT and RESI measurement of the same nucleus , as described by Wu et al . 30 . The respective Picasso hdf5 files were segmented in SMAP 43 and saved in a file format compatible for averaging by employing plugins segment - NPC , NPCsegmentCleanup and sitenumbers2loc . Model - free averaging was then performed on the resulting _ sml . mat files with default param - eters by running the particleFusion . m script in Matlab ( available with the SMAP source code ) . The averages shown correspond to the result of the final iteration , in which each point is rendered with a Gaussian of \u03c3 = 2 nm in x , y and z . Numerical simulations for CD20 distribution To interpret the results of the NND data in untreated cells , numerical simulations were performed . Briefly , two populations , one of CD20 monomers and one of dimers with a CSR distribution , were simulated and then their NNDs calculated . The algorithm can be summarized as follows : ( 1 ) Choice of parameters . Density of monomers : number of monomers per unit area ; density of dimers : number of dimers per unit area ; dimer distance : expected distance between the two molecules includ - ing the labelling construct ; uncertainty : variability in the position of each molecule due to labelling and localization errors ; labelling efficiency : fraction of ground - truth molecules that will actually be labelled and measured . The observed density , which has to match the experimental parameter , then becomes observed density = ( density of monomers + density of dimers ) \u00d7 labelling efficiency . For quanti - fication of the labelling efficiency of the DNA - conjugated GFP nano - body we used a transiently transfected CHO cell line expressing a GFP - and Alfa - tag at the C terminus of a monomeric membrane pro - tein ( for example , CD86 ) . We then labelled GFP - and Alfa - tag using their cognate nanobodies conjugated to two orthogonal docking sequences and performed two rounds of Exchange - PAINT . We then obtained the best - fitting parameters for a sample comprising pairs of GFP / Alfa - tag , and isolated Alfa - tags , similarly to how CD20 dimer / monomer analysis is performed . The ratio of these two populations is then used as an estimation of labelling efficiency . Full details of the quantification approach will be available in a manuscript currently in preparation . ( 2 ) Simulation of monomers : a set of spatial coordinates with CSR dis - tribution and given density are drawn ; simulation of dimers : a set of spatial coordinates with CSR distribution are drawn , representing the centre of each dimer . For each dimer centre , two positions are generated with a random orientation and a distance with expect - ed value dimer distance . The position of each pair of molecules is drawn , taking into account the uncertainty parameter ( drawn from a Gaussian distribution ) . ( 3 ) A random subset of \u2018detectable\u2019 molecules is taken from the ground - truth set ( fraction = labelling efficiency ) to simulate the labelling process . ( 4 ) NNDs are calculated on the subset of detectable molecules . The parameters density of monomers = 212 \u00b5m \u2013 2 , density of dimers = 0 \u00b5m \u2013 2 , uncertainty = 5 nm and labelling efficiency = 50 % were used to compare data for RTX - treated cells with a CSR distribu - tion of monomers . For the untreated case , the best - fit parameters were obtained through an iterative , nonlinear , least - squares algorithm . The experimentally observed density ( 50 molecules \u00b5m \u2013 2 ) is used for the simulation . Description of the iterative nonlinear , least - squares algorithm For every set of parameters a simulation is performed , NNDs are histo - grammed and the sum of the squared differences between the simula - tion and experimental histogram are computed . A fit consists of finding the parameters that minimize the sum of the squared differences . Parameters \u2022 D , average dimer distance ( nm ) \u2022 \u03c3 _ label , variability introduced by the labelling ( nm ) \u2022 frac _ of _ dimers , fraction of dimers ( % ) Note : frac _ of _ monomers = 100 \u2013 frac _ of _ dimers Estimation of parameters . ( 1 ) Coarse - fit over a large range of parameters to determine the range of the best - fit parameters . Range D = 1 \u2013 20 nm , \u03c3 _ label = 1 \u2013 20 nm , frac _ of _ dimers = 0 \u2013 100 % . ( 2 ) Fine - fit over a reduced parameter space around the best - fit results in the previous step . The parameters D _ opt , \u03c3 _ label _ opt and frac _ of _ dimers _ opt that best match the proposed model and the data are now found . In this case it resulted in D _ opt = 13 . 5 nm , \u03c3 _ label _ opt = 5 . 5 nm , frac _ of _ dimers _ opt = 47 % ( Fig . 4e , f ) . Estimation of parameter uncertainty . ( 1 ) M is created ( in this case , M = 100 ) , simulated ( using datasets D _ opt , \u03c3 _ label _ opt , frac _ of _ dimers _ opt ) with the same number of molecules as the experimental data ( around 21 , 000 ) . ( 2 ) M datasets are fine - fitted and the best - fit parameters D _ opt , \u03c3 _ label _ opt and frac _ of _ dimers _ opt are obtained . Three sets are obtained : D _ opt , \u03c3 _ label _ opt and frac _ of _ dimers _ opt . ( 3 ) The distributions of D _ opt , \u03c3 _ label _ opt and frac _ of _ dimers _ opt are studied . Standard deviation can be used as an estimation of the parameter uncertainties obtained in b . The uncertainties of the parameters D _ opt , \u03c3 _ label _ opt and frac _ of _ dimers _ opt are now obtained . Data availability Localization data from this study are available at Zenodo ( https : / / doi . org / 10 . 5281 / zenodo . 7795826 ) . Raw microscopy data obtained during this study are available from the corresponding author on reasonable request . Code availability RESI can be performed using Picasso v . 0 . 6 . 0 , available at https : / / github . com / jungmannlab / picasso with documentation provided at https : / / picassosr . readthedocs . io / en / latest / render . html . The custom - written scripts used in this study are available at https : / / github . com / jungmannlab / resi . 40 . Douglas , S . M . et al . Rapid prototyping of 3D DNA - origami shapes with caDNAno . Nucleic Acids Res . 37 , 5001 \u2013 5006 ( 2009 ) . 41 . Edelstein , A . D . et al . Advanced methods of microscope control using muManager software . J . Biol . Methods 1 , e10 ( 2014 ) . 42 . Wade , O . K . et al . 124 - Color super - resolution imaging by engineering DNA - PAINT blinking kinetics . Nano Lett . 19 , 2641 \u2013 2646 ( 2019 ) . 43 . Ries , J . SMAP : a modular super - resolution microscopy analysis platform for SMLM data . Nat . Methods 17 , 870 \u2013 872 ( 2020 ) . Acknowledgements We thank Y . - L . Wu and J . Ries for valuable assistance with model - free averaging . We thank J . Schmied and F . Schueder for helpful discussions . We thank M . K . Steen - Mueller and I . Glueck for proofreading the manuscript . This research was funded in part by the European Research Council through an ERC Consolidator Grant ( ReceptorPAINT , grant agreement no . 101003275 ) , the German Research Foundation through SFB1032 ( project A11 , no . 201269156 ) , the Danish National Research Foundation ( Centre for Cellular Signal Patterns , DNRF135 ) , the Human Frontier Science Program through a Young Investigator Grant ( no . HFSP RGY0065 / 2018 ) , the Volkswagen Foundation through the initiative \u2018Life ? \u2014A Fresh Scientific Approach to the Basic Principles of Life\u2019 ( grant no . 98198 ) , the Max Planck Foundation and the Max Planck Society . S . S . and T . S . acknowledge support by the QBM graduate school . S . C . M . R . , I . B . , P . R . S . , A . S . E . , E . M . U . and M . T . S . acknowledge support by the IMPRS - LS graduate school . I . B . acknowledges funding support by Roche . L . A . M . acknowledges a postdoctoral fellowship from the European Union\u2019s Horizon 20212022 research and innovation programme under Marie Sk\u0142odowska - Curie grant agreement no . 101065980 . Author contributions S . C . M . R . designed and conducted 2D and 3D DNA origami as well as Nup96 experiments , developed the analysis software and analysed DNA origami and Nup96 data . L . A . M . designed and conducted computer simulations , contributed to analysis software and analysed DNA origami , Nup96 and CD20 data . I . B . designed and conducted CD20 experiments and analysed CD20 data . P . R . S . designed and conducted 2D DNA origami experiments , contributed to analysis software and analysed DNA origami and Nup96 data . R . K . and T . S . contributed to analysis software . S . S . developed labelling probes . A . S . E . , S . S . , E . M . U . and M . T . S . performed preliminary RESI experiments . C . K . contributed to the design of studies targeting CD20 and their interpretation . S . C . M . R . , L . A . M . , I . B . , P . R . S . and R . J . interpreted data and wrote the manuscript . R . J . conceived the concept , designed experiments and supervised the study . S . C . M . R . , L . A . M . , I . B . and P . R . S . contributed equally . All authors reviewed and approved the final manuscript . Funding Open access funding provided by the Max Planck Society . Competing interests C . K . declares employment , patents ( unrelated to this work ) and stock ownership with Roche . Additional information Supplementary information The online version contains supplementary material available at https : / / doi . org / 10 . 1038 / s41586 - 023 - 05925 - 9 . Correspondence and requests for materials should be addressed to Ralf Jungmann . Peer review information Nature thanks Alistair Curd and the other , anonymous , reviewer ( s ) for their contribution to the peer review of this work . Peer reviewer reports are available . Reprints and permissions information is available at http : / / www . nature . com / reprints . Article a b c Numerical K = 1 K = 10 K = 30 K = 50 K = 80 K = 100 20 0 y ( n m ) - 20 20 0 - 20 5 0 - 5 y ( n m ) 10 20 x ( nm ) x ( nm ) 10 20 x ( nm ) 10 20 x ( nm ) 10 5 0 - 5 - 10 10 5 0 - 5 - 10 y ( n m ) y ( n m ) 3 . 0 2 . 5 2 . 0 1 . 5 1 . 0 0 . 5 R ES I ( n m ) K 0 50 100 10 nm Extended Data Fig . 1 | RESI resolution estimation . a , A grid of defined positions of binding sites is generated ( top left ) , SMLM ( DNA - PAINT ) localizations are simulated as samples from a gaussian distribution ( top right ) . Localizations for only one binding site were plotted for clarity . For each binding site , subsets of K localizations are randomly selected ( bottom left ) and averaged ( bottom right ) . One exemplary subset and its average is highlighted . b , Resulting RESI - localizations are histogrammed to produce images at different resolutions ( K values ) . c , RESI - localization precision \u03c3 RESI vs K . Analytical dependence on K ( blue line ) and numerical results ( black dots ) . A total of 1200 SMLM localizations per site are simulated . Error bars represent mean \u00b1 1 s . d . Round 2 Round 1 Rounds 1 & 2 1 nm 0 . 1 nm 0 . 1 nm 5 . 4 nm 6 . 4 nm 6 . 1 nm 5 . 4 nm 6 . 2 nm 5 . 4 nm 0 . 3 nm 0 . 2 nm 0 . 3 nm 0 . 3 nm R4 R3 R1 20 nm 5 nm a b Round 1 c d e f g RESI DNA - PAINT Round 2 Average ( n = 90 origami ) 5 . 5 nm 20 nm 20 nm 20 nm 20 nm 9 0 n m 6 0 n m Extended Data Fig . 2 | RESI in 2D DNA origami . a , DNA origami design featuring six 5 nm - spaced orthogonal docking strand pairs ( red R1 , blue R3 ) and six alignment docking strands ( green R4 ) . See Methods for sequence details . b , DNA - PAINT acquisition parameters were tuned such that 5 nm were not consistently resolvable . c , First imaging round conducted with R1 ( target ) and R4 imagers ( alignment , sites circled ) . d , Second imaging round conducted with R3 ( target ) and alignment imagers ( R4 , sites circled ) . The alignment sites were used for translational and rotational alignment between rounds . e , RESI resolves the 5 nm distances . f , The distance and orientation between R1 and R3 docking strands are consistent with the design . g , An average of 90 DNA origami structures reveals consistent results and excellent alignment performance . The numbers indicate the distance between rounds . Article Extended Data Fig . 3 | 2D DNA origami . Representative DNA origami from across the field of view of the measurement . a , Four DNA origami , shown at DNA - PAINT resolution ( upper row ) and RESI resolution ( lower row ) . The insert depicts a pair of docking strands spaced at approx . 5 nm . b , 40 additional DNA origami , shown at DNA - PAINT resolution ( upper rows ) and RESI resolution ( lower rows ) . Round 2 Round 1 Rounds 1 & 2 R4 R3 R1 ii ) R4 i ) R1 / R3 Average ii ) R4 i ) R1 / R3 x y z i ) R1 / R3 ii ) R4 x y z i ) R1 / R3 ii ) R4 ( i ) ( ii ) Sum ( 88 DNA origami ) ( i ) ( ii ) ( i ) ( ii ) 10 nm 20 nm x y z x y z 3D DNA - PAINT Round 2 Round 1 c d e b a 3D RESI x y z x y z ( i ) ( ii ) 20 nm 20 nm 20 nm g f x y z x y z 10 nm Docking strand Docking strand at opposite side 11 nm R4 R3 R1 10 nm 10 nm 20 nm 20 nm h x y z 10 nm d z = 0 . 2 nm CI = [ 0 . 0 , 2 . 1 ] nm d = 1 . 3 nm CI = [ 0 . 3 , 2 . 3 ] nm CI = [ 0 . 3 , 2 . 2 ] nm d xy = 1 . 3 nm d z = 0 . 2 nm d = 0 . 4 nm CI = ( 0 . 0 , 1 . 4 ) nm CI = [ 0 . 0 , 1 . 0 ] nm CI = [ 0 . 0 , 1 . 6 ] nm d xy = 0 . 3 nm d z = 0 . 8 nm CI = [ 0 . 0 , 2 . 3 ] nm d = 1 . 0 nm CI = [ 0 . 0 , 2 . 2 ] nm CI = [ 0 . 0 , 1 . 3 ] nm d xy = 0 . 6 nm d z = 0 . 6 nm d = 0 . 7 nm CI = [ 0 . 0 , 1 . 8 ] nm CI = [ 0 . 0 , 1 . 8 ] nm CI = [ 0 . 0 , 1 . 1 ] nm d xy = 0 . 5 nm d z = 0 . 1 nm CI = [ 0 . 0 , 1 . 1 ] nm d = 0 . 1 nm CI = [ 0 . 0 , 0 . 8 ] nm x y z x y z CI = [ 0 . 0 , 0 . 6 ] nm d xy = 0 . 1 nm 1 nm 5 nm d z = 0 . 3 nm d = 0 . 4 nm CI = [ 0 . 0 , 1 . 1 ] nm CI = [ 0 . 0 , 1 . 2 ] nm CI = [ 0 . 0 , 0 . 7 ] nm d xy = 0 . 3 nm d = 11 . 6 \u00b1 0 . 8 nm d z = 11 . 3 \u00b1 0 . 8 nm d xy = 2 . 5 \u00b1 0 . 4 nm Extended Data Fig . 4 | RESI in 3D DNA origami . a , DNA origami design featuring one pair of orthogonal docking strands ( red R1 , blue R3 ) as well as six alignment docking strands ( green R4 ) . Docking strands extend from both the top and bottom surface of the DNA origami ( insert ) . b , The design ensures that all but the R1 / R3 docking strand pair are spaced sufficiently to be resolved by DNA - PAINT . c , 3D DNA - PAINT imaging resolves R4 alignment sites , barely resolves R1 / R3 axially and does not resolve R1 / R3 laterally . d , Sequential 3D DNA - PAINT imaging with R4 sites used for alignment . e , RESI resolves R1 / R3 both axially and laterally . f , An overlay of 88 DNA origami reveals overall good alignment despite structural heterogeneity . g , Average of 88 DNA origamis . h , The particle average recovers the structure with an alignment uncertainty of 0 . 7 nm CI = [ 0 , 1 . 6 ] nm , showing a distance between the average R1 / R3 positions of 11 . 6 \u00b1 0 . 8 nm ( xy - distance : 2 . 5 \u00b1 0 . 4 nm , z - distance : 11 . 3 \u00b1 0 . 8 nm ) , matching the designed distances 20 . Same scale applies to all magnification panels . CI describes 68 % confidence interval . Article 100 nm 3D DNA - PAINT 3D RESI 43nm 22nm 0nm Z pos . 3D RESI ( zoom in for details ) 3D DNA - PAINT 3D RESI ( zoom in for details ) 3D DNA - PAINT 3D RESI ( zoom in for details ) 3D DNA - PAINT 3D RESI ( zoom in for details ) 3D DNA - PAINT a b 100 nm Extended Data Fig . 5 | 3D DNA origami . Representative 3D DNA origami from across the field of view of the measurement . a , Five DNA origami , shown at DNA - PAINT resolution ( upper row ) and RESI resolution ( lower row ) . The color scale to the right represents the z position of localizations . The measured z coordinates for each DNA origami have been shifted by a constant such that the lowest localization for a given structure is defined to be at z = 0 . This ensures full use of the color range . b , 32 additional DNA origami , shown at DNA - PAINT resolution ( upper rows ) and RESI resolution ( lower rows ) . The z positions are colored according to the color scale in panel a . 3D RESI 3D DNA - PAINT 3D RESI 3D DNA - PAINT 3D DNA - PAINT 3D DNA - PAINT 3D DNA - PAINT 3D DNA - PAINT 100 nm 3D RESI 3D RESI 3D RESI 3D RESI 3D RESI 3D DNA - PAINT 100 nm 0nm 140nm 70nm Z pos . a b Extended Data Fig . 6 | U2OS Nup96 - mEGFP . Representative NPCs from across the field of view of the measurement . a , Six NPCs , measured using DNA - PAINT ( upper row ) and RESI ( lower row ) . The color scale to the right represents the z position of localizations . The measured z coordinates for each NPC have been shifted by a constant such that the lowest localization for a given structure is defined to be at z = 0 . This ensures full use of the color range . b , 72 additional NPCs , measured using DNA - PAINT ( upper rows ) and RESI ( lower rows ) . The z positions are colored according to the color scale in panel a . Article Extended Data Fig . 7 | Averaging of Nup96 proteins . a , Model - free averaging for DNA - PAINT measurements of Nup96 ( N = 1045 NPCs ) . An angled isometric view is shown . b \u2013 d , DNA - PAINT resolves nucleoplasmic and cytoplasmic rings and recapitulates their eight - fold symmetry , but fails to resolve individual Nup96 proteins . e , Side views of all Nup96 pairs in both rings reveal the angled orientation but do not resolve individual Nup96 proteins . f , Model - free averaging for RESI measurements of Nup96 ( N = 1190 NPCs ) . g \u2013 i , RESI recapitulates nucleoplasmic and cytoplasmic rings as well as their eight - fold symmetry and resolves individual adjacent Nup96 proteins in the majority of cases . j , Side views of all eight Nup96 pairs in both rings reveal the angled orientation as well as , in some cases , adjacent individual Nup96 proteins . k , The Cryo - EM structure of the nuclear pore complex indicates that a given Nup96 protein will have neighbors spaced at 11 nm , 39 nm , 71 nm , 93 nm and 101 nm . l , Performing clustering and nearest neighbor analysis for DNA - PAINT data reveals a peak at approx . 40 nm , corresponding to the distance between two Nup96 pairs , but not below that . RESI , on the other hand , features a first peak at approx . 15 nm , corresponding to the distance between adjacent Nup96 while taking linkage error ( label size ) into account . m , Analysis of first to tenth nearest neighbor distances for RESI and DNA - PAINT recapitulates the distances from ( k ) , but only RESI resolves the smallest distance . All scale bars : 20 nm . DNA - PAINT RESI RESI ( zoom in for details ) DNA - PAINT RESI ( zoom in for details ) DNA - PAINT RESI ( zoom in for details ) DNA - PAINT 100 nm a b 5\u00c5 5\u00c5 100 nm Extended Data Fig . 8 | Sub - nm DNA origami . Representative DNA origami from across the field of view of the measurement . a , Four DNA origami , shown at DNA - PAINT resolution ( upper row ) and RESI resolution ( lower row ) . The inserts show pairs of directly adjacent docking strands resolved by RESI . b , 42 additional DNA origami , shown at DNA - PAINT resolution ( upper rows ) and RESI resolution ( lower rows ) . Article Round 1 Round 2 d DNA origami design Distance measurement in single DNA origami Round 1 ( R1 or R4 ) Round 2 ( R3 or R4 ) d = 8 . 8 \u00b1 2 . 9 \u00c5 d = 8 . 9 \u00b1 2 . 5 \u00c5 d = 10 . 6 \u00b1 2 . 7 \u00c5 d = 9 . 7 \u00b1 2 . 2 \u00c5 d = 7 . 8 \u00b1 2 . 7 \u00c5 d = 11 . 4 \u00b1 2 . 6 \u00c5 d = 1 . 2 \u00c5 CI = [ 0 , 4 . 0 ] \u00c5 d = 1 . 1 \u00c5 CI = [ 0 , 4 . 1 ] \u00c5 d = 0 . 6 \u00c5 CI = [ 0 , 4 . 4 ] \u00c5 d = 1 . 5 \u00c5 CI = [ 0 , 5 . 6 ] \u00c5 d = 1 . 1 \u00c5 CI = [ 0 , 5 . 2 ] \u00c5 d = 1 . 5 \u00c5 CI = [ 0 , 4 . 3 ] \u00c5 Round 1 Round 2 Average Sum N = 42 R1 R3 R4 a b c 20 nm 5 \u00c5 20 nm 1 nm 20 nm Extended Data Fig . 9 | Sub - nm RESI measurements . a , DNA origami featuring six alignment strands ( green R4 ) and six pairs of orthogonal docking strands ( red R1 , blue R3 ) spaced one base pair apart . b , RESI representation with RESI - localizations from round 1 in red and round 2 in blue illustrates excellent alignment . The distances between RESI - localizations from round 1 and 2 are defined as illustrated . c , Overlaying 42 DNA origami and performing a particle average recovers the structure with an alignment uncertainty of 1 . 2 \u00c5 CI = [ 0 , 4 . 6 ] \u00c5 , showing distances between the average positions of the sites at 9 . 5 \u00b1 2 . 6 \u00c5 ( mean over six distances in the average \u00b1 mean over the error - propagated uncertainties of the six distances ) . Same scale applies to all magnification panels . CI describes 68 % confidence interval . 5 \u03bcm 5 \u03bcm 20 nm 20 nm 20 nm 20 nm 5 \u03bcm 20 nm 20 nm N - th NND ( nm ) 0 20 40 60 80 100 120 800 600 400 200 0 C oun t 5000 4000 3000 2000 1000 0 C ou n t Density = 49 . 9 \u03bcm 2 Density = 128\u03bcm 2 Density = 289\u03bcm 2 N - th NND ( nm ) 0 20 40 60 80 100 120 N - th NND ( nm ) 0 20 40 60 80 100 120 0 20 40 60 800 600 400 200 C oun t 0 1st NND ( nm ) 0 20 40 60 1st NND ( nm ) 0 20 40 60 1st NND ( nm ) 800 600 400 200 0 5000 4000 3000 2000 1000 0 C ou n t 1000 800 600 400 200 C oun t 0 1000 C oun t RESI DNA - PAINT a DNA - PAINT b c N - th NND analysis 1st NND analysis d e Extended Data Fig . 10 | RESI resolves CD20 dimers in untreated CHO cells for different expression levels . a , DNA - PAINT imaging of whole mEGFP - CD20 - expressing CHO cells , labeled with anti - GFP - nanobodies , shows homogeneously distributed molecules for three independent experiments . b , Zoom - in regions of DNA - PAINT show cases in which dimers could not be resolved . c , RESI reveals sub - 10 - nm spaced receptor pairs , which are unresolvable in the DNA - PAINT cases . d , Whole - cell analysis of first nearest neighbor distances ( 1 st NNDs ) of CD20 receptors ( histograms of the distances are displayed ) . Only RESI , but not DNA - PAINT , allows the routine detection of sub - 10 - nm distances between proteins . e , RESI - localization precision below 1 nm allows for routine detection of sub - 10 - nm distances , resulting in an accurate assessment of the first NND . Article 50 nm DNA - PAINT Cluster analysis RESI DNA - PAINT 5 \u03bcm DNA - PAINT 500 nm Diffraction - limited i ii iii i ii iii RESI Single molecule clustering 100 nm c DNA - PAINT RESI a b c Extended Data Fig . 11 | RESI resolves the substructure in RTX - induced chain - like arrangements of CD20 receptors with sub - nanometer precision . a , DNA - PAINT overview image of mEGFP - CD20 expressing CHO cells treated with RTX . b , Labeling with DNA - conjugated anti - GFP - nanobodies and imaging with DNA - PAINT reveals higher - order organization after RTX - treatment . RESI ( insets i \u2013 iii ) achieves molecular resolution and thereby resolves the molecular arrangement of mEGFP - CD20 . c , DNA - PAINT imaging shows clustered CD20 molecules . Performing RESI with sequences R1 , R2 , R3 and R4 in four separate imaging rounds ( color - coded ) allows for clustering of localizations originating from a single target . From the clustered localizations , RESI - localizations were calculated , enabling true single - protein resolution . N - th NND ( nm ) 2000 1500 1000 500 0 C oun t C oun t 2000 1000 0 4000 3000 2000 1000 0 1500 500 C oun t 0 20 40 60 5 \u03bcm 5 \u03bcm 5 \u03bcm 50 nm 50 nm 50 nm 50 nm 50 nm 50 nm Density = 116\u03bcm 2 Density = 63 . 6 \u03bcm 2 Density = 90 . 8 \u03bcm 2 0 20 40 60 0 20 40 60 1st NND ( nm ) 0 20 40 60 2000 1500 1000 500 0 C oun t C oun t 2000 1000 0 4000 3000 0 20 40 60 0 20 40 60 N - th NND ( nm ) 1st NND ( nm ) 1st NND ( nm ) 2000 1000 0 1500 500 C oun t RESI N - th NND analysis 1st NND analysis DNA - PAINT a DNA - PAINT b c N - th NND ( nm ) d e Extended Data Fig . 12 | RESI reveals higher order arrangement of CD20 dimers in Rituximab - treated CHO cells . a , DNA - PAINT imaging of whole mEGFP - CD20 - expressing CHO cells , labeled with anti - GFP - nanobodies , shows clustered CD20 - molecules in Rituximab - treated cells for three independent experiments . b , Zoom - in regions of DNA - PAINT show mEGFP - CD20 clustered into chain - like arrangements . c , RESI reveals sub - 10 - nm spaced receptor pairs within the clusters , unresolvable by DNA - PAINT . d , Whole - cell analysis of first nearest neighbor distances ( 1 st NNDs ) of CD20 receptors bound to Rituximab ( histograms of the distances are displayed ) . Only RESI , but not DNA - PAINT , allows the routine detection of sub - 10 - nm distances between proteins . e , Routine detection of sub - 10 - nm distances by RESI recapitulates the first NND measured in the untreated case . Notably the NND peaks measured in the three repeats are consistent , independently of the protein density . Article a c g d h b 0 . 0 0 . 2 Circularity 0 . 4 0 . 6 0 . 8 1 1 . 0 2 . 0 N o r m . F r equen cy Simulation Data Simulation Data Number of molecules per cluster N o r m . F r equen cy 0 . 2 0 . 4 0 . 3 0 . 1 0 5 10 15 20 Simulation Data Simulation ( clustered ) Data ( clustered ) 50 nm 50 nm 200 nm 200 nm 200 nm 200 nm e f Extended Data Fig . 13 | Comparison of Rituximab treated CD20 data to simulated CD20 hexamer arrangements . a , Example of ground truth simulated CD20 hexamers ( light blue circles , simulated as triangles of dimers with intra - dimer distances of 13 . 5 nm as measured experimentally ) with random distribution and orientation on a 2D surface at the experimentally determined density . b , Label uncertainty and labeling efficiency ( black circles indicate labeled molecules ) are taken into account in the simulation for a realistic comparison . c , Simulated proteins in hexameric arrangements represented as gaussians . d , Hexamers after DBSCAN cluster analysis ( colors indicate clusters ) . e , RESI image of CD20 data after RTX - treatment . f , RESI - localizations of CD20 data after DBSCAN cluster analysis ( colors indicate clusters ) . g , Number of molecules per detected cluster for the experimental data and the simulated hexamers . h , Circularity metric of experimental data and the simulated hexamers after convex hull analysis of the clusters . We note that the sharp drop at 0 . 605 stems from the maximum circularity metric for clusters where the convex hull is defined by three molecules . Notably , the absence of a circularity peak at ~ 0 . 7 in the experimental data suggests that CD20 molecules are not arranged in isolated ring - like hexameric structures . a b c d N o r m a li z ed f r equen cy 1st - NN distance ( nm ) F r a c t i on o f non - r e s o l v a b l e m o l e c u l e s 400 600 800 1000 200 0 200 400 600 800 1000 Density = 100 / \u00b5m 2 0 25 50 75 100 125 150 200 F r a c t i on o f non - r e s o l v ab l e m o l e c u l e s D en s i t y ( \u00b5 m - 2 ) x ( nm ) y ( n m ) Resolution ( nm ) 20 40 60 80 200 400 600 800 1000 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 Density ( \u00b5m - 2 ) 0 50 100 150 200 250 300 350 400 Resolution12 nm 19 nm 27 nm 35 nm 42 nm Extended Data Fig . 14 | Stochastic labeling . a , Exemplary simulation of proteins with a Complete Spatial Random ( CSR ) distribution of a given density . b , Histogram of Nearest Neighbor Distances ( NNDs ) . The red line indicates the smallest distance ( d ) that can be resolved by DNA - PAINT for a given set of imaging parameters . The fraction of molecules with a NN below this distance threshold ( blue , shaded ) can be computed for a given density and a given DNA - PAINT resolution . c , 2D map of the fraction of non - resolvable molecules as a function of density and resolution . d , 1D cuts of c at different resolutions ( color - coded ) can be used as a user guide to estimate the number of multiplexing rounds needed to perform RESI efficiently given a certain target fraction of non - resolvable distances . Article Extended Data Table 1 | Imaging and RESI parameters Overview of DNA - PAINT image acquisition parameters alongside clustering and RESI parameters .",
    "obashi2023conformational": "Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 A conformational switch in clathrin light chain regulates lattice structure and endocytosis at the plasma membrane of mammalian cells Kazuki Obashi 1 , Kem A . Sochacki 1 , Marie - Paule Strub 1 & Justin W . Taraska 1 Conformational changes in endocytic proteins are regulators of clathrin - mediated endocytosis . Three clathrin heavy chains associated with clathrin light chains ( CLC ) assemble into triskelia that link into a geometric lattice that curves to drive endocytosis . Structural changes in CLC have been shown to regulate triskelia assembly in solution , yet the nature of these changes , and their effects on lattice growth , curvature , and endocytosis in cells are unknown . Here , we develop a new correlative \ufb02 uorescence resonance energy transfer ( FRET ) and platinum replica electron microscopy method , named FRET - CLEM . With FRET - CLEM , we measure conformational changes in clathrin at thousands of individual morphologically distinct clathrin - coated structures . We discover that the N - terminus of CLC repositions away from the plasma membrane and triskelia vertex as coats curve . Preventing this conformational switch with chemical tools increases lattice sizes and inhibits endocytosis . Thus , a speci \ufb01 c conformational switch in the light chain regulates lattice cur - vature and endocytosis in mammalian cells . Clathrin - mediated endocytosis is the primary internalization pathway in eukaryotic cells and is key to many processes , including nutrient uptake , excitability , signaling , and the recycling of membrane components 1 . Over 50 proteins have been implicated in this process 2 , 3 . Determining the nanoscale localizations 4 , 5 , numbers 6 , 7 , accumulation dynamics 8 , 9 , interactions 1 , and conformations 10 , 11 of these proteins in cells is important for understanding how the endocytic machinery operates in health and disease . Assembly and curvature of clathrin lattices are required steps to build cargo - loaded vesicles 12 \u2013 14 . The basic unit of the clathrin lattice is the clathrin triskelion 15 , 16 . This six protein complex is composed of three clathrin heavy chains and three smaller clathrin light chains ( CLCs ) that form a three - legged pinwheel ( Supplementary Fig . 1a ) 17 . In non - neuronal cells , the stoichiometry of heavy chains to light chains might not always be one - to - one 18 . The heavy chains provide the backbone , and light chains are thought to regulate the assembly ofthe lattice and effect its mechanical properties 16 . Work both in vitro and in cells has indicated its important role 19 \u2013 23 , yet the structural , and func - tional actions of the light chains are not fully understood 16 , 24 . In vitro , clathrin triskelia assemble into spherical cages 13 . Cla - thrin light chain inhibits this assembly 25 , 26 . Light chain binds to heavy chain throughmultiple interactions 16 . One speci \ufb01 c interaction occurs through an acidic patch ( EED ) near the light chain \u2019 s N - terminus and the heavy chain \u2019 s knee ( Supplementary Fig . 1a ) 27 . This interaction prevents cage assembly by regulating the heavy chain knee con - formation ( Supplementary Fig . 1b ) 17 . If CLC binds to the heavy chain through these N - terminal contacts , CLC adopts an extended con - formation , stabilizing a straight conformation of the heavy chain knee . The straight conformation cannot assume the angles needed for cage assembly . However , if this interaction is prevented , the heavy chain knee is free to move , allowing the knee to bend and assemble as a cage . Thus , conformational switching of CLC is Received : 5 May 2022 Accepted : 25 January 2023 Check for updates 1 Biochemistry and Biophysics Center , National Heart , Lung , and Blood Institute , National Institutes of Health , 50 South Drive , Building 50 , Bethesda , MD 20892 , USA . e - mail : justin . taraska @ nih . gov Nature Communications | ( 2023 ) 14 : 732 1 1 2 3 4 5 6 7 8 9 0 ( ) : , ; 1 2 3 4 5 6 7 8 9 0 ( ) : , ; proposed to control the assembly of lattices 17 . Yet , the nature of these structural changes , and their effects on lattice growth , and curvature at the membrane of living cells are mostly unknown . Recent super - resolution imaging studies have mapped changes in protein locations at distinct stages of endocytosis at the scale of tens of nanometers 4 , 5 . Proteins are , however , regulated by sub - ten nanometer intra - and intermolecular conformational changes and binding events . To understand endocytosis , these conformational changes need to be determined 3 , 13 . This is a major gap in under - standing endocytosis . Yet , the precision of super - resolution imaging cannot measure structural changes at these scales . Fluorescence ( or F\u00f6rster ) resonance energy transfer ( FRET ) , however , can \ufb01 ll this gap , mapping molecular interactions and conformational changes within and between proteins 28 . FRET ef \ufb01 ciency depends on the spectral overlap between the donor and acceptor , distances between the dipoles , and their relative orientations to each other . Speci \ufb01 cally , FRET occurs when a donor \ufb02 uorophore and acceptor are generally separated by lessthan 10 nm 29 . WhileFRETmeasurementsareusually madewithdiffraction - limitedimaging , super - resolvedFRETmethods have been reported 30 . The resolution , labeling density , and colors possible in these experiments , however , are not able to discriminate morphological differences in small organelles such as clathrin - coated structures ( CCSs ) . For example , past FRET analysis of the protein organization of CCSs in yeast was limited to a late stage of endocytosis trapped by drug treatments 31 . Thus , existing methods cannot readily relate changes in protein conformation generated fromFRET to themorphologicalstagesofclathrin - coatedpitsasthey grow and curve . To overcome this gap , we turned to correlative light and electron microscopy ( CLEM ) 32 . CLEM can directly map \ufb02 uorescence signals to single nanoscale cellular structures visualized in paired electron microscopy ( EM ) images 33 . Platinum replica transmission EM ( PREM ) of unroofedcellplasmamembranes providesa uniquely high contrast , high - resolution , and wide - \ufb01 eld view of the inner plasma membrane of cells 34 . PREM has been combined with diffraction limited 35 , 36 , super - resolution 4 , 37 \u2013 40 , and polarized total internal re \ufb02 ection \ufb02 uorescence ( TIRF ) 41 microscopy to investigate molecular mechanisms of endocy - tosis , exocytosis , and the cortical cytoskeleton . Here , we develop a correlative lifetime based - FRET ( FLIM - FRET ) and PREM method , named FRET - CLEM . This method allows us to measure changes in protein interactions and conformations at distances less than 10 nm at single structurally - de \ufb01 ned organelles in single cells . We investigate the conformational changes in clathrin light chain by mapping distance changes both parallel and per - pendicular to the plasma membrane . We \ufb01 nd that the N - terminus of CLC moves away from both the CLC C - terminus and the plane of the plasma membrane as clathrin sites gain curvature . To determine the mechanistic impact of these structural changes , we develop a method to directly manipulate the N - terminal position of the cla - thrin light chain using a chemically - inducible dimerization system . These manipulations were con \ufb01 rmed with FRET . With acute che - mical perturbation , we show that inhibiting conformational chan - ges in CLC \u2019 s N - terminus increases clathrin lattice size , impairs maturation of clathrin structures at the plasma membrane , and inhibits transferrin endocytosis . Together , these data reveal a new conformational switch in clathrin light chain that regulates endo - cytosis in living cells . Results FRET - CLEM provides nanometer - scale spatial information at single clathrin sites To determine sub - ten nanometer conformational movements during endocytosis , we established a new correlative FLIM - FRET and PREM method , which we named FRET - CLEM ( Fig . 1 ) . We chose monomeric EGFP and a dark yellow \ufb02 uorescent protein , ShadowY 42 , as an optimized FRET pair by comparing six potential \ufb02 uorescent protein ( FP ) pairs ( Supplementary Fig . 2a ) . Calculations from the emission and absorbance spectra of this pair resulted in an estimated R 0 value ( distance of 50 % FRET ef \ufb01 ciency ) of 60\u00c5 ( Supplementary Fig . 2b ) . In line with these calculations , tandemly - connected EGFP - ShadowY probes fused to CLC showed > 50 % FRET ef \ufb01 ciency on the plasma membrane ( Supplementary Fig . 3a \u2013 d ) . We used this construct as a positive FRET control . Next , to testwhether FRET couldbe localized to single CCS , we performed FRET - CLEM measurements on HeLa cells expressing EGFP - CLC or EGFP - ShadowY - CLC ( Fig . 1a ) . First , cells were unroofed to expose the inner surface of the plasma membrane and rapidly \ufb01 xed 34 . These membranes provide uniquely high contrast , low background , ultra - thin samples for \ufb02 uorescence imaging 32 . Unroofed cells were then imaged with FLIM and subsequently prepared for platinum replica electron microscopy ( PREM ) . Past work has shown that these sample preparation steps do not measurably change the morphology and structure of the membrane and its associated endo - cyticorganelles 4 , 37 , 43 . This correlative method allows us to assign single diffraction - limited \ufb02 uorescent spots from a FLIM image to a single clathrinstructurevisualizedinPREM ( Fig . 1a ) . BecausePREMhasahigh spatial resolution , single \ufb02 uorescent spots in the FLIM image can be further classi \ufb01 ed according to the nanoscale structural features of the clathrin lattice ( Fig . 1b ) 43 . We then tested whether \ufb02 uorescence life - times can be analyzed at single CCS resolution . With the total photon number detected by our FLIM acquisition parameters ( Supplementary Fig . 3e \u2013 g , see also Methods ) , \ufb02 uorescence lifetime decay curves determined from single CCSs wereclearly separated between negative EGFP - CLC and positive EGFP - ShadowY - CLC FRET controls ( Fig . 1c ) . Speci \ufb01 cally , they could be \ufb01 t with a bi - exponential where the FRET ef \ufb01 ciency was determined for each CCS ( Supplementary Fig . 3h ) 44 . However , curve - \ufb01 tting was suboptimal for small CCSs due to the lim - ited number of photons 45 . Thus , to measure as many clathrin struc - tures aspossible across a cell , the mean \ufb02 uorescence lifetimewas used to estimate FRET ef \ufb01 ciencies across all structures ( Fig . 1d ) . These results show that FRET - CLEM can be used to generate FRET - based atomic - scale distances at single CCS resolution at morphologically distinct stages of endocytosis at the plasma membrane of mamma - lian cells . CLC N - terminal region moves away from the clathrin triskelion vertex Next , we applied FRET - CLEM to study conformational changes in CLC at the plasma membrane . In vitro , clathrin triskelia assemble into empty cages and CLC regulates this assembly 27 . CLC has been pro - posed to take two different conformations when bound to triskelia , an extended and bent conformation ( Fig . 2a ) 17 . Here , we assume three models of CLC conformations in cells based on in vitro models ( Fig . 2a ) . First , weassumethatunassembledtriskelioninthecytoplasm contains extended CLC and the spherical lattices in cells resembles in vitro assembled clathrin cages and contains bent CLC . Next , we propose three possible models for light chain dynamics during cur - vature . In the \ufb01 rst , CLC does not change conformations after assembly on the plasma membrane . Here , CLC assumes the bent conformation regardless of the curvature stage of the lattice ( \ufb02 at , domed , spherical ) . In the second , the proportions of extended and bent CLC gradually shift as curvature increases during endocytosis . In the third , as yet undescribed conformations of CLC are present in \ufb02 at and domed clathrin that switches to the bent conformation in spheres . To test these models , we used FRET - CLEM . Speci \ufb01 cally , the dis - tances between the N - terminus and C - terminus of surrounding CLCs differ substantially between the extended and bent conformations according to structural models 17 , 46 ( Fig . 2b , c , and Supplementary Fig . 1 ) . Although the expression levels of the transfected probes , the ratio between endogenous and transfected CLCs , and the fraction of heavy chains binding to CLC could affect FRET ef \ufb01 ciency , FRET Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 2 ef \ufb01 ciencies of the extended conformation are always larger than that of bent conformation when the degree of these factors are stable ( Supplementary Fig . 4 ) . This condition is satis \ufb01 ed when comparisons are made within a single cell . Thus , if the population of extended and bent conformations change on the plasma membrane during lattice assembly , FRET between the N - and C - terminus of CLCs is predicted to likewise change ( model 2 in Fig . 2a ) . Alternatively , if the conformations do not change , FRET should remain the same across all clathrin sub - types at the plasma membrane ( model 1 in Fig . 2a ) . To compare models 1 and 2 , we performed FRET - CLEM mea - surements on HeLa cells expressing EGFP - CLC , or EGFP - CLC and CLC - ShadowY ( Fig . 2d and Supplementary Fig . 5a , b ) . FPs are attached to CLCthrougha \ufb02 exiblelinker ( SupplementaryFig . 6 , seealsoMethods ) . The FP - attached CLCs were the dominant species within cells with our transfectioncondition ( SupplementaryFig . 7 , seealsoMethods ) . Mean \ufb02 uorescence lifetimes from single CCSs were analyzed by grouping them according to clathrin curvature classes determined from PREM images ( \ufb02 at , domed , and sphere ; Fig . 1b ) . There was no clear rela - tionship between mean \ufb02 uorescence lifetimes and photon counts ( Supplementary Fig . 8 ) . In cells expressing both EGFP - CLC and CLC - ShadowY , \ufb02 uorescence lifetimes increased as clathrin lattices curved ( Fig . 2d ) . This substantial increase did not occur when expressing only EGFP - CLC . The neuronal isoform of CLC which has an insertion of residues near the C - terminus 16 and was used in the previous in vitro study 17 showed similar lifetime changes ( Supplementary Fig . 9 ) . Fur - thermore , similar results were obtained in SK - MEL - 2 human cells a EGFP - CLC EGFP - ShadowY - CLC c b Flat Dome Sphere PREM FLIM 2 . 5 1 . 0 N o r m . c oun t s 1 0 . 01 0 . 001 0 . 1 Time ( ns ) 0 2 4 6 8 12 10 EGFP - CLC EGFP - ShadowY - CLC EGFP - CLC EGFP - ShadowY - CLC M ean f l uo r . li f e t i m e ( n s ) 2 . 5 2 . 0 1 . 5 d 200 nm PREM image Merge 500 nm F l uo r . li f e t i m e ( n s ) FLIM image 2 . 5 1 . 0 FRET Fluor . lifetime ( ns ) High Low F R E T High Low FRET Fig . 1 | FRET - CLEM . a Correlative FLIM - FRET and PREM images of unroofed membranes of HeLa cells expressing EGFP - CLC ( top ) or EGFP - ShadowY - CLC ( bot - tom ) . FLIM images ( left ; photon counts are represented by brightness and \ufb02 uor - escence lifetimes are represented by pseudo color ) , PREM images ( center ) , and mergeimages ( right ) . n = 1cellforeachcondition . Scale500nm . b PREMimagesof CCSs on an unroofed membrane of a HeLa cell expressing EGFP - CLC that were classi \ufb01 ed as \ufb02 at , domed , or sphere and corresponding areas from a FLIM image . n = 1 cell . Scale 200nm . c Fluorescence lifetime decays from single CCSs indicated by arrows in panel a . d Mean \ufb02 uorescence lifetimes from single CCSs on an unroofed membrane of HeLa cells expressing EGFP - CLC or EGFP - ShadowY - CLC . n = 58 CCSs from 1 cell ( EGFP - CLC ) and 81 CCSs from 1 cell ( EGFP - ShadowY - CLC ) . Forboxplots , boxisinterquartilerange , centerlineismedian , centercircleismean , whiskers are minimum and maximum data points with a coef \ufb01 cient value of 1 . 5 . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 3 indicating that these changes were not cell - type speci \ufb01 c ( Supplemen - taryFig . 10a \u2013 c ) . Inlivingcells , FRETef \ufb01 cienciesbetweenEGFP - CLCand CLC - ShadowY were higher on CCSs than in the cytosol ( Supplemen - tary Fig . 11 ) . However , unlike previous measurements in solution 17 , our FRET measurements re \ufb02 ected both intra - and inter - triskelia FRET . Next , we measured EGFP - CLC \u0394 N which is a truncation mutant of the \ufb02 exibleN - terminaldomain ( residues 1 \u2013 89 ) ( Fig . 2e and Supplementary Fig . 5c , d ) . In this mutant , EGFP would be bound near the heavy chain binding helix 15 . For this truncation , \ufb02 uorescence lifetimes did not substantially change across different lattice states . These data indicate that the N - terminal position of CLC undergoes a speci \ufb01 c conforma - tional switch at clathrin lattices on the plasma membrane during Flat Dome Sphere Model 1 CLC conformation FRET ( Fluor . lifetime ) Model 2 CLC conformation FRET ( Fluor . lifetime ) Model 3 New conformation CLC conformation FRET ( Fluor . lifetime ) ? ( ? ) Cell in vitro a b Extended Bent Unassembled Cage Low ( Long ) Low ( Long ) Low ( Long ) Bent Bent Bent Bent Bent Low ( Long ) High ( Short ) Low ( Long ) Extended + Bent d e f F l uo r . li f e t i m e c hange aga i n s t a v e r age o f f l a t c l a t h r i n ( % ) 1 0 - 1 2 F D S EGFP - CLC ( N only ) EGFP - CLC CLC - ShadowY ( N - C ) F D S F D S F D S EGFP - CLC (cid:2) N ( N only ) EGFP - CLC (cid:2) N CLC - ShadowY ( N - C ) 1 0 - 1 2 1 0 - 1 2 F D S F D S EGFP - QQN ( N only ) EGFP - QQN CLC - ShadowY ( N - C ) p < 0 . 001 p = 0 . 02 p = 0 . 005 p = 0 . 001 p = 0 . 004 p < 0 . 001 p = 0 . 004 FlatDomeSphere F l uo r . li f e t i m e c hange aga i n s t a v e r age o f f l a t c l a t h r i n ( % ) F l uo r . li f e t i m e c hange aga i n s t a v e r age o f f l a t c l a t h r i n ( % ) F R E T High Low Extended Bent c N - terminus C - terminus N - term . bent CLC N - term . extendedCLC New conformation ? ( ? ) Unassembled ( cytoplasm ) Extended Extended Extended Fig . 2 | CLC conformational changes in cells . a Schematic models of CLC con - formation at different assembly states in vitro or in living cells and their expected FRETef \ufb01 cienciesbetweentheN - andC - terminusofCLCs . b Astructuralmodelwith the assumption that either extended ( light blue ) or bent conformations ( magenta ) ofCLCsassembleintothelattice . ThemodelisbasedonPDB3LVGand6WCJ . 3LVG isoverlaidwith6WCJ . c Schematicmodelsoftwoassembledtriskeliawithextended ( left ) or bent CLCs ( right ) . A CLC N - terminus position ( yellow ) and C - terminus positions of surrounding CLCs ( orange ) are shown . d FRET - CLEM was performed on HeLa cells expressing either EGFP - CLC , or EGFP - CLC and CLC - ShadowY . Mean \ufb02 uorescence lifetimes from single CCSs were analyzed by categorizing them accordingtolatticestructures ( \ufb02 at , domedandsphere ) andtheywerecomparedto theaveragevaluesof \ufb02 atstructures . n = 6cellsfrom3experiments ( EGFP - CLC ) and n = 6 cells from 4 experiments ( EGFP - CLC and CLC - ShadowY ) . e FRET - CLEM on HeLacellsexpressingeitherEGFP - CLC \u0394 N , orEGFP - CLC \u0394 NandCLC - ShadowY . n = 6 cells from 5 experiments ( EGFP - CLC \u0394 N ) and n = 6 cells from 4 experiments ( EGFP - CLC \u0394 N and CLC - ShadowY ) . f FRET - CLEM on HeLa cells expressing either EGFP - QQN ( QQN mutant of CLC ) , or EGFP - QQN and CLC - ShadowY . n = 6 cells from 3 experiments for each condition . One - way ANOVA , then Tukey \u2019 s test . Each dot is from one cell experiment and errors are SE . For box plots , box is interquartile range , center line is median , center circle is mean , whiskers are minimum and maximum data points with a coef \ufb01 cient value of 1 . 5 . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 4 endocytosis . These data are not consistent with model 1 . Furthermore , \ufb02 uorescence lifetimes from cells expressing EGFP - CLC and ShadowY - CLC changed across different stages ( Supplementary Fig . 5g , h ) . This supports the movement of the CLC N - terminal region . Next , to further test model 2 , we measured QQN mutants of clathrin light chain ( resi - dues 20 \u2013 22 were substituted from EED to QQN ) ( Fig . 2f and Supple - mentary Fig . 5e , f ) . QQN mutants are de \ufb01 cient in binding to clathrin heavy chain near the triskelion vertex due to a loss of a negatively - charged patch at the N - terminus of the light chain ( Supplementary Fig . 1a ) 27 . Thus , they are inhibited from adopting the proposed extended conformation 17 . Here , if model 2 is correct , FRET across the structural states should differ between wild - type and QQN mutants . Speci \ufb01 cally , QQN mutants are expected to show a smaller displace - ment inconformations and smaller changes in \ufb02 uorescencelifetimeas the lattice curves . However , in these mutants , we found similar \ufb02 uor - escence lifetime changes as in the wild - type protein ( Fig . 2d , f ) . This similarity in \ufb02 uorescence lifetime change was observed among the range of expression levels obtained in our experimental conditions ( Supplementary Fig . 12 ) . These data do not support model 2 . Thus , we conclude that CLC structural movements at the plasma membrane cannot be described by current in vitro models ( model 1 or 2 ) . CLC N - terminal region moves away from the plasma membrane Next , we measured FRET between EGFP positioned in different domains of clathrin light chain and the membrane resident dark FRET - based quencher dipicrylamine ( DPA ) ( Supplementary Fig . 2c ) 47 . Because the plane of the membrane is \ufb01 xed , FRET between EGFP and DPA provides relative distances perpendicular to the plane of the plasma membrane 48 , 49 . Speci \ufb01 cally , when EGFP sits near the mem - brane , the FRET ef \ufb01 ciency is high and \ufb02 uorescence lifetimes are short ( Fig . 3a ) . With DPA , CLC - EGFP showed the shortest lifetime ( Fig . 3b , c ) . ThisindicatesthattheC - terminusofCLCislocatedclosetotheplasma membrane . In contrast , EGFP - CLC showed the longest \ufb02 uorescence lifetime . EGFP - CLC \u0394 N showed an intermediate lifetime . From these data , we can position segments of CLC relative to the plane of the plasma membrane ( Fig . 3d ) . Unlike classic models 17 , we \ufb01 nd that the N - terminus of CLC bound to clathrin - coated structures at the plasma membrane is located farther into the cytoplasm than the proximal leg of clathrin heavy chain . To investigate conformational changes during lattice curvature , we performed FRET - CLEM measurements with DPA and compared those measurements to morphological changes in clathrin lattice ( Fig . 3e and Supplementary Fig . 13 ) . In these experiments , \ufb02 uores - cence lifetimes of EGFP - CLC increased as lattices curved , and the degree of change was larger than those for CLC - EGFP . Similar results were obtained for cells expressing SNAP - CLC - EGFP or EGFP - CLC - SNAP ( Supplementary Fig . 14 ) . In addition , in SK - MEL - 2 cells , \ufb02 uor - escence lifetimes of EGFP - CLC increased as lattices curved , and the degree of change was similar to that of HeLa cells ( Supplementary Fig . 10d , e ) . These data are consistent with a model where the N - terminal position of CLC changes during curvature . Thus , the combined results from EGFP - ShadowY and EGFP - DPA FRET indicate that the N - terminus of CLC extends away from the CLC C - terminus ( triskelion vertex ) and the plane of the plasma membrane during endocytosis ( Fig . 3f ) . Conformational changes in CLC regulate lattice structure and endocytosis Next , wetestedwhethertheseconformationalchangesinCLCregulate clathrin - mediated endocytosis . To manipulate the CLC N - terminal position directly , we used a rapamycin - inducible FKBP / FRB dimeriza - tion system 50 . Speci \ufb01 cally , we employed the T2098L mutant of FRB which is heterodimerized to FKBP by the rapamycin analog , AP21967 51 . We attached FKBP to the N - terminus of CLC , and one or two FRBs ( FRB\u00d72 ) to the C - terminus of the membrane - bound PH domain from PLC \u03b4 1 ( PH ) 52 ( Fig . 4a ) . We made both FRB and FRB\u00d72 probes to extend the reach of the system . To con \ufb01 rm that the N - terminal position of FKBP - EGFP - CLC changes relative to the plasma membrane after dimerization , we measured EGFP \ufb02 uorescence lifetimes without or with AP21967 in the presence of DPA ( Fig . 4b , c ) . Changes in the posi - tion of EGFP can be estimated by comparing the difference in \ufb02 uor - escence lifetimes between the control and DPA - treated membranes ( Fig . 4d ) . For both probes , differences in EGFP lifetimes with AP21967 treatment were larger than those without AP21967 . These results indicate that the CLC N - terminus moves towards the plasma mem - brane as a result of FKBP / FRB dimerization . Because PH - miRFP - FRB\u00d72 showed larger lifetime change ( Fig . 4d ) and stronger accumu - lation with AP21967 than PH - miRFP - FRB ( Supplementary Fig . 15a ) , we used PH - miRFP - FRB\u00d72 ( or PH - mCherry - FRB\u00d72 ) to alter the structure of clathrin light chain in living cells . First , usingourFKBP / FRBconstructs , weinvestigatedtheimpactof moving the CLC N - terminus towards the plasma membrane on the structure of clathrin lattices 43 . Here , unroofed membranes from cells expressing FKBP - EGFP - CLC and PH - miRFP - FRB\u00d72 and treated without or with AP21967 were imaged with PREM ( Supplementary Fig . 16 ) . Unlike the previous homodimerization system 53 , clathrin lattices were not noticeably distorted . The average size of all visible \ufb02 at and domed clathrin structures increased with AP21967 treatment [ 31463\u00b1859nm 2 ( ctrl ) and 42919\u00b12278nm 2 ( AP ) for \ufb02 at , 29821\u00b1576nm 2 ( ctrl ) and 36471\u00b12862nm 2 ( AP ) for domed , 15311\u00b1588nm 2 ( ctrl ) and 17303\u00b1691nm 2 ( AP ) for sphere ] ( Fig . 5a - c ) . We conclude that manip - ulation of the CLC N - terminal position directly effects the structure of clathrin lattices . Next , toinvestigatehowchangingCLCconformationseffectsCCS assembly and maturation , we performed live cell time lapse evanes - cent \ufb01 eld imaging . Cells expressing FKBP - EGFP - CLC and PH - miRFP - FRB\u00d72 were imaged with total internal re \ufb02 ection \ufb02 uorescence micro - scopy ( TIRF ) in the absence or presence of AP21967 ( Fig . 5d , e ) . To quantitate CCS dynamics , we analyzed the residence times of FKBP - EGFP - CLC spots on the plasma membrane . The residence times became shorter after AP21967 addition . This indicates that tethering the CLC N - terminal position towards the plasma membrane likely facilitates the disassembly of CCSs . Finally , we investigated whether manipulation of the CLC N - terminal position effects transferrin endocytosis in whole cells ( Fig . 5f , g and Supplementary Fig . 17a , b ) . As a control for binding of FRB to the CLC N - terminus , a cytosolic probe , mCherry - FRB\u00d72 , was used . Furthermore , as a control of clathrin tethering to the plasma membrane , and also FRB accumulation on CCSs , a C - terminal - linked FKBP probe , CLC - FKBP - EGFP , was tested . For all three probe pairs , FKBP / FRB dimerization was con \ufb01 rmed by clustering of FRB at CCSs ( SupplementaryFig . 15b ) andchangesinFRETef \ufb01 ciencybetweenEGFP and mCherry ( Supplementary Fig . 15c ) . Expression of these probes did not change transferrin uptake without AP21967 treatment ( Supple - mentary Fig . 17b ) . Consistent with this \ufb01 nding , the amount of surface transferrin receptors was not changed by the expression of FKBP / FRB probes ( Supplementary Fig . 17c , d ) . We found that transferrin uptake decreased with AP21967 for PH \u2013 mCherry - FRB\u00d72 but not for the con - trol mCherry - FRB\u00d72 ( Fig . 5g ) . Furthermore , for the C - terminus - attached FKBP control ( CLC - FKBP - EGFP ) , PH - mCherry - FRB\u00d72 did not alter transferrin uptake ( Fig . 5g ) . These results indicate that manip - ulation of the CLC N - terminalposition inhibited transferrinuptake and endocytosis . These data are consistent with the decrease in the resi - dence time of CCSs we observed in live cell imaging . From these col - lective data , we concluded that the conformational changes in CLC we mapped in FRET - CLEM experiments are important structural changes required for endocytosis of cargo - loaded vesicles in mammalian cells . Thus , the movement of the N - terminal domain away from the clathrin vertex and plasma membrane is a key regulatory step in clathrin - mediated endocytosis . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 5 High FRET ( Short fluor . lifetime ) PM DPA EGFP 2 . 5 2 1 . 5 1 0 . 5 0 M ean f l uo r . li f e t i m e ( n s ) 0 100 200 DPA concentration ( (cid:3) M ) EGFP - CLC (cid:2) N EGFP - CLC ( N - term ) CLC - EGFP ( C - term ) a b DPA0 (cid:3) M 20 (cid:3) M CLC - EGFP ( C - term ) EGFP - CLC (cid:2) N c d e 4 F l uo r . li f e t i m e c hange aga i n s t a v e r a ge o f f l a t c l a t h r i n ( % ) 3 1 0 - 1 2 Flat Dome Sphere Flat Dome Sphere CLC - EGFP ( C - term ) EGFP - CLC ( N - term ) p < 0 . 001 p < 0 . 001 p < 0 . 001 p < 0 . 001 p = 0 . 002 f PM Low FRET ( Long fluor . lifetime ) EGFP 2 . 5 0 . 5 Fluor . lifetime ( ns ) FRET Distance from PM Near Far High Low F R E T High Low Sphere Flat PM EGFP - CLC ( N - term ) 10 (cid:3) m EGFP - CLC ( N - term ) EGFP - CLC (cid:2) N CLC - EGFP ( C - term ) Fig . 3 | CLC N - terminal position perpendicular to the plane of the plasma membrane . a Dipicrylamine ( DPA ) is a non \ufb02 uorescent hydrophobic anion that incorporates into membranes . DPA quenches EGFP in a distance dependent man - ner by FRET . PM is the plasma membrane . b FLIM images of unroofed membranes ofHeLacellsexpressingCLC - EGFP ( left ) , EGFP - CLC \u0394 N ( center ) , orEGFP - CLC ( right ) without ( top ) or with 20 \u03bc M DPA ( bottom ) . Scale 10 \u03bc m . c Mean \ufb02 uorescence lifetimes with different DPA concentrations . n = 16 ( EGFP - CLC ) , 17 ( EGFP - CLC \u0394 N ) , and 15 cells ( CLC - EGFP ) from 3 experiments . Errors are SE . d A structural model of EGFPpositionsofCLCprobespredictedfromtheEGFP - DPAFRETexperiments . The model is based on PDB 4KW4 and 3LVG . e FRET - CLEM on HeLa cells expressing either CLC - EGFP or EGFP - CLC with DPA . DPA concentrations were 3 \u03bc M for CLC - EGFPand80 \u03bc MforEGFP - CLCtoobtain ~ 50 % FRETef \ufb01 cienciestomakethedegree of \ufb02 uorescence lifetime changes similar . Mean \ufb02 uorescence lifetimes from single CCSs were analyzed by categorizing them according to lattice structures and compared to average values of \ufb02 at structures . n = 6 cells from 4 experiments for each condition . Each dot is from one cell experiment and errors are SE . One - way ANOVA , then Tukey \u2019 s test . For box plots , box is interquartile range , center line is median , center circle is mean , whiskers are minimum and maximum data points withacoef \ufb01 cientvalueof1 . 5 . f AproposedstructuralmodelofCLCconformational changes predicted from both FRET - CLEM with EGFP - ShadowY and EGFP - DPA . The N - terminus of CLC moves away from both the CLC C - terminus ( triskelion vertex ) and the plane of the plasma membrane as clathrin lattices curve . The structural model is based on PDB 3LVG . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 6 Discussion We have explored the molecular - scale conformational changes in clathrin light chain at clathrin sites at the plasma membrane of mam - maliancells . We \ufb01 ndthattheN - terminusofCLCmovesawayfromboth the triskelion vertex and the plasma membranes as clathrin lattices curve . Blocking this movement increased clathrin lattice sizes at the plasma membrane and reduced transferrin endocytosis . Thus , speci \ufb01 c structural changes in CLC in clathrin lattices at the plasma membrane are regulators of curvature and endocytosis in living mammalian cells . Conformational changes in CLC have been proposed to regulate clathrin assembly 17 . We mapped the speci \ufb01 c changes in cells by measuring vectors in two orthogonal axes . First , EGFP - DPA FRET measurements positioned the light chain relative to the plane of the membrane ( Fig . 3d ) , and reported the movement along the axis perpendicular to the plasma membrane ( Fig . 3f ) . Unlike classic in vi - tro models 17 , we \ufb01 nd that the N - terminus of CLC is located farther into the cytoplasm than the proximal leg of clathrin heavy chain and moves away from the membrane . With intermolecular FRET between EGFPandShadowY at theN - and C - terminus ofthelight chain ( Fig . 2 ) , we could map structures along the axis parallel to the plasma mem - brane . Here again , we detected a displacement of the N - terminal domain away from the CLC C - terminus ( clathrin lattice vertex ) . Combined , these data indicate that the N - terminus of CLC moves away from both the clathrin vertex and the plane of the plasma membrane as lattices gain curvature . Although the exact position of the N - terminus in relation to the heavy chain proximal leg domain was not determined , our experi - ments could be used to provide additional insights . Speci \ufb01 cally , we modeled how FRET ef \ufb01 ciencies between EGFP - CLC / CLC - ShadowY ( Supplementary Fig . 5b ) or EGFP - CLC / ShadowY - CLC ( Supplementary Fig . 5g ) would change as the N - terminal position changes according to the x - ray structures 17 across all possible spatial positions PH domain miRFP FRB FRB CLC FKBP EGFP AP21967 a b Potential FRET pairs EGFP and miRFP EGFP and miRFP EGFP and DPA EGFP and DPA AP21967 DPA Clathrinlattice Plasmamembrane FKBP - EGFP - CLC PH - miRFP - FRB\u00d72 c d Ctrl - + DPA - + AP - + - + Ctrl AP 2 1 M ean f l uo r . li f e t i m e ( n s ) PH - miRFP - FRB PH - miRFP - FRB\u00d72 (cid:2) f l uo r . li f e t i m e ( n s ) [ D PA ( 0 (cid:3) M ) - D PA ( 20 (cid:3) M ) ] 0 1 0 . 5 Ctrl AP Ctrl AP PH - miRFP - FRB PH - miRFP - FRB\u00d72 2 . 5 1 . 5 Fig . 4 | Manipulation of CLC N - terminal position using a chemically inducible dimerization system . a Schematic models of the chemically inducible FKBP / FRB dimerization system . FKBP is attached to the N - terminus of CLC , and two FRBs ( FRB\u00d72 ) are attached to the C - terminus of PH domain from PLC \u03b4 1 . A rapamycin analog , AP21967 , induces heterodimerization between FKBP and the T2098L mutant of FRB . b Potential FRET pairs without or with AP21967 treatment either absence or presence with DPA . c Fluorescence lifetime measurements were per - formed on unroofed membranes of HeLa cells expressing FKBP - EGFP - CLC either with PH - miRFP - FRB , or PH - miRFP - FRB\u00d72 without or with 20 \u03bc M DPA . Cells were unroofed after 15min incubation with AP21967 or ethanol ( control ) . n = 20 ( PH - miRFP - FRB , control ) , 19 ( PH - miRFP - FRB , AP21967 ) , 20 ( PH - miRFP - FRB\u00d72 , control ) , and 19 cells ( PH - miRFP - FRB\u00d72 , AP21967 ) from 3 experiments . d Differences in \ufb02 uorescencelifetimes ( shown inpanel c ) without and withDPA . For box plots , box is interquartile range , center line is median , center circle is mean , whiskers are minimum and maximum data points with acoef \ufb01 cient value of 1 . 5 . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 7 A r ea o cc upa t i on ( % ) 4 0 8 d f a 50 m Tf - AF647 FKBP - EGFP - CLC PH - mCherry - FRB\u00d72 N o r m . T f - A F 647 f l uo r . i n t en s i t y 1 0 2 - + Transfection - + - + AP21967 FKBP - EGFP - CLC PH - mCherry - FRB\u00d72 FKBP - EGFP - CLC mCherry - FRB\u00d72 CLC - FKBP - EGFP PH - mCherry - FRB\u00d72 p < 0 . 001 g AP 21967 2 D p r o j e c t i on a r ea ( 10 n m \u00b2 ) 4 2 0 6 Ctrl AP 1 0 2 0 . 3 0 0 . 6 4 2 0 6 1 0 2 p < 0 . 001 p = 0 . 04 Flat Dome Sphere N u m be r ( / m 2 ) 2 0 1 0 0 . 5 0 0 . 4 0 . 2 erehpS emoD talF erehpS emoD talF Ctrl AP Ctrl AP Ctrl AP Ctrl AP Ctrl AP Ctrl AP Ctrl AP Ctrl AP No treatment AP21967 200 0 R e s i d en c e t i m e ( s ) 400 p < 0 . 001 10 0 20 30 F r eq uen cy ( % ) 10 0 20 30 76912PA tnemtaert oN e Residence time ( s ) 65 125 185 245 305 > 320 65 125 185 245 305 > 320 Residence time ( s ) C on t r o l c b 3 Fig . 5 | Manipulation of CLC conformation changed lattice structures , dynamics , and endocytosis . a \u2013 c Unroofed membranes from cells expressing FKBP - EGFP - CLC and PH - miRFP - FRB\u00d72 treated with AP21967 ( AP ) or ethanol ( con - trol ) were imaged with PREM . Two - dimension area of single CCS were manually segmented and measured . Membrane area occupation against the total analyzed membrane area ( a ) , two - dimension projection area ( b ) , and density ( c ) of \ufb02 at , domed , andsphereCCSswerecompared . Eachdotisfromonecellexperiment , and errors are SE . n = 6 cells from 3 experiments for each condition . The average measured area / cell ( mean\u00b1SE ) = 213\u00b134 ( control ) and 165\u00b110 \u03bc m 2 ( AP21967 ) . d Live cell time lapse TIRF imaging on HeLa cells expressing FKBP - EGFP - CLC and PH - miRFP - FRB\u00d72 without or with AP21967 treatment . Tracks with over 20s were analyzed and residence times were compared . n = 206 spots from 5 cells from 3 experiments ( no treatment ) and 234 spots from 5 cells from 4 experiments ( AP21967 ) . A two - sided unpaired t test was used . e Histogram of residence times . f Confocal projection images of Alexa Fluor 647 conjugated transferrin ( Tf - AF647 ) uptake in HeLa cells expressing FKBP - EGFP - CLC and PH - mCherry - FRB\u00d72 treated with AP21967 or ethanol ( control ) . Fluorescence intensity of Tf - AF647 ( arbitrary units ) isrepresentedbypseudocolor . n = 3experiments . Scale50 \u03bc m . g Transferrin uptake in HeLa cells expressing FKBP and FRB probes treated with AP21967 . Fluorescence intensities of incorporated Tf - AF647 normalized by non - transfected cells in the same sample were compared between non - transfected and transfected cells . n = 116 \u2013 198cellsfrom3experimentsforeachconditions . Forboxplots , boxis interquartile range , center line is median , center circle is mean , whiskers are minimum and maximum data points with a coef \ufb01 cient value of 1 . 5 . A two - sided unpaired t test was used . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 8 ( Supplementary Fig . 18a ) . Although the fraction of heavy chains binding to ShadowY - attached CLC might affect the overall gross ef \ufb01 - ciency of FRET , the positional dependency and direction of change would not changewithexpressionlevelsaccordingtoourcalculations . Because FRET ef \ufb01 ciencies were largely similar for both sites , the N - terminus of CLC is predicted to rest in the overlap regions that match these paired distances ( Supplementary Fig . 18b , magenta cir - cles ) . To accommodate these constraints , CLC would assume a slightly folded conformation ( orange and green in Supplementary Fig . 18c ) rather thana stretched conformation ( blue in SupplementaryFig . 18c ) . From these considerations we propose a new model of CLC con - formational changes ( Fig . 6 ) . Here , CLC assumes an extended and tightly bound conformation in unassembled triskelia in the cytoplasm similar to that seen in x - ray crystal structures and measured with FRET 17 . Next , whenCLCassemblesatthemembraneasa \ufb02 at lattice , the light chain changes conformations from this extended position into a new folded conformation . The N - terminus is then displaced deeper intothecytosolasclathrinlatticescurveintovesicles . Insupportofour data , the average position of the CLC N - terminal domain is not visible by cryo - EM in puri \ufb01 ed clathrin - coated vesicles 46 . Thissuggests thatthe N - terminal domain is \ufb02 exible and positioned away from the lattice in highly curved structures . What initiates this structural change in the light chain ? HIP and HIP1R interact with the CLC N - terminal residues 54 , 55 . Although CLC QQN mutants are thought to be de \ufb01 cient in binding HIP1 and HIP1R 55 , we found that FRET between the N - and C - terminus of the CLC QQN mutants showed FRET changes comparable to wild type proteins ( Fig . 2d , f ) . Thus , if this region is necessary for binding , HIP1R inter - actions are unlikely to drive these conformational changes . The bind - ing of other proteins to the lattice 16 , 24 , changes in protein - protein interactions caused by cargo loading 1 , or phosphorylation 21 are alter - nate mechanisms that could drive this change . Finally , physical mechanismssuchas crowding and cargoloading intothe vesicle could induce the conformational switch 2 , 13 . In vitro studies have shown that CLC increases the rigidity of clathrin lattices on a solid substrate 20 . These effects were different between different CLC isoforms 23 . In living cells , CLC has been shown to be important for endocytosis under conditions of high membrane tension 19 , 22 . Here , our FKBP / FRB experiments demonstrate that con - formational changes in CLC regulate the structure of clathin ( Fig . 5a \u2013 c ) , the dynamics of single clathrin sites in live cells ( Fig . 5d , e ) , and the endocytosis of cargo in whole cells ( Fig . 5f , g ) . One possible interpretationisthatthe latticecangrow but not fully curve due to the inhibition of the CLC conformation switch . Possibly , clathrin lattices grow larger because treatment disturbed the balance between lattice assembly and curvature 12 . These irregularly - assembled CCSs are then disassembled by proofreading mechanisms 3 . The CLCs role in endo - cytosis differs between isoforms , cargos , and cell types 19 , 21 \u2013 23 , 56 \u2013 59 and the physiological function of CLC are not fully understood 16 , 24 . Clar - i \ufb01 cation on how CLC conformations are regulated acrossisoformsand cargos will provide a uni \ufb01 ed view of the light chain \u2019 s mechanistic roles across different cells , tissues , and states . Our work has some speci \ufb01 c limitations . For example , we expressed FP - tagged probes to measure FRET . Thus , the expression ratio between endogenous and FP - labeled proteins could affect the overall FRET ef \ufb01 ciency ( Supplementary Figs . 4c and 18a ) . This , along with other photophysical issues , makes it dif \ufb01 cult to convert FRET ef \ufb01 ciencies to absolute atomic distances 60 . However , the differences in expression level do not affect the direction of FRET changes caused by conformational changes in this system ( Supplementary Figs . 4c , 12 , and 18a ) . Likewise , the fact that our two orthogonal FRET - based experiments using either FRET between two \ufb02 uorescent pro - teins or FRET between a \ufb02 uorescent protein and a probe in the membrane all showed similar vectors of movement for the light chain strongly supports our structural models . In the future , dual - tagging all alleles of the endogenous protein is a direction to further re \ufb01 ne these measurements 31 . Also it will be interesting to investigate the relationship between the stoichiometry of proteins and FRET with a combination of other optical techniques . Furthermore , the size of \ufb02 uorescent proteins , while relatively small compared to the size of the clathrin complex and approaching the size of large red organic dyes , may affect a protein \u2019 s structure . Likewise , larger probes can limit the ability to detect small or distant conformational changes 60 . Thus , the incorporation of smaller tags or arti \ufb01 cial amino acids are future directions to improve these measurements 61 , 62 . Indeed , smaller probes have been shown to better recapitulate absolute distance changes seen during conformational transitions and might have less of an in \ufb02 uence on the proteins themselves 29 . In addition , increasing photon counts with organic dyes , combined with new quantitative analysis of photon count data , will be impor - tant for improving this method . In the case of clathrin light chain , however , fusion of \ufb02 uorescent proteins has not been seen to perturb endocytosis or change clathrin lattice structures 63 . Expression of CLC probes used in FRET experiments did not change the size and number of clathrin lattice structures ( Supplementary Fig . 19 ) . FRET ef \ufb01 ciencies are determined by photophysical properties ( excitation , emission , absorbance ) , distance , and orientation of the probes 28 , 44 . Here , we used a \ufb02 exible linker to connect the FPs to CLC ( Supplementary Fig . 6 and Methods ) . Thus , it is reasonable to propose that the population of FPs in a CCS can assume many possible orien - tations and the effect of an orientation bias would be small . However , we cannot exclude the small possibility that the curvature of clathrin lattices causes slight differences in the orientations between the donor and acceptor , which may affect the overall FRET ef \ufb01 ciencies . However , the fact that the N - terminal changes was observed in both two FP - FRET experiments and FP - DPA FRET experiments , again supports the idea that orientation effects are not large or dominant . Combined , these caveats make it dif \ufb01 cult to directly equate FRET ef \ufb01 ciencies to atomic distances . ThisiscommonlythecaseinFRETstudiesdonewithincells 60 . For cells expressing only EGFP probes , although there were no substantial change in \ufb02 uorescence lifetimes among different lattice structures , clathrin vesicles tended to show slightly higher lifetimes Plasmamembrane Flat Dome Sphere Bent model Cage Cell in vitro Extended model Unassembled Clathrinlattice Proposed model Sphere Flat Fig . 6 | ModelsofconformationalswitchinCLCinlivingcells . Schematicmodels of conformationalswitchin CLCatthe plasma membrane incells . CLC assumes an extended conformation in unassembled triskelia in the cytoplasm similar to that seeninx - raycrystalstructures . Next , whenCLCassemblesatthemembraneasa \ufb02 at lattice , CLC changes conformations from the extended to a new folded con - formation . The N - terminus is then displaced deeper into the cytosol as clathrin latticescurveintovesicles . TheextendedandbentmodelsarebasedonPDB3LVG . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 9 than \ufb02 atclathrinstructures ( Fig . 2 ) . BecauseHomo - FRETisnotthought to change the overall \ufb02 uorescencelifetime 44 , this might be caused by a difference in the local environment 64 , or a systematic counting error due to the differences in the photon count rate among CCSs with different \ufb02 uorophore densities 65 . These effects , however , were small and consistent across samples . Here , we developed a new correlative FLIM - FRET and PREM method to track the conformational changes in CLC at single sites of endocytosis in cells . The N - terminus of CLC makes a dramatic movement away from the clathrin lattice vertex and deeper into the cytosol as lattices curve . These conformational dynamics are key for clathrin - mediated endocytosis . These data , combined with the rich biochemical , functional , genetic , and biophysical information on membrane traf \ufb01 c , will lead to a more robust understanding of how endocytic proteins work together during clathrin - mediated endocy - tosis to drive the internalization of cargo . More generally , our new method maps molecular interactions and conformational changes of targeted proteins at identi \ufb01 ed sites in the complex environment of the cell . Thus , FRET - CLEM can be used to investigate conformational changes of any accessible membrane - associated protein including ion channels , transporters , receptors , adhesion proteins , or enzymes , in the context of their local plasma membrane environments 29 , 32 . Methods Cell culture HeLa cells ( ATCC , CCL - 2 ) were maintained at 37\u00b0C , with 5 % CO 2 in DMEM ( Gibco , 11995073 ) supplemented with 10 % fetal bovine serum ( ATLANTA biological , S12450H ) and 1 % vol / vol penicillin / streptomycin ( Invitrogen , 15070 - 063 ) . SK - MEL - 2 cells ( ATCC , HTB - 68 ) were main - tained at 37\u00b0C , with 5 % CO 2 in EMEM ( ATCC , 30 - 2003 ) supplemented with 10 % fetal bovine serum and 1 % vol / vol penicillin / streptomycin . The cells were placed on poly - L - lysine coated coverslips ( Neuvitro , GG - 25 - 1 . 5 - PLL ) . They were transfected with 0 . 75mL of Opti - MEM ( Life Tech - nologies , 31985062 ) , 3 . 8 \u03bc L Lipofectamine 2000 ( Life Technologies , 11668027 ) , and 1 . 5 \u03bc g of DNA for 3h after being introduced to the cells . Then , transfected cells were incubated in DMEM or EMEM growth medium for 20 \u2013 30h before experiments . The cells tested negative for mycoplasma contamination . Plasmids ShadowY ( # 104621 ) , mScarlet - I ( # 85068 ) , mTurquoise2 ( # 54842 ) , miRFP703 - CLCb ( # 79997 ) , CFP - FKBP ( # 20160 ) , Nup54 - EGFP494 ( 0 ) ( # 163426 ) , Nup54 - EGFP494 ( 1 ) ( # 163427 ) , Nup54 - EGFP494 ( \ufb02 ex0 ) ( # 163429 ) , andNup54 - EGFP494 ( \ufb02 ex1 ) ( # 163430 ) werepurchasedfrom Addgene , and pSNAP - tag ( m ) was purchased from New England Bio - labs . CLCa - GFP was kindly donated by Dr . W . Almer ( Oregon Health & Science University ) . Lyn - GFP - FRB was kindly donated by Dr . T . Inoue ( Johns Hopkins University ) . mNeonGreen was kindly donated by Dr . J . Shah ( Harvard University ) . mCherry - PH was from our previous studies 66 . All plasmids used in this study were constructed using either Q5 Site Directed Mutagenesis Kit ( New England Biolabs , E0554S ) or In - Fusion HD Cloning Plus ( Clonetech , 638920 ) following manufacturer \u2019 s instructions . For tandemly - connected FPs in Supplementary Fig . 2a , N - ( residues 1 \u2013 4 for ShadowY , and 1 \u2013 5 for mCherry and mScarlet - I ) and C - terminal disordered regions ( residues 228 \u2013 238 for EGFP and mTur - quoise2 , and 226 \u2013 236 for mNeonGreen ) of FPs were deleted . And two FPs were connected with a short linker ( GGSGGS ) . For EGFP - ShadowY - CLC in Fig . 1 , tandemly - connected EGFP - ShadowY was fused to CLCa with a 10 a . a . linker . For CLCb probes used inFRET experiments , either C - terminaldisorderedregion ( residues228 \u2013 238 ) deletedorN - terminal disordered region ( residues 1 \u2013 4 ) deleted EGFP ( or ShadowY ) was connected to CLCb with a short linker ( GGSGGS ) . All plasmids were con \ufb01 rmed by sequencing ( Psomagen ) and identi \ufb01 ed as in Supple - mentary Data 1 . Fixing and unroo \ufb01 ng Cells were rinsed in intracellular buffer ( 70mM KCl , 30 mM HEPES maintained at pH 7 . 4 with KOH , 5 mM MgCl 2 , 3mM EGTA ) , and manually unroofed with 19 - gauge needle and syringe using 2 % paraf - ormaldehyde ( ElectronMicroscopySciences , 15710 ) intheintracellular buffer . After unroo \ufb01 ng , the coverslips were transferred to fresh 2 % paraformaldehyde in the intracellular buffer for 20min . They then washed with phosphate - buffered saline ( PBS ) . Imaging was performed in PBS ( pH 7 . 4 ) ( Quality Biological , 114 - 058 - 131 ) . For FKBP / FRB dimerization experiments , cells were unroofed after 15min incubation with 500 nM AP21967 ( 500 \u03bc M stock in etha - nol ) ( Takara , 635055 ) or 0 . 1 % w / v ethanol ( control ) in DMEM growth medium . Imaging of unroofed plasma membranes for CLEM Unroofed cells were stained either with 16 . 5 pmol of Alexa Fluor 350 - phalloidin ( LifeTechnologies , A22281 ) , AlexaFluor568 - phalloidin ( Life Technologies , A12380 ) , or Alexa Fluor 647 - phalloidin ( Life Technolo - gies , A22287 ) for 15min depending on spectra of expressing FPs . Then cells were rinsed with PBS . 1 mm\u00d7 1mm large montage was generated for proteins ofinterest and phalloidinusing a NikonEclipse Ti inverted microscope with a 100\u00d7 , 1 . 49 NA objective ( Nikon , SR HP Apo TIRF ) and an Andor iXon Ultra 897 EM - CCD camera under the control of Nikon Elements software . Images were obtained by TIRF illumination except for Alexa Fluor 350 which was imaged by epi illumination . This map was used to \ufb01 nd the cells expressing the target proteins for FLIM and CLEM analysis . The imaged area wasmarked with a circle ( 4 mm in diameter ) around the center of the imaged area using an objective diamond scriber ( Leica , 11505059 ) 40 , 67 . The immersion oil wascarefully removed from the bottom of the glass coverslip . The sample was subsequentlyimaged by FLIM orstored in 2 % glutaraldehyde ( Electron Microscopy Sciences , 16019 ) at 4 \u00b0C until EM sample preparation . FLIM Time - domain \ufb02 uorescence lifetime imaging was performed with a Leica Falcon SP8 confocal microscope with a 63\u00d7 , 1 . 40 NA oil immer - sion objective ( Leica , HC PL APO CS2 ) under the control of Leica LAS X software . As \ufb01 rst , a large image montage around the region of interest on the coverslip marked by a diamond scriber was acquired to \ufb01 nd the cells which were identi \ufb01 ed in the TIRF montage . Then , FLIM images were collected at a lateral spatial resolution of 80nm per pixel and withascanspeedat3 . 16 \u03bc sperpixel . Theconfocalaperturewassetata diameter of 168 \u03bc m ( 2 AU @ 510nm ) . EGFP was excited by 488nm and 498 \u2013 560nm \ufb02 uorescence was collected . A notch \ufb01 lter for 488 nm was used . To reduce the impact of pile - up effect , an excitation power was set tokeeppeak count perpulse aslessthan0 . 1 . For CLEM , theimaged sample was then stored in 2 % glutaraldehyde at 4 \u00b0C until EM sample preparation . To select a FP pair for FLIM - FRET , we made tandemly con - nected FPs and compared their FRET ef \ufb01 ciency ( Supplementary Fig . 2a ) . Based on its high FRET ef \ufb01 ciency and compatibility for our optical systems , we selected monomeric EGFP and ShadowY ( Sup - plementary Fig . 2b ) . To check the relationship between the number of frames and photobleaching or FRET ef \ufb01 ciency change , \ufb02 uores - cence intensity and FRET ef \ufb01 ciency of EGFP - ShadowY - CLC was compared with various frame numbers ( Supplementary Fig . 3e , f ) . Fluorescence intensity did not change and FRET ef \ufb01 ciency slightly decreased as the frame number increases . This might be due to photobleaching of ShadowY occurs faster than EGFP . Although the magnitude of ShadowY photobleaching should be smaller for lower FRET ef \ufb01 ciency situation like intermolecular FRET , we used 150 frames to minimize the effect of photobleaching . With 150 frames , ~ 10 , 000 photon counts per single CCS were obtained in average ( Supplementary Fig . 3g ) . With this range of photon counts , FRET ef \ufb01 ciency ( E ) of EGFP - ShadowY - CLC from single CCS could be Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 10 estimated by \ufb01 tting a \ufb02 uorescence decay curve with a bi - exponential function convolved with the Gaussian pulse response function 68 ( Supplementary Fig . 3h ) . F \u00f0 t \u00de = F 0 \u00bd P 1 H t , t 0 , \u03c4 1 , \u03c4 G (cid:1) (cid:3) + \u00f0 1 (cid:2) P 1 \u00de H t , t 0 , \u03c4 2 , \u03c4 G (cid:1) (cid:3) (cid:3) \u00f0 1 \u00de H t , t 0 , \u03c4 1 , \u03c4 G (cid:1) (cid:3) = 1 2exp \u03c4 G 2 2 \u03c4 1 (cid:2) t (cid:2) t 0 \u03c4 1 (cid:4) (cid:5) erfc \u03c4 G 2 (cid:2) \u03c4 1 t (cid:2) t 0 (cid:1) (cid:3) \ufb03\ufb03\ufb03 2 p \u03c4 1 \u03c4 G ! \u00f0 2 \u00de where F ( t ) is the \ufb02 uorescence lifetime decay curve , F 0 is the peak \ufb02 uorescence before convolution , P 1 is the fraction of the \ufb01 rst com - ponent , \u03c4 1 and \u03c4 2 are the \ufb02 uorescence lifetime of \ufb01 rst and second components , \u03c4 G is the width of the Gaussian pulse response function , t 0 is time offset , and erfc is the complementary error function . \u03c4 a = P 1 \u03c4 1 + 1 (cid:2) P 1 (cid:1) (cid:3) \u03c4 2 \u00f0 3 \u00de E = 1 (cid:2) \u03c4 a , DA \u03c4 a , D \u00f0 4 \u00de where \u03c4 a is amplitude - weighted \ufb02 uorescence lifetime , \u03c4 a , D is amplitude - weighted \ufb02 uorescence lifetime of the donor without acceptor , and \u03c4 a , DA is amplitude - weighted \ufb02 uorescence lifetime of the donor in presence of the acceptor 44 . Values from single CCSs ( SupplementaryFig . 3h ) wereconsistentwiththosefromwholeplasma membrane ( Supplementary Fig . 3a \u2013 c ) . However , in general , nearly 10 , 000 photon counts are required to \ufb01 t a bi - exponential 45 , 69 . So , curve - \ufb01 tting on the \ufb02 uorescence decay curves from very small CCSs willnotbeadequate . Thus , insteadtoestimateFRETef \ufb01 ciencythrough curve - \ufb01 tting , mean \ufb02 uorescence lifetime , the center mass of the \ufb02 uorescence lifetime decay , was used as an indicator for FRET ef \ufb01 ciency ( Fig . 1d and Supplementary Fig . 3d ) . Background was not subtracted . In our measurements , background counts were generally verylowandsignal - to - backgroundratiowas > 100inmostexperiments ( Supplementary Fig . 8b ) . Mean \ufb02 uorescence lifetime was not depen - dent on the signal - to - background ratio . For FLIM measurements with dipicrylamine ( DPA ; City Chemical LLC ) , a 20 mM stock solution of DPA in DMSO ( Sigma - Aldrich , D2650 ) was prepared fresh from powder every day and diluted to a \ufb01 nal concentration with PBS . Imaging was performed at least 10 min after addition of DPA . FRET ef \ufb01 ciency between DPA and miRFP should be very low because DPA absorbancespectra and miRFP emission spectra are suf \ufb01 ciently separated ( Supplementary Fig . 2c ) . For live cell FLIM imaging ( Supplementary Fig . 11 ) , cells were imaged in imaging buffer ( 130mM NaCl , 2 . 8 mM KCl , 5mM CaCl 2 , 1 mM MgCl 2 , 10mM HEPES , and 10 mM glucose at pH 7 . 4 ) at 21 \u00b0C . Platinum replica EM EM samples were prepared asdescribed previously 4 , 67 . Coverslips were transferred from glutaraldehyde into 0 . 1 % w / v tannic acid for 20 min . Then , they were rinsed 4 times with water , and placed in 0 . 1 % w / v uranyl acetate for 20min . The coverslips were then dehydrated , cri - tical point dried with a critical point dryer ( Tousimis Samdri , 795 ) , and coated with platinum and carbon with a freeze fracture device ( Leica , EM ACE 900 ) . The region of interest on the coverslip marked by a diamond scriber was imaged with a 20\u00d7 phase - contrast objective to obtain another map of the region imaged in \ufb02 uorescence . The replicas were lifted and placed onto formvar / carbon - coated 75 - mesh copper TEM grids ( Ted Pella , 01802 - F ) that were freshly glow - discharged with PELCO easiGlow 91000 . Again , the grid was imaged with a 20\u00d7 phase - contrastobjectiveto \ufb01 ndthesameregionthatwasoriginallyimagedin \ufb02 uorescence . Each cell of interest was located on the grid prior to EM imaging 40 . TEM imaging was performed as previously described 37 at \u00d715 , 000 magni \ufb01 cation ( 1 . 2nm per pixel ) using a JEOL 1400 and SerialEM freeware for montaging 70 . Electron microscopy montages were processed using IMOD freeware 71 . FRET - CLEM image analysis The FLIM images were aligned to the EM images using an af \ufb01 ne spatial transformation with nearest - neighbor interpolation to map the CCSs visibleinbothFLIMimagesandEMimages 36 , 41 . Sincepositionofsphere clathrin is sometimes changed during critical point drying due to a weak attachment tothemembrane , we used \ufb02 atand domed clathrinas \ufb01 ducials 67 . Rectangular ROI was created around single isolated CCS , and mean \ufb02 uorescence lifetime within each ROI was calculated . These values were analyzed by categorizing according to lattice structures , and were compared to the average value of \ufb02 at clathrin for each unroofed membrane . This is because FRET ef \ufb01 ciencies , and thus \ufb02 uorescence lifetimes , varied among cells due to the differences in expression levels of donor and acceptor probes , or incorporated density of DPA ( Supplementary Figs . 5 , 9 , 10 , 13 , and 14 ) . For FRET between EGFP and ShadowY , because ShadowY expression cannot be con \ufb01 rmed with \ufb02 uorescence , only cells with a cellular average \ufb02 uorescence lifetime of less than 2 . 1 ns were analyzed to ensure the expression of ShadowY . EM image analysis Binary masks of the \ufb02 at ( no visible curvature ) , domed ( curved but can still see the edge of the lattice ) , and sphere clathrin ( curved beyond a hemisphere such that the edge of the lattice is no longer visible ) were manually segmented ( Supplementary Fig . 19a ) 43 , 72 . The percentage of occupiedmembraneareawasde \ufb01 nedasthesumofareasfromclathrin lattices of the speci \ufb01 ed subtype divided by the total area of measured membrane . The expression of CLC probes did not change the size and density of clathrin lattice structures ( Supplementary Fig . 19b \u2013 d ) . Live cell TIRF imaging and analysis Images were acquired using a Nikon Eclipse Ti inverted microscope with a 100\u00d7 , 1 . 49 NA objective and an Andor iXon Ultra 897 EM - CCD camera . HeLa cells in the phenol red free DMEM growth medium [ DMEM ( Gibco , 31053036 ) with 10 % fetal bovine serum , 1 : 100 dilution of 100\u00d7 GlutaMAX ( Gibco , 35050061 ) , 1 : 100 dilution of 100 mM sodium pyruvate ( Gibco , 11360070 ) ] were mounted in a chamber at 37\u00b0C with a water bath and continuous \ufb02 ow of humidi \ufb01 ed 5 % CO 2 to maintain the osmolality and pH of the medium . The pixel size was 110nm . TIRF images were acquired with 100ms exposures at 0 . 5 Hz for 10 min . For AP21967 treatment , images were acquired at least 10min after the addition of 500nM AP21967 . FKBP - EGFP - CLC spots were tracked using the ImageJ plugin TrackMate 73 . Spot detection wasdone with the differenceof Gaussians approach , and tracking was done using the simple linear assignment problem ( LAP ) tracker algorithm with a linking maximum distance of 330 nm per frame . Tracks that appeared and disappeared during the whole movie with over 20s duration , and a total net displacement was less than 330 nm were selected for residence time analysis . Further , each track was visually inspected for isolation and tracking errors . Transferrin uptake assay HeLa cells were incubated in starvation medium [ DMEM containing 20mM HEPES at pH 7 . 4 and 0 . 1 % w / v bovine serum albumin ( Fisher Bioreagents , BP9703 ) ] for 1h in an CO 2 incubator . They were then incubated with 500nM AP21967 or 0 . 1 % w / v ethanol ( control ) in starvation medium for 5min . Then , 25mg / mL Alexa Fluor 647 con - jugated human transferrin ( Invitrogen , T23366 ) was added to the starvation medium 74 . After 15 min incubation , the cells were \ufb01 xed with 2 % paraformaldehyde at room temperature for 25min . They then washed with PBS . The cells were imaged with a Leica Falcon SP8 con - focal microscope with a 63\u00d7 , 1 . 40 NA oil immersion objective . Images were collected at a lateral spatial resolution of 120nm per pixel . The Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 11 confocal aperture was set at a diameter of 191 \u03bc m , and optical sections with z - spacing of 0 . 8 \u03bc m were collected . Alexa Fluor 647 was excited by 633nm and 638 \u2013 800nm \ufb02 uorescence was collected . The stack images were recombined using a sum - intensity opera - tion . And background intensity fromnon - cellregions were subtracted . Then , cell outlines were traced manually and the mean intensity ( total intensity / area ) was measured for individual cells . Fluorescence inten - sity was normalized by the average value of non - transfected cells for each coverslip . Immunocytochemistry HeLacellswere \ufb01 xedin2 % paraformaldehydeinPBSfor25min , blocked with 3 % w / v bovine serum albumin for 60min , and reacted with mouse monoclonal antibody against transferrin receptor ( Santa Cruz , sc - 65877 , 1 : 50 ) . The primary antibody was visualized by secondary anti - body staining using goat anti - mouse IgG conjugated to Alexa Fluor 647 ( Invitrogen , A21237 , 1 : 500 ) . The cells were imaged with a Leica Falcon SP8 confocal microscope with a \u00d763 , 1 . 40 NA oil immersion objective . Images were collected at a lateral spatial resolution of 120nm per pixel . The confocal aperture was set at a diameter of 191 \u03bc m , and optical sections with z - spacing of 0 . 8 \u03bc m were collected . Alexa Fluor 647 was excitedby633nmand638 \u2013 800nm \ufb02 uorescencewascollected . Images were analyzed as transferrin uptake assay . Linker \ufb02 exibility measurements with polarization - TIRF ( pol - TIRF ) microscopy HeLa cells were imaged with 3i vector TIRF microscopy system , based on an inverted microscope ( IX - 81 ; Olympus ) and equipped with 100\u00d7 , 1 . 49 NA objective ( Olympus ) . Fluorescencewas excited by 488 nm and 561 nm lasers passed through LF405 / 488 / 561 / 635 \ufb01 lters ( Semrock ) . Emitted light was then divided by the image splitter \u2019 s dichroic ( 565DCXR ) and projected side - by - side through 525Q / 50 and 605Q / 55 emission \ufb01 lters onto the chip of an EMCCD camera ( Andor , DU 897 ) . Images were acquired using SlideBook6 ( 3i ) . A sequence of 40 images was taken with pol - TIRF \ufb01 elds that were parallel ( s - pol ) or perpendi - cular ( p - pol ) to the coverslip . Each individual image had an exposure time of 100 ms . For calibration , \ufb02 uorescein solution ( Alfa Aesar , L13251 ) wasmeasured . 20 s - poland20p - polimageswere summed and background was subtracted . A p : s ratio was then calculated and nor - malized by the value of \ufb02 uorescein measurements . Past work measured the orientation of nucleoporins ( Nups ) in the nuclear pore complex using pol - TIRF microscopy and orientational sensors 75 . In these sensors , a short N - terminal \u03b1 - helix of EGFP is con - jugated directly to a C - terminal \u03b1 - helix in a protein of interest by a continuous \u03b1 - helix . When the number of amino acids in this linker helix is varied , EGFP rotates around the linker helix axis by angles dictated by \u03b1 - helical geometry . Since the excitation dipole is \ufb01 xed within EGFP , the p : s ratio shifts by changing the linker length . On the other hand , if EGFP is conjugated by a \ufb02 exible linker , the p : s ratio does not shift by changing the linker length . First , we tested if we could reproduce the previous results in our imaging system . We imaged the bottom of the nucleus of \ufb01 xed HeLa cells expressing Nup54 - EGFP494 constructs 75 ( Supplementary Fig . 6a ) . Consistent with previous data , the p : s ratio shifted by changing the linker length for the rigid linker probes but did not shift for the \ufb02 exible linker probes ( Supplementary Fig . 6b ) . Next , we applied this strategy to investigate the linker \ufb02 ex - ibility in CLC probes . We conjugated EGFP to C - terminal \u03b1 - helix domain of CLC ( at residue 206 ) by rigid or \ufb02 exible linkers ( Supple - mentary Fig . 6c ) . We then imaged unroofed HeLa cells expressing CLC orientational sensors , CLC - EGFP , or EGFP - CLC ( Supplementary Fig . 6d ) . The p : s ratio shifted with different lengths of the rigid linkers however the p : s ratio was not changed by varying lengths of the \ufb02 ex - ible linkers . The p : s ratio of CLC - EGFP and EGFP - CLC which were used inFRETexperimentsweresimilartothoseofthe \ufb02 exiblelinkersensors . Theseresultssupportthe assumptionthat thelinkersofCLC - EGFP and EGFP - CLC are \ufb02 exible . Further , since there were no signi \ufb01 cant corre - lation between EGFP intensity and the p : s ratio in CLC - EGFP and EGFP - CLC ( Supplementary Fig . 6e \u2013 g ) , linkers are likely \ufb02 exible in clathrin lattices with various shapes . Western blot HeLa cells were dissociated with trypsin and collected . After washing and centrifuge , 100 \u00b5 L RIPA lysis buffer ( Millipore 20 \u2013 188 ) supple - mented with protease inhibitor cocktail ( Halt , 78429 ) was added and samples were incubated for 1 h on ice . After vortex , cell lysates were centrifuged at 21300 \u00d7 g for 15 min at 4 \u00b0C . The supernatant was used for SDS - PAGE on 4 \u2013 12 % Bis - Tris gels ( Invitrogen , NP0335 ) with Tris - buffered saline with 0 . 1 % Tween - 20 ( TBS - T ) at 150 V . Blotting was performed by iBlot ( invitrogen ) with nitrocellulose membranes ( Invitrogen , IB301002 ) accordingly to the manufacturer \u2019 s protocol . Afterwards , membranes were incubated for 2 h in 5 % milk in TBS - T . Primary antibodies [ anti - CLCa / b ( Millipore , AB9884 , 1 : 5000 ) , anti - CLCa ( Sigma , HPA050918 , 1 : 1000 ) , or anti - CLCb ( Abnova , H00001212 - M01 , 1 : 500 ) ] were diluted in 5 % milk / TBS - T and applied on the membranes overnight . After 5 times 5 min washing with TBS - T , secondary antibodies [ anti - mouse IgG - HRP ( Jackson ImmunoR - esearch Labs , 115 - 035 - 174 , 1 : 2000 ) or anti - rabbit IgG - HRP ( Jackson ImmunoResearch Labs , 211 - 032 - 171 , 1 : 2000 ) ] were diluted in 5 % milk / TBS - T and applied on the membranes for 1 h . After washing with TBS - T , the membrane was imaged with ChemiDoc system ( Bio - Rad ) with ECL solution ( cytiva , RPN2232 ) . Then after washing , the membrane was stained with anti - \u03b2 actin - HRP ( Cell Signaling , 5125 S , 1 : 2000 ) and imaged . Estimation of expression levels of transfected constructs To estimate the expression levels of transfected constructs , we performed western blot ( Supplementary Fig . 7 ) . First , we stained with anti - CLCa / b antibody ( Supplementary Fig . 7a ) . The amount of endogenous CLCa / b decreased by transfection ( Supplementary Fig . 7d ) . The degree of decrease was lager in dual - transfection than single - transfection . The reactivity of this antibody against N - terminal - FP - attached CLC was weaker than that against C - terminal - FP - attached CLC . Thus we did not measure the ratio between transfected and endogenous CLC with this antibody . Next , we stained with either anti - CLCa ( Supplementary Fig . 7b ) or anti - CLCb ( Supplementary Fig . 7c ) speci \ufb01 c antibody . Consistent with anti - CLCa / b antibody , the amount of endogenous CLCa and CLCb decreased with transfection ( Supplementary Fig . 7e , f ) . And the amount of transfected CLCb was larger than that of endogenous CLCb ( Supplementary Fig . 7g ) . From these results , we modeled the expression of transfected constructs . CLCa is the dominant isoform in HeLa cells 18 . Thus , we assumed a ratio between CLCa and CLCb of 4 : 1 and a transfection ef \ufb01 ciency of 80 % . Under these conditions , we estimate the CLC amount in transfected cells ( Supplementary Fig . 7h ) . In this model , the total amount of CLCs increases 10 times after transfection and the ratio between transfected and endogen - ous CLCs is 25 : 1 for single transfection and 100 : 1 for double trans - fection . Thus , in our FRET experiments , we propose that most CLCs are attached to an FP . FRET simulation FRET simulations ( Supplementary Figs . 4c and 18 ) were performed using MATLAB . We used 60 \u00c5 for F\u00f6rster radius for EGFP - ShadowY FRET . For the simulation in Supplementary Fig . 4c , we determined the position of N - and C - terminus of CLC according to the structure models ( PDB 3LVG and 6WCJ ) 17 , 46 . For the simulation in Supple - mentary Fig . 18 , the lateral position of N - terminus was moved by 0 . 6 \u00c5 spacing , and N - terminus was assumed to locate axially 25 \u00c5 higher than C - terminus . When we focus on single clathrin heavy chain , there are four different states ; without CLC , endogenous Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 12 CLCa or CLCb binding , EGFP - CLC binding , or CLC - ShadowY binding ( Supplementary Fig . 4b ) . By considering these four states , we cal - culated the FRET ef \ufb01 ciency between EGFP - CLC and CLC - ShadowY or ShadowY - CLC binding to surrounding \ufb01 ve clathrin heavy chains at various heavy chain occupancy with ShadowY - attached CLC . Quanti \ufb01 cation and statistical analysis Image analysis and quanti \ufb01 cation were performed with ImageJ 76 and MATLAB . The statistical tests used for each experiment and the exact sample numbers ( n values ) are indicated in the corresponding \ufb01 gure legends . p values of < 0 . 05 were considered statistically signi \ufb01 cant . All statistical analysis and \ufb01 tting were performed using Origin 2016 ( Origin Lab ) . The exact p values are provided in Source Data \ufb01 le . Materials availability The plasmids used in the study are deposited at Addgene ( Supple - mentary Data 1 ) . Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article . Data availability The data generated in this study has been deposited in Figshare at https : / / doi . org / 10 . 25444 / nhlbi . c . 6259170 . The remaining data are availableintheArticleorSupplementary Information \ufb01 les . Sourcedata are provided with this paper . Atomic coordinates of previously determined X - rayor Cryo - EM structures are availableinthe PDB under the following accession codes : 3LVG ( clathrin heavy chains and light chains ) , 6WCJ ( clathrin - coated vesicle ) , and 4KW4 ( GFP ) . Source data are provided with this paper . Code availability MATLABcodesusedinthisstudyarespeci \ufb01 ctolab \ufb01 leformatting . The codes are available in Figshare at https : / / doi . org / 10 . 25444 / nhlbi . 14502156 . References 1 . McMahon , H . T . & Boucrot , E . Molecular mechanism and physiolo - gicalfunctionsofclathrin - mediatedendocytosis . Nat . Rev . Mol . Cell Biol . 12 , 517 \u2013 533 ( 2011 ) . 2 . Kaksonen , M . & Roux , A . Mechanisms of clathrin - mediated endo - cytosis . Nat . Rev . Mol . Cell Biol . 19 , 313 \u2013 326 ( 2018 ) . 3 . Mettlen , M . , Chen , P . H . , Srinivasan , S . , Danuser , G . & Schmid , S . L . Regulation of clathrin - mediated endocytosis . Annu Rev . Biochem 87 , 871 \u2013 896 ( 2018 ) . 4 . Sochacki , K . A . , Dickey , A . M . , Strub , M . P . & Taraska , J . W . Endocytic proteins are partitioned at the edge of the clathrin lattice in mam - malian cells . Nat . Cell Biol . 19 , 352 \u2013 361 ( 2017 ) . 5 . Mund , M . et al . Systematic nanoscale analysis of endocytosis links ef \ufb01 cient vesicle formation to patterned actin nucleation . Cell 174 , 884 \u2013 896 . e817 ( 2018 ) . 6 . Cocucci , E . , Aguet , F . , Boulant , S . & Kirchhausen , T . The \ufb01 rst \ufb01 ve secondsinthelifeofaclathrin - coatedpit . Cell 150 , 495 \u2013 507 ( 2012 ) . 7 . Akamatsu , M . et al . Principles of self - organization and load adap - tation by the actin cytoskeleton during clathrin - mediated endocy - tosis . Elife 9 , e49840 ( 2020 ) . 8 . Taylor , M . J . , Perrais , D . & Merri \ufb01 eld , C . J . A high precision survey of the molecular dynamics of mammalian clathrin - mediated endocy - tosis . PLoS Biol . 9 , e1000604 ( 2011 ) . 9 . Kukulski , W . , Schorb , M . , Kaksonen , M . & Briggs , J . A . Plasma membrane reshaping during endocytosis is revealed by time - resolved electron tomography . Cell 150 , 508 \u2013 520 ( 2012 ) . 10 . Wrobel , A . G . et al . Temporal ordering in endocytic clathrin - coated vesicle formation via AP2 phosphorylation . Dev . Cell 50 , 494 \u2013 508 e411 ( 2019 ) . 11 . Clarke , N . I . & Royle , S . J . FerriTag is a new genetically - encoded inducible tag for correlative light - electron microscopy . Nat . Com - mun . 9 , 2604 ( 2018 ) . 12 . Sochacki , K . A . & Taraska , J . W . From \ufb02 at to curved clathrin : con - trolling a plastic ratchet . Trends Cell Biol . 29 , 241 \u2013 256 ( 2019 ) . 13 . Wood , K . M . & Smith , C . J . Clathrin : the molecular shape shifter . Biochem J . 478 , 3099 \u2013 3123 ( 2021 ) . 14 . Wu , M . & Wu , X . A kinetic view of clathrin assembly and endocytic cargo sorting . Curr . Opin . Cell Biol . 71 , 130 \u2013 138 ( 2021 ) . 15 . Fotin , A . et al . Molecular model for a complete clathrin lattice from electron cryomicroscopy . Nature 432 , 573 \u2013 579 ( 2004 ) . 16 . Brodsky , F . M . Diversity of clathrin function : new tricks for an old protein . Annu Rev . Cell Dev . Biol . 28 , 309 \u2013 336 ( 2012 ) . 17 . Wilbur , J . D . et al . Conformation switching of clathrin light chain regulates clathrin lattice assembly . Dev . Cell 18 , 841 \u2013 848 ( 2010 ) . 18 . Girard , M . , Allaire , P . D . , McPherson , P . S . & Blondeau , F . Non - stoichiometric relationship between clathrin heavy and light chains revealed by quantitative comparative proteomics of clathrin - coated vesicles from brain and liver . Mol . Cell Proteom . 4 , 1145 \u2013 1154 ( 2005 ) . 19 . Boulant , S . , Kural , C . , Zeeh , J . C . , Ubelmann , F . & Kirchhausen , T . Actin dynamics counteract membrane tension during clathrin - mediated endocytosis . Nat . Cell Biol . 13 , 1124 \u2013 1131 ( 2011 ) . 20 . Dannhauser , P . N . et al . Effect of clathrin light chains on the stiffness of clathrin lattices and membrane budding . Traf \ufb01 c 16 , 519 \u2013 533 ( 2015 ) . 21 . Maib , H . , Ferreira , F . , Vassilopoulos , S . & Smythe , E . Cargoregulates clathrin - coated pit invagination via clathrin light chain phosphor - ylation . J . Cell Biol . 217 , 4253 \u2013 4266 ( 2018 ) . 22 . Biancospino , M . et al . Clathrin light chain A drives selective myosin VI recruitment to clathrin - coated pits under membrane tension . Nat . Commun . 10 , 4974 ( 2019 ) . 23 . Redlingshofer , L . et al . Clathrin light chain diversity regulates membrane deformation in vitro and synaptic vesicle formation in vivo . Proc . Natl Acad . Sci . USA 117 , 23527 \u2013 23538 ( 2020 ) . 24 . Das , J . , Tiwari , M . & Subramanyam , D . Clathrinlightchains : nottobe taken so lightly . Front . Cell Dev . Biol . 9 , 774587 ( 2021 ) . 25 . Ungewickell , E . & Ungewickell , H . Bovine brain clathrin light chains impede heavy chain assembly in vitro . J . Biol . Chem . 266 , 12710 \u2013 12714 ( 1991 ) . 26 . Liu , S . - H . , Wong , M . L . , Craik , C . S . & Brodsky , F . M . Regulation of clathrin assembly and trimerization de \ufb01 ned using recombinant triskelion hubs . Cell 83 , 257 \u2013 267 ( 1995 ) . 27 . Ybe , J . A . et al . Clathrin self - assembly is regulated by three light - chain residues controlling the formation of critical salt bridges . EMBO J . 17 , 1297 \u2013 1303 ( 1998 ) . 28 . Miyawaki , A . Development of probes for cellular functions using \ufb02 uorescent proteins and \ufb02 uorescence resonance energy transfer . Annu Rev . Biochem 80 , 357 \u2013 373 ( 2011 ) . Occupancy = \u00bd ShadowY probe binding (cid:3) No CLC binding (cid:7) (cid:8) + Endogenous CLCa = b biding (cid:7) (cid:8) + EGFP (cid:2) CLC binding (cid:7) (cid:8) + \u00bd ShadowY probe binding (cid:3) \u00f0 5 \u00de Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 13 29 . Taraska , J . W . Mapping membrane protein structure with \ufb02 uores - cence . Curr . Opin . Struct . Biol . 22 , 507 \u2013 513 ( 2012 ) . 30 . Szalai , A . M . , Zaza , C . & Stefani , F . D . Super - resolution FRET mea - surements . Nanoscale 13 , 18421 \u2013 18433 ( 2021 ) . 31 . Skruzny , M . , Pohl , E . , Gnoth , S . , Malengo , G . & Sourjik , V . Theprotein architecture of the endocytic coat analyzed by FRET microscopy . Mol . Syst . Biol . 16 , e9009 ( 2020 ) . 32 . Taraska , J . W . A primer on resolving the nanoscale structure of the plasma membrane with light and electron microscopy . J . Gen . Physiol . 151 , 974 \u2013 985 ( 2019 ) . 33 . Sochacki , K . A . & Taraska , J . W . Findyourcoat : usingcorrelativelight and electron microscopy to study intracellular protein coats . Curr . Opin . Cell Biol . 71 , 21 \u2013 28 ( 2021 ) . 34 . Heuser , J . The production of \u2018 cell cortices \u2019 for light and electron microscopy . Traf \ufb01 c 1 , 545 \u2013 552 ( 2000 ) . 35 . Collins , A . , Warrington , A . , Taylor , K . A . & Svitkina , T . Structural organization of the actin cytoskeleton at sites of clathrin - mediated endocytosis . Curr . Biol . 21 , 1167 \u2013 1175 ( 2011 ) . 36 . Bucher , D . et al . Clathrin - adaptor ratio and membrane tension regulate the \ufb02 at - to - curved transition of the clathrin coat during endocytosis . Nat . Commun . 9 , 1109 ( 2018 ) . 37 . Sochacki , K . A . , Shtengel , G . , van Engelenburg , S . B . , Hess , H . F . & Taraska , J . W . Correlative super - resolution \ufb02 uorescence and metal - replica transmission electron microscopy . Nat . Methods 11 , 305 \u2013 308 ( 2014 ) . 38 . Vassilopoulos , S . , Gibaud , S . , Jimenez , A . , Caillol , G . & Leterrier , C . Ultrastructure of the axonal periodic scaffold reveals a braid - like organization of actin rings . Nat . Commun . 10 , 5803 ( 2019 ) . 39 . Roberts , A . D . et al . Structurally distinct endocytic pathways for B cell receptors in B lymphocytes . Mol . Biol . Cell 31 , 2826 \u2013 2840 ( 2020 ) . 40 . Prasai , B . et al . The nanoscale molecular morphology of docked exocytic dense - core vesicles in neuroendocrine cells . Nat . Com - mun . 12 , 3970 ( 2021 ) . 41 . Scott , B . L . et al . Membrane bending occurs at all stages of clathrin - coat assembly and de \ufb01 nes endocytic dynamics . Nat . Commun . 9 , 419 ( 2018 ) . 42 . Murakoshi , H . & Shibata , A . C . E . ShadowY : a dark yellow \ufb02 uorescent protein for FLIM - based FRET measurement . Sci . Rep . 7 , 6791 ( 2017 ) . 43 . Sochacki , K . A . et al . The structure and spontaneous curvature of clathrin lattices at the plasma membrane . Dev . Cell 56 , 1131 \u2013 1146 e1133 ( 2021 ) . 44 . Lakowicz , J . R . Principles of Fluorescence Spectroscopy . ( Springer , 2006 ) . 45 . Barber , P . R . et al . Multiphoton time - domain \ufb02 uorescence lifetime imaging microscopy : practical application to protein \u2013 protein interactions using global analysis . J . R . Soc . Interface 6 , S93 \u2013 S105 ( 2008 ) . 46 . Paraan , M . et al . The structures of natively assembled clathrin - coated vesicles . Sci . Adv . 6 , eaba8397 ( 2020 ) . 47 . Chanda , B . , Asamoah , O . K . , Blunck , R . , Roux , B . & Bezanilla , F . Gatingchargedisplacementinvoltage - gatedionchannelsinvolves limited transmembrane movement . Nature 436 , 852 \u2013 856 ( 2005 ) . 48 . Taraska , J . W . & Zagotta , W . N . Structuraldynamicsinthegatingring of cyclic nucleotide - gated ion channels . Nat . Struct . Mol . Biol . 14 , 854 \u2013 860 ( 2007 ) . 49 . De - la - Rosa , V . , Rangel - Yescas , G . E . , Ladron - de - Guevara , E . , Rosenbaum , T . & Islas , L . D . Coarse architecture of the transient receptor potential vanilloid 1 ( TRPV1 ) ion channel determined by \ufb02 uorescence resonance energy transfer . J . Biol . Chem . 288 , 29506 \u2013 29517 ( 2013 ) . 50 . Putyrski , M . & Schultz , C . Protein translocation as a tool : the current rapamycin story . FEBS Lett . 586 , 2097 \u2013 2105 ( 2012 ) . 51 . Bayle , J . H . et al . Rapamycin analogs with differential binding spe - ci \ufb01 city permit orthogonal control of protein activity . Chem . Biol . 13 , 99 \u2013 107 ( 2006 ) . 52 . Czech , M . P . PIP2 and PIP3 . Cell 100 , 603 \u2013 606 ( 2000 ) . 53 . Moskowitz , H . S . , Heuser , J . , McGraw , T . E . & Ryan , T . A . Targeted chemicaldisruptionofclathrinfunctioninlivingcells . Mol . Biol . Cell 14 , 4437 \u2013 4447 ( 2003 ) . 54 . Chen , C . Y . & Brodsky , F . M . Huntingtin - interacting protein 1 ( Hip1 ) and Hip1 - related protein ( Hip1R ) bind the conserved sequence of clathrin light chains and thereby in \ufb02 uence clathrin assembly in vitro and actin distribution in vivo . J . Biol . Chem . 280 , 6109 \u2013 6117 ( 2005 ) . 55 . Legendre - Guillemin , V . et al . Huntingtin interacting protein 1 ( HIP1 ) regulates clathrin assembly through direct binding to the reg - ulatory region of the clathrin light chain . J . Biol . Chem . 280 , 6101 \u2013 6108 ( 2005 ) . 56 . Huang , F . , Khvorova , A . , Marshall , W . & Sorkin , A . Analysis of clathrin - mediated endocytosis ofepidermal growthfactorreceptor by RNA interference . J . Biol . Chem . 279 , 16657 \u2013 16661 ( 2004 ) . 57 . Ferreira , F . et al . Endocytosis of G protein - coupled receptors is regulated by clathrin light chain phosphorylation . Curr . Biol . 22 , 1361 \u2013 1370 ( 2012 ) . 58 . Wu , S . et al . Clathrin light chains \u2019 role in selective endocytosis in \ufb02 uences antibody isotype switching . Proc . Natl Acad . Sci . USA 113 , 9816 \u2013 9821 ( 2016 ) . 59 . Chen , P . H . et al . Crosstalk between CLCb / Dyn1 - mediated adaptive clathrin - mediated endocytosis and epidermal growth factor receptor signaling increases metastasis . Dev . Cell 40 , 278 \u2013 288 e275 ( 2017 ) . 60 . Piston , D . W . & Kremers , G . J . Fluorescent protein FRET : the good , the bad and the ugly . Trends Biochem Sci . 32 , 407 \u2013 414 ( 2007 ) . 61 . Marks , K . M . & Nolan , G . P . Chemical labeling strategies for cell biology . Nat . Methods 3 , 591 \u2013 596 ( 2006 ) . 62 . Nikic , I . , Kang , J . H . , Girona , G . E . , Aramburu , I . V . & Lemke , E . A . Labeling proteins on live mammalian cells using click chemistry . Nat . Protoc . 10 , 780 \u2013 791 ( 2015 ) . 63 . Royle , S . J . The cellular functions of clathrin . Cell Mol . Life Sci . 63 , 1823 \u2013 1832 ( 2006 ) . 64 . Suhling , K . et al . Imaging the environment of green \ufb02 uorescent protein . Biophys . J . 83 , 3589 \u2013 3595 ( 2002 ) . 65 . Datta , R . , Heaster , T . M . , Sharick , J . T . , Gillette , A . A . & Skala , M . C . Fluorescence lifetime imaging microscopy : fundamentals and advances in instrumentation , analysis , and applications . J . Biomed . Opt . 25 , 1 \u2013 43 ( 2020 ) . 66 . Trexler , A . J . , Sochacki , K . A . & Taraska , J . W . Imaging the recruitment and loss of proteins and lipids at single sites of calcium - triggered exocytosis . Mol . Biol . Cell 27 , 2423 \u2013 2434 ( 2016 ) . 67 . Sochacki , K . A . & Taraska , J . W . Correlative \ufb02 uorescence super - resolution localization microscopy and platinum replica EM on unroofed cells . Methods Mol . Biol . 1663 , 219 \u2013 230 ( 2017 ) . 68 . Yasuda , R . Imaging intracellular signaling using two - photon \ufb02 uor - escent lifetime imaging microscopy . Cold Spring Harb . Protoc . 2012 , 1121 \u2013 1128 ( 2012 ) . 69 . K\u00f6llner , M . & Wolfrum , J . How many photons are necessary for \ufb02 uorescence - lifetime measurements ? Chem . Phys . Lett . 200 , 199 \u2013 204 ( 1992 ) . 70 . Mastronarde , D . N . Automated electron microscope tomography usingrobustpredictionofspecimenmovements . J . Struct . Biol . 152 , 36 \u2013 51 ( 2005 ) . 71 . Kremer , J . R . , Mastronarde , D . N . & McIntosh , J . R . Computer visualization of three - dimensional image data using IMOD . J . Struct . Biol . 116 , 71 \u2013 76 ( 1996 ) . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 14 72 . Alfonzo - M\u00e9ndez , M . A . , Sochacki , K . A . , Strub , M . - P . & Taraska , J . W . Dual clathrin and integrin signaling systems regulate growth factor receptor activation . Nat . Commun . 13 , 905 ( 2022 ) . 73 . Tinevez , J . Y . et al . TrackMate : an open and extensible platform for single - particle tracking . Methods 115 , 80 \u2013 90 ( 2017 ) . 74 . Doyon , J . B . et al . Rapid and ef \ufb01 cient clathrin - mediated endocytosis revealed in genome - edited mammalian cells . Nat . Cell Biol . 13 , 331 \u2013 337 ( 2011 ) . 75 . Pulupa , J . , Prior , H . , Johnson , D . S . & Simon , S . M . Conformation of the nuclear pore in living cells is modulated by transport state . Elife 9 , e60654 ( 2020 ) . 76 . Schindelin , J . et al . Fiji : an open - source platform for biological - image analysis . Nat . Methods 9 , 676 \u2013 682 ( 2012 ) . Acknowledgements We thank Dr . L . D . Islas ( Universidad Nacional Aut\u00f3noma de M\u00e9xico ) and Dr . H . Murakoshi ( National Institute for Physiological Sciences ) for sharing the absorbance and emission spectra of DPA and ShadowY respectively . We thank Dr . J . Jiang ( National Heart , Lung , and Blood Institute ) for use of the ChemiDoc . We thank National Heart , Lung , and Blood InstituteLight Microscopy Core and Electron Microscopy Core for use of instruments and advice . We thank G . Haber for help with coding . We thank members of the Taraska lab for helpful discussions and edits . J . W . T . is supported by the Intramural Research Program of the National Heart , Lung , and Blood Institute , National Institutes of Health . K . O . is supported by JSPS Research Fellowship for Japanese Biomedical and Behavioral Researchers at NIH . Author contributions K . O . , K . A . S . , and J . W . T . designed the research . K . O . performed experi - ments and analysis . K . A . S . developed software for analysis . M . - P . S . helped with molecular cloning and western blots . J . W . T . supervised the project . K . O . and J . W . T . wrote the paper . All authors contributed to the interpretation of the data and commented on the manuscript . Funding OpenAccessfundingprovidedbytheNationalInstitutesofHealth ( NIH ) . Competing interests The authors declare no competing interests . Additional information Supplementary information The online version contains supplementary material available at https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 . Correspondence and requests for materials should be addressed to Justin W . Taraska . Peer review information Nature Communications thanks Sergi Padilla Parra and the other , anonymous , reviewer ( s ) for their contribution to the peer review of this work . Peer reviewer reports are available . Reprints and permissions information is available at http : / / www . nature . com / reprints Publisher \u2019 s note Springer Nature remains neutral with regard to jur - isdictional claims in published maps and institutional af \ufb01 liations . Open Access This article is licensed under a Creative Commons Attribution 4 . 0 International License , which permits use , sharing , adaptation , distribution and reproduction in any medium or format , as long as you give appropriate credit to the original author ( s ) and the source , provide a link to the Creative Commons license , and indicate if changes were made . The images or other third party material in this article are included in the article \u2019 s Creative Commons license , unless indicated otherwise in a credit line to the material . If material is not included in the article \u2019 s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use , you will need to obtain permission directly from the copyright holder . To view a copy of this license , visit http : / / creativecommons . org / licenses / by / 4 . 0 / . ThisisaU . S . Governmentworkandnotundercopyrightprotectioninthe US ; foreign copyright protection may apply 2023 Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36304 - 7 Nature Communications | ( 2023 ) 14 : 732 15",
    "packerImportanceSDICurrent1979": "The Importance of SDI for Current Awareness in Fields with Severe Scatter of Information Katherine H . Packer Faculty of Library Science , University of Toronto , Toronto M55 1A 1 , Canada Dagobert Soergel College of Library and Information Services , University of Maryland , College Park , MD 20742 Scope : The purpose of this study was to explore the use and effectiveness of current awareness ( CA ) methods , espe - cially SDI , among chemists at Canadian universities . Method : We used a questionnaire which was mailed to a sample of 170 ( response rate 80 % ) . The variables studied include per - ceived scatter of information , use of various CA methods , time spent on CA , perceived success in CA , and CA effi - ciency = successhime spent . The results of this explanatory study are tentative . Major results : Keeping up - to - date is more difficult in specialties with high scatter of information ( hypothesis 5 ) . SDI seems to be the only method capable of counteracting scatter ( hypotheses 5 - 7 ) . In contrast , if scatter is low , SDI use decreases CA efficiency ( hypothesis 6a ) ; scanning journals , the most prevalent CA method , seems to be sufficient for low - scatter specialties . High scatter does not lead to increased SDI use , in some cases due to per - ceived difficulty of profile construction ( hypothesis 3a ) . Conclusions : SDI services should concentrate efforts on serving scientists in specialties with high scatter of informa - tion by developing systems that are capable of searching across conventional disciplines and specialties , by providing assistance in profile development , and by conducting a mar - keting campaign directed toward these scientists . Introduction Selective dissemination of information ( SDJ ) , as defined by Housnian and Kaskela [ I ] , is a service \u201cthat attacks the hforniation problem by keeping him [ the scientist ] con - tinuously informed of new documents published in his areas of specialization so that he can keep abreast of the latest developments . \u201d When first introduced it was thought to hold the promise of solving the scientists\u2019 problem of keeping up - to - date . However , in the 20 years since H . P . Lulin first put forward the concept , SDI services have been used far less than predicted and not nearly enough for the promise to be realized . Received August 1977 ; revised November 27 , 1978 . 0 1979 by John Wiley & Sons , Inc . This raises the question : Is it that scientists fail to per - ceive the value of SDI to them , or is it that they identify problems in the use of SDI that have not been sufficiently recognized by the designers of SDI systems , or if recog - nized , not effectively solved ? A mail questionnaire survey of chemists and chemical engineers at Canadian universities was undertaken to shed light on this question . The major aims of this study were to ( 1 ) test the effec - tiveness of SDI as a method of current awareness and , in particular . explore the factors which influence the effective - ness of SDI ; ( 7 - ) find out the reasons why scientists do or do not subscribe to an SDI service . In addition to contributing to an understanding of the information transfer process in science it was hoped that the results of the study would be of value to managers of SDI services . In order to gain an understanding of the use of SDI services it was thought necessary to look at the entire current awareness effort of scientists . Some subject fields were known to be served by a small number of highly specialized journals . In others , particularly interdisciplinary fields , relevant articles may be published in a large number of different journals . I t was an underlying assumption of this study that such scatter of information in a subject field would play an important role in explaining the effectiveness of any current awareness method and particular attention was paid , therefore , to scatter of information as an indepen - dent variable . This study is exploratory in nature . Its results must be considered preliminary . Some hypotheses were formulated a priori , but other , new hypotheses were also formulated based on the results of the data analysis . Clear distinction between a priori and a posteriori hypotheses has been main - tained in the section dealing with the results . The remainder of this section gives an overview of the findings and draws some conclusions for the management of SDI services . The sections that follow discuss in detail the methodology used , the results , and the limitations of the study . JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE 0002 - 8231 179 / 0030 - 0125 $ 01 . OO Overview of Results Fifty - six percent of the respondents reported a scatter value , of 4 or 5 on a 5 - point scale . All respondents made use of a variety of current awareness methods . On the whole , formal methods were used slightly more than infor - mal methods . Only 19 % of the respondents subscribed to an SDI service . The mean time spent on current awareness was from 2 to 5 hours per week , and the average level of success reported was 3 on a 5 - point scale ( adequate ) . The following relationships between variables were found : ( 1 ) The more time spent on current awareness the higher the level of success achieved . ( 2 ) The more time spent on current awareness the higher the productivity ( as measured by number of journal articles published ) . ( 3 ) The greater the scatter of information , the more use made of current awareness methods and the more time spent on this activity . The respondents in high - scatter fields were no more likely to subscribe to an SDI service , however , than respondents in low - scatter fields . ( 4 ) The respondents in fields with high scatter were as successful in keeping up - to - date as respondents in fields with low scatter . However , they do so at a price . ( 5 ) Respondents in high - scatter fields had to spend more time to achieve the same level of success . Their effi - ciency ( success per time spent ) was lower . This effect is very strong for nonsubscribers to an SDI service . Among the subscribers to an SDI service the negative effect of high scatter on efficiency is less severe . ( 6 ) In the total sample subscribers to an SDI service are no more efficient in keeping up - to - date than nonsubscribers . However . when the sample was divided into a low - scatter and a high - scatter group , the effect of subscribing to an SDI service was found to have a reverse effect in the two groups . For the low - scatter group , subscribing leads to lower effi - ciency ; for the high - scatter group subscribing leads to higher efficiency . No such effects were found for other current awareness methods . The results can be summarized as follows : ( 7 ) Among the current awareness methods tested , sub - scribing to an SDI service is the only one that was found to have any capacity to counteract scatter . In spite of the apparent success of SDI services in counteracting scatter of information , as already reported above . respondents in high - scatter fields were no more likely to subscribe to an SDI service than those in low - scatter fields . These findings have practical implications for the development of SDI services . From interviews it became evident that some scientists recognize the problem of scatter of information , but have serious doubts about the possibility of designing a profile that would identify all rele - vant material and at the same time suppress irrelevant material except when interests were very narrow and specific . They viewed scatter of information as a reason for not subscribing to an SDI service . In conclusion , it must be emphasized that the finding that subscribing to an SDI service counteracts scatter of in - formation is only tentative . While tendencies are discernible from the data none of the tests produced significant results , probably because of the small number of subscribers in the sample . Implications for the Management of SDI Services SDI services should concentrate efforts on serving scien - tists in specialties with high scatter of information . Systems must be designed so that searching across conventional dis - ciplines and specialties ( often this means across data bases ) is easy and convenient . Intermediaries must be available in local libraries to assist in profile development , a task that is more difficult for users in high - scatter specialties . These intermediaries must be trained particularly with respect to this task , and , perhaps more importantly , a marketing cam - paign must be directed at prospective users in high - scatter areas to convince them of the powers of an SDI service . Otherwise many members of this group that stand to gain most from using an SDI service will not even make an attempt to do so . These measures are essential if the poten - tial of SDI services is ever to be realized , and their benefits enjoyed , by the research community . Methodology Study Population Chemists and chemical engineers in Canadian universities were selected as the study population because this group had had an SDI system available to it since 1969 ; a period of five years provided an opportunity for this group to test the system . University faculties were selected because of their accessibility . Data were collected in the Fall of 1974 . Sampling Procedure Because a careful study of the 34 institutions that com - prised the target population indicated that the variance within the institutions was at least as great , if not greater , than that between the institutions , a systematic sample was decided on . A very accurate sampling frame was established based on current calendars from all 34 universities . The in - dividual faculty members were listed by institution , within institution by rank , and within rank alphabetically . This method ensured proportional representation from both official language groups , from all regions of the country , and from all academic ranks . A 1 - in - 5 systematic sample was drawn beginning with a randomly selected number be - low 6 . This procedure produced a sample of 170 names . ( Usable data were collected for 134 or SO % , see the Re - sponse Rate section . ) Data Collection Method The method chosen for data collection was a mailed questionnaire . A small number of supplementary interviews 126 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - May 1979 were carried out to check the interpretation of the data with the respondents themselves in order to validate some of the rating scales used , to throw light on apparent anomalies , and to add depth to the study . Response Rate The response rate was high : 80 % with usable data for a total of 134 respondents . The known characteristics of the nonrespondents were checked with a view to discovering what , if any . bias they might introduce into the study . The most serious discrepancy between the two groups , respon - dents and nonrespondents , concerned the proportion sup - ported by research grants . Since several of the questions dealt with research it was perhaps to be expected that a higher proportion of those not supported by grants would fail to respond . Variables Studied The variables considered in this article are the following : ( 1 ) Perceived scatter of information in the respondent\u2019s specialty . ( 2 ) The amount of use made of each of the following current awareness methods as perceived by the respondents . ( 2 . 1 ) Scanning professional journals ( 2 . 2 ) Scanning books and review books ( 2 . 3 ) Scanning current awareness tools , e . g . , Chemical ( 2 . 4 ) Use of an automated current awareness service , ( 2 . 5 ) Reports from students and assistants ( 2 . 6 ) Personal communication with colleagues in one\u2019s own institution ( 2 . 7 ) Personal communication with colleagues in other institutions ( 2 . 8 ) Refereeing for journals Titles e . g . , CAN / SDl ( 3 ) Whether or not the respondent subscribed to an SDI ( 4 ) Time spent on current awareness as perceived by the ( 5 ) Perceived success in keeping up - to - date . ( 6 ) Perceived efficiency in keeping up - to - date ( efficien - cy = success / time spent ) . ( 7 ) Productivity . service . respondent . ( 3 ) Amount of use was measured on this scale : 0 1 2 3 4 never rarely occasionally regularly heavily A total score for current awareness activity was de - rived by adding the values for all eight current awareness tools . Partial scores for formal and informal current aware - ness methods were derived in like manner ( Chart 1 ) . ( 3 ) The variable subscriber / nonsubscriber was intro - duced as a cross - check on the responses to question 2 . 4 above on the use of automated current awareness services , and also to see whether there were any respondents who shared in the use of SDI printouts received by a colleague , and thus were SDI users without being SDI subscribers . Those who identified themselves as nonsubscribers were asked their reasons for not subscribing . ( 4 ) Time spent was measured through this questionnaire item : Please indicate on the scale below the amount of time you spend each week in keeping up - to - date on developments in your field . 1 3 3 4 5 6 under 1 i2 2 - 5 5 - 10 10 - 20 over 1 hr hr h r 11 r hr 20 hr ( 5 ) Perceived success was measured through this ques - tionnaire item : Please indicate on the scale below how well you are able to keep up in your field . 1 2 3 4 5 have serious not too adequately quite well very well trouble well ( 6 ) Perceived efficiency was computed by dividing the scale value of perceived success by the midpoint of the time - spent span ( e . g . , 3 . 5 for 3 : 2 - 5 hr ) , admittedly a rather crude procedure justified only by the exploratory nature of this study . ( 7 ) Productivity was measured using the following scale for reporting the number ofjournal articles published in the previous five years : 0 1 - 5 6 - 1 0 1 1 - 20 21 - 50 over 50 These variables were measured as follows : ( I ) Perceived scatter was measured through the follow - ing questionnaire item : Please rate this field with respect to ease of keeping up - to - date by circling the appropriate number on the scale below . Number 1 represents a field in which one can keep up - to - date by regularly checking only one or two sources , and 5 a field in which information is scattered in a great many sources . 1 2 3 4 5 In the interpretation of the results presented in the following section it is important to keep in mind the nature of these variables . They represent the perception of the re - spondent , not independent assessment ( see the Limitations of the Study section ) . Results The main purpose of the following analysis is to probe the effectiveness of current awareness tools in fields with JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - May 1979 127 CHART 1 . Use made of the various methods of current awareness ( number of respondents ) . Method Amount of Use Never Rarely Sometimes Regularly Heavily Weighted ( x 0 ) ( x 1 ) ( x 2 ) ( x 3 ) ( x 4 ) Totals Formal Journals Review books , etc . CA tools SDI service Informal Communication with other institutions Student reports Communication with own institution Refereeing 0 1 12 93 3 9 6 20 5 14 22 8 3 - 2 26 31 34 19 50 41 2 65 53 41 51 61 44 52 18 30 30 16 16 Subtotal . Formal Methods 35 10 37 9 38 6 Subtotal , Informal Methods Total 26 4 Grand Total 420 342 3 14 124 1200 287 279 269 835 2035 230 2265 severe scatter of information . The analysis will move toward this goal through steps of increasing complexity . First , distributions of individual variables in the total sample are given . Second , the effect of scatter of informa - tion on perceived success , time spent , and efficiency is analyzed . Third , the effectiveness of current awareness methods , in particular SDI , is probed by analyzing their impact on efficiency in keeping up - to - date . All hypotheses , whether a posteriori or a priori , are numbered in one sequence . Distributions of Variables in the Total Sample ( I ) Most respondents reported having to cope with a fairly scattered field . Figure 1 summarizes the responses . The mean was above the midpoint on the scale . ( 2 ) The information collected regarding the use made of the various methods of current awareness is summarized in Chart 1 . More use was made of the formal methods of com - munication than the informal , and the use made of journals is significantly heavier than that of any other method . All respondents made some use of journal scanning as a means of keeping abreast with developments in their field . Inter - views threw light on the kind of use made of the formal and informal methods of communication . Informal methods were useful for generating new ideas , in identify - ing what was being done and by whom , and for exchanging information on techniques , but evaluation of the work required that a written report be available for study . ( 3 ) Only 19 % of the respondents subscribed to an SDI service . Reasons for not doing so ranged from ignorance of the existence of such services ( 15 % ) to personal preference ( 45 % ) . Fifteen percent said they did not require such services because of the nature of their work . Twenty - five percent did not subscribe because of the cost . In spite of cost representing a considerable barrier no one interviewed favored offering the service free on the grounds that many would accept the service under such circumstances but make no use of it . Figures 2 and 3 summarize the responses for time spent on current awareness and success in keeping up - to - date . The mean time reported is from 2 to 5 hours per week , and the mean level of success is 3 ( adequately ) . ( 45 ) 50 - 40 - 30 - Number of Respondents 20 - 10 - 5 - 34 30 - 4 N : 134 5 ma . , mun Scatter of Information FIG . 1 . Scatter of information in the field of specialization . 128 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - May 1979 Number Of Respondents 0 - 1 1 - 2 53 38 2 - 5 N : 134 10 - 20 Hours Spent on Current Awareness I , IG . 2 . Time spent on current awareness . The time estimate is a very much shorter period of time than that reported by the Committee on Corporation Asso - ciates [ 2 ] in the study carried out on industrial chemists , where the average amount of time spent on current aware - ness was 7 . 5 hours per week . The method of drawing the sample , which was to have each of the company representa - tives select ten of his company\u2019s professional research che - mists in \u201cas random a manner as was convenient , \u201d may alone account for the difference . In addition , only a 50 % return is reported and no investigation was made to deter - mine whether the nonrespondents were discriminably differ - ent with respect to time spent on reading . It is also possible that the use of SDI systems and current awareness tools such as Chemical Absrructs has actually resulted in lower - ing the average amount of time spent on this activity . T o validate the data collected for use of current aware - ness methods and time spent . Kendall\u2019s tau test was per - formed . The result : tau = 0 . 31 , significant beyond 0 . 00005 . TABLE 1 . Relationship between time spent on current awareness and succe \\ s in keeping up - to - date . a Time Spent on Current Awareness Success in C A 1 - 2 hr 2 - 5 hr 5 + h r Ail Low 38 25 16 25 Adequate 49 49 52 50 High 13 2G 32 25 100 100 100 100 I7 31 53 44 134 - - - - % slues are in percent . Number of Respondents 6 7 7 30 \u201c135 Success in Keeping Up - to - Date F\u2018IG . 3 . Success in keeping up - to - date . ( 6 ) After considering the variables time spent and success individually , their relationship was investigated . The data shown in Table 1 suggested the following : A posteriori hypothesis 1 : There is a direct relationship between time spent on current awareness and level of success achieved ( tau significant at 0 . O 1 ) . ( 7 ) One further result is reported here though it does not fit in the main line of the argument . Maizell , in his study of industrial research chemists [ 3 ] , found that the most creative chemists spent more hours on the job reading the technical and scientific literature , examined more jour - nals made available to them through company library facili - ties , and were more interested in maintaining their own card indexes of current journal articles than the least creative chemists . In spite of the fact that the present study was confined to current awareness activities and therefore is in no sense a replication of Maizell\u2019s study , it was thought of some interest to find out whether the information - gather - ing patterns of university faculty in departments of chemis - try and chemical engineering resembled those of industrial research chemists . No variable \u201ccreativity\u201d was available for testing , but Maizell had found that the most creative group of chemists had a higher average number of publications than did the less creative groups , and therefore the two variables , pro - ductivity ( measured by number of journal articles published in the last five years ) and time spent on current awareness , were selected in order to test the first of his findings . JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - Mav 1979 129 TABLE 2 . Relationship between productivity and time spent on current awarenessa Productivity ( Articles Published ) 0 - 2 hr 2 - 5 hr 5 - 20 hr All Time Spent on Current Awareness TABLE 3 . Relationship between scatter and time spenta Scatter of Information Time Spent Low High All 2 hr or less ~ ~ 40 19 28 0 - 5 36 30 9 25 2 - 5 hr 40 39 39 6 - 10 28 18 19 21 11 - 20 25 28 26 26 Over 20 11 24 46 28 100 100 100 100 - - - - n 36 50 43 129 aValues are in percent . A priori hypothesis 2 : There exists a direct relationship between productivity and tirne spent on current awareness ( tau significant beyond 0 , 000 1 , see Table 2 ) . This highly significant result tends to confirm Maizell\u2019s findings . Effects of Scatter of Information A priori hypothesis 3 : There exists a direct relationship between scatter of information and both use made of current awareness methods ( tau significant beyond 0 . 001 ) and time spent on current awareness ( tau significant at 0 . 0007 , see Table 3 ) . A priori hypothesis 3a : There exists a direct relationship between scatter of information and subscribing to an SDI service ( chi square not significant ) . This point will be discussed in more detail below . If one considers just success in keeping up - to - date with - out regard to the data on time spent , one might formulate the following : A priori hypothesis 4 : There exists an inverse relationship between scatter of information and success in keeping up - to - date ( tau not significant ) . The data in Table 4 show that t . here is not even a trend in this direction . The explanation for the negative result lies in the data on time spent . Respondents in high - scatter specialties make up for the disadvantage by spending more More than 5 hr 20 42 33 100 100 100 - _ _ - n 58 14 132 aValues are in percent . time on current awareness . This discovery suggests an in - vestigation of the level of success achieved per time spent on current awareness . For this purpose the derived variable efficiency ( efficiency = success / time spent ) was introduced . The sample was divided into a low - scatter ( 1 - 3 ) group and a high - scatter ( 4 - 5 ) group . These two groups were then com - pared with respect to efficiency using the Mann - Whitney I / test . Respondents of both groups together are ranked according to their efficiency . The test then determines whether the respondents of one group are concentrated in the upper part of the rank listing to a degree that would be unlikely under the null hypothesis of no effect . The data are also shown in a 2 X 3 - table ( Table 5 ) which is , of course , much cruder than the data used for the Mann - Wliitney U test . The results of this comparison confirm the following : A priori hypothesis 5 : There is an inverse relationship between scatter of in - formation and efficiency in keeping up - to - date ( U significant at 0 . 01 ; graphically shown in Fig . 4 , bottom line ) . TABLE 4 . Relationship between scatter of information and success in keeping upto - date . a Scatter of Information Success in CA Low High All Not too well 35 19 26 Adequately 41 58 50 Well 24 23 24 100 100 100 - - - n 58 I 5 133 aValues are in percent . 130 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - May 1979 TABLE 5 . ciency in keeping up - to - date . a Relationship between scatter of information and effi - Whole Sample , Low and High Scatter Scatter of Information Efficiency Low High All Low 30 55 45 High 70 45 55 - - - 100 100 100 tz 56 14 130 ( a ) Nonsubscribers , Low and High Scatter Scatter of Information Efficiency Low High All Low 21 59 44 High 13 41 56 100 100 100 - - - n 45 54 99 ( b ) Low - Scatter Nonsubscribers and High - Scatter Subscribers Subgroup Low - Scat ter High - Sca tter Efficiency Nonsubscriber Subscriber Low High 21 13 45 55 100 n 45 100 20 ( c ) Subscribers , Low and High Scatter Scatter of Information Efficiency Low High All Low 45 45 45 High 55 55 55 - - - 100 100 100 n 11 20 31 ~ \u2018values are in percent . This result refers to the total current awareness effort . It shows that researchers in high - scatter fields need more effective methods for keeping up - to - date than they currently use to overcome the problem of high scatter . The problem is all the more pressing as the majority of respondents fall into the high - scatter group . But what methods should these scientists use ? The next section will address the effective - ness of particular current awareness methods and explore what methods are capable of counteracting scatter . Effectiveness of Particular Current Awareness Methods The effectiveness of a current awareness method can be gauged ( 1 ) by the extent to which it enables a user to keep up - to - date as measured by success , and , more importantly , ( 2 ) by the efficiency of the user in keeping up - to - date as measured by the relationship between success and time spent on keeping up - to - date . Reflection on the role of scatter led to the hypothesis that scatter of information is an important factor in explaining the effectiveness of a particular current awareness method . This is borne out through various results to be presented below . When scatter is low , i . e . , when a field is well served by a few specialist journals , scanning the appropriate journals will probably result in high efficiency ( success / time spent ratio ) . When the information is scattered in many sources , however , journal scanning becomes very time consuming and effi - ciency is low . An SDI service might be extremely helpful in this situation by pulling together relevant references from many varied sources . It is the ability to counteract scatter of information , then , that is the true test of the effectiveness of any method of current awareness . As the effectiveness of subscribing to an SDI service was the primary focus of this study , results for this method are discussed first . Subsequently , subscribing to an SDI service is compared with other methods of keeping up - to - date , particularly the use of manual current awareness tools , such as Chemical Titles or Chemical Abstracts . Effectiveness of Subscribing to an SDI Service The interaction between scatter and subscriber / nonsub - scriber can be approached from two angles . One is to repeat the analysis leading to hypothesis 5 ( high scatter of infor - mation leads to lower efficiency ) comparing appropriate subgroups of the total sample . Repeating the analysis for nonsubscribers leads to hypothesis 5a . A posteriori hypothesis 5a : For nonsubscribers to an SDI service high scatter of information leads to lower efficiency in keeping up - to - date [ U significant at 0 . 0005 , 2 value of - 3 . 4 , Table 5 ( a ) l . This is the same effect that was detected for the sample as a whole , only stronger . The next question is : Does subscribing to an SDI service alleviate the negative effects of scatter ? The most obvious way to try to answer this question is to repeat the analysis for subscribers to an SDI service . However , the data pre - sented below indicate that low - scatter respondents lose effi - ciency by subscribing to an SDI service and that the most efficient group of all are low - scatter nonsubscribers . This JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - May 1979 131 I Low Scatter 1 Hiqh Scatter 1 A l l I I 99 ( Sa ) 54 I 45 Non - subscriber I I 31 ( SC ) 20 I l1 \u201ct Subscriber ( 5 ) 74 I 130 c - All FIG . 4 . Summary of the effect of scatter of information and sub - scribing to an SDI service on efficiency in keeping up - to - date [ Tables 5 , 5 ( a ) , 5 ( b ) , 5 ( c ) , 6 , 6 ( a ) , and 6 ( h ) ] . Arrows point to group with higher efficiency ; s = significant , t = tendency , nt = not even a ten - dency . group , then , could be used as a \u201cbaseline\u201d for judging effi - ciency of other groups . Following this reasoning high - scatter subscribers are compared to low - scatter nonsub - scribers ; the results lead to hypothesis 5b . A posteriori hypothesis 5b : Subscribing to an SDI service alleviates the negative effect of scatter of information on efficiency in keeping up - to - date [ Mann - Whitney U for the scatter - efficiency relationship converts to a Z value of - 1 . 6 , not significant , see Table 5 ( b ) ] . The results of this test seem to indicate that high scatter of information leads to lower efficiency in keeping up - to - date even if a respondent subscribes to an SDI service , but the negative effect of scatter is not as strong as , for high - scatter nonsubscribers . Another , and perhaps more direct , way to approach the interaction between scatter and , subscriber / nonsubscriber is to analyze the effect of subscriber / nonsubscriber on effi - ciency , first for the sample as a whole and then for low scatter only and high scatter only . Reflection on the pur - pose of an SDI service led to hypothesis 6 . A priori hypothesis 6 : Subscribers to an SDI service achieve higher efficiency in keeping upto - date than nonsubscribers ( U not sig - nificant , but see a posteriori hypotheses 6a and 6b ) . A look at Table 6 clearly reveals that not even a tendency exists . Initially this result was viewed as surprising . Since the previous analysis had confirmed the impact of scatter on the relationship between time spent and success ( a priori hypothesis 5 ) , an explanation was sought by investigating the interaction between scatter of information and sub - scriber / nonsubscriber . To this end the comparison between subscribers and nonsubscribers with respect to efficiency TABLE 6 . efficiency in keeping up - to - date . a Whole Sample Relationship between subscribing to an SDI service and Subscriber / Nonsuhscriber Efficiency Nonsubscriber Subscriber All Low 44 45 45 High 56 55 55 100 100 100 - - - n 99 31 130 ( a ) Low Scatter SubscriberlNonsubscriber Efficiency Nonsubscriber Subscriber All Low 27 45 30 High 73 55 70 100 100 100 n 45 11 56 - - - ( b ) High Scatter ~ ~ ~ Suhscriber / Nonsubscriber Efficiency Nonsubscriber Subscriber All Low 59 45 55 High 41 55 45 100 100 100 n 54 20 74 - - - aValues are in percent was performed within the high - scatter group and within the low - scatter group . The results are as follows : A posteriori hypothesis 6a : Among chemists working in specialties with low scatter of information , subscribers to an SDI service have lower efficiency in their current awareness activities than nonsubscribers [ U significant at 0 . 02 , see Table 6 ( a ) ] . A posteriori hypothesis 6b : Among chemists working in specialties with h & h scatter of information , subscribers to an SDI service have a higher efficiency in their current awareness activities than nonsubscribers [ U not significant , transforms to Z = 1 . O , but data in Table 6 ( b ) show a tendency ] . 132 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - May 1979 LOW scatter FIG . 5 . Summary of the effect of scatter of information and the use of manual current awareness tools on efficiency in keeping up - to - date [ Tables 7 , 7 ( a ) . 7 ( b ) , 7 ( c ) , 8 . 8 ( a ) , and 8 ( b ) ] . Arrows point to group with higher efficiency ; s = significant . t = tendency . but not Fignificant , nt = not even ii tendency . High Scatter All Upon reflection these results make sense . In a field where one can keep up - to - date by scanning a few journals , reading of SDI notices will add only a few additional refer - ences and the benefit is not worth the cost ( time spent ) . ( The data suggest that in both the low - scatter and the high - scatter groups subscribers do better than nonsubscribers in terms of absolute success where time spent is not con - sidered . ) The reversed effects of subscribing in the low - scatter group and the high - scatter group explain , then , why there is no effect at all in the total sample . While , on the basis of the tests reported above , subscribing to an SDI service does not totally eliminate the effects of scatter of information , subscribers do achieve higher efficiency than nonsubscribers in fields in which information is scattered ( strong tendency ) . Figure 4 summarizes the results of the efficiency analyses for subscription to an SDI service . Effectiveness of Using Other CA Tools The efficiency analyses were now repeated for the use of manual current awareness tools which is , in a sense , the manual version of subscribing to an SDI service . SDI sub - scribers were excluded from these analyses . The results are displayed in Fig . 5 ( which also refers to the supporting data ) . Comparison with Fig . 4 brings out striking differences . In the low - scatter category neither high nor low users of manual current awareness methods enjoy an advantage in terms of efficiency [ Table 8 ( a ) ] ; therefore no \u201cbaseline\u201d group could be identified . Subscribing to an SDI service was seen to counteract somewhat the negative effect of scatter of information . No such effect is found for high use of CA tools . Among high users of CA tools those with high scatter are less efficient in keeping up - to - date [ strong tendency , Table 7 ( c ) ] . When high - scatter high users of CA tools were compared with low - scatter low users of CA tools using the Mann - Whitney U test high - scatter high users of CA tools were found to be significantly less effi - cient [ U value converts to a Z value of - 3 . 38 , significant at 0 . 0004 , Table 7 ( b ) ] . In the high - scatter group SDI sub - scribers were more efficient than nonsubscribers [ strong tendency , Table 6 ( b ) ] . No such effect is found comparing high users of CA tools with low users in the high - scatter group [ Table 8 ( b ) ] . The relationship between scatter of information and effi - ciency of keeping up - to - date was investigated for all methods of keeping up - to - date using 2 X 2 cross - tabulations and the chi square test ( which is more convenient but less powerful than the Mann - Whitney U test ) . The results are summarized in Chart 2 . These results can be summarized in the following : A posteriori hypothesis 7 : Among the current awareness methods tested , sub - scribing to an SDI service is the only one that is capa - ble of counteracting scarter . Before these findings can be accepted , however , they must stand up to the test of replication , not only for this field but for other fields and other groups of researchers . Scatter of Information and Use of an SDI Service Since a direct relationship was found between scatter of information and the amount of use made of the various methods used to keep up - to - date , it would be reasonable to expect that high perceived scatter of information in the field of specialization would encourage the use of an SDI service , particularly in the light of the finding that sub - scribing to an SDI service seems to be the most effective way to counteract scatter . A cross - tabulation of the vari - ables subscriber / nonsubscriber and scatter did not produce significant results ( a priori hypothesis3a ) . The data provided no evidence to suggest that perceived scatter of information acted as an incentive to subscribe to an SDI service . On the contrary , many of the comments volunteered by the re - spondents revealed that they viewed scatter of information as a reason for not subscribing . To quote one of the respon - dents : \u201cMy fields of interest are rather broad and such services would not be able to select appropriate material . \u201d Against this quotation two comments by librarians actively engaged in assisting SDI users in profile development may be set [ 4 ] . Most of my customers tend to have broad interdisci - plinary interests , e . g . , pollution and environmental damage . I have had extreme success with CAN / SDI for people who are in multidisciplinary areas , for example , faculty in the Kinesiology Department . These people are interested in motor learning , muscle physiology , sports psychology , anthropometry , biomechanics , and the use of two different data bases combined with different types of search strategies , title word , subject term , source author , cited author have proved highly effective in literature retrieval in this area . JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - May 1979 133 TABLE 7 . Relationship between scatter and efficiency ( subscribers to SDI excluded ) ? Whole Sample TABLE 8 . Relationship between using manual current awareness tools and efficiency ( subscribers to SDI excluded ) . a Whole Sample Use of Manual CA Tools Scatter of Information Efficiency Low High All Efficiency Low High All Low 27 59 44 High 73 41 56 100 100 100 n 44 54 98 - ~ _ _ Low 46 43 45 High 54 51 55 100 100 100 - - - n 52 46 98 ( a ) Low Use of CA Tools , Low and High Scatter ( a ) Low Scatter Scatter of Information Efficiency Low High All Use of Manual CA Tools Efficiency Low High All Low 29 25 27 Low 29 61 46 High 71 39 54 _ _ _ _ - 100 100 100 n 24 28 52 High 71 I5 73 100 100 100 - ~ _ _ n 24 20 44 ( b ) Low Use with Low Scatter and High Use of CA Tools with High Scatter ( b ) High Scatter Use of Manual CA Tools Group Low - Scatter High - Scatter Efficiency Low User High User Efficiency ~ Low High All Low 61 58 59 High 39 42 41 100 100 100 n 28 26 54 - - _ _ Low 29 58 High 71 100 ~ 42 100 n 24 26 ' values are in percent . ( c ) High Use of CA Tools , Low and High Scatter Scatter of Information the variables deal with perceptions of the respondents rather than with independent observations . It would be desirable , if laborious , to check the results in a study where scatter of information in each field of specialization was measured by a Bradford analysis , success by the fraction of information obtained by the respondent over the totality of existent relevant information as determined by an ex - haustive search , and time spent in keeping up - to - date either by observation or from diary records maintained by the respondent . Finally , as already stated , it is uncertain how far one can generalize beyond the narrow population from which the sample was drawn . Efficiency Low High All Low 25 58 43 High 75 42 57 _ _ _ _ _ _ 100 100 100 I1 20 26 46 ' Values are in percent . Limitations of the Study As this study was exploratory many of the findings re - sult from a posteriori hypotheses suggested by the data rather than from apriori hypotheses tested by the data . All Acknowledgments The contributions of Professor L . B . Heilprin , Professor J . S . Kidd , and Professor Paul Wasserman to this study are 134 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - May 1979 CHART 2 . Summary of tests on current awareness methods . Relationship between Scatter and Efficiency within Groups Making Low or High Use of a CA Method , Respectively Method Used ~ No . of No . of Low Use of Method Respondents High Use of Method Respondents Not significant Significant , inverse 22 52 Significant , inverse 107 46 Journals CA tools ( SDI users excluded ) Nearly significant , inverse Review books , etc . Not significant , inverse tendency 64 Not significant , inverse tendency 65 Not significant , inverse tendency 64 Nearly significant , inverse tendency 65 Formal methods ( SDI users excluded ) Student reports Significant , inverse 85 Not significant , inverse tendency 43 Communication , own institution Not significant , inverse tendency 86 Not significant , inverse tendency 42 Communication , other institutions Significant , inverse 87 Not significant , slight inverse tendency 42 Refereeing Significant , inverse I02 12 Not significant , inverse tendency 21 Informal methods Significant , inverse Not significant , slight inverse tendency 55 SDI Significant , inverse 97 Not significant , no tendency 32 gratefully acknowledged . Special mention should be made of the assistance received from Professor M . B . Hargrove with respect to the statistical measures employed . References 2 . American Chemical Society . Committee on Corporation Asso - ciates . Subcommittee on the Economics of Chemical Informa - tion . Cost Effectiveness of Information Systems . Washington , DC : American Chemical Society ; 1969 . 3 . Maizell , R . E . \u201cInformation gathering patterns and creativity ; a study of research chemists in an industrial research labora - tory . \u201d Doctoral dissertation , Columbia University , 1957 , un - published . 4 . Personal communications with librarians actively engaged in profile development . 1 . Housman , E . M . ; Knskela , E . D . \u201cState of the Art of Selective Dissemination of Information . \u201d IEEE Transactions of Engi - neering Writingand Speech . I11 ( 2 ) : 78 ; September 1970 . JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE - May 1979 135",
    "cataliniMicrogeographyDirectionInventive2018": "This article was downloaded by : [ 100 . 15 . 126 . 133 ] On : 06 December 2020 , At : 18 : 31 Publisher : Institute for Operations Research and the Management Sciences ( INFORMS ) INFORMS is located in Maryland , USA Management Science Publication details , including instructions for authors and subscription information : http : / / pubsonline . informs . org Microgeography and the Direction of Inventive Activity Christian Catalini To cite this article : Christian Catalini ( 2018 ) Microgeography and the Direction of Inventive Activity . Management Science 64 ( 9 ) : 4348 - 4364 . https : / / doi . org / 10 . 1287 / mnsc . 2017 . 2798 Full terms and conditions of use : https : / / pubsonline . informs . org / Publications / Librarians - Portal / PubsOnLine - Terms - and - Conditions This article may be used only for the purposes of research , teaching , and / or private study . Commercial use or systematic downloading ( by robots or other automatic processes ) is prohibited without explicit Publisher approval , unless otherwise noted . For more information , contact permissions @ informs . org . The Publisher does not warrant or guarantee the article\u2019s accuracy , completeness , merchantability , fitness for a particular purpose , or non - infringement . Descriptions of , or references to , products or publications , or inclusion of an advertisement in this article , neither constitutes nor implies a guarantee , endorsement , or support of claims made of that product , publication , or service . Copyright \u00a9 2017 , The Author ( s ) Please scroll down for article\u2014it is on subsequent pages With 12 , 500 members from nearly 90 countries , INFORMS is the largest international association of operations research ( O . R . ) and analytics professionals and students . INFORMS provides unique networking and learning opportunities for individual professionals , and organizations of all types and sizes , to better understand and use O . R . and analytics tools and methods to transform strategic visions and achieve better outcomes . For more information on INFORMS , its publications , membership , or meetings visit http : / / www . informs . org MANAGEMENT SCIENCE Vol . 64 , No . 9 , September 2018 , pp . 4348 \u2013 4364 http : / / pubsonline . informs . org / journal / mnsc / ISSN 0025 - 1909 ( print ) , ISSN 1526 - 5501 ( online ) Microgeography and the Direction of Inventive Activity Christian Catalini a a MIT Sloan School of Management , Massachusetts Institute of Technology , Cambridge , Massachusetts 02142 Contact : christian @ catalini . com , http : / / orcid . org / 0000 - 0003 - 1312 - 6705 Received : December 2 , 2016 Accepted : March 15 , 2017 Published Online in Articles in Advance : July 19 , 2017 https : / / doi . org / 10 . 1287 / mnsc . 2017 . 2798 Copyright : \u00a9 2017 The Author ( s ) Abstract . I provide novel empirical evidence grounded in an original theoretical frame - work to explain why colocation matters for the rate , direction , and quality of scien - ti\ufb01c collaboration . To address endogeneity concerns due to selection into colocation and matching , I exploit the constraints imposed on the spatial allocation of labs on the Jussieu campus of Paris by the removal of asbestos from its buildings . Consistent with search costs constituting a major friction to collaboration , colocation increases the likelihood of joint research by 3 . 5 times , an e\ufb00ect that is mostly driven by lab pairs that face higher search costs ex ante . Furthermore , separation does not negatively a\ufb00ect collaboration between previously colocated labs . However , while colocated labs grow increasingly similar in topics and literature cited , separated ones embark on less correlated research trajectories . Research outcomes , instead , seem to be mostly in\ufb02uenced by how distance a\ufb00ects exe - cution costs : after colocation , labs are more likely to pursue both lower - quality projects ( a selection e\ufb00ect ) and high - quality projects ( an e\ufb00ort e\ufb00ect ) . Opposite e\ufb00ects on quality are observed after separation . Whereas search costs a\ufb00ect which scientists are likely to collaborate together , execution costs shape the quality of their output . History : Accepted by Ashish Arora , entrepreneurship and innovation . Open Access Statement : This work is licensed under a Creative Commons Attribution - NonCommercial - ShareAlike 4 . 0 International License . You are free to download this work and share with others for any purpose , except commercially , if you distribute your contributions under the same license as the original , and you must attribute this work as \u201c Management Science . Copyright \u00a9 2017 The Author ( s ) . https : / / doi . org / 10 . 1287 / mnsc . 2017 . 2798 , used under a Creative Commons Attribution License : https : / / creativecommons . org / licenses / by - nc - sa / 4 . 0 / . \u201d Funding : This research was funded by the MIT Sloan School of Management , the Centre for Innovation and Entrepreneurship at the Rotman School of Management , and the Martin Prosperity Institute . Supplemental Material : The online appendix is available at https : / / doi . org / 10 . 1287 / mnsc . 2017 . 2798 . Keywords : colocation \u2022 idea recombination \u2022 collaboration \u2022 search costs \u2022 microgeography \u2022 low - opportunity cost time \u201cThe truth is , it may have changed my whole life in a respect . I guess if I had any long - run thoughts then , it was to make a career doing statistics , econometrics , probability models , and things like that . But when I started talking on a regular basis with Paul , and he was so full of ideas and thoughts , it was impossible not to \ufb01nd my interests moving toward more straight eco - nomics . In a way the location of that o\ufb03ce and the fact that we liked each other so much had a major in\ufb02uence on the direction my career took . \u201d \u2014Nobel laureate Robert Solow describing how his o\ufb03ce location and resulting friendship with Nobel laureate Paul Samuelson a\ufb00ected his career ( Dizikes 2011 ) 1 . Introduction When we select a location , we are committing to spending a disproportionate share of our most scarce resource , time , in that particular place . Individuals and organizations , anticipating the role geographic dis - tance will play in their allocation of time , and in de\ufb01n - ing the opportunities and talent they will have access to , pay particular attention to location decisions . As a result , location choices are highly endogenous to eco - nomic outcomes . Geographic proximity not only increases the chance of a serendipitous interaction , but also lowers the cost of a scheduled one . Both types of interactions make it substantially easier to search for new collaborators within our local environment . When colocated , joint execution costs are also lower , as coordination , moni - toring , and the transfer of complex information can all rely on more frequent , face - to - face interactions . Thus , organizations spend considerable time and re - sources optimizing the spatial allocation of their teams and invest in infrastructure to allow for interdiscipli - nary work to \ufb02ourish when new opportunities are identi\ufb01ed . This endogenously shapes the trajectory of a university , R & D lab , or start - up , as the spatial alloca - tion often results from priors about which teams and individuals will bene\ufb01t the most from proximity . When a misalignment between the objectives of an organiza - tion and its current layout emerges , e\ufb00orts are made to compensate for geographic distance , for example 4348 Catalini : Microgeography and the Direction of Inventive Activity Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) 4349 by scheduling temporary colocation ( e . g . , joint meet - ings , conferences ) and remote interactions to recre - ate the bene\ufb01ts of colocation . A recent , large - scale example is Microsoft\u2019s ambitious relocation of 1 , 200 engineers , which cut travel time between buildings for employees by 46 % with the explicit objective of encouraging face - to - face conversations and serendipi - tous interactions instead of email and Skype meetings ( Nielsen 2016 ) . Empirically , this makes it extremely challenging to understand why colocation matters and under which conditions di\ufb00erent mechanisms are responsible for the bene\ufb01ts we attribute to proximity . The objective of this paper is to focus on how search costs and joint exe - cution costs shape inventive outcomes , and to provide novel empirical evidence and a theoretical framework that can help us separate between these two compet - ing , but not necessarily exclusive , mechanisms . If search costs are a key friction to collaboration , then colocation should have a positive e\ufb00ect on the proba - bility of collaboration , but separation may not neces - sarily have a negative e\ufb00ect , as teams that are aware of each other may be able to compensate for distance through temporary colocation and remote interactions . If instead joint execution costs are driving collaboration decisions , then the e\ufb00ects observed after colocation and separation should be exact opposites of each other ( as execution costs would increase almost immediately with distance ) . Furthermore , whereas search costs do not have a strong implication for the observed value of joint projects , if lower execution costs allow teams to apply more e\ufb00ort toward advancing their ideas , then the e\ufb00ect of colocation on quality will be ambigu - ous . On the one hand , lower execution costs may induce teams to select lower - quality projects ( a selec - tion e\ufb00ect ) . On the other hand , if lower execution costs allow teams to endogenously apply more e\ufb00ort\u2014and e\ufb00ort improves the quality of the underlying idea\u2014 we may also observe an increase in right - tail outcomes after colocation ( an e\ufb00ort e\ufb00ect ) . 1 I explore the relative role of these mechanisms in a setting where the spatial allocation is constrained by reasons that are orthogonal to the outcomes of interest\u2014i . e . , where exogenous variation is injected in the process of deciding where di\ufb00erent teams are placed . Combined with a di\ufb00erence - in - di\ufb00erences approach , this mitigates the endogeneity concerns typ - ically linked to selection into colocation . The setting is the university campus of Paris Jussieu , the leading scienti\ufb01c and medical complex in France . Following a research report by INSERM on the car - cinogenic e\ufb00ects of asbestos ( June 21 , 1996 ) , the French government introduced a full ban of the \ufb01re retardant from all public buildings , including the Jussieu ones . Starting in 1997 , a separate entity ( Etablissement Pub - lic d\u2019Am\u00e9nagement Universitaire de la R\u00e9gion Ile - de - France ( EPAURIF ) ) was put in charge of the asbestos removal process on campus , which led to \ufb01ve massive waves of lab relocations over 17 years . Because of the complexity and urgency 2 of the cleaning process , labs were forced to move often on short notice and with lit - tle in\ufb02uence over their new location . As a consequence , many labs found themselves next to new neighbors , and the overall spatial allocation was severely more constrained than before . I exploit this variation , com - bined with the reconstruction of \ufb01ne - grained , longi - tudinal information on inventive outcomes , to try to understand why colocation matters for the rate , direc - tion , and quality of inventive activity . Consistent with the search costs mechanism , results show that colocation increases the likelihood that two labs will collaborate by 3 . 5 times , an e\ufb00ect that is mostly driven by pairs that faced higher search costs ex ante . Moreover , separation has a nonsigni\ufb01cant e\ufb00ect on the probability and rate of collaboration , suggesting that labs that were previously exposed to each other are later able to sustain collaboration also over dis - tance . At the same time , while colocated labs become increasingly similar in the topics they work on and references they cite , 3 separated labs embark on less correlated research trajectories . This is consistent with search costs increasing considerably even at relatively low levels of geographic distance , a result that sup - ports past literature that has linked proximity to better information di\ufb00usion ( Allen 1977 , Cowgill et al . 2009 ) , formation of social ties ( Stuart and Liu 2010 , Liu 2013 ) , and knowledge \ufb02ows ( Ja\ufb00e et al . 1993 , Thompson and Fox - Kean 2005 , Thompson 2006 ) . Furthermore , conditional on collaboration , the qual - ity of a lab pair\u2019s output also changes following a change in distance . Colocated labs are 1 . 36 times more likely to produce a paper that will end up in the highest quartile of the citation distribution , and their variance in outcomes increases . Opposite results are observed for lab pairs that are separated . This is consistent with colocation a\ufb00ecting joint execution costs , and with both a selection and an e\ufb00ort e\ufb00ect playing a role in this context . Interestingly , the collaborations resulting from interactions between labs that faced higher search costs ex ante are more likely to be of high impact , suggesting that arbitrage opportunities may exist in encouraging interactions between communities of scientists that do not overlap through other channels ( e . g . , joint confer - ences and journals ) . Taken together , the \ufb01ndings highlight that whereas search costs mostly a\ufb00ect which scientists are likely to collaborate together , execution costs shape the qual - ity of their output . By allocating space , organizations profoundly shape the evolution of scienti\ufb01c trajecto - ries and the types of opportunities that are explored by di\ufb00erent teams . This involves a trade - o\ufb00 between the perhaps more e\ufb03cient exploitation of established research paradigms and the more costly exploration Catalini : Microgeography and the Direction of Inventive Activity 4350 Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) of new ones . Colocation is an expensive way to lower search costs , as supported by the overall changes in the collaboration portfolios of the labs observed in the data : during the moves , in aggregate , labs focused more inwards , increasing within - lab research at the expense of across - lab collaborations . While this may be a result of the suboptimal set of local peers the moves o\ufb00ered the labs relative to their ideal choices , it is also a reminder of how pursuing research across disciplines is a more costly endeavor than incremental , within - discipline work both for an institution and the scientists involved . In the absence of complementary changes in incentives ( e . g . , to favor cross - disciplinary research ) , scientists will focus where search and joint execution costs are lower \ufb01rst . Although temporary forms of colocation may not be as e\ufb00ective as the longer periods studied here in leading to actual knowledge \ufb02ows and collaboration , they may still allow for cross - pollination between research trajectories and break - through discoveries at a lower cost . The layout of the remainder of the paper is as follows . In the next section , I further develop the basic hypotheses of the theoretical framework to guide empirical predictions and interpretations of the \ufb01nd - ings . Section 3 describes the empirical setting , data , and empirical strategy . Section 4 reports the main results . Section 5 concludes . 2 . Theoretical Framework The ability of an economy to generate , di\ufb00use , and recombine ideas has a profound in\ufb02uence on its ability to sustain growth ( Lucas 1988 , Romer 1990 , Weitzman 1998 ) . Our understanding of a key economic phenomenon behind agglomeration and growth\u2014 localization economies\u2014relies on , among other fac - tors , basic assumptions about how knowledge is recombined locally versus over distance . Despite frequent references in the literature on knowledge \ufb02ows and localization to the role of colo - cation in knowledge transmission and recombination ( Breschi and Lissoni 2004 , Mairesse and Turner 2005 , Singh 2005 , Agrawal et al . 2006 , Fleming et al . 2007 , Belenzon and Schankerman 2013 ) , there is more lim - ited empirical evidence on the underlying mechanisms invoked to explain its e\ufb00ects . In part , this is due to the di\ufb03culty of \ufb01nding plausibly exogenous variation in location choice . Most of the existing literature relies on observational data ( Olson and Olson 2000 , Van den Bulte and Moenaert 1998 , Kabo et al . 2014 , Fayard and Weeks 2007 , Kabo et al . 2015 , Crescenzi et al . 2017 ) , which is subject to selection bias from individuals , teams , and organizations choosing where to locate . This makes it challenging to isolate the e\ufb00ect of colo - cation from confounders and other forms of proximity , such as proximity in social space and in knowledge space . Knowledge production , moreover , is increas - ingly a collaborative process ( Wuchty et al . 2007 , Jones et al . 2008 ) between colocated and geographically dis - persed teams of scientist ( Adams et al . 2005 , Katz 1994 , Freeman et al . 2015 ) , making it di\ufb03cult to trace knowl - edge \ufb02ows through these alternative channels . Scientists are evaluated based on the quality and quantity of their output ( e . g . , in terms of scienti\ufb01c impact measured through citations , publications , out - lets ) . Two key costs shape how scientists ( and teams of scientists ) make collaboration decisions : search costs and execution costs . Search costs , by de\ufb01ning the choice set of possible collaborators , in\ufb02uence the like - lihood that any two individuals will explore a joint research project to begin with . Execution costs deter - mine if , given an idea of a certain quality , it makes sense for a speci\ufb01c team to invest time , resources , and e\ufb00ort in developing it . Colocation has a profound e\ufb00ect on both costs : ( 1 ) by increasing the chance of an interaction ( both serendipi - tous and planned ) , colocation drastically lowers search costs for new collaborators ; ( 2 ) by lowering the cost of face - to - face meetings , coordination costs , monitoring costs , and the cost of transferring complex informa - tion , colocation also reduces joint execution costs . If by applying e\ufb00ort toward a project scientists are able to improve its ultimate value , then lower execution costs , locally , endogenously change the optimal level of e\ufb00ort scientists may want to apply to local versus distant projects . In the next sections , I focus on how geographic prox - imity in\ufb02uences search and execution costs to build predictions about how microgeography a\ufb00ects the rate and type of scienti\ufb01c collaborations between teams of scientists ( labs ) . 2 . 1 . Search Costs In this context , search costs are de\ufb01ned as the frictions scientists incur in \ufb01nding new collaborators and col - laboration opportunities . A reduction in search costs should therefore induce an increase in the probability of collaboration between two labs and possibly shift over time the collaboration portfolios of the scientists a\ufb00ected . Recent experimental evidence has shown that even within the same university , search frictions can be substantial : after randomly colocating scientists in the same room for a 90 - minute information - sharing ses - sion , Boudreau et al . ( 2017 ) observe a 75 % increase in the probability that they will co - apply for a grant . An opposite increase in search costs , instead , may have a more ambiguous e\ufb00ect on the probability of collaboration , especially if we believe that once two scientists are aware of each other\u2019s research agenda , they can keep communicating new ideas cheaply over distance\u2014i . e . , when it comes to search , if social prox - imity ( e . g . , past collaboration ) can partially substitute Catalini : Microgeography and the Direction of Inventive Activity Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) 4351 for geographic proximity , then separating groups of scientists that were previously colocated may have a smaller e\ufb00ect ( or no e\ufb00ect at all ) on collaboration . 4 Search also takes place through temporary colo - cation ( e . g . , conferences ) , 5 and through codi\ufb01ed information ( e . g . , published research ) . As a result , in cases where search costs are likely to be already low ex ante because scientists attend the same conferences or read the same journals , changes in geographic proxim - ity may have a less pronounced e\ufb00ect on the likelihood of collaboration . This can be summarized in the following hypothesis : Hypothesis 1 . After colocation , scientists should be more likely to collaborate , and the e\ufb00ect will be stronger the higher search costs are ex ante . Separation , instead , should have no e\ufb00ect on the probability of collaboration . 2 . 2 . Execution Costs Joint execution costs include the cost of face - to - face meetings , coordination , monitoring , and transfer of complex information between teams of scientists . The \ufb01rst e\ufb00ect of a reduction in joint execution costs is , as in the case of search costs , an increase in the probability of collaboration ( this time possibly skewed toward preex - isting pairs ) . After colocation , scientists should collab - orate together more . 6 At the same time , while search costs may not increase immediately after separation because of the long - term bene\ufb01ts from past exposure , execution costs should increase relatively quickly , neg - atively a\ufb00ecting joint projects . To summarize : Hypothesis 2 . After colocation , scientists should be more likely to collaborate with each other ( and increase their rate of collaboration ) because of lower execution costs . After sep - aration , the opposite should be observed . Whereas the e\ufb00ect of a decrease in execution costs on the rate of collaboration is straightforward , the e\ufb00ect on the realized value of the resulting projects is ambigu - ous . If the value of a project depends both on the orig - inal idea\u2019s intrinsic quality as well as on the amount of e\ufb00ort dedicated to advancing it , then colocation may induce : ( 1 ) the development of lower - quality ideas ( a selection e\ufb00ect ) ; ( 2 ) the application of e\ufb00ort for any given idea quality ( an e\ufb00ort e\ufb00ect ) . In other words , if we assume that research projects improve when more e\ufb00ort is allocated to their development , and that it is cheaper to do so when colocated ( lower execution costs ) , then the ultimate impact of colocation on out - comes is the composition of the selection and e\ufb00ort e\ufb00ect pushing in opposite directions . On the one hand , lower execution costs locally may induce scientists to engage in marginal , lower - value projects . Intuitively , if projects below a certain value cannot be published ( i . e . , if there is a minimum thresh - old for publication ) and e\ufb00ort improves the payo\ufb00 of an idea , then the threshold for developing a research project under colocation will be lower . Conditional on an idea being developed , this should lead to an increase on the left tail of the outcome distribution . On the other hand , if scientists can improve a project more e\ufb03ciently when colocated ( e . g . , through addi - tional face - to - face meetings , better knowledge trans - fer , team coordination , etc . ) , this leads to the expected value of colocated projects being higher for any under - lying idea quality . Intuitively , scientists can achieve the same outcome with ideas of lower starting qual - ity because they can execute on them more e\ufb03ciently . Additionally , it can be shown 7 that if the distribution of idea quality is skewed toward the low end ( i . e . , if most scienti\ufb01c ideas are of low impact and a few are of very high impact ) , then the di\ufb00erence in e\ufb00ort is a ( weakly ) increasing function of idea quality ( i . e . , the di\ufb00erence in e\ufb00ort between colocated and distant projects increases with project quality ) . This leads to the following hypothesis : Hypothesis 3 . After colocation , conditional on projects being developed , we should observe more projects at the low end of the distribution ( selection e\ufb00ect ) and at the high end of the distribution ( e\ufb00ort e\ufb00ect ) . Separation should lead to opposite results . The above prediction relies on the skewed nature of inventive outcomes , on the idea that colocation o\ufb00ers lower execution costs , and on the assumption that more e\ufb00ort improves the value of a project . Empirically , a key challenge is due to the fact that whenever scientists have unrealistic expectations about the value of a project ( e . g . , when they believe a project is of publishable quality when it is not ) , or whenever projects are abandoned as scientists update their pri - ors , part of the left tail of the outcome distribution will not be observed ( as we only see published research ) , leading to underestimating the e\ufb00ects of colocation and separation on lower - quality ideas . This limits what we can learn from bibliometrics data regarding the full distribution of research outcomes . The analysis on quality in Section 4 . 3 will be therefore more informa - tive about right - tail outcomes and only provide sug - gestive evidence about the left tail . Moreover , because of lower search and execution costs , colocation is a strategic choice , and observa - tional data will overestimate the impact of these e\ufb00ects on inventive outcomes . The relocations of labs on the Jussieu campus , because of the external constraints imposed on space allocation by the asbestos removal process , can help us understand if the mechanisms suggested in the theoretical framework are correct , and improve our understanding of the conditions under which colocation is more likely to matter for collabo - rative inventive outcomes . To address additional con - cerns about selection into colocation and matching between labs , the key part of the empirical analysis Catalini : Microgeography and the Direction of Inventive Activity 4352 Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) relies on a di\ufb00erence - in - di\ufb00erences approach at the lab - pair level , as described in the next section . Lab - level portfolio choices are also presented to understand the aggregate e\ufb00ects of the moves on the trade - o\ufb00s labs make in deciding who to collaborate with . 3 . Empirical Setting The setting is the university campus of Paris Jussieu , which hosts the Universit\u00e9 Pierre et Marie Curie ( UPMC ) . According to the 2016 U . S . News and World Report rankings , UPMC is the top institution in France ( 10th in Europe , 49th on a global scale ) . It houses the faculty of sciences of Sorbonne Universit\u00e9s and approximately 31 , 000 students ( 5 , 900 master students and 3 , 000 doctoral students ) , as well as 3 , 750 research - active professors ( 80 % of its sta\ufb00 work in research cen - ters ) . 8 Three recent Nobel laureates are from UPMC : Pierre - Gilles de Gennes ( Physics , 1991 ) , Claude Cohen - Tannoudji ( Physics , 1997 ) , and Fran\u00e7oise Barr\u00e9 Sinoussi ( Medicine , 2008 ) . Strong areas of specialization include mathematics ( \ufb01fth worldwide ) , physics ( 15th ) , space science ( 20th ) , geosciences ( 21st ) , neuroscience ( 43rd ) , environment and ecology ( 49th ) , and biology and bio - chemistry ( 51st ) . The campus went through \ufb01ve massive waves of labs relocations over 17 years ( 1997 \u2013 2014 ) due to the removal of asbestos from its buildings . The moves started when the French government , reacting to a research report by INSERM on the carcinogenic e\ufb00ects of asbestos ( June 21 , 1996 ) , 9 introduced a full ban of the \ufb01re retardant material from public buildings . The Jussieu campus was built extensively using asbestos . Interviews with scientists from the labs con\ufb01rmed that given the nature and urgency 10 of the clean - ing process , labs were forced to move often under short notice and with minimal in\ufb02uence over their new location . A separate entity ( EPAURIF ) was put in charge of the cleaning process , which started with labs that were relatively easy to move ( e . g . , theoreti - cal labs in mathematics , computer science ) and only later reached labs with sophisticated instrumentation and machinery ( e . g . , applied physics ) . During the relo - cation period , entire sections of the campus were pro - gressively isolated and renovated . Because of the com - plexity and costs of the operation , lab requirements were often not a priority , resulting in many labs com - plaining about the moves . Whereas for some scientists the actual moving process was a cause of delays ( one scientist estimated a one - year delay in productivity over a 10 - year period ) , others did not \ufb01nd it disruptive at all ( a di\ufb00erent scientist , who does theoretical work , estimated just one week without lab access ) . As in other research - intensive institutions , interac - tions across labs are common , and the relocations ended up separating labs that were colocated and inter - acted before . The same moves also placed labs next to new neighbors , in some cases with positive e\ufb00ects on collaboration . In the data set , the aggregate amount of colocation between lab pairs slightly decreases during the moves , from 7 . 3 % to 6 . 9 % of pairs . 11 Whereas the mean distance between lab pairs across broad \ufb01elds of science ( e . g . , natural versus life sciences ) and within sub\ufb01elds ( e . g . , within chemistry ) is broadly stable over time , the distance across sub\ufb01elds ( e . g . , chemistry and physics ) decreases , providing plausibly exogenous variation in the composition of a lab\u2019s neighbors . The data also provides substantial variation in terms of which types of lab pairs are a\ufb00ected by the moves at any point in time . 12 The mean number of collaborations per year for colocated lab pairs increases from 0 . 034 during the pre - period to 0 . 042 during the moves . The change is substantially larger for pairs that are within a broad \ufb01eld , from 0 . 039 to 0 . 056 . An academic campus is an appealing environment for studying the role of geographic proximity on col - laboration : knowledge production is one of its de\ufb01n - ing activities , it is possible to measure collaborations and their long - run impact , knowledge \ufb02ows can be partially captured by looking at cited references , and research agendas can be mapped into knowledge space using keywords and cited references . Moreover , labs increasingly rely on a mix of proximate and distant col - laborators ( both in geographic and knowledge space ) for advancing their research agendas . 3 . 1 . Data The data set combines information on 39 , 527 pub - lications from the labs at Jussieu ( 1980 \u2013 2010 ) with \ufb01ne - grained location data over time . Publications are retrieved from SciVerse Scopus 13 and parsed to extract a\ufb03liation data . Forty - two thousand , four hundred ninety - four unique a\ufb03liation strings are cleaned and harmonized with a series of algorithmic and manual steps to match them to a speci\ufb01c lab . 14 Whenever loca - tion data is available in the papers 15 it is extracted to complement information retrieved from the UPMC archives and the EPAURIF website 16 to reconstruct the spatial allocation of labs over time . Paper a\ufb03liations that are not matched to the campus are geocoded using a combination of three di\ufb00erent services ( Google Maps API , Bing Maps API , and the Data Science Toolkit ) to identify them as either French or interna - tional a\ufb03liations . The core of the campus resembles a chessboard ( see Online Appendix Figure A - 1 ) and is composed of a series of towers connected by corridor buildings . Dis - tances are obtained by manually geocoding the loca - tion of each tower and connecting building on Google Earth . 17 Out of 39 , 527 publications , 6 % are collaborations across di\ufb00erent labs at Jussieu . In the \ufb01nal data set ( see Table 1 ) , the average minimum distance between any Catalini : Microgeography and the Direction of Inventive Activity Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) 4353 Table 1 . Summary Statistics for the Main Sample Lab - pair - year level Mean Std . dev . Min Max N LocationMinimum distance ( km ) 0 . 17 0 . 106 0 0 . 449 183 , 359 Colocated ( same building ) 0 . 072 0 . 258 0 1 295 , 435 Treatment year ( colocation ) 2 , 003 . 99 3 . 722 1 , 997 2 , 010 7 , 338 Treatment year ( separation ) 2 , 003 . 63 4 . 089 1 , 997 2 , 010 5 , 877 CollaborationProbabilityof collaboration 0 . 004 0 . 064 0 1 295 , 435 Number of collaborations 0 . 006 0 . 126 0 15 295 , 35 Quality and type of research Citation weighted collaborations 0 . 204 7 . 328 0 2 , 058 295 , 435 Maximum number of citations 0 . 164 6 . 053 0 2 , 058 295 , 435 Standard deviation of citations a 25 . 666 38 . 964 0 271 . 442 321 Keyword similarity 0 . 076 0 . 108 0 0 . 881 116 , 851 Cited references similarity 0 . 174 0 . 115 0 0 . 677 84 , 865 a Conditional on collaborating at least twice in focal year . two lab pairs is 0 . 17 km ( approximately 550 feet ) and 7 . 2 % of them are colocated at any point in time ( same tower or corridor building ) . 18 Over time , 383 lab pairs switch from not colocated to being in the same build - ing during the asbestos cleaning process ( at a rate of 16 \u2013 52 each year ) , with 37 % of the changes happening in the \ufb01rst \ufb01ve years . Two hundred forty - eight lab pairs experience the opposite change ( i . e . , are separated by the moves ) , with 34 % of the events taking place within \ufb01ve years . Collaboration across labs is a rare event , and only 0 . 4 % of all lab pairs collaborate in any given year , receiving on average 0 . 20 citations per lab - pair - year . As often observed , the distribution of citations is very skewed , with most papers receiving no citations , and the most productive lab pair totaling 2 , 058 citations from the papers published in a single year . Citation data is obtained from Scopus in 2016 , and quartiles for the citation distribution are built by year using a large sample of articles in the relevant \ufb01elds of science . Scopus data is also used to retrieve , whenever available , author and index keywords , 19 as well as the full set of references cited by each paper . Cited references and keyword data are then used to create measures of proximity in knowledge space between the labs . In particular , the cosine similar - ity between vectors of cited references ( or keywords ) used by each lab is calculated using the Scikit - Learn python module . 20 To ensure that these proximity mea - sures are not in\ufb02ated by direct collaborations ( which would share all cited references and keywords by design ) , coauthored papers between the focal labs are dropped\u2014i . e . , the vectors of cited references ( or key - words ) are based on independent publications . More - over , when the sample is split by lab pairs with above versus below the median distance in knowledge space ( e . g . , Table 7 ) , the measure is de\ufb01ned before the moves start to avoid it being in\ufb02uenced by the e\ufb00ect of colo - cation ( or separation ) . In the data set , labs that do not share any keyword have a cosine similarity of zero , the pair with the most overlap has a score of 0 . 88 , 21 and the mean keyword similarity is 0 . 08 . Similarity in cited references is on average higher ( 0 . 17 ) , but the highest observed score is lower ( 0 . 68 ) . 3 . 2 . Empirical Strategy Estimating the e\ufb00ect of colocation on inventive out - comes using observational data is likely to return biased results , the main endogeneity concern being a selection e\ufb00ect . If labs value proximity to other labs they want to collaborate with , then a basic ordinary least - squares ( OLS ) regression of the number of collab - orations on geographic distance will positively bias the e\ufb00ect of proximity on collaboration . Whereas colocated lab pairs were 2 . 75 times more likely to collaborate than not - colocated pairs before the moves started , 22 the premium is only 14 % ( and not signi\ufb01cant ) during the relocations , which is consistent with the new spatial allocation being suboptimal in the post - period . In a perfect experiment , we would randomize the location of labs and observe how collaborations between colocated pairs di\ufb00er from those between dis - tant ones . During the asbestos removal , qualitative evi - dence con\ufb01rmed that needs that were orthogonal to the research agendas of the scientists involved ( e . g . , ease of relocation of a type of lab , cost minimization for the removal operation ) constrained space assign - ment and shaped when and where di\ufb00erent labs were moved . This introduces exogenous variation in the set of neighbors a lab is o\ufb00ered . 23 To account for idiosyn - cratic reasons a lab pair may be more or less likely to collaborate in the \ufb01rst place , the analysis uses lab - pair \ufb01xed e\ufb00ects . Importantly , the pair \ufb01xed e\ufb00ects also capture the degree of in\ufb02uence a particular pair of labs may have on campus resource allocation decisions ( funding , personnel , space allocation ) . Catalini : Microgeography and the Direction of Inventive Activity 4354 Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) The econometric analysis uses a di\ufb00erence - in - dif - ferences approach 24 at the lab - pair - year level which exploits the variation in colocation ( and separation ) generated by the moves that take place during the asbestos - removal period . I estimate variations of Y ijt (cid:3) \u03b3 ij + \u03b4 t + \u03b2 AfterColocation ijt + (cid:15) ijt , ( 1 ) where Y ijt is a dummy for collaboration in focal year 25 between lab i and lab j in year t ; \u03b2 is the coe\ufb03cient of interest and the lab - pair \ufb01xed e\ufb00ects , \u03b3 ij , mean that \u03b2 is identi\ufb01ed as the within - pair e\ufb00ect on collaboration after labs become colocated because of the asbestos removal , relative to their before period ; and \u03b4 t is a year e\ufb00ect . The AfterColocation dummy is equal to 1 if the lab pair becomes colocated because of the moves , and is set back to 0 when the labs are separated again . Similarly , I estimate the e\ufb00ect of separation in the same regression using variations of Y ijt (cid:3) \u03b3 ij + \u03b4 t + \u03b2 1 AfterColocation ijt + \u03b2 2 AfterSeparation ijt + (cid:15) ijt . ( 2 ) Both equations are also estimated with nonlinear models and for di\ufb00erent outcomes : number of col - laborations in focal year , maximum number of cita - tions received ( conditional on publication ) , number of papers published in di\ufb00erent quartiles of the citation distribution , similarity in cited references space and in keyword space , probability of publishing in a journal that is new for at least one of the labs , etc . Before moving to the di\ufb00erence - in - di\ufb00erences frame - work , the paper uses regressions at the lab - year level for the same set of outcomes to explore how the relo - cation period in\ufb02uenced collaboration and the compo - sition of the labs\u2019 portfolios ( in terms of share of a\ufb03lia - tions classi\ufb01ed as same lab , di\ufb00erent lab , other French lab or international lab , etc . ) . All lab - year \u2013 level regres - sions include lab \ufb01xed e\ufb00ects and year \ufb01xed e\ufb00ects . 4 . Results The \ufb01rst section of the results focuses on assessing the overall e\ufb00ect of the moves on the labs involved . Empirically , it is important to determine if the relo - cation period had a negative e\ufb00ect on the labs\u2019 pro - ductivity , and how labs adjusted their collaboration portfolios because of the changes in their neighbors . Since this part of the analysis does not rely on the di\ufb00erence - in - di\ufb00erences empirical strategy described above , results will only be informative but not conclu - sive with respect to the mechanisms proposed in the theoretical framework . The second section will directly test Hypothesis 1 on the e\ufb00ects of a change in proximity on the probability of collaboration . This is done by looking at pairs of labs that \ufb01nd themselves in the same building because of the asbestos removal , as well as pairs that are separated because of it . In particular , separated pairs will be infor - mative about the ability of other forms of proximity ( e . g . , social proximity ) to compensate for geographic distance . Robustness is shown to highlight the absence of a pre - trend in collaboration among pairs that are going to be colocated , which is consistent with the moves not being driven by the labs\u2019 research agendas . Knowledge distance between labs is then introduced as a way to compare lab pairs with di\ufb00erent ex ante search costs , and to see if the changes in the proba - bility of collaboration are consistent with the search cost mechanism . Since Hypotheses 1 and 2 do not dif - fer in terms of what they predict we should observe after colocation , but di\ufb00er in terms of what we should see after separation , this section will also compare the relative role of search versus execution costs in de\ufb01ning who collaborates . Lastly , the section explores the e\ufb00ect of search costs over longer periods of time by using information on the references cited in the a\ufb00ected papers . The third section introduces quality ( as proxied by citations ) to test Hypothesis 3 , and to see if outcomes are consistent with lower execution costs under coloca - tion leading to both more marginal ideas being devel - oped ( selection e\ufb00ect ) , as well as higher - quality ideas ( e\ufb00ort e\ufb00ect ) . 4 . 1 . Lab - Level Results and Collaboration Portfolios One may worry that the relocations on the Jussieu campus had a negative e\ufb00ect on the productivity of the labs . This would in\ufb02uence how we would inter - pret the results from the di\ufb00erence - in - di\ufb00erences anal - ysis in the next sections of the paper . Furthermore , since labs are likely to reallocate their resources toward di\ufb00erent types of projects as their local environment changes , before moving to the lab - pair \u2013 level regres - sions , it is useful to descriptively explore how aggre - gate collaboration portfolios shifted during the study period because of the moves . The unit of analysis in Table 2 is a lab - year , lab \ufb01xed e\ufb00ects are included to account for unobservable dif - ferences between labs that are constant over time ( e . g . , \ufb01eld of science , relative scale and resources of the lab within the institution , etc . ) , and year \ufb01xed e\ufb00ects are introduced to control for changes in productivity over time ( e . g . , increased resources available to the campus , changes in the national science policy , etc . ) . 26 The key explanatory variable in the table is a dummy equal to 1 during the Relocation Period \u2014i . e . , during all of the years during which a speci\ufb01c lab is moved from its original location because of the asbestos removal ( and 0 other - wise ) . As can be seen both in terms of raw publication counts ( column ( 1 ) ) , and in terms of quality - adjusted output ( column ( 2 ) ) , the moves are not correlated with a decrease in output for the a\ufb00ected labs relative to Catalini : Microgeography and the Direction of Inventive Activity Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) 4355 Table 2 . Lab - Level Outcomes ( 1 ) ( 2 ) ( 3 ) ( 4 ) ( 5 ) Variables Publications Citation weighted pubs Mean cites Max cites Std . dev . cites Relocation Period 10 . 8003 \u2217\u2217\u2217 246 . 6649 \u2217\u2217\u2217 \u2212 0 . 0157 28 . 0384 \u2217 6 . 4865 \u2217\u2217 ( 2 . 3118 ) ( 82 . 7443 ) ( 1 . 6609 ) ( 14 . 6638 ) ( 2 . 6019 ) Lab \ufb01xed e\ufb00ects Yes Yes Yes Yes Yes Year \ufb01xed e\ufb00ects Yes Yes Yes Yes Yes Observations 3 , 951 3 , 951 3 , 951 3 , 951 3 , 951 R - squared 0 . 161 0 . 070 0 . 015 0 . 015 0 . 011 Number of labs 328 328 328 328 328 Notes . Relocation Period is a dummy equal to 1 for all of the years during which a speci\ufb01c lab is moved from its original location because of the asbestos removal . Citation data is collected from Scopus in 2016 . Robust standard errors clustered at the lab level in parentheses . \u2217 p < 0 . 1 ; \u2217\u2217 p < 0 . 05 ; \u2217\u2217\u2217 p < 0 . 01 . the controls ( i . e . , labs that have not been moved yet or that have returned to their original location ) . Interest - ingly , results on quality are suggestive of an increase on both tails of the outcome distribution : the e\ufb00ect on mean citations is negative although nonsigni\ufb01cant ( col - umn ( 3 ) ) , max citations increase ( column ( 4 ) ) , and the moves are correlated with a positive increase in the standard deviation of citations ( column ( 5 ) ) . It is important to highlight that multiple mecha - nisms , including changes in neighboring labs , infras - tructure , and research type , can explain these results . For example , labs could have changed the composi - tion of their projects in response to shifts in their local environment in a way that favored higher - variance projects\u2014e . g . , if the moves increased the likelihood that a lab had access to peers from a di\ufb00erent disci - pline , the polarization in outcomes could be a result of an increase in cross - disciplinary work . This would be consistent with past research that has shown that recombination of ideas that span more distant areas of the knowledge space exhibit higher variance ( e . g . , because the ideas involved are recombined less often and therefore represent a less explored area of the research landscape ; see Fleming and Sorenson 2004 , Singh and Fleming 2010 ) . In this context , while keep - ing the expected value of their projects constant ( i . e . , while staying on their original risk \u2013 return indi\ufb00erence curve ) , labs could have undertaken some projects that o\ufb00ered higher risk but also higher reward . Alterna - tively , the moves could have induced a temporary , one - time shift in research agendas because of the reshuf - \ufb02ing of scientists and labs ( a \u201cnovelty\u201d e\ufb00ect ) . For these reasons , the next section will move away from the lab - level analysis and control for many of these confounders by looking at variation within lab pairs . With the introduction of lab - pair \ufb01xed e\ufb00ects , the di\ufb00erence - in - di\ufb00erences approach allows to account for the idiosyncratic reasons two labs may ( or may not ) collaborate with each other , and controls for unob - served heterogeneity in the type of research a particu - lar pair may be conducting . Within a lab pair , the con - straints imposed by the moves will deliver plausibly exogenous variation in distance , allowing us to look at outcomes while keeping issues related to the matching between labs constant\u2014i . e . , within a lab pair , changes in collaboration and outcomes will be predominantly driven by shifts in proximity . Additionally , the analysis of pre - trends in collaboration within the di\ufb00erence - in - di\ufb00erences framework will allow us to visually inves - tigate the presence of selection into colocation during the moves\u2014i . e . , if labs are endogenously paired in an e\ufb00ort to improve their collaborations , this should be observable in the data . Before moving to the lab - pair analysis , it is useful to check how the relocation period shifted the collab - oration portfolios of the labs at the aggregate level . Table 3 uses the same speci\ufb01cation of the previous table to look at how the shares of a\ufb03liations on the labs\u2019 papers changed during the moves . The dependent variables here are , respectively , the share of a\ufb03liations that are international ( column ( 1 ) ) , within France but not from the institution ( column ( 2 ) ) , and within the institution ( column ( 3 ) ) . Interestingly , the moves saw a decrease in the share of international collaborators ; a Table 3 . Lab Level\u2014Overall Collaboration Portfolio ( 1 ) ( 2 ) ( 3 ) Share Share Share Variables international French within institution Relocation Period \u2212 0 . 0237 \u2217 0 . 0018 0 . 0218 ( 0 . 0121 ) ( 0 . 0137 ) ( 0 . 0146 ) Lab \ufb01xed e\ufb00ects Yes Yes Yes Year \ufb01xed e\ufb00ects Yes Yes Yes Observations 3 , 951 3 , 951 3 , 951 R - squared 0 . 144 0 . 072 0 . 207 Number of labs 328 328 328 Notes . Relocation Period is a dummy equal to 1 for all of the years dur - ing which a speci\ufb01c lab is moved from its original location because of the asbestos removal . Shares are calculated based on the Scopus author a\ufb03liations data available for each paper and then aggregated at the lab level on a yearly basis . \u201cShare French\u201d does not include the Jussieu campus and a\ufb03liated labs . Robust standard errors clustered at the lab level in parentheses . \u2217 p < 0 . 1 . Catalini : Microgeography and the Direction of Inventive Activity 4356 Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) Table 4 . Lab Level\u2014Collaboration Portfolio Within Institution ( 6 ) ( 7 ) ( 8 ) ( 9 ) ( 1 ) ( 3 ) ( 4 ) ( 5 ) Colocated Colocated Not colocated Not colocated Share ( 2 ) Not Intensive Extensive intensive extensive intensive extensive Variables across labs Colocated colocated margin margin margin margin margin margin Relocation Period \u2212 0 . 0210 \u2217\u2217 0 . 0155 \u2217\u2217\u2217 \u2212 0 . 0365 \u2217\u2217\u2217 \u2212 0 . 0035 \u2212 0 . 0175 \u2217\u2217 0 . 0030 0 . 0125 \u2217\u2217\u2217 \u2212 0 . 0065 \u2212 0 . 0300 \u2217\u2217\u2217 ( 0 . 0093 ) ( 0 . 0048 ) ( 0 . 0082 ) ( 0 . 0066 ) ( 0 . 0075 ) ( 0 . 0037 ) ( 0 . 0031 ) ( 0 . 0051 ) ( 0 . 0075 ) Lab \ufb01xed e\ufb00ects Yes Yes Yes Yes Yes Yes Yes Yes Yes Year \ufb01xed e\ufb00ects Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 3 , 951 3 , 951 3 , 951 3 , 951 3 , 951 3 , 951 3 , 951 3 , 951 3 , 951 R - squared 0 . 022 0 . 023 0 . 034 0 . 065 0 . 130 0 . 036 0 . 035 0 . 041 0 . 119 Number of labs 328 328 328 328 328 328 328 328 328 Notes . Relocation Period is a dummy equal to 1 for all of the years during which a speci\ufb01c lab is moved from its original location because of the asbestos removal . All shares are calculated based on the Scopus author a\ufb03liations data available for each paper and then aggregated at the lab level on a yearly basis . The table only includes a\ufb03liations that can be assigned to the Jussieu campus and a\ufb03liated labs . Robust standard errors clustered at the lab level in parentheses . \u2217\u2217 p < 0 . 05 , \u2217\u2217\u2217 p < 0 . 01 . nonsigni\ufb01cant , small positive change in collaborators from other French institutions ; and an increase in within - institution research . While the regression con - trols for lab \ufb01xed e\ufb00ects and year \ufb01xed e\ufb00ects , as in the previous table multiple mechanisms could explain these changes . One interpretation is that the campus lost competitiveness because of the moves , becoming more inward focused . A di\ufb00erent one is that the reshuf - \ufb02ing improved local opportunities by o\ufb00ering labs new neighbors , making outside collaboration relatively less appealing . To further test how internal collaboration was af - fected , Table 4 uses the same approach of Table 3 but only focuses on internal collaborations ( i . e . , all of the shares are now de\ufb01ned over the total number of a\ufb03l - iations assigned to the institution ) . Column ( 1 ) high - lights how the increase in within - institution research observed in column ( 3 ) of Table 3 is entirely driven by collaborations within labs : the share of across - lab collaborations actually decreases , suggesting that the moves may have worsened the local environment , forc - ing labs to increase collaboration within their unit . This is consistent with the relocations leading to a subop - timal space allocation relative to what the labs would have selected if they were in charge of it . The next two pairs of columns in Table 4 , take the share from column ( 1 ) and further decompose it across di\ufb00erent dimensions . In particular , columns ( 2 ) and ( 3 ) split the collaborations across labs ( i . e . , column ( 1 ) ) between labs that are colocated versus labs that are distant : the negative result from the \ufb01rst column comes from a drastic decrease in the share of collaborators from not colocated labs ( column ( 3 ) ) that is not compensated by an equal increase in collaborations with colocated labs ( column ( 2 ) ) . When column ( 1 ) is instead decom - posed into intensive margin versus extensive margin collaborations ( i . e . , between labs that had collaborated before the relocations versus not ) , the data show that the decay in across - lab collaborations is mostly driven by extensive margin pairs\u2014i . e . , during the relocations , labs were substantially less likely to explore research with labs they had not collaborated with before . The aggregate result of columns ( 4 ) and ( 5 ) , however , hide di\ufb00erent heterogeneous e\ufb00ects by microgeogra - phy : in columns ( 6 ) \u2013 ( 9 ) , the share of column ( 1 ) is divided into colocated ( intensive versus extensive ) and not colocated ( intensive versus extensive ) collabora - tions . Whereas the results are consistent with geogra - phy only slightly facilitating ( or obstructing ) collabora - tive work on the intensive margin ( the coe\ufb03cients are respectively positive in column ( 6 ) and negative in col - umn ( 8 ) , but in both cases not signi\ufb01cant ) , columns ( 7 ) and ( 8 ) suggest that microgeography plays a substan - tially more important role on the extensive margin of collaboration . This seems consistent with search costs having a disproportionate e\ufb00ect on de\ufb01ning who col - laborates with whom in the absence of past exposure or social proximity . This hypothesis will be further tested in the next section . Overall , at least in the Jussieu case , the increase in collaboration with newly colocated pairs ( columns ( 6 ) and ( 7 ) ) seems to be more than o\ufb00set by the decrease with not colocated labs ( columns ( 8 ) and ( 9 ) ) . Fur - thermore , the result in column ( 1 ) ( combined with the e\ufb00ect of column ( 3 ) in Table 3 ) is consistent with labs being more inward focused during the moves , potentially because they faced a less optimal set of local peers around them relative to what they would have selected in an ideal scenario . Together with the observed increased in best outcomes and variance in Table 2 , and with the rise in colocated experimentation on the extensive margin ( column ( 7 ) in Table 4 ) , this raises the question of how colocation and separation directly a\ufb00ected research outcomes once we keep lab pairs constant . Catalini : Microgeography and the Direction of Inventive Activity Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) 4357 4 . 2 . Probability of Collaboration and the Role of Search Costs In the theoretical framework , both a reduction in search costs ( Hypothesis 1 ) and in joint execution costs ( Hypothesis 2 ) predict an increase in the probability of collaboration after colocation . If joint execution costs are particularly important , then we should also observe a tangible rise in the rate of collaboration . The two mechanisms have di\ufb00erent implications for the reverse move though\u2014i . e . , for what we should observe after two labs that were previously colocated are suddenly separated . If search costs do not increase immediately after separation because of past exposure , then sepa - ration should have little to no e\ufb00ect on the probability of collaboration . If joint execution costs are instead dis - proportionately driving collaboration decisions , then separation should be accompanied with a drop in col - laboration . This section tests these hypotheses by \ufb01rst looking at within - lab - pair changes in collaboration fol - lowing changes in distance , and then by exploring het - erogeneous e\ufb00ects by pairs that faced ex ante higher versus lower search costs to compare the two mecha - nisms more directly . The \ufb01rst part of Hypotheses 1 and 2 are tested in Table 5 : after colocation , lab pairs are 3 . 5 times more likely to collaborate ( column ( 1 ) ) and collaborate on average 2 . 5 \u2013 3 . 3 times more ( columns ( 3 ) and ( 2 ) , respectively ) . Results are robust and qualitatively sim - ilar across di\ufb00erent functional forms ( OLS , Poisson , logit , rare event logit ) . Robust standard errors are clus - tered at the lab - pair level , and all regressions include lab - pair \ufb01xed e\ufb00ects as well as year \ufb01xed e\ufb00ects . The estimated e\ufb00ects do not change if lab - year trends are included in the regression . The \ufb01ndings are consistent with lower search and execution costs increasing the Table 5 . Increase in Collaboration and Colocation ( 1 ) ( 2 ) ( 3 ) Variables OLS 1 / 0 OLS # Collabs Poisson # Collabs After Colocation 0 . 0146 \u2217\u2217\u2217 0 . 0210 \u2217\u2217 0 . 8981 \u2217\u2217\u2217 ( 0 . 0054 ) ( 0 . 0094 ) ( 0 . 2814 ) Lab - pair \ufb01xed e\ufb00ects Yes Yes Yes Year \ufb01xed e\ufb00ects Yes Yes Yes Observations 295 , 435 295 , 435 10 , 383 R - squared 0 . 005 0 . 005 Number of lab pairs 35 , 805 35 , 805 587 Log likelihood \u2212 2 , 929 Notes . The dependent variable in column ( 1 ) is a dummy equal to 1 if the lab pair collaborates in the focal year . The dependent variable in columns ( 2 ) and ( 3 ) is the number of collaborations . After Colocation is equal to 1 when a lab pair becomes colocated because of the moves and 0 otherwise . The Poisson speci\ufb01cation with \ufb01xed e\ufb00ects drops all lab pairs where collaboration is never observed , hence the smaller number of observations . Robust standard errors clustered at the lab - pair level in parentheses . \u2217\u2217 p < 0 . 05 ; \u2217\u2217\u2217 p < 0 . 01 . likelihood ( column ( 1 ) ) and rate ( columns ( 2 ) and ( 3 ) ) of collaboration . It is important to stress that the presence of the lab - pair \ufb01xed e\ufb00ects takes care of the idiosyncratic in\ufb02u - ence a pair of labs may have on campus ( e . g . , in terms of funding , ability to bargain for more or better resources and infrastructure ) . Nevertheless , one may still worry that in\ufb02uential labs might have exerted pressure on EPAURIF to assign them a particular temporary loca - tion , or to change the timing of their move . From an identi\ufb01cation perspective , the main worry is that the moves were driven by preexisting collaboration pat - terns and therefore endogenous to the outcomes of interest . If that were the case , and labs were able to obtain a spot next to the labs with which they had an interest in collaborating more , then we should observe a rise in collaboration that predates the relocation . Fig - ure 1 reassures us this is not the case , as the rise in col - laboration follows colocation ( no pre - trend ) and builds progressively over the years in the post - period . In fact , there is no activity until two years or more after the move . The \ufb01gure plots the estimated coe\ufb03cient of an OLS regression with year and lab - pair \ufb01xed e\ufb00ects for all of the years before and after the move . 27 The dependent variable is the probability of collaboration ( dummy ) , and the error bars represent 95 % con\ufb01dence intervals based on robust standard errors ( clustered at the lab - pair level ) . 28 Past exposure may allow labs to contact a now - distant lab when the right opportunity emerges or , if opportu - nities have already been identi\ufb01ed , labs can make com - mitments to keep their research active through joint seminars , temporary colocation , and distant interac - tions . Therefore , if search costs are a key friction to Figure 1 . ( Color online ) Probability of Collaboration and Colocation \u2013 0 . 02 0 0 . 02 0 . 04 0 . 06 0 . 08 \u2013 10 \u2013 8 \u2013 6 \u2013 4 \u2013 2 0 2 4 6 Years to / Since move Notes . Estimated coe\ufb03cient for years before and after the move . The dependent variable is collaboration ( 1 / 0 ) . Regression includes lab - pair \ufb01xed e\ufb00ects and year \ufb01xed e\ufb00ects . Error bars represent 95 % con\ufb01dence intervals based on robust standard errors clustered at the lab - pair level . Catalini : Microgeography and the Direction of Inventive Activity 4358 Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) Table 6 . Colocation and Separation ( 1 ) ( 2 ) ( 3 ) Variables OLS 1 / 0 OLS # Collabs Poisson # Collabs After Colocation 0 . 0148 \u2217\u2217\u2217 0 . 0217 \u2217\u2217 0 . 8871 \u2217\u2217\u2217 ( 0 . 0054 ) ( 0 . 0093 ) ( 0 . 2814 ) After Separation 0 . 0101 0 . 0388 \u2212 0 . 4520 ( 0 . 0089 ) ( 0 . 0278 ) ( 0 . 3171 ) Lab - pair \ufb01xed e\ufb00ects Yes Yes Yes Year \ufb01xed e\ufb00ects Yes Yes Yes Observations 295 , 435 295 , 435 10 , 383 R - squared 0 . 005 0 . 005 Number of lab pairs 35 , 805 35 , 805 587 Log likelihood \u2212 2 , 926 Notes . The dependent variable in column ( 1 ) is a dummy equal to 1 if the lab pair collaborates in the focal year . The dependent vari - able in columns ( 2 ) and ( 3 ) is the number of collaborations . After Colocation is equal to 1 when a lab pair becomes colocated because of the moves and 0 otherwise . After Separation is equal to 1 when a previously colocated pair is separated because of the moves and 0 otherwise . The Poisson speci\ufb01cation with \ufb01xed e\ufb00ects drops all lab pairs where collaboration is never observed , hence the smaller num - ber of observations . Robust standard errors clustered at the lab - pair level in parentheses . \u2217\u2217 p < 0 . 05 ; \u2217\u2217\u2217 p < 0 . 01 . collaboration , separation should not have a strong , neg - ative e\ufb00ect on the probability of collaboration . 29 If joint execution costs constitute a barrier to collaboration instead , separated pairs should reduce their rate of collaboration . Consistent with the search costs mechanism , in Table 6 separation has a noisy , nonsigni\ufb01cant e\ufb00ect on the probability and intensity of collaboration . The co - e\ufb03cient is positive and nonsigni\ufb01cant in the OLS re - gressions ( columns ( 1 ) and ( 2 ) ) and negative and non - signi\ufb01cant in the Poisson speci\ufb01cation ( column ( 3 ) ) . 30 Meanwhile , controlling for separation leaves the results on colocation unchanged . Figure 2 , which plots the estimated regression coef - \ufb01cients 31 for the years before and after separation , highlights three facts : ( 1 ) lab pairs that are going to be separated have a higher propensity to collaborate before the move ( all of the coe\ufb03cients in the pre - period are positive and many are statistically di\ufb00erent than zero ) ; ( 2 ) the relocation process seems to gener - ate a slight , temporary decay in collaboration ( years \u2212 1 to + 1 ) ; ( 3 ) lab pairs revert to their mean collabora - tion propensity in the later periods ( although results are noisy , potentially because some pairs recover and others do not ) . Overall , results from Tables 5 and 6 are consistent with the idea that search costs may be a more impor - tant friction than joint execution costs in de\ufb01ning if two labs will collaborate or not . If the key mechanism through which colocation facilitates collaboration is by helping scientists discover new potential collaborators , then the e\ufb00ect of proximity should be strongest where Figure 2 . ( Color online ) Probability of Collaboration and Separation \u2013 0 . 05 0 0 . 05 0 . 10 \u2013 10 \u2013 8 \u2013 6 \u2013 4 \u2013 2 0 2 4 6 Years to / Since move Notes . Estimated coe\ufb03cient for years before and after the move . The dependent variable is collaboration ( 1 / 0 ) . Regression includes lab - pair \ufb01xed e\ufb00ects and year \ufb01xed e\ufb00ects . Error bars represent 95 % con - \ufb01dence intervals based on robust standard errors clustered at the lab - pair level . search costs are ex ante high and less pronounced where search costs are lower because alternative chan - nels for knowledge di\ufb00usion are likely to exist ( e . g . , joint conferences , journals , etc . ) . To proxy for the fact that two labs may be work - ing in related areas before the moves ( i . e . , within more proximate communities of science ) , Table 7 uses two sources of data : keywords listed on publications and the full set of references cited in the papers . To calculate a measure of proximity in knowledge space between two labs , the cosine distance between the vectors of keywords ( or of references cited ) used by each lab is calculated excluding all direct collaborations ( which would exhibit perfect overlap in keywords and refer - ences by design ) . Lab pairs that have an above - the - median similarity in keywords and references are clas - si\ufb01ed as facing lower search costs ex ante ( columns ( 1 ) and ( 3 ) ) , since they are more likely to overlap in top - ics and the literature they cite ( and potentially con - ferences , etc . ) . Lab pairs that have below - the - median similarity ( columns ( 2 ) and ( 4 ) ) are classi\ufb01ed as expe - riencing higher search costs ex ante , since they do not seem aware of each other\u2019s research topics and body of knowledge . Interestingly , whereas the two classi\ufb01cations assign slightly di\ufb00erent sets of labs to each bin , the inter - pretation of the main e\ufb00ect is consistent between them . When search costs are low ex ante ( columns ( 1 ) and ( 3 ) ) , colocation has a positive but nonsigni\ufb01cant e\ufb00ect on the probability of collaboration . This is consis - tent with these lab pairs already overlapping through other channels and being aware of each other\u2019s work ( as evidenced by the similar keywords and literature used ) . As additional evidence that search costs may Catalini : Microgeography and the Direction of Inventive Activity Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) 4359 Table 7 . Probability of Collaboration and Search Costs ( 1 ) ( 2 ) ( 3 ) ( 4 ) OLS 1 / 0 OLS 1 / 0 OLS 1 / 0 OLS 1 / 0 low search costs high search costs low search costs high search costs Variables ( keywords ) ( keywords ) ( cited ref . ) ( cited ref . ) After Colocation 0 . 0056 0 . 0179 \u2217 0 . 0006 0 . 0145 \u2217 ( 0 . 0126 ) ( 0 . 0093 ) ( 0 . 0268 ) ( 0 . 0077 ) After Separation \u2212 0 . 0197 \u2217 0 . 0244 \u2212 0 . 0080 0 . 0059 ( 0 . 0112 ) ( 0 . 0185 ) ( 0 . 0180 ) ( 0 . 0136 ) Lab - pair \ufb01xed e\ufb00ects Yes Yes Yes Yes Year \ufb01xed e\ufb00ects Yes Yes Yes Yes Observations 48 , 710 60 , 837 20 , 698 77 , 951 R - squared 0 . 008 0 . 016 0 . 008 0 . 012 Number of lab pairs 3 , 864 6 , 102 1 , 366 6 , 485 Notes . The dependent variable in all columns is a dummy equal to 1 if the lab pair collaborates in the focal year . In columns ( 1 ) and ( 3 ) , only pairs with above the median cosine similarity ( low search costs ) are included . In columns ( 2 ) and ( 4 ) , only pairs with below the median similarity ( high search costs ) are included . The cosine similarity measures are based on the vectors of keywords used by each lab in columns ( 1 ) and ( 2 ) , and on the vectors of cited references in columns ( 3 ) and ( 4 ) . In all cases , the measures are calculated before the moves start and do not include direct collaborations between the focal labs ( which would share mechanically all keywords and cited references ) . After Colocation is equal to 1 when a lab pair becomes colocated because of the moves and 0 otherwise . After Separation is equal to 1 when a previously colocated pair is separated because of the moves and 0 otherwise . Robust standard errors clustered at the lab - pair level in parentheses . \u2217 p < 0 . 1 . not be the key constraint for collaboration between these pairs , separation does seem to induce a decay in their likelihood of collaboration , possibly because some of these marginal , within - \ufb01eld collaborations are only sustainable when joint execution costs are low . Moreover , when search costs are high , not only do we observe a strong and positive e\ufb00ect of colocation on collaboration , but separation has only a noisy e\ufb00ect on outcomes , supporting the idea that for these pairs , the key constraint is being aware of each other\u2019s agenda . A consequence of this result is that over time , labs may become more similar because of lower search costs and repeated exposure , exhausting some of the best arbitrage opportunities in idea space . After separation , because of higher execution costs , some labs that are very similar ( columns ( 1 ) and ( 3 ) ) may reduce collab - oration with each other in favor of new opportunities with newly colocated labs that bring more novelty to their approaches ( columns ( 2 ) and ( 4 ) ) . To test if repeated exposure through colocation leads to an increase in similarity and awareness of each other\u2019s research , Figure 3 ( a ) plots the estimated regres - sion coe\ufb03cients 32 for the years before and after colo - cation using the cosine similarity in cited references as the dependent variable . The measure captures the extent of overlap in backward cites for a speci\ufb01c lab pair in a given year ( excluding joint collaborations ) . Overlap does not increase until two or more years after colocation , which is consistent with changes taking place slowly over time through knowledge \ufb02ows and direct collaboration . 33 Adding separation to the regression of similarity in cited references ( Figure 3 ( b ) ) shows that the reverse process takes place when distance increases : after sep - aration , previously colocated labs progressively grow apart , a result that is striking because in the two years before separation , they were actually more likely to cite the same references than the controls . 34 The neigh - bors of a lab , by exposing it to new knowledge , ideas , and collaboration opportunities , seem to profoundly shape its research trajectory . In this particular set - ting , whereas collaboration seems to recover from an increase in geographic distance ( Figure 2 ) \u2014possibly also because labs can make purposeful investments in temporary colocation and remote interactions to com - pensate for it\u2014knowledge \ufb02ows seem to be negatively a\ufb00ected by separation . Ironically , as separated labs embark on less correlated research trajectories , they may also set the stage for more novel recombination of ideas in the future . 4 . 3 . Quality of Collaborations and Execution Costs This last section explores the e\ufb00ect of proximity on the quality of collaborations . Hypothesis 3 predicts that if geographic proximity has a tangible e\ufb00ect on joint exe - cution costs , then after colocation , we should observe more projects both on the left tail of the quality dis - tribution ( a selection e\ufb00ect ) as well as on the right tail ( an e\ufb00ort e\ufb00ect ) . Separation instead should generate opposite e\ufb00ects . It is important to highlight that the prediction refers to the observed outcomes conditional on a project being developed , but empirically an addi - tional issue arises from the fact that not all projects are observed : e . g . , if scientists are overly optimistic about their chances of publishing their results , or if part of the projects that are initially developed are dropped before submission , then we will never see projects of Catalini : Microgeography and the Direction of Inventive Activity 4360 Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) Figure 3 . ( Color online ) Similarity in Cited References Space \u2013 0 . 06 \u2013 0 . 04 \u2013 0 . 02 0 0 . 02 0 . 04 0 . 06 0 . 08 \u2013 0 . 06 \u2013 0 . 04 \u2013 0 . 02 0 0 . 02 0 . 04 0 . 06 0 . 08 \u2013 10 \u2013 8 \u2013 6 \u2013 4 \u2013 2 0 2 4 6 Years to / Since move \u2013 10 \u2013 8 \u2013 6 \u2013 4 \u2013 2 0 2 4 6 Years to / Since move ( a ) Colocation only ( b ) Colocation vs . separation Notes . Estimated coe\ufb03cient for years before and after the move . The dependent variable is the cosine similarity ( based on cited refer - ences ) between the two labs . The higher line in the post - period is for colocation , and the lower line is for separation . Regression includes lab - pair \ufb01xed e\ufb00ects and year \ufb01xed e\ufb00ects . Error bars represent 95 % con\ufb01dence intervals based on robust standard errors clustered at the lab - pair level . the lowest quality . As a result , any e\ufb00ect on the left tail of the outcome distribution should be considered a lower estimate of the true e\ufb00ect of colocation ( or sepa - ration ) on inventive outcomes . Table 8 explores how the citation distribution 35 changes after the moves , conditional on the lab pairs collaborating . Overall , colocation seems to increase the number of collaborations that will end on both tails of the distribution ( \ufb01rst and fourth quartiles ) , and to decrease activity in the third quartile ( negative and signi\ufb01cant ) and potentially in the second quartile ( neg - ative but not signi\ufb01cant ) . A comparison between the estimated coe\ufb03cients for colocation shows that the fourth quartile ( highest quartile , column ( 4 ) ) is statisti - cally di\ufb00erent at 1 % from the one in the third quartile ( column ( 3 ) ) , and at 10 % from the one in the second quartile ( column ( 2 ) ) . It is not statistically di\ufb00erent from the coe\ufb03cient in the bottom quartile ( column ( 1 ) ) , and second and third quartiles are not statistically di\ufb00erent from each other . The noisier results in the \ufb01rst quar - tile could be due to the fact that collaborations that do not generate a publication are not observed , truncat - ing the left tail . Separation exhibits an almost symmet - ric pattern , with a positive ( although nonsigni\ufb01cant ) increase in the third quartile and a signi\ufb01cant decay in the fourth . In Table 9 , the right tail of the outcome distribution is further explored by separating lab pairs that faced high versus low search costs before the moves ( based on the same approach used in Table 7 , columns ( 1 ) and ( 2 ) ) . Whereas for pairs with low search costs , both coloca - tion and separation generate only noisy and insignif - icant results , when search costs are high , colocation leads to high impact research , and separation generates the opposite result . In the online appendix , robustness is shown using a di\ufb00erent proxy for quality ( citation weighted publications ) and by looking at the standard deviation of citations as a way to capture e\ufb00ects on both tails of the outcome distribution ( Online Appendix Table A - 2 ) . While only suggestive , the results in Table 9 support the idea that pairs that usually face high search costs ( potentially also because space allocation is generally optimized by discipline ) are not only more responsive to changes in distance ( as we have seen in Table 7 ) , but are also able to produce higher - impact research when given the opportunity to take advantage of lower execution costs . One interpretation of this result is that these col - laborations constitute arbitrage opportunities across domains of science , and that if we were able to observe these pairs for substantially longer periods of time , the positive e\ufb00ect on the right tail may revert to the mean . Another interpretation is that once faced with a more heterogeneous set of local peers , labs were encouraged to embark on higher - risk , higher - reward projects ( which would be consistent with the increase in variance ) . The result is also consistent with the view that multidisciplinary research tends to be of higher variance and , when successful , of higher impact . The e\ufb00ects are also a reminder of how search frictions can be a tangible obstacle to impactful research outside of a community of science . In these cases , colocation ( whether temporary or not ) can be used to remove some of these frictions and introduce novelty within current research agendas . 5 . Limitations This study has a number of limitations . First , only col - laborations that end up in a peer - reviewed publication are observed , meaning that the left tail of the outcome Catalini : Microgeography and the Direction of Inventive Activity Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) 4361 Table 8 . Colocation , Separation , and the Citation Distribution ( 1 ) ( 2 ) ( 3 ) ( 4 ) # Collabs # Collabs # Collabs # Collabs Variables 1st quartile ( lowest ) 2nd quartile 3rd quartile 4th quartile ( highest ) After Colocation 0 . 2135 \u2212 0 . 0719 \u2212 0 . 3937 \u2217 0 . 4250 \u2217\u2217 ( 0 . 2241 ) ( 0 . 2082 ) ( 0 . 2007 ) ( 0 . 1675 ) After Separation \u2212 0 . 1236 \u2212 0 . 2239 0 . 3725 \u2212 0 . 3403 \u2217\u2217 ( 0 . 2638 ) ( 0 . 2190 ) ( 0 . 3538 ) ( 0 . 1656 ) Lab - pair \ufb01xed e\ufb00ects Yes Yes Yes Yes Year \ufb01xed e\ufb00ects Yes Yes Yes Yes Observations 1 , 221 1 , 221 1 , 221 1 , 221 R - squared 0 . 061 0 . 081 0 . 070 0 . 199 Number of lab pairs 627 627 627 627 Notes . The dependent variable in all columns is the number of publications by the lab pair in the focal year within the given quartile of the citation distribution . Citation data is obtained from Scopus in 2016 , and quartiles for the citation distribution are built by year using a large sample of articles in the relevant \ufb01elds of science . The \ufb01rst quartile represents papers with the lowest level of citations , the fourth the highest . Results are conditional on collaboration in the focal year . After Colocation is equal to 1 when a lab pair becomes colocated because of the moves and 0 otherwise . After Separation is equal to 1 when a previously colocated pair is separated because of the moves and 0 otherwise . Robust standard errors clustered at the lab - pair level in parentheses . \u2217 p < 0 . 1 , \u2217\u2217 p < 0 . 05 . distribution is not observed . Whereas the analysis on cited references and keywords may be able to capture the outcome of some of the interactions ( planned or serendipitous ) that do not translate into a paper , the analyses on the rate and quality of research misses them . The measured e\ufb00ects on the left tail of the out - come distribution are therefore likely to represent a lower estimate of the true e\ufb00ects . Furthermore , collab - orations across labs are rare relative to collaborations within labs , which limits the set of tests that can be con - ducted on this sample ( e . g . , further splits of the sample across more \ufb01ne - grained dimensions ) . Table 9 . Colocation , Separation and Max Citations ( 1 ) ( 2 ) Max cites Max cites Variables ( low search costs ) ( high search costs ) After Colocation 27 . 9531 40 . 0822 \u2217\u2217\u2217 ( 31 . 6185 ) ( 13 . 3165 ) After Separation 6 . 6615 \u2212 36 . 4662 \u2217\u2217\u2217 ( 11 . 3936 ) ( 6 . 9199 ) Lab - pair \ufb01xed e\ufb00ects Yes Yes Year \ufb01xed e\ufb00ects Yes Yes Observations 639 529 R - squared 0 . 142 0 . 118 Number of lab pairs 293 294 Notes . The dependent variable in all columns is the maximum num - ber of citations received by the publications of the lab pair in the focal year . Citation data is obtained from Scopus in 2016 . In column ( 1 ) , only pairs with above the median cosine similarity in keywords ( low search costs ) are included . In column ( 2 ) , only pairs with below the median similarity in keywords ( high search costs ) are included . Results are conditional on collaboration in focal year . After Colocation is equal to 1 when a lab pair becomes colocated because of the moves and 0 otherwise . After Separation is equal to 1 when a previously colo - cated pair is separated because of the moves and 0 otherwise . Robust standard errors clustered at the lab - pair level in parentheses . \u2217\u2217\u2217 p < 0 . 01 . Second , location information is imprecise , 36 generat - ing noise in the exact timing of a move and potentially biasing the estimates downward . The same applies to the ability to correctly capture and match every pub - lication of the entities involved , as a\ufb03liation data is di\ufb03cult to clean and harmonize at scale at the lab level . Hopefully , as bibliometrics and algorithms for disam - biguation improve , better data will become available . Third , while the relocations on the Jussieu campus are substantially more constrained than typical obser - vational data on campus moves , they do not consti - tute random assignment . It is reassuring to see that the di\ufb00erence - in - di\ufb00erences estimates , and in particu - lar the pre - trends , do not hint at strong selection into colocation during the study period . Fourth , the data does not allow perfect separation and direct measurement of some of the mechanisms discussed in the theoretical framework ( e . g . , face - to - face interactions , serendipitous conversations , etc . ) . 6 . Conclusions The paper provides novel empirical evidence groun - ded in an original theoretical framework to explain why colocation matters for the rate , direction , and quality of scienti\ufb01c collaboration . To address endo - geneity concerns due to selection into colocation and matching , I exploit the constraints imposed on the spa - tial allocation of labs on the Jussieu campus of Paris by the removal of asbestos from its buildings . The analyses highlight under which conditions search and joint execution costs are more likely to be responsible for the patterns observed in the data . Consistent with recent experimental ( Boudreau et al . 2017 ) and observational ( Kabo et al . 2014 ) evidence , search frictions can be a substantial obstacle to col - laboration even within the boundaries of the same Catalini : Microgeography and the Direction of Inventive Activity 4362 Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) institution : labs that become colocated because of the moves are 3 . 5 times more likely to collaborate with each other . Supporting the search costs hypothesis , e\ufb00ects are driven by pairs that are more likely to face higher search costs ex ante , such as labs that do not work on similar topics and that do not cite the same lit - erature . Furthermore , separating previously colocated labs does not lead to a decay in collaboration , which is consistent with past exposure allowing scientists to compensate for distance through temporary coloca - tion and remote interactions , as well as through inven - tive and organizational networks ( Crescenzi et al . 2016 , Monge et al . 1985 , Breschi and Lissoni 2004 , Singh 2005 , Azoulay et al . 2011 ) . At the same time , over long periods of time , search costs seem to a\ufb00ect the research trajectories of sep - arated scientists : while colocated labs grow increas - ingly similar in topics and literature cited , separated ones undertake less correlated research trajectories . Ironically , by embarking on di\ufb00erent paths , labs may be setting the stage for future , high - impact idea recombinations . Lower execution costs under colocation , by endoge - nously allowing scientists to improve ideas more e\ufb03 - ciently when proximate , also have an e\ufb00ect on the quality of inventive outcomes . Whereas search costs profoundly in\ufb02uence who a scientist is more likely to collaborate with , joint execution costs appear to shape the distribution of outcomes , conditional on col - laboration . Consistent with the theoretical framework , after colocation , more collaborations are observed on both tails of the outcome distribution . The high - impact collaborations that emerge are predominantly coming from lab pairs that faced higher search costs ex ante : this points to potential gains from connecting labs within an institution that may not overlap directly through other channels , as these labs may bring novel ideas to domain - speci\ufb01c research agendas . Results are also con - sistent with past research that has shown that diverse inventive teams are correlated with novel idea recom - bination ( Singh and Fleming 2010 ) ; that breakthrough , Nobel Prize contributions are correlated with scientists being embedded in di\ufb00erent communities of research at the time of their development ( Ham and Weinberg 2016 ) ; and that winning solutions to innovation prob - lems tend to come from solvers from a di\ufb00erent \ufb01eld of technical expertise ( Jeppesen and Lakhani 2010 ) . The \ufb01ndings also point out some of the strategic trade - o\ufb00s that the spatial allocation of teams entails . By optimizing space based on current beliefs of where opportunities are , organizations are making a strategic choice that will profoundly shape their R & D trajectory . Space acts as a powerful layer of incentives , which can be used to de\ufb01ne not only the intensity of interactions , but also their quality . Organizations that either need to move away from a declining trajectory or want to explore radically novel opportunities can colocate pre - viously separated teams to encourage serendipitous ( and planned ) conversations between individuals with di\ufb00erent priors , ideas , and knowledge . When coloca - tion is not an option , other forms of temporary colo - cation could be strategically used to compensate for the lower chance of an interaction and higher search costs . Interestingly , once individuals are aware of each other , proximity plays a lesser role , and collaboration can also be sustained over distance . At the same time , this seems to come at the cost of right - tail outcomes , a result that future research may be able to unpack fur - ther , and that advances in communication technology and virtual reality may be able to undo ( e . g . , by recre - ating the bene\ufb01ts of in - person , face - to - face interactions and serendipitous conversations ) . The moves on the Jussieu campus , by injecting exo - genous variation into a process otherwise optimized by scienti\ufb01c \ufb01elds , also highlight how scienti\ufb01c com - munities , also because of endogenous space alloca - tion , can become an obstacle to breakthrough research . While it may not be optimal for an institution to relo - cate scientists to overcome these barriers , 37 forms of temporary colocation ( e . g . , joint conferences ) could be strategically used to encourage cross - pollination across disciplines . Since Marshall\u2019s seminal work ( Marshall 1890 ) on localization economies , scholars have been interested in why colocation matters for the generation of new ideas . While we do know that the spatial allocation of inventors and scientists has an impact on the di\ufb00usion of information and ultimately on innovation , we still know surprisingly little about the micro - foundations of knowledge recombination . This study is a \ufb01rst step toward helping us understand the mechanisms at work at di\ufb00erent levels of distance . Acknowledgments The author thanks Ajay Agrawal , Ashish Arora , Pierre Azoulay , Alessandro Bonatti , Kevin Boudreau , Je\ufb00 Furman , Alberto Galasso , Alfonso Gambardella , Patrick Gaule , Avi Goldfarb , Matt Grennan , Nicola Lacetera , Karim Lakhani , Mara Lederman , Chris Liu , Hong Luo , Liz Lyons , Alex Oettl , Peter Thompson , Carlos Serrano , Tim Simcoe , Olav Soren - son , Paula Stephan , Scott Stern , Jane Wu , the associate editor and referee team , seminar participants at MIT , the Univer - sity of Toronto , Bocconi University , Boston University , the Fox School of Business , Harvard Business School , IESE Busi - ness School , London Business School , the London School of Economics , the Krannert School of Management , the Scheller College of Business , SMU Cox School of Business , Universitat Pompeu Fabra , the Yale School of Management , Wilfrid Lau - rier University , the Academy of Management , the Association of American Geographers 2012 Annual Meeting , the the 2011 Economics of the Creative Economy conference , and Colle - gio Carlo Alberto ( Turin ) for comments . Mia Rozenbaum , Taranjit Singh , and Consuela - Emilia Tataru provided excel - lent research assistance . All errors remain the author\u2019s own . Catalini : Microgeography and the Direction of Inventive Activity Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) 4363 Endnotes 1 This assumes that inventive outcomes are skewed ( i . e . , most ideas are of very low quality ) , that ideas need to be of su\ufb03cient quality to be published ( minimum e\ufb00ort level ) , and that applying more e\ufb00ort improves the quality of a research project . 2 Additional , more strict European regulation was published in 2003 ( directive 2003 / 18 / EC ) , 2004 ( directive 2004 / 37 / EC ) , 2007 ( 2007 / 30 / EC ) , and 2009 ( 2009 / 148 / EC ) . 3 Five or more years after becoming neighbors , labs are 20 % more similar in terms of the references they cite . 4 This is consistent with past research that has shown that social proximity can compensate for geographic distance ( Agrawal et al . 2006 , Sorenson et al . 2006 ) . 5 For example , Chai ( 2014 ) \ufb01nds that participation in the same con - ference is positively correlated with future collaboration . 6 Additionally , if the arrival rate of ideas is higher under colocation because of lower search costs ( e . g . , serendipitous interactions during low - opportunity cost time ) , this e\ufb00ect would be ampli\ufb01ed . 7 See the Agrawal et al . ( 2016 ) model of slack time and innovation for a detailed description . In the paper , the authors assume that idea quality follows an exponential distribution , that the cost function is convex in e\ufb00ort , and that there is a minimum e\ufb00ort requirement for an idea to be developed . 8 http : / / www . upmc . fr / en / university . html ( accessed April 2 , 2016 ) . 9 Seehttp : / / www . ipubli . inserm . fr / handle / 10608 / 20 ( accessedApril2 , 2016 ) . 10 Additional , more strict European regulation was published in 2003 ( directive 2003 / 18 / EC ) , 2004 ( directive 2004 / 37 / EC ) , and 2007 ( 2007 / 30 / EC ) and 2009 ( 2009 / 148 / EC ) . 11 Some of the labs are placed in temporary sites . 12 For example , in 1997 , 38 . 9 % of pairs that experience a change in colocation are within sub\ufb01elds , and 61 . 1 % across sub\ufb01elds ; in 1998 the opposite is true , with 69 . 7 % within and 30 . 3 % across ; in 1999 , 89 . 5 % is across , 10 . 5 % is within ; in 2000 all pairs are across ; in 2001 , 66 . 6 % are across and 33 . 4 % are within ; in 2002 the opposite is true , with 68 . 8 % within and 31 . 2 % across , etc . 13 https : / / www . elsevier . com / solutions / scopus / content ( accessed July 6 , 2017 ) . 14 The initial set of a\ufb03liations is \ufb01rst run through a script that divides the strings into their main components ( i . e . , department name , insti - tute name , lab name , building name , \ufb02oor , address , zip code , city , and country ) . Normalized lists are then created for the key labs , institutes , and departments , so that each string can be mapped to a uni\ufb01ed name . Eighty - seven percent of a\ufb03liation strings are matched to a UPMC lab or entity , resulting in a \ufb01nal set of 36 , 822 a\ufb03liation strings . 15 Five thousand eight hundred forty - seven unique a\ufb03liation in - stances . Location information is extracted by searching for building numbers , names , and their potential abbreviations , and then manu - ally checked to con\ufb01rm accuracy . 16 See http : / / www . epaurif . fr / documentation / transferts ( accessed May 1 , 2015 ) . 17 Geodesic distances are calculated using a method developed by Thaddeus Vincenty and implemented in Stata by Nichols ( 2003 ) . 18 The choice of this level of analysis is driven by data constraints , as \ufb02oor - level data is not always available , particularly for early years . Same - \ufb02oor - level estimates for colocation ( within the subsample with \ufb02oor information ) are typically larger ; hence , the same tower and corridor building e\ufb00ects probably constitute a lower bound of the true e\ufb00ects of microgeography . 19 For index keywords , according to Scopus : \u201cA team of professional indexers assigns index terms to records according to the following controlledvocabularies ( inadditiontokeywordssuppliedbyauthors themselves ) : GEOBASE Subject Index ( geology , geography , earth and environmental sciences ) , EMTREE ( life sciences , health ) , MeSH ( life sciences , health ) , FLX terms , WTA terms ( \ufb02uid sciences , textile sci - ences ) , Regional Index ( geology , geography , earth and environmen - tal science ) , Species Index ( biology , life sciences ) , Ei Thesaurus ( con - trolled and uncontrolled terms ) ( engineering , technology , physical sciences ) . \u201d Source : http : / / info . sciencedirect . com / scopus / scopus - in - detail / content - coverage - guide / metadata ( accessed May 1 , 2015 ) . 20 The Scikit - Learn python module ( http : / / scikit - learn . org / stable / modules / metrics . html ) is used to calculate the L2 - normalized dot product of the two vectors of keywords x and y as cosinesimi - larity ( x , y ) (cid:3) xy T / (cid:107) x (cid:107)(cid:107) y (cid:107) . Seealsohttp : / / nlp . stanford . edu / IR - book / html / htmledition / dot - products - 1 . html ( accessed May 1 , 2015 ) . 21 If the two vectors were identical , then the cosine similarity would be equal to 1 . 22 In an OLS regression that includes lab - pair \ufb01xed e\ufb00ects and year \ufb01xed e\ufb00ects . Results are robust to excluding the \ufb01xed e\ufb00ects . 23 As a comparison , the expansion of a campus would not o\ufb00er the same degree of exogenous variation as labs would have more in\ufb02uence over the allocation of the new space , and the expansion itself could be the result of promising research being conducted on campus . 24 In other words , it does not simply rely on a single shock , but on multiple moves staggered in time . 25 But can also be the number of collaborations or key statistics linked to the type and quality of papers that emerge ( e . g . , citations , similar - ity in keyword or cited references space ) . 26 Descriptive statistics at this level of analysis are reported in Online Appendix Table A - 1 . 27 The baseline is any year more than 10 years away from the move . 28 In the appendix , robustness is provided by building the same \ufb01g - ure with the number of collaborations as the dependent variable ( Online Appendix Figure A - 2 ) , and by ignoring the \ufb01ve most recent years of moves in the data set . Whereas moves on the campus con - tinued until 2014 ( and a few are still taking place in 2016 ) , the data used in this paper cover 1980 \u2013 2010 . When estimating the e\ufb00ects , Online Appendix Figure A - 3 limits the sample to moves between 1997 and 2005 . 29 Empirically , it is important to highlight that separated lab pairs were more likely to be related both in knowledge space and research agendas ( since the campus was optimized to minimize distance by \ufb01eld before the moves started ) \u2014i . e . , these pairs are also more likely to overlap in other circumstances ( e . g . , teaching , conferences ) and have a higher baseline risk of collaboration . 30 Part of the noise in the result could be due to crowding out taking place for some of these pairs as labs shift part of their collaboration portfolios toward their new neighbors , as described in the lab - level regressions . 31 OLS regression with lab - pair \ufb01xed e\ufb00ects and year \ufb01xed e\ufb00ects . Error bars represent 95 % con\ufb01dence intervals based on robust stan - dard errors clustered at the lab - pair level . 32 OLS regression with lab - pair \ufb01xed e\ufb00ects and year \ufb01xed e\ufb00ects . Error bars represent 95 % con\ufb01dence intervals based on robust stan - dard errors clustered at the lab - pair level . 33 The e\ufb00ect is close in timing to the rise in collaboration observed in Figure 1 , which suggests that labs become more similar mostly through purposeful interactions . At the same time , it becomes pos - itive and signi\ufb01cant substantially earlier , which is consistent with at least some interactions predating collaboration or taking place in the absence of collaboration . In the online appendix , similar graphs show that colocated labs are also more likely to start publishing in a journal that is new to them but not to their collaborator ( Online Catalini : Microgeography and the Direction of Inventive Activity 4364 Management Science , 2018 , vol . 64 , no . 9 , pp . 4348 \u2013 4364 , \u00a92017 The Author ( s ) Appendix Figure A - 4 ) and exhibit increasing overlap in the key - words they use outside of direct collaborations ( Online Appendix Figure A - 5 ) . 34 The estimated coe\ufb03cient for the separation line at t \u2212 2 and t \u2212 1 is positive and signi\ufb01cant , which is consistent with the moves sepa - rating lab pairs that were actually bene\ufb01ting from colocation in the pre - period . 35 Citation data were obtained from Scopus in 2016 , and quartiles for the citation distribution are built by year using a large sample of articles in the relevant \ufb01elds of science . 36 An analysis on the subsample of data for which \ufb02oor - level infor - mation is available shows e\ufb00ects roughly twice as large as those measured in Table 5 for labs that not only share the same building , but also the same \ufb02oor . 37 Results on collaboration portfolios highlight how labs actually became more inward focused during the moves , potentially because of the worsened set of neighboring labs relative to their preferences . References Adams J , Clemmons R , Black C , Stephan P ( 2005 ) Scienti\ufb01c teams and institutional collaborations : Evidence from U . S . universities . Res . Policy 34 : 259 \u2013 285 . Agrawal A , Cockburn I , McHale J ( 2006 ) Gone but not forgotten : Labor \ufb02ows , knowledge spillovers and enduring social capital . J . Econom . Geography 6 ( 5 ) : 571 \u2013 591 . Agrawal A , Catalini C , Goldfarb A , Luo H ( 2016 ) Slack time and inno - vation . Rotman School of Management Working Paper 2599004 , University of Toronto , Toronto . Allen TJ ( 1977 ) Managing the Flow of Technology ( MIT Press , Cam - bridge , MA ) . Azoulay P , Zivin JSG , Sampat BN ( 2011 ) The di\ufb00usion of scienti\ufb01c knowledge across time and space : Evidence from professional transitions for the superstars of medicine . Technical report , National Bureau of Economic Research . Belenzon S , Schankerman M ( 2013 ) Spreading the word : Geogra - phy , policy , and knowledge spillovers . Rev . Econom . Statist . 95 ( 3 ) : 884 \u2013 903 . Boudreau KJ , Brady T , Ganguli I , Gaule P , Guinan E , Hollenberg A , Lakhani KR ( 2017 ) A \ufb01eld experiment on search costs and the formation of scienti\ufb01c collaborations . Rev . Econom . Statist . 99 ( 4 ) : 565 \u2013 576 . Breschi S , Lissoni F ( 2004 ) Knowledge networks from patent data : Methodological issues and research targets . Moed HF , Gl\u00e4nzel W , Schmoch U , eds . Handbook of Quantitative Science and Technology Research : The Use of Publication and Patent Statistics in Studies of S & T Systems ( Springer , Berlin ) , 613 \u2013 643 . Chai S ( 2014 ) Temporary colocation and collaborative discovery . Working paper , Harvard University , Cambridge , MA . Cowgill B , Wolfers J , Zitzewitz E ( 2009 ) Using prediction mar - kets to track information \ufb02ows : Evidence from Google . Das S , Ostrovsky M , Pennock D , Szymanksi B , eds . Auctions , Market Mechanisms and Their Applications , Lecture Notes of the Institute for Computer Sciences , Social Informatics and Telecommunica - tions Engineering , Vol . 14 ( Springer , Berlin ) . Crescenzi R , Filippetti A , Iammarino S ( 2017 ) Academic inven - tors : Collaboration and proximity with industry . J . Tech . Transfer 42 ( 4 ) : 730 \u2013 762 . Crescenzi R , Nathan M , Rodr\u00edguez - Pose A ( 2016 ) Do inventors talk to strangers ? Proximity and collaborative knowledge creation . Res . Policy 45 ( 1 ) : 177 \u2013 194 . Dizikes P ( 2011 ) The o\ufb03ce next door . MIT Tech . Rev . ( October 25 ) , https : / / www . technologyreview . com / s / 425881 / the - o\ufb03ce - next - door . Fayard A - L , Weeks J ( 2007 ) Photocopiers and water - coolers : The a\ufb00ordances of informal interaction . Organ . Stud . 28 ( 5 ) : 605 \u2013 634 . Fleming L , Sorenson O ( 2004 ) Science as a map in technological search . Strategic Management J . 25 ( 8 \u2013 9 ) : 909 \u2013 928 . Fleming L , King I Charles , Juda AI ( 2007 ) Small worlds and regional innovation . Organ . Sci . 18 ( 6 ) : 938 \u2013 954 . Freeman R , Ganguli I , Murciano - Goro\ufb00 R ( 2015 ) Why and wherefore of increased scienti\ufb01c collaboration . Ja\ufb00e AB , Jones BF , eds . The Changing Frontier : Rethinking Science and Innovation Policy ( Uni - versity of Chicago Press , Chicago ) , 17 \u2013 48 . Ham JC , Weinberg BA ( 2016 ) Novelty , knowledge spillovers and innovation : Evidence from Nobel laureates . ZBW GLO Discus - sion Paper 30 , Global Labor Organization . Ja\ufb00e A , Trajtenberg M , Henderson R ( 1993 ) Geographic localization of knowledge spillovers as evidenced by patent citations . Quart . J . Econom . 108 ( 3 ) : 577 \u2013 598 . Jeppesen LB , Lakhani KR ( 2010 ) Marginality and problem - solving e\ufb00ectiveness in broadcast search . Organ . Sci . 21 ( 5 ) : 1016 \u2013 1033 . Jones BF , Uzzi B , Wuchty S ( 2008 ) Multi - university research teams : Shifting impact , geography and social strati\ufb01cation in science . Science 322 ( 5905 ) : 1259 \u2013 1262 . Kabo F , Hwang Y , Levenstein M , Owen - Smith J ( 2015 ) Shared paths to the lab a sociospatial network analysis of collaboration . Envi - ron . Behav . 47 ( 1 ) : 57 \u2013 84 . KaboFW , Cotton - NesslerN , HwangY , LevensteinMC , Owen - SmithJ ( 2014 ) Proximity e\ufb00ects on the dynamics and outcomes of scien - ti\ufb01c collaborations . Res . Policy 43 ( 9 ) : 1469 \u2013 1485 . Katz J ( 1994 ) Geographical proximity and scienti\ufb01c collaboration . Scientometrics 31 ( 1 ) : 31 \u2013 43 . Liu C ( 2013 ) Brokerage by design : Formal structure , geography , and crosscutting ties . Working paper , Rotman School of Manage - ment , University of Toronto , Ontario . Lucas REJ ( 1988 ) On the mechanics of economic development . J . Monetary Econom . 22 ( 1 ) : 3 \u2013 42 . Mairesse J , Turner L ( 2005 ) Measurement and explanation of the intensity of co - publication in science : An analysis at the labo - ratory level . NBER Working Paper 11172 , National Bureau of Economic Research , Cambridge , MA . Marshall A ( 1890 ) Principles of Economics ( Macmillan , London ) . Monge PR , Rothman LW , Eisenberg EM , Miller KI , Kirste KK ( 1985 ) The dynamics of organizational proximity . Management Sci . 31 ( 9 ) : 1129 \u2013 1141 . Nichols A ( 2003 ) VINCENTY : Stata module to calculate distances on the Earth\u2019s surface . Statistical Software Components , Boston College . Nielsen C ( 2016 ) How Microsoft used an o\ufb03ce move to boost collab - oration . Harvard Bus . Rev . ( October 11 ) , https : / / hbr . org / 2016 / 10 / how - microsoft - used - an - o\ufb03ce - move - to - boost - collaboration . Olson GM , Olson JS ( 2000 ) Distance matters . Human \u2013 Computer Inter - action 15 ( 2 ) : 139 \u2013 178 . Romer PM ( 1990 ) Endogenous technological change . J . Political Econom . 98 ( 5 ) : 71 \u2013 102 . Singh J ( 2005 ) Collaborative networks as determinants of knowledge di\ufb00usion patterns . Management Sci . 51 ( 5 ) : 756 \u2013 770 . Singh J , Fleming L ( 2010 ) Lone inventors as sources of break - throughs : Myth or reality ? Management Sci . 56 ( 1 ) : 41 \u2013 56 . Sorenson O , Rivkin JW , Fleming L ( 2006 ) Complexity , networks and knowledge \ufb02ow . Res . Policy 35 ( 7 ) : 994 \u2013 1017 . Stuart T , Liu C ( 2010 ) Boundary spanning in a for - pro\ufb01t research lab : An exploration of the interface between commerce and academe . Working paper , Harvard University , Cambridge , MA . Thompson P ( 2006 ) Patent citations and the geography of knowl - edge spillovers : Evidence from inventor - and examiner - added citations . Rev . Econom . Statist . 88 ( 2 ) : 383 \u2013 388 . Thompson P , Fox - Kean M ( 2005 ) Patent citations and the geogra - phy of knowledge spillovers : A reassessment . Amer . Econom . Rev . 95 ( 1 ) : 450 \u2013 460 . Van den Bulte C , Moenaert RK ( 1998 ) The e\ufb00ects of R & D team co - location on communication patterns among R & D , marketing , and manufacturing . Management Sci . 44 ( 11 , part 2 ) : S1 \u2013 S18 . Weitzman ML ( 1998 ) Recombinant growth . Quart . J . Econom . 113 ( 2 ) : 331 \u2013 360 . Wuchty S , Jones B , Uzzi B ( 2007 ) The increasing dominance of teams in production of knowledge . Science 316 ( 5827 ) : 1036 \u2013 1039 .",
    "kyumurkov2023force": "ARTICLE Force tuning through regulation of clathrin - dependent integrin endocytosis Alexander Kyumurkov 1 * , Anne - Pascale Bouin 1 * \ue840 , Mathieu Boissan 2 , 3 , Sandra Manet 1 , Francesco Baschieri 4 \ue840 , Mathilde Proponnet - Guerault 1 , Martial Balland 5 \ue840 , Olivier Destaing 1 \ue840 , Myriam R\u00b4egent - Kloeckner 1 , Claire Calmel 2 , 3 , Alice Nicolas 6 \ue840 , Fran\u00e7ois Waharte 2 , 3 , Philippe Chavrier 7 \ue840 , Guillaume Montagnac 4 \ue840 , Emmanuelle Planus 1 * * \ue840 , and Corinne Albiges - Rizo 1 * * \ue840 Integrin endocytosis is essential for many fundamental cellular processes . Whether and how the internalization impacts cellular mechanics remains elusive . Whereas previous studies reported the contribution of the integrin activator , talin , in force development , the involvement of inhibitors is less documented . We identified ICAP - 1 as an integrin inhibitor involved in mechanotransduction by co - working with NME2 to control clathrin - mediated endocytosis of integrins at the edge of focal adhesions ( FA ) . Loss of ICAP - 1 enables \u03b2 3 - integrin - mediated force generation independently of \u03b2 1 integrin . \u03b2 3 - integrin - mediated forces were associated with a decrease in \u03b2 3 integrin dynamics stemming from their reduced diffusion within adhesion sites and slow turnover of FA . The decrease in \u03b2 3 integrin dynamics correlated with a defect in integrin endocytosis . ICAP - 1 acts as an adaptor for clathrin - dependent endocytosis of integrins . ICAP - 1 controls integrin endocytosis by interacting with NME2 , a key regulator of dynamin - dependent clathrin - coated pits fission . Control of clathrin - mediated integrin endocytosis by an inhibitor is an unprecedented mechanism to tune forces at FA . Introduction Adhesive receptors , most notably integrins , help cells to perceive their microenvironment by sensing chemical , physical , and mechanical cues of extracellular matrix ( ECM ) through adhesive machineries called focal adhesions ( FA ) , which act in concert with the actomyosin - based contractility system ( Albiges - Rizo et al . , 2009 ; Engler et al . , 2006 ) . Actomyosin - mediated con - tractility is a highly conserved mechanism for generating me - chanical stress and governing cell shape , cell migration , cell differentiation , and morphogenesis ( Murrell et al . , 2015 ) . Cel - lular contractility usually proceeds with the engagement of \u03b1 5 \u03b2 1 and \u03b1 v \u03b2 3 integrins in the context of cells exploring fibronectin ( Fn ) - based microenvironments ( Schiller et al . , 2013 ) . These two integrins cooperate for the fine tuning of FA lifetime and adhe - sion strength in response to mechanical forces ( Kuo et al . , 2011 ; Schiller et al . , 2013 , 2011 ; Zamir et al . , 2000 ; Rossier et al . , 2012 ; Roca - Cusachs et al . , 2009 ; Milloud et al . , 2017 ; De Mets et al . , 2019 ) . Reinforcement of FA and cell contractility is likely coupled to the inhibition of FA disassembly , a processes involving FA targeting by microtubules , enhanced integrin endocytosis , calpain - mediated cleavage of talin , and loss of tension upon Rho kinase inhibition ( Wehrle - Haller , 2012 ) . \u03b1 5 \u03b2 1 and \u03b1 v \u03b2 3 integrins are functionally distinct , yet it remains unclear how these two Fn - binding integrin receptors are coordinately regulated to or - chestrate assembly or disassembly of adhesion sites , to adjust adhesion lifetime , to adapt adhesion strength and force to the cellular environment . The regulation of integrin function can be achieved at several levels , including ECM ligand engagement and recruitment of intracellular adaptors including integrin activators such as talin and integrin inhibitors like ICAP - 1 . These intracellular adaptors control integrin clustering and their activation switch , which is crucial for modulating integrin ligand - binding affinity and for serving as nucleation hubs for the assembly of larger signaling and structural scaffolds linked to actomyosin fibers ( Legate and . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 University Grenoble Alpes , INSERM 1209 , CNRS UMR5309 , Institute for Advanced Biosciences , Grenoble , France ; 2 University Sorbonne , INSERM UMR _ S 938 , Saint - Antoine Research Center , CRSA , Paris , France ; 3 Laboratory of Biochemistry and Hormonology , Tenon Hospital , AP - HP , Paris , France ; 4 Inserm U1279 , Gustave Roussy Institute , Universit\u00b4e Paris - Saclay , Villejuif , France ; 5 Laboratoire Interdisciplinaire de Physique , UMR CNRS 5588 , University Grenoble Alpes , Grenoble , France ; 6 University Grenoble Alpes , CNRS , CEA / LETIMinatec , Grenoble Institute of Technology , Microelectronics Technology Laboratory , Grenoble , France ; 7 Institut Curie , UMR144 , Universit\u00b4e de Recherche Paris Sciences et Lettres , Centre Universitaire , Paris , France . S . Manet died in January 2021 . * A . Kyumurkov , and A . - P . Bouin contributed equally to this paper ; * * E . Planus , and C . Albiges - Rizo contributed equally to this paper . Correspondence to Emmanuelle Planus : mailto : emmanuelle . planus @ univ - grenoble - alpes . fr ; Corinne Albiges - Rizo : corinne . albiges - rizo @ univ - grenoble - alpes . fr Myriam R\u00b4egent - Kloeckner \u2019 s present address is CY Cergy Paris Universit\u00b4e , Cergy , France . \u00a9 2022 Kyumurkov et al . This article isdistributedunder the termsof an Attribution \u2013 Noncommercial \u2013 Share Alike \u2013 NoMirror Sites license for the first six months after the publication date ( see http : / / www . rupress . org / terms / ) . After six months it is available under a Creative Commons License ( Attribution \u2013 Noncommercial \u2013 Share Alike 4 . 0 International license , as described at https : / / creativecommons . org / licenses / by - nc - sa / 4 . 0 / ) . Rockefeller University Press https : / / doi . org / 10 . 1083 / jcb . 202004025 1 of 19 J . Cell Biol . 2023 Vol . 222 No . 1 e202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 F\u00a8assler , 2009 ) . Whereas talin is known to be important for building actin - bound mechanosensitive adhesive complexes ( Klapholz and Brown , 2017 ; Shattil et al . , 2010 ) , the contribution of integrin inhibitors such as ICAP - 1 in force transmission is still elusive . ICAP - 1 does not localize in FA , but it can accumulate at small adhesion sites or tips of filopodia ( Jacquemet et al . , 2019 ; Fournier et al . , 2002 , 2005 ; Millon - Fr\u00b4emillon et al . , 2008 ) . Thus , ICAP - 1 interacts with partners described to be within and out - side FA ( Fournier et al . , 2002 ; Millon - Fr\u00b4emillon et al . , 2008 ) . ICAP - 1 interacts with the membrane - distal NPXY motif in the cytoplasmic tail of \u03b2 1 integrin through a phosphotyrosine binding ( PTB ) domain ( Zhang and Hemler , 1999 ; Degani et al . , 2002 ) . Further , it binds to the nucleoside diphosphate kinase NDPK / NME2 / NM23 - H2 ( Fournier et al . , 2002 ) . Of note , ICAP - 1 , NME2 , and \u03b2 1 integrin are colocalized at the cell leading edge during the early stage of cell spreading on Fn , thereby suggesting a transitory function of this complex in adhesion site dynamics ( Fournier et al . , 2002 ) . However , the biological relevance of ICAP - 1 / NME2 interaction in cell adhesion remains unknown . We have shown that ICAP - 1 is a determinant for controlling FA dynamics since it impedes FA assembly by limiting both talin and kindlin interaction with \u03b2 1 integrin ( Millon - Fr\u00b4emillon et al . , 2013 ; Millon - Fr\u00b4emillon et al . , 2008 ) . Furthermore , the physical inter - action of ICAP - 1 with the \u03b2 1 integrin tail leading to the modula - tion of \u03b2 1 integrin affinity state is required for down regulation of FA assembly . We have shown that ICAP - 1 is involved in cell mechanical properties and cell differentiation in a \u03b2 1 integrin - dependent manner ( Brunner et al . , 2011 ; Faurobert et al . , 2013 ; Bouvard et al . , 2007 ; Renz et al . , 2015 ; Millon - Fr\u00b4emillon et al . , 2008 ) . We have also proposed ICAP - 1 as a key modulator of cellular mechanoresponse in a \u03b2 1 - integrin \u2013 independent manner ( Bouin et al . , 2017 ) . To delve deeper into the ICAP - 1 func - tional properties , we interrogated \u03b2 1 - integrin \u2013 dependent and \u2013 independent contributions of ICAP - 1 to the cell mechanical response . Using cellular models genetically engineered to lack only \u03b2 1 integrin or ICAP - 1 or both \u03b2 1 integrin and ICAP - 1 in conjunction with quantitative traction force microscopy , in - tegrin dynamics , and proximity ligation assays , we provide converging evidence that points to a control by ICAP - 1 of clathrin - mediated \u03b2 3 integrin endocytosis and crucial tuning of cell mechanoresponses . Results The loss of ICAP - 1 enables \u03b2 3 - integrin - mediated force generation independent of \u03b2 1 integrin To understand to what extent ICAP - 1 regulates forces , we de - signed a quantitative traction force microscopy analysis of os - teoblasts deleted for \u03b2 1 integrin and / or ICAP - 1 . Deletion of \u03b2 1 integrin or ICAP - 1 in the respective cell lines was confirmed by Western blot ( Fig . S1 , A \u2013 C ) . Based on qPCR experiments and Western blot analysis ( Fig . S1 , D and E ) , we checked that the combined deletion of \u03b2 1 integrin and ICAP - 1 did not affect the total expression of \u03b2 3 integrin . Cells were seeded on Fn - coated polyacrylamide hydrogels ( PA hydro gels ) with a Young \u2019 s mo - dulus ( E ) of 5 kPa as previously described ( Bouin et al . , 2017 ) . First , we observed that osteoblasts lacking \u03b2 1 integrin ( \u03b2 1 integrin \u2013 / \u2013 - icap - 1 + / + ) were defective in force generation ( Fig . 1 , A and B ) similarly to what was reported for other cell lines ( Danen et al . , 2002 ; Schiller et al . , 2013 ; Milloud et al . , 2017 ) . The \u03b2 1 integrin - deficient cell line developed only 50 % of the total forces generated by \u03b2 1 integrin + / + - icap - 1 + / + cells ( Fig . 1 B ) . Secondly and unexpectedly , we found that additional loss of ICAP - 1 in \u03b2 1 integrin - deficient osteoblasts restored the traction force potential ( Fig . 1 , A and B ) . This result was confirmed by the rescue of myosin light - chain phosphorylation ( pp - MLC ) as judged by immunofluorescence staining ( Fig . 1 , C \u2013 E ; and Fig . S1 F ) and Western blot analysis ( Fig . S1 G ) using phospho site - specific antibodies against myosin light chain . In addition , while \u03b2 1 integrin \u2013 / \u2013 cells displayed limited spreading as evidenced by the projected area of actin cytoskeleton , normal cell spreading was rescued in the double - mutant ( \u03b2 1 integrin \u2013 / \u2013 - icap - 1 \u2013 / \u2013 ) cells that typically assembled thick actin stress fibers highly deco - rated with ppMLC ( Fig . 1 , C \u2013 E ; and Fig . S1 F ) . Of note , the loss of ICAP - 1 increased slightly force developed by osteoblasts al - though the difference with control cells was non - significant ( Fig . 1 , A and B ) . Contractility is an important aspect of cell migration , thus set out to explore the role of ICAP - 1 in the ad - aptation of cell speed as a function of substrate stiffness . Cells were seeded on Fn - coated PA hydrogels of increasing rigidity and imaged for 3 h to monitor migration velocity ( Fig . S2 A ) . \u03b2 1 integrin + / + - icap - 1 + / + and \u03b2 1 - deficient cells showed a stiffness - dependent increase in cell velocity . However , additional loss of ICAP - 1 uncoupled the correlation between cell speed and substrate rigidity . As the first steps of Fn fibrillogenesis are known to de - pend on actomyosin contractility ( Wu et al . , 1995 ) , we investigated whether the traction force observed in \u03b2 1 integrin \u2013 / \u2013 - icap - 1 \u2013 / \u2013 cells was correlated with Fn fibrillogenesis . As previously described ( Danen et al . , 2002 ) , the lack of \u03b2 1 integrin drastically impaired the formation of a stretched meshwork of Fn fibrils ( Fig . 1 , F and G ) . However , additional loss of ICAP - 1 , which was associated with increased traction forces ( Fig . 1 A ) , rescued the formation and organization of thick and numerous Fn fibrils connecting to large \u03b2 3 integrin adhesions as compared to the short and less dense Fn fibrils observed in the case of \u03b2 1 integrin deletion alone ( Fig . 1 , F and G ) . These results indicate that ICAP - 1 plays a role in cellular mechanotransduction and that traction strength on a compliant Fn substrate does not strictly correlate with the presence of \u03b2 1 integrin and is likely related to another integrin receptor regulated by ICAP - 1 . As Fn - coated surface mediates RGD binding to \u03b1 5 \u03b2 1 and \u03b1 v \u03b2 3 integrins ( Leiss et al . , 2008 ) which can exert both specific and redundant functions ( Ballestrem et al . , 2001 ; Danen et al . , 2002 ) , we investigated whether the increase in traction forces in \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells was dependent on \u03b2 3 integrin engagement . Silencing of \u03b2 3 integrin by RNA in - terference ( Fig . 2 F ) not only led to a significant decrease in cell spreading and ppMLC staining ( Fig . 2 , A and B ) but also abol - ished traction forces generated in \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells when compared to \u03b2 1 integrin + / + - icap - 1 + / + osteoblasts ( Fig . 2 , C and D ) . These data position \u03b2 3 integrin as a major driver for the spreading and tensile phenotypes observed on \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells . In line with this scheme , the contractile phenotype Kyumurkov et al . Journal of Cell Biology 2 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure 1 . Osteoblasts are able to exert traction force on fibronectin - coated substrate in the absence of \u03b2 1 integrin and ICAP - 1 . ( A and B ) Repre - sentative traction forces maps ( A ) and quantification ( B ) of the total force applied on fibronectin - coated polyacrylamide gel with a defined rigidity of 5 kPa . Kyumurkov et al . Journal of Cell Biology 3 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 of \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells was correlated with an increase in \u03b2 3 integrin expression at the cell surface as judged by FACS analysis ( Fig . 2 E ) while the total amount of \u03b2 3 in - tegrin remained unchanged ( Fig . S1 , D and E ) . Altogether , our results suggest a new regulatory role for ICAP - 1 in ac - tomyosin contractility and force generation through \u03b2 3 integrin . \u03b2 3 - integrin - dependent traction force is associated with redistribution and higher lifetime of enlarged \u03b2 3 - integrin - positive focal adhesions Stabilization of \u03b1 v \u03b2 3 integrin - Fn bonds through actomyosin - mediated tension is required for cells to adjust cell contractil - ity to the substrate stiffness ( Schiller et al . , 2013 ; De Mets et al . , 2019 ) . Based on the increase in surface \u03b2 3 integrin expression levels in \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells ( Fig . 2 E ) , we investigated whether the contractile behavior mediated by \u03b2 3 integrin upon the loss of \u03b2 1 integrin and ICAP - 1 might be related to a change in \u03b2 3 integrin - containing FA organization . All four osteoblast cells lines were able to form \u03b2 3 integrin FAs on an Fn - coated sub - stratum as revealed by immunostaining of \u03b2 3 integrin ( Fig . 3 A ) . Nevertheless , the deletion of \u03b2 1 integrin noticeably reduced cell spreading ( Fig . 1 D ) without affecting the adhesion area of FAs occupied by \u03b2 3 integrin even though it total adhesion area was increased with respect to the cell area ( Fig . 3 , A \u2013 C ) . These results suggest that \u03b2 3 integrin alone failed to support the typical spreading of osteoblasts bound to Fn . Remarkably , the additional loss of ICAP - 1 in \u03b2 1 - integrin - deficient cells restored cell spreading and increased the mean size of \u03b2 3 integrin FAs when compared to \u03b2 1 integrin null cells ( Fig . 3 , A and C ) . We noticed an increase of \u03b2 3 integrin FA translocation to the cell center ( Fig . S2 B ) which is compatible with the rescue of Fn fibril organization in \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells ( Fig . 1 , F and G ) . These results revealed a major role for ICAP - 1 in controlling \u03b2 3 integrin clustering and function in the absence of \u03b2 1 integrin . Next , we analyzed whether the contractile behavior mediated by \u03b2 3 in - tegrin upon the loss of \u03b2 1 integrin and ICAP - 1 might be linked to a change in \u03b2 3 - integrin - containing FA dynamics . To this end , the turnover of FA in osteoblast cell lines transfected with \u03b2 3 integrin - eGFP was monitored in real time ( Fig . 3 D ) . The life time of \u03b2 3 integrin - eGFP FAs increased in association with their translocation to the cell center in \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells ( Fig . 3 , D and E ; and Fig . S2 B ) . \u03b2 3 integrin exchange rates were further analyzed by fluo - rescent recovery after photobleaching ( FRAP ) using total inter - nal reflection fluorescent ( TIRF ) microscopy on osteoblasts lines transfected with \u03b2 3 integrin - eGFP ( Fig . 3 F ) . Neither single loss of \u03b2 1 integrin nor that of ICAP - 1 had any significantly effect on \u03b2 3 integrin exchange rate in FAs . In contrast , a threefold slower \u03b2 3 integrin - eGFP exchange rate typified the oversized \u03b2 3 in - tegrin FAs observed in double \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells ( Fig . 3 F ) , as compared to \u03b2 1 null cells . Finally , the slower \u03b2 3 integrin exchange rate was associated with an increase of \u03b2 3 integrin - eGFP FA life time ( Fig . 3 , D \u2013 F ; and Videos 1 , 2 , 3 , and 4 ) . In the context of \u03b2 1 null cells , these data hint that the loss of ICAP - 1 strongly affects the clustering and dynamics of \u03b2 3 in - tegrin in FAs . These findings reveal that the control of \u03b2 3 in - tegrin clustering and dynamics by ICAP - 1 might modulate traction force generation . \u03b2 3 - integrin - dependent contractility is associated with a defect of clathrin - mediated \u03b2 3 integrin endocytosis As endocytic membrane traffic regulates the availability of cell - surface receptors and associated signaling ( Ceresa and Schmid , 2000 ; Scita and Di Fiore , 2010 ) , we hypothesized that the con - comitant increase in \u03b2 3 integrin surface expression ( Fig . 2 E ) and the decrease in \u03b2 3 integrin dynamics in \u03b2 1 / ICAP - 1 double - mutant cells ( Fig . 3 , D \u2013 F ) might stem from a defect in \u03b2 3 integrin endocytosis . To test this hypothesis , we monitored the uptake of anti - integrin \u03b2 3 antibodies by cells plated on an Fn substrate using confocal microscopy . Uptake was measured using \u03b2 3 integrin - specific antibody ( LucA . 5 ) coupled with pH - rodo , which becomes fluorescent in the acidic environment of endo - cytic vesicles ( Fig . 4 A ) . Quantification of intracellular vesicles positive for internalized \u03b2 3 integrin ( LucA5 - positive ) revealed a significant decrease in uptake ( 23 % ) in the absence ICAP - 1 as compared to \u03b2 1 integrin + / + - icap - 1 + / + cells . An even more sig - nificant decrease ( 40 % ) was measured in osteoblasts deficient for both \u03b2 1 integrin and ICAP - 1 when compared to cells depleted in \u03b2 1 integrin only ( Fig . 4 A ) . Past studies have demonstrated that surface \u03b2 3 integrins are constitutively internalized by clathrin - mediated endocytosis that depends on the AP - 2 clathrin \u03b2 1 integrin KO cells exert less force than the other osteoblasts mutants . The additional deletion of ICAP - 1 led to generation of traction forces revealing a novel pathway independent of \u03b2 1 integrins to generate traction forces on fibronectin . Scale bar , 10 \u03bc m . Error bars represent SD . N \u2265 60 cells . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( C ) Immunofluorescence staining of the ppMyosin ( ppMLC antibody , red ) and F - actin ( phalloidin , green ) in the four osteoblasts cell lines showed that deletion of ICAP - 1 alone does not change the organization of acto - myosin cytoskeleton but increases slightly the intensity and the thickness of the stress fibers ( see quantification of the ppMLC area in E ) . Deletion of \u03b2 1 integrin leads to a decrease of the cell area ( D ) and disorganization and decrease of thickness and number of the ppMLC decorated stress fibers . Scale bar , 20 \u00b5m . ( D ) Quantifiedcellspreading areafrom stainingoffluorescent F - actin . Error bars represent SD . N \u2265 429 cells . ns , adjustedP value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( E ) Quantified ppMLC staining , normalized to cell area . Error bars represent SD . N \u2265 211 cells . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( F ) Osteoblasts cells were spread in serum - free medium on uncoated glass for 24 h . Immunofluorescence staining of the extracellular fibronectin ( cellular fibronectin antibody , red ) , F - actin ( phalloidin , blue ) and \u03b2 3 integrins ( Luc . A5 antibody , green ) inthe four osteoblasts cell lines were analyzedby fluorescent confocalmicroscopy . The \u03b2 1 + / + cell lines orchestrate FN fibrillogenesis events and \u03b2 1 integrin \u2212 / \u2212 / icap - 1 + / + cells demonstrated very poor organization of synthetized fibronectin . \u03b2 1 integrin \u2212 / \u2212 / icap - 1 \u2212 / \u2212 on the other side showed significant amount of FN fibrillogenesis . Scale bar , 20 \u00b5m . ( G ) FN fibrils length of thresholded images of deposited and organized FN fibrillogenesis were processed and quantified . 20 images per condition were analyzed . Quantifications reveal that the additional loss of ICAP - 1 in \u03b2 1 KO cell line increases the fibril length compared to the \u03b2 1 KO cell line , expressing ICAP - 1 . Error bars represent SD . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . Kyumurkov et al . Journal of Cell Biology 4 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure 2 . ppMyosin area and traction force are dependent on \u03b2 3 integrin and ICAP - 1 expression . ( A ) \u03b2 1 integrin KO osteoblast cell lines were treated with \u03b2 3 integrin siRNA or scramble siRNA for 48 h and then let spread on FN coated glass for 24 h . Staining of myosin phosphorylation ( ppMLC antibody , red ) Kyumurkov et al . Journal of Cell Biology 5 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 adaptor complex ( Arjonen et al . , 2012 ; Yu et al . , 2015 ; Ezratty et al . , 2009 ) . Moreover , the large GTPase dynamins are known to be required for the scission of newly formed clathrin - coated vesicles ( CCV ) from the plasma membrane clathrin - coated pits ( CCP ) . We applied proximity ligation assay ( PLA ) , which allows the detection of close proximity , not only between ICAP - 1 and \u03b2 3 integrin ( Fig . 4 B ) but also between ICAP - 1 and CCP components such as the \u03b1 - adaptin of the clathrin adaptor complex , AP - 2 and dynamin - 2 ( Fig . 4 C ) , supporting a role for ICAP - 1 in \u03b2 3 integrin endocytosis . Indeed , the decrease of \u03b2 3 integrin endocytosis in the case of ICAP - 1 deletion was associated with an increase of CCP containing \u03b2 1 and \u03b2 3 integrins ( Fig . 4 D ) , specifying the contribution of ICAP - 1 in the formation of CCVs . Given the link between ICAP - 1 and clathrin - mediated endo - cytosis of \u03b2 3 integrin , we questioned whether the loss of clathrin might impact cell spreading , like the loss of ICAP - 1 . As expected , whereas \u03b2 1 integrin - deficient cells displayed a decrease in their spreading capacity compared to \u03b2 1 integrin + / + - icap - 1 + / + cells , additional knockdown of the clathrin heavy chain ( Fig . S3 ) re - stored their ability to spread ( Fig . 4 , E and F ) , in conjunction with the formation of larger \u03b2 3 integrin - positive FAs ( Fig . 4 , E and G ) and ppMLC - enriched actin stress fibers ( Fig . 4 , E and H ) , thus phenocopying the effects of the ICAP - 1 loss in those cells . Importantly , there was no additive effect of clathrin deletion over ICAP - 1 loss ( \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells and \u03b2 1 integrin + / + - icap - 1 \u2212 / \u2212 cells ) with respect to cell spreading , \u03b2 3 integrin - positive FA size and ppMLC staining ( Fig . 4 , E \u2013 H ) . These data demon - strate that the loss of clathrin and resulting impairment in clathrin - mediated endocytosis impacts \u03b2 3 integrin signaling . Accordingly , the increase in \u03b2 3 integrin clustering concurs with the increase in cell spreading and reorganization of the acto - myosin cytoskeleton . In addition , the non - sensitivity of cells devoid of ICAP - 1 to alteration of the clathrin - based endocytic machinery and the close proximity of ICAP - 1 with AP2 and dy - namin support a potential role of ICAP - 1 in clathrin - mediated integrin endocytosis . ICAP - 1 controls \u03b2 3 integrin endocytosis through NME - dependent scission of endocytic clathrin - coated pits Next , we investigated the mechanism by which ICAP - 1 might affect \u03b2 3 integrin endocytosis . We have previously reported an interaction between ICAP - 1 and nucleoside diphosphate kinase NME2 ( Fournier et al . , 2002 ) , known to catalyze the synthesis of nucleoside triphosphates including GTP from nucleoside diphosphates and ATP ( Boissan et al . , 2018 ) . Genetic and functional studies have demonstrated the ability of the related NME1 and NME2 NDPKs to fuel dynamin with GTP to support clathrin - dependent endocytosis ( Boissan et al . , 2014 ; Krishnan et al . , 2001 ; Dammai et al . , 2003 ; Nallamothu et al . , 2008 ) . Based on these findings , we investigated whether NME might impact \u03b2 3 integrin dynamics . As NME2 can work in a hexameric complex with NME1 to be recruited to CCPs through their physical interaction with dynamin ( Boissan et al . , 2014 ) , both NME1 and NME2 were knocked down in osteoblasts ( Fig . 5 A ) . Deletion of NME1 / 2 in \u03b2 1 - integrin - depleted osteoblasts de - creased \u03b2 3 integrin \u2013 GFP turnover as evidenced by FRAP ex - periments ( Fig . 5 B ) , thus recapitulating the deletion of ICAP - 1 . We further addressed whether ICAP - 1 might impact the locali - zation and function of NME2 before membrane scission . First , PLA showed that NME2 and \u03b1 - adaptin as well as dynamin ( Fig . 5 , C and D ) exist in close proximity to each other in agreement with the known functional link between NME and CCP ( Boissan et al . , 2014 ) . Secondly , PLA confirmed the close proximity between NME2 and ICAP - 1 in \u03b2 1 integrin + / + - icap - 1 + / + osteoblasts in contrast to ICAP - 1 \u2013 deficient cells ( Fig . 5 , C and D ) confirming the proximity of these two proteins as previously observed based on classical immunofluorescence ( Fournier et al . , 2002 ) . More importantly , we noticed a decrease of PLA signal between AP - 2 and NME2 and between dynamin and NME2 upon ICAP - 1 loss ( Fig . 5 , C and D ) , suggesting that ICAP - 1 acts as a linker to connect NME2 and CCP components . Similarly , we found that the proximity between NME2 and \u03b2 1 or \u03b2 3 integrin was disrupted in the absence of ICAP - 1 ( Fig . 5 , E and F ) , indi - cating the requirement for ICAP - 1 in keeping \u03b2 integrin and NME2 in close vicinity . Interestingly , ICAP - 1 loss had a negli - gible effect on transferrin receptor / NME2 PLA signal , confirm - ing the specificity of ICAP - 1 \u2019 s function in maintaining NME2 / integrin proximity ( Fig . 5 , E and F ) . In addition , we detected specific interactions between NME1 / 2 and ICAP - 1 , \u03b2 1 , and \u03b2 3 occurring at \u03b1 - adaptin \u2013 marked CCPs in \u03b2 1 integrin + / + - icap - 1 + / + osteoblasts ( Fig . S4 A ) . The functional link between integrin and NME was further supported by the distribution of NME / integrin complexes in the vicinity of vinculin - stained FAs ( Fig . 6 , A and B ) , showing a role for NME / integrin association close to adhesion sites . Indeed , for each proximity interaction , the fraction of hits was significantly higher ( N = 96 for \u03b2 1 integrin , P < 0 . 0001 ; N = 52 for \u03b2 3 integrin , P = 0 . 0014 ; Welch t tests ) than when using randomized images . This finding indicates that hits are unlikely due to stochastic distribution of duolink spots in the cell and thus suggests that a and \u03b2 3 integrin ( LucA . 5 , green ) was performed and analyzed by fluorescent confocal microscopy . Silencing the expression of \u03b2 3 integrin leads to the complete abolishment of the ppMLC - decorated stress fibers in the \u03b2 1 \u2212 / \u2212 / icap - 1 \u2212 / \u2212 and shrinkage of the cell area . Scale bar , 20 \u00b5m . ( B ) Quantification of the ppMLC area normalized to the cellular surface area after treatment with \u03b2 3 integrin siRNA ( si \u03b2 3 , clear colors ) or with scramble siRNA ( si Scr , dark colors ) . Customized particle analysis script from ImageJ was used after application of Unsharpen mask and Despecle filters . The error bars represent SD . N \u2265 42 cells . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( C and D ) Representative traction forces maps ( TFM ) from \u03b2 1 integrin + / + - icap - 1 \u2212 / \u2212 and \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cell lines treated with \u03b2 3 integrin siRNA ( si \u03b2 3 , bottom panels ) or with scramble siRNA ( si Scr , upper panels ) and quantification ( D ) of the total force applied ( nN ) on fibronectin - coated polyacrylamide gel with a defined rigidity of 5 kPa . The additional silencing of \u03b2 3 integrins in the \u03b2 1 \u2212 / \u2212 / icap - 1 \u2212 / \u2212 cells decimates the traction forces , confirming that , in absence of \u03b2 1 integrin and ICAP - 1 , generation of strong cellular con - tractility at adhesion sites is dependent on \u03b2 3 integrins . Scale bar , 10 \u03bc m . Error bars represent SD . N = 80 cells . * * * * , P value \u2264 0 . 0001 . ( E ) FACS analysis of median fluorescence intensity of \u03b2 3 integrin on cell surface in \u03b2 1 integrin KO osteoblast cell lines . Error bars represent SD . ( F ) Western blot showing the efficiency of \u03b2 3 integrin silencing in \u03b2 1 integrin + / + - icap - 1 \u2212 / \u2212 and \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cell lines . Source data are available for this figure : SourceData F2 . Kyumurkov et al . Journal of Cell Biology 6 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure 3 . Size and dynamic of \u03b2 3 integrin FAs are dependent on \u03b2 1 integrins and ICAP - 1 . ( A ) Staining area of \u03b2 3 integrins was carried out on osteoblast cells spread on fibronectin - coated coverglass for 4 h using LucA5 antibody . Scale bar , 20 \u00b5m . ( B ) Quantification of the adhesive area of \u03b2 3 integrins FA , Kyumurkov et al . Journal of Cell Biology 7 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 correlation exists between duolink spots and spatial distribution of FAs . Whereas the presence of ICAP - 1 was necessary to maintain NME close to both \u03b2 1 and \u03b2 3 integrins ( Fig . 5 , E and F ) , \u03b2 1 integrin was not required for NME / \u03b2 3 integrin ( Fig . 6 C ) or NME / ICAP - 1 proximity ( Fig . S4 , B and C ) . Indeed , the proximity between \u03b2 3 integrin and NME was retained even in cells lacking \u03b2 1 integrin ( Fig . 6 , C and D ) . As NME2 is crucial for GTP - loading on dynamin ( Boissan et al . , 2014 ) , and dynamin function pre - cedes the transient recruitment of the clathrin uncoating protein auxilin immediately after CCP scission ( Massol et al . , 2006 ) , we analyzed whether inhibition of NME1 / 2 could affect the burst of auxilin recruitment , a hallmark of CCP scission ( Massol et al . , 2006 ) . Consistent with this hypothesis , recruitment of auxilin to CCPs was strongly impaired when NME2 and NME1 were de - leted ( Fig . S5 , A and B ) . In addition , like NME1 / 2 loss , the loss of ICAP - 1 in osteoblasts also led to a decrease of the auxilin burst at the rim of FA ( Fig . 6 , E and F ) . Altogether , these data indicate that ICAP - 1 is required for NME function at CCPs upstream of dynamin and auxilin - dependent steps to allow optimal forma - tion of clathrin - coated vesicles . The consequences of the inhi - bition of clathrin - dependent endocytosis on force generation were assessed by TFM after clathrin and NME1 / 2 knock - down in \u03b2 1 integrin \u2212 / \u2212 cells and \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells . While deletion of clathrin and NME mimicked ICAP - 1 deletion in \u03b2 1 integrin \u2212 / \u2212 cells cultured on a stiff substrate by rescuing stress fibers and P - myosin ( Fig . 4 , F \u2013 H ) and by decreasing \u03b2 3 integrin turnover ( Fig . 5 B ) , TFM experiments performed on soft gels ( 5 kPa ) did not reveal any significant change in force generation after blocking clathrin - mediated endocytosis in cells devoid of \u03b2 1 in - tegrin . These results reveal an additional activity of ICAP - 1 on \u03b2 3 integrin in soft environments that is not supported by NME and clathrin . In summary , ICAP - 1 regulates integrin endocytosis by con - trolling NME association to integrin - containing CCPs and is required for NME - dependent scission of CCPs . More impor - tantly , ICAP - 1 might regulate cellular force by controlling in - tegrin endocytosis through NME2 - control of dynamin function . Nevertheless , our results suggest that unlike clathrin and NME1 / 2 , ICAP - 1 plays an additional role to rescue force in \u03b2 1 integrin deleted cells cultured on a soft substratum . Discussion Integrin trafficking contributes to both cell movement and ad - hesion receptor signaling ( Caswell et al . , 2009 ; Moreno - Layseca et al . , 2019 ; Lock et al . , 2019 ) . Although essential for many cel - lular processes , the sequence of molecular events during clathrin - mediated integrin endocytosis remains elusive . Our present results implicate a critical contribution of the integrin inhibitor , ICAP - 1 , in cellular mechanotransduction by working in concert with NME2 NDPK to control clathrin - mediated en - docytosis of integrins at the edge of FAs . This new mechanism illustrates the importance of integrin inhibitors in mechano - sensing and mechanoresponsiveness ( Bouin et al . , 2017 ; Lerche et al . , 2020 ) . Mechanistically , ICAP - 1 acts as an adaptor protein in integrin endocytosis . ICAP - 1 mediates cargo selection by po - sitioning NME close to the integrin receptor , together with AP2 and dynamin in order to fuel dynamin at CCP allowing execution of vesicle fission to ensure integrin turnover ( Fig . 7 ) . It has been previously described that the cell mechanical state regulates clathrin coat dynamics at the plasma membrane ( Ferguson et al . , 2017 ; Baschieri et al . , 2018 ) . Our study provides further insights into the importance of controlling integrin turnover to limit the maturation of FAs and tune cell contractility . In addition , our results add another piece of evidence to the conclusion that both \u03b2 1 and \u03b2 3 integrins can control actomyosin contractility ( Roca - Cusachs et al . , 2009 ; Schiller et al . , 2013 ) . While our results clearly confirm the efficiency of \u03b2 1 integrin deletion to abolish traction forces , they also suggest the inefficiency of \u03b2 3 integrin to take over in the presence of ICAP - 1 . Thus , suppression of ICAP - 1 appears to be necessary to unlock a potential inhibition applied to \u03b2 3 integrin to permit its translocation to the cell center , generate cell contractile response and ensure Fn fibril - logenesis . Despite the main function of \u03b2 1 integrin in traction force and fibrillogenesis , \u03b2 3 integrin has already been reported to reinforce force ( Roca - Cusachs et al . , 2009 ) and to stimulate the initial steps of \u03b1 v \u03b2 3 - mediated Fn fibril formation ( Danen et al . , 2002 ) . Moreover , the ability of \u03b1 v \u03b2 3 integrin to drive Fn fibrillogenesis is supported by previous data demonstrating the role of integrin - \u03b1 v \u03b2 3 to trigger Fn assembly and CAF invasion ( Attieh et al . , 2017 ) . Our results demonstrate that ICAP - 1 and NME share the task of \u03b2 3 integrin endocytosis control to regulate integrin cell surface and cell tensional homeostasis . Clathrin - dependent trafficking of integrins is essential for FA disassembly , a process dictated by the intracellular domains of both \u03b1 and \u03b2 integrin subunits which control the nature of the ligand and the activation status of integrin ( Margadant et al . , 2011 ; De Franceschi et al . , 2016 ; Ezratty et al . , 2009 ; Arjonen et al . , 2012 ) . By targeting the cytoplasmic domain of \u03b2 integrin subunit , ICAP - 1 directly participates to FA disassembly ( Bouvard normalized to the cellular surface area . Error bars represent SD . N \u2265 208 cells . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( C ) Quantification of the mean of \u03b2 3 integrin FAs area . The deletion of ICAP - 1 in \u03b2 1 deficient cell line ( grey box ) drives massive leap of \u03b2 3 integrinstained FAs size . Error barsrepresent SD . N \u2265 208 cells . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( D ) Representative time series of the lifetime of eGFP - \u03b2 3 integrin FAs of the four osteoblastic cell lines . Asterix points out typical eGFP - \u03b2 3 integrin FAs in the cells with typical disassembly lifetime . Note the translocation of FA to the cell center in \u03b2 1 integrin \u2212 / \u2212 / icap - 1 \u2212 / \u2212 cells . Scale bar , 2 \u00b5m . ( E ) TIRF lifetime analysis on the eGFP - \u03b2 3 integrin FAs . The lifetime of the eGFP - \u03b2 3 integrin containing FA is increased in \u03b2 1 integrin \u2212 / \u2212 / icap - 1 \u2212 / \u2212 cells . Spinning disk videos of eGFP - \u03b2 3 integrin were taken for the duration of 2 h . The adhesion lifetime was analyzed by Focal Adhesion Analysis Server ( FAAS ; Berginski and Gomez , 2013 ) and verifiedvisually . Error bars represent SD . N \u2265 50focal adhesions . ns , adjustedP value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( F ) FRAP analysis on the eGFP - \u03b2 3 integrin FAs reveal that the \u03b2 3 integrin mobility at the plasma membrane is halted in the absence of ICAP - 1 and \u03b2 1 integrin . At least six FAs per cell lines were bleached and their recovery was monitored for 5 min . Error bars represent SD . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . Kyumurkov et al . Journal of Cell Biology 8 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure 4 . The \u03b2 3 integrin dependent contractility is associated with the defect of \u03b2 3 integrin endocytosis . ( A ) The \u03b2 3 integrin uptake was measured using \u03b2 3 integrin specific antibody ( LucA . 5 ) , coupled with pH - Rhodo . The volume of the endocytotic vesicles per cell volume was measured at 37 and 4\u00b0C . Note the decrease of \u03b2 3 integrin endocytosis in cell lines deleted in ICAP - 1 . The quantification was performed after 45 min incubation . Error bars represent SD . N \u2265 127 cells . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( B ) Representative images of PLA Kyumurkov et al . Journal of Cell Biology 9 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 et al . , 2003 ) . However , as supported by 2 - hybrid approaches and in vitro interaction assays ( Fournier et al . , 2002 ) , ICAP - 1 must also interact with NME2 in order to drive integrin internaliza - tion . Different approaches including proteomic studies have highlighted that NME form a complex with both \u03b2 1 and \u03b2 3 in - tegrin complexes on the one hand ( Kuo et al . , 2011 ; Schiller et al . , 2011 ) , and with AP2 ( Marino et al . , 2013 ) on the other hand , indicating a functional contribution of NME in dynamin - mediated scission of CCPs ( Boissan et al . , 2014 ) . By uncovering the impact of integrin internalization on force generation through ICAP - 1 / NME complex formation , we provide an addi - tional mechanistic facet of the otherwise multifunctional protein ICAP - 1 , which is known to regulate FA disassembly , cell con - tractility , Fn fibrillogenesis , and cell migration ( Millon - Fr\u00b4emillon et al . , 2008 ; Faurobert et al . , 2013 ; Bouin et al . , 2017 ; Lisowska et al . , 2018 ; Brunner et al . , 2011 ) . The versatility of ICAP - 1 needs to be considered to explain the lack of rescue of force generation upon clathrin or NME1 / 2 deletion in \u03b2 1 null cells on soft substrates . While \u03b2 1 null cells do not form large \u03b2 3 FAs and do not produce traction force when plated on a soft substrate ( TFM conditions ) , ICAP - 1 deletion releases the in - teraction site for kindlin which favors \u03b2 3 integrin clustering , a step required for efficient linkage of integrin to the actin cytoskeleton and the development of traction force ( Ye et al . , 2013 ) . Our assumption is that integrin internalization by the ICAP - 1 / NME1 complex in CCPs is spatially controlled and re - quires adhesion sites with minimal clustering of integrins to be effective . According to our experiments ( Figs . 4 , F \u2013 H and 5 B ) , clathrin and NME deletion mimics ICAP - 1 deletion in stiff conditions since substrate stiffness is enough to induce in - tegrin clustering ( Fig . 4 , F \u2013 H ) . Integrin clustering might be important to correctly position the CCP machinery . This hy - pothesis is supported by the distribution of NME / integrin complexes ( Fig . 6 , A and B ) and bursts of auxilin ( Fig . 6 , E and F ) in the vicinity of vinculin - stained FA . This result empha - sizes a role for the ICAP - 1 / NME / integrin trio close to adhesion sites in a stiffness - dependent manner . Of note , ICAP - 1 knock - out revealed a phenotype in bone which might support an important role of ICAP - 1 in stiff conditions ( Bouvard et al . , 2007 ) . So far it was unknown whether individual integrin classes like \u03b1 5 \u03b2 1 and \u03b1 v \u03b2 3 interacting with the same ligand like Fn are internalized using common or specific routes . Because the de - letion of ICAP - 1 led to an increase of both AP2 / \u03b2 1 and AP2 / \u03b2 3 complexes , the partnership between ICAP - 1 and NME2 likely contributes to both \u03b2 1 and \u03b2 3 internalizations . Concomitantly , the control of CCP scission by ICAP - 1 / NME complex directly regulates cell mechanics which has previously been shown to result from the cooperation between \u03b2 1 and \u03b2 3 integrin ( Schiller et al . , 2013 ) . However , we cannot exclude a compensatory effect between \u03b2 1 and \u03b2 3 integrins to generate force since the prox - imity between NME with ICAP - 1 and \u03b2 3 integrin is not condi - tioned by the presence of \u03b2 1 integrin . Importantly , many other endocytic accessory proteins or co - adaptors associate with integrins and CCPs ( Yap and Winckler , 2015 ) . Dab2 and Numb are PTB domain - containing proteins that accumulate at or near FA shortly before their disassembly ( Chao and Kunz , 2009 ; Nishimura and Kaibuchi , 2007 ; Ezratty et al . , 2009 ; Teckchandani et al . , 2009 ) . The physiological relevance of having many adaptors or co - adaptors like ICAP - 1 , Numb , or Dab2 in the same cell might respond either to integrin specificity or to physical properties of the microenvironment . As an illus - tration , Dab2 is not found at \u03b2 3 integrin - mediated FA sites formed on RGD - coated glass . In addition , it has been shown that the development of actomyosin contractility inhibits Dab2 binding to \u03b2 3 integrin . However the loss of cell \u2013 matrix force development is a key determinant for Dab2 binding to \u03b2 3 in - tegrin clusters and the ongoing endocytosis of \u03b2 3 integrin under soft conditions such as on mobile RGD membranes ( Yu et al . , 2015 ) . This suggests a specificity of some adaptors depending on the microenvironment stiffness . So far , no direct interaction between ICAP - 1 and \u03b2 3 integrin has been demonstrated , possibly due to the lack of proper post - translational modifications in classical pull - down assays . Moreover , the nature of this inter - action might be complexified by the requirement of a third partner to bridge \u03b2 3 integrin and ICAP - 1 . However , PLA ex - periments show proximity not only between \u03b2 3 integrin and ICAP - 1 but also between \u03b2 3 integrin and NME , even in the ab - sence of \u03b2 1 integrin . ICAP - 1 might take over for Dab2 in stiffer microenvironment as ICAP - 1 and Dab2 share the same distal performedwith antibodies against ICAP - 1and \u03b2 3 integrin . Red dots denote regions of signal amplification indicating the close proximity between ICAP - 1and \u03b2 3 integrin . PLA performed on ICAP - 1 deficient cell line is used as control . Nuclei are stained in blue with DAPI . ( C ) Representative images of proximity ligation assays performed with antibodies against ICAP - 1 and AP2 or Dynamin 2 . Red dots denote regions of signal amplification indicating that ICAP - 1 belongs to clathrin endocytosis machinery . PLA performed on ICAP - 1 deficient cell line is used as control . Nuclei are stained in blue with DAPI . Scale bar , 20 \u00b5m . ( D ) PLA performedwith antibodiesagainstAP2 and \u03b2 1 integrin or \u03b2 3 integrinand quantification of thenumber of PLAspots percellshowsthat deficiency inICAP - 1leads to increase of the association of \u03b2 1 and \u03b2 3 integrins with AP2 . N \u2265 25 cells / condition from three independent experiments . Scale bars , 20 \u00b5m . Error bars represent SD . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( E ) Immunofluorescence staining of the phospho - myosin ( ppMLC antibody , red ) and \u03b2 3 integrin ( LucA . 5 antibody , green ) . Representative micrographs of the four osteoblast cell lines treated with clathrin siRNA ( si Clat , right ) or scramble siRNA ( si Scr , left ) . Inhibition of clathrin expression in \u03b2 1 integrin \u2212 / \u2212 / icap - 1 + / + cell line rescues cell spreading through \u03b2 3 integrin - mediated FA and development of acto - myosin cytoskeleton ( see also graphs F \u2013 H ) . Scale bar , 20 \u00b5m . ( F ) Quantification of cell spreading area of the four osteoblast cell lines treated with clathrin siRNA ( si Clathrin , light colors ) or scramble siRNA ( si Scr , dark colors ) . Error bars represent SD . N \u2265 27 cells . ns , adjustedP value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( G ) Quantification ofareaof \u03b2 3 integrincontaining FAs normalized to the cellular surface area of the four osteoblast cell lines treated with clathrin siRNA ( si Clathrin , light colors ) or scramble siRNA ( si Scr , dark colors ) . Error bars represent SD . N \u2265 27 cells . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( H ) Quantification of phospho - myosin staining area normalized to the cellular surface area of the four osteoblast cell lines treated with clathrin siRNA ( si Clathrin , light colors ) or scramble siRNA ( si Scr , dark colors ) . Error bars represent SD . N \u2265 27 cells . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . Kyumurkov et al . Journal of Cell Biology 10 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure 5 . ICAP - 1 is required for NME recruitment in clathrin - coated pits and for keeping the proximity between integrin and NME . ( A ) Western blot analysis showing the efficiency of SiNME1 / 2 in osteoblast cells . ( B ) TIRF / FRAP analysis shows that deletion of NME1 / 2 complex by siRNAs ( siNME1 / 2 , light Kyumurkov et al . Journal of Cell Biology 11 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 NPXY motif to interact with integrin . Regarding \u03b2 1 integrin endocytosis , Dab2 and clathrin are not found to activate the \u03b2 1 integrin clusters observed in cells upon adhesion on mobile RGD membranes ( Yu et al . , 2015 ) . This suggests that the integrin turnover might be regulated differentially with respect to the physical properties of the microenvironment , possibly involving ICAP - 1 and Dab2 in stiff and soft microenvironments , respec - tively . Whether specific mechanosensitive pathways and dis - tinct PTB domain integrin adapters are required to control integrin internalization or endocytic frustration ( Lock et al . , 2019 ; Baschieri et al . , 2018 ) , depending on the physical proper - ties of the microenvironment will be the next avenue to explore . As migratory and adhesive behavior is mediated by \u03b2 1 and \u03b2 3 integrin cooperation ( Schiller et al . , 2013 ) , the regulation of their respective intracellular trafficking in a coordinated manner might be likely essential for rapidly and efficiently adapting the responsiveness of migratory cells to extracellular guidance cues . Whether ICAP - 1 coordinates commonly \u03b2 1 and \u03b2 3 integrin en - docytic process to adapt integrin dynamics and traction force generation in a context - dependent manner is a pending question . Materials and methods Antibodies and chemicals Human plasma Fn was purchased from Sigma - Aldrich . Anti \u03b2 3 integrin antibody was purchased from Emfret ( Clone LucA . 5 , # M030 - 0 ) , the double phosphorylated ( T18 / S19 ) myosin light - chain antibody and clathrin antibody were obtained from Cell Signaling , # 3674 and # 4796 , respectively , and the unmodified myosin light - chain antibody and tubulin were purchased from Sigma - Aldrich , # M4401 and # T4026 , respectively . The poly - clonal anti - \u03b2 3 integrin was kindly provided by M . H . Ginsberg ( University of California San Diego , San Diego , CA ) . The Fn antibody was purchased from Millipore ( # AB2033 ) . The trans - ferrin antibody was purchased from Abcam ( # ab82411 ) . The HRP - conjugated antibodies were obtained from Jackson ImmunoResearch \u2014 F ( ab \u2019 ) \u2082 Anti - Rabbit HRP ( # 711 - 036 - 152 ) or anti - mouse IgG , light - chain HRP ( # 115 - 035 - 174 ) . The fluorescent secondary antibodies conjugated with AlexaFluor 488 ( # A - 11063 ) , AlexaFluor 546 ( # A - 11003 ) , or AlexaFluor 633 ( # A - 21053 ) were obtained from Thermo Fisher Scientific . Mouse monoclonal anti - NME2 was purchased from Kamiya Biomedical Company . Rabbit polyclonal pan - NME antibodies ( recognizing both NME1 and NME2 isoforms ) were prepared by affinity pu - rification using purified human recombinant NME1 and NME2 proteins coupled to NHS - activated HiTrap columns . Mouse monoclonal anti - ICAP - 1 \u03b1 antibodies ( 4D1D6 and 9B10 ) were prepared using recombinant His - tagged ICAP - 1 \u03b1 protein as an - tigen ( Fournier et al . , 2002 ) . Phalloidin coupled with Atto 647 was also purchased from Thermo Fisher Scientific ( # A22287 ) . Cell culture Immortalized osteoblasts from icap - 1 \u2212 / \u2212 ; \u03b2 1 integrin flox / flox mice were generated as described previously ( Bouvard et al . , 2007 ) . These cells were infected or not by adenoCre viruses from gene transfer vector core ( University of Iowa ) in order to ob - tain \u03b2 1 integrin - null cells . The icap - 1 null cells were incubated with retroviral particles to obtain rescued cells expressing ICAP - 1 WT . The cells were selected with 1 mg / ml puromycin to produce cell populations with heterogeneous ICAP - 1 expression levels . Cells were maintained in culture in DMEM ( # 31966 - 021 ; Life Tech - nologies ) supplemented with 10 % FBS ( # S1810 - 500 ; Dominique Dutscher ) , 100 U / ml penicillin , and 100 \u00b5g / ml streptomycin ( # P06 - 07100 ; PAN Biotech ) at 37\u00b0C in a 5 % CO 2 - humidified chamber . For all experiments , cells were washed with PBS ( # L0615 - 500 ; Dominique Dutcher ) , detached using trypsin ( # L0615 - 500 ; Dominique Dutcher ) , and treated with 1 mg / ml trypsin inhibitor ( # T6522 ; Sigma - Aldrich ) . Cells were then plated in DMEM containing 10 % FBS for 4 h and then the ap - propriate analysis was carried out . Where needed , a serum - free medium OptiMEM was used ( # 51985 - 026 ; Life Technologies ) as substitute . The four osteoblast clones \u2014 icap - 1 + / + ; icap - 1 \u2212 / \u2212 ; \u03b2 1 integrin floxed / floxed ; \u03b2 1 integrin floxed / floxed ; and icap - 1 \u2212 / \u2212 \u2014 were infected using lentiviral infection system from Invitrogen with pLenti \u2013 murine \u03b2 3 integrin - GFP vector . Western blotting analysis Cells were plated on 50 % confluence and left to spread over - night . The next day , the dishes were washed twice with ice cold PBS and lysed in cold RIPA buffer , supplemented with 1\u00d7 cOmplete protease inhibitors , 5 mM NaF , and 2 mM Na - orthovanadate . After protein quantification through Pierce BCA Protein Assay ( # 23227 ; Thermo Fisher Scientific ) , the samplers were mixed with Laemmli sample buffer ( 0 . 4 % SDS , 20 % glycerol , 120 mM Tris - Cl [ pH 6 . 8 ] , and 0 . 02 % [ w / v ] bro - mophenol blue ) and loaded on electrophoretic PAA gels . Fol - lowing the standard wet blotting protocol , the nitrocellulose colors ) or scramble siRNA ( si Scr , dark colors ) impedes the turnover of eGFP - \u03b2 3 integrins at the plasma membrane in \u03b2 1 integrin \u2212 / \u2212 / icap - 1 + / + cell line . Six FAs per cell were bleached for each experiment . eGFP - \u03b2 3 integrin recovery was monitored for 5 min . 19 cells \u2264 N \u2264 107 cells . Error bars represent SD . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( C and D ) Representative images of PLAs ( C ) and PLA assay quantification ( D ) of the number of PLA spots per cell performed with antibodies against NME and ICAP - 1 or AP2 or Dynamin 2 . Red dots denote regions of signal amplification consistent with NME / ICAP - 1 interaction , NME / AP2 interaction , and NME / dynamin 2 interaction in \u03b2 1 integrin + / + - icap - 1 + / + osteoblast cells ( left ) . PLA performed on ICAP - 1deficient cell line isusedas control . Thedeletion ofICAP - 1induces a decreaseof red dotsinallthree cases indicating the crucial role of ICAP - 1 for keeping NME in clathrin endocytosis machinery ( right ) . Nuclei are stained in blue with DAPI . Scale bar , 20 \u00b5m . Error bars represent SD . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( E and F ) Representative images ( E ) of PLAs with quantification ( F ) performed with antibodies against NME and \u03b2 1 integrin or \u03b2 3 integrin or transferrin receptor . Red dots denote regions of signal amplification consistent with NME / integrin proximity in \u03b2 1 integrin + / + - icap - 1 + / + osteoblast cells ( left ) . The deletion of ICAP - 1 induces a decrease of the number of red dots indicating the crucial role of ICAP - 1 for keeping NME and integrin vicinity ( right ) . PLA performed on ICAP - 1 deficient cell line is used as control . Nuclei are stained in blue with DAPI . Scale bar , 20 \u00b5m . Error bars represent SD . N \u2265 25 cells / condition from three independent experiments . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . Source data are available for this figure : SourceData F5 . Kyumurkov et al . Journal of Cell Biology 12 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure6 . RequirementofICAP - 1 for NME function inCCP upstreamofdynaminand auxillinsteps to allow optimal clathrin - coated vesicle budding in the vicinity offocal adhesion . ( A ) Highlyinclinedillumination fluorescence microscopy analysiswas performed at the basalfaceofadherentcell on FNcoated Kyumurkov et al . Journal of Cell Biology 13 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 membranes ( # 10600003 ; Amersham ) were probed with the appropriate primary antibodies , diluted in 5 % BSA in Tween - TBS ( TTBS ) , and incubated overnight . The membrane was sub - sequently incubated with the appropriate secondary antibodies , also diluted in 5 % BSA in TTBS for 1 h and then developed using Clarity ECL kit ( # 170 - 5061 ; Biorad ) and recorded with ChemiDoc Imaging System and analyzed with ImageLab software . Traction force microscopy The poly - acrylamide hydrogels with defined rigidity of 5 kPa and containing fluorescent microbeads ( # F8783 ; Life Technolo - gies ) were cast in 2 - well LabTeks ( # 154461 ; Thermo Fisher Sci - entific ) , coated with BindSilane ( # GE17 - 1330 - 01 ; Sigma - Aldrich ) and covered with coverslip , coated with Sigmacote ( # SL2 ; Sigma - Aldrich ) . After the polymerization of the polyacrylamide the wells were flooded with water and the coverslips were detached gently . For the functionalization a protocol from ( Przybyla et al . , 2016 ) was used . Briefly , solution of tetrame - thacrylate , N6 , and Irgacure was deposed on the gels and baked under UV light ( 312 nm ) for 5 min . Then , Fn ( 5 \u03bc g / ml ) was de - posed on the gels and incubated at 4\u00b0C overnight . Cells were allowed to adhere and spread 4 h in DMEM containing 10 % FBS and then placed in 4 % FBS . Just before the acquisition , the membrane was stained with red fluorescent membrane marker PKH26 ( # PKH26GL ; Sigma - Aldrich ) . Images were taken using spinning disk microscope iMIC EMCCD Andromeda ( FEI ) equipped with heating chamber , CO 2 installation , using 40\u00d7 magnification oil objective ( Plan - Apochromat 40\u00d7 / 1 . 4 Oil - DIC M27 WD 0 . 13 mm ) . A fluorescent image of the beads with the cell spread on and fluorescent image of the cell membrane was obtained . Then , the culture medium was replaced with pure solution of trypsin and after verification that cells were com - pletely detached the second image of the fluorescent beads were taken . Isolated cells were randomly chosen for each experi - mental condition . Force calculations were performed as previ - ously described ( Tseng et al . , 2011 ) . Briefly , the displacement fields describing the deformation of the PA substrate are de - termined from the analysis of fluorescent beads images before and after removal of the adhering cells with trypsin treatment . The displacement field is obtained by a two - step method con - sisting of particle image velocimetry followed by individual bead tracking ( Butler et al . , 2002 ; Sabass et al . , 2008 ) . A special procedure is used to evaluate displacements in the area of the adhesive pattern where gel deformation is expected to be largest . Depending on the pattern shape , traction forces may be strongly localized leading to large displacements in very small areas . In coverslide . Left panel shows cells immunostained for vinculin ( green ) and \u03b2 1 integrin / NME duolink signal ( PLA event , red ) , and the right panel shows cells stained for vinculin ( green ) and \u03b2 3 integrin / NME duolink signal ( PLA event , red ) . Scale bar , 5 \u00b5m . ( B ) Boxplot representation of the fraction of the hits es - timated for \u03b2 1 integrin / NME - vinculin and for \u03b2 3 integrin / NME - vinculin and for randomized images ( rand . ) . ( C and D ) Representative images of PLA with quantification performed withantibodies against NMEand \u03b2 3 integrin . Reddotsdenote regions of signalamplification revealingICAP - 1 / \u03b2 3 integrin proximity in \u03b2 1 integrin deficient cells . PLA performed on ICAP - 1 deficient cell line is used as control . Nuclei are stained in blue with DAPI . N \u2265 25 cells / condition from three independent experiments . Scale bar , 20 \u00b5m . ( E and F ) Spinning disk highly magnified video micrographs ( every 4 s ) of an FA , ( vinculin - green ) and auxillin bursts in red ( indicated by yellow arrow ) in \u03b2 1 integrin + / + - ICAP - 1 + / + cells . The quantification in F shows that ICAP - 1 deletion slows down the auxillin bursts / FA by 50 % . At least 35 FAs and 3 squares in at least 4 cells per condition were imaged in 3 independent experiments . Error bars represent SD . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . Figure 7 . Recruitment of NME by ICAP - 1 to regulate cellular force by promoting inactive integrin internalization . Scheme representing the sequence of molecular events during clathrin - mediated integrin endocytosis occurring at the edge of FA . ICAP - 1 acts as an adaptor protein in integrin endocytic process . ICAP - 1 mediates cargo selection by positioning NME close to integrin ( 1 ) , AP2 , and dynamin ( 2 ) in order to fuel dynamin at the clathrin - coated pits ( 3 ) allowing execution of subsequent membrane deformation and vesicle fission ( 4 ) to ensure the turnover of integrins . Kyumurkov et al . Journal of Cell Biology 14 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 this case , failure to correctly track a few beads in such areas would significantly alter the calculated force magnitude . Therefore , the pattern area is divided into smaller windows that are allowed to overlap , before applying the cross - correlation and tracking analysis . Reducing the size of the windows makes it possible to retrieve larger displacements with cross - correlation and , using overlapped windows , we can avoid missing beads close to the windows boundaries . All image processing and analysis were performed using Matlab ( Gao and Kilfoil , 2009 ) . To calculate cell - induced traction stress from displacement data , we have used the Fourier - transform traction cytometry ( FTTC ) method ( Sabass et al . , 2008 ) . We kept the regularization pa - rameter at small values ( \u03bb < \u223c 10 \u2212 9 ) in order to maintain the best spatial resolution , which is estimated to be about 50 nm in our case . PAA substrates Polyacrylamide ( PAA ) substrates were prepared on 2 - well LabTeks ( Thermo Fisher Scientific ) with 8 % acrylamide / x % bis - acrylamide and 10 mM Hepes ( pH 8 . 5 ) gel . After two sulfo - sanpah ( Thermo Fisher Scientific ) activation , gels were coated with 5 \u00b5g / ml fi - bronectin ( 1 \u00b5g / cm 2 ) at 4\u00b0C overnight . We used three concen - trations of bis - acrylamide in order to control the rigidity : 0 . 1 % for 3 kPa , 0 . 2 % for 15 kPa , and 0 . 4 % for 60 kPa . Random migration analysis For migration assays , cells were plated on a 2 - well LabTeks containing an Fn - coated PAA substrate ( 10 mm diameter , 100 nm thick ) for 3 h in CO 2 - independent DMEM containing fibronectin - free 4 % FCS . Cells are maintained at 37\u00b0C and imaged on an inverted microscope ( Axiovert 200 ; Zeiss ) equipped with a motorized stage , cooled CCD camera ( CoolSnap HQ2 ; Roper Sci - entific ) , and a live - cell imaging 10\u00d7 objective ( EC Plan - Neofluar ) for 5 h at a frequency of one image per 4 min . Cell velocity was obtained using the manual tracking plugin in ImageJ . 150 \u2013 200 cells were analyzed from at least five different positions per ex - periment , and results were issued from three independent experiments . Focal adhesion lifetime analysis Cells stably expressing \u03b2 3 integrin - GFP were spread in 2 - well LabTeks and left to spread for 4 h . Spinning disk videos were taken for the length of 2 h with 1 min frequency . The lifetime analysis was performed with FA Analysis Server ( Berginski and Gomez , 2013 ) . Immunofluorescence Cells were plated at an approximate density of 6 \u00d7 10 4 per cm 2 for 4 h , fixed with 4 % PFA , permeabilized with 0 . 3 % Triton X - 100 , and blocked with 10 % goat - serum in PBS then with appropriate primary antibodies and after rinsing , with appropriate AlexaFluor - conjugated secondary antibody and AlexaFluor - conjugated phalloidin . Finally , the coverslips were mounted in Mowiol / DAPI solution . Images were acquired on an iMIC EMCCD Andromeda ( FEI ) spinning - disk microscope equipped with an objective Plan - Apochromat 40\u00d7 / 1 . 4 Oil - DIC M27 , WD 0 . 13 mm . siRNA - mediated downregulation of proteins in osteoblast cells Cells were plated in 6 - well plates at low density \u2014 6 \u00d7 10 3 cells per cm 2 and left to spread overnight . The next day they were transfected with the appropriate siRNA using RNAiMAX system ( Thermo Fisher Scientific ) . The medium was changed the next day , and a second hit with the same siRNA was performed . The transfected cells were used in 24 h after . All used siRNA were purchased from Dharmacon as follows : siRNA against \u03b2 3 in - tegrin \u2013 L - 040746 - 01 - 0005 ; siRNA against clathrin \u2013 L - 063954 - 00 , siRNA against NME1 \u2013 L - 040142 - 00 , and siRNA against NME2 - L - 040143 - 00 ; for all experiments we used a non - targeting siRNA as control \u2013 D - 001810 - 10 - 20 . FRAP analysis Cells stably expressing \u03b2 3 integrin \u2013 GFP were spread in 2 - well LabTeks and left to spread for 4 h . FRAP videos were taken using multimodal microscope for photo manipulations equipped with TIRF 63\u00d7 objective ( \u03b1 - Plan - Apochromat 63\u00d7 / 1 . 46 Oil Korr M27 , WD 0 . 1 mm ) . The analysis was carried out using build in FRAP analysis module in the FEI offline analysis . Image analysis and statistical tests For ppMLC staining or surface analysis , we measured the nec - essary signal using a thresholding method with manual correc - tion when needed . More than 30 cells were measured in each condition that allowed us to do a non - parametric Kruskal \u2013 Wallis test ( non - parametric ) followed by Wilcoxon test with a Bon - ferroni correction when KW tests were significant ( using GraphPad ) ; experiments were carried out at least three times . For FA analysis , we measured the \u03b2 3 integrin staining signal of at least 50 cells per experiment and three independent experi - ments using the particle analyzer of ImageJ software . Particles over 10 nm 2 were analyzed . The total adhesive area per cell was analyzed by a Kruskal \u2013 Wallis test ( non - parametric ) followed by Wilcoxon test with a Bonferroni correction when KW tests were significant ; the mean area of FA was analyzed by an ANOVA - 2 analysis and Tukey \u2019 s HSD post - hoc tests ( using GraphPad ) . Fibronectin fibrillogenesis 100 , 000 cells were plated on LabTek glass slide 4 chambers and allowed to adhere for 24 h in serum - free medium made of Op - tiMEM . Then , cells were fixed with 4 % PFA . Cells were stained with appropriate primary antibodies anti - Fn , anti - \u03b2 3 integrin ( LucA . 5 ) , and with appropriate AlexaFluor - conjugated second - ary antibody and AlexaFluor - conjugated phalloidin . Then , the coverslips were mounted in Mowiol solution . The length of in - dividual fibers was determined manually in ImageJ . To assess Fn coverage , the images were processed with fast Fourier trans - form bandpass filters to visualize all fibers , and the amount of fibronectin was measured by thresholding using Fiji ( Schindelin et al . , 2012 ) . Fluorescent integrin antibody uptake assays LucA5 antibody was labeled using SiteClick Antibody Azido Modification Kit ( Thermo Fisher Scientific ) and Click - iT pHrodo iFL Red sDIBO Alkyne for Antibody Labeling ( Thermo Fisher Scientific ) following the manufacturer \u2019 s instructions . Cells were Kyumurkov et al . Journal of Cell Biology 15 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 spread on Fn - coated LabTek slides in DMEM , supplemented with 10 % FBS and 1 % P / S . Labeled antibody ( 10 \u00b5g / ml ) was added to the cells for 45 min at 37 or 4\u00b0C , and cells were fixed in 4 % PFA and permeabilized . Actin staining was performed in order to detect cell contour . Images were acquired on an iMIC EMCCD Andromeda ( FEI ) spinning - disk microscope equipped with an objective Plan - Apochromat 40\u00d7 / 1 . 4 Oil - DIC M27 , WD 0 . 13 mm , and analyzed using the 3D Object Counter in ImageJ . Statistical analysis was performed in R using one - way analysis of variance ( ANOVA ) and Tukey \u2019 s HSD post - hoc test . In situ PLA To monitor the subcellular localization of protein \u2013 protein in - teractions at single molecule resolution , an in situ PLA was performed as previously described ( S\u00a8oderberg et al . , 2006 ) . Cells grown on coverslips were fixed with cold methanol and then incubated with primary antibodies . Secondary antibodies tagged with short DNA oligonucleotides were added . Hybridization , ligation , amplification , and detection were realized according to the manufacturer \u2019 s protocol ( Sigma - Aldrich ) . Briefly , secondary antibodies were incubated in preheated humidity chamber for 1 h at 37\u00b0C . Ligation was performed with a ligase - containing li - gation solution for 30 min at 37\u00b0C . Finally , amplification step was performed with a polymerase - containing amplification so - lution for 1 h 40 min at 37\u00b0C . After the PLA reaction , coverslips were further incubated with AlexaFluor 488 - conjugated IgGs and Cy5 - conjugated IgGs to detect proteins corresponding to the primary antibodies used . PLA signal corresponds to the Cy3 fluorescence . Coverslips were analyzed on an inverted wide - field microscope ( Leica DM4B , Camera Leica DFC 3000 G , Ob - jectif X63 Leica Germany 506185 ) . Highly inclined illumination fluorescence microscopy ( Imaging , 2010 ) was used to analyze the spatial distribution of \u03b2 1 / \u03b2 3 integrin \u2013 NME proximity complexes revealed by the PLA technology in cells labeled for vinculin to assess the proximity of NME / integrin complexes with FA . Images have been analyzed by a homemade script in ImageJ ( code available in Data S1 ) to detect duolink spots ( using D . Sage watershed algorithm ; Teaching by Doing , 2003 ) and FA , and assess their spatial proximity . The principle was to enlarge ( by three pixels ) regions corresponding to detected duolink spots following the method described in Montagnac et al . ( 2013 ) . Then , the intensity of fluorescence in the second channel corresponding to the vin - culin staining is measured to determine any colocalization . A \u201c hit \u201d is obtained in case of a positive result . The ratio of those hits to the total number of duolink spots is calculated for each cell . For each proximity complexes , the fraction of hits is sig - nificantly higher than when using randomized images . This indicates that the hits are unlikely to be due to stochastic dis - tribution of duolink spots in the cell , and thus suggests that a correlation exists between duolink spots and FA spatial distribution . Auxillin flashes measurement The four osteoblast clones were electroporated using AMAXA nucleofector Kit V ( protocol X - 001 ) according to the manu - facturer \u2019 s instructions in order to transfect them with two plasmids encoding , respectively , GFP - tagged Auxillin and the mCherry - tagged Paxillin . 100 , 000 cells per condition were plated on fluorodishes . In the following days , cells were imaged at 2 s intervals for 5 min using a spinning disk microscope ( Andor ) based on a CSU - W1 Yokogawa head mounted on the lateral port of an inverted IX - 83 Olympus microscope equipped with a 60\u00d7 1 . 35 NA UPLSAPO objective lens and a laser combiner system , which included 491 and 561 nm 100 mW DPSS lasers ( Andor ) . Images were acquired with a Zyla sCMOS camera ( Andor ) . The system was steered by IQ3 software ( Andor ) . Paxillin - positive FA were manually selected in Fiji and their area was measured . Auxillin flashes appearing at the border of FAs were manually counted . Subsequently , a square - shaped region of 2 . 5 \u00d7 2 . 5 \u00b5m 2 was placed in the center of the cell in a region where FAs were absent in order to count Auxillin flashes outside of FAs . At least 35 FAs and 3 squares in at least 4 cells per condition were imaged in three independent experiments . Re - sults are expressed as Auxillin flashes per second divided by the FA area or by the area of the square - shaped region of 2 . 5 \u00d7 2 . 5 \u00b5m 2 and normalized to the value obtained in \u03b2 1 integrin + / + - icap - 1 + / + osteoblasts . Statistical analyses were performed in SigmaPlot using one - way ANOVA followed by All Pairwise Multiple Comparison Procedure ( Holm - Sidak method ) . Fluorescence - activated cell sorting Cells were gently detached with trypsin and then treated with trypsin inhibitor ( # T0256 ) . Then , they were placed in round - bottom 96 - well plate and blocked with 1 % BSA in PBS for 30 min at 37\u00b0C . Then , cells were incubated with the appropriate anti - bodies and secondary antibodies as control diluted in PBS / 1 % BSA for 30 min on ice . After subsequent incubation with sec - ondary antibodies cells were fixed in 4 % PFA for 10 min and surface staining was detected with BD Accuri C6 flow cytometer and analyzed with the provided software . Online supplemental material Fig . S1 shows the characterization of osteoblast cells used in the study . Fig . S2 shows the role of ICAP - 1 in cellular mechano - transduction . Fig . S3 shows the efficiency of clathrin silencing in osteoblast cell lines . Fig . S4 shows NME / \u03b2 3 integrin complex at \u03b1 - adaptin \u2013 marked CCPs and its independence with respect to \u03b2 1 integrin . Fig . S5 shows the requirement of NME in CCP up - stream of dynamin and auxillin steps to allow optimal clathrin - coated vesicle budding . Video 1 shows the lifetime of eGFP - \u03b2 3 integrin FAs in \u03b2 1 integrin + / + / icap - 1 + / + cells . Video 2 shows the lifetime of eGFP - \u03b2 3 integrin FAs in \u03b2 1 integrin + / + / icap - 1 \u2212 / \u2212 cells . Video 3 shows the lifetime of eGFP - \u03b2 3 integrin FAs in \u03b2 1 integrin \u2212 / \u2212 / icap - 1 + / + cells . Video 4 shows the lifetime of eGFP - \u03b2 3 integrin FAs in \u03b2 1 integrin \u2212 / \u2212 / icap - 1 \u2212 / \u2212 cells . Data S1 provides the homemade image script used in this paper . Acknowledgments We thank Isabelle Tardieux and Kate Miroshnikova for re - viewing the manuscript . We thank Daniel Bouvard and Reinhard Fassler for providing osteoblast cells , Alexei Grichine , Jacques Mazzega , and Myl\u00e8ne Pezet for their technical assistance on the Kyumurkov et al . Journal of Cell Biology 16 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 microcell Imaging platform from the Institute for Advanced Biosciences , and Agnieszka Kawska ( http : / / www . IlluScientia . com ) for artwork associated with Fig . 7 . We thank Christiane Oddou for technical assistance . We are grateful to Dr . M . H . Ginsberg for providing polyclonal antibodies specific for \u03b2 3 integrin . A . Kyumurkov and M . R\u00b4egent - Kloeckner are doctoral fellows funded from the French Minist\u00e8re de l \u2019 Enseignement Sup\u00b4erieur et de la Recherche . M . Proponnet - Guerault is a recipient of a fellowship from the Ligue Nationale contre le Cancer ( LNCC ) . This research was funded by the Agence Nationale de la Re - cherche grant ( CODECIDE ANR - 17 - CE13 - 022 , StrepB2brain , ANR - 17 - CE15 - 0026 - 01 ) , by the Fondation pour la Recherche sur le Cancer and by the Fondation pour la Recherche M\u00b4edicale grant ( DEQ20170336702 ) to C . Albiges - Rizo and the Ligue Contre le Cancer R\u00b4egionale Auvergne - Rh \u02c6 one - Alpes et Sa \u02c6 one - et - Loire ( 17021 ) to A . - P . Bouin . The authors declare no competing financial interests . Author contributions : A . Kyumurkov and A . - P . Bouin did most of the experiments , analyzed the data , and prepared the figures . M . Proponnet - Guerault contributed in performing en - docytosis experiments . M . R\u00b4egent - Kloeckner contributed to performing quantitative analyses of phase - contrast live - cell imaging data . M . Balland and A . Nicolas contributed to gener - ating TFM data . G . Montagnac , F . Baschieri , and P . Chavrier contributed to setting up Auxillin flash experiments . S . Manet contributed to FACS analysis and O . Destaing to TIRF and FRAP analyses . M . Boissan , C . Calmel , F . Waharte , and P . Chavrier contributed to performing PLA experiments . C . Albiges - Rizo , E . Planus , A . P . Bouin , G . Montagnac , P . Chavrier , O . Destaing , and M . Boissan contributed with material and critical discussions . C . Albiges - Rizo and E . Planus supervised the entire work . C . Albi - ges - Rizo and E . Planus initiated the topic and wrote the manu - script together with A . P . Bouin . Submitted : 3 April 2020 Revised : 22 July 2022 Accepted : 27 September 2022 References Albiges - Rizo , C . , O . Destaing , B . Fourcade , E . Planus , and M . R . Block . 2009 . Actin machinery and mechanosensitivity in invadopodia , podosomes and focal adhesions . J . Cell Sci . 122 : 3037 \u2013 3049 . https : / / doi . org / 10 . 1242 / jcs . 052704 Arjonen , A . , J . Alanko , S . Veltel , and J . Ivaska . 2012 . Distinct recycling of active and inactive \u03b2 1 integrins . Traffic . 13 : 610 \u2013 625 . https : / / doi . org / 10 . 1111 / j . 1600 - 0854 . 2012 . 01327 . x Attieh , Y . , A . G . Clark , C . Grass , S . Richon , M . Pocard , P . Mariani , N . Elkhatib , T . Betz , B . Gurchenkov , and D . M . Vignjevic . 2017 . Cancer - associated fibroblasts lead tumor invasion through integrin - \u03b2 3 - dependent fibro - nectin assembly . J . Cell Biol . 216 : 3509 \u2013 3520 . https : / / doi . org / 10 . 1083 / jcb . 201702033 Ballestrem , C . , B . Hinz , B . A . Imhof , and B . Wehrle - Haller . 2001 . Marching at the front and dragging behind : Differential alphaVbeta3 - integrin turnover regulates focal adhesion behavior . J . Cell Biol . 155 : 1319 \u2013 1332 . https : / / doi . org / 10 . 1083 / jcb . 200107107 Baschieri , F . , S . Dayot , N . Elkhatib , N . Ly , A . Capmany , K . Schauer , T . Betz , D . M . Vignjevic , R . Poincloux , and G . Montagnac . 2018 . Frustrated en - docytosis controls contractility - independent mechanotransduction at clathrin - coated structures . Nat . Commun . 9 : 3825 . https : / / doi . org / 10 . 1038 / s41467 - 018 - 06367 - y Berginski , M . E . , and S . M . Gomez . 2013 . The focal adhesion analysis server : A web tool for analyzing focal adhesion dynamics . F1000Res . 2 : 68 . https : / / doi . org / 10 . 12688 / f1000research . 2 - 68 . v1 Boissan , M . , G . Montagnac , Q . Shen , L . Griparic , J . Guitton , M . Romao , N . Sauvonnet , T . Lagache , I . Lascu , G . Raposo , et al . 2014 . Membrane trafficking . Nucleoside diphosphate kinases fuel dynamin superfamily proteins with GTP for membrane remodeling . Science . 344 : 1510 \u2013 1515 . https : / / doi . org / 10 . 1126 / science . 1253768 Boissan , M . , U . Schlattner , and M . L . Lacombe . 2018 . The NDPK / NME su - perfamily : State of the art . Lab . Invest . 98 : 164 \u2013 174 . https : / / doi . org / 10 . 1038 / labinvest . 2017 . 137 Bouin , A . - P . , A . Kyurmurkov , M . R\u00b4egent - Kloeckner , A . - S . Ribba , E . Faurobert , H . - N . Fournier , I . Bourrin - Reynard , S . Manet - Dup\u00b4e , C . Oddou , M . Bal - land , et al . 2017 . ICAP - 1 monoubiquitylation coordinates matrix density andrigiditysensingforcell migration throughROCK2 \u2013 MRCK \u03b1 balance . J . Cell Sci . 130 : 626 \u2013 636 . https : / / doi . org / 10 . 1242 / jcs . 200139 Bouvard , D . , A . Aszodi , G . Kostka , M . R . Block , C . Albig\u00e8s - Rizo , and R . F\u00a8assler . 2007 . Defective osteoblast function in ICAP - 1 - deficient mice . Develop - ment . 134 : 2615 \u2013 2625 . https : / / doi . org / 10 . 1242 / dev . 000877 Bouvard , D . , L . Vignoud , S . Dup\u00b4e - Manet , N . Abed , H . N . Fournier , C . Vincent - Monegat , S . F . Retta , R . F\u00a8assler , and M . R . Block . 2003 . Disruption of focal adhesions by integrin cytoplasmic domain - associated protein - 1 alpha . J . Biol . Chem . 278 : 6567 \u2013 6574 . https : / / doi . org / 10 . 1074 / jbc . M211258200 Brunner , M . , A . Millon - Fr\u00b4emillon , G . Chevalier , I . A . Nakchbandi , D . Mosher , M . R . Block , C . Albig\u00e8s - Rizo , and D . Bouvard . 2011 . Osteoblast mineral - ization requires beta1 integrin / ICAP - 1 - dependent fibronectin deposi - tion . J . Cell Biol . 194 : 307 \u2013 322 . https : / / doi . org / 10 . 1083 / jcb . 201007108 Butler , J . P . , I . M . Tolic - Norrelykke , B . Fabry , and J . J . Fredberg . 2002 . Traction fields , moments , and strain energy that cells exert on their surround - ings . Am . J . Physiol . Cell Physiol . 282 : C595 \u2013 C605 . https : / / doi . org / 10 . 1152 / ajpcell . 00270 . 2001 Caswell , P . T . , S . Vadrevu , and J . C . Norman . 2009 . Integrins : Masters and slaves of endocytic transport . Nat . Rev . Mol . Cell Biol . 10 : 843 \u2013 853 . https : / / doi . org / 10 . 1038 / nrm2799 Ceresa , B . P . , and S . L . Schmid . 2000 . Regulation of signal transduction by endocytosis . Curr . Opin . Cell Biol . 12 : 204 \u2013 210 . https : / / doi . org / 10 . 1016 / s0955 - 0674 ( 99 ) 00077 - 0 Chao , W . - T . , and J . Kunz . 2009 . Focaladhesiondisassemblyrequiresclathrin - dependent endocytosis of integrins . FEBS Lett . 583 : 1337 \u2013 1343 . https : / / doi . org / 10 . 1016 / j . febslet . 2009 . 03 . 037 Dammai , V . , B . Adryan , K . R . Lavenburg , and T . Hsu . 2003 . Drosophila awd , the homolog of human nm23 , regulates FGF receptor levels and func - tions synergistically with shi / dynamin during tracheal development . Genes Dev . 17 : 2812 \u2013 2824 . https : / / doi . org / 10 . 1101 / gad . 1096903 Danen , E . H . J . , P . Sonneveld , C . Brakebusch , R . Fassler , and A . Sonnenberg . 2002 . The fibronectin - binding integrins alpha5beta1 and alphavbeta3 differentially modulate RhoA - GTP loading , organization of cell matrix adhesions , and fibronectin fibrillogenesis . J . Cell Biol . 159 : 1071 \u2013 1086 . https : / / doi . org / 10 . 1083 / jcb . 200205014 Degani , S . , F . Balzac , M . Brancaccio , S . Guazzone , S . F . Retta , L . Silengo , A . Eva , and G . Tarone . 2002 . The integrin cytoplasmic domain - associated protein ICAP - 1 binds and regulates Rho family GTPases during cell spreading . J . Cell Biol . 156 : 377 \u2013 387 . https : / / doi . org / 10 . 1083 / jcb . 200108030 Engler , A . J . , S . Sen , H . L . Sweeney , and D . E . Discher . 2006 . Matrix elasticity directs stem cell lineage specification . Cell . 126 : 677 \u2013 689 . https : / / doi . org / 10 . 1016 / j . cell . 2006 . 06 . 044 Ezratty , E . J . , C . Bertaux , E . E . Marcantonio , and G . G . Gundersen . 2009 . Clathrin mediates integrin endocytosis for focal adhesion disassembly in migrating cells . J . Cell Biol . 187 : 733 \u2013 747 . https : / / doi . org / 10 . 1083 / jcb . 200904054 Faurobert , E . , C . Rome , J . Lisowska , S . Manet - Dup\u00b4e , G . Boulday , M . Mal - bouyres , M . Balland , A . - P . Bouin , M . K\u00b4eramidas , D . Bouvard , et al . 2013 . CCM1 - ICAP - 1 complex controls \u03b2 1 integrin - dependent endothelial contractility and fibronectin remodeling . J . Cell Biol . 202 : 545 \u2013 561 . https : / / doi . org / 10 . 1083 / jcb . 201303044 Ferguson , J . P . , S . D . Huber , N . M . Willy , E . Ayg\u00fcn , S . Goker , T . Atabey , and C . Kural . 2017 . Mechanoregulationofclathrin - mediatedendocytosis . J . Cell Sci . 130 : 3631 \u2013 3636 . https : / / doi . org / 10 . 1242 / jcs . 205930 Fournier , H . - N . , S . Dup\u00b4e - Manet , D . Bouvard , M . - L . Lacombe , C . Marie , M . R . Block , and C . Albiges - Rizo . 2002 . Integrin cytoplasmic domain - associated protein 1alpha ( ICAP - 1alpha ) interacts directly with the metastasis suppressor nm23 - H2 , and both proteins are targeted to newly formed cell adhesion sites upon integrin engagement . J . Biol . Chem . 277 : 20895 \u2013 20902 . https : / / doi . org / 10 . 1074 / jbc . M200200200 Kyumurkov et al . Journal of Cell Biology 17 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Fournier , H . - N . , S . Dup\u00b4e - Manet , D . Bouvard , F . Luton , S . Degani , M . R . Block , S . F . Retta , and C . Albiges - Rizo . 2005 . Nuclear translocation of integrin cyto - plasmic domain - associated protein 1 stimulates cellular proliferation . Mol . Biol . Cell . 16 : 1859 \u2013 1871 . https : / / doi . org / 10 . 1091 / mbc . E04 - 08 - 0744 De Franceschi , N . , A . Arjonen , N . Elkhatib , K . Denessiouk , A . G . Wrobel , T . A . Wilson , J . Pouwels , G . Montagnac , D . J . Owen , and J . Ivaska . 2016 . Se - lective integrin endocytosis is driven by interactions between the in - tegrin \u03b1 - chain and AP2 . Nat . Struct . Mol . Biol . 23 : 172 \u2013 179 . https : / / doi . org / 10 . 1038 / nsmb . 3161 Gao , Y . , and M . L . Kilfoil . 2009 . Accurate detection and complete tracking of large populations of features in three dimensions . Opt . Express . 17 : 4685 \u2013 4704 . https : / / doi . org / 10 . 1364 / OE . 17 . 004685 Imaging , M . O . 2010 . PH880 topics in physics modern optical imaging ( Fall 2010 ) . Nat . Methods . 5 : 1 \u2013 7 . https : / / doi . org / 10 . 1038 / NMETH . 1171 Jacquemet , G . , A . Stubb , R . Saup , M . Miihkinen , E . Kremneva , H . Hamidi , and J . Ivaska . 2019 . Filopodome mapping identifies p130Cas as a mechano - sensitive regulator of filopodia stability . Curr . Biol . 29 : 202 \u2013 216 . e7 . https : / / doi . org / 10 . 1016 / j . cub . 2018 . 11 . 053 Klapholz , B . , and N . H . Brown . 2017 . Talin \u2013 the master of integrin adhesions . J . Cell Sci . 130 : 2435 \u2013 2446 . https : / / doi . org / 10 . 1242 / jcs . 190991 Krishnan , K . S . , R . Rikhy , S . Rao , M . Shivalkar , M . Mosko , R . Narayanan , P . Etter , P . S . Estes , and M . Ramaswami . 2001 . Nucleoside diphosphate kinase , a source of GTP , is required for dynamin - dependent synaptic vesicle recycling . Neuron . 30 : 197 \u2013 210 . https : / / doi . org / 10 . 1016 / s0896 - 6273 ( 01 ) 00273 - 2 Kuo , J . - C . , X . Han , C . - T . Hsiao , J . R . Yates , and C . M . Waterman . 2011 . Analysis of the myosin - II - responsive focal adhesion proteome reveals a role for \u03b2 - Pix in negative regulation of focal adhesion maturation . Nat . Cell Biol . 13 : 383 \u2013 393 . https : / / doi . org / 10 . 1038 / ncb2216 Legate , K . R . , and R . F\u00a8assler . 2009 . Mechanisms that regulate adaptor binding tobeta - integrincytoplasmictails . J . CellSci . 122 : 187 \u2013 198 . https : / / doi . org / 10 . 1242 / jcs . 041624 Leiss , M . , K . Beckmann , A . Gir\u00f3s , M . Costell , and R . F\u00a8assler . 2008 . The role of integrin binding sites in fibronectin matrix assembly in vivo . Curr . Opin . Cell Biol . 20 : 502 \u2013 507 . https : / / doi . org / 10 . 1016 / j . ceb . 2008 . 06 . 001 Lerche , M . , A . Elosegui - Artola , J . Z . Kechagia , C . Guzm\u00b4an , M . Georgiadou , I . Andreu , D . Gullberg , P . Roca - Cusachs , E . Peuhu , and J . Ivaska . 2020 . Integrinbinding dynamics modulateligand - specific mechanosensing in mammary gland fibroblasts . iScience . 23 : 100907 . https : / / doi . org / 10 . 1016 / j . isci . 2020 . 100907 Lisowska , J . , C . J . R\u00a8odel , S . Manet , Y . A . Miroshnikova , C . Boyault , E . Planus , R . De Mets , H . - H . Lee , O . Destaing , H . Mertani , et al . 2018 . The CCM1 \u2013 CCM2complexcontrolscomplementaryfunctionsofROCK1andROCK2thatarerequiredforendothelialintegrity . J . Cell Sci . 131 : jcs216093 . https : / / doi . org / 10 . 1242 / jcs . 216093 Lock , J . G . , F . Baschieri , M . C . Jones , J . D . Humphries , G . Montagnac , S . Str\u00a8omblad , and M . J . Humphries . 2019 . Clathrin - containing adhesion complexes . J . Cell Biol . 218 : 2086 \u2013 2095 . https : / / doi . org / 10 . 1083 / jcb . 201811160 Margadant , C . , H . N . Monsuur , J . C . Norman , and A . Sonnenberg . 2011 . Mechanisms of integrin activation and trafficking . Curr . Opin . Cell Biol . 23 : 607 \u2013 614 . https : / / doi . org / 10 . 1016 / j . ceb . 2011 . 08 . 005 Marino , N . , J . - C . Marshall , J . W . Collins , M . Zhou , Y . Qian , T . Veenstra , and P . S . Steeg . 2013 . Nm23 - h1 binds to gelsolin and inactivates its actin - severing capacity to promote tumor cell motility and metastasis . Cancer Res . 73 : 5949 \u2013 5962 . https : / / doi . org / 10 . 1158 / 0008 - 5472 . CAN - 13 - 0368 Massol , R . H . , W . Boll , A . M . Griffin , and T . Kirchhausen . 2006 . A burst of auxilin recruitment determines the onset of clathrin - coated vesicle uncoating . Proc . Natl . Acad . Sci . USA . 103 : 10265 \u2013 10270 . https : / / doi . org / 10 . 1073 / pnas . 0603369103 De Mets , R . , I . Wang , M . Balland , C . Oddou , P . Moreau , B . Fourcade , C . Al - biges - Rizo , A . Delon , and O . Destaing . 2019 . Cellular tension encodes local Src - dependent differential \u03b2 1 and \u03b2 3 integrin mobility . Mol . Biol . Cell . 30 : 181 \u2013 190 . https : / / doi . org / 10 . 1091 / mbc . E18 - 04 - 0253 Millon - Fr\u00b4emillon , A . , D . Bouvard , A . Grichine , S . Manet - Dup\u00b4e , M . R . Block , and C . Albiges - Rizo . 2008 . Cell adaptive response to extracellular ma - trix density is controlled by ICAP - 1 - dependent beta1 - integrin affinity . J . Cell Biol . 180 : 427 \u2013 441 . https : / / doi . org / 10 . 1083 / jcb . 200707142 Millon - Fr\u00b4emillon , A . , M . Brunner , N . Abed , E . Collomb , A . - S . Ribba , M . R . Block , C . Albig\u00e8s - Rizo , and D . Bouvard . 2013 . Calcium and calmodulin - dependent serine / threonine protein kinase type II ( CaMKII ) - mediated intramolecular opening of integrin cytoplasmic domain - associated protein - 1 ( ICAP - 1 \u03b1 ) negatively regulates \u03b2 1 integrins . J . Biol . Chem . 288 : 20248 \u2013 20260 . https : / / doi . org / 10 . 1074 / jbc . M113 . 455956 Milloud , R . , O . Destaing , R . deMets , I . Bourrin - Reynard , C . Oddou , A . Delon , I . Wang , C . Albig\u00e8s - Rizo , and M . Balland . 2017 . \u03b1 v \u03b2 3 integrins negatively regulate cellular forces by phosphorylation of its distal NPXY site . Biol . Cell . 109 : 127 \u2013 137 . https : / / doi . org / 10 . 1111 / boc . 201600041 Montagnac , G . , V . Meas - Yedid , M . Irondelle , A . Castro - Castro , M . Franco , T . Shida , M . V . Nachury , A . Benmerah , J . - C . Olivo - Marin , and P . Chavrier . 2013 . \u03b1 TAT1 catalyses microtubule acetylation at clathrin - coated pits . Nature . 502 : 567 \u2013 570 . https : / / doi . org / 10 . 1038 / nature12571 Moreno - Layseca , P . , J . Icha , H . Hamidi , and J . Ivaska . 2019 . Integrin traf - ficking in cells and tissues . Nat . Cell Biol . 21 : 122 \u2013 132 . https : / / doi . org / 10 . 1038 / s41556 - 018 - 0223 - z Murrell , M . , P . W . Oakes , M . Lenz , and M . L . Gardel . 2015 . Forcing cells into shape : The mechanics of actomyosin contractility . Nat . Rev . Mol . Cell Biol . 16 : 486 \u2013 498 . https : / / doi . org / 10 . 1038 / nrm4012 Nallamothu , G . , J . A . Woolworth , V . Dammai , and T . Hsu . 2008 . Awd , the homolog of metastasis suppressor gene Nm23 , regulates Drosophila epithelial cell invasion . Mol . Cell . Biol . 28 : 1964 \u2013 1973 . https : / / doi . org / 10 . 1128 / MCB . 01743 - 07 Nishimura , T . , andK . Kaibuchi . 2007 . Numbcontrolsintegrinendocytosisfor directional cell migration with aPKC and PAR - 3 . Dev . Cell . 13 : 15 \u2013 28 . https : / / doi . org / 10 . 1016 / j . devcel . 2007 . 05 . 003 Przybyla , L . , J . N . Lakins , R . Sunyer , X . Trepat , and V . M . Weaver . 2016 . Monitoring developmental force distributions in reconstituted embry - onicepithelia . Methods . 94 : 101 \u2013 113 . https : / / doi . org / 10 . 1016 / j . ymeth . 2015 . 09 . 003 Renz , M . , C . Otten , E . Faurobert , F . Rudolph , Y . Zhu , G . Boulday , J . Duchene , M . Mickoleit , A . - C . Dietrich , C . Ramspacher , et al . 2015 . Regulation of \u03b2 1 integrin - Klf2 - mediated angiogenesis by CCM proteins . Dev . Cell . 32 : 181 \u2013 190 . https : / / doi . org / 10 . 1016 / j . devcel . 2014 . 12 . 016 Roca - Cusachs , P . , N . C . Gauthier , A . Del Rio , andM . P . Sheetz . 2009 . Clustering of alpha ( 5 ) beta ( 1 ) integrins determines adhesion strength whereas al - pha ( v ) beta ( 3 ) and talin enable mechanotransduction . Proc . Natl . Acad . Sci . USA . 106 : 16245 \u2013 16250 . https : / / doi . org / 10 . 1073 / pnas . 0902818106 Rossier , O . , V . Octeau , J . - B . Sibarita , C . Leduc , B . Tessier , D . Nair , V . Gatter - dam , O . Destaing , C . Albig\u00e8s - Rizo , R . Tamp\u00b4e , etal . 2012 . Integrins \u03b2 1and \u03b2 3 exhibit distinct dynamic nanoscale organizations inside focal adhe - sions . Nat . Cell Biol . 14 : 1057 \u2013 1067 . https : / / doi . org / 10 . 1038 / ncb2588 Sabass , B . , M . L . Gardel , C . M . Waterman , and U . S . Schwarz . 2008 . High res - olution traction force microscopy based on experimental and compu - tational advances . Biophys . J . 94 : 207 \u2013 220 . https : / / doi . org / 10 . 1529 / biophysj . 107 . 113670 Schiller , H . B . , C . C . Friedel , C . Boulegue , and R . F\u00a8assler . 2011 . Quantitative proteomics of the integrin adhesome show a myosin II - dependent re - cruitment of LIM domain proteins . EMBO Rep . 12 : 259 \u2013 266 . https : / / doi . org / 10 . 1038 / embor . 2011 . 5 Schiller , H . B . , M . - R . Hermann , J . Polleux , T . Vignaud , S . Zanivan , C . C . Friedel , Z . Sun , A . Raducanu , K . - E . Gottschalk , M . Th\u00b4ery , et al . 2013 . \u03b2 1 - and \u03b1 v - classintegrins cooperate toregulate myosin IIduring rigiditysensingof fibronectin - based microenvironments . Nat . Cell Biol . 15 : 625 \u2013 636 . https : / / doi . org / 10 . 1038 / ncb2747 Schindelin , J . , I . Arganda - Carreras , E . Frise , V . Kaynig , M . Longair , T . Pietzsch , S . Preibisch , C . Rueden , S . Saalfeld , B . Schmid , et al . 2012 . Fiji : An open source platform for biological image analysis . Nat . Methods . 9 : 676 \u2013 682 . https : / / doi . org / 10 . 1038 / nmeth . 2019 Scita , G . , and P . P . Di Fiore . 2010 . The endocytic matrix . Nature . 463 : 464 \u2013 473 . https : / / doi . org / 10 . 1038 / nature08910 Shattil , S . J . , C . Kim , and M . H . Ginsberg . 2010 . The final steps of integrin activation : The endgame . Nat . Rev . Mol . CellBiol . 11 : 288 \u2013 300 . https : / / doi . org / 10 . 1038 / nrm2871 S\u00a8oderberg , O . , M . Gullberg , M . Jarvius , K . Ridderstr \u02da ale , K . J . Leuchowius , J . Jarvius , K . Wester , P . Hydbring , F . Bahram , L . G . Larsson , and U . Landegren . 2006 . Direct observation of individual endogenous protein complexes in situ by proximity ligation . Nat . Methods . 3 : 995 \u2013 1000 . https : / / doi . org / 10 . 1038 / nmeth947 Teaching by Doing . 2003 Teckchandani , A . , N . Toida , J . Goodchild , C . Henderson , J . Watts , B . Wollsc - heid , and J . A . Cooper . 2009 . Quantitative proteomics identifies a Dab2 / integrinmoduleregulatingcellmigration . J . CellBiol . 186 : 99 \u2013 111 . https : / / doi . org / 10 . 1083 / jcb . 200812160 Tseng , Q . , I . Wang , E . Duchemin - Pelletier , A . Azioune , N . Carpi , J . Gao , O . Filhol , M . Piel , M . Th\u00b4ery , and M . Balland . 2011 . A new micropatterning method of soft substrates reveals that different tumorigenic signals can promote or reduce cell contraction levels . Lab Chip . 11 : 2231 \u2013 2240 . https : / / doi . org / 10 . 1039 / c0lc00641f Kyumurkov et al . Journal of Cell Biology 18 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Wehrle - Haller , B . 2012 . Assembly and disassembly of cell matrix adhesions . Curr . Opin . Cell Biol . 24 : 569 \u2013 581 . https : / / doi . org / 10 . 1016 / j . ceb . 2012 . 06 . 010 Wu , C , V . M . Keivens , T . E . O \u2019 Toole , J . A . McDonald , and M . H . Ginsberg . 1995 . Integrin activation and cytoskeletal interaction are essential for the assembly of a fi - bronectinmatrix . Cell . 83 : 715 \u2013 724 . https : / / doi . org / 10 . 1016 / 0092 - 8674 ( 95 ) 90184 - 1 Yap , C . C . , and B . Winckler . 2015 . Adapting for endocytosis : Roles for endo - cytic sorting adaptors in directing neural development . Front . Cell . Neurosci . 9 : 119 . https : / / doi . org / 10 . 3389 / fncel . 2015 . 00119 Ye , F . , B . G . Petrich , P . Anekal , C . T . Lefort , A . Kasirer - friede , S . J . Shattil , R . Ruppert , M . H . Ginsberg , M . Moser , and R . Fa . 2013 . The mechanism of kindlin - mediated activation of integrin \u03b1 IIb \u03b2 3 . Curr Biol . 23 : 2288 \u2013 2295 . https : / / doi . org / 10 . 1016 / j . cub . 2013 . 09 . 050 Yu , C . H . , N . B . Rafiq , F . Cao , Y . Zhou , A . Krishnasamy , K . H . Biswas , A . Ravasio , Z . Chen , Y . - H . Wang , K . Kawauchi , et al . 2015 . Integrin - beta3 clusters recruit clathrin - mediated endocytic machinery in the absence of traction force . Nat . Commun . 6 : 8672 . https : / / doi . org / 10 . 1038 / ncomms9672 Zamir , E . , M . Katz , Y . Posen , N . Erez , K . M . Yamada , B . Z . Katz , S . Lin , D . C . Lin , A . Bershadsky , Z . Kam , and B . Geiger . 2000 . Dynamics and segregation of cell - matrix adhesions in cultured fibroblasts . Nat . Cell Biol . 2 : 191 \u2013 196 . https : / / doi . org / 10 . 1038 / 35008607 Zhang , X . A . , and M . E . Hemler . 1999 . Interaction of the integrin beta1 cyto - plasmic domain with ICAP - 1 protein . J . Biol . Chem . 274 : 11 \u2013 19 . https : / / doi . org / 10 . 1074 / jbc . 274 . 1 . 11 Kyumurkov et al . Journal of Cell Biology 19 of 19 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Supplemental material Kyumurkov et al . Journal of Cell Biology S1 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure S1 . \u03b2 3 integrin expression level is unchanged upon loss of \u03b2 1 integrin . ( A \u2013 C ) Western blot of total cell lysate ( A ) and quantification ( B and C ) confirmed the deletion of \u03b2 1 integrin or ICAP - 1 in osteoblast cells . Tubulin is used as loading control . ( E ) The deletion of \u03b2 1 integrin and ICAP - 1 do not affect the expression of \u03b2 3 integrin . Actin is used as loading control . ( D ) The expression of \u03b2 3 integrin mRNA is not changed upon deletion of \u03b2 1 integrin or ICAP - 1 . Luminescence signal is normalized to the \u03b2 1 integrin + / + - icap - 1 + / + . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( F ) Intensity of ppMLC staining decorating the stress fibers , normalized to cell area . Error bars represent SD . N \u2265 211 cells . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . ( G ) The level ofthe double phosphorylation ( T18 / S19 ) of the MLC was assessed and quantified via Western blot against the total level of MLC of cell lysates of cells spread for 4 h on fibronectin - covered glass . Error bars represent SD . ns , adjusted P value > 0 . 05 ; * , P value \u2264 0 . 05 ; * * , P value \u2264 0 . 01 ; * * * , P value \u2264 0 . 001 ; * * * * , P value \u2264 0 . 0001 . Source data are available for this figure : SourceData FS1 . Kyumurkov et al . Journal of Cell Biology S2 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure S2 . ICAP - 1 is required for cells to adapt cell velocity as function of substrate stiffness independently on the presence of \u03b2 1 integrin and is involved in \u03b2 3 integrin FA translocation to the cell center . ( A ) Osteoblast cells were spread on FN - coated PAA gels of different rigidities ( 3 , 15 , and 60 kPa ) . Cell migration was monitored for 5 h using time - lapse microscopy . The cell velocity was determined by individually tracking 200 \u2013 300 cells in three inde - pendent experiments . ICAP - 1 deficient cells displayed a constant migration velocity whatever the substrate rigidity as compared to the \u03b2 1 integrin + / + - icap - 1 + / + cells , independently of the presenceof \u03b2 1 integrinhighlighting the crucial roleofICAP - 1inrigidity sensing . Error bars represent SD . N \u2265 199 cells . * , P < 0 . 05 ; * * , P < 0 . 005 ; * * * , P < 0 . 0005 . ( B ) Note that the additional deletion of ICAP - 1 in cells depleted in \u03b2 1 integrin induces an increase of \u03b2 3 integrin FA translocation to the cell center . Quantification of the percentage of FA whose distance to the nearest cell contour point is > 12 . 5 % of the average diameter of the cell . Error bars represent SD . N \u2265 208 cells . * * * * , P value \u2264 0 . 0001 . Kyumurkov et al . Journal of Cell Biology S3 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure S3 . Efficiency of clathrin silencing in osteoblast cell lines . Clathrin and actin are visualized by Western blot after treatment with siRNA against clathrin as compared to Scramble conditions . Source data are available for this figure : SourceData FS3 . Kyumurkov et al . Journal of Cell Biology S4 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Figure S4 . NME / \u03b2 3 integrin interaction occurs at \u03b1 - adaptin \u2013 marked CCPs and does not need \u03b2 1 integrin in osteoblasts . ( A ) PLA signal using NME antibody in combination with ICAP - 1 antibody ( 4d1d6 ) or ICAP - 1 antibody ( 9b10 ) or \u03b2 3 integrin antibody or \u03b2 1 integrin antibody in WT osteoblasts . Im - munolabeling of \u03b1 - adaptin was performed after PLA . Insets are higher magnification of boxed regions . Insets show PLA signals ( NME / ICAP - 1 or NME / \u03b2 1 in - tegrin or NME / \u03b2 3 integrin ) colocalizing with \u03b1 - adaptin - positive CCPs ( arrowheads ) . PLA signal indicates a close proximity , possibly in situ interaction of NME with ICAP - 1 , \u03b2 1 , and \u03b2 3 integrins at CCPs . Scale bar , 10 \u03bc m . Inset scale bar , 5 \u03bc m . ( B and C ) Representative images of PLAs ( B ) and PLA assay quantification ( C ) of the number of PLA spots per cell performed with antibodies against NME and \u03b2 3 integrin . Red dots denote regions of signal amplification consistent with NME / \u03b2 3 integrin proximity in osteoblast cells deficient or not in \u03b2 1 integrin . PLA performed on ICAP - 1 and \u03b2 1 integrin deficient cell line is used as control . Scale bar , 20 \u03bc m . Error bars represent SD . N = 15 cells / condition from three independent experiments . ns , adjusted P value > 0 . 05 . Kyumurkov et al . Journal of Cell Biology S5 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023 Video 1 . \u03b2 1 integrin + / + - icap - 1 + / + cells expressing \u03b2 3 integrin - GFP were spread in 2 - well LabTeks and left to spread for 4 h . Spinning disk videos were taken for the length of 2 h with 1 min frequency . Playback speed , 1 frame / s . Video 2 . \u03b2 1 integrin + / + - icap - 1 \u2212 / \u2212 cells expressing \u03b2 3 integrin - GFP were spread in 2 - well LabTeks and left to spread for 4 h . Spinning disk videos were taken for the length of 2 h with 1 min frequency . Playback speed , 1 frame / s . Video 3 . \u03b2 1 integrin \u2212 / \u2212 - icap - 1 + / + cells expressing \u03b2 3 integrin - GFP were spread in 2 - well LabTeks and left to spread for 4 h . Spinning disk videos were taken for the length of 2 h with 1 min frequency . Playback speed , 1 frame / s . Video 4 . \u03b2 1 integrin \u2212 / \u2212 - icap - 1 \u2212 / \u2212 cells expressing \u03b2 3 integrin - GFP were spread in 2 - well LabTeks and left to spread for 4 h . Spinning disk videos were taken for the length of 2 h with 1 min frequency . Playback speed , 1 frame / s . Provided online is a dataset . Data S1 provides the homemade image script used in this paper . Figure S5 . Requirement of NME in CCP upstream of dynamin and auxillin steps to allow optimal clathrin - coated vesicle budding . ( A ) Still image and kymographs of simultaneous two - color TIRF - M time series of mock - or siRNA - treated BSC - 1 cells overexpressing mRFP - LCa and GFP - auxilin ( 1 image / s ) . In mock - treated cells , vertical arrowhead points to a burst of GFP - auxilin coincident with mRFP - LCa signal disappearing . In siRNA - treated cells , arrowhead points to mRFP - LCa signal disappearing without associating GFP - auxilin . ( B ) Percentage of disappearing mRFP - LCa CCPs ending with a burst of GFP - auxilin in mock - and siRNA - treated BSC - 1 cells . At least 300 events were analyzed for each condition . Kyumurkov et al . Journal of Cell Biology S6 Force tuning by ICAP - 1 and NME cooperation https : / / doi . org / 10 . 1083 / jcb . 202004025 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 1 / e202004025 / 1440807 / j c b _ 202004025 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 23 F eb r ua r y 2023",
    "yuDistributedAnalogicalIdea2016": "Distributed Analogical Idea Generation with Multiple Constraints Lixiu Yu , Robert E . Kraut , Aniket Kittur Carnegie Mellon University 5000 Forbes Avenue , Pittsburgh , PA 15213 { lixiuyu , robert . kraut , nkittur } @ cs . cmu . edu ABSTRACT Previous work has shown the promise of crowdsourcing analogical idea generation , where distributing the stages of analogical processing across many people can reduce fixation , identify inspirations from more diverse domains , and lead to more creative ideas . However , prior work has only considered problems with a single constraint , while many real - world problems involve multiple constraints . This paper contributes a systematic crowdsourcing approach for eliciting multiple constraints inherent in a problem and using those constraints to find inspirations useful in solving it . To do so we identify methods to elicit useful constraints at different levels of abstraction , and empirical results that identify how the level of abstraction influences creative idea generation . Our results show that crowds find the most useful inspirations when the problem domain is represented abstractly and constraints are represented more concretely . Author Keywords Constraint ; inspiration ; problem - solving ; idea generation ACM Classification Keywords H . 5 . 3 Group and Organization Interfaces INTRODUCTION The use of analogy has historically driven innovation in science , technology , and design , in which inspirations from distant domains help problem solvers identify mechanisms that would not be apparent in the target problem\u2019s original domain [ 4 , 5 , 10 , 15 , 16 ] . For example , in 2013 , a group of engineers and a world - renowned origami expert designed a large solar array to be carried by a narrow rocket . Using origami - folding techniques they were able to fold the array to a tenth of its deployed size , solving a 50 - year old problem [ 20 ] . A rich literature in psychology , engineering , and design has investigated the methods by which individuals and small groups use analogies to innovate new approaches , solutions , and products [ 5 , 6 , 7 , 11 , 15 , 17 ] . One thread of this research has introduced systematic methods for distributing the analogical idea generation process to crowds [ 28 , 29 ] . For example , even the non - expert workers on Amazon\u2019s Mechanical Turk can search product databases ( such as the thousands of unselected product submissions to Quirky ) to find inspirations that contain useful mechanisms for solving a target problem in a different domain [ 28 , 29 ] . A key enabling factor in this approach is abstracting the target problem and its constraints into a schema : a structured representation that removes domain - specific information while keeping the essential structure of the original problem . Such domain - agnostic representations reduce fixation on surface features , and can help focus search on inspirations that have structural similarity to the target problem . While successful , such approaches have to date been limited by reliance on a single schema embodying a single set of constraints , i . e . , requirements that a solution must satisfy to be successful . In contrast , many product design and engineering problems involve multiple , often conflicting constraints [ 3 , 21 , 23 , 24 , 26 ] . For example , the design of a kindergarten chair might require multiple constraints , such as its safety ( e . g . , preventing it from tipping over or pinching fingers ) and flexibility ( e . g . , making it easy to move or stack ) . Each of these constraints could be represented as its own schema and at multiple levels of representation ( e . g . , safety vs . preventing pinched fingers ) . Generalizing distributed idea generation to more complex real - world problems requires a deeper understanding of how to elicit and represent multiple constraints . In this paper , we propose a distributed process for eliciting the constraints inherent in a problem at multiple levels of abstraction and using those constraints to find inspirations useful in solving that problem . In doing so we explore how to operationalize constraints in ways that are useful for problem solving and how this can be done using non - expert crowds . We test our approach through two randomized experiments in which crowd workers generate constraints , find inspirations , and solve problems while we manipulate Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page . Copyrights for components of this work owned by others than ACM must be honored . Abstracting with credit is permitted . To copy otherwise , or republish , to post on servers or to redistribute to lists , requires prior specific permission and / or a fee . Request permissions from Permissions @ acm . org . CSCW ' 16 , February 27 - March 02 , 2016 , San Francisco , CA , USA \u00a9 2016 ACM . ISBN 978 - 1 - 4503 - 3592 - 8 / 16 / 02 $ 15 . 00 DOI : http : / / dx . doi . org / 10 . 1145 / 2818048 . 2835201 1236 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA the level of abstraction of the problem description ( i . e . , domain independent or dependent descriptions ) and of the constraints ( i . e . , more abstract or more concrete descriptions of the constraints ) . Our results suggest that crowd workers can identify important constraints for a product and can find inspirational examples that are useful for satisfying those constraints . The best results occur by representing the problem domain at a relatively abstract level but the constraints at a more concrete level . Under these conditions , people find the most diverse set of inspirational examples , yet ones that are most relevant to the initial problem . People shown these inspirational examples generate more original and practical ideas to solve the original problem . LEVELS OF ABSTRACTION Our research goal is to identify the levels of abstraction that are useful for product design in terms of both problem context ( e . g . , designing a kindergarten chair ) and problem constraints ( e . g . , safety , flexibility ) . Abstracting a problem context into a schema in order to find analogical solutions from distant domains can be a highly effective approach for increasing innovation in domains ranging from scientific discovery to creative problem - solving to product development [ 5 , 6 , 17 ] . For example , people are better able to solve Duncker\u2019s radiation problem [ 12 ] , in which a doctor needs to destroy a tumor with a large dose of X - rays without destroying surrounding tissue , if they are given as inspiration the case of a military general who conquers a castle surrounded by mined roads by dividing his army into small groups that are light enough to evade the mines and who attack from different directions . However , this inspiration is most useful if the problem solvers represent the solution abstractly , in terms of a divide - and - converge strategy , which maps the high - level structure shared by the problem and solution description [ 13 ] . Even though the medical and military domains differ radically in surface detail , an abstract divide - and - converge schema maps the general\u2019s strategy of dividing troops in transit into small groups before converging at the castle with the tumor problem , where the doctor divides the x - ray into multiple small beams converging on the tumor . Although this example was taken from cognitive science research , both empirical observation and experimental studies in product development find a positive relationship between using analogical inspiration and the originality of the product concept [ e . g . , 17 , 22 ] . Recently , Yu et al . [ 29 ] developed an approach to distributing the schema abstraction process across multiple individuals , showing that searching for inspirations and generating ideas with crowds who had not seen the original problem domain or description decreased fixation and increased solution creativity . However , the level of abstraction of the problem constraints is also likely to affect the diversity and relevance of inspirations found . Searching for inspirations relevant to \u201csafety\u201d might generate more diverse inspirational examples than \u201cpreventing pinched fingers\u201d , but might also lead to solutions less applicable to the design of kindergarten chairs . Real world idea generation often requires solving \u201cwicked problems\u201d [ 2 ] , which can involve satisfying multiple constraints simultaneously , rather than developing a single insight that characterizes Duncker\u2019s x - ray problem . Stripping away too much detail during abstraction may cause problem solvers to consider mechanisms that are not relevant to their problem or prevent them from recognizing the precise mechanisms they need to solve a constraint in a particular domain . For example , in Duncker\u2019s radiation problem , the relevance of solutions is constrained because x - rays travel in straight lines ; analogies from ant colonies involving ants who can follow arbitrary paths might thus not be relevant to reasoning about x - rays . Even though abstracting away problem context allows problem solvers to identify a wider range of potential solutions , actually applying the solutions to the original problem domain may require more detailed knowledge of the constraints that must be met . Thus we hypothesize that while describing a problem context may be most useful if the domain or context is abstracted away ( to reduce functional fixation ) , the representation of constraints may be most useful when they are concrete enough to suggest specific potential mechanisms ( e . g . , \u201cprevent tipping over\u201d rather than \u201csafety\u201d ) while simultaneously being general enough to allow multiple possible mechanisms ( e . g . , preventing something from tipping over by attaching it to a stable object or introducing a low center of gravity ) . To test these hypotheses , we first developed a procedure to elicit constraints from crowd workers for designing a kindergarten chair . Using these constraints , we conducted an experiment in which participants found inspirations using either an abstract or concrete representation of the problem context and constraints ( in a 2x2 design ) , and then subsequent participants used these inspirations to generate new solutions to the design problem . Below we describe how we elicited constraints using crowds , and then describe the experiment procedure and results . ELICITING CONSTRAINTS One key question we aim to address here is how to elicit and represent constraints in a way that can be used by non - expert crowd workers to effectively search for inspirations , which in turn would be useful in solving a target problem . We conducted a series of pilot studies asking workers from Amazon\u2019s Mechanical Turk [ 18 ] to generate constraints for designing a good kindergarten chair , using the probe question : \u201c We are planning to design creative chairs for kindergarten kids . What aspects or constraints should be considered ? \u201d The responses we got back varied significantly in their level of detail . Some of them were quite abstract , such as \u201c Safety . As we all know kids is very fragile so the safety is the most important thing to have \u201d . Others were much more concrete , such as \u201c chairs should have to be shockproo f \u201d . We found two prevalent types of constraints , which corresponded to the level of abstraction 1237 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA workers used to describe the constraint . Abstract constraints such as \u201cthey should be safe for kids to sit on\u201d , can be satisfied in many different ways . In contrast , more concrete constraints , such as adding stabilizers , round off sharp corners , implied a much smaller number of ways they could be satisfied . Based on these findings , we iteratively developed a process to elicit abstract and concrete design constraints ( see Table 1 ) . To elicit abstract constraints , we first recruited 27 crowd workers to brainstorm constraints for kindergarten chairs . As noted previously , their responses varied in abstractness . To make the response set more homogeneous and abstract , we asked a new group of crowd workers to summarize each response in a word or two : \u201c Which design aspect of the chair is the following suggestion referring to ? Summarize it in one or two words \u201d . We then treated these keywords as the abstract constraints , because they shifted the representation of the initially brainstormed constraints to a more abstract level . After a consolidation step , in which the experimenters eliminated duplicates and synonyms , there were seven abstract constraints : safety , flexibility , pedagogy aesthetics , cost , comfort , and ease of cleaning . To elicit concrete constraints , we asked seven new groups of roughly 15 crowd workers each to generate concrete constraints for each abstract constraint , by framing them in terms of lower - level requirements . In this task , the crowd workers received one of the abstract constraint key words , such as safety , and described components of safety or another abstract constraint should be considered when designing a kindergarten chairs . The instructions were to eliciting more concrete version of the safety constraint were , \u201c We are planning to design kindergarten chairs for 5 - 8 year - old children . In this task , we would like you to brainstorm the constraints kindergarten chairs should meet . Please identify a constraint by filling out the following sentences . Note that constraints are requirements on certain dimensions rather than solutions . Regarding safety , a kindergarten chair should meet the requirement that _ _ _ _ _ _ _ _ _ _ _ _ \u201d . This task returned a set of lower - level , more concrete constraints for each abstract constraint . For example , for the safety constraint , participants identified concrete constraints such as \u201cchairs shouldn\u2019t tip over easily\u201d and \u201cchairs shouldn\u2019t have sharp edges\u201d . While more concrete than general categorical constraints such as safety , such constraints still support being instantiated in multiple ways ; for example , a chair could be prevented from tipping over by lowering its center of gravity or fixing it to the floor . As each group described similar concrete constraints in somewhat different ways , we presented the lower - level constraints for an abstract constraint to a new group of crowd workers and asked them to select the ones that were similar and summarize them in terms of a single constraint . In the domain - independent condition we identified a list of domain related terms such as \u201cchair\u201d , \u201ckindergarten\u201d , \u201ckids\u201d , and \u201cclassroom\u201d . The experimenters removed these words and replaced them with more general , domain - independent terms , such as \u201cobjects\u201d . EXPERIMENT 1 : SEARCHING FOR INSPIRATION The previous section described a process through which crowd workers identified a variety of important constraints for a product category at multiple levels of abstraction . Moreover , simple transformations , in particular , substituting domain - specific vocabulary with more general terms , provided a method to express constraints independent of the problem domain . We used these more or less abstract constraints presented with and without reference to the problem domain as inputs for the experiment described below . The goal of this experiment was to better understand conditions under which abstract representations of problems aid or hinder people\u2019s ability to identify solutions to them . We distinguish between the functional requirements of a problem or its constraints , on one hand , and the problem Steps Constraint examples Step 1 : Eliciting abstract constraints 1 ) chairs shouldn\u2019t hurt kids ; 2 ) chairs should have to be shockproof \u00e0\uf0e0 Safety 1 ) comfortable , because kids need to be healthy ; 2 ) Cozy , we all like cozy furniture \u00e0\uf0e0 Comfort Step 2 : Eliciting concrete constraints Safety \u00e0\uf0e0 1 ) The chair should not have sharp edges ; 2 ) The chair won ' t tip over Comfort \u00e0\uf0e0 1 ) The chair should be easy for kids to get on and off ; 2 ) The chair should have appropriate angle giving the correct strain free posture Domain - independent condition Safety : 1 ) The object should not have sharp edges ; 2 ) The object won ' t tip over Comfort : 1 ) The object should be easy to get on and off ; 2 ) The object should have appropriate angle giving the correct strain free posture . Table 1 . Workflow for generating constraints . Constraints Domain Abstract Concrete Abstract Domain - independent abstract constraint Domain - independent concrete constraint Concrete Domain - dependent abstract constraint Domain - dependent concrete constraint Table 2 . Conditions in Experiment 1 . 1238 SESSION : CROWD INNOVATION AND CROWDFUNDING context or domain , on the other . We hypothesize that abstracting away the problem domain will improve search for useful inspiration by preventing functional fixation , leading searchers to find a richer and more diverse set of potentially relevant solutions . However , we hypothesize that concrete representations of constraints will be more useful than abstract representations , because they help searchers more completely understand the type of problem they are trying to solve and to evaluate the relevance of potential solutions for that problem . To test these hypotheses , we designed a 2X2 experiment with the four conditions shown in Table 2 . In Experiment 1 , searchers were given descriptions of the kindergarten chair design problem varying in abstractness and domain specificity and asked to find examples that could inspire solutions . The description of the constraints that solutions needed to meet were presented either more abstractly ( e . g . , ways to make objects safe ) or concretely ( e . g . , ways to make objects that won\u2019t tip over ) . Similarly , the description of the problem domain was also presented relatively abstractly ( i . e . , in domain independent manner , without mentioning kindergarten chairs ) or more concretely ( i . e . , in a domain dependent manner , mentioning kindergarten chairs ) . Henceforth , laid out in Table 2 , for clarity we refer to the abstractness of constraints using the terms \u201cabstract\u201d and concrete , and refer to the abstractness of domains using the terms \u201cdomain - independent\u201d to refer to more abstract domain and the term \u201cdomain - dependent\u201d to refer to the more concrete domains . Participants Overall , 158 Amazon Mechanical Turk workers participated in the experiment . Forty - nine percent were women , and 91 % were native English speakers . Their average age was 32 and ranged from 18 to 58 . Design and Procedure After participants accepted the task , they were randomly assigned to one of the four experimental conditions in Table 2 . They were asked to search for an inspirational source for product design tasks . The task charged them with satisfying two abstract constraints , safety and flexibility , or four concrete constraints : two related to safety ( Tipping over , Pinching fingers ) and the other two related to flexibility ( Moving around , Folding ) . The task in the domain - dependent abstract constraint condition was described as designing a kindergarten chair\u2019s that was either safe or flexible . For example , the instructions for this task with the safety constraint was . \u201c We are looking for inspirations for novel and useful designs for kindergarten chairs that are safe for kids to use . Go to the Internet and retrieve a picture or description of something that has been designed to be safe to use and that could be relevant to the design of kindergarten chairs . \u201d The task in the domain - dependent concrete constraint condition was described as designing a kindergarten chair in terms of one of the following four concrete constraints ( tipping over , pinching fingers , moving around , or folding ) . For example , \u201c Go to the Internet and retrieve a picture or description of something that has been designed to easily move around and that could be relevant to the design of kindergarten chairs . \u201d Note that participants in all domain - dependent conditions were explicitly instructed to not restrict their search to the world of kids\u2019 chairs , and that their results would be judged partially on novelty : \u201c Please search broadly . The examples can come from the domains that are outside of kids chairs . Your search results will be judged on whether it can inspire novel and useful design . \u201d In the domain - independent conditions , no mention was made of kindergarten chair . Specifically , we replaced domain related terms such as \u201cchair\u201d or \u201ckids\u201d with more general terms such as \u2018people\u201d or \u201cobjects\u201d . For example , instructions in the domain - independent abstract constraint condition were , \u201c We are looking for inspirations for novel and useful designs for objects that are safe to use . Go to the Internet and retrieve a picture or description of something that has been designed to be safe to use \u201d . The instructions in the domain - independent concrete constraint conditions were , \u201cGo to the Internet and retrieve a picture or description of something that has been designed to prevent tipping over . Please search broadly . The examples can come from any domains . Your search results will be judged on whether it can inspire novel and useful design . \u201d Participants were asked to provide a link to the inspiration they found , an explanation why they selected this inspirational example , and a description of how they searched for it . Participants returned roughly 40 examples in each condition . Measuring Distance and Diversity To measure the distance between the original problem domain of kindergarten chairs to the examples searchers returned , two judges blind to experimental condition rated each example on a 7 - point Likert scale : \u201c How different is the above kindergarten chair design problem from the domain of the example ? \u201d By this metric , for example , a glass was considered further from a kindergarten chair than was a bench . The judges achieved high inter - rater reliability ( Intraclass Correlation Coefficient ( ICC ) = 0 . 85 ) [ 9 ] . The final distance score was calculated by averaging the scores of the two judges . Because the distribution of the scores was bimodal , we converted distance from the kindergarten - chair domain into a binary variable based on a median split . Examples above the median score were considered \u201cfar\u201d ( 1 ) and those below the median were considered \u201cnear\u201d ( 0 ) . 1239 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA We also calculated a diversity score by counting how many unique domains were found within in each condition . In counting unique domains we combined those from highly similar inspirations ; for example , all the examples shown in the first column of Table 3 were classified as coming from the chair domain while the last two examples from the fourth column , ( the Russian eggs and the cup ) , where from two different domains . Analysis and Results Summary statistics for the data are shown in Table 4 . For analysis purposes , the four conditions were represented as two dummy variables : domain abstractness and constraint concreteness . Domain independent descriptions were coded as abstract ( 1 ) while domain - dependent descriptions were coded as 0 . Problem descriptions with constraints described at a less abstractness level ( e . g . , tipping over , pinching fingers , moving around , and folding up ) were coded as having concrete constraints ( 1 ) , while descriptions with more abstract constraints ( e . g . , safety or flexibility ) were coded as 0 . We ran a logistic regression analysis with binary Distance as the dependent variable regressed on domain abstractness , constraint concreteness , and their interaction . The results show the odds of finding an example from a far domain were substantially higher when the domain was described abstractly rather than concretely ( odd ratio = 214 . 86 , p < 0 . 001 ) . This is consistent with previous literature showing that representing a problem abstractly reduces fixation and increases the likelihood of considering far domains . There was no significant effect of constraint concreteness on domain distance ( b = 1 . 20 , p = 0 . 15 ) , nor a significant interaction effect ( b = - 1 . 66 , p = . 0 . 14 ) . For domain diversity , we see a similar main effect of domain abstractness : the odds of an example coming from a unique domain were substantially higher when problems were described in an abstract , domain independent way ( odds ratio = 26 . 31 , p < 0 . 001 ) . In addition , we also observe a significant negative interaction between domain abstractness and constraint concreteness , with the effect of domain abstractness significantly higher when constraints were also presented abstractly ( b = - 1 . 77 , p < 0 . 05 ) . These results indicate that Domain - independent abstract constraint lead to the highest diversity of domains explored . These results are visually reflected in Table 3 , which shows examples of inspirations found in the four conditions . As these examples illustrate , in the domain - dependent conditions , most examples were about chairs and furniture . In contrast , the domain - independent conditions returned a wider variety of potential examples , including glasses , 3D printer techniques , and kids tools . We also analyzed how participants searched for inspirations in the different conditions . In the domain - dependent conditions , most search involved domain terms . For example , participants described the ways they searched for examples , \u201cI was searching google for \" fun kids chairs \" \u201d , \u201ccreative seating for classrooms\u201d , and \u201csearch for tip proof chair\u201d . When participants searched with domain - independent abstract constraints , the ways of conducting search were vague . For example , \u201c I have search on Google for different kind of images related to novel inspiration for kids \u201d and \u201c I did an image search for adaptable images for inspiration \" . In contrast , when people searched with domain - independent concrete constraints , the search terms were much more concrete while without the fixation on chairs , including \u201c I found the example by typing designs for objects that don ' t pinch fingers \u201d , and \u201c I googled objects that don ' t tip over easily \u201d . These different ways of searching resulted from different representations explained the findings . Domain - dependent , abstract constraint Domain - dependent , concrete constraint Domain - independent , abstract constraint Domain - independent , concrete constraint Table 3 . Examples of inspirations found when the domains and constraints were presented concretely and abstractly . Conditions N Avg Distance % far % unique domains Domain - dependent , abstract constraint 40 2 . 36 5 % 30 % Domain - dependent , concrete constraint 40 2 . 69 15 % 35 % Domain - independent , abstract constraint 37 5 . 80 92 % 92 % Domain - independent , concrete constraint 41 6 . 20 88 % 71 % Table 4 . Distance of inspirations from the kindergarten furniture domain and their diversity in Experiment 1 . 1240 SESSION : CROWD INNOVATION AND CROWDFUNDING EXPERIMENT 2 : GENERATING CREATIVE SOLUTIONS The results from Experiment 1 suggest that the level of abstraction of the problem domain is a key driver of exploration : the more abstract the domain description , the more distant and diverse the domains explored . However , simply exploring more distant and diverse domains is not sufficient to find useful inspirations ; to be useful , the solutions suggested by the inspirations must also be relevant and applicable to the target domain ( here , kindergarten chairs ) . Observation of the rightmost two columns of Table 3 suggests that while both domain - independent conditions led to inspirations in distant domains , those found in the abstract - constraint condition ( third column ) have various safety designs that do not seem relevant to chair safety . In contrast , the inspirations found with concrete constraints ( fourth column ) suggest a variety of relevant mechanisms to prevent tipping over . We designed an experiment to test the usefulness of the inspirations in generating creative solutions to the target problem . In the experiment , we asked people to design kindergarten chairs after showing them four examples randomly selected from one of the four conditions in the previous experiment . There were four concrete constraints ( Tipping over , Pinching fingers , Moving around , Folding ) in the concrete - constraint conditions ; therefore , we randomly selected one example from each constraint . Because there were only two constraints ( safety and flexibility ) in the abstract - constraint conditions , we selected two examples from each constraint . Thus all conditions received four examples split amongst the two abstract constraints , with a further split for the concrete - constraint conditions . Participants Overall , 147 Amazon Mechanical Turk workers participated in the experiment . Forty - one percent were women , and 91 % were native English speakers . Their average age was 33 and ranged from 19 to 83 . Design and Procedure The same four conditions in Table 2 were used in this experiment , with the addition of a baseline control in which participants were not shown any examples . Participants first saw the instructions given to Turkers searching for inspirations in the domain - dependent , concrete constraint condition in Experiment 1 : \u201c This task asks you to design a flexible and safe kindergarten chair . The chair should be flexible enough to be used and stored in a variety of ways in a kindergarten class . It should also be safe for kids ( 4 - 5 year olds ) to use . Specifically , a flexible and safe kindergarten chair should meet the following criteria : 1 ) The chair can be easily folded up ; 2 ) The chair can be moved around easily by 4 - 5 year olds ; 3 ) The chair is stable and won\u2019t tip over easily ; 4 ) The chair won\u2019t pinch kids\u2019 fingers . \u201d After seeing the instruction , participants saw four examples and were asked to think about how these examples could be valuable : \u201c In a previous task , we asked other Turkers to search for examples that could provide inspiration for designing such a kindergarten chair . Below are four examples and the Turkers\u2019 explanations for why they provided them . Please look at these examples carefully and explain how the ideas from this example could help design a kindergarten chair . \u201d Participants were then asked to \u201c sketch the design of a flexible and safe kindergarten chair . Use the above examples as inspiration if you think they are helpful \u201d . Rating Did the examples found with different representations influence the quality of the new designs differently ? We draw on previous research on operationalizing measures of creativity , which suggests that ideas are more creative the more original and practical they are [ 1 , 8 , 25 , 27 ] . Two judges , the first author and an undergraduate student , blind to experimental condition rated each design on Likert scales measuring originality and practicality . Although neither judge was a professional designer , because the design challenge was about a consumer product we believe non - experts have sufficient knowledge to evaluate the novelty and practicality of the solutions . Prior work also shows high agreement between designers and non - designers in judging consumer products such as chair design [ e . g . , 30 , page 50 - 51 ] . To achieve more objective judgment , we considered the requirements of kindergarten chairs , and , based on these requirements , created specific criteria for judging originality and practicality . Originality was defined as an idea that was not obvious and differed from existing products on the market . To judge originality , we selected four most frequently found chair examples in the previous experiment as the typical chair design . The raters were asked to judge originality on 7 - point Likert scales by comparing the designs to these four typical chairs , as shown in Table 5 . Practicality was defined as how much the new design met the four constraint criteria and how realistic it would be to design , manufacture , and use the idea . To measure practicality we used a combination of five three - point Likert scales judging each of the four constraints and the realism constraint described above . Below shows examples of the specific instructions : Tipping over : how much does the design solve the tipping over issue ? Not at all ( 1 ) , sort of ( 2 ) , completely ( 3 ) Realism : Can it be realistically designed , produced , and Table 5 . Typical chairs used in judging originality . 1241 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA used ? Not at all ( 1 ) , sort of ( 2 ) , completely ( 3 ) The final practicality score was calculated by adding all the five scores together , creating a 15 - point scale . After several rounds of training and discussion , the judges achieved ICC inter - rater reliabilities of 0 . 75 and 0 . 79 for originality and practicality respectively . The designs in Table 6 show two examples with their corresponding scores on practicality and originality . Because large differences in participants\u2019 sketching skills could distort judges\u2019 estimates of idea quality , the judges were told to use the sketches to assist understanding the design idea but not to base their judgments on sketch quality . In the experiment , participants were asked to explain how their ideas were inspired by the examples if they were so inspired . We used the information to judge whether a new design was inspired by an inspirational example . It was coded as 1 if the participant mentioned an inspiration ; otherwise it was coded as 0 . Analysis and Results The five conditions differ in terms of whether showing examples and the sources of the shown examples . We examined practicality and originality of the new designs as well as whether a new design was inspired by the examples shown to them , inspiration . The means , standard deviations of practicality and originality , and the percentage of designs that were inspired by an example are shown in Table 7 . The distribution of the originality scores was left skewed . We performed a square root transformation to normalize the data and used the transformed data for analysis . We first compared all the experimental conditions to the baseline condition on practicality and originality separately through two regression analyses . The results show that only the ideas produced by the Domain - independent concrete constraint condition had significantly higher practicality and originality than the baseline condition , as shown in Table 7 . To further examine the factors affecting the quality of ideas , we applied a similar dummy coding as in the previous experiment . Regression analyses , regressed on domain abstractness , constraint concreteness , and their interaction , showed that domain abstractness and constraint concreteness alone did not predict practicality ( b = - 0 . 64 , p = 0 . 19 for domain abstractness ; b = - 0 . 14 , p = 0 . 78 for constraint concreteness ) nor originality ( b = 0 . 02 , p = 0 . 89 for domain abstractness ; b = 0 . 07 , p = 0 . 53 for constraint concreteness ) . However , we observed a significant positive interaction between domain abstractness and constraint concreteness for both practicality ( b = 3 . 10 , p < 0 . 01 ) and originality ( b = 0 . 38 , p < 0 . 05 ) , as shown in Figure 1 and 2 . A logistic regression analysis examining inspiration revealed a similar finding : while domain abstractness and constraint concreteness alone did not predict practicality ( b = - 0 . 80 , p = 0 . 14 for domain abstractness ; b = - 0 . 25 , p = 0 . 65 for constraint concreteness ) nor originality ( b = 0 . 02 , p = 0 . 89 for domain abstractness ; b = 0 . 07 , p = 0 . 53 for constraint concreteness ) , we observed a significant positive The chair is made out of flexible material . The back part of the chair is made out of a soft cotton material and so is the cushion for where you sit . There is an addition of two more legs outside of the chair . It can easily be folded upright thanks to the chair being out of a more flexible material . The joints of the chair each have one of the protective joint covers mentioned by the other Turkers that prevents fingers from being pinched as the chair folds down . It was inspired by the first image , the second and the last because it seemed to make the most sense ( Practicality : 13 ; Originality : 5 ) The slide example reminded me how much fun I use to have as a child playing on the swing set . So I thought , \" wouldn ' t it be great to have a swing chair ? \" The chair would be big enough for one child to sit and there would be enough soft pillow / padding in it to be very comfortable . The swing would be designed to allow enough freedom of movement to get a swinging motion but limited so that it doesn ' t swing to wide or to high . The chair ergonomics would function like a cradle and if desired a child could sleep in it like a mini hammock . ( Practicality : 10 ; Originality : 7 ) Table 6 Experiment 2 : Design examples 1242 SESSION : CROWD INNOVATION AND CROWDFUNDING interaction between domain abstractness and constraint concreteness ( b = 1 . 95 , p < 0 . 01 ) . Overall these results suggest creative ideas are most likely to be inspired by examples found using an abstract problem context ( i . e . , removing domain specific information ) but constraints that are more concrete and suggest mechanisms by which their requirements can be met ( e . g . , prevent from tipping over ) rather than more abstract requirements ( e . g . , safety ) . The examples in Table 6 offer some insights into these findings . In the first example , the participant borrowed three mechanisms from three of the four examples : the extra supporting legs for stability ( the first image ) , the buffer on the joint ( the second image ) , and the folding method ( the last image ) . These mechanisms can solve three constraints : tipping over , pinching fingers , and folding . Such successful transfer not only resulted in a practical design but also increased originality , because these mechanisms may have been widely used in the diverse domains people explored , but they were rarely seen in kindergarten chairs . In contrast , examples found in the domain - dependent conditions have no such benefits because most of them belong to the chair domains ; examples found with domain - independent abstract constraints are not as useful because many of their mechanisms are not applicable to chairs . In Experiment 2 , we were also interested in the integration of multiple constraints . Applying multiple examples to different constraints is challenging for several reasons . One is that some constraints conflict with each other . Once you adapt a mechanism from one inspiration , it might become difficult to integrate a mechanism from another . For example , to prevent chairs from pinching , the chair can be designed without joints , but this design could make it difficult to integrate mechanisms involving folding , which are useful for satisfying the storage constraint . A second explanation for the difficulty in incorporating mechanisms from multiple inspirations is high cognitive load and limited working memory . Indeed , in Experiment 2 , we found that even though some participants could successfully apply mechanisms from all four examples , the majority of participants used fewer than four , as illustrated in Table 6 . We will return to this observation in the Discussion . DISCUSSION In this paper we examined how to extend previous crowdsourcing idea generation research beyond problems involving a single constraint to those involving multiple constraints . In particular , we contributed a process for eliciting multiple constraints for a problem and investigated how the level of abstraction of those constraints influenced creative idea generation . During the first step of the process , crowd workers transformed an ill - formed , open - ended problem ( e . g . , design a creative kindergarten chair ) to a better - structured statement comprising concrete constraints ( e . g . , design a chair that is easily movable , stackable and won\u2019t tip over ) . Other crowd workers searched for inspirational examples of products in remote domains that could satisfy the constraints ( e . g . , mechanisms that prevent objects from tipping over ) . Yet other crowd Conditions Practicality Originality % Inspiration N Mean SD Mean SD Baseline ( no examples ) 9 . 70 1 . 79 2 . 04 1 . 49 23 Domain - dependent , abstract constraint 9 . 45 1 . 78 2 . 03 1 . 17 26 % 31 Domain - dependent , concrete constraint 9 . 31 2 . 22 2 . 20 1 . 32 31 % 29 Domain - independent , abstract constraint 8 . 81 1 . 49 2 . 06 1 . 32 44 % 32 Domain - independent , concrete constraint 11 . 77 * * * 2 . 10 3 . 48 * * * 1 . 61 71 % * * 31 p < . 01 = * * , p < . 001 = * * * Table 7 . The effects of abstractness of examples on design quality in Experiment 2 . Figure 1 Practicality : Interaction between abstraction of the domain and constraints , with 95 % CIs . Figure 2 Originality : Interaction between abstraction of the domain and constraints , with 95 % CIs . 1243 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA workers were able to draw inspiration from these examples to generate creative conceptual designs for chairs that met these constraints . Their product designs were more creative ( i . e . , more practical and original ) if the search for inspirations started with a problem representation that abstracted away the problem domain ( e . g . , kindergarten chairs ) , but kept the constraints concrete , compared to searches that started with a representation that mentioned the problem ( e . g . , kindergarten chairs ) or that abstracted the nature of the constraints . The contributions of this paper take a step towards moving distributed analogical idea generation in the direction of more complex , real - world design problems that often involve multiple constraints . However , an important open question is the extent to which more complex real - world problem - solving challenges , such as in developing a car or a mobile phone , would work with this approach . Furthermore , although we demonstrated this process with non - experts , there may be significant benefits in moving beyond the relatively na\u00efve problem solvers ( i . e . , untrained Mechanical Turk workers ) used here to trained problem solvers and domain experts ( e . g . , engineers ) . Some aspects of the current process may strongly benefit from employing more sophisticated workers at different stages . For example , it may be more useful to elicit constraints from domain experts ( e . g . , teachers , parents or children for the kindergarten chair challenge ) , as in traditional requirements analysis , to recruit experts in other fields for the search step to scour their domain archives for relevant mechanism , and to employ traditional product designers in the final , integration steps . We consider our research only a first step . Additional research is needed at both the theoretical and practical levels . Progress in theory development requires crisper ways to decompose the nature of a problem . Our research differentiated the problem context ( kindergarten chair versus other problems ) from problem constraints ( e . g . , flexibility and safety ) , but we suspect this decomposition is too crude . For example , constraints themselves are likely to differ on multiple dimensions , such as their concreteness ( which we examined in the current research ) , the degree to which they are integral to the overall problem or optional , their breadth of applicability and their complexity . These attributes are likely to influence the most effective way to communicate constraints to problem solvers and the degree to which analogical transfer will be a fruitful approach for finding mechanisms to satisfy them . At the methodological level , we need to develop more robust ways to elicit constraints and requirements from analysts . Although requirements analysis is a key feature of HCI practice , we believe there is no single best way to elicit and present requirements in such as way to be useful to problem solvers . At the application level , we believe the most interesting research involves both understanding and developing processes to help problem solvers to integrate multiple constraints . Having multiple constraints creates two challenges during integration : the large amount of information to process and the possible conflict between constraints . For example , a kindergarten chair may have dozens of constraints and people may be able to identify multiple analogs ( examples of methods for satisfying each constraint ) , some of which will conflict with each other . For example , to make a chair easy to clean , workers may find analogs like a kitchen blender that uses a \u201cdisassembly\u201d schema , so that parts from a bulky object can be removed and cleaned separately . However , applying this to a chair ( e . g . , by designing detachable legs ) could undermine its stability and provide gaps that could catch fingers . In the research presented here , the people who searched for inspiration received each constrain separately , with integration deferred to people in the problem - solving stage who needed to integrate multiple mechanisms . However , this approach is not necessarily the best one . This points to a fundamental tradeoff in distributed analogical processing between taking advantage of decomposing and distributing a problem across many people ( which can improve the capacity and diversity of exploration ) and dealing with the lack of context of other parts of the process that such decomposition incurs . Studying the conditions under which this tradeoff is optimized for design problems with multiple constraints would be fruitful follow - up research . ACKOWLEDGEMENTS This work was supported by NSF grants IIS - 1526665 , IIS - 1149797 , IIS - 1217559 , OCI - 0943148 , IIS - 0968484 , IIS - 1111124 , Bosch , Google , and Microsoft , Heinz College Center for the Future of Work , and Carnegie Mellon\u2019s Center for the Future of Work . We thank reviewers for useful feedback . REFERENCES 1 . Teresa Amabile . 1996 . Creativity in Context . Westview Press . 2 . Richard Buchanan . 1992 . Wicked problems in design thinking . Design issues , 5 - 21 . 3 . Balakrishnan Chandrasekaran . 1990 . Design problem solving : A task analysis . AI magazine 11 , 4 : 59 . 4 . Joel Chan , Susannah B . F . Paletz , and Christian D . Schunn . 2012 . Analogy as a strategy for supporting complex problem solving under uncertainty . Memory & cognition 40 , 8 : 1352 \u2013 1365 . 5 . Darren W . Dahl and Page Moreau . 2002 . The Influence and Value of Analogical Thinking during New Product Ideation . Journal of Marketing Research 39 , 1 : 47 \u2013 60 . 6 . Kevin Dunbar and Isabelle Blanchette . 2001 . The in vivo / in vitro approach to cognition : The case of analogy . Trends in cognitive sciences 5 , 8 : 334 \u2013 339 . 7 . Karl Duncker and Lynne S . Lees . 1945 . On problem - solving . Psychological monographs 58 , 5 : i - 113 . 1244 SESSION : CROWD INNOVATION AND CROWDFUNDING 8 . Ronald A . Finke , Thomas B . Ward , and Steven M . Smith . 1992 . Creative cognition : Theory , research , and applications . The MIT Press . 9 . R . A . Fisher . 1954 . Statistical methods for research workers . Oliver and Boyd . 10 . Dedre Gentner . 2002 . Analogy in scientific discovery : The case of Johannes Kepler . In Model - Based Reasoning . Springer , 21 \u2013 39 . 11 . Dedre Gentner , Sarah Brem , Ron Ferguson , Philip Wolff , Arthur B . Markman , and K . D . Forbus . 1997 . Analogy and creativity in the works of Johannes Kepler . Creative thought : An investigation of conceptual structures and processes , 403 \u2013 459 . 12 . Mary L . Gick and Keith J . Holyoak . 1980 . Analogical problem solving . Cognitive psychology 12 , 3 : 306 \u2013 355 . 13 . Mary L . Gick and Keith J . Holyoak . 1983 . Schema induction and analogical transfer . Cognitive psychology 15 , 1 : 1 \u2013 38 . 14 . Ashok K . Goel and Sambasiva R . Bhatta . 2004 . Use of design patterns in analogy - based design . Advanced Engineering Informatics 18 , 2 : 85 \u2013 94 . 15 . Andrew Hargadon and Robert I . Sutton . 1997 . Technology brokering and innovation in a product development firm . Administrative science quarterly , 716 \u2013 749 . 16 . Mary B . Hesse . 1966 . Models and analogies in science . University of Notre Dame Press , Notre Dame , IN . 17 . J . S . Linsey , A . B . Markman , and K . L . Wood . 2012 . Design by analogy : A study of the WordTree method for problem re - representation . Journal of Mechanical Design 134 , 4 : 041009 . 18 . Winter Mason and Siddharth Suri . 2012 . Conducting behavioral research on Amazon\u2019s Mechanical Turk . Behavior research methods 44 , 1 , 1 \u2013 23 . 19 . Scarlett R Miller , Brian P Bailey . 2014 . Searching for inspiration : an in - depth look at designers\u2019 example finding practices . ASME 2014 International Design Engineering Technical Conference and Computers and Information in Engineering Conference . 20 . Meg Monk . ( 2013 , December ) . BYU engineers use origami to make more space in space . The Digital Universe , Retrieved from http : / / universe . byu . edu / 2013 / 12 / 12 / byu - engineers - use - origami - to - make - more - space - in - space / 21 . Donald A . Norman . 2002 . The design of everyday things . Basic books . 22 . Tony Poze . 1983 . Analogical connections\u2014The essence of creativity . The Journal of creative behavior 17 , 4 , 240 \u2013 258 . 23 . Tom Ritchey . 2006 . Problem Structuring Using Computer - Aided Morphological Analysis . The Journal of the Operational Research Society 57 , 7 , 792 \u2013 801 . 24 . David Serrano . 1987 . Constraint management in conceptual design . Retrieved January 13 , 2015 from http : / / dspace . mit . edu / handle / 1721 . 1 / 14689 25 . Jami J . Shah , Steve M . Smith , and Noe Vargas - Hernande . Metrics for measuring ideation effectiveness . Design studies 24 , 2 ( 2003 ) , 111 \u2013 134 . 26 . Patrick C . Shih , Gina Venolia , and Gary M . Olson . 2011 . Brainstorming under constraints : why software developers brainstorm in groups . In Proceedings of the 25th BCS Conference on Human - Computer Interaction , British Computer Society , 74 \u2013 83 . Retrieved January 13 , 2015 from http : / / dl . acm . org / citation . cfm ? id = 2305331 27 . Robert J . Sternberg . 1998 . Handbook of creativity . Cambridge University Press . 28 . Lixiu Yu , Aniket Kittur , and Robert E . Kraut . 2014 . Searching for analogical ideas with crowds . In Proceedings of the 32nd annual ACM conference on Human factors in computing systems ( CHI\u201914 ) , ACM , 1225 \u2013 1234 . http : / / dl . acm . org / citation . cfm ? id = 2557378 29 . Lixiu Yu , Robert E . Kraut , and Aniket Kittur . 2014 . Distributed analogical idea generation : innovating with crowds . In Proceedings of the ACM conference on human factors in computing systems ( CHI\u201914 ) , ACM , 1245 - 1254 . http : / / dx . doi . org / 10 . 1145 / 2556288 . 25573781 30 . Lixiu Yu . 2012 . Crowd Idea Generation . Ph . D Dissertation . Stevens Institute of Technology , Hoboken , NJ . 1245 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA",
    "chiuInvestigatingEffectsOppositely2012": "Journal ofEngineering Design Vol . 23 , No . 4 , April 2012 , 271 - 296 Q Taylor & Francis ~ Taylor & . FrantisCroup N - o N ~ < l ) . . c o . . . . u o N o \" \" \" r ' I o - Investigating effects of oppositely related semantic stimuli on design concept creativity Q Department ofMechanical and Industrial Engineering , Ryerson University , 350 Victoria Street , Toronto , ON , Canada M5B 2K3 ; b Department ofMechanical and Industrial Engineering , University ofToronto , 5 King ' s College Road , Toronto , ON , Canada M5S 3G8 ( Received 25 February 2011 . . final version received 4 July 2011 ) We are motivated to investigate methods to increase creativity in conceptual design since creativity is essential to design success , and no other stage influences final design success as much as conceptual design . Existing work supports that design stimuli may encourage creative concept generation , but does not give details on how to systematically generate stimuli . The established relationship between language and cognition , and the systematic nature of language prompt us to examine its use as design stimuli . Language relationships such as opposition provide a systematic method of generating non - obvious semantic stimuli for design problems . In this paper , we presenttwo experiments , apen - and - paperand averbal protocol study , where participants usedoppositely related and similarlyrelated word stimuliinconceptualdesign . Wefound that designers using oppositely related word stimuli developed more creative concepts . Language analysis revealed how opposite stimulieliciteddesignerbehaviours thatmay encourage and supportcreative concept generation . Our empirical results combined with linguistic theory lead us to propose a model explaining the interactions and effects of opposite - stimulus words on concept creativity . This knowledge can be used to facilitate more creative , and ultimately , more successful design . Keywords : language ; creativity ; design stimuli ; concept generation ; engineering design 1 . Introduction We have been studying and quantifying the application of natural language , i . e . human language , not artificial language , to the process ofstimulating creative conceptual design . We are motivated to study language in relation to design because language appears ubiquitously , is inherent in people , and is connected to cognition . Researchers have established a linkbetween language and cognition , although the exact relationship is disputed . Some have shown that language affects cognition ( Levinson 1996 ) , while others have shown that language reflects cognition ( Pinker 2007 ) . Design is a cognitively intensive activity requiring such tasks as information gathering , spatial manipulation , searching , decision - making , etc . ( Simon 1969 , Gero et af . 1994 , Coley et af . 2007 ) . The cognition required of design suggests that we can take advantage of the relationship between language and cognition to facilitate and understand design . * Corresponding author . Email : shu @ mie . utoronto . ca ISSN 0954 - 4828 print / ISSN 1466 - 1837 online \u00a9 2012 Taylor & Francis http : / / dx . doi . orgl 10 . 1080109544828 . 2011 . 603298 http : / / www . tandfonline . com 272 I . Chiu and LH . Shu In this paper , we present and describe work on stimulating creative concept generation using language . Creativity is an important measure of design effectiveness ( Kan and Gero 2007 ) . While customers seek creative designs , they may not explicitly indicate creativity as a requirement ( Cross 2006 ) . There is general agreement that creativity is essential to design ( Gordon 1961 , de Bono 1992 , Altshuller and Shulyek 1996 , Hubka and Eder 1996 , Adams 2001 , Cross 2006 , Kan and Gero 2007 , Shai et al . 2009 , Yang 2009 , Brown 2008 , etc . ) , and some assert that design necessarily entails creativity ( Holt 1993 , Hubka and Eder 1996 , Hatchuel and WeiI2009 ) . We are specifically interested in the concept generation stage of design because the early stages of design have the largest influence on the final design ( Ullman 2003 , Keller et al . 2009 ) , and creative concepts result in creative and successful designs . We investigate the effects of language stimuli on concept creativity in experiments , where we provided participants with stimulus words that were either oppositely or similarly related to the desired functions ofthe solution . Stimulus words were systematically generated , using a thesaurus and WordNet . Words were chosen as stimuli because words are the smallest unit of language that carry meaning , and thus a natural starting point for our investigations . Verbs were the chosen part of speech for the stimuli because verbs denote abstract actions or functions and not specific forms . Word stimuli provided were verbs in root form , e . g . ' remove ' . Many agree that verbs should be used to model functions for design ( cf . Pahl and Beitz 1996 , Stone and Wood 2000 ) . The opposite / similar relationship was used to generate stimulus words because it is a systematic and well - understood relationship and one of only two valid verb relationships . While we may never fully understand the effects of language on design cognition , insights gained from our empirical results combined with linguistic theory enable us to propose a model of the language and design cognition interactions required to generate more creative concepts . In this paper , we will first review related work in language , design and creativity . Then we will describe and discuss our experiments with respect to the specific effects of different types of word stimuli on concept creativity ( Experiments 1 and 2 ) and on designer behaviour ( Experiment 2 ) . Finally , we will discuss our experiments within an empirical and theoretical context and propose a model explaining the effects of opposite / similar word stimuli on creative concept generation . 2 . Background 2 . 1 . Language and design The connection between language and cognition was first observed by the ancient Greeks , who used the same word , logos , to denote both reasoning and language ( Kalmar and Davidson 1997 ) . The Sapir - Wharf hypothesis dating from the early twentieth century argues that the language of a person determines how he or she understands the world ( Ratner and Gleason 1993 ) . While there is debate over whether cognition influences language ( Levinson 1996 , Li and Gleitman 2002 , etc . ) , or language influences cognition , many agree that language and cognition are closely related ( e . g . Pinker , Fodor , Boaz from Saeed 2003 ) . Researchers such as Chomsky ( 1968 ) , Jackendoff ( 1983 ) and Pinker ( 2007 ) argue that language provides insight into human cognitive processes . In our work , we explore the use of language to stimulate creative design . Language , specifically at the word level , appears ideal as design stimuli , and fits well within early stages of the design process , where the problem is likely to be ill - defined , and exploration of the solution space encouraged . Words are the smallest unit of language that carry meaning , and thus an appropriate starting point for investigating language as design stimuli . While words impose a pre - existing symbol system on the user that can be shared and manipulated ( Bruner 1964 ) , they are not necessarily fixed to a particular form ( Segers 2004 ) , and ambiguity enables ~ . . ~ ~ ~ ~ ~ . _ - ~ - - - - - - - - - - - - - - - - - - - - - - - - Journal ofEngineering Design 273 freedom of interpretation ( Miller et al . 1993 ) . Not only is language connected to reasoning , words also appear connected to our knowledge of the world . In one model of lexical memory , when a specific word is found in the intemallexicon of permanent memory , simultaneously retrieved are properties associated with that word , e . g . meaning , spelling , pronunciation , etc . At the same time , properties not strictly linguistic are also retrieved , e . g . if the word ' elephant ' is retrieved , it might also trigger the common knowledge that elephants never forget ( Carroll 1999 ) . The knowledge retrieved is not necessarily just part of the word - meaning , but may be related to the conceptual knowledge of the world in general . While words seem simple and familiar , word meaning can be complex and ambiguous ( Carroll 1999 ) . A designer ' s familiarity with the complexities oflanguage may make language a useful design tool . While natural language is familiar and ubiquitous to most , it is not generally considered a conventional engineering tool . However , many researchers are investigating the use of natural language as a formal tool to support the engineering design process . For example , language has been used as input in requirements gathering ( Nuseibeh and Easterbrook 2000 ) , concept generation and synthesis ( Thomas and Carroll 1984 , Hacco and Shu 2002 , Nagai and Taura 2006 , Chiu and Shu 2007 , Tseng et at . 2008 ) , design modelling and representation ( Stone and Wood 2000 ) , and design analysis ( Dong et al . 2003 , Dong 2006 ) . Some concept generation and creativity methods explicitly use language as stimuli in an attempt to increase creativity . Thomas and Carroll ( 1984 ) demonstrated that participants who were given 20 pages of semi - random stimulus words generated more creative concepts compared with par ticipants not given stimulus words . Nagai and Taura ( 2006 ) investigated the interpretation of noun - noun combinations for promoting creativity in concept synthesis . Some work in biomimetic design uses functional keywords to systematically retrieve analogies from biological corpora for use as stimuli in engineering design ( Hacco and Shu 2002 , Chiu and Shu 2007 , Cheong et at . 2011 ) . Methods implicitly using language to stimulate creative design include synectics and random input . In synectics , Gordon ( 1961 ) proposes the use of metaphors and similes , which are figures of speech , to draw parallels and connections between disparate topics or domains . Metaphors and similes promote analogical thinking and can allow solutions to be applied from one domain to another . The random input method involves randomly selecting stimulus , e . g . a picture from a catalogue , to relate back to the design problem through a series of word associations . The process of relating the problem to random stimulus , which may be non - obvious and unexpected , may provide new perspectives and thus stimulate creative design ( de Bono 1970 ) . The above examples show that language is frequently used to facilitate design . However , fewer researchers have attempted to examine and model the effect of language on design . Nagai and Noguchi ( 2003 ) developed a model of the thinking path required for creative thinking . In an experiment where designers used keywords while designing , Nagai and Noguchi observed that difficult and remote keywords caused designers to extend their thinking pathways . They concluded that extending thinking pathways may help to realise creative concepts . In our experiments , we observed that oppositely related keywords result in designer behaviours that may increase concept creativity . We then propose a model to explain the effects ofoppositely related stimuli on increasing concept creativity . 2 . 2 . Opposition in design We study opposite versus similar - language stimuli because opposition and opposing relation ships are common in language and reasoning . Antonyms , or opposite words , are universally found in language , and most people demonstrate good intuition in recognising antonym / synonym pairs ( Fellbaum 1993 , Murphy 2003 ) . The antonym / synonym relationship is also one of only 274 1 . Chiu and L . R . Shu two valid verb relationships , with the other being the hypernymjhyponym relationship . The hypernymjhyponym relationship can be thought of as a super - ordinatejsub - ordinate relationship , where words are hierarchically related in either a more general or more specific manner . Design methods using opposition include TRIZ , design - by - analogyjcontrast , argumentation and argumentative negotiation ( Rittel and Webber 1984 , Altshuller and Shulyak 1996 , Fantoni et al . 2006 , Jin et al . 2007 ) . In TRIZ , the Russian abbreviation for the theory of inventive problem solving , the problem at hand is first phrased in contradictions to identify parameters to be improved and those degraded as a consequence . Fantoni et al . ( 2006 ) proposed a method of design - by analogyjcontrast involving the use of synonyms and antonyms as design stimuli . Argumentation and argumentative negotiation involve the verbalisation ofcontradictory demands and then a move towards agreement to produce novel solutions in collaborative engineering ( Rittel and Webber 1984 , Jin et al . 2007 ) . Hubka and Eder ( 1996 ) even speculate that opposites and dissimilarities may contribute to creativity in engineering design through the resolution of ' cognitive dissonance ' , e . g . resolving ideas from intuitive versus intellectual modes of thinking . Festinger , the originator of the theory of cognitive dissonance , theorised tensions occur when an individual becomes simultaneously aware of two inconsistent thoughts . To resolve the tensions , the individual must implement change ( Myers 1999 ) , thus creating new , alternative solutions . In design , resolving tensions between the problem statement and other factors , e . g . the degraded parameter versus the improved parameter in TRIZ , and stimuli oppositely related to the design problem , may lead to creative design . 2 . 3 . Measuring creativity in design Many approaches have been developed to assess the creativity ofa concept or idea . In general , most agree that creativity is multi - dimensional ( Torrance 1974 , Amabile 1983 , Shah et al . 2000 , etc . ) and that using a single measure ofcreativity may result in identifying strange or even incorrect ideas as being creative . For example , in Wilson ' s ( 1951 ) method ofstatistical infrequency , infrequent ideas are considered novel , and therefore creative . Two generally agreed upon measures of creativity within science and engineering are novelty and usefulness , defined below : Novelty : A creative idea must contain some degree of newness , originality or surprise ( Wilson 1951 , Torrance 1974 , Hubka and Eder 1996 , Howard et al . 2008 , Shai et al . 2009 , Brown 2008 ) . Usefulness : A creative idea must contain some degree of appropriateness and value ( Besemer and Treffinger 1981 , Amabile 1989 , de Bono 1992 , Akin andAkin 1998 , Howard et al . 2008 , Shai etal . 2009 , etc . ) . Amabile ( 1989 ) defines appropriateness in the sciences as being correct . Usefulness is especially emphasized in the engineering literature , often in the context of functionality ( Pahl and Beitz 1996 , Dieter 2000 , Shah et al . 2000 , Ullman 2003 ) . Some researchers assert that creative ideas only need to be novel and useful , e . g . Amabile ( 1989 ) and Akin and Akin ( 1998 ) . However , novelty and usefulness alone may not sufficiently measure creativity , especially at the abstract conceptual level . Many suggest that the wholeness , clarity , elaboration , or cohesiveness of an idea must also be considered : Cohesiveness : A creative idea must contain some degree of wholeness , elaboration , detail , style and clarity ( Torrance 1974 , Besemer and Treffinger 1981 , Hubka and Eder 1996 , Adams 2001 , Kudrowitz and Wallace 2010 ) . Other measures of creativity may also be used to assess the effectiveness of a creativity method or individual . These measures can include fluency , or the quantity of concepts , generated by an individual or the creativity method ( Torrance 1974 , Shah et at . 2000 , Yang 2009 ) and variety , or the number of different categories of concepts generated ( Torrance 1974 , Shah et at . 2000 ) . This paper focuses on the use of the direct measures of novelty , usefulness and cohesiveness for the evaluation of each individual concept . 2 . 4 . Previous work Journal ofEngineering Design 275 Previous exploratory experiments showed that designers provided with opposite - stimulus words tend to generate concepts that were more novel . These previous experiments are summarised below . The first exploratory experiment was a pen - and - paper study where participants indicated their concepts on worksheets . Forty - two participants were provided with opposite and simi lar words simultaneously for a series of four problems and then instructed to use the words as stimuli for design . For three of the problems , participants who chose at least one opposite stimulus word generated more novel concepts ( Chiu and Shu 2008a ) . Novelty was determined using statistical infrequency , where less frequently occurring concepts were deemed more novel ( Wilson 1951 , Shah et al . 2000 ) . However , because all participants were provided with opposite - and similar - stimulus words simultaneously , results may have been confounded , i . e . were participants affected by only the opposite words , or by the pairs of opposite / similar words ? From this experiment , we were also unable to gain further insight into how par ticipants used language stimuli due to the nature of data collected from pen - and - paper experiments . The second exploratory experiment was a small - scale between - subjects experiment where par ticipants verbalised all thoughts as they designed . Six participants received either opposite words or similar words while generating concepts for one problem . Those receiving opposite stimuli generated more novel concepts , however the difference was not significant . Novelty in this exper iment was determined by two independent human raters . Two raters were recruited because two is the minimum number of human raters required for judgement tasks ( Landis and Koch 1977 ) . Spearman ' s correlation between the two raters was r = 0 . 51 , p = 0 . 054 \" - ' 0 . 05 , showing a large , borderline significant agreement between the two raters ( Chiu and Shu 2008b ) . No training was provided to the raters in this exploratory experiment . Using the session transcripts from the second experiment , we were able to conduct an explicit content analysis to examine participants ' language use and behaviours with respect to different stimulus types . Specifically , we examined the stimulus part - of - speech ( paS ) , to determine whether a given stimulus word was used as a noun , verb or adjective . While stimulus words were given as root verbs , e . g . ' remove ' and not ' removing ' or ' removed ' , participants were not told that the stimuli were verbs . We found that opposite - stimulus participants used stimulus words significantly more often as verbs than similar - stimulus participants . Furthermore , in this second exploratory experiment , we found that opposite - stimulus participants tended to introduce more new words and phrases ( NWPs ) in their concept generation process , but not significantly so . NWPs are identified by comparing words and phrases given in a problem statement with the words and phrases generated by participants in relationship to the stimulus words . NWPs may represent new concept elements that have been expressed , or lexicalised , within concept generation . Increased NWP introduction appears linked to the use of stimuli as verbs rather than nouns . Increased NWPs appear advantageous for concept generation because NWPs may form the basis ofcreative concepts . An expanded verbal protocol experiment , i . e . in terms of participants and problems , would allow for more thorough analysis . Overall , previous experiments showed that opposite stimuli appear to increase concept novelty , one of the measures that contribute to total concept creativity . These preliminary results serve as motivation for current experiments ( described in Sections 3 and 4 ) that overcome limitations of previous work described above . On the basis of our preliminary results , we hypothesise the following : ( 1 ) Opposite - stimulus words increase concept creativity in terms of all creativity measures ; ( 2 ) Opposite - stimulus words used as verbs increase the introduction of NWPs _ _ _ _ _ _ _ _ _ _ _ _ u _ 276 I . Chiu and LH . Shu The following sections will describe two experiments where we focus on further quantifying the above two hypotheses . 3 . Experiment 1 : a between - subjects pen - and - paper experiment Experiment 1 was a fully between - subjects pen - and - paper experiment where participants gener ated concepts for a series of four problems : ( 1 ) Bushing problem ( 2 ) Egg problem ( 3 ) Grinding problem and ( 4 ) Sunflower problem . Participants generated concepts under one of the following experimental conditions : ( 1 ) Opposite stimulus ; ( 2 ) Similar stimulus . Four independent raters were recruited to judge the concepts based on the creativity components of novelty , usefulness and cohesiveness . For two of the experiment problems , the Sunflower and Egg problems , raters judged opposite - stimulus concepts as being significantly more novel , useful and cohesive than similar - stimulus concepts . For the other two problems , the Bushing and Grinding problems , there were no significant differences in any of the creativity component scores . Details are given below . 3 . 1 . Method 3 . 1 . 1 . Participants and procedure Participants consisted of nine graduate engineering students from the Department of Mechanical and Industrial Engineering at the University ofToronto . At the time ofthe experiment , participants were enrolled in a graduate design course . Participants consisted of two females and seven males with an average age of 27 . 2 years ( sd = 5 . 38 ) . Eight out of nine participants indicated they had industry design experience ranging from a few months to five years . Participants were randomly assigned to one of two experimental conditions where they gen erated concepts using either opposite stimuli or similar stimuli . All participants completed four problems , but in a different , random order as determined by a random number generator . Partici pants were allotted a total time of 10 min per problem to review given stimulus words ; to select stimulus words ; and to generate and describe their concepts using selected stimulus words on provided worksheets . The total experiment duration was 40 min . The worksheets were collected for analysis . The exact instructions are given below : This is an experiment investigating the use of stimuli in concept generation . The following are four unrelated design problems . For each design problem , a set of related word stimuli is supplied . For each problem : ( I ) Review the problem and associated word stimuli . ( 2 ) Perform a functional decomposition , e . g . what needs to be done ? ( 3 ) Select the word ( s ) you want as stimuli and indicate your selection ( s ) . Each stimulus set is only relevant to the associated design problem . ( 4 ) Use selected word ( s ) to develop concepts to solve the design problem . Ifyou determine that you cannotcomplete a concept using your selected word ( s ) , you may select another word . ( 5 ) Please consider each problem in the order given . You have 10minutes for each problem . Record all concepts . You may write , sketch , calculate , etc . , on your worksheets . Pen - and - paper experiments are a fairly time - efficient method for collecting design data . Others have used pen - and - paper experiments to study various aspects ofconceptual design such as effects of random stimulus words , analogical similarity , generation of form alternatives and sketching abilities ( Thomas and Carroll 1984 , Tseng et at . 2008 , Yang 2009 , Corremans 2011 ) . Journal ofEngineering Design 277 3 . 1 . 2 . Experiment problems and stimuli Table I . Summary of problems and stimulus sets for Experiment I . 3 . 1 . 3 . Concept evaluation Concept evaluation for our previous studies ( described in Section 2 . 4 ) was limited to the evaluation of concept novelty , using statistical infrequency or human raters . Since the creativity literature suggests that creativity is more than merely novelty , and that creativity is ultimately a human judgment , e . g . from a customer ' s point of view ( Cross 2006 , Brown 2008 ) , we recruited human Opposite Fill InsertJoinCombine Roughen AddClogReplace Empty WithdrawDisconnectDivide Stimulus words Similar Smooth SubtractCleanRemove Problem description and decomposition Sunflower seed oil is a nutritious and valuable commodity in sub - Saharan West Africa . Mechanical presses to make oil from the shelled seeds exist locally , but machines to remove the shells do not . At present , there exists no alternative to the laborious and time - consuming process of shelling the sunflower seeds individually by hand before loading them into the press . Develop a concept for shelling sunflower seeds that can be used locally with minimal resources ( DTM 2006 ) Grinding of metals is quite common to obtain a fine surface finish and tight tolerances . But when grinding soft materials such as rubber or plastic , the grinding wheels quickly become clogged . Repeated dressings ( sharpening and shaping of the grinding wheel ) do not help . Develop concepts that will enable surface finishing ( with or without grinding wheels ) to be used on soft materials ( Kosse 2004 ) Participants were provided with four problems and related - stimulus sets on worksheets . Stimulus words were obtained by first performing a functional decomposition of the problem . Functional decomposition identifies the functions , i . e . physical actions or behaviours , required to transform an initial state into the desired final state ( Pahl and Beitz 1996 ) . A high - level functional decomposition yielded functional keywords that were then used to generate similar - or opposite - stimulus words using a thesaurus ( Merriam - Webster . com 2008 ) and WordNet 3 . 0 ( Princeton University 2008 ) . Problems and stimuli are summarised in Table 1 . For the Sunflower problem , the required high - level functions to transform whole seeds into oil were ' extracting ' the seeds from the shell , and then ' separating ' the seeds from the shell fragments for the production of oil . From the original functional keyword ' extract ' , we generated using a thesaurus and WordNet similar words ' empty ' and ' withdraw ' , and opposite words ' insert ' and ' fill ' . From ' separate ' , similar words generated were ' disconnect ' and ' divide ' , and opposite words were ' join ' and ' combine ' . The original functional keywords were not provided to participants . Problem Sunflower - seed shelling Soft - material grinding Egg orientation Develop concepts to automatically orient raw chicken eggs with the pointed ends all facing one direction ( Kosse 2004 ) SelectDetectPivot Move RejectMiss FixRestrain Bushing - and - pin assembly Parts that are automatically mated , e . g . a bushing and a pin , must be positioned so that their axes coincide . Using chamfers on mating parts does not solve the alignment problem . Develop a concept to centre mating parts that does not require high positioning accuracy ( Kosse 2004 ) StraightenMatchInjectInstall SkewMixEjectExtract 278 0 1 2 Not novel I . Chiu and L . H . Shu Figure I . Rating scale for novelty . Reprinted with pennission . Copyright ASME 20IO . Table 2 . Example anchor concepts for Sunflower problem . raters to judge concept creativity for this study . Human raters assessed concept creativity using all three components described previously : Novelty , usefulness and cohesiveness . Each component was scored on an II - point scale between a and 10 . The rating scale for novelty is illustrated in Figure 1 , with the same numeric scale applied to usefulness and cohesiveness . For this experiment , four independent raters were recruited to score each concept . Raters con sisted of one female and three males , all with an engineering background and interest in design . Raters were not provided with the identity of the designers , nor the stimulus condition under which the concepts were generated . Concepts were presented to raters in random order . Raters were provided with low anchor and high anchor concepts , i . e . ' not novel / useful / cohesive ' and ' very novel / useful / cohesive ' concepts obtained from previous experiments , and instructed to evaluate all concepts based on those anchors . This method of anchoring and obtain ing human judgements is known as direct scaling and is commonly used in psychophysics , a branch of psychology that deals with relating physical stimuli with mental phenomena ( Engen 1972 ) . The direct scaling method has been applied to rating pleasantness of smells , perception of heaviness , and even to rating of emotions and beauty . Examples of anchor concepts for the Sunflower problem are provided in Table 2 . Low / high Low High Novelty Place whole seeds in machine . Divide the shell by machine with chisel / knife edge and open like hinge securing shell . Empty the seeds from the shell and discard the shells from the hinge device Combine two wheels with sticky surfaces to crack the seeds . Shells stick to the surface and get carried away Usefulness Machine grips two sides ofseed . Seed is cut into half ( divide ) . The seed is emptied using vibration ( empty ) . Shells are discarded . Separate seeds processed as before Whole seeds go into a hopper . Use edible solvent to dissolve shells . Clean seeds . Dry seeds . Grind to get oil Cohesiveness Divide the shell into two . Empty shell and remove seed from shell Refer to high usefulness concept . High usefulness concept is also a high cohesiveness concept 3 . 2 . Results 3 . 2 . 1 . Inter - rater agreement Kendall ' s W concordance coefficient was calculated to measure inter - rater agreement . Kendall ' s W measures the agreement ofmore than two raters scoring N entities and ranges from 0 , signifying no agreement or random ratings , to 1 , signifying consensus between the multiple raters . As Kendall ' s W is related to Spearman ' s correlational coefficient , Kendall ' s W can be interpreted similarly to a correlation coefficient , where 0 . 1 , 0 . 3 and 0 . 5 are small , medium and large agreements , respectively ( Siegal 1956 ) . The inter - rater agreements are given in Table 3 . Although the coefficients above suggest medium to large degrees of agreement , the sample sizes are too small to draw confident conclusions ( p \u00bb 0 . 05 ) . Upon aggregating ratings for all concepts from all four problems , the coefficients indicate medium to large degrees of agreement that are significant , or borderline significant , as seen in Table 4 . Journal ofEngineering Design 279 Table 3 . Inter - rater agreement for Experiment 1 concepts . Problem Novelty Usefulness Cohesiveness 3 . 2 . 2 . Creativity results Therefore , based on the aggregated results , there is significant agreement among the raters ( p approximately 0 . 05 or less ) . In this section , we will first present examples of participant concepts and resulting rater scores to illustrate the experimental and concept rating methodology . Then , the overall results of the quantitative comparisons are provided . Three concept examples from the Sunflower problem are given in Table 5 along with associated average rater scores . Concepts are unedited and as collected directly from the worksheets . Stimulus words used by participants are italicised for reference . W = 0 . 6181 X 2 ( 8 ) = 14 . 83 P = 0 . 06 W = 0 . 5278 x 2 ( l4 ) = 22 . 17 P = 0 . 08 W = 0 . 4634 x 2 ( l5 ) = 20 . 85 P = 0 . 14 W = 0 . 497 X 2 ( 12 ) = 17 . 89 P = 0 . 12 Cohesiveness W = 0 . 4908 X 2 ( 52 ) = 76 . 5592 P = 0 . Ql5 Usefulness W = 0 . 5482 X 2 ( 8 ) = 13 . 16 P = 0 . 11 W = 0 . 6502 X 2 ( 14 ) = 27 . 31 P = 0 . 02 W = 0 . 5024 x 2 ( l5 ) = 22 . 61 P = 0 . 09 W = 0 . 3261 X 2 ( 12 ) = 11 . 74 P = 0 . 47 W = 0 . 4426 X 2 ( 52 ) = 69 . 052 P = 0 . 06 Novelty W = 0 . 507 X 2 ( 52 ) = 79 . 091 P = 0 . 009 W = 0 . 3711 X 2 ( 8 ) = 8 . 91 P = 0 . 35 W = 0 . 5531 x 2 ( l4 ) = 23 . 23 p = 0 . 06 W = 0 . 6004 x 2 ( l5 ) = 27 . 02 p = 0 . 03 W = 0 . 4839 x 2 ( l2 ) = 17 . 42 P = 0 . 13 Bushing and pin ( N = 9 ) Egg ( N = 15 ) Problem Grinding ( N = 16 ) Sunflower ( N = 13 ) Aggregated results ( N = 53 ) Table 4 . Kendall ' s W for aggregated results of Experiment I concepts . Table 5 . Example concepts and creativity scores for the Sunflower problem . Avg . rater score ( N = 4 ) Example Concept Participant experimental condition Nov . Use . Coho Concept 2 : To divide : Orient the seed so that it lies flat , then use a knife to split along the flat side of the shell . Or cut into half , then use vibration or gravity to empty , blow it out to extract inner part Concept 13 : ( 1 ) Air hose insert into seed . ( 2 ) Airjills the seed , cracking the shell using pressure . ( 3 ) The newly cracked seed is collected and joined with the rest of the seeds from before Concept 16 : Second concept - Grind shell and seed together . Then find a fluid , ( hopefully water will do ) that has the proper density to float the lighter shells to the surface and leave the heavier seeds on the bottom Average concept creativity for all Sunflower concepts ( N = 13 ) Similar Opposite Opposite 2 8 5 . 25 4 . 35 2 . 5 4 . 5 7 . 5 4 . 88 2 . 25 5 . 25 7 . 5 5 . 0 280 I . Chiu and L . R . Shu opp Sim . Opp . Cohesiveness Cohesiveness . 00 ' - . . ~ - ~ - - , - - - ~ . _ - 4 . 00 2 . 00 6 . 00 4 . 00 . 00 . L\u2022 . ~ I \u2022 . . ~ \u2022 . \u2022 . \u2022 . 2 . 00 Usefulness 4 . 00 opp Sim . Opp . Usefulness 2 . 00 6 . 00 6 . 00 opp Nov ~ lty r I - , ~ I - I i I , I I I ! i I 1 sim opp Sim . Opp . . 00 \" . - . \" - - ~ _ . - \" - - ~ . . . . . . . . - . 00 ' ; ' \u00a7 ' \" 6 . 00 \u2022\u2022 OJ ! . . . = S - : : I = ~ . 2 : c : z = 9 ; OJ ! 1lol OJ ! ~ ~ 2 . 00 Concept 2 was generated under the similar - stimulus condition , while concepts 13 and 16 were generated under the opposite - stimulus condition . Note that concept 16 does not incorporate stimu lus words . However , the participant had indicated that this was his second concept for this problem , and stimulus words were used in his first concept for this problem ; demonstrating he had reviewed the stimulus words . Concept 2 was rated as the least creative concept , and concepts 13 and 16 were rated as being above - average concepts . While raters judged concept 13 to be the most novel , concept 16 was rated as more useful and cohesive . This may reflect that raters considered batch processing of multiple seeds simultaneously ( concept 16 ) to be more practical than processing of individual seeds ( concept 13 and concept 2 ) . For each problem , rater scores for concepts were averaged to facilitate analysis , and independent T - tests were conducted to compare concept novelty , usefulness and cohesiveness . Rater scores were averaged because our main interest is in designer behaviour , not rater behaviour . For two of the experiment problems , the Sunflower and Egg problems , raters judged opposite - stimulus concepts as being more novel , useful and cohesive than similar - stimulus concepts . The T - tests showed that these differences were either significant , or borderline significant , in all but the Egg problem novelty scores . This can be seen in Figure 2 and Table 6 . Sim . Opp . Sim . Opp . Sim . Opp . Figure 2 . Graph of novelty , usefulness and cohesiveness results for the Sunflower and Egg problems . Table 6 . T - test results for Sunflower and Egg concept creativity . Sunflower Egg Creativity Sim . Mean Opp . Mean Sim . Mean Opp . Mean Component Rating ( N = 7 ) Rating ( N = 6 ) t ( ll ) p ( one - tail ) Rating ( N = 9 ) Rating ( N = 6 ) t ( 13 ) p ( one - tail ) Novelty 3 . 51 5 . 33 - 1 . 80 0 . 048 5 . 00 5 . 83 - 0 . 96 0 . 16 Usefulness 3 . 93 6 . 00 - 3 . 32 0 . 004 3 . 75 5 . 58 - 2 . 98 0 . 006 Cohesiveness 4 . 15 5 . 97 - 2 . 25 0 . 02 3 . 97 5 . 50 - 1 . 80 0 . 048 Journal ofEngineering Design 281 For the other two problems , the Bushing and Grinding problems , there were no significant differences in the creativity component scores , and thus different stimuli had no significant effects . 3 . 2 . 3 . Language results 3 . 3 . Experiment 1 discussion While pen - and - paper experiments are not the most conducive to examining language use , we were able to examine aggregated frequency ofstimulus use by counting and categorising stimulus words used by the participants . Many instances ofstimulus words found on worksheets were merely listed and not used in a phrase or sentence . When these ' unknown ' uses of stimuli are subtracted from the frequency totals , opposite - stimulus concepts incorporate significantly more stimulus words , t ( 5l ) = 2 . 791 , p = 0 . 0035 < 0 . 05 , than similar - stimulus concepts . This can be seen in Figure 3 , and suggests that using more opposite - stimulus words may result in more creative concepts . There was strong inter - rater agreement with regard to aggregated concept creativity measures . This is illustrated by the below - average usefulness scores for concepts involving individual seed processing ( concepts 2 and 13 ) , versus a higher score for concepts involving batch processing of multiple seeds ( concept 16 ) . This may reflect that batch processing is seemingly more practical . However , it should be noted that strong agreement does not necessarily mean that the raters were correct ( Siegal 1956 ) , nor does it necessarily account for other factors which may have influenced concept creativity . Despite any drawbacks associated with human ratings of creativity , creativity is a human judgment , and what is considered creative is not necessarily ' correct ' or ' incorrect ' . For two of the problems , the Sunflower and Egg problems , opposite - stimulus concepts were significantly more creative than similar - stimulus concepts . However , for the Bushing and Grinding problems , opposite - and similar - stimulus concepts were found to be equally creative . Results may reflect the difference in problem type and problem novelty . The Sunflower and Egg problems are general - domain problems that may be more novel to the participants , while the Bushing and Grinding problems are more technically oriented and should be familiar to most engineers , i . e . participants and raters . The results shown above suggest that opposite - stimulus words can Error Bars show 95 % CI of Mean 1 . 50 l : ~ l : ~ l : ~ 1 . 00 o ~ l : QI : : : : l cr ~ ~ . 50 Z : E N - o N . . . . Q ) . g . . . . U o N o \" T M o - Opp Stim Sim Figure 3 . Aggregated frequency of stimulus word - use with no unknown uses of stimulus words . 282 I . Chiu and LB . Shu stimulate more creative concepts for general - domain problems or problems that may be more novel to the designers . In cases where problems are more technically oriented , experience may neutralise the effects of stimuli . Both rater and participant experience affect creativity ; experience is an important factor in individual creativity ( Amabile 1989 , Akin 1990 ) . It is difficult to determine if participant concepts are based on designs encountered elsewhere , and difficult to separate historically creative concepts , concepts that never existed before in the world , from personally creative concepts , concepts that never existed before in the participant ' s mind , but existing in the world ( Amabile 1983 ) . In addition , a more experienced rater may tend to judge a participant ' s personally creative concepts as less creative even though they are creative and novel in the context of the experiment . A less experienced rater may judge the same concept as more creative . This may have been the case with the technically oriented problems , where both participants and raters should be familiar with a variety of alignment / insertion and material removal solutions . Raters may also judge concepts to be less creative if the participant explicitly refers to existing designs , even if concepts appear creative . Overall , it would be difficult to control for participant and rater experience . Regarding stimulus - word use , results show that opposite - stimulus concepts incorporated more stimulus words within the concept . This supports that opposite - stimulus participants were better able to use stimuli to introduce and develop new ideas in concept generation , which suggests that opposite - stimulus words are more useful for stimulating creative concepts . A more comprehensive verbal protocol , or talk - out - loud , experiment may provide further insight into how opposite stimulus words increase concept creativity . The next section describes such an experiment . 4 . Experiment 2 : An expanded verbal protocol experiment Experiment 2 is a verbal protocol experiment involving 14 participants generating concepts for three problems under one of the following experimental conditions : ( 1 ) No stimulus ( control ) ; ( 2 ) Similar stimulus ; ( 3 ) Opposite stimulus . Language use in relation to participant behaviour , as well as concept creativity , is examined in detail in Experiment 2 . 4 . 1 . Method 4 . 1 . 1 . Participants and procedure All 14 participants are fluent English speakers recruited from the Department of Mechanical and Industrial Engineering at the University of Toronto . Participants consisted of 13 males and one female , ranging from fourth - year undergraduate students to second - year PhD students . In individual sessions , participants first completed three training problems to habituate them to verbalising . Then , participants were instructed to verbalise all thoughts as they completed a series of three design problems . Fifteen minutes were allotted for each problem for a total experiment duration of45 min . Ten participants were provided with stimulus words , either opposite or similar words , while four were not provided with any stimuli . Ofthe 10 stimulus participants , five switched stimulus type between problems . Table 7 details the experimental design . Similar to the previous pen - and - paper experiment , participants in verbal protocol experiments were provided with worksheets presenting the problem statements and stimulus words . In contrast to pen - and - paper experiments , participants in verbal protocol experiments were also instructed to Journal ofEngineering Design 283 Table 7 . Experimental design for Experiment 2 . Stimulus participants Control participants Condition subtotals Prob . TH IS VT SW lL DRO DR UG MM DH DL 1M AF AP Opp . Sim . None Bushing S S S S S S 0 0 0 0 N N N N 4 6 4 Snow 0 0 0 S S S S 0 S 0 N N N N 5 5 4 Coal S S S S S S 0 0 0 0 N N N N 6 4 4 Notes : Reprinted with permission . Copyright ASME 2010 . S : Similar stimuli ; 0 : Opposite stimuli ; N : No stimuli . verbalise all thoughts as they worked on the design tasks . The sessions were recorded and fully transcribed for analysis . Verbal protocols are common for studying cognitive processes such as human machine interac tion ( Bainbridge et al . 1968 ) , medical decision - making ( Lutfey et al . 2008 ) and are considered a relatively objective and appropriate method for studying phenomena in design ( Cross et al . 1996 , Hubka and Eder 1996 , McNeill et al . 1998 , Chakrabarti et al . 2004 , Cross 2006 , Visser 2006 ) . In fact , verbalisation may be the most popular method for studying design cognition ( Coley et al . 2007 ) . However , there are some debates associated with verbal protocol studies . Nisbett and Wil son ( 1977 ) have questioned the accuracy of the data obtained from verbalisations as they have found that verbal reports do not necessarily match recordings of the reported event . Ericsson and Simon ( 1993 ) , on the other hand , contend that as long as verbalisations are immediate and do not require recall from memory , verbalisations accurately describe events being reported . Addi tionally , since verbal protocol experiments are a time and resource intensive method , the number of participants involved is usually small . A survey of design studies using this method reveals that the typical number of participants is low , e . g . 4 participants in a design cognition modelling study ( Benami and Jin 2002 ) , 8 participants in a personal creativity and design activities study ( Kim et al . 2011 ) , 10 participants in a design education study ( Atman and Bursic 1996 ) , and 20 participants in a design stimulation study ( Jin and Benami 2010 ) . Our sample size of 14 is reasonable considering the range of sample sizes typical of verbal protocol experiments . 4 . 1 . 2 . Problems and stimuli Participants were provided with three problems and related stimulus sets on worksheets . Specif ically , the problems were ( 1 ) Bushing - and - pin assembly ( 2 ) Snow insulation of houses and ( 3 ) Coal storage . Again , stimulus sets for the opposite and similar stimulus conditions were verbs in the root form generated by using a combination of a thesaurus ( Merriam - Webster . com 2008 ) and WordNet 3 . 0 ( Princeton University 2008 ) , starting from the original functional keywords . Some keywords do not have antonyms in the resources consulted , e . g . ' to insulate ' from the snow insulation problem , so opposite stimuli were generated based on opposition to the problem itself , e . g . ' to pack ' , as the problem specifically stated that ' packing ' of snow is undesirable . As gen erating opposite and similar verbs is not possible for all keywords , and oppositely related words are sparse for verbs to start with , this strategy was used for the other problems as well . Problems and stimulus sets are given in Table 8 . 4 . 1 . 3 . Concept identification and evaluation To reduce bias , design sessions were transcribed by an independent transcriptionist and an indepen dent concept reviewer was recruited to identify and code concepts from the free - form transcripts and worksheets . The following is a transcript excerpt representing approximately 30 s of one experiment session . Lines are numbered for referencing during analysis . 284 I . Chiu and L . H . Shu Table 8 . Summary of problems and stimulus sets for Experiment 2 . Prob . Bushing - and - pin assembly Snow Coal Problem description Parts that are automatically mated , e . g . a bushing and a pin , must be positioned so that their axes coincide . Using chamfers on mating parts does not solve the alignment problem . Develop a concept to centre mating parts that does not require high positioning accuracy ( Kosse 2004 ) In Canada , snow is readily available in winter and has good insulating qualities due to the amount of air in it . However , if the snow is packed to the point , it becomes ice , it is less insulating due to the loss of air . Come up with a concept to enable snow to be used as an additional layer of insulation for houses in the winter Clean coal and clean coal combustion technologies make it possible to generate cleaner electricity . That , combined with the increasing cost of oil and natural gas , power plant operators may consider converting or reconverting their power plants from oil or natural gas back to coal . However , there may not be enough land area near the plant that can be used for on - the - ground coal storage . Propose alternative solutions to a conventional coal pile ( adapted from Dieter 2000 ) Opposite stimulus words Original keywords : Opposite of align and insert Stimulus words : Change , disorder , disarrange , scramble , randomise , misalign , tumble , skew , move , expel , pull , eject , evict Original keywords : pack and compact Stimulus words : Arrange , bundle , change , compress , constrict , contract , force , impact , move , push , squeeze , tighten , wad Original keyword : opposite of store Stimulus words : Abandon , discard , discharge , dispense , disperse , dispose , distribute , export , remove , scatter , spread , waste Similar stimulus words Original keywords : align and insert Stimulus words : Inject , transplant , sandwich , connect , skew , mount , misalign , attach , join , reorient , adjust , modify , match Original keywords : insulate and surround Stimulus words : Blanket , control , cover , defend , enclose , immerse , pack , preserve , prevent , restrain , restrict , submerge , touch Original keyword : store Stimulus words : Accumulate , collect , displace , distribute , feed , give , heap , keep , place , supply , transfer , withhold Note : Reprinted with permission . Copyright ASME 2010 . ( l ) I think the obvious . . . ( 2 ) the first thing that comes to mind is that you ' d like blanket the house . . . uh ( 3 ) essentially blanket the house in a layer , in a thin layer ofsnow . . . um ( 4 ) If the snow is packed to the point that it becomes ice . . . ( 5 ) I guess , you ' d obviously try to figure out what amount ofpacking you ' d have to do ( 6 ) to restrict the snow from becoming ice due to over packing . Finished transcripts were corrected for minor spelling errors , e . g . ' chamfer ' for ' camphor ' , ' pedal ' for ' petal ' , but were otherwise not annotated nor changed . Concepts were identified by reviewing transcripts and worksheets . Concepts identified by the independent reviewer were compared with the concepts identified by the investigators who also - - - _ . _ - _ . Journal ofEngineering Design 285 reviewed all worksheets and transcripts . In cases where there was disagreement between the inde pendent reviewer and investigators , both identified concepts were added to the set ofconcepts . For example , in a transcript segment from the Coal problem , both the independent reviewer and the investigators identified an ' underground storage concept ' in which the coal would be stored under ground . However , only the independent reviewer also identified a ' storage pile ' as a concept . The investigators did not regard the ' storage pile ' as a concept because the problem statement specifi cally required concepts other than a ' conventional coal pile ' . Although the investigators disagreed with the independent reviewer with regard to the identification ofthe ' storage pile ' as a concept , the ' storage pile ' concept was added to the set of concepts to be evaluated by the raters . This concept identification process helped to ensure that all possible concepts were included for evaluation . The independent reviewer also summarised participants ' instances of similar concepts into a single concept type . For example , in one transcript segment for the Coal problem , multiple refer ences to the ' tower concept ' or ' condominium concept ' are in fact only one concept that involved storing coal in a tall structure . A total of 195 concepts were identified between all participants and problems , 59 , 59 and 77 concepts for the Bushing , Snow and Coal problems , respectively . Concept creativity was evaluated using the anchoring and direct - scaling method developed and described in Experiment 1 . Three raters were recruited and consisted ofone female and two males , all with knowledge and interest in engineering design . Raters were not provided with the identity of the designers , nor the stimulus condition under which the concepts were generated . Concepts were presented to the raters in random order . 4 . 1 . 4 . Analysis First , inter - rater agreeability was calculated using Kendall ' s W . Then , rater scores for each concept were averaged , and all concept scores for the same participant were aggregated to facilitate analysis . This produces an aggregated novelty , usefulness and cohesiveness score for each of the 14 participants . Aggregated scores were analysed using a mixed - model analysis of variance ( ANaYA ) . As in Experiment 1 , rater scores were averaged to facilitate analysis because our main interest is not in rater behaviour , but in participant behaviour . Because five participants switched stimulus type between problems during the experiment ( identified as TH , JS , YT , MM and DR in Table 7 ) , pseudo - replicates were created to model these participants as independently contributing to each stimulus condition , effectively increasing the sample size to 19 from 14 . This is a common technique to deal with scenarios where not all participants contribute independently to only one experimental condition over multiple trials . Generally , the use of pseudo - replicates results in a conservative estimate of differences ( L . Duquette , personal communication , 2009 ) . 4 . 2 . Results 4 . 2 . 1 . Inter - rater agreement Kendall ' s W was calculated for each of the problems to examine inter - rater agreement and shown in Table 9 . All values for W show a medium to large agreement between the raters , and are statistically significant , p < 0 . 05 . Therefore , there is significant agreement between the raters . 4 . 2 . 2 . Creativity results Overall , raters judged opposite - stimulus concepts to be more creative than similar - stimulus con cepts . This difference was significant , or borderline significant , for all three creativity metrics . See - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 286 I . Chiu and L . H . Shu Table 9 . Kendall ' s W for Experiment 2 concepts . Problem Novelty Usefulness Cohesiveness Bushing ( N = 59 ) W = 0 . 62 W = 0 . 53 W = 0 . 57 X 2 ( 58 ) = 143 . 61 X 2 ( 58 ) = 122 . 76 X 2 ( 58 ) = 131 . 28 p < 0 . 0001 p < 0 . 0001 p < 0 . 0001 Snow ( N = 59 ) W = 0 . 55 W = 0 . 60 W = 0 . 65 X 2 ( 58 ) = 127 . 25 X 2 ( 58 ) = 138 . 49 X 2 ( 58 ) = 150 . 17 p < 0 . 0001 p < 0 . 0001 p < 0 . 0001 Coal ( N = 77 ) W = 0 . 51 W = 0 . 41 W = 0 . 47 X 2 ( 76 ) = 153 . 99 X 2 ( 76 ) = 126 . 13 x 2 ( 76 ) = 143 . 68 p < 0 . 0001 p = 0 . 0003 p < 0 . 0001 Notes : Reprinted with permission . Copyright ASME 2010 _ Shaded rows show significant or borderline significant differences . a Adjusted for effects of problem order - Table 10 and Figure 4 for planned contrasts and interaction graphs of all creativity results . Note the lines in the interaction graph do not signify a relationship , but assist in visualising variable effects and interactions . For novelty , a significant main effect was found for Stimulus Type F ( 2 , 27 . 58 ) = 7 . 09 , p = 0 . 003 , p < 0 . 05 . Planned contrasts comparing individual experiment conditions , e . g . opposite - stimulus concepts versus no - stimulus concepts , show a significant difference between opposite - stimulus and similar - stimulus concepts , t ( 27 . 65 ) = - 3 . 02 , p = 0 . 0025 , p < 0 . 05 . For usefulness and cohesiveness , planned contrasts show that opposite - stimulus concepts are border line significantly more useful and cohesive than similar - stimulus concepts , t ( 31 . 5l ) = - 1 . 61 , P = 0 . 059 , p ' \" ' - ' 0 . 05 and t ( 29 . 42 ) = - 1 . 62 , p = 0 . 058 , p ' \" ' - ' 0 . 05 , respectively . Problem order , the order in which problems were completed , was found to have an effect on cohesiveness and was corrected in the planned contrasts . For novelty , usefulness and cohesiveness , planned contrasts show no significant difference between opposite - stimulus and no - stimulus ( control ) concepts . Overall , the ANOVAs and planned contrasts support the original hypothesis that opposite stimulus concepts are more novel , useful and cohesive than similar - stimulus concepts . However , we also observe two results that contradict much of the literature with regard to design stimulation : Contrast t - and p - values Expt . Condo 2 Comparison of estimated means Expt . Condo 1 Planned contrast results for concept novelty , usefulness and cohesiveness . Nov . Use . Coho Table 10 . ( 1 ) Opposite - stimulus and no - stimulus concepts were found to be equally creative ; ( 2 ) No - stimulus ( control ) concepts were found to be more creative than similar - stimulus concepts . These results will be further discussed in a later section . Journal ofEngineering Design 287 6 . , - - - - - - - - . . , . - - - - - - - t ' 4 + - - - - - - - - r ~ ~ - - - = - - - . . ~ ~ i ~ . . . . . . None - Sim . - Opp . 2 + - - - - - - - - - - - - - - Coal Saow Problem B . lhla & o - - - - - - - - - - - - - ~ Ii , . . - - - - - - - . - - - - - - - . - - 5 - 1 - - - - - - - - - - - - - - o + - - - - - - - - - - - - - ~ Ii - . - - - - ' - - . - - - - - - - - . 0 + - - - - - - - . - - - - . . . . . . , . . - - - - . . . . I r . 5 I ! 4 t - - ; : ~ ~ < . . . . . ~ . . . . . . . - . . . - . . = = , . . . . . . = . ~ = = . . . - . . . ~ . . . . . . . ~ . . . - j i : t - - - - - - - - - - - - ~ ~ Figure 4 . Graphs of concept novelty , usefulness and cohesiveness ratings . Note the lines in the interaction graph do not signify a relationship , but assist in visualising variable effects and interactions . Reprinted with permission . Copyright ASME201O . An explicitcontent analysis was performed using experimenttranscripts to determine how stimulus participants used stimulus words in the concept generation process . A parameter of interest is the POS in which the stimulus words were used , and therefore , stimulus words used by participants were categorised as a noun , verb or adjective . Another parameter of interest corresponds to the NWPs introduced by stimulus words . NWPs are of interest because they may indicate that new concept elements were being introduced into the concept generation process . In tum , new concept elements may form the basis of creative concepts . The investigators were able to identify NWPs by comparing words and phrases in the problems and stimulus sets with the words and phrases generated by the participants . In an example from the Snow problem , given the stimulus word ' constrict ' , sentence A below contains an NWP ( underlined ) , while sentence B does not contain anNWP . 4 . 2 . 3 . Language results ( A ) Constrict the motion of heat [ from leaving the house ] ( B ) . . . constrict snow . In sentence A , the phrase ' motion of heat ' , associated with the stimulus word ' constrict ' , was not given as part of the design problem nor stimulus set . However in sentence B , the word ' snow ' , also associated with the stimulus word ' constrict ' , was given as part of the design problem . Overall , results suggest ( l ) verbs introduce more NWPs and ( 2 ) opposite - stimulus verbs intro duce the most NWPs for two of the problems , the Bushing and Snow problems . This can be seen in the interaction graphs in Figures 5 - 7 . For each problem , a two - way ANOVA was used to compare the effect of Stimulus POS ( verb , noun or adjective ) and Stimulus Type ( opposite or similar stimuli ) on NWPs . First , theANOVA is ~ . ~ _ . . . . . _ - _ . _ - - - - - - - - - - - - - - - - - 288 . . . S c \" \" E ~ 4 ~ \" E 1 3 ~ . . . ~ 2 I . Chiu and L . H . Shu . . . . . . . . . . . . . . . . . . . . . . . . . . . POS _ NEW - - - N - V A Figure 5 . Bushing language results comparing effects of pas and Stimulus Type on estimated marginal means ofNWP . N . . . . . . o N . . . v . g - U o N o ~ M o . . . . . . sim StimType opp used to determine if there are Stimulus POS effects on NWP introduction . We found that Stimulus POS has a significant effect on the introduction ofNWPs , F ( 2 , 16 ) = 7 . 28 , p = 0 . 006 , F ( 2 , 16 ) = 40 . 42 , P = 0 . 000 and F ( 2 , 16 ) = 10 . 24 , p = 0 . 001 , for the Bushing , Snow and Coal problems respectively . For all three problems , planned contrasts show that stimuli used as verbs introduce significantly more NWPs , or borderline significantly more NWPs , as seen in Tables 11 - 13 . Second , the ANOVA is used to determine if there are Stimulus Type and Stimulus Type * POS interaction effects on NWPs . For the Bushing problem , Figure 5 suggests that opposite stimuli introduce the most NWPs . However , there is no significant effect of Stimulus Type on NWPs nor Stimulus Type * POS interaction , meaning that one Stimulus Type POS , e . g . opposite - stimulus verbs , does not significantly introduce more or fewer NWPs . 10 \" n\u00b7N - v A \u2022\u2022 , \u2022\u2022\u2022\u2022 # . . . . . . . . . - ' \" . . . . . . . . . . . . . . . . . . . . . o sim StimTvpe opp Figure 6 . Snow language results comparing effects of pas and Stimulus Type on estimated marginal means of NWP . ~ ~ - ~ - - ~ . _ - - - - - - - - - - - - - - - - - - - Journal ofEngineering Design 289 Figure 7 . Coal language results comparing effects of POS and Stimulus Type on estimated marginal means of NWP . Table II . Bushing problem contrast results . POS _ New \u00b7 . . . \u00b7N - V A F - and p - values F - and p - values F ( l , 8 ) = 4 . 61 , p = 0 . 064 - 0 . 05 F ( I , 8 ) = 2 . 80 , p = 0 . 13 opp F ( l , 8 ) = 41 . 76 , p = 0 . 000 < 0 . 05 F ( I , 8 ) = 2 . 90 , p = 0 . 13 StimType Noun : 2 . 30 Noun : 2 . 30 , . . . . . . . . . . . . . . . . . . . . . . . . . . ( : : Noun : 2 . 20 Noun : 2 . 20 sim Mean NWP introduction o 2 10 . . . 8 c = E 16 ' e - ~ \" D 4 ! l . . E i . . , Mean NWP introduction Verb : 5 . 10 Adj : 1 . 10 Table 12 . Snow problem contrast results . Verb : 7 . 10 Adj : 1 . 20 Table 13 . Coal problem contrast results . Mean NWP introduction F - and p - values Verb : 7 . 20 Adj : 2 . 70 Noun : 2 . 70 Noun : 2 . 70 F ( I , 8 ) = 47 . 58 , p = 0 . 000 < 0 . 05 F ( l , 8 ) = 2 . 90 , p = 0 . 13 For the Snow problem , Figure 6 suggests that opposite stimuli introduce the most NWPs . There is a significant effect of Stimulus Type on the introduction ofNWPs , F ( 1 , 8 ) = 5 . 26 , p = 0 . 051 , . . . , . 0 . 05 . Additionally , there is a significant interaction of Stimulus Type * PaS , F ( 2 , 16 ) = 13 . 88 , p = 0 . 000 . Contrasts for Stimulus Type * PaS interaction comparing opposite - stimulus verbs to other pas show significant differences , F ( l , 8 ) = 18 . 94 , p = 0 . 002 , indicating that different stimulus pas have different effects depending on the Stimulus Type . In this case , it indicates that opposite - stimulus verbs introduce significantly more NWPs . For the Coal problem , unlike the Bushing and Snow problems , Figure 7 suggests that similar stimuli introduce more NWPs rather than opposite stimuli . The ANaVA shows a significant effect 290 I . Chiu and L . H . Shu of Stimulus Type on NWP introduction , F ( l , 8 ) = 16 . 76 , p = 0 . 003 , while there is no significant interaction of Stimulus Type * POS . These results are discussed in detail in the next section . 4 . 3 . Experiment 2 discussion Experiment 2 shows opposite - stimulus concepts were significantly , or borderline significantly , more novel , useful and cohesive than similar - stimulus concepts . These results support our original hypothesis that opposite stimuli result in more creative concepts . These results are consistent with previous results obtained from Experiment 1 . However , surprisingly , Experiment 2 also shows that stimulus concepts in general were not more creative than no - stimulus ( control ) concepts . This is contrary to the literature and results of other design stimulation studies . It is unclear why no - stimulus concepts were judged more creative than some stimulus con cepts , specifically similar - stimulus concepts . Further work is required to explain the discrepancy between results found in this experiment , and results reported by others . However , we have noted that many other design stimulation experiments were pen - and - paper , e . g . Tseng et at . ( 2008 ) . The discrepancy observed in Experiment 2 may have resulted from a limitation of experimental methodology ; it is possible that the requirement to use stimuli in addition to verbalising and designing increased the cognitive workload of stimulus participants past optimal performance . It is well - known that increased mental workload can decrease human performance ( Wickens and Hollands 2000 , Drews et at . 2009 ) , where human performance is creativity in this case . Table 14 enumerates the tasks performed by no - stimulus participants versus stimulus participants and shows that stimulus participants perform an additional task compared to no - stimulus participants . A cognitive workload assessment , e . g . NASA Task Load Index , can be used to determine if stimulus use in verbal protocol design experiments will increase cognitive workload ( and thus decrease performance in terms of creativity ) . For example , the NASA Task Load Index asks participants to rate perceived workload in several dimensions during or after completion of a task ( Hart and Staveland 1988 ) . Other methods can also be used to determine the cognitive workload associated with design tasks . For example , Tang and Zeng ( 2009 ) investigated the use of body movements to quantify a designer ' s mental stress during the design process . As it may not be possible to equalise the workload between the stimulus and control conditions , i . e . comparing designs generated using stimulus versus no stimulus , this may be a serious limitation of verbal protocol experiments as applied to design research . Review of other verbal protocol experiments reveal at least one other reported case in which an experimental manipulation ( which should improve design overall ) did not produce better design concepts despite an increase of other metrics ( Atman and Bursic 1996 ) . In Atman and Bursic ' s study , reading a short design text before verbalising and designing improved metrics that should indicate improved design , e . g . time spent on designing , but did not result in better quality designs . More details with respect to potential limitations of verbal protocol design experiments are described by Chiu and Shu ( 2010 ) . Language analyses were conducted to gain insight into how designers used stimulus words in concept generation . First , stimulus words used in concept generation were examined to determine Table 14 . Task comparison between control and stimulus participants . No - stimulus tasks 1 . Design 2 . Verbalise Stimulus tasks 1 . Design 2 . Verbalise 3 . Use stimuli Note : Reprinted with permission . Copyright ASME 2010 . - - - - . _ - - - - - - . _ - - - - - - - - - - - - - - - - - - Journal ofEngineering Design Table 15 . Breakdown of average unknown usage of stimulus words . Average Average unknown unknown usage of usage of opposite similar Comparison Problem stimulus words stimulus words t - values p - values Bushing 26 . 0 16 . 3 t ( 8 ) = 1 . 10 0 . 15 Snow 18 . 6 17 . 8 t ( 8 ) = 0 . 12 0 . 45 Coal 31 . 0 17 . 0 t ( 8 ) = 2 . 07 0 . 04 291 the POS in which they were used , Le . verb , noun , adjective or unknown , and then any associated NWPs were examined . Language analyses showed that verbs in general introduced more NWPs than nouns or adjectives . Furthermore , results indicate that opposite verbs may introduce the highest number of NWPs , and correspond to increased creativity measures as determined by the independent raters . Combined results of the Bushing and Snow problems suggest that opposite - stimulus verbs may be the mechanism causing participants to introduce more NWPs , thus resulting in more creative concepts . However , the results of the Coal problem were contrary to those of the other two prob lems , in that similar - stimulus verbs introduced more NWPs than opposite - stimulus verbs . Despite the contradicting NWP introduction results for the Coal problem , opposite - stimulus concepts were still judged as more creative than similar - stimulus concepts . An explanation for this inconsistent result may be found by examining the high rate of unknown stimulus use in the Coal problem . Re - examining stimulus - use frequency for the Coal problem shows that opposite - stimulus partici pants had significantly more instances ofunknown stimulus use , e . g . stimulus words listed without context and hence with unknown POS , on average than in any other experimental condition or problem , t ( 8 ) = 2 . 07 , p = 0 . 04 . The average unknown stimulus POS per problem is shown in Table 15 . Stimulus words used as an unknown POS do not introduce NWPs , but frequent instances of recorded unknown stimuli may indicate that the participant frequently looked at specific stimulus words and likely thought about those words as they designed . It is possible that participants were fatigued during the Coal problem , the last problem in the experiment , and did not verbalise all thoughts related to the task . 5 . Overall discussion Our experiments showed that oppositely related stimuli can increase concept creativity . Results suggest that use ofopposite stimuli is most effective for general domain problems , or problems that are novel to the designer , e . g . the Sunflower , Egg and Snow problems . Experiment 2 also suggests that stimuli in general may be detrimental to designer performance , but this may be a limitation of the experimental method . Despite any potential methodological limitations , Experiment 2 did allow a comparative study ofhow language stimuli may support more creative concept generation . Experiment 2 results show that opposite - stimulus words are associated with more NWPs , which further suggests that opposite - stimulus verbs may force introduction of NWPs so that they are correct and consistent in the context of the problem . While participants used stimuli as verbs and nouns often ( and adjectives infrequently ) , verbs may be better at introducing NWPs because verbs are more flexible than nouns ( Gentner and France 1988 ) . Verbs are also known to unconsciously and automatically evoke concepts corresponding to the semantic filler roles typically associated with the event they denote ( Lyons 1977 , McRae et al . 2005 ) . Common semantic filler roles include 292 1 . Chiu and LH . Shu Table 16 . Semantic properties ofnouns and verbs , adapted from Croft ( 1991 ) . Property Noun Verb Semantic class Object Action Pragmatic function Reference Predication Stativity State Process Persistence Persistent Transitory Valencyjrelationality 0 1 + patients ( often direct objects ) , agents ( often subjects ) , instruments and locations . For example , the verb ' hammer ' will commonly evoke ' carpenter ' as an agent role , and ' nail ' as an object role . In the Snow problem , opposite - stimulus participants were provided with the stimulus word ' constrict ' . Those using it as a verb automatically realised that the verb ' constrict ' had to ' constrict something ' . ' Constrict snow ' ( to the point where it turns to ice ) was inconsistent with the problem because participants were explicitly instructed not to compress snow to the point of ice . However , introducing an NWP in ' constrict the motion of the heat ' , is both new , and consistent . The verb ' constrict ' implies that something must be constricted , and it is flexible enough to allow for different arguments while still ' making sense ' . Examining semantic properties of nouns and verbs further explain the relationship between stimulus verbs and NWPs . Table 16 summarises some key properties and differences between nouns and verbs . Table 16 shows verb properties that may benefit conceptual design more than noun properties . First , nouns are used to reference objects , which are usually fixed and static , while verbs represent transitory actions and processes ( Croft 1991 ) . Fixedness is contrary to the purpose of conceptual design , which is to expand the solution space , and not to fix , or limit the space . The abstractness of actions and processes allows verbs to avoid naming an actual design solution , which renders verbs more advantageous for expanding the solution space . Second , nouns have a valency of zero , while verbs have at least a valency of one . The valency , or relationality , of a word refers to the implied entities associated with the use of the word . When a noun is used , there is no other implied entity , e . g . ' book ' does not imply the existence of any other entity . However , when a verb is used , at least one other entity is implied , e . g . ' hit ' implies the existence of a hitter and the object hit ( Lyons 1977 , Croft 1991 , McRae et al . 2005 ) . For verbs , the implied entities are the semantic filler roles . Similar stimulusword used consistently with current problem Novel and creative concepl Figure 8 . Explanatory model of opposite - verb NWP introduction and its effect on concept creativity . Journal ofEngineering Design 293 Examining the semantic properties of nouns versus verbs aids the explanation of the empirical results obtained in Experiment 2 . Both linguistic theory and empirical results suggest that verbs used as design stimuli may increase concept creativity . Furthermore , empirical results show that oppositely related verbs may be the most effective at stimulating NWPs , which may lead to more creative concepts . With similarly related verbs , NWP introduction is unnecessary because existing problem - statement words and phrases capture the current problem state , e . g . problem objects like ' snow ' or ' bushings ' . Similar stimuli are already consistent with the problem and do not need to be resolved by the introduction of NWPs . Figure 8 models the effect of opposite - stimulus verbs on the generation of creative concepts . 6 . Summary and concluding remarks We investigate the effects of language on design because connections between language and cognition may be used to facilitate creative and successful design . Specifically , we study oppo sitely related words because they may stimulate creative concept generation by being unexpected and non - obvious , while being available for systematic retrieval in lexical resources . Through experiments , we observed the effects of oppositely and similarly related stimulus words on con cept creativity and designer behaviours with respect to the different stimulus types in concept generation . Empirical results support our original hypotheses and show that opposite stimuli may increase : ( l ) Concept novelty as well as other creativity metrics ; ( 2 ) Introduction of NWPs that may form the basis of novel and creative concepts . Empirical results combined with linguistic theory allow us to speculate on the mechanism in which opposite stimuli interacts with the conceptual design process to produce more creative concepts . While similar - stimulus words can be used ' as - is ' to reason about the problem and concepts consistently , opposite - stimulus words must be used with NWPs to maintain consistency within the problem . These new words or phrases may be key to the formation of more creative concepts . Results also reinforced that problem novelty and designer experience may be a factor ( Exper iment 1 ) , and that methodological issues with verbal protocols may interfere with results ( Experiment 2 ) . However , overall , our results show that opposite stimuli appear a practical means of stimulating creative design , that is simple to implement , e . g . using flashcards or worksheets . Unlike some creative design methods , e . g . TRIZ , which require training and related materials , e . g . contradiction tables or software , opposite stimuli only require the generation of words that are oppositely related to the problem . These words , which are familiar to most as antonyms , can be obtained from conventional thesauri or WordNet . The ease of obtaining oppositely related words , i . e . in antonyms , increases the ease by which this method can be integrated into engineering practice . While the designer ' s previous experience with similar design problems may offset the creative advantages offered by opposite - stimulus words , opposite - stimulus words are suitable for instances of conceptual design where the engineer may not have already gained familiarity with the new domain . It is unlikely that the effects of language on design cognition and creativity can be fully under stood in the near future . However , insights from our experiments combined with linguistic theory allow us to propose an explanatory model of interaction between language stimuli and design cognition . This knowledge can be used to facilitate more creative and successful design . 294 Acknowledgements I . Chiu and L . H . Shu N , . . . , o N . . . . v . g . . . . . U o N o \" \" \" M o , . . . , We gratefully acknowledge the Natural Sciences and Engineering Research Council of Canada ( NSERC ) for funding support . We would also like to acknowledge the participants and raters who provided their time . References Adams , J . L . , 2001 . Conceptual blockbusting , a guide to better ideas 4 / e . New York , NY : Perseus Press . Akin , 0 . , 1990 . Necessary conditions for design expertise and creativity . Design Studies , II ( 2 ) , 107 - 113 . Akin , O . and Akin C . , 1998 . On the process of creativity in puzzles , inventions and designs . Automation in Construction , 7 ( 2 - 3 ) , 123 - 138 . Altshuller , G . S . and Shulyak , L . , 1996 . Andsuddenly the inventorappeared : TRIZ , the theory ofinventive problem solving , 2 / e . Worcester , MA : Technical Innovation Center . Amabile , T . M . , 1983 . The social psychology ofcreativity . New York : Springer - Verlag . Amabile , T . M . , 1989 . Growing up creative , nurturing a lifetime ofcreativity . New York , NY : Crown Publishers , Inc . Atman , C . J . and Bursic K . , 1996 . Teaching engineering design : can reading a textbook make a difference ? Research in Engineering Design , 8 ( 4 ) , 240 - 250 . Bainbridge , L . , et al . , 1968 . A study of real - time human decision making using a plant simulator . Operational Research Quarterly , Special Conference Issue , 19 , 91 - 106 . Reprinted in E . Edwards and EP . Lees , eds . ( 1974 ) . The human operator in process control . London : Taylor and Francis . Benami , O . and Jin , Y . , 2002 . Creative stimulation in conceptual design . Proceedings ofthe ASME DETCjCIE , Montreal , Canada , DETC2002 / DTM - 34023 . Besemer , S . P . and Treffinger , D . l . , 1981 . Analysis ofcreative products : review and synthesis . Journal ofCreative Behavior , IS ( 3 ) , 158 - 178 . de Bono , E . , 1970 . Lateral thinking : creativity step by step . New York , NY : Harper & Row . de Bono , E . , 1992 . Serious creativity . New York , NY : HarperCollins . Brown , D . C . , 2008 . Guiding computational design creativity research . Proceedings ofNSF International Workshop on Studying Design Creativity ' 08 , University of Provence , France . Bruner , J . S . , 1964 . The course of cognitive growth . American Psychologist , 19 ( 1 ) , I - IS . Carroll , D . W . , 1999 . Psychology oflanguage , 2 / e . Pacific Grove , CA : Brooks / Cole Publishing Company . Chakrabarti , A . , Morgenstern , S . , and Knaab , H . , 2004 . Identification and application of requirements and their impact on the design process : a protocol study . Research in Engineering Design , 15 ( I ) , 22 - 39 . Chiu , I . and Shu , L . H . , 2007 . Biomimetic design through natural language analysis to facilitate cross - domain information retrieval . Artificial Intelligence for Engineering Design , Analysis and Manufacturing , 21 ( I ) , 45 - 59 . Chiu , I . and Shu , L . H . , 2008a . Use of opposite - relation lexical stimuli in concept generation . Annals ofthe CIRP , 57 ( 1 ) , 149 - 152 . Chiu , I . and Shu , L . H . , 2008b . Effects of dichotomous lexical stimuli in concept generation . Proceedings of ASME international design engineering technical conferences , New York City , NY , USA , 3 - 6 August 2008 . DETC2008 49372 ( DTM ) . Chiu , I . and Shu , L . H . , 20I O . Potential limitations ofverbal protocols in design experiments . Proceedings ofASMEinterna tionaldesign engineering technical conferences , Montreal , Quebec , Canada , 15 - 18August 2010 . IDETC2010 - 28675 ( DTM ) . Cheong , H . , et al . , 2011 . Biologically meaningful keywords for functional terms of the functional basis . Journal of Mechanical Design , 133 ( 2 ) , 021007 : I - II . Chomsky , N . , 1968 . Language and mind . New York , NY : Harcourt Brace Jovanovich . Coley , E , Houseman , 0 . , and Roy , R . , 2007 . An introduction to capturing and understanding the cognitive behaviour of design engineers . Journal ofEngineering Design , 18 ( 4 ) , 311 - 325 . Corremans , J . AM . , 2011 . Measuring the effectiveness of a design method to generate form alternatives : an experiment performed with freshman students product development . Journal ofEngineering Design , 22 ( 4 ) , 259 - 274 . Croft , W . , 1991 . Syntactic categories and grammatical relations : the cognitive organization ofinformation . Chicago , IL : University of Chicago Press . Cross , N . , 2006 . Designerly ways ofknowing . London : Springer - Verlag . Cross , N . , Christiaans , H . , and Drost , K . , 1996 . Introduction : the Delft protocols workshop . In : N . Cross , H . Christiaans , and K . Drost ' s , eds . Analysing design activity . West Sussex : John Wiley & Sons , 1 - 16 . Dieter , G . E . , 2000 . Engineering design : a materials and processing approach . 3rd ed . New York , NY : McGraw - Hill . Dong , A . , Hill , AW . , andAgogino , AM . , 2003 . A document analysis method for characterizing design team performance . Journal ofMechanical Design , 126 ( 3 ) , 378 - 385 . Dong , A , 2006 . Concept formation as knowledge accumulation : a computational linguistics study . Artificial Intelligence for Engineering Design Analysis & Manufacturing , 20 ( 1 ) , 35 - 53 . Drews , EA , et aI . , 2009 . Text messaging during simulated driving . Human Factors : The Journal ofthe Human Factors and Ergonomics Society [ online ] . Available from : http : / / hfs . sagepub . com [ Accessed 17 January 2010 ] . DTM , 2006 . Design that matters design challenge portfolio : shelling machines [ online ] . Available from : http : / / www . designthatmatters . org / [ Accessed 17 October 20061 . Journal ofEngineering Design 295 Engen , T . , 1972 . Psychophysics II . Scaling methods . In : J . w . Kling , L . A . Rigg , Woodworth & Scho1sberg , eds . Experimental psychology . 3rd ed . New York , NY : Holt Rinehart & Winston , 47 - 86 . Ericsson , KA . and Simon , H . A . , 1993 . Protocol analysis : verbal reports as data . Cambridge , MA : MIT Press . Fantoni , G . , Taviani , C , and Santoro , R . , 2006 . Design by functional synonyms and antonyms : a structured creative technique based on functional analysis . Proceedings ofthe Institution ofMechanical Engineers , Part B , Journal of Engineering Manufacture , 221 ( 4j2007 ) , 673 - 683 . Fellbaum , C . , 1993 . English verbs as a semantic net , in five papers on WordNet , 40 - 61 [ online ] . Available from : ftp : jjftp . cogsci . princeton . edujpubjwordnetj5papers . ps [ Accessed 7 July 2008 ] . Gentner , D . and France , 1 . , 1988 . The verb mutability effect : studies of the combinatorial semantics of nouns and verbs . In : S . Small , G . Cottrell , and M . Tanenhaus , eds . Lexical ambiguity resolution . Los Altos , CA : Morgan Kaufmann , 343 - 382 . Gero , J . S . , Sushil , J . L . , and Kundu , S . , 1994 . Evolutionary learning of novel grammars for design . Artificial Intelligence for Engineering Design Analysis & Manufacturing , 8 ( 2 ) , 83 - 94 . Gordon , W . J . J . , 1961 . Synectics , the development ofcreative capacity . New York , NY : Harper & Row . Hacco , E . and Shu , L . , 2002 . Biomimetic concept generation applied to design for remanufacture . Proceedings ofASME design engineering technical conferences , Montreal , DETC2002jDFM - 341 . Hart , S . G . and Staveland , L . E . , 1988 . Development of NASA - TLX ( task load index ) : results of empirical and theoretical research . In : P . A . Hancock and N . Meshkati , eds . Human mental workload . ElsevierScience : NorthHolland , 139 - 183 . Hatchuel , A . and Weil , B . , 2009 . C - K design theory : an advanced fortnulation . Research in Engineering Design , 19 ( 4 ) , 181 - 192 . Holt , K , 1993 . Computer - aided creativity in engineering design . Journal ofEngineering Design , 4 ( 4 ) , 371 - 376 . Howard , T . J . , Culley , S . J . , and Dekoninck , E . , 2008 . Describing thecreativedesign process by the integrationofengineering design and cognitive psychology literature . Design Studies , 29 ( 2 ) , 160 - 180 . Hubka , V . and Eder , W . E . , 1996 . Design science - introduction to the needs , scope and organization ofengineering design knowledge . London : Springer - Verlag . Jackendoff , R . , 1983 . Semantics and cognition . Cambridge , MA : MIT Press . Jin , Y . and Benami , 0 . , 2010 . Creative patterns in conceptual design , artificial intelligence for engineering design . Analysis and Manufacturing , 24 ( 2 ) , 191 - 209 . Jin , Y . , Geslin , M . , and Lu , S . C - Y . , 2007 . Impact of argumentative negotiation on collaborative engineering . Annals of the CIRP , 56 ( 1 ) , 181 - 184 . Kalmar , 1 . and Davidson , D . , 1997 . Anthropological linguistics and semiotics . 2nd ed . Toronto , Canada : Quirk Press . Kan , J . T . w . and Gero , J . S . , 2007 . Using the FBS ontology to capture semantic design infortnation in design protocol studies . In : J . McDonnell and P . Lloyd , eds . DTRS7 . London : University of the Arts , 155 - 165 . Keller , R . , Eckert , CM . , and Clarkson , P . J . , 2009 . Using an engineering change methodology to supportconceptual design . Journal ofEngineering Design , 20 ( 6 ) , 571 - 587 . Kim , Y . S . , Jin , S . T . , and Lee , S . w . , 2011 . Relations between design activities and personal creativity modes . Journal of Engineering Design , 22 ( 4 ) , 235 - 257 . Kosse , V . , 2004 . Solving problems with TRIZ : an exercise handbook . 2nd ed . Southfield , M1 . : Ideation Int ' l Inc . Kudrowitz , B . M . and Wallace , D . R . , 2010 . Assessing the quality of ideas from prolific , early - stage product ideation . Proceedings ofthe ASME IDETCjCIE 2010 , Montreal , Quebec , Canada , August 15 - 18 , 2010 ( DETC201O - 28991 ) . Landis , J . R . and Koch , G . G . , 1977 . The measurement of observer agreement for categorical data . Biometrics , 33 ( 1 ) , 159 - 174 . Levinson , S . , 1996 . Language and space . Annual Review ofAnthropology , 25 , 353 - 382 . Li , P . and Gleitman , L . , 2002 . Turning the tables : language and spatial reasoning . Cognition , 83 ( 3 ) , 265 - 294 . Lutfey , KE . , et ai . , 2008 . How are patient characteristics relevant for physicians ' clinical decision making in diabetes ? An analysis of qualitative results from a cross - national factorial experiment . Social Science & Medicine , 67 ( 9 ) , 1391 - 1399 . Lyons , J . , 1977 . Semantics . vol . 2 . Cambridge , NY : Cambridge University Press . McNeill , T . , Gero , J . S . , andWarren , J . , 1998 . Understanding conceptualelectronicdesign using protocol analysis . Research in Engineering Design , 10 ( 3 ) , 129 - 140 . McRae , K , etai . , 2005 . A basisfor generating expectancies for verbs from nouns . Memory & Cognition , 33 ( 7 ) , 1174 - 1184 . Merriam - Webster . com , 20as . Merriam - Webster online dictionary [ online ] . Available from : http : jjmerriam - webster . com [ Accessed 23 September 2008 ] . Miller , G . , et ai . , 1993 . Introduction to WordNet : an on - line lexical database , in Five papers on WordNet , 1 - 25 [ online ] . Available from : ftp : jjftp . cogsci . princeton . edujpubjwordnetj5papers . ps [ Accessed 7 July 2008 ] . Murphy , M . L . , 2003 . Semantic relations and the lexicon : antonymy , synonymy and other paradigms . Cambridge : Cambridge University Press . Myers , D . G . , 1999 . Social psychology , 6je . Boston , MA : McGraw - Hill Publishers . Nagai , Y . and Noguchi , H . , 2003 . An experimental study on the design thinking process started from difficult keywords : modeling the thinking process of creative design . Journal ofEngineering Design , 14 ( 4 ) , 429 - 437 . Nagai , Y . and Taura , T . , 2006 . Fortnal description of concept - synthesizing process for creative design . In : J . S . Gero , ed . Design computing and cognition ' 06 . Dordrecht , The Netherlands : Springer , 443 - 460 . Nisbett , R . and Wilson , T . , 1977 . Telling more than we can know : verbal reports on mental processes . Psychological Review , 84 ( 3 ) , 231 - 259 . Nuseibeh , B . and Easterbrook , S . , 2000 . Requirements engineering : a roadmap . In : A . CW . Finkelstein , ed . Thefuture of software engineering . Limerick , Ireland : IEEE Computer Society Press , 35 - 46 . 296 I . Chiu and L . H . Shu Pahl , G . and Beitz , w . , 1996 . Engineering design , a systematic approach . In : K Wallace , L . Blessing , and F . Bauert , trans . , K Wallace , eds . 2 / e . , London : Springer - Verlag London Ltd . Pinker , S . , 2007 . The stuffofthought , language as a window into human nature . New York , NY : Viking . Ratner , N . and Gleason , J . , 1993 . An introduction to psycholinguistics . In : J . B . Gleason and N . B . Ratner , eds . Psycholinguistics . Orlando , FL : Harcourt Brace College Publishers , 1 - - 40 . Rittel , H . W . J . and Webber , M . M . , 1984 , Planning problems are wicked problems . In : N . Cross ' , ed . Developments in design methodology . Bath : John Wiley & Sons , 135 - 144 . Saeed , J . I . , 2003 , Semantics . 2nd ed . Oxford : Blackwell Publishing . Segers , N . , 2004 . Computational representations of words & representations of words & associations in architectural design , development of a system support creative design . Thesis ( PhD ) . Bouwstenen 78 , Technische Universiteit Eindhoven . Shah , J . , Kulkarni , S . , and Vargas - Hernandez , N . , 2000 . Evaluation of idea generation methods for conceptual design : effectiveness metrics & design of experiments . Journal ofMechanical Design , 122 ( 4 ) , 377 - 384 . Shai , 0 . , Reich , Y , and Rubin , D . , 2009 . Creative conceptual design : extending the scope by infused design . Computer Aided Design , 41 ( 4 ) , 117 - 135 . Siegal , S . , 1956 . Non parametric statisticsfor the behavioral sciences . New York , NY : McGraw - Hill , Inc . Simon , H . , 1969 . The sciences ofthe artificial . Cambridge , MA : MIT Press . Stone , R . B . and Wood , KL . , 2000 . Development of a functional basis for design . Journal ofMechanical Design , 122 ( 4 ) , 359 - 369 . Tang , Y , and Zeng . , Y , 2009 . Quantifying designer ' s mental stress in the conceptual design process using kinesics study . Proceedings ofInternational Conference on Engineering Design , ICED ' 09 , 24 - 27 August 2009 , Stanford University , Stanford , CA . Thomas , J . C and Carroll , J . M . , 1984 . The psychological study of design . In : N . Cross ' s , ed . Developments in design methodology . Chichester : John Wiley & Sons , 221 - 235 . Torrance , E . P . , 1974 . Torrance tests ofcreative thinking . Bensenville , IL : Scholastic Testing Service , Inc . Tseng , I . , et ai . , 2008 . Overcoming blocks in conceptual design : the effects of open goals and analogical similarity on idea generation . Proceedings ofASME IDETCjCIE , Brooklyn , NY , August 3 - 6 , DEC2008 - 49276 . Ullman , D . , 2003 . The mechanical design process . 3rd ed . New York , NY : McGraw - Hill . Visser , W . , 2006 . The cognitive artifacts ofdesigning . Mahwah , NJ : Lawrence Erlbaum Associates , Publishers . Wickens , CD . and Hollands , J . G . , 2000 . Engineering psychology and human performance . 3rd ed . Upper Saddle River , NJ : Prentice Hall . Wilson , R . C , 1951 . An operational definition of originality . American Psychologist , 6 ( 6 ) , 297 . . Princeton University , 2008 . About WordNet . WordNet . , Princeton University . Available from : http : / / WordNet . Princeton . edu [ Accessed 7 July 2008 ] . Yang , M . C , 2009 . Observations on concept generation and sketching in engineering design . Research in Engineering Design , 20 ( 1 ) , I - II .",
    "srinivasanAnalogiLeadImprovingSelection2023": "AnalogiLead : Improving Selection of Analogical Inspirations with Chunking and Recombination Arvind Srinivasan \u2217 arvindrb @ umd . edu arvind @ cheenu . net University of Maryland College Park , Maryland , USA Joel Chan \u2217 joelchan @ umd . edu University of Maryland College Park , Maryland , USA Figure 1 : Our proposed System . The interface is split into three rows . In the first row , the users select magnets ( 2 . 1 ) from the Problem and their corresponding Analogy , mixing them or adding their own in the Playspace ( 2 . 2 ) to generate insightful questions / recombinations ( 2 . 3 ) to facilitate divergent thinking and as a result , aid in recognition of beneficial analogies . ABSTRACT Analogical reasoning , a process that integrates potential leads across domains and disciplines , has been proven to contribute to breakthrough innovations . Selecting the right analogical leads is crucial , as it determines the quality and effectiveness of the gen - erated ideas . However , identifying relevant analogical leads can Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page . Copyrights for third - party components of this work must be honored . For all other uses , contact the owner / author ( s ) . C & C \u201923 , June 19 \u2013 21 , 2023 , Virtual Event , USA \u00a9 2023 Copyright held by the owner / author ( s ) . ACM ISBN 979 - 8 - 4007 - 0180 - 1 / 23 / 06 . https : / / doi . org / 10 . 1145 / 3591196 . 3596817 be challenging and may be missed due to premature rejection or design fixation . To address this problem , our system , \" AnalogiLead \" , draws on the cognitive mechanisms of chunking and recombina - tion as a medium of interaction for selecting beneficial analogies . Users interact with meaningful chunks or segments from a design problem and analogy , represented as interactive tiles called \" mag - nets \" , and evaluate the analogies by recombining the \" magnets \" into brainstorming questions . These mechanisms are designed to foster playful and divergent exploration of analogical leads ( vs . restric - tive , relevance - based screening ) , to reduce premature rejection of analogical leads and foster more analogical innovations . 338 C & C \u201923 , June 19 \u2013 21 , 2023 , Virtual Event , USA Srinivasan CCS CONCEPTS \u2022 Human - centered computing \u2192 Graphical user interfaces ; Web - based interaction ; User interface design ; \u2022 Computing methodologies \u2192 Machine learning ; KEYWORDS Creativity Support , Machine Learning , User Interface Design , In - formation Seeking & Search ACM Reference Format : Arvind Srinivasan and Joel Chan . 2023 . AnalogiLead : Improving Selection of Analogical Inspirations with Chunking and Recombination . In Creativity and Cognition ( C & C \u201923 ) , June 19 \u2013 21 , 2023 , Virtual Event , USA . ACM , New York , NY , USA , 4 pages . https : / / doi . org / 10 . 1145 / 3591196 . 3596817 1 INTRODUCTION Analogical reasoning has been recognized as a powerful tool for generating breakthrough innovations . Analogical thinking involves drawing connections between seemingly disparate domains to gen - erate new ideas and approaches by identifying structural similarities [ 11 , 13 , 24 , 25 ] . Far analogies , in particular , have been instrumental in some of the most significant scientific and technological break - throughs throughout history [ 6 , 12 , 21 , 35 , 41 ] . However , selecting relevant analogical leads can be challenging due to the preference of memory retrieval for near , within - domain analogies that share object attributes [ 14 , 15 , 19 , 28 ] . Furthermore , analogical processing can be demanding on cognitive resources , often exhausting working memory , especially when multiple rela - tions need to be processed at once [ 18 ] . This can sometimes result in sub - optimal outcomes , as seen in the development of the first microwave oven , which was inspired by the analogy between radar and cooking but took several years to develop into a viable product , due to an improper understanding of users needs [ 10 ] . Existing approaches to analogical processing have attempted to address these challenges through instruction [ 16 , 30 , 37 ] or non - overgeneralized [ 39 ] abstract representations of problems and ana - logical examples [ 27 , 42 ] . Computational models and systems have also been proposed to facilitate this process through various tech - niques of abstraction and representation [ 5 , 9 , 11 , 27 , 33 ] . However , mere exposure to beneficial analogical leads alone is not enough to ensure their adoption in problem - solving [ 23 ] : factors such as expertise [ 2 , 36 ] , presence of usable anchors [ 3 ] , optimal represen - tations [ 7 , 8 , 11 , 17 ] , and diversity of solutions [ 27 ] can affect the premature rejection of potential leads , leading to design fixation [ 1 , 26 , 32 ] . Fully - automated analogical search engines such as SOLVENT [ 5 ] and ProbMap [ 34 ] have proposed breaking down a problem into functional components such as Stakeholders , Purpose and Mecha - nism to facilitate computational retrieval of new analogical leads . Given prior research on the creativity benefits of interacting with ideas in terms of their conceptual \" chunks \" [ 29 , 43 ] , we hypoth - esize that these conceptual \" chunks \" may also be beneficial for structuring interactions with existing analogical leads in a way that supports creative exploration and ( re ) interpretation ( and thus less premature rejection ) of analogical leads . In addition , aligning non - identical relational predicates has been suggested to be facilitated by re - representation [ 14 , 31 ] , which can assist in identifying and retrieving relevant analogies in complex systems . Earlier studies conducted in learning environments have explored questions as a possible mechanism to facilitate this process of re - representation and recombination ; attempting to facilitate the creation of new linkages between unlinked or previously linked components tar - geting specific applications [ 20 ] . According to Herring et al . [ 22 ] , designers were found to utilize examples by re - appropriating and recombining solution components to generate novel ideas . Addi - tionally , studies conducted in learning environments have found that questions play a vital role in promoting creative thinking . In particular , open - ended questions were found to greatly increase divergence in thinking [ 40 ] . A recent study [ 38 ] that proposed a system to semi - automatically generate external stimuli in the form of questions found that individuals , who were exposed to these generated questions produced better and more versatile ideas that those who were not . Our approach uses Large Language Models [ 4 ] to make this process easier and more accessible . Building on these insights , we designed a new system that lever - ages chunking and recombination mechanisms to support the recog - nition and adoption of far - domain analogies , facilitating their trans - fer across domains , and mitigating design fixation . By addressing these challenges , our proposed system aims to improve the ef - fectiveness of analogical reasoning in generating breakthrough innovations . 2 THE ANALOGILEAD SYSTEM The AnalogiLead System is designed to foster creative idea genera - tion and problem - solving through its three key sections : Magnets , Playspace , and Recombinations . 2 . 1 Magnets The Magnets in the AnalogiLead System are carefully curated and pre - defined sentence fragments or phrases that represent common functional constraints or attributes related to the problem domain . Expanding upon prior studies , we break down the problem into four components , namely : Stakeholder . This functional constraint highlights the benefi - ciaries who will be affected by the assigned design problem . Context . This functional constraint highlights the context in which the aforementioned stakeholders face the assigned design problem . Goal . This functional constraint highlights the goal that the stakeholders need to achieve to solve the assigned design problem . Obstacle . This functional constraint highlights an obstacle that hinders the stakeholders from achieving their goal for the assigned design problem . Alongside these functional constraints , the recommended analogi - cal leads add an additional constraint to highlight the solution : Solution . This functional constraint highlights the solution pro - posed to solve the goal of the recommended analogical lead for the given design problem . 339 Improving Selection of Analogies with Chunking and Recombination C & C \u201923 , June 19 \u2013 21 , 2023 , Virtual Event , USA These Magnets are designed to be flexible and versatile , allow - ing users to select and combine them in various ways to create prompts for idea generation . The Magnets serve as building blocks or \" chunks \" of ideas that can be easily manipulated and rearranged to construct meaningful questions via . the Playspace . 2 . 2 Playspace The Playspace is where users can experiment with the selected Mag - nets to create questions . The system uses Generative Pretrained Models ( GPT ) to automatically generate a wide range of questions by recombining the selected Magnets in real - time . The generated questions serve as prompts for the user to explore different angles and perspectives on the problem at hand , sparking creative thinking and encouraging divergent ideas . The user can iterate and experi - ment with different combinations of Magnets in the Playspace to generate a multitude of questions that prompt unique insights and solutions , promoting divergent thinking . 2 . 3 Recombinations The Recombinations section of the AnalogiLead System provides users with the option to further customize and refine the generated questions . Users can edit and modify the questions to better align with their thought process , specific problem domain , or desired outcome . This customization capability allows users to fine - tune the prompts to their specific needs , ensuring that the generated questions are relevant , meaningful , and tailored to their unique requirements . The Recombinations section empowers users with creative control , enabling them to craft prompts that are aligned with their creative goals and objectives . The combination of Magnets , Playspace , and Recombinations in the AnalogiLead System creates a dynamic and iterative process for idea generation and problem - solving , ultimately facilitating in selecting the most beneficial analogies for a given problem . The system leverages the power of Generative Pretrained Models to generate a wide range of prompts , while providing users with the flexibility to customize and refine the prompts to suit their creative process . This human - in - the - loop approach fosters exploration of di - verse ideas and encourages unique approaches to problem - solving , making the AnalogiLead System a powerful tool for stimulating creativity and improving selection of analogical inspirations while preventing possibilities for premature rejection and design fixation . 3 TECHNICAL OVERVIEW 3 . 1 Prompt Engineering Prompt : By combining the following list of words to - gether , generate [ n ] meaningful questions with insightfulrelationships : [ word1 , word2 , . . . , wordN ] Output : 1 . TheinterfaceusestheOpenAIlanguagemodel text - davinci - 002 for generating recombinations in a zero - shot learning context . Longer prompts including functional constraints resulted in con - fusing outputs , so the prompt structure was simplified . Descriptive words like \" insightful \" and \" meaningful \" were added to prioritize recombinations . The parameters for generating responses , such as temperature , frequency penalty , and presence penalty , were ad - justed to balance creativity and coherence . A temperature value of 0 . 75 was used for moderate randomness , and frequency penalty and presence penalty values of 0 . 5 and 0 . 25 , respectively , were used to encourage unique and diverse responses without overly strict penalties . 3 . 2 Software Implementation This interface is developed using the React Framework in JavaScript , which is a popular choice for building user interfaces . React provides a set of reusable primitives that can be used to create interactive web applications . React Hooks are used to simplify the code making it easier to manage and reuse code across components . To store data , the system uses Firestore Backend as a service , which is a cloud - based NoSQL database offered by Google . Firestore provides scalable storage and real - time synchronization for mobile , web , and server applications . 4 DEMONSTRATION For the purposes of this demonstration , a preset design problem will be presented , each with six corresponding handpicked analogies . Three analogies would be from a domain with varying degrees of closeness to the problem while the remaining would be from a far , yet structurally related domain . The demo will invite engagement from the participants by giving them the freedom to play around with the magnets of the problem and analogies , mix them or add their own magnets to the Playspace to generate recombinations . We also hope to expand the demo , to accommodate for users problems instead of preset - ones if possible , to promote greater engagement with the audience . Since the proposed system is a Web - interface and given the virtual nature of the event , there is no additional requirement needed for setup . ACKNOWLEDGMENTS This research was supported in part by ONR N000142012506 . REFERENCES [ 1 ] Marine Agogu\u00e9 , Akin Kazak\u00e7i , Armand Hatchuel , Pascal Le Masson , Benoit Weil , Nicolas Poirel , and Mathieu Cassotti . 2014 . The Impact of Type of Examples on Originality : Explaining Fixation and Stimulation Effects . The Journal of Creative Behavior 48 , 1 ( 2014 ) , 1 \u2013 12 . https : / / doi . org / 10 . 1002 / jocb . 37 [ 2 ] Olufunmilola Atilola , Megan Tomko , and Julie S . Linsey . 2016 . The effects of representation on idea generation and design fixation : A study comparing sketches and function trees . Design Studies 42 ( Jan 2016 ) , 110 \u2013 136 . https : / / doi . org / 10 . 1016 / j . destud . 2015 . 10 . 005 [ 3 ] David E . Brown and John Clement . 1989 . Overcoming misconceptions via ana - logical reasoning : abstract transfer versus explanatory model construction . In - structional Science 18 , 4 ( Dec 1989 ) , 237 \u2013 261 . https : / / doi . org / 10 . 1007 / BF00118013 [ 4 ] Tom B . Brown , Benjamin Mann , Nick Ryder , Melanie Subbiah , Jared Kaplan , Prafulla Dhariwal , Arvind Neelakantan , Pranav Shyam , Girish Sastry , Amanda Askell , Sandhini Agarwal , Ariel Herbert - Voss , Gretchen Krueger , Tom Henighan , Rewon Child , Aditya Ramesh , Daniel M . Ziegler , Jeffrey Wu , Clemens Winter , Christopher Hesse , Mark Chen , Eric Sigler , Mateusz Litwin , Scott Gray , Benjamin Chess , Jack Clark , Christopher Berner , Sam McCandlish , Alec Radford , Ilya Sutskever , and Dario Amodei . 2020 . Language Models are Few - Shot Learners . arXiv : 2005 . 14165 [ cs ] ( Jun 2020 ) . http : / / arxiv . org / abs / 2005 . 14165 00030 arXiv : 2005 . 14165 Citation Key : brownLanguageModelsAre2020 . [ 5 ] Joel Chan , Joseph Chee Chang , Tom Hope , Dafna Shahaf , and Aniket Kittur . 2018 . SOLVENT : A mixed initiative system for finding analogies between research papers . ( 2018 ) . https : / / doi . org / 10 / ggv7np 340 C & C \u201923 , June 19 \u2013 21 , 2023 , Virtual Event , USA Srinivasan [ 6 ] Robert S . Cohen and John J . Stachel . 1979 . Newton and the Law of Gravitation [ 1965d ] . In Selected Papers of L\u00e9on Rosenfeld , Robert S . Cohen , Marx W . Wartof - sky , Robert S . Cohen , and John J . Stachel ( Eds . ) . Vol . 21 . Springer Netherlands , Dordrecht , 58 \u2013 87 . https : / / doi . org / 10 . 1007 / 978 - 94 - 009 - 9349 - 5 _ 9 [ 7 ] Karl Duncker . 1945 . On problem - solving . Psychological Monographs 58 , 5 ( 1945 ) , i \u2013 113 . https : / / doi . org / 10 . 1037 / h0093599 Citation Key : dunckerProblemsolv - ing1945 . [ 8 ] Michael C Frank and Michael Ramscar . 2003 . How do Presentation and Context Influence Representation for Functional Fixedness Tasks ? . In Proceedings of the Annual Meeting of the Cognitive Science Society . Citation Key : frankHowPresen - tationContext2003 . [ 9 ] Katherine Fu , Joel Chan , Jonathan Cagan , Kenneth Kotovsky , Christian Schunn , and Kristin Wood . 2013 . The Meaning of \u201cNear\u201d and \u201cFar\u201d : The Impact of Structuring Design Databases and the Effect of Distance of Analogy on De - sign Output . Journal of Mechanical Design 135 , 2 ( Feb 2013 ) , 021007 . https : / / doi . org / 10 . 1115 / 1 . 4023158 [ 10 ] Giovanni Gavetti , Daniel A . Levinthal , and Jan W . Rivkin . 2005 . Strategy making innovelandcomplexworlds : thepowerofanalogy . StrategicManagementJournal 26 , 8 ( Aug 2005 ) , 691 \u2013 712 . https : / / doi . org / 10 . 1002 / smj . 475 [ 11 ] Dedre Gentner . 1983 . Structure - Mapping : A Theoretical Framework for Anal - ogy * . Cognitive Science 7 , 2 ( Apr 1983 ) , 155 \u2013 170 . https : / / doi . org / 10 . 1207 / s15516709cog0702 _ 3 [ 12 ] D . Gentner , S . Brem , R . W . Ferguson , P . Wolff , A . B . Markman , and K . D . For - bus . 1997 . Analogy and Creativity in the Works of Johannes Kepler . American Psychological Association , Washington D . C . , 403 \u2013 459 . Citation Key : gentner - AnalogyCreativityWorks1997 . [ 13 ] Dedre Gentner and Arthur B . Markman . 1997 . Structure mapping in analogy and similarity . American Psychologist 52 , 1 ( Jan 1997 ) , 45 \u2013 56 . https : / / doi . org / 10 . 1037 / 0003 - 066X . 52 . 1 . 45 [ 14 ] D . Gentner , M . J . Rattermann , and K . D . Forbus . 1993 . The Roles of Similarity in Transfer : Separating Retrievability From Inferential Soundness . Cognitive Psychology 25 , 4 ( Oct 1993 ) , 524 \u2013 575 . https : / / doi . org / 10 . 1006 / cogp . 1993 . 1013 [ 15 ] Mary L . Gick and Keith J . Holyoak . 1980 . Analogical problem solving . Cognitive Psychology 12 , 3 ( 1980 ) , 306 \u2013 355 . https : / / doi . org / 10 . 1016 / 0010 - 0285 ( 80 ) 90013 - 4 03379 . [ 16 ] Mary L . Gick and Keith J . Holyoak . 1983 . Schema induction and analogical transfer . CognitivePsychology 15 , 1 ( Jan1983 ) , 1 \u2013 38 . https : / / doi . org / 10 . 1016 / 0010 - 0285 ( 83 ) 90002 - 6 [ 17 ] Sam Glucksberg and Robert W . Weisberg . 1966 . Verbal behavior and problem solving : Some effects of labeling in a functional fixedness problem . Journal of Experimental Psychology 71 , 5 ( 1966 ) , 659 \u2013 664 . https : / / doi . org / 10 . 1037 / h0023118 Citation Key : glucksbergVerbalBehaviorProblem1966 . [ 18 ] Graeme S . Halford , Rosemary Baker , Julie E . McCredden , and John D . Bain . 2005 . How Many Variables Can Humans Process ? Psychological Science 16 , 1 ( Jan 2005 ) , 70 \u2013 76 . https : / / doi . org / 10 . 1111 / j . 0956 - 7976 . 2005 . 00782 . x [ 19 ] KristianJ . Hammond , ColleenM . Seifert , andKennethC . Gray . 1991 . Functionality inAnalogicalTransfer : AHardMatchIsGoodtoFind . TheJournaloftheLearning Sciences 1 , 2 ( 1991 ) , 111 \u2013 152 . https : / / www . jstor . org / stable / 1466674 [ 20 ] Andrew Hargadon and Robert I . Sutton . 1997 . Technology Brokering and Inno - vation in a Product Development Firm . Administrative Science Quarterly 42 , 4 ( Dec 1997 ) , 716 . https : / / doi . org / 10 . 2307 / 2393655 [ 21 ] J . W . Herivel . 1960 . Newton\u2019s Discovery of the Law of Centrifugal Force . Isis 51 , 4 ( Dec 1960 ) , 546 \u2013 553 . https : / / doi . org / 10 . 1086 / 349412 [ 22 ] Scarlett R . Herring , Chia - Chen Chang , Jesse Krantzler , and Brian P . Bailey . 2009 . Getting inspired ! : understanding how and why examples are used in creative design practice . In Proceedings of the 27th international conference on Human factors in computing systems . ACM , 87 \u2013 96 . https : / / doi . org / 10 . 1145 / 1518701 . 1518717 Citation Key : herringGettingInspiredUnderstanding2009 . [ 23 ] F . W . HesseandD . Klecha . 1990 . Useofanalogiesinproblemsolving . Computersin Human Behavior 6 , 1 ( Jan 1990 ) , 115 \u2013 129 . https : / / doi . org / 10 . 1016 / 0747 - 5632 ( 90 ) 90034 - E [ 24 ] Keith J . Holyoak and Kyunghee Koh . 1987 . Surface and structural similarity in analogical transfer . Memory & Cognition 15 , 4 ( Jul 1987 ) , 332 \u2013 340 . https : / / doi . org / 10 . 3758 / BF03197035 [ 25 ] Keith J . Holyoak and Paul Thagard . 1989 . Analogical Mapping by Constraint Satisfaction . Cognitive Science 13 , 3 ( Jul 1989 ) , 295 \u2013 355 . https : / / doi . org / 10 . 1207 / s15516709cog1303 _ 1 [ 26 ] David G . Jansson and Steven M . Smith . 1991 . Design fixation . Design Studies 12 , 1 ( 1991 ) , 3 \u2013 11 . https : / / doi . org / 10 . 1016 / 0142 - 694X ( 91 ) 90003 - F Citation Key : janssonDesignFixation1991 . [ 27 ] Hyeonsu B . Kang , Xin Qian , Tom Hope , Dafna Shahaf , Joel Chan , and Aniket Kittur . 2022 . Augmenting Scientific Creativity with an Analogical Search En - gine . ACM Transactions on Computer - Human Interaction ( Mar 2022 ) . https : / / doi . org / 10 . 1145 / 3530013 Just Accepted Citation Key : kangAugmentingScien - tificCreativity2022 . [ 28 ] Mark T . Keane . 1997 . What makes an analogy difficult ? The effects of order and causal structure on analogical mapping . Journal of Experimental Psychology : Learning , Memory , and Cognition 23 , 4 ( 1997 ) , 946 \u2013 967 . https : / / doi . org / 10 . 1037 / 0278 - 7393 . 23 . 4 . 946 [ 29 ] G . Knoblich , S . Ohlsson , H . Haider , and D . Rhenius . 1999 . Constraint relax - ation and chunk decomposition in insight problem solving . Journal of Exper - imental Psychology : Learning , Memory , and Cognition 25 , 6 ( 1999 ) , 1534 \u2013 1555 . https : / / doi . org / 10 . 1037 / 0278 - 7393 . 25 . 6 . 1534 00691 Citation Key : knoblichCon - straintRelaxationChunk1999 . [ 30 ] NicolasKokkalis , ThomasK\u00f6hn , JohannesHuebner , MoontaeLee , FlorianSchulze , and Scott R . Klemmer . 2013 . Taskgenies : Automatically providing action plans helps people complete tasks . ACM Transactions on Computer - Human Interaction ( TOCHI ) 20 , 5 ( 2013 ) , 27 . https : / / doi . org / 10 . 1145 / 2513560 00000 Citation Key : kokkalisTaskgeniesAutomaticallyProviding2013 . [ 31 ] K . J . Kurtz and J . Loewenstein . 2007 . Converging on a new role for analogy in problem solving and retrieval : When two problems are better than one . Memory & Cognition 35 , 2 ( 2007 ) , 334 \u2013 341 . https : / / doi . org / 10 . 3758 / BF03193454 Citation Key : kurtzConvergingNewRole2007 . [ 32 ] Keelin Leahy , Shanna R . Daly , Seda McKilligan , and Colleen M . Seifert . 2020 . Design Fixation From Initial Examples : Provided Versus Self - Generated Ideas . Journal of Mechanical Design 142 , 101402 ( Apr 2020 ) . https : / / doi . org / 10 . 1115 / 1 . 4046446 00000 Citation Key : leahyDesignFixationInitial2020 . [ 33 ] J . S . Linsey , A . B . Markman , and K . L . Wood . 2012 . Design by Analogy : A Study of the WordTree Method for Problem Re - Representation . Journal of Mechanical Design 134 , 4 ( 2012 ) , 041009 . https : / / doi . org / 10 . 1115 / 1 . 4006145 Citation Key : linseyDesignAnalogyStudy2012 . [ 34 ] Stephen MacNeil , Zijian Ding , Kexin Quan , Ziheng Huang , Kenneth Chen , and Steven P . Dow . 2021 . ProbMap : Automatically Constructing Design Galleries throughFeatureExtractionandSemanticClustering . In AdjunctProceedingsofthe 34th Annual ACM Symposium on User Interface Software and Technology ( Virtual Event , USA ) ( UIST\u201921Adjunct ) . AssociationforComputingMachinery , NewYork , NY , USA , 134 \u2013 136 . https : / / doi . org / 10 . 1145 / 3474349 . 3480203 [ 35 ] George Mestral . 1961 . Separable fastening device . https : / / patents . google . com / patent / US3009235A / en [ 36 ] Laura R . Novick . 1988 . Analogical transfer , problem similarity , and expertise . Journal of Experimental Psychology : Learning , Memory , and Cognition 14 , 3 ( 1988 ) , 510 \u2013 520 . https : / / doi . org / 10 . 1037 / 0278 - 7393 . 14 . 3 . 510 [ 37 ] Lindsey E . Richland and Ian M . McDonough . 2010 . Learning by analogy : Dis - criminating between potential analogs . Contemporary Educational Psychology 35 , 1 ( Jan 2010 ) , 28 \u2013 43 . https : / / doi . org / 10 . 1016 / j . cedpsych . 2009 . 09 . 001 [ 38 ] Dominik Siemon , Taras Rarog , and Susanne Robra - Bissantz . 2016 . Semi - Automated Questions as a Cognitive Stimulus in Idea Generation . In 2016 49th Hawaii International Conference on System Sciences ( HICSS ) . IEEE . https : / / doi . org / 10 . 1109 / hicss . 2016 . 39 [ 39 ] Rand J . Spiro , Richard L . Coulson , P . J . Feltovich , and Daniel K . Anderson . 1988 . Cognitive flexibility theory advanced knowledge acquisition in ill - structured do - mains / . Number 441 in Reading Research and Education Center Report . Cham - paign , IL . 20 pages . 01887 Citation Key : spiroCognitiveFlexibilityTheory1988 . [ 40 ] ToyinTofade , JamieElsner , andStuartT . Haines . 2013 . BestPracticeStrategiesfor EffectiveUseofQuestionsasaTeachingTool . AmericanJournalofPharmaceutical Education 77 , 7 ( Sep 2013 ) , 155 . https : / / doi . org / 10 . 5688 / ajpe777155 [ 41 ] Terry Winograd and Fernando Flores . 2008 . Understanding computers and cogni - tion : a new foundation for design ( 24th printing ed . ) . Addison - Wesley , Boston . [ 42 ] Lixiu Yu , Aniket Kittur , and Robert E . Kraut . 2014 . Searching for Analogical Ideas withCrowds . In Proceedingsofthe32NdAnnualACMConferenceonHumanFactors in Computing Systems ( CHI \u201914 ) . ACM , New York , NY , USA , 1225 \u2013 1234 . https : / / doi . org / 10 . 1145 / 2556288 . 2557378 Citation Key : yuSearchingAnalogicalIdeas2014 . [ 43 ] Pengyi Zhang and Dagobert Soergel . 2020 . Cognitive mechanisms in sensemak - ing : A qualitative user study . Journal of the Association for Information Science and Technology 71 , 2 ( 2020 ) , 158 \u2013 171 . https : / / doi . org / 10 . 1002 / asi . 24221 Citation Key : zhangCognitiveMechanismsSensemaking2020 . Received 17 April 2023 341",
    "schwartzSymmetricPatternsCoordinations2016": "Proceedings of NAACL - HLT 2016 , pages 499 \u2013 505 , San Diego , California , June 12 - 17 , 2016 . c (cid:13) 2016 Association for Computational Linguistics Symmetric Patterns and Coordinations : Fast and Enhanced Representations of Verbs and Adjectives Roy Schwartz , 1 Roi Reichart , 2 1 Institute of Computer Science , The Hebrew University 2 Faculty of Industrial Engineering and Management , Technion , IIT { roys02 | arir } @ cs . huji . ac . il roiri @ ie . technion . ac . il Ari Rappoport 1 Abstract State - of - the - art word embeddings , which are often trained on bag - of - words ( BOW ) con - texts , provide a high quality representation of aspects of the semantics of nouns . However , their quality decreases substantially for the task of verb similarity prediction . In this paper we show that using symmetric pattern contexts ( SPs , e . g . , \u201cX and Y\u201d ) improves word2vec verb similarity performance by up to 15 % and is also instrumental in adjective similarity pre - diction . The unsupervised SP contexts are even superior to a variety of dependency con - texts extracted using a supervised dependency parser . Moreover , we observe that SPs and dependency coordination contexts ( Coor ) cap - ture a similar type of information , and demon - strate that Coor contexts are superior to other dependency contexts including the set of all dependency contexts , although they are still inferior to SPs . Finally , there are substantially fewer SP contexts compared to alternative rep - resentations , leading to a massive reduction in training time . On an 8G words corpus and a 32 core machine , the SP model trains in 11 min - utes , compared to 5 and 11 hours with BOW and all dependency contexts , respectively . 1 Introduction In recent years , vector space models ( VSMs ) have become prominent in NLP . VSMs are often eval - uated by measuring their ability to predict human judgments of lexical semantic relations between pairs of words , mostly association or similarity . While many datasets for these tasks are limited to pairs of nouns , the recent SimLex999 word similar - ity dataset ( Hill et al . , 2014 ) also consists of sim - ilarity scores for verb and adjective pairs . State - of - the - art VSMs such as word2vec skip - gram ( w2v - SG , ( Mikolov et al . , 2013a ) ) and GloVe ( Pennington et al . , 2014 ) excel at noun - related tasks . However , their performance substantially decreases on verb similar - ity prediction in SimLex999 , and their adjective rep - resentations have rarely been evaluated ( Section 2 ) . In this paper we show that a key factor in the re - duced performance of the w2v - SG model on verb representation is its reliance on bag - of - words ( BOW ) contexts : contexts of the represented words that con - sist of words in their physical proximity . We investi - gate a number of alternative contexts for this model , including various dependency contexts , and show that simple , automatically acquired symmetric pat - terns ( SPs , e . g . , \u201cX or Y\u201d , ( Hearst , 1992 ; Davidov and Rappoport , 2006 ) ) are the most useful contexts for the representation of verbs and also adjectives . Moreover , the SP - based model is much more com - pact than the alternatives , making its training an or - der of magnitude faster . In particular , we train several versions of the w2v - SG model , each with a different context type , and evaluate the resulting word embeddings on the task of predicting the similarity scores of the verb and adjective portions of SimLex999 . Our results show that SP contexts ( SG - SP ) obtain the best results on both tasks : Spearman\u2019s \u03c1 scores of 0 . 459 on verbs and 0 . 651 on adjectives . These results are 15 . 2 % and 4 . 7 % better than BOW contexts and 7 . 3 % and 6 . 5 % better than all dependency contexts ( DepAll ) . Moreover , the number of SP contexts is substantially 499 smaller than the alternatives , making it extremely fast to train : 11 minutes only on an 8G word cor - pus using a 32 CPU core machine , compared to 5 and 11 hours for BOW and DepAll , respectively . Recently , Schwartz et al . ( 2015 ) presented a count - based VSM that utilizes SP contexts ( SRR15 ) . This model excels on verb similarity , outperform - ing VSMs that use other contexts ( e . g . , BOW and DepAll ) by more than 20 % . In this paper we show that apart from its SP contexts , the success of SRR15 is attributed in large to its explicit representation of antonyms ( live / die ) ; turning this feature off reduces its performance to be on par with SG - SP . As op - posed to Schwartz et al . ( 2015 ) , we keep our VSM \ufb01xed across experiments ( w2v - SG ) , changing only the context type . This allows us to attribute our im - proved results to one factor : SP contexts . We further observe that SP contexts are tightly connected to syntactic coordination contexts ( Coor , Section 3 ) . Following this observation , we compare the w2v - SG model with three dependency - based context types : ( a ) Coor contexts ; ( b ) all dependency links ( DepAll ) ; and ( c ) all dependency links exclud - ing Coor links ( Coor C ) . 1 Our results show that training with Coor contexts is superior to training with the other context types , leading to improved similarity prediction of 2 . 7 - 4 . 1 % and 4 . 3 - 6 . 9 % on verbs and adjectives respectively . These results demonstrate the prominence of Coor contexts in verb and adjective representation : these contexts are even better than their combination with the rest of the dependency - based contexts ( the DepAll contexts ) . Nonetheless , although Coor con - texts are extracted using a supervised dependency parser , they are still inferior to SP contexts , extracted automatically from plain text ( Section 3 ) , by 4 . 6 % and 2 . 2 % for verb and adjective pairs . 2 Background Word Embeddings for Verbs and Adjectives . A number of evaluation sets consisting of word pairs scored by humans for semantic relations ( mostly as - sociation and similarity ) are in use for VSM evalua - tion . These include : RG - 65 ( Rubenstein and Good - enough , 1965 ) , MC - 30 ( Miller and Charles , 1991 ) , WordSim353 ( Finkelstein et al . , 2001 ) , MEN ( Bruni 1 Coor \u222a Coor C = DepAll , Coor \u2229 Coor C = \u2205 et al . , 2014 ) and SimLex999 ( Hill et al . , 2014 ) . 2 Nouns are dominant in almost all of these datasets . For example , RG - 65 , MC - 30 and Word - Sim353 consist of noun pairs almost exclusively . A few datasets contain pairs of verbs ( Yang and Pow - ers , 2006 ; Baker et al . , 2014 ) . The MEN dataset , al - though dominated by nouns , also contains verbs and adjectives . Nonetheless , the human judgment scores in these datasets re\ufb02ect relatedness between words . In contrast , the recent SimLex999 dataset ( Hill et al . , 2014 ) contains word similarity scores for nouns ( 666 pairs ) , verbs ( 222 pairs ) and adjectives ( 111 pairs ) . We use this dataset to study the effect of context type on VSM performance in a verb and adjective simi - larity prediction task . Context Type in Word Embeddings . Most VSMs ( e . g . , ( Collobert et al . , 2011 ; Mikolov et al . , 2013b ; Pennington et al . , 2014 ) ) de\ufb01ne the context of a target word to be the words in its physical prox - imity ( bag - of - words contexts ) . Dependency con - texts , consisting of the words connected to the tar - get word by dependency links ( Grefenstette , 1994 ; Pad\u00f3 and Lapata , 2007 ; Levy and Goldberg , 2014 ) , are another well researched alternative . These works did not recognize the importance of syntactic coor - dination contexts ( Coor ) . Patterns have also been suggested as VSM con - texts , but mostly for representing pairs of words ( Turney , 2006 ; Turney , 2008 ) . While this approach has been successful for extracting various types of word relations , using patterns to represent single words is useful for downstream applications . Re - cently , Schwartz et al . ( 2015 ) explored the value of symmetric pattern contexts for word representation , an idea this paper develops further . A recently published approach ( Melamud et al . , 2016 ) also explored the effect of the type of con - text on the performance of word embedding models . Nonetheless , while they also explored bag - of - words and dependency contexts , they did not experiment with SPs or coordination contexts , which we \ufb01nd to be most useful for predicting word similarity . Limitations of Word Embeddings . Recently , a few papers examined the limitations of word em - bedding models in representing different types of se - 2 For a comprehensive list see : wordvectors . org / 500 mantic information . Levy et al . ( 2015 ) showed that word embeddings do not capture semantic relations such as hyponymy and entailment . Rubinstein et al . ( 2015 ) showed that while state - of - the - art embed - dings are successful at capturing taxonomic infor - mation ( e . g . , cow is an animal ) , they are much less successful in capturing attributive properties ( ba - nanas are yellow ) . In ( Schwartz et al . , 2015 ) , we showed that word embeddings are unable to distin - guish between pairs of words with opposite mean - ings ( antonyms , e . g . , good / bad ) . In this paper we study the dif\ufb01culties of bag - of - words based word embeddings in representing verb similarity . 3 Symmetric Patterns ( SPs ) Lexico - syntactic patterns are templates of text that contain both words and wildcards ( Hearst , 1992 ) , e . g . , \u201cX and Y\u201d and \u201cX for a Y\u201d . Pattern instances are sequences of words that match a given pattern , such that concrete words replace each of the wild - cards . For example , \u201c John and Mary \u201d is an instance of the pattern \u201cX and Y\u201d . Patterns have been shown useful for a range of tasks , including word relation extraction ( Lin et al . , 2003 ; Davidov et al . , 2007 ) , knowledge extraction ( Etzioni et al . , 2005 ) , senti - ment analysis ( Davidov et al . , 2010 ) and authorship attribution ( Schwartz et al . , 2013 ) . Symmetric patterns ( SPs ) are lexico - syntactic pat - terns that comply to two constraints : ( a ) Each pat - tern has exactly two wildcards ( e . g . , X or Y ) ; and ( b ) When two words ( X , Y ) co - occur in an SP , they are also likely to co - occur in this pattern in oppo - site positions , given a large enough corpus ( e . g . , \u201c X or Y \u201d and \u201c Y or X \u201d ) . For example , the pattern \u201c X and Y \u201d is symmetric as for a large number of word pairs ( e . g . , ( eat , drink ) ) both members are likely to occur in both of its wildcard positions ( e . g . , \u201ceat and drink\u201d , \u201cdrink and eat\u201d ) . SPs have shown useful for tasks such as word clustering ( Widdows and Dorow , 2002 ; Davidov and Rappoport , 2006 ) , semantic class learning ( Kozareva et al . , 2008 ) and word classi\ufb01cation ( Schwartz et al . , 2014 ) . In this paper we demonstrate the value of SP - based contexts in vector representa - tions of verbs and adjectives . The rationale behind this context type is that two words that co - occur in an SP tend to take the same semantic role in the sen - tence , and are thus likely to be similar in meaning ( e . g . , \u201c ( John and Mary ) sang\u201d ) . SP Extraction . Many works that applied SPs in NLP tasks employed a hand - crafted list of patterns ( Widdows and Dorow , 2002 ; Dorow et al . , 2005 ; Feng et al . , 2013 ) . Following Schwartz et al . ( 2015 ) we employ the DR 06 algorithm ( Davidov and Rap - poport , 2006 ) , an unsupervised algorithm that ex - tracts SPs from plain text . We apply this algorithm to our corpus ( Section 4 ) and extract 11 SPs : \u201cX and Y\u201d , \u201cX or Y\u201d , \u201cX and the Y\u201d , \u201cX or the Y\u201d , \u201cX or a Y\u201d , \u201cX nor Y\u201d , \u201cX and one Y\u201d , \u201c either X or Y\u201d , \u201cX rather than Y\u201d , \u201cX as well as Y\u201d , \u201c from X to Y\u201d . A description of the DR 06 algorithm is beyond the scope of this paper ; the interested reader is referred to ( Davidov and Rappoport , 2006 ) . SP Contexts . We generate SP contexts by taking the co - occurrence counts of pairs of words in SPs . For example , in the SP token \u201c boys and girls \u201d , the term girls is taken as an SP context of the word boys , and boys is taken as an SP context of girls . We do not make a distinction between the differ - ent SPs . E . g . , \u201c boys and girls \u201d and \u201c boys or girls \u201d are treated the same . However , we distinguish be - tween left and right contexts . For example , we gen - erate different contexts for the word girls , one for left - hand contexts ( \u201c girls and boys \u201d ) and another for right - hand contexts ( \u201c boys and girls \u201d ) . SPs and Coordinations . SPs and syntactic coor - dinations ( Coors ) are intimately related . For exam - ple , of the 11 SPs extracted in this paper by the DR 06 algorithm ( listed above ) , the \ufb01rst eight represent co - ordination structures . Moreover , these SPs account for more than 98 % of the SP instances in our corpus . Indeed , due to the signi\ufb01cant overlap between SPs and Coors , the former have been proposed as a sim - ple model of the latter ( Nakov and Hearst , 2005 ) . 3 Despite their tight connection , SPs sometimes fail to properly identify the components of Coors . For example , while SPs are instrumental in captur - ing shallow Coors , they fail in capturing coordina - tion between phrases . Consider the sentence John 3 Note though that the exact syntactic annotation of coordi - nation is debatable both in the linguistic community ( Tesni\u00e8re , 1959 ; Hudson , 1980 ; Mel\u2019\u02c7cuk , 1988 ) and also in the NLP com - munity ( Nilsson et al . , 2006 ; Schwartz et al . , 2011 ; Schwartz et al . , 2012 ) . 501 walked and Mary ran : the SP \u201cX and Y\u201d captures the phrase walked and Mary , while the Coor links the heads of the connected phrases ( \u201c walked \u201d and \u201c ran \u201d ) . SPs , on the other hand , can go beyond Coors and capture other types of symmetric structures like \u201c from X to Y\u201d and \u201cX rather than Y\u201d . Our experiments reveal that both SPs and Coors are highly useful contexts for verb and adjective rep - resentation , at least with respect to word similarity . Interestingly , Coor contexts , extracted using a super - vised dependency parser , are less effective than SP contexts , which are extracted from plain text . 4 Experiments Model . We keep the VSM \ufb01xed throughout our experiments , changing only the context type . This methodology allows us to evaluate the impact of dif - ferent contexts on the VSM performance , as context choice is the only modeling decision that changes across experimental conditions . Our VSM is the word2vec skip - gram model ( w2v - SG , Mikolov et al . ( 2013a ) ) , which obtains state - of - the - art results on a variety of NLP tasks ( Baroni et al . , 2014 ) . We employ the word2vec toolkit . 4 For all context types other than BOW we use the word2vec package of ( Levy and Goldberg , 2014 ) , 5 which aug - ments the standard word2vec toolkit with code that allows arbitrary context de\ufb01nition . Experimental Setup . We experiment with the verb pair ( 222 pairs ) and adjective pair ( 111 pairs ) portions of SimLex999 ( Hill et al . , 2014 ) . We re - port the Spearman \u03c1 correlation between the ranks derived from the scores of the evaluated models and the human scores provided in SimLex999 . 6 We train the w2v - SG model with \ufb01ve different context types : ( a ) BOW contexts ( SG - BOW ) ; ( b ) all dependency links ( SG - DepAll ) ( c ) dependency - based coordination contexts ( i . e . , those labeled with conj , SG - Coor ) ; ( d ) all dependency links except for coordinations ( SG - Coor C ) ; and ( e ) SP contexts . Our training corpus is the 8G words corpus gener - 4 https : / / code . google . com / p / word2vec / 5 https : / / bitbucket . org / yoavgo / word2vecf 6 Model scores are computed in the standard way : applying the cosine similarity metric to the vectors learned for the words participating in the pair . Model Verb Adj . Noun Time # Cont . SG - BOW 0 . 307 0 . 604 0 . 501 320 13G SG - DepAll 0 . 386 0 . 586 0 . 499 551 14 . 5G SG - Coor 0 . 413 0 . 629 0 . 428 23 550M SG - Coor C 0 . 372 0 . 56 0 . 494 677 14G SG - SP 0 . 459 0 . 651 0 . 415 11 270M SRR15 0 . 578 0 . 663 0 . 497 \u2014 270M SRR15 \u2212 0 . 441 0 . 68 0 . 421 \u2014 270M Table 1 : Spearman\u2019s \u03c1 scores on the different portions of SimLex999 . The top part presents results for the word2vec skip - gram model ( w2v - SG ) with various context types ( see text ) . The bottom lines present the results of the count SP - based model of Schwartz et al . ( 2015 ) , with ( SRR15 ) and without ( SRR15 \u2212 ) its antonym detection method . The two rightmost columns present the run time of the w2v - SG mod - els in minutes ( Time ) and the number of context in - stances used by the model ( # Cont . ) . 10 For each Sim - Lex999 portion , the score of the best w2v - SG model across context types is highlighted in bold font . ated by the word2vec script . 7 Models ( b ) - ( d ) require the dependency parse trees of the corpus as input . To generate these trees , we employ the Stanford POS Tagger ( Toutanova et al . , 2003 ) 8 and the stack version of the MALT parser ( Nivre et al . , 2009 ) . 9 The SP contexts are generated using the SPs extracted by the DR 06 algorithm from our training corpus ( see Section 3 ) . For BOW contexts , we experiment with three win - dow sizes ( 2 , 5 and 10 ) and report the best results ( window size of 2 across conditions ) . For depen - dency based contexts we follow the standard con - vention in the literature : we consider the immedi - ate heads and modi\ufb01ers of the represented word . All models are trained with 500 dimensions , the de - fault value of the word2vec script . Other hyper - parameters were also set to the default values of the code packages . Results . Table 1 presents our results . The SG - SP model provides the most useful verb and adjective representations among the w2v - SG models . Com - pared to BOW ( SG - BOW ) , the most commonly used 7 code . google . com / p / word2vec / source / browse / trunk / demo - train - big - model - v1 . sh 8 nlp . stanford . edu / software / tagger . shtml 9 http : / / www . maltparser . org / index . html 502 context type , SG - SP results are 15 . 2 % and 4 . 7 % higher on verbs and adjectives respectively . Com - pared to dependency links ( SG - DepAll ) , the im - provements are 7 . 3 % and 6 . 5 % . For completeness , we compare the models on the noun pairs portion , observing that SG - BOW and SG - DepAll are \u223c 8 . 5 % better than SG - SP . This indicates that different word classes require different representations . The results for SG - Coor , which is trained with syntactic coordination ( Coor ) contexts , show that these contexts are superior to all the other depen - dency links ( SG - Coor C ) by 4 . 1 % and 6 . 9 % on verbs and adjectives . Importantly , comparing the SG - Coor model to the SG - DepAll model , which aug - ments the Coor contexts with the other syntactic de - pendency contexts , reveals that SG - DepAll is ac - tually inferior by 2 . 7 % and 4 . 3 % in Spearman \u03c1 on verbs and adjectives respectively . Interestingly , Coor contexts , which are extracted using a super - vised parser , are still inferior by 4 . 6 % and 2 . 2 % to SPs , which capture similar contexts but are extracted from plain text . Table 1 also shows the training times of the vari - ous w2v - SG models on a 32G memory , 32 CPU core machine . SG - SP and SG - Coor , which take 11 min - utes and 23 minutes respectively to train , are sub - stantially faster than the other w2v - SG models . For example , they are more than an order of magnitude faster than SG - BOW ( 320 minutes ) and SG - Coor C ( 677 minutes ) . This is not surprising , as there are far fewer SP contexts ( 270M ) and Coor contexts ( 550M ) than BOW contexts ( 13G ) and Coor C con - texts ( 14G ) ( # Cont . column ) . Finally , the performance of the SG - SP model is still substantially inferior to the SRR15 SP - based model ( Schwartz et al . , 2015 ) . As both models use the same SP contexts , this result indicates that other modeling decisions in SRR15 lead to its superior performance . We show that this difference is mostly attributed to one feature of SRR15 : its method for detecting antonym pairs ( good / bad ) . Indeed , the SRR15 model without its antonym detection method ( SRR15 \u2212 ) obtains a Spearman \u03c1 of 0 . 441 , compared to 0 . 459 of SG - SP on verb pairs . For adjectives , however , SRR15 \u2212 is 1 . 7 % better than SRR15 , in - 10 We compare the w2v - SG models training time only . SRR15 and SRR15 \u2212 are count - based models and have no training step . creasing the difference from SG - SP to 2 . 9 % . 11 5 Conclusions We demonstrated the effectiveness of symmetric pattern contexts in word embedding induction . Ex - periments with the word2vec model showed that these contexts are superior to various alternatives for verb and adjective representation . We further pointed at the connection between symmetric pat - terns and syntactic coordinations . We showed that coordinations are superior to other syntactic con - texts , but are still inferior to symmetric patterns , al - though the extraction of symmetric patterns requires less supervision . Future work includes developing a model that successfully combines the various context types ex - plored in this paper . We are also interested in the representation of other word classes such as adverbs for which no evaluation set currently ex - ists . Finally , the code for generating the SG - SP embeddings , as well as the vectors experimented with in this paper , are released and can be down - loaded from http : / / www . cs . huji . ac . il / \u223c roys02 / papers / sp _ sg / sp _ sg . html Acknowledgments This research was funded ( in part ) by the Intel Col - laborative Research Institute for Computational In - telligence ( ICRI - CI ) , the Israel Ministry of Science and Technology Center of Knowledge in Machine Learning and Arti\ufb01cial Intelligence ( Grant number 3 - 9243 ) . The second author was partially funded by the Microsoft / Technion research center for elec - tronic commerce and the Google faculty research award . References Simon Baker , Roi Reichart , and Anna Korhonen . 2014 . An unsupervised model for instance level subcatego - rization acquisition . In Proc . of EMNLP . Marco Baroni , Georgiana Dinu , and Germ\u00e1n Kruszewski . 2014 . Don\u2019t count , predict ! a systematic compari - son of context - counting vs . context - predicting seman - tic vectors . In Proc . of ACL . 11 We report results for our reimplementation of SRR15 and SRR15 \u2212 . 503 Elia Bruni , Nam - Khanh Tran , and Marco Baroni . 2014 . Multimodal distributional semantics . JAIR . Ronan Collobert , Jason Weston , L\u00e9on Bottou , Michael Karlen , Koray Kavukcuoglu , and Pavel Kuksa . 2011 . Natural language processing ( almost ) from scratch . JMLR , 12 : 2493 \u2013 2537 . Dmitry Davidov and Ari Rappoport . 2006 . Ef\ufb01cient un - supervised discovery of word categories using sym - metric patterns and high frequency words . In Proc . of ACL - COLING . Dmitry Davidov , Ari Rappoport , and Moshe Koppel . 2007 . Fully unsupervised discovery of concept - speci\ufb01c relationships by web mining . In Proc . of ACL . Dmitry Davidov , Oren Tsur , and Ari Rappoport . 2010 . Enhanced sentiment learning using twitter hashtags and smileys . In Proc . of COLING . Beate Dorow , Dominic Widdows , Katarina Ling , Jean - Pierre Eckmann , Danilo Sergi , and Elisha Moses . 2005 . Using Curvature and Markov Clustering in Graphs for Lexical Acquisition and Word Sense Dis - crimination . Oren Etzioni , Michael Cafarella , Doug Downey , Ana - Maria Popescu , Tal Shaked , Stephen Soderland , Daniel S Weld , and Alexander Yates . 2005 . Unsuper - vised named - entity extraction from the web : An exper - imental study . Arti\ufb01cial intelligence , 165 ( 1 ) : 91 \u2013 134 . Song Feng , Jun Seok Kang , Polina Kuznetsova , and Yejin Choi . 2013 . Connotation lexicon : A dash of sentiment beneath the surface meaning . In Proc . of ACL . Lev Finkelstein , Evgeniy Gabrilovich , Yossi Matias , Ehud Rivlin , Zach Solan , Gadi Wolfman , and Eytan Ruppin . 2001 . Placing search in context : The concept revisited . In Proc . of WWW . Gregory Grefenstette . 1994 . Explorations in automatic thesaurus discovery . Boston : Kluwer . Marti A . Hearst . 1992 . Automatic acquisition of hy - ponyms from large text corpora . In Proc . of COLING \u2013 Volume 2 . Felix Hill , Roi Reichart , and Anna Korhonen . 2014 . Simlex - 999 : Evaluating semantic models with ( gen - uine ) similarity estimation . arXiv : 1408 . 3456 [ cs . CL ] . Richard A . Hudson . 1980 . Arguments for a Non - transformational Grammar . Chicago : University of Chicago Press . Zornitsa Kozareva , Ellen Riloff , and Eduard Hovy . 2008 . Semantic class learning from the web with hyponym pattern linkage graphs . In Proc . of ACL - HLT . Omer Levy and Yoav Goldberg . 2014 . Dependency - based word embeddings . In Proc . of ACL . Omer Levy , Steffen Remus , Chris Biemann , and Ido Da - gan . 2015 . Do supervised distributional methods re - ally learn lexical inference relations ? In Proc . of NAACL . Dekang Lin , Shaojun Zhao , Lijuan Qin , and Ming Zhou . 2003 . Identifying synonyms among distributionally similar words . In Proc . of IJCAI . Oren Melamud , David McClosky , Siddharth Patwardhan , and Mohit Bansal . 2016 . The role of context types and dimensionality in learning word embeddings . In Proc . of NAACL . Igor Mel\u2019\u02c7cuk . 1988 . Dependency Syntax : Theory and Practice . State University of New York Press . Tomas Mikolov , Ilya Sutskever , Kai Chen , Greg S Cor - rado , and Jeff Dean . 2013a . Distributed representa - tions of words and phrases and their compositionality . In Proc . of NIPS . Tomas Mikolov , Wen - tau Yih , and Geoffrey Zweig . 2013b . Linguistic regularities in continuous space word representations . In Proc . of NAACL - HLT . George A Miller and Walter G Charles . 1991 . Contex - tual correlates of semantic similarity . Language and cognitive processes . Preslav Nakov and Marti Hearst . 2005 . Using the web as an implicit training set : application to structural ambi - guity resolution . In Proc . of HLT - EMNLP . Jens Nilsson , Joakim Nivre , and Johan Hall . 2006 . Graph transformations in data - driven dependency parsing . In Proc . of ACL - COLING . Joakim Nivre , Marco Kuhlmann , and Johan Hall . 2009 . An improved oracle for dependency parsing with on - line reordering . In Proc . of IWPT . Sebastian Pad\u00f3 and Mirella Lapata . 2007 . Dependency - based construction of semantic space models . Compu - tational Linguistics . Jeffrey Pennington , Richard Socher , and Christopher D . Manning . 2014 . Glove : Global vectors for word rep - resentation . In Proc . of EMNLP . Herbert Rubenstein and John B . Goodenough . 1965 . Contextual correlates of synonymy . Communications of the ACM . Dana Rubinstein , Ef\ufb01 Levi , Roy Schwartz , and Ari Rap - poport . 2015 . How well do distributional models cap - ture different types of semantic knowledge ? In Proc . of ACL . Roy Schwartz , Omri Abend , Roi Reichart , and Ari Rap - poport . 2011 . Neutralizing linguistically problematic annotations in unsupervised dependency parsing eval - uation . In Proc . of ACL - HLT . Roy Schwartz , Omri Abend , and Ari Rappoport . 2012 . Learnability - based syntactic annotation design . In Proc . of COLING . Roy Schwartz , Oren Tsur , Ari Rappoport , and Moshe Koppel . 2013 . Authorship attribution of micro - messages . In Proc . of EMNLP . 504 Roy Schwartz , Roi Reichart , and Ari Rappoport . 2014 . Minimally supervised classi\ufb01cation to semantic cat - egories using automatically acquired symmetric pat - terns . In Proc . of COLING . Roy Schwartz , Roi Reichart , and Ari Rappoport . 2015 . Symmetric pattern based word embeddings for im - proved word similarity prediction . In Proc . of CoNLL . Lucien Tesni\u00e8re . 1959 . \u00c9l\u00e9ments de syntaxe structurale . Paris : K1incksieck . Kristina Toutanova , Dan Klein , Christopher D Manning , and Yoram Singer . 2003 . Feature - rich part - of - speech tagging with a cyclic dependency network . In Proc . of NAACL . Peter D . Turney . 2006 . Similarity of semantic relations . Computational Linguistics . Peter D . Turney . 2008 . A uniform approach to analo - gies , synonyms , antonyms , and associations . In Proc . of COLING . Dominic Widdows and Beate Dorow . 2002 . A graph model for unsupervised lexical acquisition . In Proc . of COLING . Dongqiang Yang and David M . W . Powers . 2006 . Verb similarity on the taxonomy of wordnet . In Proc . of GWC . 505",
    "chanImportanceIterationCreative2015": "The importance of iteration in creative conceptual combination Joel Chan a , \u21d1 , Christian D . Schunn b a LRDC Rm 823 , University of Pittsburgh , 3939 O\u2019Hara St , Pittsburgh , PA 15260 , USA b LRDC Rm 821 , University of Pittsburgh , 3939 O\u2019Hara St , Pittsburgh , PA 15260 , USA a r t i c l e i n f o Article history : Received 15 December 2014 Revised 14 August 2015 Accepted 22 August 2015 Available online 29 August 2015 Keywords : Creativity Problem solving Conceptual combination a b s t r a c t Theories of creative conceptual combination hypothesize that , to generate highly creative concepts , one should attempt to combine source concepts that are very different from each other . While lab studies show a robust link between far combinations and increased novelty of concepts , empirical evidence that far combinations lead to more creative concepts ( i . e . , both more novel and of higher quality ) is mixed . Drawing on models of the creative process , we frame conceptual combination as a divergent process , and hypothesize that iteration is necessary to convert far combinations into creative concepts . We trace conceptual genealogies of many hundreds of concepts proposed for a dozen different problems on a large - scale Web - based innovation platform , and model the effects of combination distance on creative outcomes of concepts . The results are consistent with our predictions : ( 1 ) direct effects of far combina - tions have a mean zero effect , and ( 2 ) indirect effects of far combinations ( i . e . , building on concepts that themselves build on far combinations ) have more consistently positive effects . This pattern of effects is robust across problems on the platform . These \ufb01ndings lend clarity to theories of creative conceptual combination , and highlight the importance of iteration for generating creative concepts . (cid:1) 2015 Elsevier B . V . All rights reserved . 1 . Introduction How are creative outcomes produced ? Conceptual combination is one strategy that has been examined in some depth . It is deceptively simple and process - free in de\ufb01nition : it involves two or more concepts combined into a new concept . Real - world examples of the products of conceptual combination abound , from \u2018\u2018mash - ups\u201d and hip - hop sampling in music , to \u2018\u2018fusion\u201d cooking , to compound engineered products ( like the Apple iPhone , and com - ponent / module reuse in engineering ) . Lab studies have identi\ufb01ed a number of different cognitive processes for combining concepts , including property transfer ( transferring properties from \u2018\u2018helper\u201d concepts to a head concept , e . g . , \u2018\u2018pet - bird\u201d = \u2018\u2018bird you keep in the house and feed when hungry\u201d ) , hybridization ( interpreting a new concept as a \u2018\u2018cross\u201d or \u2018\u2018blend\u201d between the constituent concepts , e . g . , \u2018\u2018saw - scissors\u201d = \u2018\u2018dual purpose tool that both cuts and saws\u201d ) , and relational linking ( constituent concepts play distinct roles in a thematic relation , e . g . , pet - bird = \u2018\u2018bird for grooming pets\u201d ) . Here , we are particularly interested in how conceptual combination distance \u2014 the degree of semantic distance between the component concepts \u2014 in\ufb02uences the creativity of the produced concepts . Speci\ufb01cally , many theorists and eminent cre - ators ( Blasko & Mokwa , 1986 ; Koestler , 1964 ; Mednick , 1962 ; Rothenberg , 1979 ) contend that far combinations are more likely to lead to creative outcomes than near combinations , and numer - ous anecdotes of eminent creative accomplishments are consistent with this claim ( Johansson , 2006 ; Rothenberg , 1995 ; Ward , 2001 ) . Is this hypothesis supported by empirical evidence ? Lab studies have consistently shown that far combinations \u2014 where constituent concepts are semantically distant from each other ( e . g . , \u2018\u2018kitchen utensil\u201d and \u2018\u2018bird\u201d vs . \u2018\u2018kitchen utensil\u201d and \u2018\u2018plate\u201d ) \u2014 lead to more novel combinations ( Doboli , Umbarkar , Subramanian , & Doboli , 2014 ; Gielnik , Frese , Graf , & Kampschulte , 2011 ; Mobley , Doares , & Mumford , 1992 ; Nagai , Taura , & Mukai , 2009 ; Wilkenfeld & Ward , 2001 ; Wilkenfeld , 1995 ; Wisniewski , 1997 ) . A major factor in why this effect occurs is that people generate attributes of the product concept that are emergent , i . e . , not characteristic of its constituent concepts . For example , one might say that a \u2018\u2018kitchen - utensil bird\u201d is a bird that has a strong jaw for hammering ( where neither property is likely to be listed as characteristic of either kitchen utensils or birds when considered separately ) . Emergent attributes can be generated through \ufb01rst identifying alignable con\ufb02icts through analogical mapping ( Hampton , 1997 ) and performing causal reasoning to generate attributes to reconcile those con\ufb02icts ( Kunda , Miller , & Claire , 1990 ) . Another reason novel concepts are more likely to http : / / dx . doi . org / 10 . 1016 / j . cognition . 2015 . 08 . 008 0010 - 0277 / (cid:1) 2015 Elsevier B . V . All rights reserved . \u21d1 Corresponding author at : 2504B Newell - Simon Hall , Carnegie Mellon University , 5000 Forbes Ave , Pittsburgh , PA 15213 , USA . E - mail addresses : joelchuc @ cs . cmu . edu ( J . Chan ) , schunn @ pitt . edu ( C . D . Schunn ) . Cognition 145 ( 2015 ) 104 \u2013 115 Contents lists available at ScienceDirect Cognition journal homepage : www . elsevier . com / locate / COGNIT emerge from combining dissimilar concepts is that people are more likely to think of abstract relations and attributes of constituent concepts ( e . g . , using metaphor ) when those concepts are distantly related ( Mumford , Baughman , Maher , Costanza , & Supinski , 1997 ) . In contrast to the link between combination distance and nov - elty that has been well established in the lab , the impact of combi - nation distance on idea creativity is less clear . Most major models of creativity agree that products are creative if they are both novel and good ( of high quality , useful ; Boden , 2004 ; Finke , Ward , & Smith , 1996 ; Hennessey & Amabile , 2010 ; Runco , 2004 ; Sawyer , 2012 ; Shah , Vargas - Hernandez , & Smith , 2003 ) . However , rela - tively few studies of conceptual combination and creativity have actually measured quality or creativity . Two lab studies have shown that more distant combinations lead to lower quality ideas ( Baughman & Mumford , 1995 ; Mobley et al . , 1992 ) , while one lab study has shown that it has no signi\ufb01cant effect , but trending toward higher quality ( Doboli et al . , 2014 ) . Thus , the connection to quality is unclear . Four lab studies have examined effects on cre - ativity ( i . e . , the joining of novelty and quality ) : two found positive effects ( Howard - Jones , Blakemore , Samuel , Summers , & Claxton , 2005 ; Zeng , Proctor , & Salvendy , 2011 ) , while the other two found no effect ( Jang , 2014 ; Siangliulue , Arnold , Gajos , & Dow , 2015 ) , with Siangliulue et al . ( 2015 ) showing a trend in favor of lower diversity leading to higher creativity . The relatively small number of studies with mixed results leaves us with uncertainty about the relationship between concept similarity in conceptual combination and creativity . One interpre - tation of these mixed \ufb01ndings is that far combinations lead only to increased novelty per se , not necessarily increased creativity . A related controversy exists in the literature on analogical distance , where studies are divided on whether the most creative analogi - cally inspired ideas come from analogies outside of the problem domain ( in other words , from far analogies ) . Some researchers argue that the best interpretation of the data is that there is no clear / general advantage of far analogies for creative ideation ( e . g . , Chan , Dow , & Schunn , 2015 ; Dunbar , 1997 ; Perkins , 1983 ; Weisberg , 2009 , 2011 ) . Is a similar conclusion ( combination dis - tance does not in\ufb02uence creativity ) warranted based on the extant empirical data on combination distance ? We believe it is plausible , but argue that alternative theoretical interpretations should \ufb01rst be ruled out before accepting it . In this paper , we develop and test one theoretically motivated alternative explanation for the con - \ufb02icting \ufb01ndings : the bene\ufb01ts of combination distance depend on how much convergence has happened from the point of combina - tion . We argue that , to detect the bene\ufb01ts of combination distance , we need to observe and evaluate the resulting solution path further down its path of development ( vs . early on in its development ) . To develop our alternative explanation , we draw on a generally shared process model of creativity as involving \ufb01rst , divergent ( generating new ideas ) , then convergent ( selecting and building on the best ideas ) processes ( Amabile , 1983 ; Finke et al . , 1996 ; Sawyer , 2012 ; Simonton , 2011 ; Wallas , 1926 ; Warr & O\u2019Neill , 2005 ) . For example , Amabile\u2019s ( 1983 ) prominent process model prominently includes a movement from divergent processes ( response generation ) to convergent processes ( response valida - tion ) . Similarly , the Geneplore model ( Finke et al . , 1996 ) speci\ufb01es a Generate phase ( initial generation of candidate ideas ) followed by an Explore phase ( extensive exploration of those ideas ) . Sim - plistically , one can view the creative process as linearly progress - ing from a divergent to a convergent phase . Realistically , creators often go through many divergent - convergent cycles when devel - oping creative products ( Herring , Jones , & Bailey , 2009 ; Jin & Chusilp , 2006 ) . They also sometimes interleave divergent and con - vergent processes throughout , but transition from earlier periods with more divergence to later periods with less divergence ( Atman et al . , 2007 ; Ball , Evans , Dennis , & Ormerod , 1997 ; Goel & Pirolli , 1992 ; Shih , Venolia , & Olson , 2011 ) , where convergence on a few promising prototypes becomes necessary to move for - ward . Overall , there is theoretical consensus that divergent and convergent processes are distinct and jointly necessary for success - ful creative production , and the creative process moves from an emphasis on divergent processes early on to convergent processes later on . This theoretical framework provides a principled justi\ufb01cation for the hypothesis that far combinations should lead to more cre - ative ideas . If creativity is the production of artifacts that are both new and valuable , then at least some novelty is necessary to create new value . It follows , then , that a creative process that lacked divergence entirely ( e . g . , only selected from existing ideas ) would be highly unlikely to produce a creative idea . Relatedly , models of \ufb01rm innovation often focus on the tradeoff between exploring uncertain new opportunities and exploiting existing / old certainties ( March , 1991 ) . In such models , an exclusive focus on exploitation might be bene\ufb01cial in the short run , but usually leads to an even - tual loss of competitive advantage in dynamically competitive environments . We claim that far conceptual combinations in particular \u2014 given the usual nature of their conceptual products \u2014 are a primarily divergent process for generating new ideas . Therefore , incorporating them into the creative process should eventually increase the likelihood of a highly creative idea , even if they only raise the novelty of ideas considered ( but hold quality constant ) . By contrast , near conceptual combinations could serve both divergent and convergent thinking purposes . Importantly , understanding far conceptual combination as pri - marily a divergent process can help explain the con\ufb02icting \ufb01ndings on far combinations and creative outcomes . Within this framing , we can draw on the literature on divergent / convergent creative processes to suggest multiple reasons why combination distance might not have an immediate bene\ufb01t for creativity . First , some researchers argue that a good divergent process increases quality variance in order to make it more likely that the best ideas will be generated ( Girotra , Terwiesch , & Ulrich , 2010 ; Terwiesch & Ulrich , 2009 ) . Therefore , far combination will likely produce both good and bad ideas . Some form of selection process should then be necessary to separate the good ideas from the bad ideas . Sec - ondly , if we conceive of a solution space for creative problems as possessing no more than a few \u2018\u2018peaks\u201d ( i . e . , really good ideas ) , then statistically there should be many more mediocre or bad ideas than good ideas . It follows from this sparse quality peaks perspective that initial forays into very new regions of the space , if they are \u2018\u2018blind\u201d ( Simonton , 2011 , 2012 ) , will more likely land on mediocre or bad ideas than good ones on the \ufb01rst try . Thus , some time must be allowed to pass in order for some convergent process to select and re\ufb01ne the \u2018\u2018good novel\u201d ideas ( i . e . , to move from the low qual - ity initial landing spot in a novel conceptual region to the nearby high quality variants in that conceptual region ) . Finally , models and studies of idea generation consistently \ufb01nd that better ideas overall ( i . e . , combinations of both novelty and quality ) tend to be generated later down a solution path ( Basadur & Thompson , 1986 ; Benedek & Neubauer , 2013 ; Kohn , Paulus , & Choi , 2011 ; Krynicki , 2014 ; Nijstad , De Dreu , Rietzschel , & Baas , 2010 ; Nijstad & Stroebe , 2006 ; Parnes , 1961 ; Parnes & Meadow , 1959 ; Paulus , Kohn , Arditti , & Korde , 2013 ; Rhodes & Turvey , 2007 ; Rietzschel , Nijstad , & Stroebe , 2007 ) . These theoretical insights suggest a potential resolution to the mixed \ufb01ndings regarding combination distance and idea creativity : to observe the bene\ufb01ts of combination distance , one needs to examine its effects well into the convergent phase of the creative process . Given the high - quality - variance nature of far conceptual combination as a creative strategy , a longer convergent phase J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 105 ( i . e . , with iteration ) may be necessary to convert initially highly novel and highly variable quality ideas to creative solutions ( high on both novelty and quality ) . Therefore , we predict that , if we sep - arately observe the creativity of direct and indirect conceptual descendants of far combinations , we will see a positive effect for indirect , but not direct , descendants . We should be clear that we are not making the trivial claim that raw ideas must \ufb01rst be elaborated and iterated on to produce a cre - ative \ufb01nal product ( i . e . , a main effect of time on idea creativity ) . What we claim is that initial raw ideas from far combinations in particular are likely to have high novelty , but uncertain utility , and that iterations on these raw ideas ( not prototype testing ) is necessary to get to creative raw ideas ( i . e . , concepts that are both highly novel and have high potential utility ) . In other words , we are making an interaction prediction : the bene\ufb01ts of far over near combinations on idea creativity will only emerge at later time points . Anecdotal accounts of creative discovery by conceptual combination do not clearly specify whether dissimilar conceptual combinations lead to immediately creative raw ideas : some speak of it purely in terms of generating \u2018\u2018fresh\u201d ( i . e . , novel ) ideas , being agnostic about the expected potential utility of ideas . Others simply claim that dissimilar conceptual combinations directly yield more creative raw ideas with above average probability . For example , Ward\u2019s ( 2001 ) analysis of the origins of science \ufb01ction author Stephen Donaldson\u2019s award - winning The Chronicles of Thomas Covenant the Unbeliever fantasy series involves no speci\ufb01 - cation of iteration between the initial leprosy - unbelief conceptual combination and the \ufb01nal idea that formed the overarching theme of the series . In this paper , we empirically test our theoretically - driven pre - diction that observations of the bene\ufb01ts of conceptual combination distance vary with the genealogical lag between source and target ideas . Taking a genealogical approach , we trace lagged effects of conceptual combination distance on the creativity of direct and indirect conceptual descendants on a real - world innovation plat - form . We also address some key methodological issues in prior studies ( to increase our con\ufb01dence in our theoretically motivated hypothesis testing ) . First , all prior studies examined relatively few creativity problems , and were conducted only in the lab ( and there - fore under somewhat arti\ufb01cial conditions , and often with toy prob - lems ) . Some key studies ( Doboli et al . , 2014 ; Jang , 2014 ; Zeng et al . , 2011 ) also had relatively low Ns , making null effects ambiguous and raising potential concerns about effect sizes being signi\ufb01cantly overestimated ( Button et al . , 2013 ) , or even incorrectly estimated as positive / negative in sign ( Gelman & Weakliem , 2009 ) . Studies are needed that examine ( 1 ) a range of problems with ( 2 ) large Ns , ( 3 ) under more realistic conditions , and ( 4 ) use creativity rather than just novelty or quality in isolation to estimate the general effect of combination distance on creativity . Therefore , the research reported in this paper is conducted on a diverse range of problems , with large numbers of participants working on real world challenges , testing hypotheses with respect to a creative outcome measure that combines both novelty and quality . 2 . Methods 2 . 1 . Dataset We examine the relationship between combination distance and creative outcomes in the context of OpenIDEO ( www . openideo . com ) , a Web - based innovation platform that addresses a range of social and environmental problems ( e . g . , managing e - waste , increasing accessibility in elections ; see Appendix A for more details on the diverse set of problems sampled for our study ) . Expert designers from IDEO \u2014 a design consulting \ufb01rm renowned for its creativity \u2014 guide platform contributors through a structured design process to produce solutions for these problems that are ultimately implemented for real - world impact ( \u2018\u2018Impact Stories , \u201d n . d . ) . We focus our analysis in this study on processes and outcomes in three crucial early phases in the process . (cid:1) First , in the inspiration phase ( (cid:3) 1 . 5 to 4 weeks ) , contributors help to de\ufb01ne the problem space and identify promising solu - tion paths by posting inspirations : descriptions of solutions to analogous problems and case studies of stakeholders . (cid:1) In the concepting phase that follows ( for the next 2 \u2013 6 weeks ) , contributors post concepts : speci\ufb01c proposed solutions to the stated problem . Crucially , contributors cite concepts or inspira - tions that serve as sources of inspiration for their idea : this pro - vides our process data for conceptual combination . Fig . 1 shows an example inspiration and concept for an OpenIDEO problem about managing electronic waste . They are representative of the typical length and level of speci\ufb01cation of inspirations and concepts on the platform . (cid:1) In the shortlist phase , a subset of concepts for each problem is shortlisted by an expert panel ( composed of the OpenIDEO designers and a set of domain experts / stakeholders ) for further re\ufb01nement , based on their creative potential . (cid:1) In later stages , these concepts are re\ufb01ned and evaluated in more detail , and then a subset of them is selected for implementation . We focus on the \ufb01rst three stages given our focus on the use of conceptual combination for generating creative raw ideas ( the later stages involve many other design processes , such as prototyping ) . For more details on the dataset , see Chan ( 2014 ) . 2 . 2 . Sample The full dataset for this study consists of 2341 concepts and 4557 inspirations posted for 12 distinct problems by 1190 unique contributors . All inspirations and concepts were downloaded ( with administrator permission ) , and a HTML parser was used to extract the following metadata : ( 1 ) Concept / inspiration author ( who posted the concept / inspiration ) . ( 2 ) Number of comments . ( 3 ) Shortlist status ( yes / no ) . ( 4 ) List of cited sources of inspiration . ( 5 ) Full - text of concept / inspiration . The current study was conducted with two subsamples of this larger dataset . Speci\ufb01cally , our analysis focused on concepts that ( for subsample 1 ) directly cited at least 2 inspirations or ( for sub - sample 2 ) indirectly cited at least 2 inspirations . We de\ufb01ne in the next section how we operationalize indirect citations . Our sam - pling criteria re\ufb02ect our focus on measuring the effects of concep - tual combination distance , which is not measurable with fewer than two sources . The \ufb01rst subsample includes 456 concepts posted by 239 contributors , collectively citing 2167 unique inspira - tions . The second subsample includes 522 concepts posted by 281 authors , collectively citing 2556 unique inspirations . We were able to obtain professional expertise information ( e . g . , personal websites , online portfolios , pro\ufb01le pages on company names ) posted in the public OpenIDEO pro\ufb01les of 90 contributors ( approximately 1 / 3 of the authors in the sub - samples ) . In this sub - sample of the contributors , at least 1 / 3 are professionals in design - related disciplines ( e . g . , user experience / interaction design , communication design , architecture , product / industrial design , entrepreneurs and social innovators , etc . ) and / or domain experts or stakeholders ( e . g . , urban development researcher contributing to the vibrant - cities challenge , education policy researcher 106 J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 contributing to the youth - employment challenge , medical profes - sional contributing to the bone - marrow challenge ) . Thus , from a contributor perspective , our sample includes a range of creative / design expertise , from novice to expert . 2 . 3 . Measures 2 . 3 . 1 . Conceptual genealogies To examine the effects of indirect conceptual descendants , we constructed conceptual genealogies for all concepts in the sample . These genealogies were constructed via breadth - \ufb01rst search through the citation graph gathered in initial data collection : this search \ufb01rst returned all sources that a concept built upon , and then returned all sources that each of these sources built upon ( whether they were concepts or inspirations ) , traversing the conceptual tree to its endpoint . If duplicate entries were encountered , that source was credited at its \ufb01rst appearance in the graph : for instance , if an inspiration I was a direct source for a concept C ( at level 1 ) , and also for another concept / inspiration at level 2 , it would only be counted once as a level 1 source for C . In this study , we de\ufb01ned indirect conceptual descendants as inspirations from levels 2 to 4 of each concept\u2019s genealogy ( see Fig . 2 ) : this range choice re\ufb02ects our goal of examining the effects of sources that are \u2018\u2018just recent enough\u201d to have discernible effects ( we may not be able to Notice from Fig . 2 that indirect sources would also include inspirations cited by cited concepts ( i . e . , the sources of concepts that acted as immediate sources for the root concept ) . One way to think about this relationship of the root con - cept with these indirect sources of other concepts is that ( at least part of ) the insights / information / ideas contained in those inspira - tions are \u2018\u2018passed on\u201d to the root concept through their incorpora - tion into the concepts immediately cited by the root concept . 2 . 3 . 2 . Creativity of concepts Concept creativity is operationalized as whether a concept was shortlisted . In OpenIDEO , concepts are selected for the shortlist by a panel of expert judges , including the original stakeholders who posted the problem and a team of OpenIDEO designers . Both groups of judges have signi\ufb01cant expertise that quali\ufb01es them to judge the concepts\u2019 creativity : the stakeholders have spent signi\ufb01 - cant time searching for and learning about existing approaches , and the OpenIDEO designers , in addition to their expertise in the general domain of creative design , have spent considerable time upfront with the stakeholders , learning about and de\ufb01ning the problem space . An expert panel is widely considered a \u2018\u2018gold standard\u201d for mea - suring the creativity of ideas ( Amabile , 1982 ; Baer & McKool , 2009 ; Brown , 1989 ; Sawyer , 2012 ) . Further , addressing our need for a creativity measure that jointly considers novelty and quality , we learned from conversations with the OpenIDEO team that the panel\u2019s judgments combine consideration of both novelty and usefulness / appropriateness ( here operationalized as potential for impact ; A . Jablow , personal communication , May 1 , 2014 ) . Addi - tionally , since problems posted on OpenIDEO are unsolved , success - fulconceptsmustbedifferentfrom ( and , perhapsmoreimportantly , signi\ufb01cantly better than ) existing unsatisfactory solutions . To validate the reported focus of the IDEO panel , we obtained independent external expert judgments on Likert - like scales on separate dimensions of novelty , impact , and feasibility for a subset of concepts in the OpenIDEO data . Speci\ufb01cally , we collected approximately four expert judgments per concept for \ufb01ve of the Fig . 1 . Example inspiration ( left ) and concept ( right ) for an OpenIDEO problem about managing electronic waste . Fig . 2 . Illustrated example of \u2018\u2018indirect\u201d sources as sources in levels 2 \u2013 4 of a speci\ufb01c target concept\u2019s genealogy . Teal circles denote concepts ; maroon circles denote inspirations . Note that some indirect sources in this example serve as direct sources for the earlier concepts in this genealogy . J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 107 challenges ( N = 318 concepts ) . Re\ufb02ecting the complex and multi - disciplinary nature of the challenges , the expert\u2019s ratings had mod - erate levels of agreement ( ICCs of . 46 , . 57 , and . 63 for novelty , impact , and feasibility , respectively ) . All three ratings were posi - tively associated with short - list status ( novelty r pb = 0 . 09 , p = . 10 ; impact r pb = 0 . 21 , p = . 00 ; novelty r pb = 0 . 17 , p = . 00 ) . Fitting a sim - ple logistic regression of shortlist on the three dimensions shows that impact is a strong predictor ( b = . 51 , p = . 02 ) . Feasibility is mar - ginally predictive ( b = . 33 , p = . 07 ) , while novelty has a positive but nonsigni\ufb01cant estimate ( b = . 11 , p = . 63 ) , allaying potential concerns about panel bias against novel concepts . 2 . 3 . 3 . Combination distance A standard approach to measuring combination distance would be to obtain pairwise conceptual distance judgments between all conceptual descendants . However , the scale of the present study ( more than 2000 inspirations would require more than 2 million pairwisecomparisons ) presentsformidablechallengestomeasuring distance using human judgments , from not only a cost standpoint , but also an effectiveness perspective , since the quality of human judgments can deteriorate severely if the workload is too high . We therefore took a computational approach to measuring combination distance . Speci\ufb01cally , we employed Latent Dirichlet Allocation ( LDA ; Blei , Ng , Jordan , & Lafferty , 2003 ) \u2014 an unsuper - vised machine learning technique for learning topical structures from unstructured texts \u2014 to learn the semantic space of ideas posted on OpenIDEO , and used that space to estimate semantic dis - tance between inspirations . LDA treats documents as mixtures of latent \u2018\u2018topics\u201d ( occurring with different \u2018\u2018weights\u201d in the mixture ) , and uses Bayesian statistical learning algorithms to infer the latent topical structure of the corpus ( and the topical mixtures for each document ) from the co - occurrence patterns of words across docu - ments . With this inferred topical structure , we can then derive con - ceptual similarity between any pair of documents by computing the cosine between their topical mixtures ( which are represented as vectors of topic - weights ) . Essentially , documents that share dominant topics in similar relative proportions ( e . g . , primarily about recycling and electronics ) are the most similar . This similar - ity is measured by computing the cosine between their topic - weight vectors , yielding a similarity score between 0 and 1 ( where values closer to 1 indicate greater similarity ) . LDA has been successfully used to model semantics in many other contexts , including modeling semantic memory representa - tion phenomena ( Grif\ufb01ths , Steyvers , & Tenenbaum , 2007 ) , support - ing knowledge discovery and information retrieval in repositories of scienti\ufb01c papers ( Grif\ufb01ths & Steyvers , 2004 ) , and analyzing topical dynamics in social media use ( Schwartz et al . , 2013 ) . Our application of LDA in the OpenIDEO corpus was validated by exam - ining correlations with human judgments on two sub - samples in the corpus . We collected Likert - scale pairwise similarity judg - ments for inspirations from 3 research assistants for the \ufb01rst sub - sample , and 5 from the second . Inter - rater agreement was acceptable , aggregate consistency intraclass correlation coef\ufb01cient ( ICC ( 2 , 3 ) ) = . 46 for the \ufb01rst sub - sample , and ICC ( 2 , 5 ) = . 74 for the second sub - sample . The correlation of the LDA cosines with the mean human - judged pairwise similarities was high in both sub - samples , at r = . 54 and r = . 51 , respectively . Notably , these agreement levels were better than the highest correlation between individual human raters in both subsamples ( r = . 39 and r = . 48 , respectively ) , reinforcing the value of automatic coding methods for this dif\ufb01cult task . Further details on our implementation and validation of LDA are available in Chan ( 2014 ) . Combination distance ( hereafter denoted COMB - DIST DIR for direct sources , and COMB - DIST IND for indirect sources ) was mea - sured for each concept as the mean of the reversed pairwise cosi - nes between inspirations cited by that concept ( i . e . , subtracting from 0 , to derive distance rather than similarity ) . Fig . 3 shows an example near and far COMB - DIST DIR from the data . Note that conceptual combination research has tended to focus on pairs of concepts being combined , but here there were often more than just two concepts being combined , especially in the indi - rect source set . Real world problems are complex , and often involve many subproblems , such that diverse sources must be brought together to solve each of the subproblems . We used mean pairwise distance ( the most natural conceptualization of combination distance ) to characterize the general diversity among the sources . But it could be that problem solvers brought the most similar pieces together in pairs , or were most in\ufb02uenced by the maximum distance among sources . Therefore we also explored using min and max distance measures rather than mean distance in our anal - yses . These approaches produced similar results ( with slightly more statistical noise ) , suggesting the patterns we found were not due to idiosyncrasies of how we conceptualized combination distance . However , the added noise when considering min and max distance also suggested our initial intuitions that mean distance is the most appropriate operationalization of combination distance were correct . Therefore , we only report the mean distance results . 2 . 3 . 4 . Control measures To improve our estimates of the effects of combination distance per se in this multi - faceted naturalistic dataset , we measured and accounted for other important factors that may in\ufb02uence concept creativity ( i . e . , we statistically controlled for likely confounds ) . 2 . 3 . 4 . 1 . Feedback . Feedback can be an important contributor to the quality of a concept . Feedback can provide encouragement , raise issues / questions , or provide speci\ufb01c suggestions for improvement , all potentially signi\ufb01cantly enhancing the quality of the concept . Further , feedback may be an alternate pathway to success via com - bination distance , in that concepts that build on far combinations may attract more attention and therefore higher levels of feedback , which then improve the quality of the concept . On OpenIDEO , con - cepts receive feedback in the form of comments . We operationalize feedback ( labeled here as FEEDBACK ) as the number of comments received by a given concept . 2 . 3 . 4 . 2 . Quality of cited sources . Concepts that build on existing high - quality concepts ( e . g . , those that end up being shortlisted ) may be more likely to be shortlisted : contributors might incorpo - rate lessons learned from the mistakes and shortcomings , good ideas , and feedback in these high - quality concepts . We operationalize source quality ( SOURCEQUAL ) as the number of shortlisted concepts a given concept builds upon . 2 . 3 . 4 . 3 . Conceptual source distance from problem . Finally , building on a prior study in the OpenIDEO context that showed a positive effect of sources that were conceptually closer to the problem domain ( Chan et al . , 2015 ) , we also control for the distance of sources from the problem domain . Source distance ( here labeled SP - DIST ) is measured for each concept by taking the mean of the reverse cosines between cited inspirations and the problem . 2 . 4 . Analytic approach Our goal is to model the creative outcomes of concepts posted by contributors for 12 different problems as a function of combina - tion distance , controlling for other factors . However , contributors are not cleanly nested within problems , nor vice versa ; concepts are cross - classi\ufb01ed within both authors and challenges ( see Fig . 4 ) . This cross - classi\ufb01ed structure violates assumptions of uni - form independence between concepts : concepts posted by the same contributor or within the same problem are likely to be 108 J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 correlated with each other on various dimensions , most impor - tantly overall quality . Failing to account for this non - independence could lead to overestimates of the statistical signif - icance of model estimates ( i . e . , make unwarranted claims of statis - tically signi\ufb01cant effects ) . This issue is exacerbated when testing for small effects within large datasets . Additionally , while we are primarily interested in concept - level outcomes , we need to model between - contributor effects to separate out contributor - effects ( e . g . , higher / lower creativity , effort ) from the impact of sources on individual concepts . Therefore , we employ generalized linear mixed models ( GLMMs ) to model both \ufb01xed effects ( of our inde - pendent and control variables ) and random effects ( potential vari - ation of the outcome variable attributable to contributor - or problem - variation and also potential between - problem variation in the effect of combination distance ) on shortlist status ( a binary variable , which requires logistic , rather than linear , regression ) . The following is the general structure of these models ( in mixed model notation ) : g i \u00f0 contributorjchallenge k \u00de \u00bc c 00 \u00fe X q c q 0 X qi \u00fe u 0 contributorj \u00fe u 0 challenge k where (cid:1) g i \u00f0 contributorjchallengek \u00de is the predicted log odds of being shortlisted for the i th concept posted by the j th contributor in the k th challenge (cid:1) c 00 is the grand mean log odds for all concepts (cid:1) c q 0 is a vector of q predictors ( q = 0 for our null model ) (cid:1) u 0 contributorj and u 0 challengek model between - contributor and between - challenge variability in mean c 00 We \ufb01t our GLMMs using the glmer function in the lme4 pack - age ( Bates , Maechler , Bolker , & Walker , 2013 ) in R ( R Core Team , 2013 ) , using full maximum likelihood estimation by the Laplace approximation . Our general modeling strategy is as follows . First , we \ufb01t a reduced model with crossed random effects of challenge and con - tributor , and \ufb01xed effects only of our control measures ( i . e . , feed - back , source quality , and source problem - distance ) . Because these are theoretically motivated predictors , we leave them in the model regardless of statistical signi\ufb01cance . This reduced model serves as a more realistic baseline than the null model ; we com - pare the reduced model to a second ( \ufb01xed - slope ) model with the added \ufb01xed effect of combination distance . Finally , we \ufb01t a third ( random - slope ) model with an added parameter u 1 challengek to model potential challenge - level random effects on the mean effect of combination distance . To select our \ufb01nal model , we choose the model that meets three criteria : ( 1 ) signi\ufb01cantly reduces deviance from the null model ( low standard for explanatory power ) , ( 2 ) signi\ufb01cantly reduces deviance compared to the reduced model from the previous step ( higher standard for explanatory power ) , and ( 3 ) has a lower Akaike Information Criterion ( AIC ) than the previous step to avoid over\ufb01tting . Fig . 3 . Example near ( left ) and far ( right ) combinations according to COMB - DIST DIR . Fig . 4 . Illustrated cross - classi\ufb01ed structure of data . J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 109 3 . Results 3 . 1 . Direct effects of combination distance We \ufb01rst examine the hypothesis that combining diverse sources leads directly to ideas that are more creative . Recall that this analysis is with the sub - sample of 456 concepts . 3 . 1 . 1 . Descriptive statistics Table 1 summarizes the descriptive statistics and intercorrela - tions between the variables . There are statistically signi\ufb01cant pos - itive correlations between the control variables and Pr ( shortlist ) ; hence the importance of including them in the models . There are no strong inter - correlations between the predictor variables , alleviating potential concerns about multicollinearity ; a variance in\ufb02ation analysis also shows that having COMB - DIST DIR and SP - DIST in the same model should not introduce multicollinearity , with variance in\ufb02ation factors of 1 . 16 for both variables . 3 . 1 . 2 . Statistical models We \ufb01t a series of generalized linear mixed models using full maximum likelihood estimation by the Laplace approximation , with concepts cross - classi\ufb01ed within both contributors and problems . We rescale COMB - DIST DIR ( multiplying it by 10 ) for easier interpretation ( a more meaningful \u2018\u20181 - unit\u201d change ) . Table 2 presents the model estimates and \ufb01t statistics for the GLMMs . The \ufb01rst model is a baseline model \ufb01tted with the control variables as predictors . This model yields a large and statistically signi\ufb01cant reduction in deviance compared to the null model , v 2 ( 2 ) = 64 . 70 , p = 0 . 00 . Adding a \ufb01xed slope for COMB - DIST DIR to this model does not provide any meaningful reduction in deviance , with the likelihood ratio being essentially zero , v 2 ( 1 ) = 0 . 00 , p = 0 . 92 , and an increase in the AIC . The point estimate for the effect of a change of . 10 in COMB - DIST DIR ( remember that it is rescaled in this model ) is also essentially zero ( see Fig . 5 ) , albeit with a fairly wide con\ufb01dence interval . To ensure that this wide con\ufb01dence interval is not due to one or two outlier problems overwhelming an overall positive or negative trend across the problems , we estimate an additional model with a random slope for COMB - DIST DIR . Visually inspecting the posterior modes for the slope of COMB - DIST DIR for each problem ( see Fig . 5 ) , we see an even scatter about the mean value , with 6 challenges having a positive sign for the coef\ufb01cient for diversity , and 6 challenges having a negative sign . A binomial sign test with a null hypothesis of equal probability for positive and negative effects of diversity estimates a two - tailed p - value of 1 . 00 of observing either 6 or fewer positive or 6 or more negative signs in 12 \u2018\u2018trials\u201d ( although this binomial test outcome should be intuitively obvious to the reader , we present it here to parallel the same test conducted on the indirect problem - speci\ufb01c slopes ) . The graph in Fig . 5 gives the impression that there are clear effects of combination distance , which can be positive or negative Table 1 Descriptive statistics and intercorrelations between variables for COMB - DIST DIR . Variable Descriptives Correlations M ( SD ) FEEDBACK SOURCE SHORT SP - DIST COMB - DIST DIR SHORTLIST 0 . 16 ( 0 . 36 ) 0 . 33 * * * 0 . 11 * * (cid:4) 0 . 10 * (cid:4) 0 . 01 FEEDBACK 9 . 14 ( 9 . 92 ) 0 . 12 * * 0 . 02 0 . 05 SOURCESHORT 0 . 61 ( 1 . 07 ) (cid:4) 0 . 05 0 . 10 * SP - DIST (cid:4) 0 . 13 ( 0 . 62 ) 0 . 29 * * * COMB - DIST DIR 2 . 02 ( 1 . 25 ) \u2013 m p < . 10 . * p < . 05 . * * p < . 01 . * * * p < . 001 . Table 2 Model estimates and \ufb01t statistics for cross - classi\ufb01ed multilevel logistic regressions of Pr ( shortlist ) on COMB - DIST DIR , with comparison to baseline model ( controls ) . Baseline model ( controls ) COMB - DIST DIR , \ufb01xed slope COMB - DIST DIR random slope Fixed effects c 00 , intercept (cid:4) 3 . 08 [ (cid:4) 3 . 37 , (cid:4) 2 . 12 ] (cid:4) 3 . 05 [ (cid:4) 3 . 99 , (cid:4) 2 . 12 ] (cid:4) 3 . 03 [ (cid:4) 4 . 11 , (cid:4) 1 . 95 ] c 10 , FEEDBACK 0 . 10 * * * [ 0 . 07 , 0 . 12 ] 0 . 10 * * * [ 0 . 07 , 0 . 13 ] 0 . 10 * * * [ 0 . 07 , 0 . 13 ] c 20 , SOURCESHORT 0 . 25 m [ (cid:4) 0 . 10 , 0 . 35 ] 0 . 25 m [ (cid:4) 0 . 03 , 0 . 52 ] 0 . 26 m [ (cid:4) 0 . 03 , 0 . 54 ] c 30 , SP - DIST (cid:4) 0 . 49 m [ (cid:4) 0 . 71 , 0 . 10 ] (cid:4) 0 . 50 m [ (cid:4) 1 . 05 , 0 . 04 ] (cid:4) 0 . 54 * [ (cid:4) 1 . 08 , (cid:4) 0 . 00 ] c 40 , COMB - DIST DIR 0 . 01 [ (cid:4) 0 . 27 , 0 . 30 ] 0 . 03 [ (cid:4) 0 . 28 , 0 . 33 ] Random effects u 0 authorj 0 . 47 0 . 47 0 . 44 u 0 challengek 0 . 71 0 . 71 1 . 63 u 1 challengek 0 . 05 Model \ufb01t statistics I Deviance 323 . 57 323 . 57 321 . 74 AIC 335 . 57 337 . 57 339 . 74 \u2044\u2044 p < . 01 . 95 % CI ( Wald ) = [ lower , upper ] . m p < . 10 . I Generalized linear mixed models ( including multilevel logistic regressions ) are \ufb01tted using maximum likelihood estimates that are not , strictly speaking , minimizing variance , and typical \ufb01t statistics like R 2 are not considered meaningful for assessing the \ufb01t of these models . Therefore , here we report common alternative \ufb01t statistics more closely related to the model estimation procedures , i . e . , deviance and AIC . * p < . 05 . * * * p < . 001 . Fig . 5 . Model - \ufb01tted relationship between combination distance and Pr ( shortlist ) . Fitted values evaluated at mean values of FEEDBACK , SOURCEQUAL , and SP - DIST . Greyed lines are \ufb01tted from posterior mode estimates of the slope of COMB - DIST DIR for each problem . 110 J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 depending upon the problem . However , it is important to note that there is not strong evidence in support of signi\ufb01cant problem mod - eration : the model estimates low problem - variance , does not meaningfully decrease variance from the \ufb01xed slope model , v 2 ( 2 ) = 1 . 83 , p = . 23 ( p - value is halved , heeding common warnings that a likelihood ratio test discriminating two models that differ on only one variance component may be overly conservative , e . g . , Pinheiro & Bates , 2000 ) , and also further increases AIC , calling into question whether the estimated problem variation is meaningful . In other words , the most parsimonious interpretation given these data is that direct combination distance has no effect ( i . e . , we select the baseline model as our \ufb01nal model ) , although it is possible that an even larger dataset would \ufb01nd problem - speci\ufb01c effects . Most importantly , these data argue strongly against a general bene\ufb01t of combination distance of direct sources on idea creativity . 3 . 2 . Indirect effects of conceptual diversity We now turn to the analysis of the effects of combination distance of indirect sources . Recall that this analysis is with the sub - sample of 522 concepts . 3 . 2 . 1 . Descriptive statistics Descriptive statistics and bivariate correlations are given in Tables 3 and 4 . There are no strong correlations among the predic - tors , giving little cause for concerns about multicollinearity . 3 . 2 . 2 . Statistical models As before , we estimate a series of generalized linear mixed models to analyze the relationship between COMB - DIST IND and Pr ( shortlist ) . We exclude COMB - DIST DIR from our models for a num - ber of reasons . First , it does not add predictive value ( as we saw in the preceding analysis ) . Second , including it as predictor would exclude concepts that had indirect inspirations as sources , but had less than two direct inspiration sources , reducing our N to 381 for no predictive gain . Finally , our estimates of the effects of COMB - DIST IND do not change with COMB - DIST DIR in the model . The model estimates are given in Table 5 . As before , we begin with a baseline controls model , which gives a large and statistically signi\ufb01cant reduction in deviance compared to the null model , v 2 ( 3 ) = 63 . 70 , p = 0 . 00 . In contrast to COMB - DIST DIR , adding a \ufb01xed slope for COMB - DIST IND to the baseline model yields a marginally signi\ufb01cant reduction in deviance , v 2 ( 1 ) = 3 . 26 , p = 0 . 07 , and a decrease in AIC , mitigating concerns about over\ufb01tting . The model estimates that a . 10 change in COMB - DIST IND corre - sponds to an increase of approximately . 45 in the log - odds of being shortlisted ( see Fig . 6 ) . Holding all the other predictors at their mean values , changing from a COMB - DIST IND of (cid:4) 0 . 20 ( close to the mean value in the sample ) to (cid:4) 0 . 10 ) increases Pr ( shortlist ) from 0 . 13 to 0 . 19 . Again , the CI for the effect is relatively wide . However , in contrast to COMB - DIST , the estimated positive effect Table 3 Descriptive statistics for indirect combination distance measures . Variable Valid N Min Max Mean Median SD SHORTLIST 522 0 1 0 . 15 0 0 . 36 FEEDBACK 522 0 67 9 . 01 6 10 . 02 SOURCESHORT 522 0 11 0 . 67 0 1 . 06 SP - DIST 522 (cid:4) 2 . 93 1 . 67 (cid:4) 0 . 11 (cid:4) 0 . 01 0 . 73 COMB - DIST IND 522 (cid:4) 0 . 73 (cid:4) 0 . 02 (cid:4) 0 . 18 (cid:4) 0 . 14 0 . 10 Table 4 Bivariate correlations for indirect combination distance measures . Variable FEEDBACK SOURCE SHORT SP - DIST COMB - DIST IND SHORTLIST 0 . 34 * * * 0 . 13 * * (cid:4) 0 . 11 * 0 . 04 FEEDBACK 0 . 11 * (cid:4) 0 . 01 0 . 13 * * SOURCESHORT (cid:4) 0 . 05 0 . 19 * * * SP - DIST (cid:4) 0 . 02 m p < . 10 . * p < . 05 . * * p < . 01 . * * * p < . 001 . Table 5 Model estimates and \ufb01t statistics for cross - classi\ufb01ed multilevel logistic regressions of Pr ( shortlist ) on COMB - DIST IND , with comparison to baseline model ( controls only ) . Baseline model ( controls ) With COMB - DIST IND , \ufb01xed slope With COMB - DIST IND random slope Fixed effects c 00 , intercept (cid:4) 2 . 80 [ (cid:4) 3 . 44 , (cid:4) 2 . 16 ] (cid:4) 1 . 98 [ (cid:4) 3 . 10 , (cid:4) 0 . 86 ] (cid:4) 2 . 12 [ (cid:4) 3 . 10 , (cid:4) 0 . 86 ] c 10 , FEEDBACK 0 . 09 * * * [ 0 . 06 , 0 . 12 ] 0 . 09 * * * [ 0 . 07 , 0 . 12 ] 0 . 09 * * * [ 0 . 07 , 0 . 12 ] c 20 , SOURCESHORT 0 . 16 [ (cid:4) 0 . 08 , 0 . 39 ] 0 . 12 [ (cid:4) 0 . 12 , 0 . 35 ] 0 . 12 [ (cid:4) 0 . 12 , 0 . 35 ] c 30 , SP - DIST (cid:4) 0 . 44 * [ (cid:4) 0 . 82 , (cid:4) 0 . 07 ] (cid:4) 0 . 45 * [ (cid:4) 0 . 83 , (cid:4) 0 . 06 ] (cid:4) 0 . 45 * [ (cid:4) 0 . 83 , (cid:4) 0 . 06 ] c 40 , COMB - DIST IND 0 . 45 m [ (cid:4) 0 . 04 , 0 . 94 ] 0 . 34 m [ (cid:4) 0 . 04 , 0 . 94 ] Random effects u 0 authorj 0 . 12 0 . 13 0 . 12 u 0 challengek 0 . 60 0 . 88 1 . 35 u 1 challengek 0 . 03 Model \ufb01t statistics Deviance 372 . 65 369 . 39 369 . 13 AIC 384 . 65 383 . 39 387 . 13 \u2044\u2044 p < . 01 . 95 % CI ( Wald ) = [ lower , upper ] . m p < . 10 . * p < . 05 . * * * p < . 001 . Fig . 6 . Model - \ufb01tted relationship between combination distance of indirect sources and Pr ( shortlist ) . Fitted values evaluated at mean values of FEEDBACK , SOURCEQUAL , and SP - DIST . Greyed lines are \ufb01tted from posterior mode estimates of the slope of COMB - DIST IND for each problem . J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 111 of COMB - DIST IND did not appear to vary by problem . An additional model with a random slope for COMB - DIST IND estimates very low problem - variance , does not meaningfully decrease variance from the \ufb01xed slope model , v 2 ( 2 ) = 0 . 26 , p = . 44 ( p - value is halved ) , and also further increases AIC . Therefore , we select the \ufb01xed slope model as our \ufb01nal model for this analysis . Importantly , when estimating the posterior modes for the effect of diversity for each problem , we see that none of the 12 challenges has a negative estimate ( see Fig . 6 ) . A binomial sign test with a null hypothesis of equal probability for positive and negative effects of diversity estimates a two - tailed p - value of 0 . 0005 of observing either 0 or fewer positive or 12 or more negative signs in 12 \u2018\u2018tri - als\u201d . In other words , the effect of combination distance of indirect sources is consistent across problems in a way that is very unlikely to have arisen by chance . 3 . 3 . Iteration chain depth and earliness of inspirations One plausible alternative explanation for our \ufb01ndings is that concepts that cite indirect sources with high combination distance are better not because of the combination distance of those sources , but because those sources were inspirations posted earlier in the challenge . These earlier inspirations might lead to better concepts for a variety of reasons unrelated to combination dis - tance , e . g . , they might be of higher quality because they are posted by \u2018\u2018early adopters\u201d to the challenge who are more motivated , or they might articulate base concepts more clearly ( e . g . , because they have more time for iteration ) . Depth in iteration chain ( our main variable of interest ) in fact had a small positive association with earliness in time ( i . e . , when the inspiration was posted ) . At the inspiration level , inspirations that were posted earlier are slightly more likely to show up deeper in iteration chains , r = 0 . 13 , p < . 001 . Inspirations that are cited as both immediate and indirect sources ( M = 9 . 5 days into the chal - lenge , SE = 0 . 16 ) tend to be posted earlier than inspirations that are only ever cited as immediate sources ( M = 12 . 2 days , SE = 0 . 49 ) . However , the mean earliness of concepts\u2019 cited inspirations is not predictive of their creative success . Estimating a generalized linear mixed model with feedback , source quality , and mean inspi - ration earliness as \ufb01xed effects and challenge and contributor as random effects , we \ufb01nd that the estimated effect of earliness is near - zero , b = 0 . 03 , 95 % CI = [ (cid:4) 0 . 01 , 0 . 08 ] , p = 0 . 17 . This model does not signi\ufb01cantly improve \ufb01t over the reduced control variables model ( i . e . , with just \ufb01xed effects of feedback , source quality , and challenge and contributor random effects ) , LRT v 2 ( 1 ) = 1 . 76 , p = 0 . 18 ( AIC = 716 . 86 vs . 716 . 62 ) . Similarly , mean depth of cited inspirations in a chain is not predictive of concepts\u2019 creative suc - cess . Adding mean chain depth to the reduced control variables model does not signi\ufb01cantly improve \ufb01t , LRT v 2 ( 1 ) = 1 . 46 , p = 0 . 23 ( AIC = 717 . 17 vs . 716 . 62 ) , and the model estimates no reli - able effect of mean chain depth , b = (cid:4) 0 . 16 , 95 % CI = [ (cid:4) 0 . 42 , 0 . 09 ] , p = 0 . 22 . Therefore , mere earliness of cited inspirations cannot explain the interaction between iteration depth and the effect of inspiration diversity on creative outcomes . 4 . Discussion 4 . 1 . Summary Our goal in this paper was to examine conceptual combination as a strategy for generating creative ideas . Theories of conceptual combination and creativity suggested the hypothesis that distant conceptual combinations are especially likely to lead to highly cre - ative ideas , but the past empirical support for this hypothesis has been uneven . Drawing on broader theories of the creative process ( Amabile , 1983 ; Finke et al . , 1996 ; Sawyer , 2012 ; Simonton , 2011 ; Wallas , 1926 ; Warr & O\u2019Neill , 2005 ) , we formulated a theoretical framework that situates distant conceptual combination within the creative process as a divergent creative strategy . This theoreti - cal framework yielded a novel re\ufb01nement of the distant combina - tion hypothesis : the bene\ufb01ts of conceptual combination distance are more likely to be seen with a genealogical lag between source and target ideas . In other words , we predicted that it takes time for distant combinations to yield their creative fruits . The current study\u2019s \ufb01ndings provided empirical support for this re\ufb01ned hypothesis . As predicted , analyzing combination distance of indirect sources indeed yielded different results than direct sources . Speci\ufb01cally , we found that the mean effect of direct com - bination distance , though slightly trending in a positive direction , was essentially zero ( with some potential problem variation ) . In contrast , combination distance of indirect sources was a positive predictor of creative outcomes , and this effect was robust across problems . Thus , distant combinations do appear to be especially likely to lead to highly creative outcomes , but only if they are \u2018\u2018indirect\u201d ( i . e . , sources of one\u2019s sources ) . Importantly , we also demonstrate that the contrast between direct and indirect sources is not explained by the mere earliness of the indirect sources . 4 . 2 . Strengths and limitations Before we draw out the larger implications of this study , we \ufb01rst note some strengths and limitations of this study . First , we note that we were able to strike a favorable tradeoff between external validity ( real designers solving real creative problems ) and statistical power , which is rare in creativity research ( i . e . , typ - ical studies of real designers have smaller Ns than lab studies , not larger Ns , as in the current study ) . This feature narrows the gap considerably from the current \ufb01ndings to generalizations in real - world creative cognition ( Dunbar & Blanchette , 2001 ; Henrich , Heine , & Norenzayan , 2010 ) . Another strength of our study is our creative outcome measure , which combines both novelty and qual - ity , and follows the gold standard expert panel approach . This allows us to think more holistically about the effects of combina - tion distance on creativity , not just novelty and / or quality in isolation ( a major gap in prior work ) . One limitation of the current work is that our correlational study design does not allow us to make strong causal claims . Relat - edly , because our data source was preexisting naturalistic behav - ioral traces online , we were not able to precisely isolate cognitive processes at a \ufb01ne - grained level , as one might be able to in the lab - oratory , or obtain psychological control measures ( e . g . , partici - pants\u2019 familiarity with inspirations ) . These are legitimate concerns , and to some degree are inherent tradeoffs of an in vivo vs . in vitro approach ( Dunbar & Blanchette , 2001 ) . However , three features of our study mitigate concerns about spurious statistical associations . First , our \ufb01ndings align well with prior theory and laboratory \ufb01ndings on the potential bene\ufb01ts of combination dis - tance . Indeed , the external validity from the in vivo approach strengthens our con\ufb01dence that the laboratory \ufb01ndings generalize to real - world creative cognition . Second , unlike some other obser - vational designs , our study does include a temporal asymmetry between the predictor and outcome variables ( we know that sources were built upon before shortlisting ) , which is a notable indicator of causal direction . Finally , our statistical analysis accounts for problem variation , contributor effects , and a variety of other important control variables , mitigating concerns over endogeneity . Nevertheless , future randomized experiments are necessary to fully establish causality . Additionally , some might be concerned that the unique context of the study \u2014 e . g . , Web - based context , focus on socially relevant problems \u2014 might limit generalizability to other creative prob - lems . This is a legitimate concern , given the recent controversy in 112 J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 the literature over the extent to which creative processes are domain - general or domain - speci\ufb01c ( Plucker , Beghetto , Sternberg , Grigorenko , & Singer , 2004 ; Simonton , 2009 ) . We have reason to believe that our results should generalize well to creativity in design - related domains ( e . g . , engineering / product / architectural design ) given the nature and diversity of problems in our sample . Solutions to these problems drew on a wide range of domain knowledge ( e . g . , public policy , human \u2013 computer - interaction , social and decision sciences , education ) and likely involved reasoning over multiple levels of systems ( e . g . , individual decision - making , communities ) . Thus , the cognitive processes and knowledge involved in generating concepts in our study are likely to have signi\ufb01cant overlap with other design - related domains . Nevertheless , we encourage further studies that explore how these \ufb01ndings might generalize to ( or be different in ) other forms of creative thought , such as artistic creativity and scienti\ufb01c discovery . Similarly , because of the public , Web - based context of the study , contributors were probably not posting every idea they had , regard - less of quality . This is a different dynamic from typical lab studies ( e . g . , of brainstorming ) where participants are asked to write down all ideas that come to mind . It is likely that in a less \ufb01ltered context , themoderatingeffectsofiterationmightbemuchmorepronounced ( but still be fundamentally similar ) : our \ufb01ndings suggest that 2 \u2013 4 iterations on a concept combined from distant concepts are suf\ufb01 - cient to make it creative , but more iterations may be required in more realistic settings where less self - \ufb01ltering is going on . 4 . 3 . Conclusions Returning to the overall question of the nature of human cre - ative cognition , our \ufb01ndings align with other studies and argu - ments that have highlighted the importance of iteration , broadly construed , in the creative process ( Chan & Schunn , 2015 ; Dow , Heddleston , & Klemmer , 2009 ; Mecca & Mumford , 2013 ; Nijstad et al . , 2010 ; Rietzschel et al . , 2007 ; Weisberg , 2011 ; Wilkenfeld & Ward , 2001 ; Yu & Nickerson , 2011 ) . This emerging body of work suggests that iteration provides pathways to not only higher qual - ity ideas ( Dow et al . , 2009 ; Yu & Nickerson , 2011 ) but also more novel ideas ( Nijstad et al . , 2010 ; Rietzschel et al . , 2007 ) , and is an advanced application of cognitive strategies like analogy ( Chan & Schunn , 2015 ) . Here , we add to this body of work , showing that iteration is a critical partner process in creative conceptual combi - nation . Again , we emphasize that this line of research concerns the importance of iteration for the development of ideas , not \ufb01nal products . The emerging picture is that good ideas rarely come in singular creative leaps , fully formed like Athena from the brow of Zeus ; instead , they more often come from the sweat of his brow building on the labors of others . Our \ufb01ndings are consistent with the predominant view of the optimal temporal ordering of divergence and convergence ( itera - tion ) in the creativity literature . Most theories of creativity posit a \u2018\u2018\ufb02are and focus\u201d view of creativity : diverge \ufb01rst , then converge . Similarly , many authors and practitioners warn of the dangers of premature solution selection ( e . g . , Brown & Katz , 2009 ; Vogel , Cagan , & Boatwright , 2005 ) . Berg ( 2014 ) showed in a series of ele - gant experiments that ideas that build \ufb01rst on very novel ideas that are then \u2018\u2018infused\u201d with more well - trodden idea components end up with a more optimal combination of novelty and quality than ideas that begin with well - trodden ideas and then are \u2018\u2018infused\u201d with novelty . While there is consensus in the literature on the optimal order of divergence and convergence , one might wonder about the optimal balance between divergence and convergence . In the pre - sent work , we observed a monotonically positive relationship between indirect combination distance and creativity . Insofar as far combinations increase divergence ( and \u2018\u2018variation\u201d in the variation - selection view of things ) , this relationship makes sense . However , this monotonically positive relationship stands in con - trast with some other work in the context of group brainstorming that has found no overall bene\ufb01t of increased divergence ( through nominal groups ) for the creativity of the \ufb01nal idea that is selected ( e . g . , Rietzschel , Nijstad , & Stroebe , 2010 ) . It might be fruitful to examine whether different divergent strategies ( e . g . , brainstorm - ing , distant conceptual combinations ) have different trajectories ( e . g . , non - monotonic , polynomial ) in terms of their effects on the \ufb01nal idea . A genealogical methodological approach ( as exempli\ufb01ed in the current study ) might be helpful for exploring this new space of questions . Acknowledgment This research was supported in part by a National Science Foun - dation # 1360013 Doctoral Dissertation Improvement Grant to the \ufb01rst author . Appendix A . OpenIDEO data additional details To give a better sense of the range of problems addressed on the platform , the following are the problem titles ( as seen by partici - pants ) for the 12 problems in the study sample : 1 . How might we increase the number of registered bone mar - row donors to help save more lives ? 2 . How might we inspire and enable communities to take more initiative in making their local environments better ? 3 . How can we manage e - waste & discarded electronics to safe - guard human health & protect our environment ? 4 . How might we better connect food production and consumption ? 5 . How can technology help people working to uphold human rights in the face of unlawful detention ? 6 . How might we identify and celebrate businesses that inno - vate for world bene\ufb01t and inspire other companies to do the same ? 7 . How might we use social business to improve health in low - income communities ? 8 . How might we increase social impact with OpenIDEO over the next year ? 9 . How might we restore vibrancy in cities and regions facing economic decline ? 10 . How might we design an accessible election experience for everyone ? 11 . How might we support web entrepreneurs in launching and growing sustainable global businesses ? 12 . How can we equip young people with the skills , information and opportunities to succeed in the world of work ? References Amabile , T . M . ( 1982 ) . Social psychology of creativity : A consensual assessment technique . Journal of Personality and Social Psychology , 43 ( 5 ) , 997 \u2013 1013 . Amabile , T . M . ( 1983 ) . The social psychology of creativity : A componential conceptualization . Journal of Personality and Social Psychology , 45 ( 2 ) , 357 . Atman , C . J . , Adams , R . S . , Cardella , M . E . , Turns , J . , Mosborg , S . , & Saleem , J . ( 2007 ) . Engineering design processes : A comparison of students and expert practitioners . Journal of Engineering Education , 96 , 359 \u2013 379 . Baer , J . , & McKool , S . S . ( 2009 ) . Assessing creativity using the consensual assessment technique . In C . S . Schreiner ( Ed . ) , Handbook of research on assessment technologies , methods , and applications in higher education ( pp . 65 \u2013 77 ) . PA : Hershey . Ball , L . J . , Evans , J . S . B . , Dennis , I . , & Ormerod , T . C . ( 1997 ) . Problem - solving strategies and expertise in engineering design . Thinking and Reasoning , 3 ( 4 ) , 247 \u2013 270 . http : / / doi . org / 10 . 1080 / 135467897394284 . J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 113 Basadur , M . , & Thompson , R . ( 1986 ) . Usefulness of the ideation principle of extended effort in real world professional and managerial creative problem solving \u2044 . The Journal of Creative Behavior , 20 ( 1 ) , 23 \u2013 34 . http : / / dx . doi . org / 10 . 1002 / j . 2162 - 6057 . 1986 . tb00414 . x . Bates , D . , Maechler , M . , Bolker , B . , & Walker , S . ( 2013 ) . lme4 : Linear mixed - effects models using Eigen and S4 . R package version 1 . 0 - 5 . < http : / / CRAN . R - project . org / package = lme4 > . Baughman , W . A . , & Mumford , M . D . ( 1995 ) . Process - analytic models of creative capacities : Operations in\ufb02uencing the combination - and - reorganization process . Creativity Research Journal , 8 ( 1 ) , 37 \u2013 62 . Benedek , M . , & Neubauer , A . C . ( 2013 ) . Revisiting Mednick\u2019s model on creativity - related differences in associative hierarchies . Evidence for a common path to uncommon thought . The Journal of Creative Behavior , 47 ( 4 ) , 273 \u2013 289 . http : / / dx . doi . org / 10 . 1002 / jocb . 35 . Berg , J . M . ( 2014 ) . The primal mark : How the beginning shapes the end in the development of creative ideas . Organizational Behavior and Human Decision Processes , 125 ( 1 ) , 1 \u2013 17 . http : / / dx . doi . org / 10 . 1016 / j . obhdp . 2014 . 06 . 001 . Blasko , V . J . , & Mokwa , M . P . ( 1986 ) . Creativity in advertising : A Janusian perspective . Journal of Advertising , 15 ( 4 ) , 43 \u2013 72 . Blei , D . M . , Ng , A . Y . , Jordan , M . I . , & Lafferty , J . ( 2003 ) . Latent Dirichlet allocation . Journal of Machine Learning Research , 993 \u2013 1022 . Boden , M . A . ( 2004 ) . The creative mind : Myths and mechanisms . New York . Brown , R . T . ( 1989 ) . Creativity : Whatarewetomeasure ? InJ . A . Glover , R . R . Ronning , & C . R . Reynolds ( Eds . ) , Handbook of Creativity ( pp . 3 \u2013 32 ) . New York , NY . Brown , T . , & Katz , B . ( 2009 ) . Change by design : How design thinking transforms organizations and inspires innovation . New York : Harper Business . Button , K . S . , Ioannidis , J . P . A . , Mokrysz , C . , Nosek , B . A . , Flint , J . , Robinson , E . S . J . , et al . ( 2013 ) . Power failure : Whysmall sample size undermines the reliability of neuroscience . Nature Reviews Neuroscience , 14 ( 5 ) , 365 \u2013 376 . http : / / dx . doi . org / 10 . 1038 / nrn3475 . Chan , J . ( 2014 ) . The impact of sources of inspiration on the genesis of creative ideas ( Doctoral Dissertation ) . University of Pittsburgh . < http : / / d - scholarship . pitt . edu / 22743 / > . Chan , J . , Dow , S . P . , & Schunn , C . D . ( 2015 ) . Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? Design Studies , 36 , 31 \u2013 58 . Chan , J . , & Schunn , C . ( 2015 ) . The impact of analogies on creative concept generation : Lessons from an in vivo study in engineering design . Cognitive Science , 39 ( 1 ) , 126 \u2013 155 . Doboli , A . , Umbarkar , A . , Subramanian , V . , & Doboli , S . ( 2014 ) . Two experimental studies on creative concept combinations in modular design of electronic embedded systems . Design Studies , 35 ( 1 ) , 80 \u2013 109 . http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 10 . 002 . Dow , S . P . , Heddleston , K . , & Klemmer , S . R . ( 2009 ) . Theef\ufb01cacy ofprototyping under time constraints . In Proceedings of the 7th ACM conference on creativity and cognition . Dunbar , K . N . ( 1997 ) . Howscientiststhink : On - linecreativityandconceptualchangein science . InT . B . Ward , S . M . Smith , & J . Vaid ( Eds . ) , Creativethought : Aninvestigation ofconceptualstructures andprocesses ( pp . 461 \u2013 493 ) . Washington D . C . Dunbar , K . N . , & Blanchette , I . ( 2001 ) . The in vivo / in vitro approach to cognition : The case of analogy . Trends in Cognitive Sciences , 5 ( 8 ) , 334 \u2013 339 . Finke , R . A . , Ward , T . B . , & Smith , S . M . ( 1996 ) . Creative cognition : Theory , research , and applications . Cambridge , MA : MIT Press . Gelman , A . , & Weakliem , D . ( 2009 ) . Of beauty , sex and power . American Scientist , 97 , 310 \u2013 316 . Gielnik , M . M . , Frese , M . , Graf , J . M . , & Kampschulte , A . ( 2011 ) . Creativity in the opportunity identi\ufb01cation process and the moderating effect of diversity of information . Journal of Business Venturing , 27 ( 5 ) , 559 \u2013 576 . http : / / dx . doi . org / 10 . 1016 / j . jbusvent . 2011 . 10 . 003 . Girotra , K . , Terwiesch , C . , & Ulrich , K . T . ( 2010 ) . Idea generation and the quality of the best idea . Management Science , 56 , 591 \u2013 605 . Goel , V . , & Pirolli , P . ( 1992 ) . The structure of design problem spaces . Cognitive Science , 16 , 395 \u2013 429 . Grif\ufb01ths , T . L . , & Steyvers , M . ( 2004 ) . Finding scienti\ufb01c topics . Proceedings of the National Academy of Sciences of the United States of America , 101 ( Suppl 1 ) , 5228 \u2013 5235 . http : / / dx . doi . org / 10 . 1073 / pnas . 0307752101 . Grif\ufb01ths , T . L . , Steyvers , M . , & Tenenbaum , J . B . ( 2007 ) . Topics in semantic representation . Psychological Review , 114 ( 2 ) , 211 \u2013 244 . http : / / dx . doi . org / 10 . 1037 / 0033 - 295X . 114 . 2 . 211 . Hampton , J . A . ( 1997 ) . Emergent attributes in combined concepts . In T . B . Ward , S . M . Smith , & J . Viad ( Eds . ) , Creative thought : An investigation of conceptual structures and processes ( pp . 83 \u2013 110 ) . Washington , DC : American Psychological Association . Hennessey , B . A . , & Amabile , T . M . ( 2010 ) . Creativity . Annual Review of Psychology , 61 , 569 \u2013 598 . Henrich , J . , Heine , S . J . , & Norenzayan , A . ( 2010 ) . The weirdest people in the world ? Behavioral and Brain Sciences , 33 , 1 \u2013 23 . Herring , S . R . , Jones , B . R . , & Bailey , B . P . ( 2009 ) . Idea generation techniques among creative professionals . In System sciences , 2009 . HICSS\u201909 . 42nd Hawaii international conference on ( pp . 1 \u2013 10 ) . IEEE . Howard - Jones , P . A . , Blakemore , S . - J . - . J . , Samuel , E . A . , Summers , I . R . , & Claxton , G . ( 2005 ) . Semantic divergence and creative story generation : An fMRI investigation . Cognitive Brain Research , 25 ( 1 ) , 240 \u2013 250 . Impact Stories . ( n . d . ) . < http : / / www . openideo . com / content / impact - stories > . Jang , S . ( 2014 ) . The effect of image stimulus on conceptual combination in the design idea generation process . Archivesof Design Research , 112 ( 4 ) , 59 . http : / / dx . doi . org / 10 . 15187 / adr . 2014 . 11 . 112 . 4 . 59 . Jin , Y . , & Chusilp , P . ( 2006 ) . Study of mental iteration in different design situations . Design Studies , 27 ( 1 ) , 25 \u2013 55 . Johansson , F . ( 2006 ) . The Medici effect : What elephants and epidemics can teach us about innovation . Boston , MA : Harvard Business Press . Koestler , A . ( 1964 ) . The act of creation . Oxford , England : Macmillan . Kohn , N . W . , Paulus , P . B . , & Choi , Y . ( 2011 ) . Building on the ideas of others : An examination of the idea combination process . Journal of Experimental Social Psychology , 47 ( 3 ) , 554 \u2013 561 . Krynicki , F . ( 2014 ) . Methods and models for quantitative analysis of crowd brainstorming . University of Waterloo . Kunda , Z . , Miller , D . T . , & Claire , T . ( 1990 ) . Combining social concepts : The role of causal reasoning . Cognitive Science , 14 ( 4 ) , 551 \u2013 577 . http : / / dx . doi . org / 10 . 1207 / s15516709cog1404 _ 3 . March , J . G . ( 1991 ) . Exploration and exploitation in organizational learning . Organization Science , 2 ( 1 ) , 71 \u2013 87 . Mecca , J . T . , & Mumford , M . D . ( 2013 ) . Imitation and creativity : Bene\ufb01cial effects of propulsion strategies and speci\ufb01city . The Journal of Creative Behavior . http : / / dx . doi . org / 10 . 1002 / jocb . 49 . Mednick , S . A . ( 1962 ) . The associative basis of the creative process . Psychological Review , 69 ( 3 ) , 220 \u2013 232 . Mobley , M . I . , Doares , L . M . , & Mumford , M . D . ( 1992 ) . Process analytic models of creative capacities : Evidence for the combination and reorganization process . Creativity Research Journal , 5 ( 2 ) , 125 \u2013 155 . Mumford , M . D . , Baughman , W . A . , Maher , M . A . , Costanza , D . P . , & Supinski , E . P . ( 1997 ) . Process - based measures of creative problem - solving skills : IV . Category combination . Creativity Research Journal , 59 . Nagai , Y . , Taura , T . , & Mukai , F . ( 2009 ) . Concept blending and dissimilarity : Factors for creative concept generation process . Design Studies , 30 ( 6 ) , 648 \u2013 675 . Nijstad , B . A . , De Dreu , C . K . W . , Rietzschel , E . F . , & Baas , M . ( 2010 ) . The dual pathway to creativitymodel : Creative ideation as a function of \ufb02exibility and persistence . European Review of Social Psychology , 21 , 34 \u2013 77 . http : / / dx . doi . org / 10 . 1080 / 10463281003765323 . Nijstad , B . A . , & Stroebe , W . ( 2006 ) . How the group affects the mind : A cognitive model of idea generation in groups . Personality and Social Psychology Review , 10 ( 3 ) , 186 \u2013 213 . http : / / dx . doi . org / 10 . 1207 / s15327957pspr1003 _ 1 . Parnes , S . J . ( 1961 ) . Effects of extended effort in creative problem solving . Journal of Educational Psychology , 52 ( 3 ) , 117 \u2013 122 . Parnes , S . J . , & Meadow , A . ( 1959 ) . Effects of \u2018\u2018brainstorming\u201d instructions on creative problem solving by trained and untrained subjects . Journal of Educational Psychology , 50 ( 4 ) , 171 \u2013 176 . Paulus , P . B . , Kohn , N . W . , Arditti , L . E . , & Korde , R . M . ( 2013 ) . Understanding the group size effect in electronic brainstorming . Small Group Research , 44 ( 3 ) , 332 \u2013 352 . http : / / dx . doi . org / 10 . 1177 / 1046496413479674 . Perkins , D . N . ( 1983 ) . Novel remote analogies seldom contribute to discovery . The Journal of Creative Behavior , 17 ( 4 ) , 223 \u2013 239 . Pinheiro , J . C . , & Bates , D . M . ( 2000 ) . Linear mixed - effects models : Basic concepts and examples . Springer . Plucker , J . A . , Beghetto , R . A . , Sternberg , R . J . , Grigorenko , E . L . , & Singer , J . L . ( 2004 ) . Why creativity is domain general , why it looks domain speci\ufb01c , and why the distinction does not matter . In Creativity : From potential to realization ( pp . 153 \u2013 167 ) . Washington , DC , US : American Psychological Association . R Core Team ( Ed . ) . ( 2013 ) . R : A language and environment for statistical computing . Vienna , Austria : R Foundation for Statistical Computing . < http : / / www . R - project . org / > . Rhodes , T . , & Turvey , M . T . ( 2007 ) . Human memory retrieval as L\u00e9vy foraging . Physica A : Statistical Mechanics and Its Applications , 385 ( 1 ) , 255 \u2013 260 . Rietzschel , E . F . , Nijstad , B . A . , & Stroebe , W . ( 2007 ) . Relative accessibility of domain knowledge and creativity : The effects of knowledge activation on the quantity and originality of generated ideas . Journal of Experimental Social Psychology , 43 ( 6 ) , 933 \u2013 946 . Rietzschel , E . F . , Nijstad , B . A . , & Stroebe , W . ( 2010 ) . The selection of creative ideas after individual idea generation : Choosing between creativity and impact . British Journal of Psychology , 101 ( 1 ) , 47 \u2013 68 . http : / / dx . doi . org / 10 . 1348 / 000712609X414204 . Rothenberg , A . ( 1979 ) . The emerging goddess : The creative process in art , science , and other \ufb01elds . Chicago : University of Chicago Press . Rothenberg , A . ( 1995 ) . Creative cognitive processes in Kekule\u2019s discovery of the structure of the benzene molecule . The American Journal of Psychology , 419 \u2013 438 . Runco , M . A . ( 2004 ) . Creativity . Annual Review of Psychology , 55 , 657 \u2013 687 . Sawyer , R . K . ( 2012 ) . Explaining creativity : The science of human innovation ( 2nd ed . ) . New York : Oxford University Press . Schwartz , A . H . , Eichstaedt , J . C . , Kern , M . L . , Dziurzynski , L . , Ramones , S . M . , Agrawal , M . , . . . Ungar , L . H . ( 2013 ) . Personality , gender , and age in the language of social media : The open - vocabulary approach . PLoS ONE , 8 ( 9 ) , e73791 . http : / / dx . doi . org / 10 . 1371 / journal . pone . 0073791 . Shah , J . J . , Vargas - Hernandez , N . , & Smith , S . M . ( 2003 ) . Metrics for measuring ideation effectiveness . Design Studies , 24 ( 2 ) , 111 \u2013 134 . Shih , P . C . , Venolia , G . , & Olson , G . M . ( 2011 ) . Brainstorming under constraints : Why software developers brainstorm in groups . In Proceedings of the 25th BCS conference on human - computer interaction ( pp . 74 \u2013 83 ) . Swinton , UK , UK : British Computer Society . < http : / / dl . acm . org / citation . cfm ? id = 2305316 . 2305331 > . Siangliulue , K . , Arnold , K . C . , Gajos , K . Z . , & Dow , S . P . ( 2015 ) . Toward collaborative ideation at scale : Leveraging ideas from others to generate more creative and diverse ideas . In Proceedings of CSCW\u201915 . 114 J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 Simonton , D . K . ( 2009 ) . Varieties of ( scienti\ufb01c ) creativity : A hierarchical model of domain - speci\ufb01c disposition , development , and achievement . Perspectives on Psychological Science , 4 ( 5 ) , 441 \u2013 452 . Simonton , D . K . ( 2011 ) . Creativity and discovery as blind variation : Campbell\u2019s ( 1960 ) BVSR model after the half - century mark . Review of General Psychology , 15 ( 2 ) . Simonton , D . K . ( 2012 ) . Creativity , problem solving , and solution set sightedness : Radically reformulating BVSR . The Journal of Creative Behavior , 46 ( 1 ) , 48 \u2013 65 . Terwiesch , C . , & Ulrich , K . T . ( 2009 ) . Innovation tournaments : Creating and selecting exceptional opportunities . Boston , MA : Harvard Business Press . Vogel , C . M . , Cagan , J . , & Boatwright , P . B . H . ( 2005 ) . The design of things to come : How ordinary people create extraordinary products . NJ : Upper Saddle River . Wallas , G . ( Eds . ) . ( 1926 ) . The art of thought . New York , NY . Ward , T . B . ( 2001 ) . Creative cognition , conceptual combination , and the creative writing of Stephen R . Donaldson . The American Psychologist , 56 ( 4 ) , 350 \u2013 354 . Warr , A . , & O\u2019Neill , E . ( 2005 ) . Understanding design as a social creative process . In Proceedings of the 5th conference on creativity and cognition ( pp . 118 \u2013 127 ) . New York , NY , USA : ACM . http : / / dx . doi . org / 10 . 1145 / 1056224 . 1056242 . Weisberg , R . W . ( 2009 ) . On \u2018\u2018out - of - the - box\u201d thinking in creativity . In A . B . Markman & K . L . Wood ( Eds . ) , Tools for innovation ( pp . 23 \u2013 47 ) . New York , NY . Weisberg , R . W . ( 2011 ) . Frank Lloyd Wright\u2019s Fallingwater : A case study in inside - the - box creativity . Creativity Research Journal , 23 ( 4 ) , 296 \u2013 312 . http : / / dx . doi . org / 10 . 1080 / 10400419 . 2011 . 621814 . Wilkenfeld , M . J . ( 1995 ) . Conceptual combinations : Does similarity predict emergence ? Wilkenfeld , M . J . , & Ward , T . B . ( 2001 ) . Similarity and emergence in conceptual combination . Journal of Memory and Language , 45 ( 1 ) , 21 \u2013 38 . Wisniewski , E . J . ( 1997 ) . Conceptual combination : Possibilities and esthetics . In T . B . Ward , S . M . Smith , & J . Viad ( Eds . ) , Creative thought : An investigation of conceptual structures and processes ( pp . 51 \u2013 81 ) . Washington , DC : American Psychological Association . Yu , L . , & Nickerson , J . V . ( 2011 ) . Cooks or cobblers ? Crowd creativity through combination . In Proceedings of the SIGCHI conference on human factors in computing systems ( pp . 1393 \u2013 1402 ) . ACM . Zeng , L . , Proctor , R . W . , & Salvendy , G . ( 2011 ) . Fostering creativity in product and service development : Validation in the domain of information technology . Human Factors , 53 ( 3 ) , 245 \u2013 270 . J . Chan , C . D . Schunn / Cognition 145 ( 2015 ) 104 \u2013 115 115",
    "kangAugmentingScientificCreativity2022": "57 Augmenting Scientific Creativity with an Analogical Search Engine HYEONSU B . KANG , Carnegie Mellon University , USA XIN QIAN , University of Maryland , College Park , USA TOM HOPE , Allen Institute for AI and The University of Washington , USA DAFNA SHAHAF , Hebrew University of Jerusalem , Israel JOEL CHAN , University of Maryland , College Park , USA ANIKET KITTUR , Carnegie Mellon University , USA Analogies have been central to creative problem - solving throughout the history of science and technology . As the number of scientific articles continues to increase exponentially , there is a growing opportunity for finding diverse solutions to existing problems . However , realizing this potential requires the development of a means for searching through a large corpus that goes beyond surface matches and simple keywords . Here we contribute the first end - to - end system for analogical search on scientific articles and evaluate its effectiveness with scientists\u2019 own problems . Using a human - in - the - loop AI system as a probe we find that our system facilitates creative ideation , and that ideation success is mediated by an intermediate level of matching on the problem abstraction ( i . e . , high versus low ) . We also demonstrate a fully automated AI search engine that achieves a similar accuracy with the human - in - the - loop system . We conclude with design implications for enabling automated analogical inspiration engines to accelerate scientific innovation . CCS Concepts : \u2022 Human - centered computing \u2192 Interactive systems and tools ; Empirical studies in HCI ; Additional Key Words and Phrases : Computational analogies , innovation , scientist users , interactive analog - ical search engine , sequence - to - sequence modeling , word embeddings , think - aloud studies ACM Reference format : Hyeonsu B . Kang , Xin Qian , Tom Hope , Dafna Shahaf , Joel Chan , and Aniket Kittur . 2022 . Augmenting Sci - entific Creativity with an Analogical Search Engine . ACM Trans . Comput . - Hum . Interact . 29 , 6 , Article 57 ( November 2022 ) , 36 pages . https : / / doi . org / 10 . 1145 / 3530013 This work was supported by Center for Knowledge Acceleration , National Science Foundation ( FW - HTF - RL , grant no . 1928631 ; IIS , grant no . 1816242 ; SHF , grant no . 1814826 ) , the European Research Council ( ERC ) under the European Union\u2019s Horizon 2020 research and innovation programme ( grant no . 852686 , SIAM ) and NSF - BSF grant no . 2017741 . This work is also based upon work supported by the Google Cloud Research Credits program with the award GCP19980904 . Authors\u2019 addresses : H . B . Kang and A . Kittur , Carnegie Mellon University , 5000 Forbes Ave , Pittsburgh , PA 15213 ; emails : { hyeonsuk , nkittur } @ cs . cmu . edu ; X . Qian and J . Chan , University of Maryland , College Park , MD 20742 ; emails : { xinq , joelchan } @ umd . edu ; T . Hope , Allen Institute for AI and The University of Washington , Seattle , WA 98103 ; email : tomh @ allenai . org ; D . Shahaf , Hebrew University of Jerusalem , Israel ; email : dshahaf @ cs . huji . ac . il . This work is licensed under a Creative Commons Attribution International 4 . 0 License . \u00a9 2022 Copyright held by the owner / author ( s ) . 1073 - 0516 / 2022 / 11 - ART57 https : / / doi . org / 10 . 1145 / 3530013 ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 2 H . B . Kang et al . 1 INTRODUCTION Analogical reasoning has been central to creative problem solving throughout the history of sci - ence and technology [ 32 , 43 , 50 , 54 , 60 , 86 ] . Many important scientific discoveries were driven by analogies : the Greek philosopher Chrysippus made a connection between observable water waves and sound waves ; an analogy between bacteria and slot machines helped Salvador Luria advance the theory of bacterial mutation ; a pioneering chemist Joseph Priestly suggested charges attract or repel each other with an inverse square force by an analogy to gravity . Today the potential for finding analogies to accelerate innovation in science and engineering is greater than ever before . As of 2009 fifty million scientific articles had been published , and the number continues to grow at an exceedingly fast rate [ 12 , 28 , 68 , 85 ] . These articles represent a potential treasure trove for finding inspirations from distant domains and generating creative solutions to challenging problems . However , searching analogical inspirations in a large corpus of articles remains a longstand - ing challenge [ 34 , 44 , 83 , 99 ] . Previous systems for retrieving analogies have largely focused on modeling analogical relations in non - scientific domains and / or in limited scopes ( e . g . , structure - mapping [ 36 \u2013 38 , 42 , 106 ] , multiconstraint - based [ 33 , 59 , 65 ] , connectionist [ 57 ] , rule - based reason - ing [ 3 , 15 , 16 , 110 ] systems ) , and the prohibitive costs of creating highly structured representations prevented hand - crafted systems ( e . g . , DANE [ 65 , 109 ] ) from having a broad coverage of topics and being deployed for realistic use . Conversely , scalable computational approaches such as keyword or citation based search engines have been limited by a dependence on surface or domain similar - ity . Such search engines aim at maximizing similarity to the query which is useful when trying to know what has been done on the problem in the target domain but less useful when trying to find inspiration outside that domain ( for example , for Salvador Luria\u2019s queries : \u201chow do bacteria mu - tate ? \u201d or \u201cwhy are bacterial mutation rates so inconsistent ? \u201d , similarity maximizing search engines may have found Luria and Delbr\u00fcck\u2019s earlier work on E . coli [ 81 ] but may have failed to recognize more distant sources of inspiration such as slot machines as relevant ) . Recently a novel idea for analogical search was introduced [ 61 ] . In this idea what would other - wise be a complex analogical relation between products is pared down to just two components : purpose ( what problem does it solve ? ) and mechanism ( how does it solve that problem ? ) . Once many such purpose and mechanism pairs are identified , products that solve a similar problem to the query but using diverse mechanisms are searched to help broaden the searcher\u2019s perspective on the problem and boost their creativity for coming up with novel mechanism ideas . Anecdotal ev - idence suggests that this approach may also be applicable to the domain of scientific research . For example , while building lighter and more compact solar panel arrays has been a longstanding challenge for NASA scientists , recognizing how the ancient art form of origami may be applied to create folding structures led to an innovation to use compliant mechanisms to build not just com - pact but also self - deployable solar arrays [ 27 , 89 , 118 ] ( diagrammatically shown in Figure 1 ) . The first remaining challenge of analogical search in the scholarly domain is how we might represent scientific articles as purpose and mechanism pairs at scale and search for those that solve sim - ilar purposes using different mechanisms . Recent advances in natural language processing have demonstrated that neural networks that use pre - trained embeddings to encode input text can offer a promising technique to address it . Pre - trained embeddings are real - valued vectors that represent tokens ( Tokenization means breaking a piece of text into smaller units ; Tokens can be words , char - acters , sub - words , or n - grams . ) , in a high - dimensional space ( e . g . , typically dimensions of a few dozens to a few thousands ) and are shown to capture rich , multifaceted semantic relations between words [ 8 , 100 ] . Leveraging them , neural networks may be trained to identify purposes and mech - anisms from text [ 61 , 62 ] to enable search - by - analogy ( i . e . , different mechanisms used for similar purposes ) . Once candidate articles are retrieved , searchers may use them to come up with novel ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 3 Fig . 1 . A diagram of two different yet analogical approaches ( dashed arrow ) for building lighter and more compact solar arrays , and their representations in purposes and mechanisms . classes of mechanisms or apply them directly to their own research problems to improve upon the current state . Prior studies in product ideation showed that users of analogical search systems could engage with the results to engender more novel and relevant ideas [ 21 , 48 , 74 ] . Here , we study the remaining open questions as to whether such findings also generalize to the scientific domains of innovation and how they may differ . In this article , we present a functioning prototype of an analogical search engine for scientific articles at scale and investigate how such a system can help users explore and adapt distant inspi - rations . In doing so our system moves beyond manually curated approaches that have limited data ( e . g . , crowdsourced annotations in [ 21 ] with \u223c 2 , 000 articles ) and machine learning approaches that have been limited to simple product descriptions [ 48 , 61 , 62 ] . Using the prototypical system , we explore how it enables scientists to interactively search for inspirations for their personalized re - search problems in a large ( \u223c 1 . 7 M ) article corpus . We investigate whether scientists can recognize mapping of analogical relations between the results returned from our analogical search engine and their query problems , and use them to come up with novel ideas . The scale of our corpus al - lows us to probe realistic issues including noise , error , and scale as well as how scientists react to a search engine that does not aim at providing only the most similar results to their query . In order to accomplish these goals we describe how we address several technical issues in the design of an interactive - speed analogical search engine , ranging from developing a machine learn - ing model for extracting purposes and mechanisms in scientific text at a token level granularity , the pipeline for constructing a similarity space of purpose embeddings , and enabling these embed - dings to be queried at interactive speeds by end users through a search interface . We construct the similarity space by putting semantically related purpose embeddings in close indices from each other such that related purposes can be searched at scale . In addition to the technical challenges there are several important questions around the design of analogical search engines that we explore here . A core conceptual difference that distinguishes analogical search engines from other kinds is that the analogs they find for a search query need to maintain some kind of distance from the query , rather than simply maximizing the similarity with it . However , only certain kinds of distance may support generative ideation while others have a detrimental effect . Another question remains as to how much distance is appropriate when it comes to finding analogical inspirations in other domains . While landmark studies of analogical innovation suggest that highly distant domains can provide particularly novel or transformative innovations [ 46 , 47 , 55 ] , recent work suggests the question may be more nuanced and that inter - mediate levels of distance may be fruitful for finding ideas that are close enough to be relevant but sufficiently distant to be unfamiliar and spur creative adaptation [ 22 , 39 , 49 ] . Using a concrete ex - ample from one of our participants who studied ways to facilitate heat transfer in semiconductors , a keyword search engine might find commonly used mechanisms appropriate for direct applica - tion ( e . g . , tweaking the composition of the material ) while an analogical search engine might find ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 4 H . B . Kang et al . similar problems in more distant domains which suggest mechanisms that inspire creative adap - tation ( e . g . , nanoscale fins that absorb heat and convert it to mechanical energy ) . Though more distant conceptual combinations may not always lead to immediately feasible or useful ideas , they may result in outsized value after being iterated on [ 9 , 23 , 75 ] . In the following sections , we explore the technical and design challenges for an analogical search engine and how users interact with such a system . First , we describe the development of a human - in - the - loop search engine prototype , in which most elements of the system are functional but human screeners are used to remove obvious noise from the end results in order to maximize our ability to probe how users interact with potentially useful analogical inspirations . Using this prototype we characterize how researchers searching for inspirations for their own problems gain the most benefit from articles that partially match their problem ( i . e . , match at a high level purpose but mismatch at a lower level specifications of the purpose ) , and that the benefits are driven not by direct application of the ideas in the article but by creative adaptation of those ideas to their target domain . Subsequently we describe improvements to the system to enable a fully automated , interactive - speed prototype and case studies with researchers using the system in a realistic way involving reformulation of their queries and self - driven attention to the results . We synthesize the findings of the two studies into design implications for next - generation analogical search engines . Through extensive in - depth evaluations using an ideation think - aloud protocol [ 35 , 107 ] with PhD - level researchers working on their own problems , we evaluate the degree to which inspira - tions spark creative adaptation ideas in a realistic way on scientists\u2019 own research problems . Unlike previous work which has often used undergraduate students in the classroom or lab [ 109 ] , and of - ten evaluated systems on predetermined problems [ 40 ] , this study design provides our evaluation with a high degree of external validity and allows us to deeply understand the ways in which en - countering our results can engender new ideas . Our final , automated search engine demonstrates how the human - in - the - loop filtering can be removed while achieving a similar accuracy . We con - clude with the benefits , design challenges , and opportunities for future analogical search engines from case studies with several researchers . To encourage innovation in this domain , we release our corpus of purpose and mechanism embeddings . 1 2 SYSTEM DESIGN The design of our analogical search engine for scientific articles involves three main system re - quirements . First , a computational pipeline for automatically identifying purposes ( what problems does it solve ? ) and mechanisms ( how does it solve those problems ) at scale ( e . g . , millions of arti - cles ) , in a token - level granularity from scientific abstracts . Second , an efficient retrieval algorithm for incorporating the identified purpose and mechanism texts into the system to enable search - by - analogy ( i . e . , article abstracts that contain similar purposes to a query problem but different mechanisms ) . Third , end - user interactivity for querying problems of interest ( e . g . , \u201ctransfer heat in semiconductors , \u201d \u201cgrow plants using nanoparticle fertilizers\u201d ) . We describe the system design in detail in the following subsections . 2 . 1 Stage One . Training Seq2Seq Models for Identifying Purpose and Mechanism Tokens 2 . 1 . 1 Overview of Modeling . In the first stage of the system , purpose and mechanism tokens are identified from article abstracts ( Figure 2 , 1 (cid:4) ) . Research article abstracts often include descriptions of the most important purpose or the core problem addressed in an article and the proposed mecha - nism or the approach taken to address the problem , making them good candidates for identification 1 https : / / github . com / hyeonsuukang / augmenting _ tochi22 . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 5 Fig . 2 . Components of our system design that address the three core challenges . 1 (cid:4) Purpose and mechanism tokens are extracted from article abstracts at scale . We develop Seq2Seq classifiers to classify tokens into purpose , mechanism , or neither , going beyond previous approaches that worked on sentences or relied on crowds . 2 (cid:4) We embed the extracted purpose texts using a pre - trained language model ( Google\u2019s USE [ 20 ] ) and train a tree - based index of vectors to place high semantic similarity vectors in close neighborhoods for efficient lookup . 3 (cid:4) When the user query arrives at the system , it is first embedded with USE . This query embedding is then used to lookup the pre - computed tree indices for high similarity purpose embeddings . Article abstracts for the corresponding purpose embeddings are retrieved from Google Datastore . In the first system , additional human filtering is performed to remove obviously irrelevant results that may have been included due to model errors . Finally , a set of articles with similar purposes to the query but different mechanisms are returned to the users for ideation . and extraction of tokens corresponding to them . For example , for a similar problem of facilitating heat transfer , Article A may propose an approach that modifies the structure of the material used at the interface between crystalline silicon ( semiconductor material ) and the substrate , while Ar - ticle B may propose a more distant mechanism ( due to the mismatch on scale ) of fin - based heat sinks commonly used for electronic devices . The goal of this first stage is to automatically identify and extract tokens that correspond to the similar purpose ( e . g . , \u201cfacilitate heat transfer\u201d ) as well as the mechanisms ( e . g . , \u201cmodifying the structure of the material used at the interface between crystalline silicon\u201d vs . \u201cfin - based heat sinks\u201d ) from the abstracts A and B . One relevant automated approach for identifying purposes and mechanisms from scientific ab - stracts is DISA [ 63 ] , which formulates the task as supervised sentence classification . However , we found that many key sentences in abstracts include both purpose and mechanism , breaking the assumptions of a sentence - level classifier ( e . g . , \u201cIn this article , [ a wavelet transforms - based method ] for [ filtering noise from images ] is presented . \u201d ) . To overcome this limitation we follow [ 62 ] and frame purpose and mechanism identification as a sequence - to - sequence ( Seq2Seq ) learning task [ 5 , 101 ] and develop deep neural networks with inductive biases capable of learning token - level patterns in the training dataset . Our dataset consists of crowdsourced annotations from Chan et al . ( the dataset is constructed via application of [ 21 ] to a larger corpus of around 2 , 000 article abstracts largely in computer science domains ) ( Table 1 ) . We train the models to classify input features ( tokens or spans of tokens ) as either purpose ( PP ) , mechanism ( MN ) or neither . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 6 H . B . Kang et al . Table 1 . Summary Statistics of the Training and Validation Datasets : The Number of Purpose ( PP ) and Mechanism ( MN ) Tokens , the Number and avg . Token Length of Article Abstracts Kind ( # of articles ) Avg . length # of PP # of MN Train ( 2021 ) 196 65 , 261 120 , 586 Validation ( 50 ) 170 1 , 510 1 , 988 We train two deep neural networks ( Models 1 and 2 ) , achieving increasing accuracy of classifi - cation . The first model is based on a Bi - directional LSTM ( BiLSTM ) architecture for sequence tagging [ 56 , 64 ] , in which the forward ( the beginning of the sequence to the end ) and the backward passes condition each token position in the text with its left and right context , respectively . A main source of improvement of Model 2 over Model 1 is the ability to more selectively attend to infor - mative tokens in a sentence rather than treating each token in a sequence as independent of each other ( as a hypothetical example , an extremely effective model based on this approach may assign more weights to the tokens \u201cselectively attend to informative tokens\u201d , as they represent the core mechanism described in the previous sentence ) and to leverage the regularities of co - occurrence with surrounding words through the self - attention mechanism [ 108 ] . 2 . 1 . 2 Seq2Seq Model Implementation Details . We implement the BiLSTM architecture of Model 1 in PyTorch [ 87 ] . We use pre - trained GloVe [ 88 ] word embeddings with 300 dimensions , consistent with prior work [ 11 , 78 , 88 ] to represent each token in the sequence as 300 - dimensional input vectors for the model . We train the model with a cross entropy loss objective for per - token classification in the three ( PP , MN , Neither ) token classes . For Model 2 , we adapt the SpanRel [ 67 ] architecture and implement it on AllenNLP [ 41 ] . We im - plement a self attention mechanism that tunes weights for the core word in each span as well as the boundary words that distinguish the context of use , consistent with [ 79 ] . We use the pre - trained ELMo 5 . 5B [ 90 ] embeddings for token representation following the near state - of - the - art perfor - mance reported in [ 67 ] on the scientific Wet Lab Protocol dataset . We train the model using a simi - lar procedure as Model 1 . We leave detailed training parameters for Models 1 and 2 to the Appendix . 2 . 1 . 3 Introducing Human - in - the - loop Filtering for Model 1 . The final classification performance ( F1 - scores ) of Model 1 on the validation set is 0 . 509 ( Purpose ) , 0 . 497 ( Mechanism ) , and 0 . 801 ( nei - ther ) . We found that the limited accuracy contributed to how the system retrieves irrelevant search results . Because reactions to obviously irrelevant results are not useful , we added a human - in - the - loop [ 31 ] filtering stage . The filtering proceeded as follows : members from the research team in - putted problem queries received from study participants into the system . Once the model produced matches , they went over from the top of the sorted list and removed only those that are irrelevant to the problem context . They continued filtering until at least 30 articles with reasonable purpose similarity were collected . After Winsorizing at top and bottom 10 % [ 115 ] , the human filterers re - viewed 45 articles per query ( SD : 27 . 6 , min : 6 , max : 138 ) for 5 queries ( SD : 2 . 4 , min : 2 , max : 9 ) to collect 33 ( SD : 3 . 5 , min : 30 , max : 40 ) purpose - similar articles ( about 12 / 45 = 26 % error rate ) . In Study 1 we show that the limited retrieval accuracy of Model 1 is sufficient for use as a probe with this additional human - in - the - loop filtering . In Study 2 and case studies , we demonstrate how this filtering can be removed with Model 2 while achieving a similar accuracy . 2 . 1 . 4 Scaling Model Inference . In order to have sufficient coverage to return diverse results , we collected an initial corpus of 2 . 8 million research articles from Springer Nature . 2 After 2 https : / / dev . springernature . com / . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 7 Table 2 . Corpus Used in the Deployed Search Engine and its Topical Distribution : Computer Science ( CS ) , Engineering ( Eng ) , Biomedicine ( BioMed ) , and Business and Engineering ( B & Eng ) Domain CS Eng BioMed B & Eng Total Count 675 K 568 K 336 K 145 K 1 . 7 M deduplication ( based on Digital Object Identifier using BigQuery 3 ) and filtering only articles with at least 50 characters in the abstract we were left with 1 . 7 million articles in four subjects ( Table 2 ) . We stored the resulting corpus in Google Cloud storage buckets . 4 To scale the classification of the Seq2Seq models we used the Apache Beam API 5 on Google Cloud Dataflow 6 to parallelize the operation . 2 . 2 Stage Two . Constructing a Purpose Similarity Space 2 . 2 . 1 Overview . In the second stage , the identified purpose texts are incorporated into the sys - tem to enable search - by - analogy of articles that solve similar problems using different mechanisms , at an interactive speed ( Figure 2 , 2 (cid:4) ) . Relevant previous approaches include Hope et al . [ 61 ] which first clusters similar purposes ( through k - means with pruning ) and subsequently samples within each cluster of similar purposes to maximize the diversity of mechanisms ( via a GMM approxi - mation algorithm [ 92 ] ) , or [ 62 ] which employs similarity metrics to balance the similarity to a purpose query and the distance to a mechanism query ( and vice versa ) . In contrast , from pilot tests in our corpus we discovered that even close purpose matches of scientific articles already had high variance in terms of the mechanisms they propose . We hypothesize that this may be the case due to the enormous span of possible research topics and the relative sparseness of their coverage in our corpus , and / or due to the emphasis on novelty in scientific research that discourages future articles which might contribute relatively small variations to an existing mechanism . We leave ex - ploration of these hypotheses for future work and simplify our sampling of the scientific articles to the one based solely on the similarity of purpose , sufficient for ensuring diversity . In order to support fast retrieval ( e . g . , sub - second response time ) of articles with similar purposes at scale ( e . g . , millions of articles ) , we pre - train Spotify\u2019s Annoy 7 indices of nearest neighboring pur - poses . Annoy trains a neural network to assign an embedding vector corresponding to a purpose to an index in the high - dimensional space that brings it close to other indices of purpose vectors that have similar meaning ( see Section 2 . 2 . 3 for details of the metric used for the similarity of mean - ing ) . Annoy uses random projection and tree - building ( see [ 1 , 2 ] ) to create read - only , file - based indices . Because it decouples creation of the static index files from lookup , it enables efficient and flexible search by utilizing many parallel processes to quickly load and map indices into memory . 2 . 2 . 2 Interactive Speed . Additionally Annoy minimizes its memory footprint in the process . This efficiency , critical for real - time applications such as ours , was further validated during our test of the end - to - end latency on the Web , with the average response taking 2 . 4 s ( SD = 0 . 56 s ) . 8 The level of latency we observed was sufficiently low to enable interactive search by end users ( both human - in - the - loop filterers in Study One and researcher participants in case studies ) . 3 https : / / cloud . google . com / bigquery . 4 https : / / cloud . google . com / storage . 5 https : / / beam . apache . org / . 6 https : / / cloud . google . com / dataflow / . 7 https : / / github . com / spotify / annoy . 8 We tested with 20 topically varied search queries that have not previously been entered to the engine to test the latency end - users experience and to exclude the effect of caching from it . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 8 H . B . Kang et al . 2 . 2 . 3 Implementation Details . To construct the similarity space , we first encode the purpose texts into high - dimensional embedding vectors which then can be used to compute pairwise se - mantic similarity . Here , the choice of an encoding algorithm depends on three main constraints . First , the pairwise similarity , when computed , should correlate well with the human - judged se - mantic similarity between the purposes . Second , similarity calculation between varying lengths of texts should be possible because extracted purposes can differ in length . Third , computationally efficient methods are preferred for scaling . To meet these requirements , we chose Universal Sentence Encoder ( USE ) 9 to encode purposes into fixed 512 - dimensional vectors . USE trains a transformer architecture [ 108 ] on a large corpus of both unsupervised ( e . g . , Wikipedia ) and su - pervised ( e . g . , Stanford Natural Language Inference dataset [ 13 ] ) data to produce a neural network that can encode text into vectors that meaningfully correlate with human judgment ( e . g . , evaluated on the semantic textual similarity benchmark [ 19 ] ) . USE can handle texts of varying lengths ( e . g . , from short phrases to sentences to paragraphs ) , and with high efficiency [ 20 ] , thereby making it suitable for our system . We pre - compute pairwise similarity of the purpose embeddings and store the indices in neigh - borhoods of high similarity for fast retrieval of similar purposes . As mentioned before , we train the Annoy indices on Google Cloud AI Platform . 10 We use 1\u2014the Euclidean distance of normalized vectors ( i . e . , given two vectors u and v , distance ( u , v ) = (cid:2) ( 2 ( 1 \u2212 cos ( u , v ) ) ) ) as a similarity metric ( using a Euclidean distance based metric for nearest neighbor clustering shows good performance , see [ 4 ] for a related discussion on the impact of the distance metric on the retrieval performance ) . We set the hyper - parameter k specifying the number of trees in the forest to 100 ( larger k \u2019s result in more accurate results but also decreases performance ; see [ 2 ] for further details ) . Empirically , 100 seemed to strike a good balance between the precision - performance tradeoff , thus we did not experiment with this parameter further . 2 . 3 Stage Three . Retrieving the Results In the last stage , the front - end interface interacts with end users and receives problem queries . These queries are then relayed to the back - end for retrieval of articles that solve similar problems using different mechanisms . The retrieved articles are presented on the front - end for users to review ( Figure 2 , 3 (cid:4) ) . When a user query is received , the back - end first encodes it using the same encoding algorithm used as the construction method of the purpose similarity space ( i . e . , USE ) . Using this query embedding , the back - end searches the pre - trained similarity space for articles with similar purposes . The articles with high purpose similarity are then returned to and displayed on the front - end . We describe the actual interfaces used in the studies in the corresponding design sections ( Sections 3 . 2 . 4 and 3 . 2 . 5 ) . Together the design of our system enabled what is to our knowledge the first functioning pro - totype of an interactive analogical search engine for scientific articles at scale . In the following sections , we report on how such a search engine can help researchers find analogical articles that facilitate creative ideation . 3 STUDY 1 : CREATIVE ADAPTATION WITH A HUMAN - IN - THE - LOOP ANALOGICAL SEARCH ENGINE In Study 1 , we set out to establish the viability of an analogical search engine using a human - in - the - loop probe in the domain of scholarly recommendations . We investigate whether analogical search returns a distinct and novel set of articles compared to keyword search results , and capture 9 https : / / tfhub . dev / google / universal - sentence - encoder - large / 5 . 10 https : / / cloud . google . com / ai - platform . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 9 participants\u2019 reaction to each result in a randomized order , blind to condition . To deeply under - stand the process of ideation using analogical articles we ask participants to come up with new ideas for their own research projects after reviewing each article . Using this data we code ideation outcomes in depth to explore the various ways in which analogical distance can shape ideation outcomes , such as inspiring direct transfer of solutions , or sparking adaptation of ideas into novel combinations . 3 . 1 Coding Ideation Outcomes We are interested in studying whether an analogical search engine provides distinctive and com - plementary value to other commonly used search approaches that rely on surface similarity . In particular , our focus is on the inspirational value rather than the immediate relevance of search results or the direct usefulness of solutions . The highest value of creative inspiration often comes from creatively adapting ideas to reformulate a problem and recognizing new bridges to previously unknown domains that open up entirely new spaces of ideas . For example , recognizing a connec - tion from the ancient art form of origami to fold intricate structures with article and building a sufficiently compact , deployable solar panel arrays and radiation shields led NASA to hire origami experts [ 27 , 89 , 118 ] . Our approach to measuring ideation outcome is through the use of a quaternary variable cat - egorizing the types of ideation . To capture the inspirational value of analogical search and move beyond the measurements focused on the immediate relevance or the direct usefulness we distin - guish the Creative Adaptation and Direct Application types of ideation . In our studies these two types corresponded to think - alouds that resulted in novel ideas whereas the rest ( Background and None ) corresponded to think - alouds in which no new ideas were produced . \u2014 Creative Adaptation : Novel mechanism ideas that involve substantial adaptation of the information provided in the article . These ideas are typically associated with a higher uncer - tainty of success due to the less familiar nature of the domains involved . \u2014 Direct Application : More directly applicable ideas that involve less adaptation than Cre - ative Adaptation . These ideas are typically associated with a lower uncertainty of success because researchers are more familiar with the domains . \u2014 Background : The information provided in the article is good for background reading ( e . g . , to learn about other domains ) . \u2014 None : Did not result in new ideas nor was useful for background reading . Creative Adaptation ideas generally involved a substantial amount of adaptation , while Direct Application ideas were closer to the source domain and more directly applicable . For example , using the data from one of our participants , applying the techniques for manipulating thermal conductance at solid - solid interfaces was considered a direct application idea for P1 ( Figure 3 , left ) because he was familiar with the concept of controlling the interfacial thermal conductiv - ity given the relevant approaches he developed in his current and past research projects . Thus the connections to the source problem were directly recognizable . On the other hand , creating a fin - based wall structure for heat transfer was an example of creative adaptation idea ( Figure 3 , right ) because of its novelty and the participant\u2019s unfamiliarity in relevant domains . The unfa - miliarity and uncertainty was generally more associated with analogs for creative adaptation than direct application . On the other hand , the unfamiliarity also sometimes acted as a barrier to partici - pants\u2019 openness and subsequent ideation . Though challenging , in order to recognize novel connec - tions to the source problem the participants may need to suspend their early rejection of a seem - ingly foreign idea and its surface - level mismatches and engage in deeper processing which could lead to re - imagination and re - formulation of the research problem at hand . To code the Creative ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 10 H . B . Kang et al . Fig . 3 . Example articles for the purpose of facilitating heat transfer heat in semiconductors . ( Top ) A Direct Application article involves directly applicable ideas and techniques for manipulating the interface material and structure to control thermal conductance . ( Bottom ) A Creative Adaptation example involves transferring a distant idea ( fin - based design for heat sinks ) and creatively adapting it into the target problem context ( designing nano - scale fins that could absorb heat and convert it to useful energy ) . Figure credits : contact configurations and interface resistance from [ 116 ] , fin - based heat sink from [ 104 ] , nano - fins from [ 94 ] . Adaptation and Direct Application types of ideation outcomes , the coders took into consideration different linguistic and contextual aspects of the descriptions of the ideas and their think - aloud process ( details in Section 3 . 2 . 3 ) . 3 . 2 Design of the Study 3 . 2 . 1 Participants . We recruited eight graduate ( four women ) researchers in the fields of sci - ences and engineering via email advertisement at a private R1 U . S . institution . Four were senior PhD students ( 3rd year or above and one recently defended their thesis ) and the rest was 2nd year or below . Disciplinary backgrounds of the participants included : Mechanical ( 3 ) , Biomedical ( 2 ) , Environmental ( 1 ) , Civil ( 1 ) , and Chemical Engineering ( 1 ) . Once a participant signed up for the study , we asked them to describe their research problems and send the research team search queries they use to look for inspirations on popular search engines such as Google Scholar . 11 Members of the research team screened articles with relevant purposes using these queries on the filtering in - terface ( Figure 4 , left ) . Despite our efforts to collect articles over diverse topical areas , the search engine did not contain enough articles for two of the participants who work on relatively novel fields ( e . g . , \u201cmachine learning methods of 3D bioprinting\u201d ) . These participants were interviewed on their current practices for reviewing prior works and coming up with new ideas for research and were not included in the subsequent analyses . 3 . 2 . 2 Study Procedure and Keyword - search Control . The rest of the participants were then in - vited to in - person interviews . To ensure that participants would be exposed to a sufficiently diverse set of analogical mechanisms and to maximize our power to observe the ideation process , we gen - erated a list of top 30 results from the analogical search engine using the search queries provided 11 https : / / scholar . google . com / . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 11 Fig . 4 . The front - end interfaces . ( Left ) Human reviewers used this filtering interface to input search queries received from the participants and remove articles with obviously irrelevant purposes . To assist the reviewers\u2019 filtering process , model predicted purpose ( e . g . , the noise reduction and time , highlighted in red at the bottom of the filtering interface ) and mechanism ( highlighted in green ) tokens were also provided along with the title and the abstract text . The background color turned green when the \u201cSimilar\u201d button is clicked and red when the \u201cDissimilar\u201d button is clicked . ( Right ) The ideation task interface was populated with a list of human filtered articles for review by the participants in Study 1 ( the order of articles was randomized ) . by the study participants . As a control condition we also included top 15 results from a keyword - based search engine using the standard Okapi BM25 algorithm [ 82 ] ( k 1 = 1 . 2 , b = 0 . 75 ) using the same search queries as the analogical search engine . The order of results in the list was ran - domized and participants were blind to condition . To account for the difference in the quantity of exposure in the analysis , we normalized the ideation outcomes by the number of results returned in each condition . Using this list we employed a think - aloud protocol [ 80 , 107 ] in which partic - ipants were presented with the title , abstract , and other metadata of articles and asked to think aloud as they read through them with the goal of generating ideas useful for their research using our Web - based interface ( Figure 4 , right ) . Although time consuming , this approach allowed us to capture rich data on participants\u2019 thought process and how those processes changed and evolved as participants considered how an article might relate to their own research problems . In addition , we asked the participants to make a judgment on the novelty of each article on a 3 - point Likert - scale . After participants finished reviewing the 45 articles , we interviewed them about their overall thoughts on the results\u2019 relevance and novelty and whether there were any surprising or unique results . Each interview lasted about one and a half hours and the participants were compensated $ 15 / hr for their participation . 3 . 2 . 3 Data and Coding . In total , our data consisted of 267 article recommendations for six par - ticipants and their Likert - scale questionnaire responses measuring the content novelty , after re - moving 3 within - condition duplicates ( these articles included cosmetic changes such as different ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 12 H . B . Kang et al . capitalization in the title or abstract ) . One participant ran out of time towards the end of the inter - view and only provided novelty measures for the last 17 article recommendations in the random - ized list . Thus , 250 transcripts of participants\u2019 think - aloud ideation after reading each article were used for analyzing ideation outcomes . To code the distance between the Creative Adaptation and Direct Application types of ideation outcomes , the coders took into consideration ( 1 ) the verbs used to describe the ideas ( e . g . , \u201cdesign\u201d , \u201cdevelop\u201d , or \u201cinvent\u201d were generally associated more with distant ideas compared to \u201capply\u201d , \u201cuse\u201d , \u201cadopt\u201d ; see Table 3 ) ; ( 2 ) the context of ideas such as participants\u2019 expression of unfamiliarity or uncertainty of the domain involved ( e . g . , \u201cI\u2019m not really sure\u201d vs . \u201cI\u2019m familiar with this domain\u201d ) ; and ( 3 ) participants\u2019 perceived immediacy of the idea\u2019s applicability ( i . e . , ideas perceived by participants as more immediately applicable were associated with direct application but not creative adaptation ideas ) . Two of the authors coded a fraction of the data together ( 13 / 250 , 5 . 2 % ) and then independently coded the rest blind - to - condition , using the four ideation outcomes types described in Section 3 . 1 and with the following protocol : The coders first judged the existence of an idea . If there was , then its type was further distinguished between Creative Adaptation and Direct Application using the linguistic and contextual descrip - tions described above ( e . g . , Creative Adaptation ideas were more frequently associated with the \u201cdesign\u201d words , higher unfamiliarity and uncertainty of the domains , and less immediate applica - bility , compared to Direct Application ideas ) . In case there was no concrete idea in the data , coders further distinguished between the Background vs . None cases . The agreement between coders was significant , with Cohen\u2019s \u03ba = 0 . 89 ( near perfect agreement ) for the four categories of ideation outcome . Given the high level of agreement between the coders , any disagreements were resolved via discussion on a case - by - case basis . 3 . 2 . 4 Apparatus 1 : The Human - in - the - loop Filtering Interface . In Study 1 , members of the re - search team first received search queries from study participants and reviewed the model - produced purpose matches to filter irrelevant articles using a filtering interface ( Figure 4 , left ) . This additional step was introduced to ensure that articles with obviously dissimilar purposes are not returned to study participants . Reviewers determined whether each article contained a clearly irrelevant pur - pose in which case it was removed by clicking the Dissimilar button at the bottom of the article . On the other hand when the Similar button was clicked it turned the background of the article green in the interface and increased the number of the articles collected so far . Reviewers continued the screening process until at least 30 articles with reasonable purpose similarity were collected . 3 . 2 . 5 Apparatus 2 : The Ideation Task Interface . The filtered articles were then displayed as a randomized list of articles to study participants ( Figure 4 , right ) . In addition to the content and metadata of articles ( e . g . , authors , publication date , venue ) , each article was presented with a Likert - scale question for measuring content novelty and a text input for ideation . 3 . 2 . 6 Limitations . To reduce potential biases , our coders were blind to experimental conditions and relied on participants\u2019 statements of ideas\u2019 novelty and usefulness ( e . g . , \u201cI\u2019ve never seen some - thing like this before , \u201d \u201cthis is not a domain I would\u2019ve searched if I used Google Scholar\u201d ) , and achieved a high inter - rater reliability . We believe coders had a reasonable understanding of how participants arrived at specific ideas from descriptions of their current and past research topics , think - alouds , and end - of - experiment discussions . Despite this , we also acknowledge the limita - tions of this approach and discuss how future research may improve upon it ( see Section 7 . 2 . 1 ) . 3 . 2 . 7 On Reporting the Results . We report the result of our studies below . To denote statis - tical significance we use the following notations : \u2217 ( \u03b1 = 0 . 05 ) , \u2217\u2217 ( \u03b1 = 0 . 01 ) , \u2217\u2217\u2217 ( \u03b1 = 0 . 001 ) , \u2217\u2217\u2217\u2217 ( \u03b1 = 0 . 0001 ) . Alpha levels were adjusted when appropriate in post - hoc analyses using Bon - ferroni correction . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 13 Fig . 5 . ( Left ) Participants judged analogy articles significantly more novel . The mean response to the question \u201cHave you seen this article before ? \u201d was significantly higher in Analogy : 2 . 7 ( SD : 0 . 48 ) than in Keyword : 2 . 3 ( SD : 0 . 55 ) . ( Right ) There were significantly more overlapping words between search query terms provided by participants and the title and abstract text of articles : Keyword : 4 . 1 ( SD : 1 . 74 ) vs . Analogy : 1 . 6 ( SD : 1 . 42 ) . 3 . 3 Result Finding novel articles for creative ideas . Our key measure of success is how article recommen - dations from the analogy search engine ( hereinafter analogy articles ) help scientists generate cre - ative ideas for their own research problems . To this end , we investigate ( a ) whether analogy articles are novel and complementary to the articles found from the keyword - search baseline ( hereinafter keyword articles ) and ( b ) whether analogy articles resulted in more creative adaptation ideas than direct application ideas in ideation . 3 . 3 . 1 Analogy Articles Differed from Keyword Articles and were Judged more Novel . The viability of our approach is based on the assumption that the analogy search pipeline returns a different distribution of results than a keyword - based baseline . This assumption appeared to hold true : the keyword - search and analogy - search conditions resulted in almost completely disjoint sets of ar - ticle recommendations . Out of the total 267 articles , the overlap between analogy and keyword articles was only one . Analogy articles appeared to represent a complementary set of results users would be unlikely to encounter through keyword - based search . To further examine this assumption we had participants rate the novelty of the results by asking them \u201c have you seen this article before ? \u201d on a 3 - point Likert scale response options of 1 : \u201c Yes , I have seen this article before \u201d , 2 : \u201c Yes , not exactly this article but I have seen similar ideas before \u201d , and 3 : \u201c No , I have not seen anything like this before \u201d . Participants found articles recommended in the analogy condition to contain significantly more novel ideas ( 2 . 7 , SD : 0 . 48 ) compared to the keyword condition ( 2 . 3 , SD : 0 . 55 ) ( Welch\u2019s two - tailed t - test , t = \u2212 5 . 53 , p = 1 . 33 \u00d7 10 \u2212 7 ) ( Figure 5 , left ) . Participants thought the \u201cvariance in results is much higher than using other search engines\u201d ( P5 ) and \u201cthere\u2019re a lot of bordering domains . . . which can be useful if I want to get ideas in them\u201d ( P4 ) . This difference was also reflected in the content of articles , with keyword articles having signifi - cantly more overlapping terms with participant - provided query terms ( 4 . 1 , SD : 1 . 74 ) than analogy articles ( 1 . 6 , SD : 1 . 42 ) ( Welch\u2019s two - tailed t - test , t ( 145 . 27 ) = 11 . 70 , p = 1 . 10 \u00d7 10 \u2212 22 ) ( Figure 5 , right ) . 12 More occurrences of familiar query terms in keyword articles\u2019 titles and abstracts may have led participants to perceive them as more familiar . 12 We measured the term overlap between participants\u2019 queries and the content of articles ( title and abstract ) . To preprocess text , we used NLTK [ 10 ] to tokenize articles\u2019 content , remove stopwords , digits , and symbols , and lemmatize adjectives , verbs , and adverbs . Finally , using the processed tokens we constructed a set of unique terms for each article and the query which was then compared to find overlapping terms . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 14 H . B . Kang et al . Fig . 6 . Frequency of the ideation outcome types by condition . Darker colors represent higher rates . Creative adaptation is 5 . 3 times more frequent among analogy articles ( 53 in Analogy vs . 10 in Keyword ) , while most of direct application is from keyword articles ( 3 in Analogy vs . 16 in Keyword ) . The distributions differed significantly ( chi - squared test , \u03c7 2 ( 3 ) = 52 . 12 , p < 1 . 0 \u00d7 10 \u2212 10 overall and \u03c7 2 ( 1 ) = 28 . 41 , p = 9 . 84 \u00d7 10 \u2212 8 for the contrast between the rates of creative adaptation and direct application ideas ) . 3 . 3 . 2 Analogy Articles Resulted in more Creative Adaptation Ideas than Direct Application Ideas . We found that the distribution of ideation outcome types differed significantly between analogy and keyword articles ( \u03c7 2 ( 3 ) = 52 . 12 , p < 1 . 0 \u00d7 10 \u2212 10 ) . Participants came up with more creative adaptation ideas ( N = 53 ; 32 % of total ) over direct application ideas ( N = 3 ; 2 % ) using analogy articles . In contrast , keyword articles resulted in more direct application ideas ( N = 16 ; 19 % ) than creative adaptation ideas ( N = 10 ; 12 % ) ( Figure 6 ) . The difference between creative adaptation and direct application was significant ( \u03c7 2 ( 1 ) = 28 . 41 , p = 9 . 84 \u00d7 10 \u2212 8 ) . To illustrate more concretely the divergent patterns of ideation leading to Creative Adaptation and Direct Application ideas , we describe vignettes from three participants ( Table 3 ) . While Direct Application ideas represented close - knit techniques and mechanisms directly useful for the source problem ( described with verbs such as apply and adopt ) , Creative Adaptation type ideas were more distant from the source problem and could be characterized with the use of different verbs asso - ciated with significant adaptation ( design and invent ) . For example , P1\u2019s research focused on the methods for improving nanoscale heat transfer in semiconductor materials . Previously he devel - oped mechanisms for manipulating the thermal conductivity at solid - solid interfaces , specifically by adjusting the semiconductor wall structures . Thus , an article reporting experimental results of manipulating thermal conductance on planar metallic contact points was deemed a directly useful article that might contain helpful techniques . On the other hand , an analogy article which dealt with the heat transfer phenomenon at a macroscale , using fin - based heat sink designs for electronic devices , gave him a new inspiration : to adapt fins for nanoscale heat transfer in semiconductors to not only transfer heat but also convert it into a useful form of mechanical energy . Despite the mismatch on scale ( [ macroscale ] (cid:2) [ microscale ] ) , challenging the assumption of the typical size of a fin - based design engendered an idea to creatively adapt it to convert heat into energy through an array of tiny fins , rather than merely dissipating it into space as in the original formulation of the problem . P1 also found another analogy article focused on thermal resistance at a liquid - solid interface useful for future ideation because despite its surface dissimilarities , there was a poten - tial mapping that may open up a new space of ideas ( e . g . , [ liquid ] (cid:2) [ polymer substrate ] , [ solid ] (cid:2) [ germanium ] , yet the pairwise relation [ liquid : solid ] \u2194 [ polymer substrate : germanium ] may be analogous and interesting ) : \u201cThis is liquid . . . but it\u2019s about liquid - solid interface which can be useful . . . because for the substrate that sits on top of silicon or germanium you use polymers which have liquid - like properties\u201d ( P1 ) . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 15 Table 3 . Examples of Direct Application and Creative Adaptation Types for Three Participants ( PID ) PID Research Problem Type Article Title \u2192 New Idea ( paraphrased ) 1 Improve nanoscale heat transfer in semiconductormaterial Direct Application Experimental investigation of thermal con - tact conductance for nominally flat metallic contact \u2192 Apply the techniques in the arti - cle to manipulate thermal conductance at the solid - solid interface Creative Adaptation Investigation on periodically developed heat transfer in a specially enhanced channel \u2192 Design nanoscale \u201cfins\u201d to absorb heat and convert it to mechanical energy 2 Grow plants better by optimizing entry of nanoparticle fertilizers into the plant Direct Application Nanoinformatics : Predicting Toxicity Us - ing Computational Modeling \u2192 Apply the computational modeling from the arti - cle for predicting toxicity of candidate nanoparticles Creative Adaptation Identification of Plant Using Leaf Image Analysis \u2192 Invent a hyperspectral 3D imaging mechanism for plants that opti - cally senses , traces , and images plant cells in 3 - dimensional structures 3 Enhance the evaporationefficiencyof thin liquid films in heat pipes and thermosyphons Direct Application Thin film evaporation effect on heat trans - port capability in a grooved heat pipe \u2192 Adopt the techniques in the article for manipulating the solid interface\u2019s surface properties to balance the film thickness and disjoining pressure Creative Adaptation Alkaline treatment kinetics of calcium phosphate by piezoelectric quartz crystal impedance \u2192 Design novel liquid film ma - terials for manipulating hydrophobicity to change disjoining pressure Each participant\u2019s research problem is described in the Problem column . While the topics of research problems vary , Creative Adaptation ideas are more distant in terms of content compared to the source problem than Direct Application ideas are , and may be characterized by the use of different sets of verbs ( { design , invent } in Creative Adaptation ideas versus { apply , adopt } in Direct Application ideas ) . In the case of P2 , an article focused on computational methods for toxicity prediction was deemed directly helpful because \u201cif certain nanomaterials are toxic to certain microorganisms that eat plants or kill them but safe for the plant , we can target these organisms using the nanomate - rials as pesticide . Another way this can be helpful is in predicting the chance of toxicity of the nanoparticles in our fertilizers\u201d ( P2 ) . Whereas an analogy article that uses image analysis for plant identification reminded her of \u201chyperspectral imaging in plants , like a CT scan for plants . So mak - ing a hyperspectral 3D model using something like this . . . to optically sense and trace plant cells ( such that the entry of fertilizer nanoparticles into plant cells can be monitored , a sub - problem of P2\u2019s research problem ) would be pretty cool . \u201d ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 16 H . B . Kang et al . Table 4 . Examples of Different Purpose - match Types Purpose - Match PID Participant Comment Full 2 \u201cIt\u2019s a little bit old ( from 2010 ) but I have read articles from that era . I love this . . . because the article mentions everything else and espe - cially one word which is \u201cdisjoining pressure\u201d\u2014if I were to publish my current project that\u2019s going to be the core topic . \u201d Part 1 \u201cThough I\u2019m not familiar with GFRP - GFRP . . . but I can see that they\u2019re referring to glass fiber reinforced plastic , so this is something not crystalized material . . . learning about this kind of materials is in - teresting . \u201d None 3 \u201cI don\u2019t know what a lot of words mean . I don\u2019t typically work with animals cells . \u201d Purpose - Match shows the level of purpose - match between a recommended article and each participant\u2019s research problem ( see Table 3 for descriptions of research problems ) . Fully matching purposes are those that match at both high - ( more abstract ) and low - levels ( specific details ) . Partial matches only match at the high - level abstraction and differ in details . The Participant Comment column shows relevant excerpts from the participant . As a third example , P6\u2019s research focused on recording and simulating electrical activity us - ing microelectrode arrays . To him , an analogy article about printing sensors for electrocardio - gram ( ECG ) recording seemed to present an interesting idea despite its mismatch in terms of scale ( [ nanoscale ] (cid:2) [ macroscale ] ) and manufacturing mechanism ( [ fabrication ] (cid:2) [ printing ] ) , be - cause the pairwise relation between [ nanoscale : fabrication ] \u2194 [ macroscale : printing ] engendered a reflection on the relative advantages of different methods and future research directions ) : \u201cInter - esting idea ! Instead of nanoscale fabrication , printing can be a good alternative for example for rapid prototyping . But I think the resolution won\u2019t be enough ( for use ) in nanoscale . . . works for this particular article\u2019s goal , but an idea for future research is whether we can leverage the benefit of both worlds\u2014rapid printing and precision of nanoscale fabrication\u201d ( P6 ) . 3 . 3 . 3 The Level of Purpose - match had Different Effects on the Ideation Outcome . Suggested in these examples is a certain kind of distance the ideas in analogy articles maintain in order to spur creative adaptation . We hypothesize that some amount of difference in purpose facilitates creative adaptation . This process may involve a curvilinear relationship between the degree of purpose mismatch and the resulting ideation outcome , with too much or too little deviation leading to a little - to - no benefit or even an adverse effect on the ideation outcome , a pattern that is consistent with the findings in the literature of creativity and learning outcomes ( e . g . , Csikszentmihalyi\u2019s optimal difficulty [ 25 ] ) . For this analysis , we coded each article based on three levels of purpose - match to the source problem : \u2014 Full : Both high - and low - level purposes match . \u2014 Part : Only the high - level abstract purpose matches . Explicit descriptions of the high - level purpose exist in either title and abstract of the article . At the same time , certain low - level aspects of the participant\u2019s research problem are mismatched as evidenced by relevant com - ments from the participant . \u2014 None : Neither high - nor low - level purposes match . Examples of these types of purpose - match are provided in Table 4 . High - level match can be considered as a first - order criterion of purpose match and low - level match as a second - order cri - terion : If the article does not have overlapping terms in terms of its purpose with the user query cast at a high level ( e . g . , transfer heat , grow plants ) then the low - level match does not matter , but if the article\u2019s purpose matches at the high level , its low - level alignment ( e . g . , specific aspects of ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 17 Table 5 . Regression Table of Three Mediation Analyses using Purpose - match , Novelty and Pid ( Participant ID ) as Mediators between Condition and the Binary Creative Adaptation Outcome Variable Effect of Condition Unique Effect Indirect Effect CI 95 % Mediator on Mediator ( a ) of Mediator ( b ) ( a \u00d7 b ) Lower Upper Purpose - match \u2212 0 . 42 \u2217\u2217\u2217\u2217 0 . 21 \u2217\u2217\u2217\u2217 \u2212 0 . 09 \u2217\u2217\u2217\u2217 \u2212 0 . 14 \u2212 0 . 05 ( . 08 ) ( . 05 ) 0 . 40 \u2217\u2217\u2217\u2217 \u2212 0 . 06 \u2212 0 . 02 \u2212 0 . 07 0 . 02 Novelty ( . 07 ) ( . 05 ) Pid \u2212 0 . 02 0 . 03 \u2217 \u2212 0 . 001 \u2212 0 . 02 0 . 02 ( . 22 ) ( . 02 ) Purpose - match was the only significant mediator between Condition and Creative Adaptation ( indirect effect = \u2212 0 . 09 , significant using a bootstrapping method [ 91 ] with 1 , 000 iterations , p < 2 \u00d7 10 \u2212 16 ) . the purpose , such as its scale or materialistic phase ) will additionally determine full ( i . e . , aligned in both high - and low - level aspects of the purpose ) vs . partial match ( i . e . , aligned only in the high - level but not low - level aspects of the purpose ) . Therefore , the coding procedure was symmetrical to the procedure described for coding four types of ideation outcome , with the high - level purpose match deciding between { Full , Part } and None match types , while the low - level purpose further distinguishing between Full vs . Partial match . Following this procedure , two independent coders achieved an inter - rater reliability Cohen\u2019s \u03ba = 0 . 72 ( substantial agreement ) and disagreements were resolved with case - by - case discussion . We used the mediation package 13 [ 105 ] to conduct a mediation analysis between the condition , the kind of purpose - match , and the binary Creative Adaptation ideation outcome . The analysis showed that the effect of condition ( Keyword vs . Analogy ) on the binary outcome of creative adap - tation was mediated by the degree of purpose - match , but not by the novelty of content , suggesting that the difference between full vs . partial matching on purpose is much more significant than the variance in the content novelty . We come back to this result in the discussion ( Section 7 . 2 . 3 ) . Table 5 presents the result of the mediation analyses . The regression coefficient between creative adaptation and condition was significant as was the regression coefficient between the degree of purpose match and creative adaptation . The indirect effect was ( \u2212 0 . 42 ) \u00d7 ( 0 . 21 ) = \u2212 0 . 09 . We tested the significance of this indirect effect using a bootstrapping procedure [ 91 ] ( p < 2 \u00d7 10 \u2212 16 ) , by com - puting the unstandardized indirect effects for each of 1 , 000 bootstrapped samples as well as the 95 % confidence interval ( CI ) . 14 Partial purpose matches in both keyword and analogy articles led to creative adaptation , but the rate was significantly higher with analogy articles . As expected , the ratio of direct application de - creased from the keyword articles that fully match in purpose ( Keyword Full , 68 % ) to the keyword articles that partially match in purpose ( Keyword Part , 6 % ) ( Figure 8 ) . At the same time , the rate of creative adaptation increased from the keyword articles that fully match in purpose ( Keyword Full , 0 % ) to the keyword articles that partially match in purpose ( Keyword Part , 21 % ) . However , the rate of creative adaptation differed significantly between the keyword and analogy articles , with the rate more than doubling among the analogy articles over keyword articles ( Analogy Part 47 % vs . Keyword Part 21 % ) . Homing in on the partial matches , these articles led to creative adap - tation ideas significantly more often in analogy search ( 47 % ) than keyword search ( 21 % ) ( Welch\u2019s 13 https : / / cran . r - project . org / web / packages / mediation / index . html . 14 Alternatively , it is possible that the mediating effect of the degree of purpose - match on the likelihood of creative adapta - tion outcome is moderated by novelty . However , the result of our analysis showed that this was unlikely : The effect was insignificant using the bootstrapping method \u2212 0 . 04 , ( p = 0 . 12 , 95 % CI = [ \u2212 0 . 09 , 0 . 01 ] ) . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 18 H . B . Kang et al . Fig . 7 . Proportion of creative adapta - tion ideas among the partial purpose - match articles . Creative Adaptation was significantly more frequent among the analogy articles ( 47 % ) than keyword articles ( 21 % ) ( Welch\u2019s two - tailed t - test , p = 9 . 0 \u00d7 10 \u2212 4 . Fig . 8 . The rate of ideation outcome types in full and partial pur - pose matches . Among the keyword articles as the purpose mis - match increases , the rate of creative adaptation also increases from 0 % to 21 % ( middle ) . However , this rate is significantly higher among the analogy articles ( 47 % ) than the keyword articles ( 21 % ) . Note that while purpose mismatches led to more creative adapta - tion among analogy articles , a large portion of them also resulted in no ideation outcome ( 38 % ) . two - tailed t - test , t ( 112 . 22 ) = \u2212 3 . 40 , p = 9 . 0 \u00d7 10 \u2212 4 , Figure 7 , left ) . While the partial purpose mis - match was highly associated with creative adaptation ideas , it could be a double - edged sword . Among the analogy articles , 38 % of the partial mismatches resulted in no useful ideation outcome as opposed to the 47 % that resulted in creative adaptation ( Figure 8 , Analogy Part ) . Therefore , knowing what mismatches are beneficial to creative adaptation has important implications for facilitating generative misalignment for ideation . 4 STUDY 2 : ENABLING A FULLY AUTOMATED ANALOGICAL SEARCH ENGINE 4 . 1 Motivation and Structure of the Study The findings of Study 1 suggest potential benefits of an analogical search engine for scientific re - search , but a core limitation of interactivity due to the human - in - the - loop system design prevented its use as a more realistic probe for understanding researchers\u2019 natural interaction with analogical results . Specifically , the results of Study 1 are limited by the lack of participants\u2019 ability to refor - mulate search queries and the study design that involved returning only a fixed number of articles that blended both keyword and analogy articles in a randomized order . These factors significantly deviate from realistic usage scenarios of a deployed analogical search engine and prevent us from observing the full scope of user interaction . In order to move beyond these limitations , first we need a fully automated pipeline that removes the need for human - in - the - loop filtering , thus al - lowing us to enable query reformulation and interaction with corresponding search results . To achieve this , we improved the model accuracy on extracting purposes and mechanisms from arti - cle abstracts by training a more sophisticated neural network that leverages more nuanced linguis - tic patterns . Specifically , we implemented an attention mechanism within a span - based Seq2Seq ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 19 model ( Model 2 ) such that it could learn words that frequently co - occur to describe coherent pur - poses or mechanisms in article abstracts , and as a result , learning more informative words for our purpose ( see Appendix for details of implementation ) . Through evaluating the system backed by this improved pipeline , we demonstrate how it can remove the human - in - the - loop while maintain - ing similar levels of accuracy . In the following sections , we report the evaluation results that show ( 1 ) an improved token - level prediction accuracy using the span - based Model 2 ; ( 2 ) rankings of the results aligning well with human - judgment of purpose - match from Study 1 ; and ( 3 ) top ranked results of the system maintaining a similar rate of partial purpose matches relative to that of the human - in - the - loop system from Study 1 . The interactivity enabled by the automated analogical search pipeline further allows us to ob - serve its use in more realistic scenarios . To probe how researchers would interact with an analog - ical search engine and what challenges they might face in the process , we ran case studies with six researchers ( Section 5 ) . From these studies , we uncover potential challenges ( Section 5 ) and synthesize design implications for future analogical search engines ( Section 6 ) . 4 . 2 Result 4 . 2 . 1 Improved Token - level Prediction of a Span - based Model . First we compared the span - based Model 2 with five other baselines to evaluate the token - level classification performance ( Table 6 ) . Model 2\u2019s overall F1 score was the highest at 0 . 65 ( Purpose ; PP : 0 . 65 , Mechanism ; MN : 0 . 64 , an 0 . 14 - and 0 . 14 - absolute - point increase from Model 1 , respectively ) on the validation set which represents an overall 0 . 15 - absolute - point increase from Model 1 used for the initial human - in - the - loop analogical search engine . 4 . 2 . 2 Pipeline with a Span - based Model Reflected Human Judgment for Ranking the Results . The improved token - level prediction performance materialized as an increase in the pipeline\u2019s ability to judge the degree of purpose match . For this evaluation , we first recorded every query provided by Study 1 participants that human - in - the - loop filterers used to search and filter the relevant articles . Then , we simulated the search condition of the filterers for the automated pipeline by providing it input as the exact queries they used . We capped the number of top search results sufficiently large at 1 , 000 for each query . From these top 1 , 000 results , we selected articles that also appeared in the human - in - the - loop system and collected the corresponding human - vetted judgments of high or low purpose - match . For each of these articles , we also collected its corresponding rank positions on the new ( automated ) pipeline\u2019s list of results . We compared the mean ranks of articles that are judged by human filterers as high purpose match to those of low purpose matches . The result showed that the new pipeline indeed was able to distinguish between the two groups of articles ; low purpose matches ( i . e . , articles that were deemed not relevant and subsequently filtered by the judges in Study 1 ) were placed at significantly lower positions on the list than high purpose matches ( i . e . , unfiltered articles in Study 1 ) . The mean rank for low purpose matches was 465 while for high purpose matches it was 343 ( Figure 9 ) . This difference was significant ( t ( 192 . 49 ) = 3 . 29 , p = 0 . 0012 . Welch\u2019s two - tailed t - test . ) . 4 . 2 . 3 Different Model Performance on Finding Articles that Fully or Partially Match on Purpose . Data and coding . In addition to the overall rankings reflecting human - vetted judgments we also found that the proportion of partial purpose matches was significant among the top - ranked results . We sourced top 20 results for each participant\u2019s research problem with the automated system ( Model 2 ) using the exact queries and order used by the human - in - the - loop filterers in Study 1 . We compared this to four other approaches : ( 1 ) the human - in - the - loop system in Study 1 ( BiLSTM with filtering ) , ( 2 ) a BiLSTM - based system excluding the human - in - the - loop from 1 ( BiL - STM ) , ( 3 ) randomly sampled articles ( Random ) , and ( 4 ) a keyword - based search results , which was ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 20 H . B . Kang et al . Table 6 . F1 Scores of Different Models , Sorted by the Overall F1 Score of Purpose ( PP ) and Mechanism ( MN ) Detection Model Embedding All PP MN ( finetuned ) 1 . Model 2 [ 67 ] ELMo ( N ) 0 . 65 0 . 65 0 . 64 2 . BiLSTM ELMo ( N ) 0 . 63 0 . 67 0 . 59 3 . BiLSTM SciBERT ( N ) 0 . 62 0 . 69 0 . 55 4 . BiLSTM - CRF [ 90 ] ELMo ( N ) 0 . 58 0 . 59 0 . 57 5 . BiLSTM GloVe ( Y ) 0 . 55 0 . 56 0 . 53 6 . Model 1 GloVe ( N ) 0 . 50 0 . 51 0 . 50 The span - based Model 2 gave the best overall F1 score ( blue ) . In comparison , the average agreement ( % ) between two experts\u2019 and crowdworkers\u2019 annotations was 0 . 68 ( PP ) and 0 . 72 ( MN ) [ 21 ] . We used AllenNLP [ 41 ] to implement the baseline models 1 \u2013 5 . Fig . 9 . Mean ranks of human - judged high and low purpose match articles from the span - based pipeline . Low matches were ranked significantly lower ( the rank num - ber was higher ) , on average at 465 th ( SD : 261 . 92 ) than high matches at 343 th ( SD : 279 . 48 ) . used as control in Study 1 ( Keyword ) . There were no overlapping articles between Model 2 and other conditions except for the Keyword condition which had 1 overlapping article . To code the degree of purpose match , we blended the results of Model 2 , biLSTM , and Random conditions . Two of the authors coded a fraction of the data together blind - to - condition ( 7 . 4 % , N = 20 / 270 ) following the same procedure used in Study 1 . Then they independently coded the rest blind - to - condition achieving an inter - rater agreement of \u03ba = 0 . 80 ( substantial agreement ) . We resolved any disagreement through discussion on an individual case basis . Result . We found that the Model 2 - based system achieved a parity with the human - in - the - loop system ( Study 1 ) for finding purpose matches ( Figure 10 ) , with more than half of the system\u2019s top 20 results judged to be partial purpose matches . In contrast , when human - in - the - loop filtering was removed from the BiLSTM - based system , the frequency of partial purpose matches decreased from 58 % to 37 % while the frequency of no matches increased from 40 % to 59 % . Random sampling resulted in mostly irrelevant results , with no alignment on purpose with the source problem . An interesting point of comparison is between the keyword - based and Model 2 - based search results . Keyword search mostly outperformed Model 2 - based system by finding full purpose matches at a much higher rate ( 23 % in keyword search vs . 4 % in the Model 2 - based system ) , with similar rates of partial purpose matches ( 58 % vs . 55 % ) , and significantly less no purpose matches ( 19 % vs . 41 % ) . On average the purpose match score was the highest in keyword - search followed by the Model 2 - based and the human - in - the - loop systems ( Figure 11 ) . Combined with the results of Study 1 , this suggests the complementary value of analogical search : The higher rate of full - matches in keyword - search may be good when searchers know what they are looking for , such as in direct search tasks and foraging from familiar sources of ideas . Nonetheless , because analogy articles were both deemed significantly more novel by the scientists and had little - to - no overlap with keyword - search articles , they augmented keyword - based search results with a complementary set of articles that introduce useful mismatches in their purposes . This set of articles may open up new domains of ideas that scientists may not have been aware of , and encourage creative adaptation . 5 CASE STUDIES WITH RESEARCHERS To further understand what potential interaction challenges prevent future analogical search en - gines from reaching their full potential , we ran case studies with six participants . To this end , ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 21 Fig . 10 . Distribution of Full , Part , and None purpose matches among the five sourcing mechanisms : BiL - STM with filtering represents the human - in - the - loop system ( Study 1 ) ; Model 1 represents a system based on the BiLSTM model alone , without human - in - the - loop filtering ; Model 2 represents the fully automated system ; Random represents randomly sampled articles ; Keyword represents keyword - based search ( Control in Study 1 ) . Model 2 and BiLSTM with filtering showed a similar distribution of purpose matches , and more partial purpose matches than BiLSTM alone . Random showed mostly no matches . The Keyword condition re - sulted in the highest number of fully matched articles and the lowest number of no matches , suggesting that keyword - based search may be an effective mechanism for direct search tasks , but potentially less effective for inspirational / exploratory search tasks . we developed a frontend interface that includes a text input for reformulating purpose queries ( Figure 12 , right ) . This frontend interfaced with our automated , Model 2 - based backend to display a ranked list of analogical results for a given purpose query . Leveraging the fully automated search engine , we also removed the structure of Study 1 that asked participants to engage with each result they encountered , thus allowing us to observe which results researchers more naturally attend to and engage with . In sum , the design of our case studies differ from Study 1 in three aspects : ( 1 ) participants interacted with only the analogical search results ranked in the order of purpose similarity , without blended keyword - based search results ; ( 2 ) participants reviewed search results returned for their queries and reformulated the queries when needed ; and ( 3 ) participants looked for articles that interest them and may serve as sources of inspiration for their research problems at their own pace , without being explicitly asked to engage with each result they encounter . The primary goal of our case studies was to identify generalizable challenges that analogical search engines may face in interactive use , thus providing us insights on how future engines may be designed and improved . Specifically , we were interested in the challenges related to ( 1 ) how researchers recognize relevance of analogical search results and ( 2 ) how researchers formulate and reformulate purpose search queries while interacting with analogical search results . 5 . 1 Participants and Design Participants were asked to formulate purpose queries for their own research problems and interact with the results to find interesting articles . If an article gave them a new idea relevant to their research project , they were asked to write a short project proposal in a shared Google Doc and explain how the article helped them to come up with the idea . Interviews were conducted via Zoom and lasted for roughly an hour . Participants were paid $ 20 in compensation . One participant was an assistant professor in mechanical engineering at a public R1 U . S . university and five were PhD researchers in the fields of sciences and engineering at a private R1 U . S . university . Two were senior PhD students ( 3rd year or above ) and the rest were 2nd year or below . Disciplinary backgrounds of the participants included Chemical ( 2 ) , Civil ( 3 ) , and Mechanical Engineering ( 1 ) . We note that one participant previously took part in Study 1 , whose research focus was the same in terms of its general domain . However , the participant\u2019s ideas and the specific articles of interest that led to them did not have overlap between the two studies . Table 7 describes participants\u2019 research problems . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 22 H . B . Kang et al . Fig . 11 . The distribution of mean purpose match scores over different conditions ( mappings : None (cid:7)\u2192 0 , Part (cid:7)\u2192 1 , and Full (cid:7)\u2192 2 ) . The mean purpose - match score of the system backed by Model 2 ( 0 . 63 , SD : 0 . 56 ) is significantly higher than that of the system used in Study 1 without the human - in - the - loop ( BiLSTM , \u03bc = 0 . 45 , SD : 0 . 58 ) ( Welch\u2019s two - tailed t - test , t ( 237 . 87 ) = 2 . 49 , p = 0 . 0135 ) , similar to that of the system with the human - in - the - loop ( BiLSTM with filtering , \u03bc = 0 . 62 , SD : 0 . 52 ) ( t ( 244 . 65 ) = 0 . 25 , p = 0 . 80 ) , and sig - nificantly lower than that of the keyword - based search ( Keyword , \u03bc = 1 . 04 , SD : 0 . 65 ) ( t ( 159 . 38 ) = \u2212 4 . 57 , p = 0 ) . Fig . 12 . The search interface used for case stud - ies featured an input for query reformulation which participants used to iteratively reformu - late their queries . Apparatus : Search interface . The improved performance of Model 2 backed the fully automated pipeline without human filtering . The search interface interacting with this back - end included a text input for reformulating purpose search queries as well as a list view of search results that showed a sorted list of articles with similar purposes ( Figure 12 ) . 5 . 2 Result 5 . 2 . 1 Overall Impressions . Overall participants described their experience with the analogical search engine in a positive light ( e . g . , \u201chelps me think at a broad topic or a big picture level\u201d\u2014P2 ; \u201cfind some very interesting and useful ideas , the design is also very simple , good when focusing on key areas of research\u201d\u2014P5 ; and \u201cvery interested now what the future of this engine would look like\u201d\u2014P3 ) , but a deeper look suggested that the success of ideation depended on how well searchers were able to engage with analogical results that deviate from their expectations : \u201cIt\u2019s surprising that the engine recommends examples like these\u201d\u2014P3 ; \u201cIf I input the same search queries on Google Scholar it\u2019d not normally return these things . . . this search engine works in a different way\u201d\u2014P1 . 5 . 2 . 2 \u201cNot the Kind of Article I\u2019d Look for but . . . \u201d : The Challenge of Early Rejections . Unlike similarity - maximizing search engines , the diversity in analogical search results can lead to pre - mature rejection of alternative mechanism ideas . One of the factors contributing to premature rejection of alternatives may be the tendency for adherence to a set of existing ideas or concepts , as studied in the literature of design fixation ( e . g . , [ 66 ] ) . In our study , the participants found the variety of domains featured in search results confusing , and it sometimes prevented them from en - gaging with the ideas therein . For example , P3 , whose research studies ways to manage or reduce task complexity for nuclear power plant operators , expected to see results similar to Google Scholar ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 23 Table 7 . Case Study Participants\u2019 Descriptions of Own Research Problems PID Participants\u2019 Description of Research Problem 1 Improve heat pipe evaporation 2 Computer simulations for fluids in nanoscale and uncovering their heat - transfer properties 3 Developing a model to identify complex steps in Nuclear Power Plant ( NPP ) operation , and understanding what task features and structures cause the complexity and how this influences the operators\u2019 performance 4 Designing simulators for training bridge inspectors 5 Developing algorithms and extensible frameworks for detecting personal protective equipment ( PPE ) in construction sites to improve the safety of construction workers 6 Convergence rates of optimization algorithms under multiple initial starting positions which are typically in the domains of operational and managerial sciences , but was surprised by unfamiliar domains represented in search results : \u201cThese ( distributed networked systems design or path planning for automated robots ) are not the kinds of fields that I normally read in , if I found them elsewhere I would\u2019ve probably thought they\u2019re irrelevant and skipped\u201d ( P3 ) . Ranging from unfamiliar terms ( P1 , P4 , P5 ) to unfamiliar categories of approaches ( e . g . , \u201cNot sure what \u2018Gauss \u2013 Newton approach for solving constrained optimization\u2019 is\u201d\u2014P6 ) , or high - level research directions ( e . g . , \u201cthis is different from my research direction , people who work on this direction might find it interesting , though\u201d\u2014P1 ) , participants saw the diversity of results as a challenge for engagement . P1 pointed out a perceived gap between the expectation of least effort and the cognitive processing required when engaging with analogical ideas and adapting them : \u201cAs I understand it , I think this search engine is trying to present articles from related but different fields to let people make connections . But people expect less friction . ( The result is ) something interesting but I can\u2019t directly write it into a project proposal . . . I think it would be challenging to make people get interested in investing time to read the articles in depth to come up with connections . I wonder what would happen if this was hosted just as an online website ( instead of the study context ) \u201d \u2014P1 On the other hand , analogs that did get examined more deeply could ultimately lead to creative adaptation . For example , P3 mapped task scheduling among computer processes to task assign - ment among the nuclear power plant operators , and came up with an idea to adapt algorithmic scheduling used in real - time distributed systems to a scheduling mechanism that could be useful in her research context . Represented symbolically this process was akin to ideating what might best fill in the \u201c ? \u201d in the relational structure [ scheduling algorithm : processes in distributed sys - tems ] \u2194 [ ? : nuclear power plant operators ] : \u201cI think the algorithms proposed in this article could be useful for calculating the operator task execution time , the power plant system\u2019s response time , and the time margin between the execution time and the system response time . . . so that the next task assignment can factor in these margins and things related to workers\u2019 well - being like rest and time required between switching tasks\u201d ( P3 ) . Participants seemed to recognize a small number of core relations as kernel for creative adap - tation . In the example of P3 , scheduling processes in the distributed systems article piqued her in - terest and led her to connect them with similar concepts in the literature she was already familiar with : \u201cYou need to make that connection . . . I saw parallels between ( distributed systems domain ) concepts like [ scheduling ] and [ tasks ] and [ scheduling tasks for the operators ] \u201d ( P3 ) . Similarly , P5 recognized a similarity between [ monitoring people\u2019s performance ] in fitness training and ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 24 H . B . Kang et al . Fig . 13 . Diagram showing different abstraction levels of purposes and their relations . Node A (cid:4) corresponds to a more specific query than its higher - level representation , denoted as B (cid:4) . Similarly , node C (cid:4) represents a more specific purpose representation of A (cid:4) , accessible via the A (cid:4) \u2192 abstraction B (cid:4) \u2192 specification C (cid:4) path . [ monitoring whether construction workers are wearing personal protective equipment ] in con - struction sites . He then adapted the idea of tracking heat emission in the fitness context to his own : \u201cI like the idea of [ monitoring heat emissions ] in fitness training . . . maybe I can also detect heat emissions from construction workers to see if they are wearing the safety vests or masks while also monitoring the site conditions and worker efficiency . It also gives me an idea to moni - tor the CO 2 emissions from workers so as to improve the robustness of detection\u201d ( P5 ) . In this case , monitoring and the physical nature of the activities involved helped P5 see the connection useful for creatively adapting the source idea . 5 . 2 . 3 \u201cI don\u2019t Know What to Type in\u201d : The Challenge of Query ( Re - ) Formulation . Another chal - lenge participants faced was that they were not used to formulating their search queries in terms of high level purposes of their research . On average participants entered 5 . 2 queries ( Min : 1 , Max : 18 , SD : 5 . 87 ) , 87 % ( 27 ) of which were in the form of a single noun phrase ( e . g . , \u201cheat pipe evaporation , \u201d\u2014 P1 , \u201ctask complexity\u201d\u2014P3 , \u201ctheoretical optimization convergence for non - convex functions\u201d\u2014P6 ) or a comma - separated set of multiple noun phrases ( e . g . , \u201cheat transfer , nanoscale , fluid\u201d\u2014P2 ) that represented specific aspects related to research purposes rather than the core purposes themselves . For example , the purpose of \u201cheat pipe evaporation\u201d may be to transfer heat , and the purpose of searching for \u201ctheoretical optimization convergence for . . . \u201d may be to detect when optimization converges or diverges , or to effectively sample unknown ( non - convex ) distributions . One of the reasons why participants formulated search queries in this way may be wrongly assuming that the search engine used keyword matching to find results . For example , extensive prior experience with search engines that highlight matching keywords in abstracts ( e . g . , Google Scholar ) in response to users\u2019 search queries can reinforce such assumptions among the users . In addition , participants\u2019 domain knowledge useful for judging which of the returned articles are relevant may have led them to notice a set of keywords the inclusion of which strongly signifies the relevance of a article . In contrast , the analogical search results often seemed to not feature such directly similar terms and this contributed to the difficulty of judging whether a result is relevant and how : \u201cI find these articles not very related to my search query at first . It\u2019d be better if you can use some graph or some pictures to indicate how these articles can relate to my keywords\u201d ( P5 ) ; \u201cI\u2019d not consider . . . ( because ) they are totally different , right ? They look irrelevant . . . until I think about it I can realize that it\u2019s useful . But if you give me the article , at first I don\u2019t realize that\u201d ( P3 ) . While it may not feel as compelling or natural to participants , formulating and abstracting queries at a high level may lead to searching more distant results that are analogous at a higher level . For example , by querying \u201cdetect personal protective equipment\u201d instead of \u201cpersonal ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 25 protective equipment construction , \u201d P5 found novel mechanisms of detection , such as general im - age segmentation algorithms or an approach to monitoring heat in the context of fitness training not specific to construction sites and personal protective equipment but nonetheless useful for creative adaptation . Querying \u201cscheduling tasks\u201d instead of \u201ctask complexity\u201d for P3 resulted in finding scheduling algorithms in distributed computer systems that otherwise P3 would not have encountered , while \u201cassigning tasks\u201d led to novel auction mechanisms which made her think about a system in which each power plant operator can bid for a task as opposed to being assigned one . Schematically , Figure 13 shows how formulating queries at a higher level of abstraction than spec - ifying the problem context in full details ( A (cid:4) \u2192 B (cid:4) ) may lead to discovering novel mechanisms that are relevant at the high level of abstraction , and in more distant ways from the original problem formulation ( B (cid:4) \u2192 C (cid:4) ) . 6 DESIGN IMPLICATIONS From both the case studies\u2019 and Study 1\u2019s participants\u2019 reflection on the challenges of interacting with analogical search results , common themes emerged . Here , we present three design implica - tions for future analogical search systems synthesized from these results . We use subscripts to denote which study participants participated in when appropriate . 6 . 1 Support Purpose Representation at Different Levels of Abstraction Analogical search engines should support users to formulate their purpose queries at different levels of abstraction . Additionally the search engine may prompt users to consider abstracting or specifying their purpose queries in the first place , and explain how it might help bring new in - sights into their problems . As seen in the case studies ( Section 5 . 2 . 3 ) , scientists recognized their purpose queries may be represented at multiple levels , but prior experiences with similarity maxi - mizing search engines may also have anchored them around pre - existing rigid formulation of pur - poses . Prompting users to consider their research problems at multiple levels may work against this rigidity , and providing candidate suggestions at varying levels may further reduce the cog - nitive demand . Moving up on the hierarchy to abstract purpose queries may be possible through removing parts of the query words that correspond to specific constraints , or by replacing them with more general descriptions . For example two participants of Study 1 had an identical purpose representation at a high level ( \u201cfacilitate heat transfer\u201d ) despite the differences in materialistic phases targeted in each purpose : solid material and semiconductors for P1 Study 1 and liquid thin films for P3 Study 1 . Furthermore , we also observed that looking for only the exact match of a purpose can lead to missed opportunities . For example , although \u201cfins represent a different idea for transferring the heat\u201d and \u201cthey ( fins ) don\u2019t match in terms of the scale\u2014macro , not nano , \u201d it nevertheless made P1 Study 1 wonder \u201cwhat if we could design nanoscale wall structures that act like fins that convert heat to mechanical energy ? \u201d A corollary to this observation is that sometimes the superpositions of misalignment with just the right amount can lead to interesting results . For P4 Study 1 , an ar - ticle presenting experimental techniques for piezoelectric properties was interesting despite its misalignment such as [ simulation - based ] ( source ) (cid:2) [ experimental ] ( analog ) and [ dielectric prop - erties ] ( source ) (cid:2) [ piezoelectric properties ] ( analog ) : \u201cThough it\u2019s an experimental study , it\u2019s very close in terms of the material and phenomenon so likely to be helpful . Because we might be able to pick up some trends like , if we increased the temperature , the dielectric response gets stronger or weaker , inferred from the experimental piezoelectric responses , which can then be used to corrob - orate simulation results or help configure its parameters\u201d ( P4 Study 1 ) . However , too much deviation seemed detrimental to its potential for inspiration : \u201c [ Molecular dynamic simulation ] is the same tool , but ( this article studies ) [ thermal ] ( not [ dielectric ] ) properties on [ polymer composites ] . . . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 26 H . B . Kang et al . [ polymer composites ] are harder to model\u201d ( P4 Study 1 ) . In sum , analogical search engines should support not only the capability to \u201cnarrow it down\u201d with specific constraints , but also ways to re - lax them to broaden the search space when suitable , thus making feasible the sweet spot between too little ( i . e . , similarity maximization and trivial matches ) and too much deviation ( i . e . , critical misalignment and unusable analogs ) . 6 . 2 Support Iterative Steering from Critical Misalignment and Towards Generative Misalignment Analogical search engines should recognize that important constraints may be discovered by users only after seeing misaligned analogs , and support this discovery process by presenting effective ex - amples of misalignment to users . Analogs that deviate on some aspects of the source problem but preserve important relations may be particularly conducive to analogical inspiration that opens up not just individual solutions , but entirely new domains of solutions . However , at the same time scientists also found it challenging to know how to come up with effective search queries because combinations of misalignment can sometimes lead to an unintended intersection of do - mains : \u201cI feel like I\u2019m tricking the machine because [ thin film ] is often used with [ solids ] , and the term [ pressure ] also appears a lot in [ manufacturing ] . . . so combining them gives a subset of arti - cles concerned with heat transfer in solid materials and in manufacturing\u201d ( P3 Study 1 ) ; \u201con Google Scholar also , I get a lot of polymer strings and get ( irrelevant ) results like we use an [ electric ] device to study [ vibration and stress ] of [ polymers ] . . . the machine is picking up [ electric ] and [ properties ] such as vibration and stress in the context of studying polymers but what I really want is [ electric properties ] of [ polymers ] not [ electronic devices ] to study the [ mechanical properties ] of [ poly - mers ] \u201d ( P4 Study 1 ) . Nonetheless , seeing misaligned analogs can be an effective way of reasoning about salient constraints and reflecting on hidden assumptions . For example , while evaluating arti - cles about designing microelectrode arrays , P6 Study 1 said : \u201cNow I think about this ( result ) , I assumed a lot of things when typing that search query . . . though impedance and topology are my main focus in microelectrode arrays , the coating , size , interface between a cell membrane and electrodes / sensors , biocompatibility , softness of electrodes , fabrication process , material of the platform : silicon or poly - mer or graphene , form factor : attaching electrodes to a shank - like structure or a broom - like structure , degree of invasiveness , are all part of the possible areas of research and it makes sense that they showed up\u2014there is no way the machine would have known that from my query . \u201d This excerpt illustrates how knowing what the necessary specifications are and which constraints need to be abstracted to cast a wide - enough net to catch interesting ideas appeared to be a difficult task for scientists , especially when they had to recall important attributes rather than simply recognize them from ex - amples of misalignment . Prior work in cognitive sciences also shows how dissimilarity associated with various factors in analogical mappings [ 45 ] can pressure working memory [ 111 ] , increase cognitive load [ 102 ] , and increase response time taken to produce correct mappings for analogy problems [ 71 ] . Therefore , analogical search engines should help to reduce the cognitive effort re - quired in the process , for example by proactively retrieving results that are \u201cusefully\u201d misaligned such that searchers can better recognize ( as opposed to having to recall ) salient constraints and re - fine their problem representation . This process is deeply exploratory [ 93 , 114 , 117 ] in nature , and suggest the importance of both providing end - users a sense of progress over time [ 103 ] as well as adequate feedback mechanisms for the machine to adjust according to the changing end - user search intent [ 72 , 95 , 96 ] . For example , while the machine may \u201ccorrectly\u201d recognize a significant anaogical relevance at a higher level of purpose representation and recommend macro - scale mech - anisms to a scientist who studies nano - scale phenomena ( P1 Study 1 ) or solid and semiconductor - based cooling mechanisms to a scientist in liquid and evaporative cooling systems ( P3 Study 1 ) , these ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 27 analogs may be critically misaligned on the specific constraints of the problem ( i . e . , the scale or materialistic phase ) and thus considered by end - users as useless and even harmful . 6 . 3 Support Reflection and Explanation of Analogical Relevance Throughout the process of analogical search , human - AI coordination is critical for success , and an important aspect is how deeply the human users can reflect on the retrieved analogs [ 53 ] and recognize how different notions of relevance may exist for their own problem context , despite po - tential dissimilarity on the surface . Looking at previous examples of the tools and techniques de - veloped for targeted reflection support may be useful to this end . For example , ImageCascade [ 76 ] provides intelligent support such as automatically generated mood - boards and semantic labels for groups of images to help designers communicate their design intent to others . Another system , Card Mapper , visualizes relative co - occurrences of design concepts using proximity in the design space [ 26 ] . Similarly representing the space of analogical ideas using spatial encoding of similarity between two analogs , or designing information that supports getting a sense of the space of search results\u2014e . g . , semantic category labels similar to ImageCascade\u2019s or the distribution of the domains that analogs are pulled from\u2014may be an avenue for fruitful future research . The explanation of relevance is also important especially when there is a risk of early rejection ( Section 5 . 2 . 2 ) . Using examples from the case studies , one approach to explaining relevance might be to surface a small number of core common features between an analog and a problem query . Such common features were considered useful by scientists for making analogical connections , and they could creatively adapt them for their own research problem context . When common features are not directly re - trieved , generation of more elaborate explanations may be required . We refer to [ 6 , 14 , 70 , 98 ] for those interested in future design considerations of automatically generated recommendation expla - nation . Further complementing the direct explanation of relevance approach , techniques such as prompting or reminding the searchers of previously rejected or overlooked ideas may also trigger deeper reflection and delay premature rejection of the ideas based solely on their surface dissim - ilarity . Participants from both studies commented that the critical first step towards analogical inspiration may be raising similarly enough attention and interest above the initial \u201chump\u201d of cog - nitive demand . Gentle reminders ( e . g . , \u201cAsk me later if this would be interesting and also provide a list of items\u201d\u2014P1 Case Studies ) or resurfacing previously rejected articles in light of new information ( P1 Case Studies , P3 Case Studies ) may help with users cross this barrier . 7 DISCUSSION 7 . 1 Summary of Contribution With the exponential growth of research output and the deepening specialization within different fields , encouraging analogical inspiration for scientific innovation that connects distant domains becomes ever more challenging . Our human - in - the - loop and fully automated analogical search en - gines represent an approach for supporting such analogical inspirations for challenging scientific problems . We have demonstrated in Study 1 that our human - in - the - loop system finds novel results that participants would be unlikely to encounter from keyword - based search , and that these results lead to high levels of creative adaptation . Through a mediation analysis we also showed that this success was driven by the analogical search engine\u2019s ability to find partial purpose matches ( e . g . , matching at the high - level purpose but differs at the low - level details ) . We saw the nuanced effects of partial purpose alignment on the results\u2019 goodness as analogs for inspiration . Through quali - tative observations , we described how certain attributes of analogical mapping were perceived as more salient by participants , and that mismatches on them can have either a positive ( i . e . , genera - tive insights ) or a negative impact ( i . e . , critical misalignment ) on creative adaptation . In contrast , ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 28 H . B . Kang et al . keyword - based search resulted in more full purpose matches and a higher level of direct applica - tion . The value of keyword - based search and analogy - based search thus may complement each other , while keyword - based search can help researchers find \u201cexactly that\u201d , analogy - based search can help researchers to switch from a preservative mode ( i . e . , aiming to find results with maximal similarity to the query ) to a generative mode ( i . e . , aiming to find analogs that are relevant despite the surface dissimilarity ) of searching , and ultimately lead them to recognize unusual relations and come up with ways to creatively adapt existing ideas for novel domains . We also demonstrated how improving the Seq2Seq purpose and mechanism identification model can remove the human - in - the - loop but maintain a similar level of accuracy on purpose - match by human judges . This improvement enabled us to develop a fully automated analogical search sys - tem to use as a probe to study searchers\u2019 more natural interaction with analogical results . Through a series of evaluation we first show that our automated analogical search pipeline can emulate hu - man judgment of purpose match and that it finds partial purpose matches in top ranked results with a similar rate compared to the human - in - the - loop system used in Study 1 . Then through case studies we find generalizable challenges that future analogical search engines may face , such as early rejection of alternative mechanism ideas and the difficulty of abstracting and representing purposes at the right level . From our studies we synthesize design implications for future analog - ical search engines , such as supporting purpose representations at different levels of abstraction , supporting the iterative process of steering away from critically misaligned analogs and towards a fertile land of generative misalignment , and providing explanations on why certain analogical search results may be relevant . We envision that future studies will shed light on deeper cognitive sources of the challenges identified here . A fruitful avenue of research may be studying how the dual processing theory [ 69 , 112 ] underlies or interacts with analogical search interaction . Studying also how simplification heuristics [ 84 ] may negatively bias interaction with analogical results and how it may be reduced for expert user populations may be an interesting future direction [ 17 , 77 ] . 7 . 2 Limitations and Future Work 7 . 2 . 1 Experimental Design and Improving its Validity . Our findings have several limitations . First the design of our studies may be improved to increase the experimental validity . We believe that our coders of the ideation outcomes had a reasonable understanding of participants\u2019 research context from descriptions of current and past research topics , think - alouds with 45 articles , and end - of - experiment discussions , and that the procedure of coding reduced potential biases ( e . g . , the coders were blind to experimental conditions , relied on participants\u2019 statements of novelty and distance ) . Despite this , it is possible that they judged ideas differently from domain experts , for example coding more or fewer ideas as creative adaptation , or pre - filtering useful ideas in the human - in - the - loop stage . In addition , other quality dimensions such as potential for impact or domain - expert - judged idea quality are largely inaccessible within the studies presented here . Fu - ture research may improve on these limitations by iterating on the experimental design , collecting data for triangulating the results and capturing other quality dimensions of the generated ideas . Additionally , future work may add ablation studies to quantify the effects of human filtering in Study 1 on the ideation outcome as well as sensitivity studies to relate how much the increased token - level classification performance of trained models may reduce the burden of human filter - ing . Furthermore , additional experiments with baselines other than keyword - based search using the whole abstract will help pinpoint the potential advantages of representing and matching arti - cles using extracted purposes and mechanisms . For example , Chan et al . [ 21 ] found that embed - ding all words from an abstract ( using GloVe embeddings ) resulted in retrieval of fewer analogi - cal items than when extracted purposes and mechanisms were used . Replicating this result with ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 29 additional approaches such as contextualized word embeddings and pre - trained language models ( e . g . , ELMo [ 90 ] , BERT [ 29 ] , and SciBERT [ 7 ] ) will be valuable . 7 . 2 . 2 Potential Sampling Bias . The sampling strategy in Study 1 was purposefully unbalanced , where analogical articles were sampled twice as much as keyword articles to ensure participants\u2019 exposure to sufficiently diverse results . This was crucial for uncovering potential benefits and challenges of our analogical search engine and investigating its viability . This ratio was chosen purposefully , to balance the statistical power for detecting potentially significant differences be - tween the conditions , while also limiting the number of articles that each participant had to review . Given the cognitive burden of reviewing an article while thinking aloud , we decided on 45 in total with the 2 : 1 ratio to fit the practical time limits of interviews . However , this may have led to unan - ticipated effects on ideation outcomes despite having accounted for the difference in sample sizes by measuring the outcomes in ratios . For example , when the results were combined into a single blinded list , the over - representation of analogical results over more purpose - aligned keyword re - sults may have shifted the users\u2019 overall perceived value of the list to be more or less positive . Users\u2019 perception of diverse results may have been further affected by their relative over - representation . For example , increased cognitive load for processing analogical mapping [ 51 , 52 , 102 ] may suggest that results that fully match on the purpose search query may have been perceived even more favorably than analogical results , due to a negative spill over effect from the rest of the articles in the list , which were less likely matched on the purpose . Investigating whether such factors led to compounding effects beyond our ratio - based measures of usefulness remains an open question for future work . 7 . 2 . 3 Controlling the Diversity of Search Results . Our work is also limited by the lack of con - trollability in sampling the search results beyond purpose similarity . As described in Section 2 . 2 . 1 , from pilot tests in our corpus we discovered that even close purpose matches of scientific articles already had high variance in terms of the mechanisms they proposed which allowed us to focus our approach to sampling based solely on purpose similarity . The simplicity of this approach also means fewer hyper parameters in the sampling mechanism compared to other approaches [ 61 , 62 ] . However , all the approaches including this work thus far lacked a mechanism for explicitly con - trolling the diversity in retrieved search results which remains a fruitful avenue for future work . For example , prior research has uncovered the nuanced effects of distance ( e . g . , near vs . far sources of inspiration [ 24 , 97 ] ) , suggesting the benefit of targeting analogs at different distance from the source problem for the right context . Future research may also uncover further complexities in the relationship between novelty and purpose - match . The result of our mediation analysis ( Table 5 ) showed that the novelty of content among the search results in Study 1 was not a significant factor to the same extent that the three levels of purpose match was . However , the relationship between novelty and purpose match may be more complex than the levels of manipulation presented in this work . For example , [ 30 ] suggested a greater importance of novelty than usefulness for predicting creativity scores . Future work may design mechanisms to manipulate the variance in content nov - elty and alignment in the purpose - mechanism schema to uncover dynamics between the two that go beyond the results from mediation analyses presented here ( Section 3 . 3 . 3 ) . Furthermore , chal - lenges with the abstraction of purposes remain open , for example , how core versus peripheral at - tributes of research purposes may be identified , and how they may be selectively matched at a spe - cific level of the conceptual hierarchy . Finally , not all query formulations are created equal in terms of their suitability for analogical search . We observed in the case studies that participants wanted to express different query intent via reformulation ( Section 5 . 2 . 3 ) . While participants could refor - mulate their search queries and examine the returned results from our analogical search engine in real - time , it was unclear whether and how specific query formulations may lead to more or less ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 30 H . B . Kang et al . diverse results , and how subsequent queries may be updated after reviewing them . As such , sys - tems that assist users in the potentially tedious process of query reformulation [ 113 ] ( for example , by way of automatic query expansion [ 18 ] ) in the context of analogical search will be important . 7 . 2 . 4 Studying the Effect of Larger Context of Scientific Innovation on Analogical Innovation . Due to our focus on ideation outcomes , our results do not explain how these ideas may be integrated , developed , and shared across the research communities . Studying the lifetime of ideas that goes beyond their inception will deepen our understanding of the factors that currently make analogical innovation such a rare event in sciences ( for example , Hofstra et al . suggested that more seman - tically distant conceptual combinations receive far less uptake [ 58 ] ) . Through interviewing our study participants and other colleagues in academia we found emerging structures related to this challenge . Our interviews informed us that in general the context in which a scientist exists\u2014such as the scientist\u2019s role in a project , the maturity of a project , and the broader academic culture\u2014can ultimately change how they interact with and seek analogical inspirations . For example , a third - year PhD student studying chemical engineering commented \u201cIn the current stage of my project it\u2019s more about parameter - tuning\u2014running multiple experiments at once and comparing which configuration works the best . . . If I were a first year PhD student maybe I would be in a broader field and exploration . \u201d In contrast , a PhD in biology who recently defended noted that \u201canalogical inspirations would perhaps be more useful if you\u2019re looking for a postdoc or a faculty position . \u201d In addition , the underlying career incentive structures in academia may also affect researchers\u2019 perception of and openness to analogical inspirations . One of the study participants commented \u201csince I\u2019m already a third year PhD student and my project is further along and more firmed up , I\u2019m not really looking for really far inspirations . . . first we push the specific way we have in mind with many iterations on the experiments until , say , publication . \u201d In addition to the career - wise incentives there are other types of competitive resourcefulness ( e . g . , social resources such as the advisors\u2019 and colleagues\u2019 expertise that participants can easily tap into ; physical and other forms of resources such as tangible artifacts like previously developed code packages or experimental pro - cesses and setups ) . These factors can influence scientists\u2019 perception of their advantage and lead them to interpret analogical inspirations as more or less useful , feasible , and directly applicable to their research . This observation is further suggested by survey results that asked our participants : \u201c Could this article be useful to you ? , \u201d their ratings were significantly higher for keyword articles than analogy articles despite them having come up with creative adaptation ideas more often with analogy articles . Therefore , future work that studies incentive structures , the quality of ideation outcome , their feasibility , the differences in research context e . g . , frames of research contribution such as discovery - oriented vs . novel system development - oriented , and taking a longitudinal ob - servation of the variation in such factors will add a significant depth to our understanding . 8 CONCLUSION In this article we present our novel human - in - the - loop and fully automated analogical search en - gines for scientific articles . Through a series of evaluations , we found that analogous articles that our systems retrieved were novel and useful for sparking creative adaptation ideas . However , sig - nificant work is needed to continue this trajectory , including additional understanding of the con - text and incentives of scientists as well as advances in the data pipeline and interaction methods beyond those described here for a system to maximize its real - world impact . We imagine a future in which scholars and designers could find inspirations based on deep analogical similarity that can drive innovation across fields . We hope this work will encourage sci - entists , designers , and system builders to collaborate across disciplinary boundaries to accelerate the rate of scientific innovation . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 31 APPENDIX A REPRODUCIBILITY Training and validation datasets . The original annotation dataset from [ 21 ] also includes Back - ground and Findings annotations which we exclude due to their relatively high confusion rates among the annotators with the Purpose and Mechanism classes and to balance the number of available training examples per annotation class . Model parameter selection . We experimented with changing the model capacity relative to the signal present in the training dataset by tuning the number of hidden layers and the nodes used in each model architecture . For Model 1 , we found a hidden layer of 100 nodes was sufficient . We opti - mized this model using the cross - entropy loss and the Adam optimizer [ 73 ] with a 0 . 0001 learning rate . For Model 2 , we found three hidden layers with 256 nodes led to an improved accuracy on the validation set . We trained this model with an L2 regularizer ( \u03b1 = 0 . 01 ) , dropouts with the rate of 0 . 3 , and the Adam optimizer with a 0 . 001 learning rate . Span - based model architecture . We adapt SpanRel [ 67 ] as architecture for the span - based Model 2 . SpanRel combines the boundary representation ( BiLSTM ) and the content representa - tion with a self - attention mechanism for finding the core words . More specifically , given a sentence x = [ e 1 , e 2 , . . . , e n ] , of n token embeddings , a span s i = [ \u03c9 s i , \u03c9 s i + 1 , . . . , \u03c9 f i ] is a concatenation of the content representation z i c ( weighted average across all token embeddings in the span ; SelfAttn ) and the boundary representation z i b of the start ( s i ) and end positions ( f i ) of the span : u 1 , u 2 , . . . , u n = BiLSTM ( e 1 , e 2 , . . . , e n ) , z c i = SelfAttn ( e s i , e s i + 1 , . . . , e f i ) , z b i = [ u s i ; u f i ] , z i = [ z c i ; z b i ] . We use the contextualized ELMo 5 . 5B embeddings 15 for token representation , following the near state - of - the - art performance reported on the named entity recognition task on the Wet Lab Proto - col dataset in [ 67 ] . We refer to [ 67 , 79 ] for further details . Other parameters . We use GloVe vectors for input feature representation for Model 1 with 300 dimensions , consistent with the prior work [ 11 , 78 , 88 ] . For Model 2 , we use the contextualized ELMo 5 . 5B embeddings as described above which have pre - determined 1 , 024 dimensions . We use USE [ 20 ] for encoding purposes . A USE embedding vector has pre - determined 512 dimensions . ACKNOWLEDGMENTS We thank our study participants for their valuable insights and feedback . REFERENCES [ 1 ] Random projection in Locality - sensitive hashing . 2008 . Retrieved January 23 , 2022 from https : / / en . wikipedia . org / wiki / Locality - sensitive _ hashing # Random _ projection . [ 2 ] Annoy : How it works . 2018 . Retrieved January 23 , 2022 from https : / / github . com / spotify / annoy # how - does - it - work . [ 3 ] Kevin D . Ashley . 1991 . Reasoning with cases and hypotheticals in HYPO . International Journal of Man - machine Stud - ies 34 , 6 ( 1991 ) , 753 \u2013 796 . [ 4 ] Yoram Bachrach , Yehuda Finkelstein , Ran Gilad - Bachrach , Liran Katzir , Noam Koenigstein , Nir Nice , and Ulrich Paquet . 2014 . Speeding up the Xbox recommender system using a euclidean transformation for inner - product spaces . In Proceedings of the 8th ACM Conference on Recommender Systems . Association for Computing Machinery , New York , NY , 257 \u2013 264 . DOI : https : / / doi . org / 10 . 1145 / 2645710 . 2645741 15 https : / / allennlp . org / elmo . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 32 H . B . Kang et al . [ 5 ] Dzmitry Bahdanau , Kyunghyun Cho , and Yoshua Bengio . 2016 . Neural machine translation by jointly learning to align and translate . arXiv : 1409 . 0473 . Retrieved from https : / / arxiv . org / abs / 1409 . 0473 . [ 6 ] Gagan Bansal , Tongshuang Wu , Joyce Zhou , Raymond Fok , Besmira Nushi , Ece Kamar , Marco Tulio Ribeiro , and Daniel Weld . 2021 . Does the whole exceed its parts ? The effect of AI explanations on complementary team perfor - mance . In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems . 1 \u2013 16 . [ 7 ] Iz Beltagy , Kyle Lo , and Arman Cohan . 2019 . SciBERT : A pretrained language model for scientific text . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Confer - ence on Natural Language Processing . Association for Computational Linguistics , Hong Kong , China , 3615 \u2013 3620 . DOI : https : / / doi . org / 10 . 18653 / v1 / D19 - 1371 [ 8 ] Yoshua Bengio , R\u00e9jean Ducharme , Pascal Vincent , and Christian Janvin . 2003 . A neural probabilistic language model . The Journal of Machine Learning Research 3 ( 2003 ) , 1137 \u2013 1155 . [ 9 ] Justin M . Berg . 2014 . The primal mark : How the beginning shapes the end in the development of creative ideas . Organizational Behavior and Human Decision Processes 125 , 1 ( 2014 ) , 1 \u2013 17 . DOI : https : / / doi . org / 10 . 1016 / j . obhdp . 2014 . 06 . 001 [ 10 ] Steven Bird and Edward Loper . 2004 . NLTK : The natural language toolkit . In Proceedings of the ACL Interactive Poster and Demonstration Sessions . Association for Computational Linguistics , Barcelona , Spain , 214 \u2013 217 . Retrieved from https : / / www . aclweb . org / anthology / P04 - 3031 . [ 11 ] Piotr Bojanowski , Edouard Grave , Armand Joulin , and Tomas Mikolov . 2017 . Enriching word vectors with subword information . Transactions of the Association for Computational Linguistics 5 ( 2017 ) , 135 \u2013 146 . DOI : https : / / doi . org / 10 . 1162 / tacl _ a _ 00051 [ 12 ] Lutz Bornmann and R\u00fcdiger Mutz . 2015 . Growth rates of modern science : A bibliometric analysis based on the number of publications and cited references . Journal of the Association for Information Science and Technology 66 , 11 ( 2015 ) , 2215 \u2013 2222 . [ 13 ] Samuel R . Bowman , Gabor Angeli , Christopher Potts , and Christopher D . Manning . 2015 . A large annotated cor - pus for learning natural language inference . In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing . Association for Computational Linguistics , Lisbon , Portugal , 632 \u2013 642 . DOI : https : / / doi . org / 10 . 18653 / v1 / D15 - 1075 [ 14 ] Zana Bu\u00e7inca , Maja Barbara Malaya , and Krzysztof Z . Gajos . 2021 . To trust or to think : Cognitive forcing functions canreduceoverrelianceonAIinAI - assisteddecision - making . ProceedingsoftheACMonHuman - computerInteraction 5 , CSCW1 ( 2021 ) , 1 \u2013 21 . [ 15 ] Jaime G . Carbonell . 1983 . Learning by analogy : Formulating and generalizing plans from past experience . In Proceed - ings of the Machine Learning . Springer , 137 \u2013 161 . [ 16 ] Jaime Guillermo Carbonell . 1985 . Derivational analogy : A theory of reconstructive problem solving and expertise ac - quisition . Technical Report . Carnegie - mellon Univ . Pittsburgh PA Dept . of Computer Science . [ 17 ] E . Moulton Carol - anne , Glenn Regehr , Maria Mylopoulos , and Helen M . MacRae . 2007 . Slowing down when you should : A new model of expert judgment . Academic Medicine 82 , 10 ( 2007 ) , S109 \u2013 S116 . [ 18 ] Claudio Carpineto and Giovanni Romano . 2012 . A survey of automatic query expansion in information retrieval . ACM Computing Surveys 44 , 1 ( 2012 ) , 1 \u2013 50 . [ 19 ] Daniel Cer , Mona Diab , Eneko Agirre , I\u00f1igo Lopez - Gazpio , and Lucia Specia . 2017 . SemEval - 2017 task 1 : Semantic textual similarity multilingual and crosslingual focused evaluation . In Proceedings of the 11th International Workshop on Semantic Evaluation . Association for Computational Linguistics , Vancouver , Canada , 1 \u2013 14 . DOI : https : / / doi . org / 10 . 18653 / v1 / S17 - 2001 [ 20 ] DanielCer , YinfeiYang , Sheng - yi Kong , NanHua , NicoleLimtiaco , RhomniSt . John , NoahConstant , MarioGuajardo - Cespedes , Steve Yuan , Chris Tar , Yun - Hsuan Sung , Brian Strope , and Ray Kurzweil . 2018 . Universal sentence encoder . CoRR abs / 1803 . 11175 . http : / / arxiv . org / abs / 1803 . 11175 [ 21 ] Joel Chan , Joseph Chee Chang , Tom Hope , Dafna Shahaf , and Aniket Kittur . 2018 . SOLVENT : A mixed initiative system for finding analogies between research papers . In Proceedings of the ACM on Human - Computer Interaction . DOI : https : / / doi . org / 10 . 1145 / 3274300 [ 22 ] Joel Chan , Steven P . Dow , and Christian D . Schunn . 2015 . Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? Design Studies 36 ( 2015 ) , 31 \u2013 58 . DOI : https : / / doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 [ 23 ] Joel Chan and Christian D . Schunn . 2015 . The importance of iteration in creative conceptual combination . Cognition 145 ( 2015 ) , 104 \u2013 115 . DOI : https : / / doi . org / 10 . 1016 / j . cognition . 2015 . 08 . 008 [ 24 ] Joel Chan , Pao Siangliulue , Denisa Qori McDonald , Ruixue Liu , Reza Moradinezhad , Safa Aman , Erin T . Solovey , Krzysztof Z . Gajos , and Steven P . Dow . 2017 . Semantically far inspirations considered harmful ? Accounting for cog - nitive states in collaborative ideation . In Proceedings of the 2017 ACM SIGCHI Conference on Creativity and Cognition . 93 \u2013 105 . [ 25 ] Mihaly Csikszentmihalyi and Mihaly Csikzentmihaly . 1990 . Flow : The Psychology of Optimal Experience . Harper & Row New York . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 33 [ 26 ] Dimitrios Darzentas , Raphael Velt , Richard Wetzel , Peter J . Craigon , Hanne G . Wagner , Lachlan D . Urquhart , and Steve Benford . 2019 . Card mapper : Enabling data - driven reflections on ideation cards . In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems . 1 \u2013 15 . [ 27 ] Nicola Davis . 2017 . Nasa needs you : Space agency to crowdsource origami designs for shield . ( 2017 ) . Re - trieved 03 March , 2021 from https : / / www . theguardian . com / science / 2017 / jul / 20 / nasa - needs - you - space - agency - to - crowdsource - origami - designs - for - shield . [ 28 ] Derek J . de Solla Price . 1965 . Networks of scientific papers . Science 149 , 3683 ( 1965 ) , 510 \u2013 515 . DOI : https : / / doi . org / 10 . 1126 / science . 149 . 3683 . 510 [ 29 ] Jacob Devlin , Ming - Wei Chang , Kenton Lee , and Kristina Toutanova . 2019 . BERT : Pre - training of deep bidirectional transformers for language understanding . In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics : Human Language Technologies . Association for Computational Linguistics , Minneapolis , Minnesota , 4171 \u2013 4186 . DOI : https : / / doi . org / 10 . 18653 / v1 / N19 - 1423 [ 30 ] Jennifer Diedrich , Mathias Benedek , Emanuel Jauk , and Aljoscha C . Neubauer . 2015 . Are creative ideas novel and useful ? Psychology of Aesthetics , Creativity , and the Arts 9 , 1 ( 2015 ) , 35 . [ 31 ] Steven Dow , Blair MacIntyre , Jaemin Lee , Christopher Oezbek , Jay David Bolter , and Maribeth Gandy . 2005 . Wizard of Oz support throughout an iterative design process . IEEE Pervasive Computing 4 , 4 ( 2005 ) , 18 \u2013 26 . [ 32 ] K . N . Dunbar . 1997 . How scientists think : On - line creativity and conceptual change in science . In Proceedings of the Creative Thought : An Investigation of Conceptual Structures and Processes . T . B . Ward , S . M . Smith , and J . Vaid ( Eds . ) , Washington D . C . , 461 \u2013 493 . [ 33 ] Chris Eliasmith and Paul Thagard . 2001 . Integrating structure and meaning : A distributed model of analogical map - ping . Cognitive Science 25 , 2 ( 2001 ) , 245 \u2013 286 . [ 34 ] Maryam Fazel - Zarandi and Eric Yu . 2008 . Ontology - based expertise finding . In Proceedings of the Practical Aspects of Knowledge Management . Takahira Yamaguchi ( Ed . ) , Springer , Berlin , 232 \u2013 243 . [ 35 ] Marsha E . Fonteyn , Benjamin Kuipers , and Susan J . Grobe . 1993 . A description of think aloud method and protocol analysis . Qualitative Health Research 3 , 4 ( 1993 ) , 430 \u2013 441 . [ 36 ] Kenneth Forbus . 2001 . Exploring Analogy in the Large . MIT Press . [ 37 ] Kenneth D . Forbus , Ronald W . Ferguson , and Dedre Gentner . 1994 . Incremental structure - mapping . In Proceedings of the 16th Annual Conference of the Cognitive Science Society . 313 \u2013 318 . [ 38 ] Kenneth D . Forbus , Ronald W . Ferguson , Andrew Lovett , and Dedre Gentner . 2017 . Extending SME to handle large - scale cognitive modeling . Cognitive Science 41 , 5 ( 2017 ) , 1152 \u2013 1201 . [ 39 ] KatherineFu , JoelChan , JonathanCagan , KennethKotovsky , ChristianSchunn , andKristinWood . 2013 . Themeaning of \u201cnear\u201d and \u201cfar\u201d : The impact of structuring design databases and the effect of distance of analogy on design output . Journal of Mechanical Design 135 , 2 ( 2013 ) , 021007 . [ 40 ] Katherine Fu , Joel Chan , Christian Schunn , Jonathan Cagan , and Kenneth Kotovsky . 2013 . Expert representation of design repository space : A comparison to and validation of algorithmic output . Design Studies 34 , 6 ( 2013 ) , 729 \u2013 762 . DOI : https : / / doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 [ 41 ] Matt Gardner , Joel Grus , Mark Neumann , Oyvind Tafjord , Pradeep Dasigi , Nelson F . Liu , Matthew Peters , Michael Schmitz , and Luke Zettlemoyer . 2018 . AllenNLP : A deep semantic natural language processing platform . In Proceed - ings of the Workshop for NLP Open Source Software . Association for Computational Linguistics , Melbourne , Australia , 1 \u2013 6 . DOI : https : / / doi . org / 10 . 18653 / v1 / W18 - 2501 [ 42 ] Dedre Gentner . 1983 . Structure - mapping : A theoretical framework for analogy . Cognitive Science 7 , 2 ( 1983 ) , 155 \u2013 170 . [ 43 ] D . Gentner , S . Brem , R . W . Ferguson , P . Wolff , A . B . Markman , and K . D . Forbus . 1997 . Analogy and creativity in the works of Johannes Kepler . In Proceedings of the Creative Thought : An Investigation of Conceptual Structures and Processes . T . B . Ward , J . Vaid , and S . M . Smith ( Eds . ) , American Psychological Association , Washington D . C . , 403 \u2013 459 . [ 44 ] Dedre Gentner and Russell Landers . 1985 . Analogical reminding : A good match is hard to find . In Proceedings of the Unknown Host Publication Title . IEEE , 607 \u2013 613 . [ 45 ] Dedre Gentner and Linsey Smith . 2012 . Analogical reasoning . Encyclopedia of Human Behavior 2 ( 2012 ) , 130 \u2013 136 . [ 46 ] Mary L . Gick and Keith J . Holyoak . 1980 . Analogical problem solving . Cognitive Psychology 12 , 3 ( 1980 ) , 306 \u2013 355 . [ 47 ] Mary L . Gick and Keith J . Holyoak . 1983 . Schema induction and analogical transfer . Cognitive Psychology 15 , 1 ( 1983 ) , 1 \u2013 38 . DOI : https : / / doi . org / 10 . 1016 / 0010 - 0285 ( 83 ) 90002 - 6 [ 48 ] Karni Gilon , Joel Chan , Felicia Y . Ng , Hila Liifshitz - Assaf , Aniket Kittur , and Dafna Shahaf . 2018 . Analogy mining for specific design needs . In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems . ACM , New York , NY , 11 pages . DOI : https : / / doi . org / 10 . 1145 / 3173574 . 3173695 [ 49 ] Milene Gon\u00e7alves , Carlos Cardoso , and Petra Badke - Schaub . 2013 . Inspiration peak : Exploring the semantic distance between design problem and textual inspirational stimuli . International Journal of Design Creativity and Innovation 1 , 4 ( 2013 ) , 215 \u2013 232 . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 34 H . B . Kang et al . [ 50 ] Howard E . Gruber and Paul H . Barrett . 1974 . Darwin on Man : A Psychological Study of Scientific Creativity . E . P . Dutton . [ 51 ] Graeme S . Halford . 1992 . Analogical reasoning and conceptual complexity in cognitive development . Human Devel - opment 35 , 4 ( 1992 ) , 193 \u2013 217 . [ 52 ] Graeme S . Halford , William H . Wilson , and Steven Phillips . 1998 . Processing capacity defined by relational com - plexity : Implications for comparative , developmental , and cognitive psychology . Behavioral and Brain Sciences 21 , 6 ( 1998 ) , 803 \u2013 831 . [ 53 ] Ning Hao , Yixuan Ku , Meigui Liu , Yi Hu , Mark Bodner , Roland H . Grabner , and Andreas Fink . 2016 . Reflection enhances creativity : Beneficial effects of idea evaluation on idea generation . Brain and Cognition 103 ( 2016 ) , 30 \u2013 37 . [ 54 ] M . Hesse . 1966 . Models and Analogies in Science . Notre Dame , IN . [ 55 ] Mary B . Hesse . 1966 . Models and Analogies in Science . Notre Dame , IN . [ 56 ] Sepp Hochreiter and J\u00fcrgen Schmidhuber . 1997 . Long short - term memory . Neural Computation 9 , 8 ( 1997 ) , 1735 \u2013 1780 . [ 57 ] Douglas R . Hofstadter and Melanie Mitchell . 1995 . The copycat project : A model of mental fluidity and analogy - making . Advances in Connectionist and Neural Computation Theory 2 ( 1995 ) , 205 \u2013 267 . [ 58 ] Bas Hofstra , Vivek V . Kulkarni , Sebastian Munoz - Najar Galvez , Bryan He , Dan Jurafsky , and Daniel A . McFarland . 2020 . The diversity \u2013 innovation paradox in science . Proceedings of the National Academy of Sciences 117 , 17 ( 2020 ) , 9284 \u2013 9291 . [ 59 ] KeithJ . HolyoakandPaulThagard . 1989 . Analogicalmappingbyconstraintsatisfaction . CognitiveScience 13 , 3 ( 1989 ) , 295 \u2013 355 . [ 60 ] K . J . Holyoak and P . Thagard . 1996 . The analogical scientist . In Proceedings of the Mental Leaps : Analogy in Creative Thought . K . J . Holyoak and P . Thagard ( Eds . ) , Cambridge , MA , 185 \u2013 209 . [ 61 ] Tom Hope , Joel Chan , Aniket Kittur , and Dafna Shahaf . 2017 . Accelerating innovation through analogy mining . In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . ACM , New York , NY , 235 \u2013 243 . DOI : https : / / doi . org / 10 . 1145 / 3097983 . 3098038 [ 62 ] Tom Hope , Ronen Tamari , Hyeonsu Kang , Daniel Hershcovich , Joel Chan , Aniket Kittur , and Dafna Shahaf . 2022 . Scaling creative inspiration with fine - grained functional facets of product ideas . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . Association for Computing Machinery , New York , NY , 15 . DOI : https : / / doi . org / 10 . 1145 / 3491102 . 3517434 [ 63 ] Hen - Hsen Huang and Hsin - Hsi Chen . 2017 . DISA : A scientific writing advisor with deep information struc - ture analysis . In Proceedings of the 26th International Joint Conference on Artificial Intelligence . 5229 \u2013 5231 . DOI : https : / / doi . org / 10 . 24963 / ijcai . 2017 / 773 [ 64 ] Zhiheng Huang , Wei Xu , and Kai Yu . 2015 . Bidirectional LSTM - CRF models for sequence tagging . CoRR . [ 65 ] John E . Hummel and Keith J . Holyoak . 2003 . A symbolic - connectionist theory of relational inference and generaliza - tion . Psychological Review 110 , 2 ( 2003 ) , 220 . [ 66 ] David G . Jansson and Steven M . Smith . 1991 . Design fixation . Design Studies 12 , 1 ( 1991 ) , 3 \u2013 11 . [ 67 ] Zhengbao Jiang , Wei Xu , Jun Araki , and Graham Neubig . 2020 . Generalizing natural language analysis through span - relation representations . In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics . Association for Computational Linguistics , Online , 2120 \u2013 2133 . DOI : https : / / doi . org / 10 . 18653 / v1 / 2020 . acl - main . 192 [ 68 ] Arif E . Jinha . 2010 . Article 50 million : An estimate of the number of scholarly articles in existence . Learned Publishing 23 , 3 ( 2010 ) , 258 \u2013 263 . [ 69 ] Daniel Kahneman . 2011 . Thinking , Fast and Slow . Macmillan . [ 70 ] Hyeonsu B . Kang , Rafal Kocielnik , Andrew Head , Jiangjiang Yang , Matt Latzke , Aniket Kittur , Daniel S . Weld , Doug Downey , and Jonathan Bragg . 2022 . From who you know to what you read : Augmenting scientific recommendations with implicit social networks . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . Association for Computing Machinery , New York , NY , 23 . DOI : https : / / doi . org / 10 . 1145 / 3491102 . 3517470 [ 71 ] Mark T . Keane , Tim Ledgeway , and Stuart Duff . 1994 . Constraints on analogical mapping : A comparison of three models . Cognitive Science 18 , 3 ( 1994 ) , 387 \u2013 438 . [ 72 ] Diane Kelly and Jaime Teevan . 2003 . Implicit feedback for inferring user preference : A bibliography . In Proceedings of the Acm Sigir Forum . ACM New York , NY , 18 \u2013 28 . [ 73 ] Diederik P . Kingma and Jimmy Ba . 2014 . Adam : A method for stochastic optimization . ( 2014 ) . In 3rd International Conference on Learning Representations , ICLR 2015 , San Diego , CA , USA , May 7 - 9 , 2015 , Yoshua Bengio and Yann LeCun ( Eds . ) . http : / / arxiv . org / abs / 1412 . 6980 [ 74 ] Aniket Kittur , Lixiu Yu , Tom Hope , Joel Chan , Hila Lifshitz - Assaf , Karni Gilon , Felicia Ng , Robert E Kraut , and Dafna Shahaf . 2019 . Scaling up analogical innovation with crowds and AI . Proceedings of the National Academy of Sciences 116 , 6 ( 2019 ) , 1870 \u2013 1877 . [ 75 ] MadelineK . Kneeland , MelissaA . Schilling , andBarakS . Aharonson . 2020 . Exploringunchartedterritory : Knowledge search processes in the origination of outlier innovation . Organization Science 31 , 3 ( 2020 ) , 535 \u2013 557 . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . Augmenting Scientific Creativity with an Analogical Search Engine 57 : 35 [ 76 ] Janin Koch , Nicolas Taffin , Michel Beaudouin - Lafon , Markku Laine , Andr\u00e9s Lucero , and Wendy E . MacKay . 2020 . ImageSense : An intelligent collaborative ideation tool to support diverse human - computer partnerships . Proceedings of the ACM on Human - Computer Interaction 4 , CSCW1 ( 2020 ) , 1 \u2013 27 . [ 77 ] Kathryn Ann Lambe , Gary O\u2019Reilly , Brendan D . Kelly , and Sarah Curristan . 2016 . Dual - process cognitive interven - tions to enhance diagnostic reasoning : A systematic review . BMJ Quality & Safety 25 , 10 ( 2016 ) , 808 \u2013 820 . [ 78 ] Thomas K . Landauer and Susan T . Dumais . 1997 . A solution to Plato\u2019s problem : The latent semantic analysis theory of acquisition , induction , and representation of knowledge . Psychological Review 104 , 2 ( 1997 ) , 211 . [ 79 ] Kenton Lee , Luheng He , Mike Lewis , and Luke Zettlemoyer . 2017 . End - to - end neural coreference resolution . In Pro - ceedings of the 2017 Conference on Empirical Methods in Natural Language Processing . Association for Computational Linguistics , Copenhagen , Denmark , 188 \u2013 197 . DOI : https : / / doi . org / 10 . 18653 / v1 / D17 - 1018 [ 80 ] ClaytonLewis . 1982 . Usingthe\u201cThinking - aloud\u201dMethodinCognitiveInterfaceDesign . IBMTJWatsonResearchCenter Yorktown Heights , NY . [ 81 ] Salvador E . Luria and Max Delbr\u00fcck . 1943 . Mutations of bacteria from virus sensitivity to virus resistance . Genetics 28 , 6 ( 1943 ) , 491 . [ 82 ] Christopher D . Manning , Prabhakar Raghavan , and Hinrich Sch\u00fctze . 2008 . Introduction to Information Retrieval . Cam - bridge University Press . [ 83 ] David W . McDonald and Mark S . Ackerman . 2000 . Expertise recommender : A flexible recommendation system and architecture . In Proceedings of the 2000 ACM Conference on Computer Supported Cooperative Work . Association for Computing Machinery , New York , NY , 231 \u2013 240 . DOI : https : / / doi . org / 10 . 1145 / 358916 . 358994 [ 84 ] Henry Mintzberg , Duru Raisinghani , and Andre Theoret . 1976 . The structure of \u201cunstructured\u201d decision processes . Administrative Science Quarterly 21 ( 1976 ) , 246 \u2013 275 . [ 85 ] Richar Van Noorden . 2014 . Global scientific output doubles every nine years . ( May 2014 ) . Retrieved 03 March , 2021 from http : / / blogs . nature . com / news / 2014 / 05 / global - scientific - output - doubles - every - nine - years . html . [ 86 ] R . Oppenheimer . 1956 . Analogy in science . American Psychologist 11 , 3 ( 1956 ) , 127 \u2013 135 . [ 87 ] Adam Paszke , Sam Gross , Francisco Massa , Adam Lerer , James Bradbury , Gregory Chanan , Trevor Killeen , Zeming Lin , Natalia Gimelshein , Luca Antiga , Junjie Bai , and Soumith Chintala . 2019 . Pytorch : An imperative style , high - performance deep learning library . arXiv : 1912 . 01703 . Retrieved from https : / / arxiv . org / abs / 1912 . 01703 . [ 88 ] Jeffrey Pennington , Richard Socher , and Christopher Manning . 2014 . Glove : Global vectors for word representation . In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing . Association for Compu - tational Linguistics , Doha , Qatar , 1532 \u2013 1543 . DOI : https : / / doi . org / 10 . 3115 / v1 / D14 - 1162 [ 89 ] Edwin A . Peraza - Hernandez , Darren J . Hartl , Richard J . Malak Jr , and Dimitris C . Lagoudas . 2014 . Origami - inspired active structures : A synthesis and review . Smart Materials and Structures 23 , 9 ( 2014 ) , 094001 . [ 90 ] Matthew Peters , Mark Neumann , Mohit Iyyer , Matt Gardner , Christopher Clark , Kenton Lee , and Luke Zettlemoyer . 2018 . Deep contextualized word representations . In Proceedings of the 2018 Conference of the North American Chap - ter of the Association for Computational Linguistics : Human Language Technologies . Association for Computational Linguistics , New Orleans , Louisiana , 2227 \u2013 2237 . DOI : https : / / doi . org / 10 . 18653 / v1 / N18 - 1202 [ 91 ] Kristopher J . Preacher and Andrew F . Hayes . 2004 . SPSS and SAS procedures for estimating indirect effects in simple mediation models . Behavior Research Methods , Instruments , and Computers 36 , 4 ( 2004 ) , 717 \u2013 731 . DOI : https : / / doi . org / 10 . 3758 / BF03206553 [ 92 ] Sekharipuram S . Ravi , Daniel J . Rosenkrantz , and Giri Kumar Tayi . 1994 . Heuristic and special case algorithms for dispersion problems . Operations Research 42 , 2 ( 1994 ) , 299 \u2013 310 . [ 93 ] Daniel M . Russell , Mark J . Stefik , Peter Pirolli , and Stuart K . Card . 1993 . The cost structure of sensemaking . In Pro - ceedings of the INTERACT\u201993 and CHI\u201993 Conference on Human Factors in Computing Systems . ACM , 269 \u2013 276 . [ 94 ] Neil Savage . 2016 . Nanofins Make a Better Hologram . ( 2016 ) . Retrieved from https : / / spectrum . ieee . org / tech - talk / semiconductors / optoelectronics / nanofins - make - a - better - hologram . [ 95 ] Tobias Schnabel , Paul N . Bennett , and Thorsten Joachims . 2019 . Shaping feedback data in recommender systems with interventions based on information foraging theory . In Proceedings of the 12th ACM International Conference on Web Search and Data Mining . 546 \u2013 554 . [ 96 ] Tobias Schnabel , Gonzalo Ramos , and Saleema Amershi . 2020 . \u201cWho doesn\u2019t like dinosaurs ? \u201d Finding and eliciting richerpreferencesforrecommendation . In Proceedingsofthe14thACMConferenceonRecommenderSystems . 398 \u2013 407 . [ 97 ] Pao Siangliulue , Joel Chan , Krzysztof Z . Gajos , and Steven P . Dow . 2015 . Providing timely examples improves the quantityandqualityofgeneratedideas . In Proceedingsofthe2015ACMSIGCHIConferenceonCreativityandCognition . 83 \u2013 92 . [ 98 ] Alison Smith - Renner , Ron Fan , Melissa Birchfield , Tongshuang Wu , Jordan Boyd - Graber , Daniel S . Weld , and Leah Findlater . 2020 . No explainability without accountability : An empirical study of explanations and feedback in inter - active ml . In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems . 1 \u2013 13 . ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 . 57 : 36 H . B . Kang et al . [ 99 ] L . Streeter and K . Lochbaum . 1988 . An expert / expert - locating system based on automatic representation of semantic structure . In Proceedings of the 4th Conference on Artificial Intelligence Applications . IEEE Computer Society , Los Alamitos , CA . DOI : https : / / doi . org / 10 . 1109 / CAIA . 1988 . 196129 [ 100 ] Ilya Sutskever , Oriol Vinyals , and Quoc V . Le . 2014 . Sequence to sequence learning with neural networks . In Proceed - ings of the Advances in Neural Information Processing Systems . Z . Ghahramani , M . Welling , C . Cortes , N . Lawrence , and K . Q . Weinberger ( Eds . ) , Vol . 27 . Curran Associates , Inc . Retrieved from https : / / proceedings . neurips . cc / paper / 2014 / file / a14ac55a4f27472c5d894ec1c3c743d2 - Paper . pdf . [ 101 ] Ilya Sutskever , Oriol Vinyals , and Quoc V . Le . 2014 . Sequence to sequence learning with neural networks . In Proceed - ings of the 27th International Conference on Neural Information Processing Systems\u2014Volume 2 . MIT Press , Cambridge , MA , 3104 \u2013 3112 . [ 102 ] John Sweller , Paul Chandler , Paul Tierney , and Martin Cooper . 1990 . Cognitive load as a factor in the structuring of technical material . Journal of Experimental Psychology : General 119 , 2 ( 1990 ) , 176 . [ 103 ] Jaime Teevan , Christine Alvarado , Mark S . Ackerman , and David R . Karger . 2004 . The perfect search engine is not enough : A study of orienteering behavior in directed search . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . Association for Computing Machinery , New York , NY , 415 \u2013 422 . DOI : https : / / doi . org / 10 . 1145 / 985692 . 985745 [ 104 ] ThermoCool . 2021 . SKIVED FIN HEAT SINKS . ( 2021 ) . Retrieved 03 March , 2021 from https : / / thermocoolcorp . com / project / skived - fins / . [ 105 ] Dustin Tingley , Teppei Yamamoto , Kentaro Hirose , Luke Keele , and Kosuke Imai . 2014 . Mediation : R package for causal mediation analysis . ( 2014 ) . Journal of Statistical Software 59 , 5 ( 2014 ) . [ 106 ] Peter D . Turney . 2008 . The latent relation mapping engine : Algorithm and experiments . Journal of Artificial Intelli - gence Research 33 ( 2008 ) , 615 \u2013 655 . [ 107 ] M . W . Van Someren , Y . F . Barnard , and J . A . C . Sandberg . 1994 . The think aloud method : A practical approach to modelling cognitive . London : AcademicPress ( 1994 ) . [ 108 ] Ashish Vaswani , Noam Shazeer , Niki Parmar , Jakob Uszkoreit , Llion Jones , Aidan N . Gomez , undefinedukasz Kaiser , and Illia Polosukhin . 2017 . Attention is all you need . In Proceedings of the 31st International Conference on Neural Information Processing Systems . Curran Associates Inc . , Red Hook , NY , 6000 \u2013 6010 . [ 109 ] SwaroopVattam , BryanWiltgen , MichaelHelms , AshokK . Goel , andJeannetteYen . 2011 . DANE : Fosteringcreativity in and through biologically inspired design . In Proceedings of the Design Creativity 2010 . Springer , 115 \u2013 122 . [ 110 ] Manuela M . Veloso and Jaime G . Carbonell . 1993 . Derivational analogy in PRODIGY : Automating case acquisition , storage , and utilization . In Proceedings of the Case - based Learning . Springer , 55 \u2013 84 . [ 111 ] James A . Waltz , Albert Lau , Sara K . Grewal , and Keith J . Holyoak . 2000 . The role of working memory in analogical mapping . Memory and Cognition 28 , 7 ( 2000 ) , 1205 \u2013 1212 . [ 112 ] Peter C . Wason and J St B . T . Evans . 1974 . Dual processes in reasoning ? Cognition 3 , 2 ( 1974 ) , 141 \u2013 154 . [ 113 ] Ryen W . White , Paul N . Bennett , and Susan T . Dumais . 2010 . Predicting short - term interests using activity - based search context . In Proceedings of the 19th ACM International Conference on Information and Knowledge Management . 1009 \u2013 1018 . [ 114 ] Ryen W . White and Resa A . Roth . 2009 . Exploratory search : Beyond the query - response paradigm . Synthesis Lectures on Information Concepts , Retrieval , and Services 1 , 1 ( 2009 ) , 1 \u2013 98 . [ 115 ] Rand R . Wilcox and H . J . Keselman . 2003 . Modern robust data analysis methods : Measures of central tendency . Psy - chological Methods 8 , 3 ( 2003 ) , 254 . [ 116 ] Michael Yovanovich , Richard Culham , and Peter Teertstra . 2004 . Calculating interface resistance . ( 2004 ) . Retrieved from http : / / www . thermalengineer . com / library / calculating _ interface _ resistance . htm . [ 117 ] Xiaolong Zhang , Yan Qu , C . Lee Giles , and Piyou Song . 2008 . CiteSense : Supporting sensemaking of research litera - ture . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . Association for Computing Machinery , New York , NY , 677 \u2013 680 . DOI : https : / / doi . org / 10 . 1145 / 1357054 . 1357161 [ 118 ] Shannon A . Zirbel , Mary E . Wilson , Spencer P . Magleby , and Larry L . Howell . 2013 . An origami - inspired self - deployable array . In Proceedings of the ASME 2013 Conference on Smart Materials , Adaptive Structures , and Intelligent Systems . American Society of Mechanical Engineers Digital Collection . Received 10 July 2021 ; revised 27 January 2022 ; accepted 31 March 2022 ACM Transactions on Computer - Human Interaction , Vol . 29 , No . 6 , Article 57 . Publication date : November 2022 .",
    "savinainenUsingBridgingRepresentation2005": "Using a Bridging Representation and Social Interactions to Foster Conceptual Change : Designing and Evaluating an Instructional Sequence for Newton\u2019s Third Law ANTTI SAVINAINEN Kuopio Lyseo High School , Puijonkatu 18 , FIN - 70110 Kuopio , Finland PHILIP SCOTT Centre for Studies in Science and Mathematics Education , School of Education , University of Leeds , LEEDS LS2 9JT , UK JOUNI VIIRI Department of Education , University of Joensuu , PL 111 , FIN - 80101 Joensuu , Finland Received 1 September 2003 ; revised 25 April 2004 ; accepted 15 June 2004 DOI 10 . 1002 / sce . 20037 Published online 19 November 2004 in Wiley InterScience ( www . interscience . wiley . com ) . ABSTRACT : This paper offers an account of , and \ufb01ndings from , an approach to designing and evaluating an instructional sequence set in the context of Newton\u2019s third law . The design of the sequence draws upon conceptual change theory and the concept of the \u201cbridging analogy\u201d is extended to introduce the notion of a \u201cbridging representation . \u201d Attention is also given in the instructional design to the proposed social interactions between teacher and students as the teaching and learning activities are played out or \u201cstaged\u201d in the classroom . A range of instruments is used to measure the extent of student learning , and evidence is presented to indicate that the designed sequence leads to enhanced learning gains when compared to those achieved with an equivalent group of students . Although set in the context of teaching and learning about mechanics a number of general points are made in relation to both instructional design and perspectives on conceptual change . C (cid:1) 2004 Wiley Periodicals , Inc . Sci Ed 89 : 175 \u2013 195 , 2005 INTRODUCTION Findings from many studies over the past three decades have shown that students have pre - instructional conceptions or beliefs about the phenomena addressed in science lessons ( see , for example , bibliographies in Duit ( 2004 ) ; McDermott & Redish ( 1999 ) ) . Often Correspondence to : A . Savinainen ; e - mail : antti . savinainen @ kuopio . \ufb01 C (cid:1) 2004 Wiley Periodicals , Inc . 176 SAVINAINEN ET AL . these student ideas differ from accepted scienti\ufb01c views and are resistant to change . This state of affairs poses a great challenge for science teaching : how can conceptual change be fostered ? Theories of conceptual change attempt to describe possible learning pathways from students\u2019 pre - instructional conceptions to the science conceptions to be learned ( Duit & Treagust , 2003 ) . Consequently , the design of instructional sequences can be informed by conceptual change theories and the research on students\u2019 pre - instructional conceptions , or alternative conceptions , in the target domain . However , it is not enough to pay attention only to the design of the instructional sequences : the teacher\u2019s role in staging those activities must also be considered ( Leach & Scott , 2003 ) . In this paper we show how an instructional sequence was designed and staged to improve science teaching , taking Newton\u2019s third law as an example . The design and evaluation is informed by the conceptual change literature , and the staging of the activities is informed by social perspectives on conceptual change . Firstly , we discuss what is involved in learn - ing Newtonian mechanics from a conceptual change perspective . Secondly , we present an instructional approach for teaching Newton\u2019s third law . Here we highlight the underlying principles of the design of the instructional sequence , relate our instructional approach to previous teaching approaches regarding the third law , and introduce the notion of a bridg - ing representation . Thirdly , we present an empirical study of the instructional approach in which its effectiveness in promoting learning is evaluated . We use the notion of contextual coherence to evaluate the extent of each student\u2019s conceptual change in this domain . The contextual coherence is a measure of the extent to which a student can apply a concept or a physical principle in a variety of familiar and novel situations : in such a way that the student\u2019s understanding is not adversely in\ufb02uenced by contextual features that are irrelevant from the physicist\u2019s point of view ( Savinainen & Viiri , 2003 ) . LEARNING THE NEWTONIAN CONCEPT OF FORCE : A CONCEPTUAL CHANGE PERSPECTIVE Many articles have been published on students\u2019 understandings of the force concept ( see , for example , Clement , 1982 ; Hake 1998 ; Halloun & Hestenes , 1985 ; Hestenes , Wells , & Swackhamer , 1992 ; Minstrell , 1982 ) . Several articles focus particularly on students\u2019 understandings of Newton\u2019s third law ( see , for example , Brown , 1989 ; Montanero et al . , 2002 ; Terry & Jones , 1986 ) . The \ufb01ndings commonly indicate that most students have a poor understanding of Newton\u2019s third law and of the force concept in general . In the light of such \ufb01ndings it appears that Newton\u2019s laws are dif\ufb01cult both to teach and to learn . Students tend to have a limited number of fundamental alternative conceptions about forces . They might think of forces as being things in themselves , as events , and as properties of objects ( Terry & Jones , 1986 ) . Perhaps the most common view among students is that of force as an innate or acquired property of objects , which implies that forces are not seen as arising from an interaction between objects . For example , Brown ( 1989 ) used a multiple - choice question in which the student is asked to compare the forces a 16 - pound bowling ball and a 4 - pound stationary pin exert on each other when the ball strikes the pin . Only 5 % of students answered the question correctly ( stating that the forces will be equal ) after a full year of traditional high school physics instruction . Most students seemed to think in terms of a \u201cdominance principle , \u201d where the bowling ball is \u201cmore forceful\u201d because it is moving , is heavier , and is more able to cause damage than the pin . Additionally , many students do not believe that an inanimate and inert object can exert a force . For instance , they may think that a table does not exert a force on a book lying on it , it is just \u201cin the way\u201d ( Minstrell , 1982 ) . The students\u2019 idea of force as an acquired property has close links with the pre - Galilean notion of impetus : it is conceived to be an inanimate \u201cmotive power\u201d or \u201cintrinsic force\u201d BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 177 that keeps things moving ( Hestenes , Wells , & Swackhamer , 1992 ; Sequira & Leite , 1991 ) . In this view , for an object to move it must be supplied with impetus . This is , of course , inconsistent with Newton\u2019s \ufb01rst law as well as with the character of forces arising due to interactions . Hestenes , Wells , and Swackhamer ( 1992 ) argue that the impetus concept of motion and the dominance principle are the most dif\ufb01cult and usually the last of the alternative conceptions to be overcome in the transition to Newtonian thinking . At a more general level , it can be stated that students\u2019 understandings of the force concept are very often context dependent , meaning that a student may show correct understanding in some exercises involving the force concept but fail to apply this in other contexts ( Bao et al . , 2002 ; Finegold & Gorsky , 1991 ; Palmer , 1997 ; Savinainen & Scott , 2002 ; Steinberg & Sabella , 1997 ) . Steinberg and Sabella ( 1997 ) argue that \u201cdifferent contexts and presentations can trigger different responses from a given student , even if the underlying physics is identical . \u201d Bao et al . ( 2002 ) identify four contextual features ( which they call physical features ) that students use in their reasoning regarding Newton\u2019s third law : velocity , mass , pushing , and acceleration . For instance , \u201cpushing\u201d can imply for some students that the object that \u201cpushes\u201d exerts a larger force . So these students recognize that both objects exert a force on each other but they fail to appreciate the fact that the forces arising from an interaction are always symmetrical . Furthermore , students may use combinations of different contextual features in their reasoning and may consider that the features have different levels of signi\ufb01cance for speci\ufb01c questions . We interpret this as a lack of contextual coherence ( Savinainen & Viiri , 2003 ) . These \ufb01ndings indicate that there is a big gap between everyday views ( or alternative con - ceptions ) and scienti\ufb01c views in the case of the force concept . According to the conceptual change literature , it seems that learning in this area requires a form of strong restructur - ing or reconceptualization ( Carey , 1985 ; Dykstra , Boyle , & Monarch , 1992 ; Tyson et al . , 1997 ) . The change from seeing force as an innate property to seeing forces as interactions between objects requires more than just extending existing ideas , it requires changing points of view . Furthermore , it seems that these changes often involve revisions to basic beliefs about the nature of the world : for instance , that inert and inanimate bodies can exert forces ( Minstrell , 1982 ) . Chi , Slotta , and de Leeuw ( 1994 ) refer to such conceptual changes in terms of ontological shifts . The notion of force as an innate property of an object , for in - stance , falls into the ontological category of \u201cmatter , \u201d whereas the scienti\ufb01c view of force as being due to interactions belongs to the category of \u201cprocesses . \u201d According to Chi and colleagues , conceptual change takes place when a concept is shifted between categories ( in this case , when the learner sees forces in terms of interactions and processes rather than in terms of object properties and matter ) . In summary , we see the learning problem in this area as one which entails a signi\ufb01cant movement from everyday to scienti\ufb01c views . In other words , it entails a big learning demand ( Leach & Scott , 2002 ) . Next we discuss how this demand is to be addressed through instruction . AN INSTRUCTIONAL APPROACH FOR TEACHING NEWTON\u2019S THIRD LAW Underlying Principles According to Leach and Scott ( 2002 ) , instructional aims can be identi\ufb01ed by comparing students\u2019 everyday ways of knowing and the scienti\ufb01c ideas to be taught . The differences between everyday and scienti\ufb01c ways of thinking and talking give rise to a \u201clearning de - mand\u201d that is speci\ufb01c to an individual learning in particular domains of science ( Leach & 178 SAVINAINEN ET AL . Scott , 1995 ) . The learning demand may be due to differences in the conceptual tools used , differences which relate to ontological assumptions , and differences in the epistemological underpinnings of the knowledge being used . Based on this approach , instructional design starts with identi\ufb01cation of the school science knowledge to be taught , and consideration of how this area of science is conceptualized in the everyday language of students . Then follows the learning demand analysis , which involves identifying the nature of any differences between the everyday and scienti\ufb01c views . The aspects of school science to be taught and students\u2019 typical views of the force concept are compared in Table 1 ( this learning demand analysis focuses on the aspects which are relevant to Newton\u2019s third law ) . The next step is to develop an instructional sequence , involving a sequence of activities plus information about the staging of those activities , to address each aspect of the learning demand . The staging involves all of the interactions between teacher and students and these are multimodal in nature ( see Kress et al . , 2001 ) since the ideas are communicated through talk , various images , representations , gestures and so on . By these means the school science point of view is gradually developed ( see Mortimer & Scott , 2003 ) during the sequence of lessons with the aim of rendering it intelligible and plausible to the students . To achieve this , the teacher needs to be aware of the existing understandings of the students and to develop convincing lines of argument to engage with those existing understandings . Social perspectives on conceptual change have identi\ufb01ed peer discussions as playing a potentially important role in helping students to explore meanings ( Jones & Carter , 1998 ) . Moreover , group discussions make it possible for students to express their ideas and beliefs , helping them to increase their metaconceptual awareness ( Mason , 1998 ; Vosniadou et al . , 2001 ) . Mason ( 1998 ) argues that in peer discussions the authority for learning and knowing is shared between teacher and students and that this enhances the intrinsic motivation for learning ( see also , Pintrich , Marx , & Boyle , 1993 ) . Review of Previous Approaches to Teaching Newton\u2019s Third Law Most textbooks introduce Newton\u2019s third law after Newton\u2019s \ufb01rst and second laws ( see , for instance , Giancoli , 1998 ; Halliday , Resnick , & Walker , 2001 ) . This order does not give a central role to forces as interactions , which we have identi\ufb01ed as being the crucial feature TABLE 1 Learning Demand Analysis : The Force Concept Aspects of School Science To Be Addressed Students\u2019 Typical Everyday Views Ontological aspect Force is the property of an interaction between two objects . Force is an innate or acquired property of objects ( impetus ) . Conceptual aspect Interaction between two objects implies that they exert forces on each other : forces always come in pairs . Inert or inanimate objects cannot exert forces . Epistemological aspect The notion of symmetrical interaction between two objects ( i . e . , Newton\u2019s third law ) is generally applicable to all situations . Newton\u2019s third law is used in some situations but not others ( where , for example , the dominance principle may be applied ) depending on the contextual features of the situation at hand . BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 179 of the force concept . Moreover , a careful conceptual analysis reveals that Newton\u2019s third law is more complex than one might assume from the standard statement ( Giancoli , 1998 , p . 83 ) : \u201cWhenever an object exerts a force on a second object , the second object exerts an equal and opposite force on the \ufb01rst . \u201d It should be noted that this statement does not explicitly address the notion of interaction . Asthenotionofinteractioniscentraltounderstandingtheforceconcept , someresearchers recommend that the force concept might be more effectively taught by emphasizing forces as interactions between objects ( Brown , 1989 ; Hellingman , 1989 , 1992 ; Jim\u00b4enez & Perales , 2001 ; Reif , 1995a ; Reif & Heller , 1982 ) . For instance , Reif ( 1995a ) recommends analyzing a physical system by describing both motion and interactions . He argues that identifying interactions before forces helps to avoid the inclusion of nonexistent forces . Just \u201ctelling\u201d students that forces arise due to interactions is not going to be very effec - tive . This became clear in our earlier study that was very successful in addressing many dimensions of the force concept , but not Newton\u2019s third law ( Savinainen & Scott , 2002 ) . In that study , many students had serious dif\ufb01culties , after being taught , in generalizing from Newton\u2019s third law to both the accelerated and uniform velocity cases , with many students believing that Newton\u2019s third law does not hold in a dynamic situation . In other words , they lacked contextual coherence in applying Newton\u2019s third law . It is worth noting that the force concept was introduced as a measure of the strength of interaction at the very beginning of the teaching in that study but this idea was not systematically used and developed in the teaching . We believe that this was a signi\ufb01cant reason why many students failed to understand Newton\u2019s third law . Some research studies have demonstrated that analogies or visualizable models can be helpful in teaching Newton\u2019s third law . For instance , Jim\u00b4enez and Perales ( 2001 ) report that the symbolic representation of interactions ( SRI ) notation signi\ufb01cantly improved students\u2019 comprehension of the third law ( the evidence for this was published in Spanish : Jim\u00b4enez & Perales , 1999 ) . In the modeling method of instruction a similar ( but not identical ) rep - resentational tool is referred to as a \u201csystem schema\u201d ( Hestenes , 1996 ; Turner , 2003 ) . Another example is provided by Clement et al . ( 1987 ) who used the simple compression of a spring as a successful anchoring example in teaching Newton\u2019s third law ( see also , Camp & Clement , 1994 ) . The use of bridging analogies and how they link to our notion of a bridging representation is discussed in the next section . Bridging Analogies and Bridging Representations in Promoting Conceptual Change Analogies and \u201canchoring examples\u201d that draw upon students\u2019 intuitive knowledge ( Brown , 1992 ; Brown & Clement , 1989 ; Treagust , Harrison , & Venville , 1996 ) appear to be powerful tools for learning science , though there are also studies which show that they are not always successful in bringing about conceptual change ( Duit et al . , 2001 ) . Brown and Clement ( 1989 ) provide an example of building on students\u2019 existing ideas through a \u201cbridging strategy\u201d which involves four steps : 1 . A target question is used to make explicit students\u2019 alternative conceptions relating to the topic under consideration ( e . g . , forces acting on a book on a table ) . 2 . An analogous case , an anchoring example , is suggested by the instructor ( e . g . , a hand pushing down on a spring ) . 3 . The instructor asks students to make an explicit comparison between the anchor and target cases in order to establish the analogy relation . 180 SAVINAINEN ET AL . 4 . If students do not accept the analogy , the instructor attempts to \ufb01nd an intermediate bridging analogy between the target and anchor ( e . g . , a book on top of a spring , and then on top of a noticeably \ufb02exible board ) . This speci\ufb01c example of a bridging analogy ( using the springy piece of wood ) utilizes a physical system to prompt an intermediate step in thinking . The bridge can also be in the form of a diagrammatic representation . The symbolic representation of interactions ( SRI ; developed by Jim\u00b4enez and Perales , 2001 ) makes identi\ufb01cation of the mechanical interaction between pairs of objects explicit ( this is elaborated later in this paper ; see Figure 2 ) . This can be contrasted with a standard diagrammatic representation in physics instruction\u2014 the free - body diagram\u2014which concentrates on the forces acting on one target object , and consequently does not make the concept of interaction explicit to students . In our view the SRI diagram provides a bridge linking concrete physical situations , more abstract free - body diagrams involving vector notation and the equations of Newton\u2019s laws . In this sense we see the pedagogic function of the SRI diagram as providing a bridge , which we refer to as a \u201cbridging representation\u201d ( Figure 1 ) . The notion of bridging representation is , we believe , very useful and has close links with the concept of a bridging analogy . It plays a central role in our instructional approach to the force concept in general and Newton\u2019s third law in particular . Next we turn to explaining how the bridging representation permits us to emphasize forces as interactions in an instructional sequence on the force concept for upper secondary school students . The Structure and Staging of the Instructional Sequence The instructional sequence includes three general features \ufb01rst developed in an earlier study ( see Savinainen & Scott , 2002 ) . Firstly , there is a strong conceptual focus : before problem solving a thorough qualitative treatment was implemented , as recommended by Van Heuvelen ( 1991 ) . Secondly , the students were provided with plenty of opportunities to explore meanings in carefully framed peer discussions monitored by the teacher . This is in line with research \ufb01ndings which show that teachers have a crucial role to play in promoting productive discussions ( Meloth & Deering , 1999 ; Mercer & Wegerif , 1999 ) . Thirdly , the use of multiple representations and linking between them was explicitly developed throughout the teaching . There is a growing recognition in science education that understanding and conceptually integrating multiple representations play a crucial role in learning scienti\ufb01c concepts ( e . g . , Goldman , 2003 ; van Someren et al . , 1998 ) . Furthermore , Hestenes ( 1996 ) argues that students\u2019 ability to understand physics depends on the representational tools at their disposal . We have used the notion of representational coherence in characteriz - ing students\u2019 ability to use multiple representations correctly and to move between them ( Savinainen & Viiri , 2003 ) . At the start of the instructional sequence , the concept of force was introduced by the teacher ( author AS ) in the context of contact interaction . The students were asked to press down on a table with their thumbs and to observe what happened . This touching was Figure 1 . The SRI diagram as a bridging representation . BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 181 characterized as an interaction between thumb and table . The students were also asked to press their books and notebooks to \ufb01nd out if they too were deformed . This made it easier to believe that the table does deform in an interaction with another object . Minstrell ( 1982 ) has shown that this is far from evident to students . Next the students were asked to press the table gently and then hard , and to observe if there were any changes in the deformation of their thumb . This simple activity gave them a sense that the strength of an interaction can vary and at this point \u201cforce\u201d was de\ufb01ned as a measure of the strength of the interaction . An ordinary scale and a spring balance were then introduced as tools to measure the strengths of interaction , i . e . forces . These activities can be interpreted as building on students\u2019 existing ideas and extending them to new domains in the spirit of a bridging analogy . Next the SRI diagram was introduced as a tool to represent interactions . In the diagram double - sided arrows are used to show an interaction between two objects . It was emphasized that both objects participate in the interaction and that the interaction is symmetrical . In other words , as the force is a measure of the interaction , the same amount of force is necessarily exerted on both objects . As an example of the use of the SRI diagram , Figure 2 illustrates the situation of a hand pushing a block on the horizontal surface of a table . Figure 3 presents a possible corresponding free - body diagram . In the SRI diagram , the block is located at the center because it is the target object in this example . It should be noted that this use of SRI diagrams is not identical to that in the original paper ( Jim\u00b4enez & Perales , 2001 ) . In our version , the perpendicular and parallel components of contact interactions are identi\ufb01ed , to facilitate a better correspondence with free - body diagrams . While the division between contact and distance forces ( or \ufb01eld forces ) is not justi\ufb01ed in the context of modern physics , we believe that it is a useful distinction in an introductory course on mechanics ( Physics Textbook Review Committee , 1998 ) . In Figure 2 the double - headed arrows emphasize that interactions are always symmetrical . This was reinforced in the teaching by explicating all of the interactions in the diagram . For instance , \u201cthe interaction between the hand and the block is symmetrical , hence the Figure 2 . A SRI diagram of a block sliding on a horizontal table . Interaction type is indicated by C or D ( contact or distant interaction ) . 182 SAVINAINEN ET AL . Figure 3 . A free - body diagram representing a block sliding on a horizontal table with constant acceleration . Friction , normal force , and gravitational forces are denoted by the symbols used in Giancoli ( 1998 ) . Directions of velocity and acceleration are also shown . force exerted by the hand on the block has the same magnitude as the force exerted by the block on the hand , but the opposite direction . \u201d The same SRI diagram represents the situation in which the block remains at rest ( friction would be static ) . Furthermore , it is not possible to tell whether the block is accelerating or not . This point was stressed by asking the students to draw SRI diagrams representing the block moving at constant velocity and with decreasing / increasing velocity . This was an illuminating moment for some students , who realized that there was no difference whatsoever in terms of the SRI diagrams . It illustrates the point that it is crucial to consider the same concept from a number of different\u2014but related\u2014perspectives in order to develop a coherent understanding of forces as interactions ( Maloney , 1990 ) . Magnitudes of forces cannot be deduced from the SRI diagram , whilst free - body dia - grams , which were introduced after the SRI diagrams , show the magnitudes of interactions via the length of force vectors ( Figure 3 ) . A free - body diagram allows determination of the direction of possible acceleration from the sum of the force vectors , i . e . the net force . The students were required to draw both SRI and free - body diagrams for multiple situations and were asked to compare the diagrams in peer discussions . This approach provided students with plenty of opportunities to explore meanings together . After the pairs had reached a con - sensus on the diagrams and their verbal explanations , the teacher provided his explanations and checked how many of the pairs had a correct solution and correct explanation . Using the SRI diagram makes it possible to address all the aspects of the learning demand ( see Table 1 ) . The SRI diagram provides a visualizable tool for identifying and representing interactions between objects , which helps students to perceive forces as the property of an interaction instead of a property of an object ( ontological aspect ) . It also shows by means of the double - headed arrows that an interaction between two objects is symmetrical ( conceptual aspect ) . Furthermore , applying the SRI diagram in a variety of situations helps students to realize that Newton\u2019s third law is indeed valid in all situations regardless of contextual features ( epistemological aspect ) . The use of SRI diagrams was discontinued around the middle point of teaching the force concept , since it was felt by the teacher ( author AS ) that the students had gained enough experience of thinking of forces as interactions , and the bridging representation was no longer needed . However , interactions were addressed throughout the course when analyzing free - body diagrams : the students were asked to identify all objects touching the target object , since touching implies interaction . BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 183 DESIGN OF THE STUDY Teaching Context and Research Questions The study group consisted of Finnish students ( aged 16 , n = 23 ) following a preparatory International Baccalaureate program during school year 2002 \u2013 2003 . A very similar pilot group ( n = 22 ) was involved in an earlier study by Savinainen and Scott ( 2002 ) during 2000 \u2013 2001 . The students in the pilot group and those in the study group had encountered mechanics in their lower secondary school studies . Both groups had available to them an American physics textbook based on algebra and trigonometry ( Giancoli , 1998 ) , and they were taught by the same teacher ( author AS ) . All teaching took place through the medium of English . The study group utilized the SRI diagram whereas the pilot group did not ; this was the essential difference in the teaching of the groups . Partially due to the introduction of a new tool ( the SRI diagram ) , slightly more time was devoted to teaching the force concept to the study group than to the pilot group . Our research questions are as follows : 1 . What is the effect of the instructional sequence , which utilizes the SRI Bridging Representation , on the development of the study group students\u2019 ability to apply Newton\u2019s third law in a range of contexts ( i . e . , on their contextual coherence in applying Newton\u2019s third law ) ? 2 . How do the results of the study group compare with those of the pilot group ? Data - Gathering Instruments Since we believe that a thorough evaluation of conceptual understanding should be broadly based , several instruments were used in this study to probe students\u2019 contextual coherence regarding Newton\u2019s third law , i . e . their ability to apply the third law in differ - ent contexts . These instruments included multiple - choice questions , an extended written response question , and interview questions . Multiple - Choice Questions . The \u201cforce concept inventory\u201d ( FCI ) has four multiple - choice questions addressing Newton\u2019s third law ( Halloun et al . , 1995 ) . Two of the questions have the same context of a car pushing a truck ( questions 15 and 16 ) . Question 4 addresses a collision between a large truck and a small car . Question 28 involves two students sitting on identical of\ufb01ce chairs and the heavier student pushes the other with his feet . This question involves two contextual features\u2014mass and pushing\u2014at the same time . Bao et al . ( 2002 ) criticize this question , arguing that if a student answers that the forces are not equal , it is impossible to know ( without asking the students ) whether an incorrect response is generated through consideration of mass , of pushing , or of both . To study the effects of the above - mentioned contextual features , Bao et al . ( 2002 ) de - veloped their \u201cSurvey on Newton\u2019s Third Law . \u201d Each question in it measures students\u2019 reasoning related to a single contextual feature of the third law . The effects of velocity , mass , pushing , and acceleration are evaluated with three questions for each feature . In ad - dition there are four questions involving other contextual features , making 16 questions altogether . Bao et al . validated the instrument by interviewing students . Figure 4 provides a sample question from the survey that involves only the contextual feature of velocity . The correct answer is shown in bold print . The Survey contains some questions that are similar to , but not identical with , those in another research - based multiple - choice test on the force concept , the \u201cforce and motion conceptual evaluation\u201d ( FMCE ) ( Thornton & Sokoloff , 1998 ) . The FMCE has ten questions 184 SAVINAINEN ET AL . Figure 4 . A sample question from the survey on Newton\u2019s third law ( Bao et al . , 2002 ) . This question involves only the contextual feature of velocity . The correct answer is shown in bold print . on the third law , which introduce variations of contextual features of mass , velocity , and acceleration within three main contexts . All of these multiple - choice questions ( FCI , Survey on Newton\u2019s Third Law , and FMCE ) were used in this study ( see Table 2 ) . Extended Written Response Question . In addition to the multiple - choice questions , an extended written response question was used to gain insights into the students\u2019 reasoning ( Figure 5 ) . The main contextual feature of the written question is mass , though the car TABLE 2 Research Instruments Administered to the Study Group Research Instrument Pilot Group Study Group Comments Pre - FCI ( in Finnish ) Yes Yes At the beginning of teaching FMCE Yes Yes Two months after the pre - FCI Post - FCI ( in English ) Yes Yes After completing the teaching of the concept of force ( a month after the FMCE ) Interview questions No Yes Five students were interviewed a week after the post - FCI Extended Response written question Yes Yes A month after the post - FCI Survey on Newton\u2019s third law No Yes After completion of teaching the law of conservation of momentum , which was derived using Newton\u2019s third law ( three months after the post - FCI ) BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 185 Figure 5 . The extended response written question used in the study ( Reif , 1995b , p . 86 ) . and the bug have different magnitudes of velocities as well . The written question involves identifying the correct answer and justifying it . Essentially the same question was answered by the pilot group in the earlier study ( Savinainen & Scott , 2002 ) . Interview Questions . Five students from the study group were chosen for interview on the basis of their success in the early part of the course : they represented the top , middle , and lower levels of the group . A semistructured interview was used and the students were asked to \u201cthink aloud\u201d whilst answering the interview questions , being free to change their answers during the interview if they so wished . The interviews were conducted by author AS . The interview questions relating to Newton\u2019s third law are presented in Figure 6 . In these questions , parts ( a ) and ( b ) have the same general context but the state of the system is changed . Part ( a ) addresses constant velocity , whereas part ( b ) addresses varying velocity ( acceleration ) . From the physics point of view , these questions resemble the situations in the FCI questions 15 and 16 , where the car is pushing a truck with increasing and constant velocity , respectively . Administration of the Research Instruments to the Study Group A Finnish translation of the 1995 version of the Force Concept Inventory ( Halloun , et al . , 1995 ; Koponen , Jauhiainen , & Lavonen , 2000 ) was administered prior to instruction because the students\u2019 competence in English was not suf\ufb01ciently good at the beginning of the program . The students were allowed to see the correct answers after the instruments had been administered but all question and answer materials were subsequently collected up . \u2217 After the pretest , all instruments and questions were in English . An outline schedule for the administration of the research instruments , over a period of six months , is presented in Table 2 . The pilot group students in the earlier study answered the pre - FCI , FMCE , and post - FCI , and essentially the same written question . Measure of the Students\u2019 Contextual Coherence It is possible for a group of students to achieve a relatively high average score for the multiple - choice questions used , even if only a few students correctly answer all of the questions addressing Newton\u2019s third law . Hence , the following measure ( based on four \u2217 Of course , they were not allowed to see the correct answers of the Pre - FCI . 186 SAVINAINEN ET AL . Figure 6 . The interview questions ( derived from McDermott et al . , 1998 , pp . 28 \u2013 29 ) . criteria ) of each student\u2019s contextual coherence was used . A student exhibits contextual coherence if 1 . all four FCI questions addressing the third law are correctly answered ; 2 . at least 9 out of 10 FMCE questions on the third law are correctly answered ( one lapse was allowed since it could be due to carelessness ) ; 3 . the written question has the correct answer with correct reasoning ; 4 . at least 15 out of 16 Survey questions are correctly answered . The above criteria were applied to the different instruments separately . This measure of contextual coherence is taken as an indicator of conceptual change . Because FCI question 15 concerning a dynamic situation proved to be hard for many pilot group students , a written explanation of the reasoning accompanied this question in the post - FCI of the study group . A short text was added to the Survey as well : \u201cExplain brie\ufb02y how you arrived at your answers and what physical principle ( s ) you applied . \u201d RESULTS Multiple - Choice Tests The averages of the FCI and FMCE results are presented in Table 3 . The statistical signi\ufb01cance of possible differences between the groups was tested with the nonparamet - ric Mann - Whitney U test of two independent samples . ( A nonparametric test was used rather than the t - test because the data did not follow the normal distribution . ) There is no BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 187 TABLE 3 Newton\u2019s Third Law Results for the Pilot Group ( n = 22 ) and Study Group ( n = 23 ) N3 Law Tests Pilot Group Study Group ( No . of Questions ) Average ( SD ) Average ( SD ) p - value Pre - FCI ( 4 ) 23 . 9 ( 30 . 4 ) 30 . 4 ( 23 . 8 ) 0 . 194 FMCE ( 10 ) 62 . 2 ( 28 . 1 ) 78 . 7 ( 23 . 8 ) 0 . 038 Post - FCI ( 4 ) 75 . 0 ( 28 . 9 ) 93 . 5 ( 13 . 5 ) 0 . 012 Survey ( 16 ) \u2013 98 . 6 ( 4 . 6 ) \u2013 p - value is calculated using Mann - Whitney U - test for two independent samples . statistically signi\ufb01cant difference between the pre - FCI test results , whereas the FMCE and post - FCI results of the study group are better ( p - values are below 0 . 05 ) . While the averages provide a way of comparing differences between the groups , they do not directly reveal information about the students\u2019 contextual coherence . Figure 7 presents the percentages of students who demonstrated contextual coherence , from different in - struments , according to the criteria given in the preceding section . The pilot group did not answer the questions in the Survey on Newton\u2019s third law . The study group students\u2019 written explanations in the Survey all supported the correct answers . In the pilot group , two students answered all the FCI questions on Newton\u2019s third law correctly at the beginning of the teaching , whereas in the study group none did . There is a clear line of progression in the development of contextual coherence in the study group and most of the students reached contextual coherence at the level of identi\ufb01cation , i . e . they were able to identify the correct answers from non - Newtonian alternatives . Only two study group students failed to answer all the Survey questions correctly . These two had only a few wrong answers but they revealed that these students , just like many in the pilot group , were still struggling to differentiate between Newton\u2019s second and third laws . This conclusion was supported by their written justi\ufb01cations accompanying the Survey . For instance , one student who incorrectly answered two questions addressing a dynamical situation wrote ( the student\u2019s language has not been edited ) : Ifthereisnoacceleration , there\u2019snonetforce , massdoesn\u2019tmatter . Whenyoupushsomeone with some amount of force , other one pushes you with the same amount force . Figure 7 . Percentage of students exhibiting contextual coherence in the questions addressing Newton\u2019s third law for the pilot and study groups . The tests are in chronological order . 188 SAVINAINEN ET AL . It seems that this student thinks that Newton\u2019s third law is valid only if there is no acceleration . Extended Response Written Question Students\u2019 responses to the extended response written question support the \ufb01ndings from the multiple - choice tests since 82 % of students in the study group answered it correctly with correct reasoning . To give some idea of the students\u2019 answers to the written question , we present one correct and one incorrect response from the study group ( the students\u2019 language has not been edited ) . Correct response ( underlining done by the student ) : According to Newton\u2019s third law , forces always come as pairs and objects interacting always exert equal amount of force on each other , no matter what . So the car and the bug exert equal magnitude of force on each other . This response shows a clear understanding of forces as interactions and recognizes that contextual features of an interaction are irrelevant from the point of view of Newton\u2019s third law . Incorrect response : As it says in N III Law , when two object interacts they exert equal amount of force in each other only magnitude varies . In this case magnitude of force exerted on the bug by the car is larger than the magnitude of the force exerted by the bug . The student recalls Newton\u2019s third law correctly but she interprets the term \u201csame amount\u201d incorrectly , enabling her to hold both Newton\u2019s third law and the dominance principle at the same time . Interview Questions Given that all \ufb01ve students interviewed from the study group had correctly answered the four questions on Newton\u2019s third law in the post - FCI a week earlier , the interview results were surprising . Three students out of \ufb01ve answered both questions correctly . One student who ended up with the correct answer was unsure about the Newton\u2019s third law force pair in the case of gravitational force : Butagain , there\u2019sthe . . . Thatmean , meangravitationalforce . Something . . . Ohwell . . . The amount of force the gravitational force exerts on the rocket , it exerts back on the , on the gravitational force . But that doesn\u2019t make any sense . When , when I put it like that . But , could the proper explanation be that it exerts the same amount of force on the Earth ? The fourth student answered both questions incorrectly in the same way . She stated that the normal force and gravitational force exerted on the crate inside the rocket constitute a Newton\u2019s third law force pair regardless of the state of motion . She recalled a slogan used during the lessons \u201cyou can\u2019t touch without being touched , \u201d but this did not help her in answering the questions . The \ufb01fth student did not seem to understand the question at all . When prompted she recalled relevant aspects from the third law but she was unable to apply these aspects in the interview questions : BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 189 Was it that , umm , they come in pairs and they\u2019re symmetrical and . . . [ and a bit later ] I think it was , umm , that forces come in pairs and they\u2019re symmetrical . And , umm , that . . . Was it that you can\u2019t touch without being touched ? It is worth noting that in contrast to the multiple - choice questions , the interview questions went beyond identifying magnitudes of a given interaction force pair . The questions were posedinacontextthatalsorequiredtheuseofNewton\u2019s\ufb01rstandsecondlaws . Inaddition , the interview questions addressed Newton\u2019s third law in the case of a noncontact gravitational force , which was not addressed in any other instruments used in this study . DISCUSSION Evaluation of the Results The \ufb01rst research question addresses the development of the study group students\u2019 con - textual coherence concerning Newton\u2019s third law . Both the study and pilot groups followed the same teaching approach but in the case of the study group the notion of forces as in - teractions was emphasized , through the use of the SRI bridging representation . Learning outcomes were measured in terms of contextual coherence : the student was required to answer correctly almost all the questions , which involved different contexts . The results demonstrate that the instructional approach proved to be very successful in promoting con - textual coherence of the third law , at least at the level of identi\ufb01cation : the students could identify the correct answers from non - Newtonian alternatives and justify their reasoning . The second research question addresses possible differences between the study and pilot groups : the study group students\u2019 results were better than those of the pilot group , the differ - ence being statistically signi\ufb01cant . Similar results have been achieved by author AS with par - allel groups of students following the Finnish upper secondary school syllabus ( Savinainen , 2004 ) . Our \ufb01ndings therefore provide strong support for Jim\u00b4enez and Perales ( 2001 ) , who reported that the SRI notation signi\ufb01cantly improved their students\u2019 comprehension of the third law . Further support can be found in the Finnish syllabus \u201cbaseline\u201d study of post - FCI results . Almost 400 Finnish upper secondary students from different parts of Finland were sampled , and it was found that only 28 % correctly answered the questions on Newton\u2019s third law in all the various contexts addressed by the FCI ( Jauhiainen , Koponen , & Lavonen , 2001 ) . In the light of these results , even the instructional approach for the pilot group , which did not utilize the SRI diagrams , can be considered quite successful , since 45 % of those students exhibited contextual coherence in Newton\u2019s third law at the end of the course . Minstrell ( 1982 ) showed that students may have dif\ufb01culties with the idea of inanimate objects , such as a table , exerting forces . The students in this study had no trouble in accept - ing the idea once they started thinking of forces as interactions . The dominance principle was prevalent in the students\u2019 responses regarding the third law at the beginning of the course . At the end of the course , almost all the students recognized that an interaction is symmetrical regardless of contextual features in the given tasks ( i . e . , they exhibited con - textual coherence in this respect ) . They had , however , much more dif\ufb01culty with Newton\u2019s second law ( Savinainen , 2004 ) . Hence , our \ufb01ndings suggest that the dominance principle is not necessarily \u201cone of the last misconceptions to be overcome in transition to Newtonian thinking\u201d ( Hestenes , Wells , & Swackhamer , 1992 ) . It appears that the prevalence of the dominance principle depends on the instructional sequence . Dykstra , Boyle , and Monarch ( 1992 ) present a successful instructional sequence in which Newton\u2019s \ufb01rst and second laws are dealt with before the third law . Given their instructional sequence , it is likely that students\u2019 would sort out the third law last . 190 SAVINAINEN ET AL . Evaluation of Research Design and Methodology In this study , there are questions that need to be addressed regarding the possible causes of the learning gains observed . Leach and Scott ( 2002 ) conclude that it is not legitimate to claim that \u201cone instructional sequence is better than another in promoting student learning , without taking into account the part played by the teacher . \u201d In this study most of the factors affecting the learning outcomes were kept as constant as was possible : the same experienced teacher ( author AS ) taught approximately the same course to similarly selected preparatory year International Baccalaureate students . Also , almost all the research instruments used for the evaluation of student learning were the same . The major difference between the two instructional approaches was the use of SRI diagrams directly demonstrating the simulta - neous and symmetric nature of interactions . Another difference was linguistic : the terms \u201caction\u201d and \u201creaction\u201d were not employed in the revised approach , as recommended by Hellingman ( 1989 ) , whereas these terms were mentioned but not systematically used with the pilot group . There are also concerns relating to the use of multiple - choice questions , which require only identi\ufb01cation of correct answers . It could be argued that the study group students just learned to do this without real understanding . Given that the students completed many similar types of multiple - choice tests , this is a reasonable worry . However , there are several reasons why the concern is not warranted : 1 . During the course , the pilot group students answered almost as many questions on the third law as the study group students , and saw the correct answers to the FMCE afterwards , but this was not enough to change the incorrect views of over half of the group . 2 . The extended response written question and the written parts of the multiple - choice tests were used to gain insight into students\u2019 reasoning and overall student\u2019s written justi\ufb01cations supported their choices in the multiple - choice tests . 3 . Data were collected over a relatively long time span ( six months ) . An additional point of concern is that the multiple - choice questions used do not address all aspects of Newton\u2019s third law . For instance , they do not test whether a student understands that forces arising from an interaction between two objects are always exactly opposite in direction . Furthermore , the questions involve only verbal representation . There is some evidence that students may not master the third law in vector diagram representation equally well as in verbal representation ( Meltzer , 2002 ) . The interview results suggest that the multiple - choice questions used have limitations : in general , these questions seem to do a good job at the level of identifying the correct answer , but they do not necessarily predict students\u2019 understanding in more complex situations . Other interviews conducted by author AS with different groups of students lend support to this suggestion ( Savinainen , 2004 ) . However , one could also question the extent to which the interviews provide reliable knowledge on students\u2019 thinking . For instance , Schoultz , S\u00a8alj\u00a8o , and Wyndham ( 2001 ) provide evidence suggesting that students\u2019 responses in in - terview studies should be regarded as situated and dependent on the tools available as resources for reasoning . Furthermore , Welzel and Roth ( 1998 ) conclude that interviews can only provide clues on ongoing , dynamic cognitive processes . In this study , the interview questions deliberately avoided referring to the concept of interaction or to the SRI diagram : hence , the interview situation did not make the above - mentioned tools readily available for the students . From this perspective , it could be speculated that the outcome of the inter - views could have been different had these tools been cued in the interviews . Indeed , it is BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 191 regrettable that only \ufb01ve students were interviewed , and only once , in the case of Newton\u2019s third law ( the same students were interviewed several times on Newton\u2019s \ufb01rst and second laws ) . Nevertheless , our study lends support to the claim made by Dufresne , Leonard , and Gerace ( 2002 ) that multiple sources of information about students\u2019 knowledge and under - standing are needed if we are to obtain a reliable assessment of students\u2019 knowledge and understanding . Whilst acknowledging all of these questions relating to methodological issues , we would nevertheless support the proposition that the instructional approach developed here , based on the SRI representation , offers an effective way of teaching Newton\u2019s third law . Why Was the Bridging Representation Helpful ? It is interesting to re\ufb02ect on the role of the bridging representation in fostering con - ceptual change . There are several possible reasons why the SRI diagram\u2014the bridging representation\u2014was effective in supporting conceptual change in the case of Newton\u2019s third law . Firstly , it helps students to perceive forces as symmetric interactions and provides the means to identify other objects with which the target object is interacting . Secondly , the potential value of the bridging representation can be seen in terms of diSessa\u2019s ( 1993 ) phenomenological primitives ( p - prims ) . The p - prims or core intuitions are not misconcep - tions as such but they can bring about a misconception when applied in physical situations ( Brown , 1993 ) . The most relevant p - prim here is \u201cmore agency means more effect\u201d : the dominance principle can be interpreted as a manifestation of this particular core intuition or p - prim when applied in , say , collisions . As pointed out earlier , almost all the students in this study gave up the dominance principle after utilizing the bridging representation . This suggests that the bridging representation helps to override the core intuition related to the dominance principle . On the other hand , this core intuition ( \u201cmore agency means more effect\u201d ) has a meaningful function in analyzing , for instance , a collision between a heavy truck and a light car : the forces on both vehicles have the same magnitude but the resulting accelerations and damages are not the same , since the masses and the internal structures of the vehicles are different . Thirdly , the bridging representation provides a complementary approach to free - body diagrams , with the potential to support the construction of deeper understandings since information is integrated by the students from more than one representation ( Ainsworth , 1999 ) . Treagust , Chittleborough , and Mamiala ( 2003 ) , for instance , found that effective learning in chemical explanations requires simultaneous use of multiple representations . In the instructional approach developed in this study , students were encouraged to combine diagrammatic representations ( the SRI and free - body diagrams ) with verbal representations , thereby helping to promote representational coherence . It would also be reasonable to assume that the bridging representation fosters students\u2019 ability to identify the forces that must be included in the free - body diagram . However , this was not investigated in this study . Fourthly , the use of the bridging representation in this study \ufb01ts well with Rosenshine , Meister , and Chapman\u2019s ( 1996 ) characterization of scaffolding , drawn from sociocultural perspectives on learning . They interpret scaffolding as \u201ctemporary frameworks or supports usedtoassiststudentsduringinitiallearningofacomplexskillorcognitivestrategy . \u201dItmight thereforebearguedthatthebridgingrepresentationhelpstoscaffoldstudents\u2019understandingofNewton\u2019sthirdlawenablingthemtomovefromconcretephysicalsituationstowardsmoreabstractdiagrammaticandmathematicalrepresentationsasdepictedinFigure1 . The bridging representation acted as a temporary scaffold since the teacher ( author AS ) discontinued its use around the middle point of teaching . It is interesting to note that even after this point the students did not completely abandon it . In retrospective comments , the 192 SAVINAINEN ET AL . interviewed students indicated that they did not draw the diagram anymore but that they used it as an aid to their reasoning when identifying forces . They also commented that the linguistic aspect was important ( i . e . , stressing that forces are due to interactions ) but that the diagram had a crucial role in their understanding and applying Newton\u2019s third law . Finally , the bridging representation was used in a highly interactive manner based on peer discussions . The students had time and opportunity to talk through their developing understandings , with the support of the teacher , as recommended by Scott ( 1998 ) . It is worth noting that the same interactive teaching approach but without an effective tool for representation of interactions was not so successful in promoting learning of the third law in our earlier study ( Savinainen & Scott , 2002 ) . It is , moreover , doubtful that the emphasis on forces as interactions would be very effective in a passive lecture format ( Hake , 1998 ) . Reflections on the Nature of Conceptual Change The results from this study can be interpreted in terms of a conceptual change that involves the students reassigning the force concept to the ontological category of \u201cprocesses\u201d and starting to think of forces in terms of interactions between objects . Furthermore this would qualify as an example of strong restructuring or a \u201cbig\u201d change ( Tyson et al . , 1997 ) . Our data lend support ( see Figure 7 ) to the view that big conceptual changes are often evolutionary rather than revolutionary in nature , involving a gradual process ( Vosniadou & Ioannides , 1998 ) . This suggests that students\u2019 learning progresses from initial , scienti\ufb01cally more or less incorrect views via some intermediate or \u201ctransitional states\u201d towards the scienti\ufb01c perspective ( Thornton , 1995 ) . Our results also support Harrison , Grayson , and Treagust\u2019s ( 1999 ) conclusion that students\u2019 conceptual change requires time and explicit attention for developing the concepts : this was addressed in our approach by applying the \u201cconcepts \ufb01rst\u201d principle ( Van Heuvelen , 1991 ) . Pintrich , Marx , and Boyle ( 1993 ) stress that motivational factors play a signi\ufb01cant role in conceptual change . Students need to be cognitively engaged in order for conceptual change to occur in the \ufb01rst place . The instructional approach taken in our study relied heavily on peer discussions as ways of exploring meanings and it was important that the students recognized and accepted the worth of peer discussions ( Crouch & Mazur , 2001 ; Gunstone , McKittrick , & Mullhall , 1999 ) . Care was taken at the beginning of instruction to motivate the students to participate in peer discussion . The teacher explained that knowing physics means being able to \u201ctalk physics\u201d as well as being able to use other representations in analyzing physical situations . It was also made clear ( as recommended by Crouch and Mazur , 2001 ) that conceptual understanding is required in the examinations . It was gratifying to note that the students were very willing to participate in peer discussions almost from the very beginning . Final Reflections This article provides an account of a study in which various theoretical perspectives have been drawn upon in designing an instructional sequence that was then implemented in the classroom and demonstrated to give rise to enhanced learning outcomes when com - pared with those from a parallel class of students . We believe that this kind of research , involving careful attention to detail in relation both to instructional design principles and to consideration of the learning demands developed in speci\ufb01c areas of subject matter , is of considerable importance in the gradual and sustained development of effective teaching approaches . We believe that our study provides a successful example of how to close the gap ( Duit & Treagust , 2003 ) between \ufb01ndings on conceptual change and instructional practice . BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 193 REFERENCES Ainsworth , S . E . ( 1999 ) . The functions of multiple representations . Computer & Education , 33 ( 2 / 3 ) , 131 \u2013 152 . Bao , L . , Zollman , D . , Hogg , K . , & Redish , E . F . ( 2002 ) . Model analysis of \ufb01ne structures of student mod - els : An example with Newton\u2019s third law . American Journal of Physics , 70 , 766 \u2013 778 . Available at URL : http : / / www . physics . ohio - state . edu / \u223c lbao / papers . htm ( accessed on 22 . 4 . 2004 ) . Brown , D . E . ( 1989 ) . Students\u2019 concept of force : The importance of understanding Newton\u2019s third law . Physics Education , 24 , 353 \u2013 358 . Brown , D . E . ( 1992 ) . Using examples and analogies to remediate misconceptions in physics : Factors in\ufb02uencing conceptual change . Journal of Research in Science Teaching , 29 , 17 \u2013 34 . Brown , D . E . ( 1993 ) . Refocusing core intuitions : A concretizing role of analogy in conceptual change . Journal of Research in Science Teaching , 30 , 1273 \u2013 1290 . Brown , D . E . , & Clement , J . ( 1989 ) . Overcomingmisconceptionsbyanalogicalreasoning : Abstracttransferversus explanatory model construction . Instructional Science , 18 , 237 \u2013 261 . Camp , C . W . , & Clement , J . J . ( 1994 ) . Preconceptions in mechanics . Lessons dealing with students\u2019 conceptual dif\ufb01culties . Dubuque , IA : Kendall / Hunt Publishing Company . Carey , S . ( 1985 ) . Conceptual change in childhood . Cambridge , MA : MIT Press . Chi , M . T . H . , Slotta , J . D . , & de Leeuw , N . ( 1994 ) . From things to processes : A theory of conceptual change for learning science concepts . Learning and Instruction , 4 , 27 \u2013 43 . Clement , J . ( 1982 ) . Students\u2019 preconceptions in introductory mechanics . American Journal of Physics , 50 , 66 \u2013 71 . Clement , J . ( 1987 ) with the assistance of Brown , D . , Camp , C . , Kudukey , J . , Minstrell , J . , Palmer , D . , Schultz , K . , Shimabukuro , J . , Steinberg , M . , and Veneman , V . Overcoming students\u2019 misconceptions in physics : The role of anchoring intuitions and analogical validity . In J . D . Novak ( Ed . ) , Proceedings of the Second International Seminar , Misconceptions and Educational Strategies in Science and Mathematics ( Vol III , pp . 84 \u2013 97 ) . Cornell University . Crouch , C . , & Mazur , E . ( 2001 ) . Peerinstruction : Tenyearsofexperienceandresults . AmericanJournalofPhysics , 69 , 970 \u2013 977 . diSessa , A . ( 1993 ) . Towards an epistemology of physics . Cognition and Instruction , 10 , 105 \u2013 225 . Dufresne , R . , Leonard , W . , & Gerace , W . ( 2002 ) . Making sense of students\u2019 answers to multiple - choice questions . The Physics Teacher , 40 , 174 \u2013 180 . Duit , R . ( 2004 ) . Bibliography\u2014STCSE students\u2019 and teachers\u2019 conceptions and science education . Online at < http : / / www . ipn . uni - kiel . de / aktuell / stcse / stcse . html > . Accessed on 22 . 4 . 2004 . Duit , R . , & Treagust , D . F . ( 2003 ) . Conceptual change : A powerful framework for improving science teaching and learning . International Journal of Science Education , 25 , 671 \u2013 688 . Duit , R . , Roth , W . - M . , Komorek , M . , & Wilbers , J . ( 2001 ) . Fostering conceptual change by analogies\u2014between Scylla and Charybdis . Learning and Instruction , 11 , 283 \u2013 303 . Dykstra , D . I . , Boyle , F . C . , & Monarch , I . A . ( 1992 ) . Studying conceptual change in learning physics . Science Education , 76 ( 6 ) , 615 \u2013 652 . Finegold , M . , & Gorsky , P . ( 1991 ) . Students\u2019 concepts of force as applied to related systems : A search for consistency . International Journal of Science Education , 13 ( 1 ) , 97 \u2013 113 . Giancoli , D . ( 1998 ) . Physics\u2014Principles with applications ( 5th ed . ) . New York : Prentice - Hall International . Goldman , S . R . ( 2003 ) . Learning in complex domains : When and why do multiple representations help ? Learning and Instruction , 13 , 239 \u2013 244 . Gunstone , R . , McKittrick , B . , & Mullhall , P . ( 1999 ) . Structuredcognitivediscussionsinseniorhighschoolphysics : Student and teacher perceptions . Research in Science Education , 29 ( 4 ) , 527 \u2013 546 . Hake , R . ( 1998 ) . Interactive - engagement vs traditional methods : A six - thousand - student survey of mechan - ics test data for introductory physics courses . American Journal of Physics , 66 , 64 \u2013 74 . Available at URL : http : / / www . physics . indiana . edu / \u223c sdi / welcome . html # z44 ( accessed on 22 . 4 . 2004 ) . Halliday , D . , Resnick , R . , & Walker , J . ( 2001 ) . Fundamentals of Physics ( sixth ed . ) . New York : Wiley . Halloun , I . , Hake , R . , Mosca , E . , & Hestenes , D . ( 1995 ) . Force concept inventory ( Revised 1995 ) . Password protected and available at URL : http : / / modeling . la . asu . edu / R & E / Research . html ( accessed on 22 . 4 . 2004 ) . Halloun , I . , & Hestenes , D . ( 1985 ) . Common sense concepts about motion . American Journal of Physics , 53 , 1056 \u2013 1065 . Harrison , A . , Grayson , D . , & Treagust , D . ( 1999 ) . Investigating a grade 11 students\u2019 evolving conceptions of heat and temperature . Journal of Research in Science Teaching , 36 , 55 \u2013 87 . Hellingman , C . ( 1989 ) . Do forces have twin brothers ? Physics Education , 24 , 36 \u2013 40 . Hellingman , C . ( 1992 ) . Newton\u2019s third law revisited . Physics Education , 27 , 112 \u2013 115 . Hestenes , D . ( 1996 ) . Modeling method for physics teachers . Proceedings of the International Conference on Un - dergraduatePhysicsEducation ( CollegePark ) . AvailableatURL : http : / / modeling . la . asu . edu / modeling - HS . html ( accessed on 22 . 4 . 2004 ) . 194 SAVINAINEN ET AL . Hestenes , D . , Wells , M . , & Swackhamer , G . ( 1992 ) . Force concept inventory . The Physics Teacher , 30 , 141 \u2013 158 . Jauhiainen , J . , Koponen , I . , & Lavonen , J . ( 2001 ) . The force concept inventory in diagnosing the conceptual understanding of Newtonian mechanics in Finnish upper secondary schools . In M . Ahtee , O . Bj\u00a8orkvist , E . Pehkonen , & V . Vatanen ( Eds . ) , Research on mathematics and science education\u2014From beliefs to cognition , fromproblemsolvingtounderstanding ( pp . 101 \u2013 114 ) . Jyv\u00a8askyl\u00a8a : InstituteforEducationalResearch , University of Jyv\u00a8askyl\u00a8a . Jim\u00b4enez , J . D . , & Perales , F . J . ( 1999 ) . Es viable utilizar en la E . S . O . una representaci\u00b4on simb\u00b4olica de la fuerza como interacci\u00b4on ? La did\u00b4actica de las ciencias experimentales . Tendencias actuales ( pp . 589 \u2013 604 ) . La Coru\u02dcna : Servicio de Publicaciones de la Universidad . Jim\u00b4enez , J . D . , & Perales , F . J . ( 2001 ) . Graphic representation of force in secondary education : Analysis and alternative educational proposals . Physics Education , 36 , 227 \u2013 235 . Jones , M . G . , & Carter , G . ( 1998 ) . Small groups and shared constructions . In J . Mintzes , J . Wandersee , & J . Novak ( Eds . ) , Teaching science for understanding\u2014A human constructivist view ( pp . 261 \u2013 279 ) . Cornell University , Ithaca , NY : Academic . Koponen , I . , Jauhiainen , J . , & Lavonen , J . ( 2000 ) . A Finnish translation of the 1995 version of the force concept inventory , available upon request . Department of Physics , University of Helsinki . Kress , G . , Jewitt , C . , Ogborn , J . , & Tsatsarelis , C . ( 2001 ) . Multimodal teaching and learning : The rhetorics of the science classroom . London : Continuum . Leach , J . , & Scott , P . ( 1995 ) . The demands of learning science concepts : Issues of theory and practise . School Science Review , 76 ( 277 ) , 47 \u2013 52 . Leach , J . , & Scott , P . ( 2002 ) . Designing and evaluating science teaching sequences : An approach drawing upon the concept of learning demand and a social constructivist perspective on learning . Studies in Science Education , 38 , 115 \u2013 142 . Leach , J . , & Scott , P . ( 2003 ) . Individual and sociocultural views of learning in science education . Science and Education , 12 ( 1 ) , 91 \u2013 113 . Maloney , D . ( 1990 ) . Forces as interactions . The Physics Teacher , 28 , 386 \u2013 390 . Mason , L . ( 1998 ) . Sharing cognition to construct scienti\ufb01c knowledge in school context : The role of oral and written discourse . Instructional Science , 26 , 359 \u2013 389 . McDermott , L . C . , & Redish , E . F . ( 1999 ) . Resource letter PER - 1 : Physics education research . American Journal of Physics , 67 , 755 \u2013 767 . Available at URL : http : / / www . physics . umd . edu / rgroups / ripe / papers / rl . htm ( accessed on 22 . 4 . 2004 ) . McDermott , L . C . , Schaffer , P . , & the PERG at the University of Washington ( 1998 ) . Tutorials for introductory physics\u2014Workbook ( preliminary ed . ) Englewood Cliffs , NJ : Prentice - Hall . Meloth , M . , & Deering , P . ( 1999 ) . The role of the teacher in promoting cognitive processing during collaborative learning . In A . M . O\u2019Donnell & A . King ( Eds . ) , Cognitive perspectives on peer learning ( pp . 235 \u2013 255 ) . Mahwah , NJ : Erlbaum . Mercer , N . , & Wegeriff , R . E . ( 1999 ) . Is \u2018explanatory talk\u2019 productive talk ? In K . Littleton & P . Light ( Eds . ) , Learning with computers ( pp . 79 \u2013 101 ) . London : Routledge . Meltzer , D . ( 2002 ) . Issues related to data analysis and quantitative methods in PER . In S . Franklin , K . Cummings , & J . Marx ( Eds . ) , Proceedings of the 2002 Physics Education Research Conference ( pp . 21 \u2013 24 ) . New York : PERC Publishing . Minstrell , J . ( 1982 ) . Explaining \u201cat rest\u201d condition of an object . The Physics Teacher , 20 , 10 \u2013 14 . Montanero , M . , Suero , M . I . , Perez , A . L . , & Pardo , P . J . ( 2002 ) . Implicit theories of static interactions between two bodies . Physics Education , 37 , 318 \u2013 323 . Mortimer , E . , & Scott , P . ( 2003 ) . Meaning making in secondary science classrooms . Maidenhead / Philadelphia , PA : Open University Press / McGraw Hill . Palmer , D . ( 1997 ) . The effect of context on students\u2019 reasoning about forces . International Journal of Science Education , 6 , 681 \u2013 696 . Physics Textbook Review Committee ( 1998 ) . Quibbles , misunderstandings , and egregious mistakes . The Physics Teacher , 37 , 297 \u2013 305 . Pintrich , P . R . , Marx , R . W . , & Boyle , R . A . ( 1993 ) . Beyond cold conceptual change : The role of motivational beliefs and classroom contextual factors in the process of conceptual change . Review of Educational Research , 63 , 167 \u2013 200 . Reif , F . ( 1995a ) . Millikan Lecture 1994 : Understanding and teaching important scienti\ufb01c thought processes . American Journal of Physics , 63 , 17 \u2013 32 . Reif , F . ( 1995b ) . Understanding basic mechanics\u2014Workbook . New York : Wiley . Reif , F . , & Heller , J . I . ( 1982 ) . Knowledge structure and problem solving in physics . Educational Psychologist , 17 , 102 \u2013 127 . BRIDGING REPRESENTATION IN FOSTERING CONCEPTUAL CHANGE 195 Rosenshine , B . , Meister , C . , & Chapman , S . ( 1996 ) . Teaching students to generate questions : A review of the intervention . Review of Educational Research , 66 ( 2 ) , 181 \u2013 221 . Savinainen , A . ( 2004 ) . High school students\u2019 conceptual coherence of qualitative knowledge in the case of the force concept . Dissertations 41 , Department of Physics , University of Joensuu . Savinainen , A . , & Scott , P . ( 2002 ) . UsingtheForceConceptInventorytomonitorstudentlearningandtoplanteach - ing . Physics Education , 37 , 53 \u2013 58 . Available at URL : http : / / kotisivu . mtv3 . \ufb01 / physics / ( accessed on 22 . 4 . 2004 . ) Savinainen , A . , & Viiri , J . ( 2003 ) . Using the Force Concept Inventory to characterise students\u2019 conceptual coherence . In L . Haapasalo & K . Sormunen ( Eds . ) , Towards meaningful mathematics and science education , Proceeding on the IXX Symposium of Finnish Mathematics and Science Education Research Associa - tion . Bulletin of Faculty of Education , no . 86 , University of Joensuu ( pp . 142 \u2013 152 ) . Available at URL : http : / / kotisivu . mtv3 . \ufb01 / physics / ( accessed on 22 . 4 . 2004 ) . Schoultz , J . , S\u00a8alj\u00a8o , R . , & Wyndham , J . ( 2001 ) . Heavenly talk : Discourse , artifacts , and children\u2019s understanding of elementary astronomy . Human Development , 44 , 103 \u2013 118 . Scott , P . ( 1998 ) . Teacher talk and meaning making in science classrooms : A review of studies from a Vygotskian perspective . Studies in Science Education , 32 , 45 \u2013 80 . Sequira , M . , & Leite , L . ( 1991 ) . Alternative conceptions and history of science in physics teacher education . Science Education , 75 , 45 \u2013 56 . Steinberg , R . , & Sabella , M . ( 1997 ) . Performance on multiple - choice diagnostics and complementary exam problems . The Physics Teacher , 35 , 150 \u2013 155 . Terry , C . , & Jones , G . ( 1986 ) . Alternative frameworks : Newton\u2019s third law and conceptual change . European Journal of Science Education , 8 , 291 \u2013 298 . Thornton , R . K . ( 1995 ) . Conceptual dynamics : Changing students views of force and motion . In C . Tarsitani , C . Bernandini , & M . Vincentini ( Eds . ) , Thinking physics for teaching ( pp . 157 \u2013 183 ) . London : Plenum . Thornton , R . , & Sokoloff , D . ( 1998 ) . AssessingstudentlearningofNewton\u2019slaws : Theforceandmotionconceptual evaluation and the evaluation of active learning laboratory and lecture curricula . American Journal of Physics , 66 , 338 \u2013 351 . Treagust , D . F . , Chittleborough , G . , & Mamiala , T . L . ( 2003 ) . The role of submicroscopic and symbolic represen - tations in chemical explanations . International Journal of Science Education , 25 , 1353 \u2013 1368 . Treagust , D . F . , Harrison , A . G . , & Venville , G . J . ( 1996 ) . Using analogical teaching approach to engender conceptual change . International Journal of Science Education , 18 , 213 \u2013 229 . Turner , L . ( 2003 ) . System schemas . The Physics Teacher , 41 , 404 \u2013 408 . Tyson , L . , Venville , G . , Harrison , A . , & Treagust , D . ( 1997 ) . A multidimensional framework for interpreting conceptual change events in the classroom . Science Education , 81 , 387 \u2013 404 . Van Heuvelen , A . ( 1991 ) . Overview , case study physics . American Journal of Physics , 59 ( 10 ) , 898 \u2013 907 . van Someren , M . W . , Reimann , P . , Boshuisen , H . P . A . , & de Jong , T . ( Eds . ) ( 1998 ) . Learning with multiple representations . Oxford : Elsevier . Vosniadou , S . , & Ioannides , C . ( 1998 ) . From conceptual development to science education : A psychological point of view . Internal Journal of Science Education , 20 , 1213 \u2013 1230 . Vosniadou , S . , Ioannides , C . , Dimitrakopoulou , A . , & Papademetriou , E . ( 2001 ) . Designinglearningenvironments to promote conceptual change in science . Learning and Instruction , 11 , 381 \u2013 419 . Welzel , M . , & Roth , W . - M . ( 1998 ) . Do interviews really assess students\u2019 knowledge ? International Journal of Science Education , 20 , 25 \u2013 44 .",
    "bernsteinNetworkCentralizationCollective2023": "This article was downloaded by : [ 129 . 2 . 89 . 115 ] On : 03 May 2024 , At : 15 : 27 Publisher : Institute for Operations Research and the Management Sciences ( INFORMS ) INFORMS is located in Maryland , USA Organization Science Publication details , including instructions for authors and subscription information : http : / / pubsonline . informs . org Network Centralization and Collective Adaptability to a Shifting Environment Ethan S . Bernstein , Jesse C . Shore , Alice J . Jang To cite this article : Ethan S . Bernstein , Jesse C . Shore , Alice J . Jang ( 2023 ) Network Centralization and Collective Adaptability to a Shifting Environment . Organization Science 34 ( 6 ) : 2064 - 2096 . https : / / doi . org / 10 . 1287 / orsc . 2022 . 1584 Full terms and conditions of use : https : / / pubsonline . informs . org / Publications / Librarians - Portal / PubsOnLine - Terms - and - Conditions This article may be used only for the purposes of research , teaching , and / or private study . Commercial use or systematic downloading ( by robots or other automatic processes ) is prohibited without explicit Publisher approval , unless otherwise noted . For more information , contact permissions @ informs . org . The Publisher does not warrant or guarantee the article\u2019s accuracy , completeness , merchantability , fitness for a particular purpose , or non - infringement . Descriptions of , or references to , products or publications , or inclusion of an advertisement in this article , neither constitutes nor implies a guarantee , endorsement , or support of claims made of that product , publication , or service . Copyright \u00a9 2022 The Author ( s ) Please scroll down for article\u2014it is on subsequent pages With 12 , 500 members from nearly 90 countries , INFORMS is the largest international association of operations research ( O . R . ) and analytics professionals and students . INFORMS provides unique networking and learning opportunities for individual professionals , and organizations of all types and sizes , to better understand and use O . R . and analytics tools and methods to transform strategic visions and achieve better outcomes . For more information on INFORMS , its publications , membership , or meetings visit http : / / www . informs . org Network Centralization and Collective Adaptability to a Shifting Environment Ethan S . Bernstein , a , * Jesse C . Shore , b , * Alice J . Jang c * Contributed equally a Harvard Business School , Organizational Behavior Unit , Boston , Massachusetts 02163 ; b Questrom School of Business , Boston University , Boston , Massachusetts 02215 ; c Pamplin College of Business , Virginia Polytechnic Institute and State University , Blacksburg , Virginia 24061 Contact : e @ hbs . edu , https : / / orcid . org / 0000 - 0001 - 9819 - 0639 ( ESB ) ; jesse . c . shore @ gmail . com , https : / / orcid . org / 0000 - 0003 - 3812 - 0262 ( JCS ) ; ajjang @ vt . edu , https : / / orcid . org / 0000 - 0003 - 1140 - 0598 ( AJJ ) Received : September 15 , 2019 Revised : June 7 , 2020 ; April 15 , 2021 ; August 20 , 2021 ; January 19 , 2022 Accepted : February 4 , 2022 Published Online in Articles in Advance : May 31 , 2022 https : / / doi . org / 10 . 1287 / orsc . 2022 . 1584 Copyright : \u00a9 2022 The Author ( s ) Abstract . We study the connection between communication network structure and an organization \u2019 s collective adaptability to a shifting environment . Research has shown that network centralization \u2014 the degree to which communication \ufb02 ows dispropor - tionately through one or more members of the organization rather than being more equally distributed \u2014 interferes with collective problem - solving by obstructing the integration of existing ideas , information , and solutions in the network . We hypothesize that the mecha - nisms responsible for that poor integration of ideas , information , and solutions would nevertheless prove bene \ufb01 cial for problems requiring adaptation to a shifting environment . We conducted a 1 , 620 - subject randomized online laboratory experiment , testing the effect of seven network structures on problem - solving success . To simulate a shifting environ - ment , we designed a murder mystery task and manipulated when each piece of information could be found : early information encouraged an inferior consensus , requiring a collective shift of solution after more information emerged . We \ufb01 nd that when the communication network within an organization is more centralized , it achieves the bene \ufb01 ts of connectivity ( spread of novel better solutions ) without the costs ( getting stuck on an existing inferior sol - ution ) . We also \ufb01 nd , however , that these bene \ufb01 ts of centralization only materialize in net - works with two - way \ufb02 ow of information and not when information only \ufb02 ows from the center of the network outward ( as can occur in hierarchical structures or digitally mediated communication ) . We draw on these \ufb01 ndings to reconceptualize theory on the impact of cen - tralization \u2014 and how it affects conformity pressure ( lock - in ) and awareness of diverse ideas ( learning ) \u2014 on collectiveproblem - solving thatdemands adaptation . History : This paper has been accepted for the Organization Science Special Issue on Experiments in Organizational Theory . Open Access Statement : This work is licensed under a Creative Commons Attribution 4 . 0 International License . You are free to copy , distribute , transmit and adapt this work , but you must attribute this work as \u201c Organization Science . Copyright \u00a9 2022 The Author ( s ) . https : / / doi . org / 10 . 1287 / orsc . 2022 . 1584 , used under a Creative Commons Attribution License : https : / / creativecommons . org / licenses / by / 4 . 0 / . \u201d Funding : The authors acknowledge \ufb01 nancial support from the Division of Research at the Harvard Busi - ness School for making this research possible . Keywords : network centralization \u2022 adaptation \u2022 organizational structure \u2022 communication networks \u2022 social networks \u2022 problem - solving \u2022 dynamic environments \u2022 collective intelligence How does the centralization of an organization \u2019 s communication network affect its members \u2019 capacity for collective problem - solving ? Because problem - solving is key to organizational success ( Grant 1996 , Nickerson and Zenger 2004 ) , because organizational problem - solving is increasingly collective ( Hong and Page 2004 , Wuchty et al . 2007 , Hackman 2011 ) , and because communication network structure is a key factor in making collective problem - solving effective ( Mason and Watts 2012 , Becker et al . 2017 , Argote et al . 2018 ) , organizational theorists and managers are interested in the link between characteristics of a communication network ( like centralization ) and suc - cessful collective problem - solving . \u201c Network centralization , \u201d following Freeman ( 1978 ) , has been de \ufb01 ned as the degree to which communication \ufb02 ows disproportionately through one or more members of the organization rather than equally through all ( e . g . , Huang and Cummings 2011 ) . Although it has been measured in a variety of ways , from degree ( Shaw 1954 ) to eigenvector centrality ( Bonacich 1987 ) , a highly cen - tralized network is always characterized by ( a ) direct ties between a \u201c core \u201d individual or group and most ( or all ) individuals and ( b ) a lack of direct ties between those not 2064 ORGANIZATION SCIENCE Vol . 34 , No . 6 , November \u2013 December 2023 , pp . 2064 \u2013 2096 ISSN 1047 - 7039 ( print ) , ISSN 1526 - 5455 ( online ) https : / / pubsonline . informs . org / journal / orsc D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . in the core ( i . e . , the \u201c periphery \u201d ) . Such a network all but guarantees , even without status - based hierarchy , 1 that the core will have access to all critical information and can \u201c put the pieces together \u201d ( Bavelas 1950 , Tushman 1979 , Butts et al . 2007 ) , whereas the periphery will have access only to whatever the core has the bandwidth and chooses to share . The hub - and - spoke structure \u2014 with a core leader ( the \u201c hub \u201d ) who brokers all interactions amongst peripheral members ( the \u201c spokes \u201d ) \u2014 is the most extreme of centralized networks , but others ( e . g . , a leadership team with multiple core members ) operate similarly , with the core responsible for integrating the work of the periphery . Prior research \ufb01 nds that such network centraliza - tion of communication harms collective problem - solving because people in the periphery cannot directly share ideas , information , and / or solutions with each other , whereas the core can become a bottleneck to \ufb01 nd - ing a good solution or can lead the whole network astray should it adopt a bad one ( Tushman 1979 , Huang and Cummings 2011 , Becker et al . 2017 , Argote et al . 2018 ) . Some organizations have , therefore , decided that collective problem - solving calls for decentralized communication structures , especially to draw on fast information \ufb02 ows and diverse expertise from across the organization for collective problem - solving requiring adaptation ( Gulati 2007 , Bernstein et al . 2016 , Lee and Edmondson 2017 , Reitzig 2022 , Bernstein 2022 ) . However , prior studies have not considered adapta - tion problems , which have different features than the previously studied collaboration problems . Instead , prior studies have focused on tasks that , although rep - resentative of daily collaborative work , do not re \ufb02 ect the rapidly shifting environments , sudden develop - ments , and need to draw on diverse ideas ( Lee and Edmondson 2017 ) that characterize adaptation tasks . The purpose of this paper is to investigate whether communication network centralization might make organizations not less but rather more capable of col - lective problem - solving requiring adaptation . We build on a tradition of research about how network structure relates to social in \ufb02 uence and the diversity of ideas ( Lazer and Friedman 2007 , Mason et al . 2008 , Mason and Watts 2012 , Shore et al . 2015 , Bernstein et al . 2018 ) to investigate whether the characteristics of centralized networks that impede traditional collabo - rative problem - solving may nevertheless improve col - lective adaptation . Speci \ufb01 cally , we consider whether the inability of peripheral individuals to communicate with each other directly \u2014 although posing an integra - tion risk to collaborative problem - solving \u2014 could actually be bene \ufb01 cial for solving adaptation problems because it could reduce conformity pressure and promote independent thinking in the periphery . Simi - larly , we consider whether the fact that the core is connected to everybody in the periphery \u2014 although posing a bottleneck risk to collaborative problem - solving \u2014 could be bene \ufb01 cial in adaptation problems because the core could learn from differing solutions in the periphery and spread the one they think best . To test these ideas , we ran a 1 , 620 - person experiment on a new experimental platform that we developed for randomized experiments with communication network structure as the independent variable ( i . e . , investigating how network structure causally affects collective per - formance ) . To collectively solve problems , participants could seek and share information \u2014 much as they would do at work with enterprise social network tech - nologies ( e . g . , Slack , Teams , Workspace , Yammer , etc . ) \u2014 and they could enter solutions and access their neighbors \u2019 solutions . We designed the experimental platform to permit the creation of a shifting environ - ment ( analogous to dynamic environments faced by product managers and organization leaders ) by con - trolling when certain information can be discovered . We used this feature to create a collaborative murder mystery task with a pivotal phase in which partici - pants had to collectively shift away from a premature ( although seemingly reasonable ) consensus on an infe - rior solution , adapt to new information , and adopt a new solution . To study the main effect of communi - cation network centralization on problem - solving and to describe the mechanisms through which it works , we tested seven realistic network structures ( four varia - tions on centralized network structures and three decentralized controls ) ( see Figure 5 ) . We randomly assigned participants to nodes , so that our results would isolate the effects of overall network structure from individual attributes rather than our \ufb01 ndings about centralization and problem - solving being con - founded with individual attributes or competency . We \ufb01 nd that centralized communication networks solved adaptation problems better than decentralized networks did . They were less likely to get stuck on a premature consensus and more likely to disseminate the superior solution . They exhibited the bene \ufb01 cial aspects of social in \ufb02 uence in problem - solving ( expo - sure to differing interpretations ) without the harmful aspects ( lock - in on an existing solution because of con - formity pressure ) . By testing both directed network ties ( one way from core to periphery ) and undirected network ties ( two way between core and periphery ) , we also provide a major caveat with respect to that main \ufb01 nding . The bene \ufb01 ts to centralization rely on central nodes learn - ing from the more - independent peripheral nodes . In real - world organizations , however , central nodes often in \ufb02 uence without being in \ufb02 uenced by the periphery , as in traditional bureaucratic / administra - tive pyramidal structures ( Puranam 2018 ) . Such struc - tures tend to prevent the aforementioned bene \ufb01 ts of centralization . Our results , therefore , reinforce the Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2065 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . importance of nurturing behaviors ( Detert and Burris 2007 , Morrison 2011 , Detert et al . 2013 ) to ensure that communication \ufb02 ows both ways . This paper seeks to make three contributions by addressing an open question in organizational theory and organization design about the impact of central - ization on collective problem - solving in tasks requir - ing adaptation and , therefore , on collective intelligence ( Woolley et al . 2010 , Malone 2018 ) . First , we demon - strate that , contrary to popular opinion , centralization is bene \ufb01 cial to tasks requiring adaptation because it promotes social learning but avoids lock - in . Second , by cleanly decoupling the effects of centralization and of hierarchy , we observe the pure effect of structurally - imposed centralization on problem - solving , which is useful for advancing scholarship on both centraliza - tion and hierarchy . Finally , because managers increas - ingly use enterprise social networking technologies ( Majchrzak et al . 2009 , Leonardi 2014 ) to bypass tradi - tional centralized structures ( Gulati and Puranam 2009 ) , our \ufb01 ndings have practical implications for using com - munication technologies to promote \u2014 and not under - mine \u2014 organizational adaptability . Hypothesis Development : The Effect of Centralization on Collective Problem - Solving in Adaptation Tasks The Harmful Effects of Centralization in Previously - Studied Tasks Generally , prior studies have found that structural centralization impedes using knowledge and ideas already in the network . More speci \ufb01 cally , studies of collaboration tasks \ufb01 nd that centralization impedes the collective ability to integrate the information and ideas held in the periphery , whereas studies of wisdom - of - the - crowd tasks \ufb01 nd that collective judg - ments suffer from the excessive in \ufb02 uence of central individuals ( see Table 1 ) . What Is Different in Adaptation Tasks ? Given the long - standing belief that the effect of struc - tural characteristics ( like centralization ) on outcomes will be contingent upon the type of problem to be solved ( Tushman 1979 ) , it may be the case that collective problem - solving tasks requiring adaptation \u2014 which differ in several key ways from tasks requiring collabo - ration or coordination \u2014 respond differently to central - ization , even as the mechanisms identi \ufb01 ed in prior research remain the same ( see Table 2 ) . We , therefore , reinterpret the key harmful mechanisms identi \ufb01 ed in prior research in light of literature on social in \ufb02 uence , social network structure , and adaptability in social net - works . In doing so , we theorize that those same mecha - nisms may improve performance on tasks requiring collective adaptation \u2014 that is , collectively shifting from one solution to a dissimilar , but superior , one . Scholars have found that various adaptation prob - lems share a common trait that distinguishes them from tasks requiring collaboration or coordination ; the existing understanding of a problem can make it very hard to \ufb01 nd other ways to solve it \u2014 that is , to adapt \u2014 because high - quality new solutions are likely to be very different from the current solution ( as is the case , for example , with successful responses to disruptive inno - vations ( Christensen 2013 ) ) . The landscape of solutions to adaptation problems would , therefore , \ufb01 t what scholars call \u201c rugged \u201d ( Levinthal 1997 ) \u2014 meaning that the performance \u201c peaks \u201d ( i . e . , best solutions ) may be very distant from each other in terms of their compo - nent parts . More generally , this happens when a solu - tion has many interdependent components , such that someone trying to solve that problem is likely to go over multiple hills and valleys ( a bumpy , iterative proc - ess of trying diverse combinations of solution compo - nents ) to get from one good solution to the next . There are many examples of such interdependent components \u2014 and therefore \u201c rugged \u201d solution land - scapes \u2014 challenging organizations that attempt to adapt to a shifting environment ( Lavie et al . 2010 ) . In steel production ( Levinthal 1997 , Prennushi et al . 1997 ) , for example , as modern factory work practices ( incentive pay , team - based work , \ufb02 exible job assign - ments , employee training ) began to replace more traditional practices ( hourly wages , narrow job de \ufb01 ni - tions , close supervision ) , factories failed to experience improved performance unless coherent , interdepend - ent sets of modern work practices were adopted together . Experimenting with only one deviation from traditional working structures ( e . g . , shifting from nar - row job de \ufb01 nitions to \ufb02 exible assignments ) might have produced no performance gain ; it might even have seen a decline . To modernize effectively , many interdependent work practices needed to be adjusted in tandem ( Bloom et al . 2012 ) , making the future solu - tion very dissimilar from the past one . Such changes are often described \u2014 and experienced \u2014 as transfor - mational rather than incremental ( Immelt 2017 , Reeves et al . 2018 ) . Similarly , components of a corpo - rate strategy are often highly interdependent . Thanks to tracking customer preferences through personaliza - tion technologies and web search history , Google \u2019 s simultaneous investments in web search , ad auctions , and personalization technologies jointly contribute to its successful overall business model precisely because they are interdependent ; Google \u2019 s auctions maximize the probability of clicking , which they are able to do by tracking customer preferences and interests and personalizing their ads . If one of these strategic ele - ments was no longer viable ( perhaps through regula - tion ) , the whole business model would have to adapt . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2066 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . Conceptually , although not all adaptation problems necessarily require such transformational rather than incremental change , this paper follows prior research in focusing on those that do . Such challenges of solving adaptation problems have been viewed in terms of the trade - off in individuals \u2019 choices between exploration ( searching for new solu - tions ) and exploitation ( re \ufb01 ning a preexisting consen - sus ) ( March 1991 , Levine et al . 2018 ) . However , even holding individual exploratory behavior constant , network structure \u2014 such as centralization \u2014 can make it more or less likely for an organization facing adapta - tion problems to shift from a good solution to a supe - rior ( but dissimilar ) one by ( a ) reducing the probability of sticking to an inferior consensus and ( b ) spreading a superior solution to everyone once it is found . In the following sections , we theorize the role centralization might play in each of those two . Table 1 . Negative Effects of Centralization on Integration and Wisdom of the Crowd Tasks Type of task Task examples Findings \u2014 mechanism by which centralization harms performance Collaboration tasks : Tasks that bene \ufb01 t from integration of the collaborators \u2019 diverse contributions of ideas and information Tushman ( 1979 ) : Research & Development ( R & D ) projects of different types and complexity Work group hierarchical structure and work group task were correlated , implying a contingency theory of centralization . High - performing research groups were less centralized than high - performing technical service groups performing routine work Cummings and Cross ( 2003 ) : Project - based work , requiring coordination and integration of ideas Centralization was negatively associated with performance Huang and Cummings ( 2011 ) : Project - based work , requiring coordination and integration of ideas Centralization prevents peripheral individuals from integrating each other \u2019 s information . Core acts as a bottleneck Argote et al . ( 2018 ) : Laboratory experiment on the generation of new features in a collaborative programming task Centralized networks generated fewer new features Wisdom of the crowd tasks : Tasks based on averaging a group of individual estimates Becker et al . ( 2017 ) : Estimation of various quantities ( coins in a jar , calories in a meal , etc . ) Centralized networks were more likely to decrease in performance than decentralized networks because of the excessive in \ufb02 uence of central individuals Golub and Jackson ( 2010 ) : Mathematical modeling Crowd wisdom occurs only when no single individual is too in \ufb02 uential Yan et al . ( 2021 ) : Crowdsourced clothing design More centralized crowds that did not have shared task experience were worse at predicting market demand for their designs Table 2 . Distinguishing Collective Problem - Solving Tasks Requiring Adaptation from Previously Studied Collective Problem - Solving Tasks Requiring Collaboration Collaboration tasks Adaptation tasks Task context Stable environment Shifting environment Task objective Integration / coordination to reach consensus Departure from prior consensus to a new , superior solution Nature of the information landscape Information available for problem - solving remains relatively static Information available for problem - solving is revealed over time Nature of the solution landscape Relatively smooth with independent components ; improving from one solution to the next - better one generally involves following an incremental path Relatively rugged with interdependent components ; improving from one solution to the next - better one generally involves following a transformational path Success enabled by Free \ufb02 ow of communication , allowing nodes to share and build off of information and interpretations already existing amongst them Preservation of diverse , novel options and multiple interpretations , unconstrained by preexisting knowledge and the consensus interpretation ( s ) Indicators of success Fully elaborated solution Shift from one solution to a new solution Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2067 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . The Effect of Centralization on the Probability of Sticking to an Inferior Consensus in Adaptation Tasks How could network structure affect the probability of sticking to an inferior consensus ? Conformity pres - sure from peers is a determinant of whether organiza - tion members are collectively adaptable or likely to get jointly stuck on a current consensus . Being aware of a network neighbor \u2019 s solution can lead an individ - ual to adopt a matching solution , and this \u201c social in \ufb02 u - ence \u201d can be a positive or negative force . It is often individually rational to copy peers \u2019 solutions when one is not con \ufb01 dent in one \u2019 s own , but this can have the effect of entrenching the current inferior consensus \u2014 because it does not allow more diverse solutions to be considered \u2014 as easily as it could lead to bene \ufb01 cial learning ( Banerjee 1992 , Bikhchandani et al . 1992 ) . For our purposes , social in \ufb02 uence can be divided roughly into two components : conformity pressure to adopt peers \u2019 solutions ( contributing to entrenched consensus ) and transfer of substantive information between nodes ( contributing to bene \ufb01 cial learning ) . In general , holding transfer of information ( and therefore learning ) constant , the stronger the social conformity pressure , the harder it will be to collectively shift from one consensus to a new superior one . We , therefore , begin by focusing \ufb01 rst on conformity pressure ( that increases the probability of sticking to an inferior con - sensus ) and return to the transfer of information in the following section on the probability of spreading superior solutions . Conformity Pressure from Neighbors : Number Effects Prior research shows that network structure can increase conformity pressure when more of a person \u2019 s network neighbors adopt a given solution . Results from Centola ( 2010 ) and Ugander et al . ( 2012 ) suggest that two neighbors with the same solution are approxi - mately twice as in \ufb02 uential as one , with each additional neighbor having a decreasing marginal in \ufb02 uence . See the four curves in Figure 1 for a visualization , with higher curves representing more neighbors . Conformity Pressure from Neighbors : Share Effects When the solutions to a problem are mutually exclu - sive ( e . g . , in a jury , the options are to convict or acquit but not both ) , the share of neighbors who have adopted each solution is important ; there is a strong majority effect , resulting in an S curve of probability of adoption ( see , e . g . , equations 15 and 16 in Stasser and Davis 1981 ) for each individual behavior ( or judgment ) . Conformity Pressure from Neighbors : Combining Number and Share Effects Together , these two effects result in a response curve ( the probability of adopting an alternative solution due to social in \ufb02 uence ) that depends on both number ( different curves in Figure 1 ) and share ( horizontal axis in Figure 1 ) of neighbors . Figure 1 depicts the case in which two competing alternatives are equal in Figure 1 . Relative Amounts of Conformity Pressure to Shift to a Competing and Mutually Exclusive Solution to a Problem Vary by Number and Share of Adopting Peers 1 neighbor 2 neighbors 3 neighbors 8 neighbors 0 . 0 0 . 3 0 . 6 0 . 9 share of neighbors adopting c on f o r m it y p r e ss u r e ( a r b it r a r y un it s ) Notes . Separate curves , as inferred from prior literature ( Stasser and Davis 1981 , Centola 2010 , Ugander et al . 2012 ) , are drawn for groups with one , two , three , and eight ( other ) peers , with points plotted at each whole number of peers . Note that exact parameters vary by setting and that this visualization is intended only to convey qualitatively how number and share matter in in \ufb02 uence . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2068 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . intrinsic in \ufb02 uence such that differences are a conse - quence only of number and of share of peers who adopt each alternative . These results suggest that for network structure to aid in adaptability , it must put people in network positions such that either ( a ) they have relatively few neighbors ( and are thus on a lower curve in Figure 1 ) or ( b ) their neighbors are unlikely to all adopt the same solution ( and thus are located to the left side of their curve ) . How Centralization Might Reduce Conformity Pressure in Adaptation Tasks If we logically apply the results on number and share effects reviewed to centralized networks , they suggest centralization might reduce conformity pressure in adaptation tasks because of the way that number and share effects work in tandem with one another in cen - tralized networks . With respect to number effects , central nodes are by de \ufb01 nition relatively few in number , meaning that peripheral nodes would experience relatively little con - formity pressure by virtue of sitting on a relatively low curve in Figure 1 . In this context , by \u201c relatively low , \u201d we mean low relative to other possible network positions and structures with the same total number of nodes \u2014 in other words , other ways of organizing the same number of people . In short , although central nodes in \ufb02 uence many people , that in \ufb02 uence may not be very strong because there are so few of them ( number effects ) . On the other hand , number effects would predict that central nodes in a centralized network could be highly in \ufb02 uenced by the peripheral nodes \u2014 central nodes in a centralized network have relatively many neighbors and thus sit on a relatively high conformity pressure curve in Figure 1 . However , share effects might limit that con - formity pressure , as peripheral nodes are likely to adopt a variety of different solutions because of ( a ) the low conformity pressure peripheral nodes feel from central nodes and ( b ) the fact that these peripheral nodes are not able to exert conformity pressure on each other ( because they are not directly connected ) . This means that despite sitting on a high curve ( because of number effects ) , central nodes are likely to \ufb01 nd themselves on the left side of their curve ( Figure 1 ) , where a relatively small share of their neighbors has adopted any given solution . Moreover , because conformity is a self - reinforcing phenomenon , low conformity pressure on the core further reduces conformity pressure on the periphery ( see Figure 2 , block A ) . The logic we have built above , based on number and share effects , to theorize how centralization could reduce conformity pressure in adaptation tasks \ufb01 nds some support , albeit by analogy , in prior theoretical and empirical research on network clustering ; that is , when one \u2019 s network neighbors are also neighbors to each other ( Hansen 1999 , Burt 2004 , Centola 2010 ) . Because both number and share of neighbors matter , the formation and maintenance of a single consensus solution are also more likely with network clust - ering . Network clustering makes it more likely for an individual to have more reinforcing exposures to the same solution ( s ) , and thus it both hastens the spread of \u201c complex contagions \u201d ( Centola 2010 ) and impedes deviation from an established consensus ( Stasser and Davis 1981 ) . Clustering ( and its global analog , density \u2014 having many network connections within a set of nodes ) , thus , has been found to inhibit a net - work \u2019 s collective ability to consider different solutions to a complex problem , even as it aids in collecting more information ( Lazer and Friedman 2007 , Shore et al . 2015 ) . Although a high level of centralization does not necessarily imply general lack of clustering , it does imply a lack of clustering and density of con - nections among peripheral nodes , which are connected to central nodes but not to each other . To make our logic more concrete , we built a sim - ulation model of how the number and share effects found in prior literature would work in the context of various network structures . We simulated the probabil - ity of shifting from one consensus solution to another , assuming the conformity pressure rules depicted in Figure 1 and that in each time period , people have a small random chance of changing their solutions ( see Appendix A for details and Figure 5 for network struc - tures ) . Our simulation found centralized networks less likely to get stuck on an initial consensus than other network structures . Speci \ufb01 cally , agents in a complete clique retained the initial consensus 92 % of the time , agents in a locally clustered network retained the ini - tial consensus 80 % of the time , peripheral nodes in a hub - and - spoke network retained the initial consensus 57 % of the time , and peripheral nodes in a core - periphery network retained the initial consensus 72 % of the time . Thus , prior research on network structure and our own simulations based on the microfoundations of social in \ufb02 uence support our logic that members of centralized networks feel less conformity pressure than those in other empirically common network structures and are , therefore , more likely to choose different solutions than their peers when solving com - plex problems . That is consistent with prior work \ufb01 nding that the performance of centralized networks on nonadaptation tasks suffers because peripheral nodes are not tightly integrated ( Huang and Cum - mings 2011 ) , but it is precisely that lack of mutual in \ufb02 uence and resultant conformity pressure that could be a blessing for adaptation tasks because it allows more diverse solutions to be considered collectively . We therefore hypothesize : Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2069 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . Hypothesis 1 . In problems requiring adaptation , central - ization reduces the probability of sticking to an inferior consensus . The Effect of Centralization on the Probability of Spreading Superior Solutions in Adaptation Tasks For collective adaptation , it is not enough to avoid get - ting stuck on an inferior solution . Good solutions \u2014 which can be hard to come by \u2014 must spread . In gen - erating Hypothesis 1 , we argued that centralization might reduce conformity pressure , but we must also argue that it does not impair the positive side of social in \ufb02 uence : the bene \ufb01 cial transfer of information and knowledge ( i . e . , bene \ufb01 cial learning ) ( e . g . , Argote and Ingram 2000 ) . There has been signi \ufb01 cant study of how to promote the spread of established , successful solu - tions in an organization ( e . g . , Reagans and McEvily 2003 ) . It can be particularly valuable to learn from peers when one lacks relevant experience ( Aranda et al . 2017 ) . Given that bene \ufb01 cial transfer of information and counterproductive conformity pressure both result from the spread of solutions through social in \ufb02 uence , is it possible to predictably get the bene \ufb01 ts of social in \ufb02 uence without the costs ? Bernstein et al . ( 2018 ) show that it is indeed possible , in that case by having intermittent breaks in interaction ( i . e . , alternating indi - vidual work with connectivity ) . In this way , people experience independence , which generates diversity of solutions without conformity pressure , while also preserving opportunities to interact , which allows them to learn from that diversity . Theoretically , centralization may create structural differentiation somewhat as intermittent interaction does , creating the conditions for both diversity of ideas ( because of relatively little conformity pressure ) and the spread of good ideas ( through the central nodes ) . Although we have argued that central nodes exert relatively little conformity pressure , they still expose many peripheral nodes to their solutions , creating the opportunity for their ideas to spread . Burt \u2019 s research program on \u201c brokerage \u201d and \u201c struc - tural holes \u201d has long argued that those who bridge otherwise disconnected people are structurally privi - leged , being more exposed to different ways of think - ing ( e . g . , Burt 2004 ) . Again , in Figure 1 , central nodes would sit to the left side , where a small share of their neighbors has adopted any given solution ( Figure 2 , B ) . Knowledge of others \u2019 solutions without conform - ity pressure to adopt a speci \ufb01 c solution compels cen - tral nodes to exercise judgment , albeit with the bene \ufb01 t of being able to ask others to explain their various positions . Thus , in a setting requiring adaptation , it would seem that central nodes are especially well positioned to learn from the periphery ( Figure 2 , C ) and to spread what they learn more widely . In essence , centralized networks may , like intermittent interaction , be able to achieve the in \ufb02 uence necessary for spreading good solutions without creating conformity pressure that would keep those good solutions from being generated by peripheral members in the \ufb01 rst place . We therefore hypothesize : Hypothesis 2 . In problems requiring adaptation , central - ization promotes the spread of superior solutions . Taken together , centralization \u2019 s hypothesized lower probability of sticking to an inferior consensus and its hypothesized increased probability of spreading supe - rior solutions should result in a higher average per - formance . We therefore hypothesize : Hypothesis 3 . In problems requiring adaptation , central - ized networks cause more people to be better on average . We add a \ufb01 nal hypothesis to capture a particularly important boundary condition . The logic behind the bene - \ufb01 ts of centralization for collective adaptation and , there - fore , behind our \ufb01 rst three hypotheses , relies on central nodes being in \ufb02 uenced by peripheral nodes ( Figure 2 , B ) . In practice , communication in some centralized net - works is unidirectional ( for example , in traditional , bureaucratic , hierarchical organizations and self - organ - ized online communities ) ; the core in \ufb02 uences but is not in \ufb02 uenced by the periphery . Thus , it is important to dis - tinguish between centralization with two - way communi - cation between core and periphery and centralization with only one - way communication from the core to the periphery . For the latter , we do not expect the bene \ufb01 ts stipulated in Hypotheses 1 \u2013 3 . Hypothesis 4 . In problems requiring adaptation , central - ized networks with one - way communication from the core to the periphery do not reduce the probability of sticking to an inferior consensus , do not promote the spread of superior solutions , and do not cause more people to be better on average . Experimental Design Using a new online laboratory platform , 2 we con - ducted a randomized experiment of the effect of net - work centralization on problem - solving in a setting that requires adapting from one answer to another because of the emergence of new information ( see Figure 5 for network treatments ) . The Task In developing a suitable experimental task , we sought a task design focused on the collective ability of the organization that was solving the problem \u2014 in the context of different communication structures \u2014 to shift from one idea to a dissimilar one , with the main Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2070 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . dependent variable de \ufb01 ned as the number of partici - pants per trial who were correct in the end . We also sought a protocol that allowed experimental control of important variables and yet was not overly stylized , such that the task was similar to real problem - solving work in shifting environments and that the means for accomplishing the task within the platform had paral - lels in the real world . Following the long tradition of murder mystery protocols in group research ( Stasser and Stewart 1992 , Stasser and Titus 2003 ) and more recently in network research ( Shore et al . 2015 ) , we adopted a traditional \ufb01 ctional murder mystery task ( akin to the popular game Clue or Cluedo ) , in which the task required piecing together facts ( \u201c clues \u201d ) to identify the culprit . However , we added a twist : red herring clues that implicate the wrong suspect and the timed release of critical clues . Information leading to the correct answer would only become available after partici - pants had already come to a reasonable but incorrect conclusion or consensus . Because we were interested in how organizational structure affects collective adaptation to a shifting environment , we measured performance in terms of the number of individuals who \ufb01 nished each trial with the correct answer \u2014 that is , who successfully adapted . Our results are , thus , at the network level , but they aggregate individual - level data , capturing how network structure affects the distribution of the number of participants who \ufb01 nd the best answer . Thus , we build on the paradigm of social organization and collective intelligence research on settings in which individuals work on separate instances of the same task but nonetheless in \ufb02 uence each other such that collective , network - level outcomes differ from those of sets of disconnected individuals ( also known as \u201c nominal groups \u201d ) . As with past research in this paradigm ( e . g . , Lazer and Friedman 2007 , Mason and Watts 2012 , Levine et al . 2014 , Shore et al . 2015 , Becker et al . 2017 ) , our approach allows disaggregated meas - urement of differing solutions and focuses on how social organization affects the emergence of consen - sus \u2014 that is , on the ability to learn from and build on each other \u2019 s solutions . Other work has adopted an alternative paradigm \u2014 with individuals working on a shared task \u2014 and has focused on determinants of Figure 2 . ( Color online ) Diagram of Mechanisms Underlying Performance in Adaptation Tasks Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2071 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . group or team performance , with attention to process gains and losses ( e . g . , Woolley et al . 2010 , Huang and Cummings 2011 , Argote et al . 2018 ) . We return to this distinction in paradigms as a limitation in our discus - sion section . Our experimental interface ( Figure 3 ) allows partici - pants to ( a ) search for clues about the mystery by entering keywords into a search bar , ( b ) share any clues they \ufb01 nd with network neighbors ( with or with - out free - text messages ) , ( c ) send messages \u2014 without clues \u2014 to neighbors , ( d ) reply to or forward messages received from neighbors , ( e ) register an answer ( along with a measure of one \u2019 s con \ufb01 dence in it ) , and ( f ) see the answers ( and con \ufb01 dence levels ) registered by neighbors . Other than the use of one \u2019 s time and the fact that payment depends on registering answers , there are no explicit costs or incentives for any of these actions ( see Appendix B for details on payment ) . Structure of Task At the beginning of the trial , 3 every participant received an initial clue in the form of a \u201c message from headquarters \u201d announcing that the victim is missing and presumed dead . Subsequent clues could be found by a search using keywords from the initial clue , which in turn produced clues about other people with whom the victim had interacted ( i . e . , possible sus - pects ) . Using their names as search terms produced more information about them , including possible motives or connections . Each trial lasted 15 minutes and had three phases in which new information became accessible . The \ufb01 rst set of clues established a red herring by bringing up a speci \ufb01 c suspect and establishing a motive . Five minutes into the \ufb01 rst phase , the second phase began , with a new message from headquarters \u2014 broadcast to all participants \u2014 revealing new information ( and therefore , new search terms ) that led to information not previously accessible . With suf \ufb01 cient search and deduction , it was now possible \u2014 although not straight - forward \u2014 to rule out the original red herring and dis - cover the real culprit . At the 10 - minute mark , the third phase began . A new \u201c message from headquarters \u201d was broadcast with a critical clue that ruled out the most prominent red herring and drew attention to the real culprit . Figure 4 shows the percentage of participants who found the correct answer , the red herring answer , or one of four other answers over time and illustrates the predominant switch toward the correct answer in the last \ufb01 ve minutes . During phase 2 , a majority had reg - istered the red herring , and fewer than 10 % had iden - ti \ufb01 ed the correct answer even though it was logically possible to do so . For 95 % of the participants for whom more than half of their visible peers had entered an answer prior to the third phase , the majority of those visible answers were matching and incorrect ; that is , those participants saw a consensus but on a wrong answer . For 69 % of the participants for whom more than half of their visible peers had entered an answer prior to the third phase , all visible answers were match - ing and incorrect . During the third phase , when it was logically straightforward to deduce the correct answer , a sizeable minority retained incorrect answers . Network Treatments We examine six network treatments with nine nodes ( participants ) each . The two centralized networks that are the focus of our study are the hub - and - spoke net - work and the core - periphery network . The hub - and - Figure 3 . ( Color online ) Screenshot of the Experimental Interface Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2072 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . spoke is the maximally centralized network , with a single central node to which all others are connected , like a group with a single coordinator who brokers interactions between the others . As the most central - ized structure possible , the hub - and - spoke network ( or close variants ) has been used in prior experimental studies of networks ( e . g . , Bavelas 1950 , Becker et al . 2017 , Argote et al . 2018 ) . It is also a representation of the communication pattern commonly associated with hierarchical organization and has been \u2014 and in some cases , remains \u2014 a default in organizations ( Figure 5 ) . The core - periphery network is a less centralized but common form of centralization found in self - organized social networks ( see , e . g . , Colizza et al . 2006 , Kane and Alavi 2007 , Cattani and Ferriani 2008 , Connelly et al . 2011 , Dahlander and Frederiksen 2012 ) and in organizations ( e . g . , a leadership team that brokers interactions between the other group mem - bers ) . It consists of a central group ( the core ) \u2014 rather than a single central individual \u2014 and peripheral indi - viduals connected to the core but not to each other ( Borgatti and Everett 2000 ) . The existence of a central group is of practical importance because of the greater conformity pressure felt in network positions with a clustered pattern of ties ( Centola 2010 ) . We tested two versions of the core - periphery net - work with directed ( one - way ) ties in order to study the mechanisms underlying the performance effects of centralization that we postulated in deriving our hypotheses . One , the \u201c out core - periphery \u201d network , differs from the undirected version of the core - periphery network in that the core has few incoming ties from the periphery . This structure occurs in self - organized networks in which communication and in \ufb02 uence do not \ufb02 ow from the periphery to the core ; for example , digital social media in which one user can unilaterally follow another ( Frahm and Shepe - lyansky 2012 ) or hierarchical organizations in which the leader communicates to the members but the members do not speak upward to the leader ( or alter - natively , the leader fails to listen ) . The second directed core - periphery network \u2014 the \u201c in core - periphery \u201d net - work \u2014 reverses the pattern : the core has many incom - ing ties from the periphery but few outgoing ties ( e . g . , a leadership team that gathers information from but does not communicate back to other organization members , as in organized crime ( Aven 2010 ) and the U . S . Central Intelligence Agency ( Hackman 2011 ) ) . In our analyses , we compare these directed core - periphery networks with both the undirected central - ized networks and the control networks . Finally , we tested three decentralized network struc - tures as controls . The locally clustered network consists , in our case , of three interconnected cliques of three indi - viduals , akin to interconnected teams ( a team of teams ) or groups of friends . The complete clique is a group of nine people , all of whom are connected to each other , like a large work group ( e . g . , a so - called \u201c all hands on deck \u201d collaboration ) or people using a shared digital communication channel . Finally , we constructed a syn - thetic treatment of \u201c isolates , \u201d which are groups of nine individuals who were resampled with replacement from the set of all participants in the core - periphery network and who had no incoming ties ( positions 5 , 7 , and 9 ) . Isolates represent the disconnected nature of \u201c crowds \u201d ( Afuah and Tucci 2012 ) or participants in a competition ( Boudreau et al . 2011 ) . These three decentralized controls represent what we believe are empirically plausible alternatives to centralization , with different levels of social in \ufb02 uence . Although random network structures ( e . g . , Becker et al . 2017 ) or extremal network structures ( e . g . , Mason and Watts 2012 ) have sometimes been used as controls for network experiments , we wanted controls that could also function as plausible null hypotheses \u2014 plausible alternative modes of organiz - ing for problem - solving that exist in the real world . See Appendix E for more detail on the network treatments . We collected 30 independent trials for each experi - mental condition and 90 bootstrap - sampled synthetic trials of isolates . Network treatments were collected in randomly permuted order to avoid systematic bias because of time effects . Fewer than 2 % of participants dropped out midway , but this resulted in approxi - mately 16 % of trials being affected by dropout at Figure 4 . ( Color online ) Percentage of Participants with Each Answer over Time 0 . 0 0 . 2 0 . 4 0 300 600 seconds pe r c en t age answer type red herring correct other Note . \u201c Other \u201d includes four incorrect suspects . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2073 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . some point . In estimating the causal effects of treat - ments , we analyzed all trials , regardless of dropout , to avoid inadvertently conditioning on posttreatment latent variables ( Montgomery et al . 2018 ) . 4 Participants were given no information about their network structure or their position within it other than the fact that communication ties were not neces - sarily symmetrical ; the people they followed ( received information from ) might not follow them back . Participant Pipeline We recruited participants through Amazon Mechani - cal Turk ( MTurk ; https : / / www . mturk . com ) , located in the United States . Building on work by Mason and Suri ( 2012 ) , we constructed a pipeline using a task sequence consisting of three stages : ( Stage 1 ) a 3 - minute individual task in which participants learn to search for clues and enter their answers as well as learn the basic format of a murder mystery task , ( Stage 2 ) a 5 - minute task in which three participants learn to use the social features of the interface , and ( Stage 3 ) the main 15 - minute trial for nine people . Requiring two tasks to be complete before entering the main trial played the dual roles of ensuring that participants were familiar with the platform before the main experiment and \ufb01 ltering out participants who were inattentive , disinterested , unable to com - plete the task , or otherwise likely to drop out of a syn - chronous trial . Low attention among MTurk workers is a rational response to the preponderance of low - paying and poorly designed tasks ( Bigham 2014 ) . However , dropouts and inattentive participants affect the performance of everybody else in the trial , so min - imizing them was a priority . In addition , random dropout ( i . e . , dropout not caused by something spe - ci \ufb01 c to network treatment ) is not a problem from a Figure 5 . Experimental Treatments Notes . ( Left panel ) Visualization of network structures . An arrow from node i to node j indicates that node j can receive clues from node i and see its answers but not the other way around . ( Right panel ) Adjacency matrix visualizations of the same treatments . A \ufb01 lled - in square in position ( i , j ) indicates that node j can receive clues from node i and see its answers . All ties in the hub - and - spoke , core - periphery , locally clustered , and complete clique are bidirectional , with the sole exception of the ties between 1 , 4 , and 7 in the locally clustered network , which are unidirectional for the purpose of preserving degree . Most ties in the out core - periphery and in the in core - periphery are one way between the core and periph - eral nodes , as indicated by the arrows , with the exception of the ties between nodes 1 and 4 , nodes 2 and 6 , and nodes 3 and 8 for the purposes of preserving degree and external validity . See Appendix E for more detail . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2074 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . statistical inference perspective but does introduce prob - lems of validity of interpretation ; if the central node in the hub - and - spoke network drops out , this creates a large de facto change in network structure . We thus took pains to create an effective \ufb01 lter \u2014 prior to the main task \u2014 that was still fair to those whom we \ufb01 ltered out . Participants were paid according to the length of time that they had the correct answers entered into the user interface ( including both the trivial attention check question and the main question of \u201c whodunnit \u201d ) . Median wages were $ 8 . 47 per hour . See Appendix B for full details on the subject pipeline , including task speci - \ufb01 cations , \ufb01 ltering and representativeness of subjects , and payment . The Nature of Communication and Collaboration Among Participants Participants could only receive information from each other or see each other \u2019 s answers if they shared a net - work tie . A network tie from person i to person j meant that j could always see i \u2019 s answers to the prob - lem . Sharing clues or free - text messages was also only possible with a network tie but was not required or explicitly incentivized . However , the instructional text that we provided articulated the collaborative nature of the experiment . Speci \ufb01 cally , the instructions for stage 1 state that \u201c [ t ] he real HIT is a collaborative game \u201d ( \u201c HIT \u201d\u2014 human intelligence task \u2014 is the standard term for a paid task on MTurk ) , and we provided the following information on interacting with other subjects for the stage 3 ( main ) trial : \u201c Like last time , you will be solving a \ufb01 ctional mur - der mystery . Some clues will be given to you , and you must use a search tool to \ufb01 nd other clues . 1 . This time , you will be working with other Turkers in real time . Be sure to watch for mes - sages from other players . Everybody will be given a fake name ; your real identity will be hidden from the other players . 2 . Note that you can only send messages to people listed in the messaging window . Some people you \u201c follow \u201d don \u2019 t \u201c follow \u201d you back . [ \u2026 ] The purpose of the study is to learn about how people solve challenging problems together . You will be collaborating with other participants to solve a \ufb01 ctional murder mystery game . You will receive some clues , and you will have to search for other clues to be used to solve the game . You will be paid based on the time you have entered the correct answers . \u201d In practice , subjects did use the communication and collaboration features frequently . The mean number of messages passed per trial was 75 . 7 ( 25th percentile ( Q1 ) (cid:2) 59 , 75th percentile ( Q3 ) (cid:2) 90 ) ; most messages shared clues that someone had found without any additional text comments . Note that search actions themselves were not visible to peers , but the clues found through search could be voluntarily shared . Participants did use free - text communication as well , with a mean of 14 . 5 messages per trial ( Q1 (cid:2) 7 , Q3 (cid:2) 20 ) . Most free - text messages were collaborative in the sense that participants discussed clues and their implications . Occasionally , process - oriented and con - versational messages were also included . A represen - tative selection of messages includes the following : \u2022 \u201c So the major project launch was in Honolulu , yes ? Where Ms . Moore and Ms . Jones were \u201d \u2022 \u201c Seems a lot of people had dealings with Anderson \u201d \u2022 \u201c Mr . Brown is the only one who got hurt twice by hall \u201d \u2022 \u201c Looks like you have info on Hall and I have info on the suspects \u201d \u2022 \u201c Have you found anything regarding a murder weapon ? \u201d \u2022 \u201c Davis has a gun , but was gone . So that leaves Brown and Lewis \u201d The spread of these free - text messages could be tracked as easily as the spread of the of \ufb01 cial clues , giv - ing us a comprehensive view of how information \ufb02 owed through our network treatments . We provided a channel of nontextual communica - tion for registering answers ; participants were also prompted for a con \ufb01 dence level ( which defaulted to 50 % ) . Our intention was not to measure an outcome variable but rather to enable more effective collective problem - solving . Previous research ( Bahrami et al . 2010 , Engel et al . 2014 , Madirolas and De Polavieja 2014 , Becker et al . 2017 ) shows the importance of com - municating con \ufb01 dence levels to collective intelligence outcomes . We believe that this feature also increases the external validity of the experiment , as most collab - orative settings allow some means of expressing con \ufb01 - dence in a solution . Results Our dependent variable is the number of participants correct at the end of the trial ; thus , it is measured at the network level but is a count of individual - level outcomes within each trial . 5 We study the effects of social in \ufb02 uence within networks and thus expect both the mean and the shape of the distribution of our dependent variable to vary with network treatment . For example , in a trial of isolates , we expect the distri - bution of the number of participants correct per trial to be unimodal and binomially distributed around some mean ( because participants solve the problem independently ) , but in a trial of a complete - clique net - work , we expect the distribution to be bimodal Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2075 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . ( because participants solve the problem with social in \ufb02 uence from every other participant , thus prompting an expectation of high correlation among their ans - wers that will , therefore , be either mostly correct or mostly incorrect ) . Our hypothesis tests rely not only on the means of these distributions but also on their quantiles in order to capture differences in collective behavior , as described . Main Result In this experiment , which required adapting to a shift - ing environment by changing from one temporarily justi \ufb01 able answer to a superior one , centralized networks performed best . They minimized the proba - bility of getting stuck on an incorrect consensus ( Hy - pothesis 1 ) while maximizing the spread of the correct answer ( Hypothesis 2 ) , thus getting the bene \ufb01 ts of collaboration without the costs ( Hypothesis 3 ) but only with two - way versus one - way communication ( Hypothesis 4 ) . We found no signi \ufb01 cant performance differences caused by network position ; the key causal variable was the whole - network structure . Basic descriptive statistics are shown in Table 3 . Figure 6 provides an overview of the distribution of how many participants were correct at the end of the trial \u2014 by experimental condition \u2014 in the form of empirical cumulative distribution functions ( eCDFs ) , which plot the percentage of trials that ended with a certain number of people correct or fewer . The farther the eCDF curves are to the right side of the panel , the better the performance in terms of the percentage of participants correct per trial . Panel ( a ) shows the eCDFs for the nominal groups of isolates ( minimal social in \ufb02 uence ) and complete clique ( maximal social in \ufb02 uence ) conditions . Compared with isolates , the com - plete clique tended to have more trials with very few participants correct ( marked \u201c costs \u201d ) and more trials with many participants correct ( marked \u201c bene \ufb01 ts \u201d ) . Subsequent panels overlay the eCDF from another net - work treatment on those of the isolates and complete clique for easier visual interpretation . Panel ( b ) shows results for the locally clustered network , which has fewer costs and fewer bene \ufb01 ts than the complete clique . Panels ( c ) and ( d ) in Figure 6 show results for the centralized network treatments ( see Figure 11 for results for the directed core - periphery treatments that serve as tests of mechanisms ) . Panel ( c ) shows that the hub - and - spoke network had few trials with few participants correct ( it did not experience the costs of social in \ufb02 uence ) but many trials with many par - ticipants correct ( it did bene \ufb01 t from social in \ufb02 uence ) . Panel ( d ) shows that the core - periphery did suffer some of the costs \u2014 although less than the complete clique treatment did \u2014 while enjoying similar bene \ufb01 ts . Tests of proportions show no signi \ufb01 cant differences in the costs of social in \ufb02 uence ( as measured by the proportion of trials with zero or one participant cor - rect at the end ) between the isolates and both central - ized network treatments . We thus reject the null for Hypothesis 1 . We also found no signi \ufb01 cant differences in the bene \ufb01 ts of social in \ufb02 uence ( as measured by the proportion of trials with greater than six participants correct at the end ) between the complete clique and both centralized network treatments . We thus reject the null for Hypothesis 2 . However , the out core - periphery had a greater proportion of trials with zero or one person correct than did the isolates ( p (cid:2) 0 . 03 , one - tailed test ) and a smaller proportion of trials with greater than six people correct than did the complete clique ( p (cid:2) 0 . 06 , one - tailed test ) , consistent with Hypothesis 4 . Models of the number of participants correct at the end of a trial put these comparisons in a regression framework , reported in Table 4 . Quantile regressions ( which model conditional quantiles instead of the more typical conditional mean ) provide a perspective on the costs and bene \ufb01 ts of social in \ufb02 uence that is complementary to that of the tests of proportions reported above . We expect outcomes to vary , even with the same treatment . If people become stuck on the wrong answer , there will be more trials toward the bottom of the performance distribution with very few participants correct at the end of the trial . There - fore , we use the 25th percentile of the distribution ( models 1 and 5 ) to provide another test of Hypothesis 1 . Similarly , if the correct answer spreads widely , it will increase the number of trials toward the top of the performance distribution with very many participants correct at the end of the trial . Therefore , we use the 75th percentile ( models 4 and 8 ) to provide another test Table 3 . Descriptive Statistics by Network Treatment Hub - and - spoke Core - periphery Isolates Locallyclustered Completeclique Out core - periphery In core - periphery N 30 30 90 30 30 30 30 Mean ( no . correct / network ) 4 . 7 4 . 367 3 . 189 3 . 2 3 . 533 3 . 133 4 Std . dev . ( no . correct / network ) 2 . 706 3 . 079 1 . 381 2 . 497 3 . 391 2 . 417 1 . 965 Median ( no . correct / network ) 4 . 5 4 3 3 1 2 . 5 4 Notes . N refers to the number of independent networks . Other statistics are in terms of the number of individuals per network . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2076 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . of Hypothesis 2 . For completeness , we model the median of the distribution of the dependent variable ( models 2 and 6 ) . Quantile regressions for count varia - bles ( all of the models ) are \ufb01 t through the method of Machado and Silva ( 2005 ) in the R package lqmm ( Ger - aci and Bottai 2014 ) . Regression models for the mean number of participants correct ( models 3 and 7 ) test Hypothesis 3 and were \ufb01 tted as quasi - Poisson generalized linear models to account for overdispersion from social in \ufb02 uence . Appendix D discusses model choice at greater length . Models 1 \u2013 4 use the hub - and - spoke as the refer - ence category ; models 5 \u2013 8 duplicate models 1 \u2013 4 but use the core - periphery network as the reference category . In each model , the only parameters included are the constant ( intercept ) and dummy variables corresponding to the network treatment . Hypothesis 4 is tested by means of the estimated parameters for the out core - periphery and in core - periphery networks . More on the mechanisms underlying performance in these directed networks can be found in the \u201c Mechanism Check \u201d section . Figure 6 . ( Color online ) Empirical Cumulative Distribution Functions of the Number of Individuals Correct per Trial costs benefits 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 number correct ( of 9 ) c u m u l a t i v e pe r c en t age o f t r i a l s Isolates Complete clique ( a ) 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 number correct ( of 9 ) c u m u l a t i v e pe r c en t age o f t r i a l s Isolates Complete clique Locally - clustered ( b ) 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 number correct ( of 9 ) c u m u l a t i v e pe r c en t age o f t r i a l s Isolates Complete clique Hub\u2212and\u2212spoke ( c ) 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 number correct ( of 9 ) c u m u l a t i v e pe r c en t age o f t r i a l s Isolates Complete clique Core\u2212periphery ( d ) Notes . Panel ( a ) shows only the isolates ( no social in \ufb02 uence ) and complete clique ( maximal social in \ufb02 uence ) conditions . The complete clique condition has more trials with few participants correct ( the \u201c costs \u201d of social in \ufb02 uence because of conformity pressure ) and more trials with many participants correct ( the \u201c bene \ufb01 ts \u201d of social in \ufb02 uence because of learning the right answer from peers ) . Panel ( b ) adds the locally clustered network , which is the third control condition , which has less cost and less bene \ufb01 t than the complete clique but maintains the tradeoff between them . Panel ( c ) shows that the hub - and - spoke network produced the bene \ufb01 ts of social in \ufb02 uence without the costs . Panel ( d ) shows that the core - periphery network produced the bene \ufb01 ts with reduced costs . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2077 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . The hub - and - spoke network dominated the three controls : it had fewer very poor trials ( with the major - ity wrong ) than the complete clique and locally clus - tered networks ( model 1 ) and more very good trials ( with the majority correct ) than the isolates ( model 4 ) . The average number of participants correct ( Table 2 and model 3 of Table 3 ) was higher for the hub - and - spoke ( 4 . 7 of 9 ) than for the isolates ( 3 . 19 ) , the complete clique ( 3 . 53 ) , and the locally clustered network ( 3 . 2 ) . Quantile regression results for the core - periphery network were more equivocal . Panel ( d ) of Figure 6 shows an effect similar to that of the hub - and - spoke but with several more trials with few people correct because of the presence of a central cluster . The core - periphery network thus outperformed the isolates and the locally clustered network in disseminating the correct answer ( 75th percentile , model 8 ) and the mean ( model 7 ) , but it was only directionally consis - tent with the hub - and - spoke network and not statisti - cally signi \ufb01 cantly better with respect to getting stuck on the wrong consensus ( model 5 ) . Given that the core - periphery network is less centralized than the hub - and - spoke , a smaller effect is expected and still consistent with our hypotheses . All that said , the core - periphery has the same proportion of trials with zero or one person correct as the isolates condition , consis - tent with a \ufb01 nding of minimal costs because of social in \ufb02 uence ( low probability of sticking to an inferior consensus ) . Figure 7 documents the size of the causal effects of the treatments , relative to the controls , in terms of \u201c common language effect size , \u201d which expresses the probability that the treatment would perform better than the controls if one trial in the treatment condition and one in a control condition were run ( or picked at random from the data ) . 6 It thus accounts for variance in outcomes for control and treatment networks . Compared with the isolates , locally clustered net - works , and complete cliques , the hub - and - spoke would perform better 60 % , 62 % , and 59 % of the time , respectively , and the control network would perform better 29 % , 30 % , and 33 % of the time , respectively . Compared with the isolates , locally clustered net - works , and complete cliques , the core - periphery would perform better 56 % , 56 % , and 52 % of the time , respectively , and the control network would perform better 35 % , 35 % , and 36 % of the time , respectively . Additionally , the distribution of performance of the hub - and - spoke network is \ufb01 rst - order stochastically dominant over that of the locally clustered network , meaning that the dominant distribution should be chosen by all decision makers ( who want high per - formance ) because for any level of performance , x , the probability that the performance of the hub - and - spoke Table 4 . Number of Participants Correct at End of Trial Dependent variable 25th percentile Median Mean 75th percentile 25th percentile median mean 75th percentile ( 1 ) ( 2 ) ( 3 ) ( 4 ) ( 5 ) ( 6 ) ( 7 ) ( 8 ) Hub - and - spoke 0 . 099 0 . 083 0 . 074 0 . 068 ( 0 . 566 ) ( 0 . 354 ) ( 0 . 149 ) ( 0 . 247 ) Core - periphery \u2212 0 . 341 \u2212 0 . 046 \u2212 0 . 074 0 . 058 ( 0 . 536 ) ( 0 . 268 ) ( 0 . 149 ) ( 0 . 196 ) Isolates 0 . 036 \u2212 0 . 327 \u2212 0 . 388 * * * \u2212 0 . 502 * * * 0 . 133 \u2212 0 . 261 \u2212 0 . 314 * * \u2212 0 . 466 * * ( 0 . 165 ) ( 0 . 222 ) ( 0 . 126 ) ( 0 . 187 ) ( 0 . 555 ) ( 0 . 163 ) ( 0 . 129 ) ( 0 . 190 ) Locally clustered \u2212 0 . 602 * * \u2212 0 . 399 \u2212 0 . 384 * * \u2212 0 . 323 * \u2212 0 . 513 \u2212 0 . 329 * \u2212 0 . 311 * \u2212 0 . 291 ( 0 . 295 ) ( 0 . 248 ) ( 0 . 162 ) ( 0 . 193 ) ( 0 . 607 ) ( 0 . 197 ) ( 0 . 165 ) ( 0 . 197 ) Complete clique \u2212 0 . 947 * * * \u2212 0 . 863 \u2212 0 . 285 * 0 . 024 \u2212 0 . 854 \u2212 0 . 898 * \u2212 0 . 212 0 . 070 ( 0 . 266 ) ( 0 . 658 ) ( 0 . 157 ) ( 0 . 186 ) ( 0 . 593 ) ( 0 . 520 ) ( 0 . 160 ) ( 0 . 189 ) Out core - periphery \u2212 0 . 524 * \u2212 0 . 468 \u2212 0 . 405 * * \u2212 0 . 467 * * \u2212 0 . 414 \u2212 0 . 388 \u2212 0 . 332 * * \u2212 0 . 429 * * ( 0 . 277 ) ( 0 . 519 ) ( 0 . 163 ) ( 0 . 194 ) ( 0 . 598 ) ( 0 . 500 ) ( 0 . 166 ) ( 0 . 197 ) In core - periphery 0 . 110 \u2212 0 . 095 \u2212 0 . 161 \u2212 0 . 257 0 . 177 \u2212 0 . 020 \u2212 0 . 088 \u2212 0 . 214 ( 0 . 231 ) ( 0 . 242 ) ( 0 . 152 ) ( 0 . 192 ) ( 0 . 581 ) ( 0 . 189 ) ( 0 . 155 ) ( 0 . 195 ) Constant 0 . 891 * * * 1 . 468 * * * 1 . 548 * * * 1 . 867 * * * 0 . 795 1 . 403 * * * 1 . 474 * * * 1 . 831 * * * ( 0 . 140 ) ( 0 . 218 ) ( 0 . 103 ) ( 0 . 178 ) ( 0 . 548 ) ( 0 . 157 ) ( 0 . 107 ) ( 0 . 182 ) Observations 270 270 270 270 270 270 270 270 Note . All coef \ufb01 cients are on the log scale . * p < 0 . 1 ; * * p < 0 . 05 ; * * * p < 0 . 01 . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2078 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . was at least x was greater than or equal to the proba - bility that the performance of the locally clustered net - work was at least x and , for some x , the performance of the hub - and - spoke network was strictly better . The core - periphery , although not differing as greatly in mean performance from the controls , is still \ufb01 rst - order stochastically dominant over the locally clustered and complete clique controls . With respect to the hub - and - spoke network , our experimental data are suf \ufb01 cient to reject the null for Hypothesis 1 ( reduced probability of getting collectively stuck on an inferior answer as measured by a test of proportions and 25th percentile regression , model 1 ) , Hypothesis 2 ( maintained spread of a superior answer as measured by a test of proportions and the 75th per - centile regression , model 4 ) , and Hypothesis 3 ( overall positive average effect , model 3 ) . For the less - centralized core - periphery network , we reject the null for Hypothe - sis 1 ( test of proportions ) , Hypothesis 2 ( test of propor - tions and model 8 ) , and Hypothesis 3 ( model 7 ) . Overall , given that quantile regression effect sizes are smaller but directionally consistent for the less - centralized net - work \u2014 and given our results from tests of propor - tions \u2014 we interpret the data overall as supporting Hypotheses 1 \u2013 3 . Tests of proportions show that the out core - periphery was signi \ufb01 cantly worse than the isolates in terms of the costs of social in \ufb02 uence and worse than the complete clique in terms of the bene \ufb01 ts . It also had a substantially and statistically signi \ufb01 cantly lower mean and 75th percentile than either of the two undirected centralized treatments ; we , therefore , reject the null for Hypothesis 4 ( see results from tests of proportions and parameter estimate for out core - periphery in models 3 , 4 , 7 , and 8 ) . Mechanisms We have shown that centralized networks can adapt better to new information and spread the correct solu - tion . We also posited certain mechanisms : speci \ufb01 cally , that the inability of peripheral nodes to communicate directly with each other could reduce conformity pressure and allow those individuals to adapt to new information and that the privileged position of the central nodes could allow them to spread good ideas that arose in the periphery . We now present step - by - step evidence of those mechanisms . Boxes A \u2013 C of Figure 2 identify the mechanisms implied by our hypotheses : that is , a more detailed account of what we should expect to observe in our data if the hypothesized outcomes do , in fact , occur . Note that although these mechanisms were part of the logic motivating our hypotheses at the whole - network level , they were never articulated as separate hypotheses prior to running the experiment , although we did intentionally collect detailed data for the purpose of exploring them . Note also that because results at the posi - tional or individual levels are endogenously correlated , an experiment like ours cannot show them as causal , and they should be treated as descriptive correlations . Mechanism A : Peripheral Nodes Are Relatively Independent We begin by providing evidence that peripheral nodes feel less conformity pressure to stay with the consensus solution than those in other network posi - tions . Figure 8 plots the probability of having the same answer as one \u2019 s neighbors for the correct answer ( upper panels ) and for the red herring ( lower panels ) . Peripheral positions covary less with their neighbors than other positions do . Speci \ufb01 cally , the left side of Figure 7 . ( Color online ) Visualization of \u201c Common Language \u201d Effect Size : Choosing One Random Trial of a Centralized Network ( Left Panel : Hub - and - Spoke , Right Panel : Core - Periphery ) and One Random Trial of a Control Network Structure , What Are the Chances That the Treatment Will Be Better Than , Equal to , or Worse Than the Control in Terms of the Number of Participants Correct at the End of the Trial ? 0 . 33 0 . 08 0 . 59 0 . 3 0 . 08 0 . 62 0 . 29 0 . 11 0 . 6 complete clique locally\u2212clustered isolates 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 probability of result c on t r o l hub\u2212and\u2212spoke better equal control better p ( hub\u2212and\u2212spoke network better than controls ) 0 . 36 0 . 12 0 . 52 0 . 35 0 . 09 0 . 56 0 . 35 0 . 09 0 . 56 complete clique locally\u2212clustered isolates 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 probability of result c on t r o l core\u2212periphery better equal control better p ( core\u2212periphery network better than controls ) Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2079 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . Figure 8 . Probability of Having the Same Answer as Neighbors by Position During Phase 3 of the Experiment Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2080 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . each panel shows that peripheral nodes are more likely than other positions to adopt a given answer ( cor - rect or not ) , even if no neighbor has adopted it . Simi - larly , they are less likely than other positions to adopt an answer that all of their neighbors have adopted . Figure 8 shows that both the number and share of neighbors with a given answer matter in deriving the probability of adopting a matching answer , as pre - dicted by others ( Stasser and Davis 1981 , Centola and Macy 2007 ) and shown in Figure 1 . For example , a member of the periphery of the hub - and - spoke net - work has only a single neighbor ( the central node ) ; this is 100 % of what that person can see . Thus , that single answer has a higher weight than one of several answers visible to a position with multiple neighbors . Conversely , when a given share of neighbors adopts an answer , the conformity pressure increases with the actual number of neighbors ; for example , when all of one \u2019 s eight neighbors agree , the probability of match - ing them is 85 % \u2013 100 % , whereas the probability of matching one \u2019 s only neighbor is approximately 55 % . Mechanism B : The Core Sees More Correct Answers , Even When Wrong Itself If peripheral nodes are less subject to conformity pres - sure , then the core should see more diverse solutions . The upper panels in Figure 9 visualize the moving average of the number of unique answers visible to individuals in each network position over time . Although the general trend is similar across positions , the core nodes in the hub - and - spoke and core - periphery conditions saw more unique answers on average ( 3 . 12 and 2 . 65 , respectively ) than did individuals in the complete clique , locally clustered , and isolates con - ditions ( 2 . 44 , 1 . 58 , and 0 , respectively ) . Isolates are not shown in the \ufb01 gure because , by construction , they saw no peer answers . The \u201c Mechanism Check \u201d section dis - cusses results for the directed core - periphery networks . Beyond seeing more diverse answers , core individuals saw more correct answers , even when they were wrong themselves . Of participants with the wrong answer reg - istered in the \ufb01 nal phase , those in the core position of the hub - and - spoke saw 0 . 267 more correct answers among neighbors than those in the complete clique , whereas those in the core of the core - periphery network saw 0 . 203 more ( p (cid:2) 0 : 008 and 0 . 011 , respectively ) . Com - paring core nodes of centralized networks with nodes in the locally clustered network , the core node in the hub - and - spoke saw 1 . 074 more correct alters , and core nodes in the core - periphery network saw 1 . 009 more . Mechanism C : The Core Learns from the Periphery Figure 8 demonstrates that those who saw more correct answers were more likely to be correct themselves . Given the nature of our data , we cannot conclusively show who learned from whom , but to connect mecha - nism C to the main result , we can show that the effects occurred in the sequence described : core nodes see the correct answer in the periphery and adopt it , and then it spreads farther to other peripheral nodes . Put another way , ( a ) individuals in the core adopted the correct answer after some in the periphery did and before others in the periphery did , and ( b ) more participants were correct in those trials in which the core adopted the correct answer . To assess ( a ) , we conducted a random permutation test of the order in which individuals in a given trial adopted a correct answer that they stuck with until the end . We randomly permuted the sequence of adopting the correct answer \u2014 coding anybody who did not end the trial with the correct answer as zero and everyone else with the whole number indicating when in the sequence they adopted the correct answer \u2014 in groups of 30 trials . We repeated this process 10 , 000 times to create a null distribution to which to compare the real data . The central node in the hub - and - spoke was statistically signi \ufb01 cantly ( p (cid:2) 2 \u00d7 10 \u2212 4 ) more likely to be the third person in the trial to adopt the correct answer , and members of the core in the core - periphery network were signi \ufb01 cantly ( p (cid:2) 0 . 01 ) more likely to be the fourth . As for point ( b ) , in the hub - and - spoke network , when the central node adopted the correct answer and maintained it to the end , 64 % of peripheral nodes that had not yet adopted the correct answer when the central node did went on to do so as compared with 29 % for cases in which the central node had not yet adopted the correct answer ( the difference is statisti - cally signi \ufb01 cant at p (cid:2) 10 \u2212 6 ) . In core - periphery net - works , the probability that a peripheral node is correct at the end goes up sharply if two or three core nodes have already adopted the correct answer . The proba - bility of completing the trial with the correct answer is 13 % if no core nodes have already adopted it but 36 % if two core nodes have and 57 % if three core nodes have ( p (cid:2) 0 . 022 and 2 . 92e - 04 , respectively ) . Figure 10 illustrates both of these \ufb01 ndings for the hub - and - spoke network . Mechanism Check The individual - level data discussed in reporting mec - hanisms A \u2013 C in Figure 2 are endogenously correlated within each trial . Thus , although we have detailed de - scriptive data , they do not support causal inferences . However , data from the directed core - periphery network treatments can be studied to increase the plausibility of those causal mechanisms , because those treatments exo - genously manipulate key variables in the mechanism diagram . Because the out core - periphery structure does not allow the core to see the periphery ( and therefore Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2081 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . Figure 9 . ( Color online ) Core Nodes See Diverse Answers , Including Correct Answers , Even When They Are Wrong Themselves U n i que s o l u t i on s C o rr e c t a l t e r s Notes . ( Upper panels ) Moving average of the number of unique answers visible to each position over the course of the experiment . ( Lower panels ) Box plot of the number of correct answers visible to individuals who did not have the correct answer registered during the last minute of the experiment . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2082 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . communication is only one way ) , all statements logi - cally downstream from \u201c core sees periphery \u201d ( includ - ing all of mechanisms A , B , and C ) should be false for that treatment condition ( per Hypothesis 4 ) . For mecha - nism A , as we argued , the core \u2019 s exposure to the pos - sibly differing answers in the periphery lessens the conformity pressure on the periphery ( because of the self - reinforcing nature of conformity ) , so when the core cannot see the periphery , we should not expect the periphery to be as independent . Figure 8 con \ufb01 rms that . As for mechanism B , the core , unable to see the periphery , cannot learn from it . Thus , the core should see a less diverse set of answers \u2014 including fewer cor - rect answers \u2014 than it would in the hub - and - spoke net - work or the ( undirected ) core - periphery network . Figure 9 con \ufb01 rms that . For mechanism C , individuals in the core of the out core - periphery should , therefore , be less likely to be correct than individuals in the core positions of undirected centralized networks . Indeed , we \ufb01 nd that they were correct 32 % of the time compared with 57 % for the core of the hub - and - spoke network and 51 % for the core of the undirected core - periphery network ( p - val - ues from \u03c7 2 tests are 0 . 029 and 0 . 014 respectively ( esti - mated on those who completed the trial ) ) . With fewer trials in which members of the core choose the correct answer , there are fewer in which the core spreads the correct answer to the periphery ( see Figure 11 and results from Table 4 ) , indicating little social learning . The in core - periphery network had a different pat - tern of results . The core could see the periphery but not vice versa . The mechanism diagram , therefore , predicts that the core would see many correct answers among the independent peripheral nodes . Indeed , because peripheral individuals in the in core - periphery had fewer network neighbors than peripheral individu - als in other networks \u2014 or none at all \u2014 they were even more independent ( see Figure 9 , lower panels ) . Individ - uals in the core , like those in the undirected centralized networks , should , therefore , be correct more frequently ; in fact , they were 49 % of the time . Despite these struc - tural bene \ufb01 ts , the core \u2019 s correct answers could not spread as widely , because three peripheral individuals had no incoming information from the core and three only had one link from a single member of the core . Figure 6 and Table 4 show that although results were not signi \ufb01 cantly different from those for decentralized networks , there were fewer trials with many correct individuals , and the 75th percentile of the number of participants correct per trial was only 5 compared with 6 . 47 for the hub - and - spoke and 6 . 24 for the undirected core - periphery network . Discussion and Conclusion Prior research has found centralization harmful for collective problem - solving and collective intelligence because it impedes the integration of ideas and infor - mation held by peripheral individuals . We argue \u2014 and our experimental results show \u2014 that these same effects of centralization nevertheless have bene \ufb01 ts for tasks requiring collective adaptation : that is , for solving problems that require a collective shift from one solution to another in response to a shifting environment . These bene \ufb01 ts \ufb02 ow from the way social in \ufb02 uence is shaped by two de \ufb01 ning features of highly centralized networks , namely that peripheral nodes ( a ) are not Figure 10 . ( Color online ) Illustration of the Time Course of Final Correct Answers for Hub - and - Spoke Network Trials core incorrect core correct 300 600 900 123456789 123456789 time of final correct answer c u m u l a ti v e nu m b e r c o rr ec t Position : periphery core Notes . If the central node is correct , it has typically adopted the correct answer after it had already been found in the periphery and then spreads that answer widely . ( Upper panel ) Trials in which the central node adopted the correct answer . ( Lower panel ) Trials in which the central node did not \ufb01 nish with the correct answer . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2083 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . ( densely ) interconnected with each other and ( b ) are connected with central nodes ( the core ) . Peripheral nodes are , therefore , relatively independent because they do not in \ufb02 uence each other , whereas the core nodes that do in \ufb02 uence them are , in turn , exposed to many \u2014 possibly differing \u2014 solutions . This lack of con - formity pressure makes the periphery more adaptable to new information and less likely to retain a consen - sus ( but inferior ) solution . Core nodes are , therefore , exposed to the diverse solutions of many of these inde - pendent peripheral nodes and thus can learn from the nodes with superior solutions . Once the core nodes adopt a superior solution , it spreads quickly to the periphery . Thus , by limiting conformity pressure but retaining ef \ufb01 cient connectivity , centralization struc - tures use social in \ufb02 uence to promote learning good solutions without undue pressure to stay on a current inferior consensus . However , for the bene \ufb01 ts of cen - tralization to occur , there must be two - way communi - cation between core and periphery . Boundary Conditions and Limitations Our experimental design , although enabling causal inference , does come with important limitations too . Figure 11 . ( Color online ) Empirical Cumulative Distribution Functions of the Number of Individuals Correct per Trial 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 number correct c u m u l a t i v e pe r c en t o f t r i a l s Isolates Complete clique Out core\u2212periphery ( a ) 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 number correct c u m u l a t i v e pe r c en t o f t r i a l s Isolates Complete clique In core\u2212periphery ( b ) 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 number correct c u m u l a t i v e pe r c en t o f t r i a l s Core\u2212periphery Out core\u2212periphery ( c ) 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 number correct c u m u l a t i v e pe r c en t o f t r i a l s Core\u2212periphery In core\u2212periphery ( d ) Notes . Panels ( a ) and ( b ) show results for the out core - periphery and in core - periphery along with the isolates ( no social in \ufb02 uence ) and complete clique ( maximal social in \ufb02 uence ) conditions , similar to Figure 6 . Panels ( c ) and ( d ) show the directed core - periphery networks along with the undirected version . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2084 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . First , it was an intentional feature of our design \u2014 a randomized experiment \u2014 that the core was not made up of \u201c special \u201d individuals but instead was randomly drawn . Thus , our design separates the phenomenon of interest \u2014 network centralization \u2014 from confounds , such as hierarchy and power . However , this depar - ture from realism for the sake of valid causal inference attaches a caveat to our \ufb01 ndings . In practice , central individuals are not typically randomly assigned to such positions but rather gain them through experi - ence , in \ufb02 uence , or other qualities . Although our study is focused on advancing theory in a \ufb01 eld that has evolved through theoretical elaborations ( for another example , see Schilke et al . 2018 ) , it is not clear whether , in real life , central nodes would play the observed role of identifying and spreading good ideas generated by peripheral nodes or if instead would privilege their own ideas ( or indeed if peripheral nodes would decline to volunteer those very ideas out of deference ( Bernstein 2012 ) ) . Put another way , our results point to the necessity of humility among cen - tral individuals ; collective adaptation and perform - ance require them not only to identify and spread the periphery \u2019 s good ideas but also to encourage the periphery \u2019 s voice and speaking - up behaviors ( Detert and Burris 2007 , Morrison 2011 , Detert et al . 2013 ) . Second , our experiment tested centralization at the interindividual level , but in real organizations , cen - tralization can also be interteam or interunit . It is not clear from this experiment how centralization would operate at those other levels of analysis , especially if peripheral units were not separate individuals but rather , say , teams ( i . e . , centralization in a team of teams ) . Ugander et al . ( 2012 ) suggest that core individ - uals may treat teams analogously to individuals in terms of how they feel social in \ufb02 uence , but it may be that clustering within peripheral teams is suf \ufb01 cient to reduce peripheral nodes \u2019 freedom from conformity pressure ( our mechanism A ) . We leave this as an open question that would be very challenging to study with a randomized , controlled experiment . Third , our task allowed each participant to enter his or her own answer to the problem . This allows us to measure in \ufb02 uence among participants , which was a focus of our study and is a realistic representation of many problem - solving settings . However , certain types and scales of problem - solving involve all partic - ipants working on a single shared solution to a single instance of the problem ; for example , multiple pro - grammers contributing code to a single piece of soft - ware . ( Conversely , if the multiple contributions were enabled by code forks and pull requests , as on Github . com , that would again be the paradigm we study . ) It is unclear a priori how this difference in problem structure would change our results , but it is possible that working on a shared solution would increase conformity pressure ( or more generally , lock - in on a current solution ) , which could reduce or elimi - nate the bene \ufb01 ts we see . Implications for Research on Network Structure We contribute to the literature on network structure , in which a major topic of inquiry is how to leverage social in \ufb02 uence for improved collective problem - solving and collective intelligence . Prior work has pre - sented two distinct views of the effects of connectivity , and thus social in \ufb02 uence , in problem - solving ; it can harm performance ( the \u201c herding \u201d perspective ) or help it ( the \u201c learning \u201d perspective ) . In the former , connec - tivity is conceptualized as a threat , because like sheep in a herd , individuals follow the group rather than exercising individual judgment , leading often to suboptimal decisions . In the latter , connectivity is con - ceptualized as a bene \ufb01 t because good solutions to problems are hard for individuals to generate and most often need to be learned from peers . We show that \u2014 in the context of adaptation \u2014 centralization bypasses the trade - offs between these negative and positive aspects of connectivity and that it is possible to structure social in \ufb02 uence to produce collective learning ( from exposure to others \u2019 diverse ideas ) while minimizing herding ( from conformity pressure to stick to a current consensus ) . In essence , for prob - lems requiring adaptation ( speci \ufb01 cally , collective search of a rugged solution space ) , our \ufb01 ndings on centralization contradict the preexisting view that network structure implies a \u201c trade - off between main - taining the diversity necessary for obtaining high per - formance in a system and the rapid dissemination of high - performing strategies \u201d ( Lazer and Friedman 2007 ; see also Uzzi 1997 , Hansen 1999 ) . This conclu - sion joins , and strengthens , nascent related work \ufb01 nd - ing that intermittent interaction ( the on - off cycling of connectivity in a network over time ) can produce superior solutions in collective problem - solving ( Bern - stein et al . 2018 ) . It also connects with established work in innovation - related tasks showing that net - works that enable both diversity and learning ( includ - ing small world networks and other structures that combine local clustering with ties to different knowl - edge pools ) are most conducive to success for individ - ual nodes within the network ( Schilling and Phelps 2007 , Tortoriello 2015 , Ter Wal et al . 2016 ) ( without investigating , as we do , the collective success of the whole network ) . Our contributions to the literature on network struc - ture also include a novel perspective on centralization . Our experimental \ufb01 ndings run exactly opposite to the academic ( Lazer and Friedman 2007 , Becker et al . 2017 ) and popular ( Bernstein et al . 2016 ) narrative of centralization suppressing solution diversity . It is pos - sible that the everyday experience of minimal diversity Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2085 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . of thought in centralized structures is misattributed to centralization and actually is a result of other forces . Another possibility is that individual experience of diversity is lessened in centralized networks precisely because they isolate peripheral nodes from each other , thus maintaining the bene \ufb01 cial aggregate diversity required for adaptability . Also , of course , it is possible that this popular negative view of centralization is simply wrong . Implications for Research on Organizational Adaptation We contribute to scholarship on network - level adapta - tion \u2014 and therefore to the literature on exploration and exploitation in searching rugged solution spaces ( Lazer and Friedman 2007 , Mason and Watts 2012 ) \u2014 by clarifying how networks shape outcomes in \u201c complex \u201d problem - solving by creating the conditions for knowledge transfer without conformity pressure . The literature on network structure has investigated the distinction between \u201c simple \u201d and \u201c complex \u201d problem - solving ( Tushman 1979 , Mason and Watts 2012 ) but has achieved only limited consensus as to what makes a problem complex , with a consequent lack of coherent theory on how group or network structure helps solve complex problems ( Almaatouq et al . 2020 ) . Prior research has generally focused either on how networks shape the diversity of available knowledge or on conformity pressure , complex conta - gions , and norms . Bringing them together as we do in this work highlights the type of complexity that is amenable to network interventions \u2014 when good solu - tions are dissimilar to each other and thus require col - lective adaptation ( rather than only balancing the exploration and exploitation behavior of individuals ) , such as the adaptation requirements experienced by many organizations during the COVID - 19 pandemic . Implications for Research on Hierarchy vs . Centralization Third , we contribute to the literature on hierarchy and organization design . Empirically , centralization is con - founded by hierarchy ( Lin 1999 ) , which is the differ - entiation of individuals by power , rank , or status . Hierarchy and centralization are both common in organizations ( Tiedens and Fragale 2003 , Gruenfeld and Tiedens 2010 , Valentine 2018 ) and have been his - torically linked with the same generally negative effects . Both are generally seen to be good for co - ordination ( Galinsky et al . 2012 ) but bad for colla - borative performance ( Van der Vegt et al . 2010 , Greer et al . 2018 ) . Because of their generally confounded nature , hierarchy and social network centralization are unlikely to be distinguishable via observational data , perpetuating the view that they have similar implications for problem - solving . We , however , mani - pulate centralization without hierarchical differences among participants to focus on the in \ufb02 uence of com - munication network structure per se , rather than of power or status hierarchy . We then \ufb01 nd that central - ization improves performance on tasks requiring adaptation \u2014 our primary contribution to the literature on hierarchy and organization design . However , our results may also imply that hierarchy \u2014 because of the centralized network structures it typically induces \u2014 is likely to be effective at adapting to a changing envi - ronment , so long as the leaders can bring themselves to listen to the periphery , encourage speaking - up behaviors , and then seek and spread good ideas to the rest of the organization ( Detert and Burris 2007 , Mor - rison 2011 , Detert et al . 2013 ) . Alternatively , it may be possible to achieve those bene \ufb01 ts for adaptability by keeping centralization while limiting hierarchy \u2014 as with the random individuals in our centralized networks . Implications for Research on Communication Technologies Fourth , we contribute to the literature on using tech - nology to structure communication . Social communi - cation technologies often create decentralized infor - mation \ufb02 ows ( Majchrzak et al . 2009 , Leonardi 2014 ) . Our results imply that such technology may , there - fore , have distinct advantages for integration of knowledge for problems that do not require adapta - tion but may be counterproductive for the problems that do require adaptation they are supposed to address . Connecting otherwise disconnected people through technology may result in \u201c recombinant innovation \u201d in the short run ( Leonardi 2014 ) , as peo - ple are able to bring together ideas already in the net - work . However , as our results suggest , it also causes people \u2019 s solutions to become more similar to each other , which limits adaptability ( including innovative - ness ) after the initial set of fruitful recombinations has been exhausted . Especially when much work is done virtually ( Neeley 2021 ) , organizational leaders may want to ensure that decentralized information \ufb02 ows do not undermine the ability to solve problems requir - ing adaptation . Conclusion : Centralization Improves Performance on Tasks Requiring Collective Adaptation Centralization has previously been found to impede the integration of knowledge and ideas held by indi - viduals in the network periphery . By designing an experimental protocol that manipulates when each piece of information can and cannot be found , we test Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2086 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . the effects of centralization in a different class of prob - lems . We show experimentally that centralization enhances collective adaptability in shifting environ - ments , so long as the central people in the network can both learn from and in \ufb02 uence the periphery , iden - tifying and spreading the good ideas discovered there . Acknowledgments The authors extend gratitude to the four editors and \ufb01 ve anonymous reviewers for investing considerable time to provide developmental , insightful , and enthusiastic com - ments throughout the review process , even during a pan - demic . The authors thank Akshaya Varghese and Cara Mazzucco for their research assistance , the participants at the Organization Science Special Issue Conference for helpful input at a critical stage in this paper \u2019 s develop - ment , and the Workshop on Information Systems and Economics program committee for nominating an earlier version of this manuscript for best paper . The novel Net - LabExperiments . org platform would never have been built without the dedication of the Harvard Business School Research Computing Services ( RCS ) team , particularly Troy Adair , Bob Freeman , and Gabe Mansur . Appendix A . Nodes in Centralized Structures Are Less Subject to Conformity Pressure Based on the assumptions documented in Figure 1 about the relative amount of in \ufb02 uence that people feel as a con - sequence of the number and share of their neighbors hav - ing an alternative opinion as well the assumption that , in each time step , people have a small random chance of changing their opinions , we simulated a random process of social in \ufb02 uence on the network structures we studied here . Our simulation proceeded as follows : Algorithm A . 1 ( Simple Simulation of Social Influence on Networks ) each node starts with opinion A at t (cid:2) 1 ; possible opin - ions are { A , B , C } ; s \u2190 0 : 04 1 + e \u2212 1 : 5 ( numNeighbors \u2212 0 : 9 ) # vector of node susceptibilities by number of neighbors ; r \u2190 0 : 001 # probability of switching to random opinion independent of in \ufb02 uence at each time step , t ; for t \u2190 2 to 500 do for O \u2208 { A , B , C } do in \ufb02 uence of opinion O \u2190 1 1 + e \u2212 10 ( shareOfNeighborsWithOpinion \u2212 : 5 ) ; chance of switching to O because of in \ufb02 uence \u2190 s \u2217 influence of opinion O sum ( influence of all opinions ) (cid:1) (cid:3) ; end determine opinion for time step t through random draw , according to probabilities calculated above ; end With 1 , 000 simulated trials of 500 time steps each , we found that centralized networks retained an initial consen - sus less than other networks . Speci \ufb01 cally , members of a complete clique had opinion A ( the initial consensus ) 92 . 3 % of the time , and members of the locally clustered network had opinion A 79 . 7 % of the time , whereas peripheral nodes in the hub - and - spoke network and core - periphery network had opinion A only 56 . 6 % and 71 . 6 % of the time , respectively . Each peripheral node in a centralized network is exposed to the same in \ufb02 uences from the center of the net - work . However , the in \ufb02 uence is not very strong because ( a ) central nodes are few in number and ( b ) central nodes are likely to be exposed to differing opinions in the periphery and themselves have a higher probability of changing opinions . Appendix B . Additional Details of Subject Pipeline Participants were eligible to advance to stage 2 if they could complete stage 1 by ( a ) following instructions to operate the search and answer functions and ( b ) deducing the correct answer to the stage 1 problem from the six clues that could be found ( details on the logical task are in this appendix ) and entering it into the answer \ufb01 eld . Participants were eligible to advance to stage 3 ( the main trial ) if they could complete stage 2 by ( a ) deducing the correct answer to the stage 2 problem , a slightly longer task completed synchronously with two other partici - pants , and ( b ) answering an attention check question cor - rectly . Participants who failed to pass a stage 1 or stage 2 trial were allowed to try again up to four times before becoming disquali \ufb01 ed from further attempts . The participant pipeline exhibited a steep drop - off at each stage . Approximately 43 % of the stage 1 ( instruc - tional ) trials were passed ; of these , 74 % opted in to a stage 2 trial ( thus the number of people opting in to the stage 2 trial was 43 % \u00d7 74 % \u2248 32 % of those who had opted in to stage 1 trials ) . Approximately 80 % of second - stage trials were passed ; of these , 78 % opted in to the main stage 3 trial . Overall , only 20 % of participants progressed all the way through a stage 3 trial , which we attribute to effective \ufb01 ltering of attentive , interested , and capable participants . We did not collect any data that would allow us to draw conclusions about the representativeness of our experimental sample . Research has shown that MTurk workers are generally representative of the U . S . popula - tion , although somewhat more politically liberal ( Levay et al . 2016 ) and with somewhat higher negative affect and lower social engagement ( McCredie and Morey 2019 ) . That said , they are more representative of the general population than university subject pools ( Berinsky et al . 2012 ) , and results of experiments run on MTurk match results from of \ufb02 ine experiments ( Horton et al . 2011 ) . It is dif \ufb01 cult to say how well our sample represents the broader population of members of organizations or , for that matter , to de \ufb01 nitively identify the attributes of such a population . We would argue , however , that our multi - stage pipeline , which involved solving complex and uncertain problems with others and appearing at a pre - scheduled time ( for the synchronous start of the experi - ment ) , selected participants for the main trial who are likely to be more similar to typical members of organiza - tions than the broader population of MTurk workers . Online labor market workers review the \u201c requesters \u201d posting tasks to help identify reputable employers ( Gray Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2087 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . et al . 2016 ) ; our experiment retained a strongly positive reputation . On Turkerview . com , our experiments had an average reported overall wage of $ 11 . 11 per hour , which includes pay for time in the waiting room and in the three stages of our experiment . Our experiments received the overall rating of \u201c Workers feel this requester pays fairly \u201d as well as the \u201c good communication , \u201d \u201c approves quickly , \u201d \u201c no rejections , \u201d and \u201c no blocks \u201d tags . We used MTurk \u2019 s quali \ufb01 cations functionality ( assigning a worker a custom quali \ufb01 cation upon successfully completing a task to allow him or her to see further tasks in a series ) to qualify and disqualify workers for speci \ufb01 c tasks . We never blocked workers or rejected their work , as these actions can seri - ously harm their ability to qualify for other work on MTurk . In addition to building on the experience of other re - searchers , many details of our subject pipeline were re - \ufb01 ned through a piloting process conducted in tandem with debugging and testing our new experimental plat - form , such that we could improve participants \u2019 under - standing of the task and platform and calibrate the degree to which our task was engaging and paid well enough . The improvements we made during the pilot phase , using feedback from pilot participants , to make the task clear , engaging , and worthwhile helped us ( a ) attract enough participants and ( b ) keep their attention focused on the experiment . Appendix B . 1 . Details of the Stage 1 Trial Stage 1 was a single - person trial and served to teach partici - pants how to use the basic functions in the experimental interface ( speci \ufb01 cally , searching for clues and entering their answers ) and also to ensure they could both perform basic logical deduction and were paying suf \ufb01 cient attention to do so . We provide details on the logical task here . Participants received the following clue as part of the instructions : \u201c The murderer is either Mr . Smith , Mr . Lee , or Ms . White . The murder weapon was poison . \u201d The instructions continue : Other clues must be found by using the search bar to search for keywords . DO IT NOW : Use the search bar to find more information about each suspect . Tip # 1 : searches should be single words , so type \u201c Smith \u201d instead of \u201c Mr . Smith \u201d . Tip # 2 : you can search the same keyword more than once to find more clues . Tip # 3 : some clues contain useful information and some don \u2019 t . The clues that could be found with the search tool were as follows : \u2022 Mr . Smith was out of town at the time of the murder . \u2022 Mr . Smith was very sad to learn about the murder . \u2022 Ms . White was with Mr . Smith at the time of the murder . \u2022 Ms . White grew up in California . \u2022 Mr . Lee \u2019 s \ufb01 ngerprints were found at the scene of the crime . \u2022 Mr . Lee had poison at his house that matched the poison used in the murder . To pass the stage 1 trial , participants had to correctly identify \u201c Mr . Lee \u201d as the culprit . Stage 1 trials were offered to all MTurk workers . Approximately 40 % of these trials were passed to Stage 2 . Appendix B . 2 . Details of the Stage 2 Trial Stage 2 was a synchronous three - person trial , and it served to teach participants how to use the social features of the experimental interface , including sharing clues with their peers with or without free - text annotations , as well as using free - text messages without sharing clues . The logical puzzle was similar in dif \ufb01 culty to that in stage 1 . At the beginning of the trial , each participant received the following message : \u2022 Message from headquarters : Mr . Walker was found strangled in his bed . Additionally , at the beginning of the session , each par - ticipant was given one of the following clues that was not given to their peers ( i . e . , each of the three partici - pants received a different one of the following clues ) : \u2022 Mr . Walker recently had romantic affairs with Mr . Brown \u2019 s wife and Mr . Taylor \u2019 s wife . \u2022 Mr . Walker had recently insulted Mr . Martin and Mr . Taylor in public . \u2022 Mr . Walker had recently \ufb01 red Mr . Clark and Mr . Taylor from his company . The individualized clues each mention Mr . Taylor ( the real culprit ) as well as an additional suspect not men - tioned in their peers \u2019 clues . Thus , they each had informa - tion and the starting point for additional searches that could be shared with their peers , but this still put all experimental participants on even footing with respect to completing the task . Clues that could be found by search were as follows : \u2022 Mr . Brown was out of town on the day of Mr . Walker \u2019 s death because of a last - minute business trip . \u2022 Mr . Martin was in a car accident and was unconscious at the hospital at the time of Mr . Walker \u2019 s death . \u2022 Security camera footage shows that Mr . Clark spent the night in a hotel and did not leave until after Mr . Walker \u2019 s murder . \u2022 Mr . Taylor could not provide an alibi for the time of Mr . Walker \u2019 s death . \u2022 Mr . Taylor \u2019 s friends say that he hasn \u2019 t been himself recently . \u2022 Witnesses saw Mr . Taylor leaving Mr . Walker \u2019 s house after dark on the day of his death . Clues found by search exonerated all suspects except for Mr . Taylor . In addition to introducing the social features , stage 2 tri - als also acted to \ufb01 lter out participants in several ways . First , stage 2 trials required a synchronous start , which we achieved by emailing individuals who had passed the stage 1 trial and announcing start times ( we provided sev - eral start times through the day , announced times a day in advance , and kept the schedule consistent from day to day ) . Only individuals who would appear at previously scheduled start times could move on to the main trial ( which required nine people in the waiting room to begin ) . Second , to pass this stage and be quali \ufb01 ed to take part in the main trial , participants had to not only identify the correct suspect but also correctly answer an attention Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2088 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . check question ( identify the name of the victim , which is given in the \ufb01 rst shared clue , as well as all of the individ - ualized clues at the start of each session ) . To reduce the probability of participants copying and pasting answers posted elsewhere on the internet , we used multiple ver - sions of the clue set , each with different suspect and vic - tim names . Thus , this stage acted as an additional \ufb01 lter on atten - tion , ability to use the platform , ability to solve simple logical puzzles , and ability to join a trial with a synchro - nous start . Appendix B . 3 . Clues and Design of the Main Trial Problem This section lists the clues that were used in the main trial . Note that like the stage 2 trials , we used multiple versions of this problem set to reduce the possibility that participants could learn from any answers posted else - where on the internet . We randomized names and gen - ders of suspects , murder weapons , city names , location names , event names , and times of day ( while maintaining the internal logic of time - based alibis ) . The design of the problem draws heavily on the one used in Shore et al . ( 2015 ) , which also used keyword - based search and logically interrelated clues requiring interpretation and deduction . We designed the three - phase structure of the information set for the present experiment with the intention of modeling a shifting information environment . The main trial was 15 minutes long and consisted of three effective phases , each 5 minutes long . At the beginning of the trial , all participants received the following message : \u2022 Mr . Hall is missing and presumed dead . Your task is to identify the murderer . Mr . Hall was scheduled to be at a political fundraiser , a major project launch , and a confer - ence in the days around his disappearance . The name of the victim ( Mr . Hall ) as well as the names of the three events listed ( political fundraiser , major project launch , and conference ) provide initial keywords for search - ing . In this \ufb01 rst phase , one suspect \u2019 s name ( Mr . Brown ) is associated with two motives and appears to have multiple connections with the events around the victim \u2019 s death . Clues that could be found by search starting at the beginning of the trial were the following : \u2022 The political fundraiser was scheduled in Chicago . \u2022 The major project launch was scheduled in Honolulu . \u2022 The conference was scheduled in London . \u2022 The conference apparently went well . \u2022 Two witnesses reported that Mr . Hall extorted large sums of money from Ms . Moore , Ms . Jones , and Mr . Brown . \u2022 Ms . Taylor told investigators that Mr . Hall cut Mr . Brown , Mr . Davis , and Ms . Taylor out of a lucrative busi - ness deal . \u2022 Ms . Miller reports that Ms . Miller and Mr . Brown were involved in business with The Panoid Group . \u2022 Ms . Miller reports that The Panoid Group was involved in secret negotiations with Mr . Hall . \u2022 Mr . Davis and Ms . Taylor were con \ufb01 rmed to be in Lon - don since before the victim went missing . \u2022 Mr . Davis and Ms . Taylor were traveling together . \u2022 Mr . Brown , Mr . Davis , and Ms . Taylor have been having frequent private meetings recently . \u2022 Anonymous sources say Ms . Taylor was interested in undermining the conference . \u2022 Ms . Miller claims to have been in Chicago since before Mr . Hall went missing . \u2022 Ms . Moore and Ms . Jones were con \ufb01 rmed to be in Hono - lulu around the time of Mr . Hall \u2019 s disappearance . \u2022 Mr . Brown and Mr . Lewis were con \ufb01 rmed to be in Hon - olulu around the time of Mr . Hall \u2019 s disappearance . \u2022 Ms . Miller and Mr . Brown were friends . \u2022 Search of Ms . Miller \u2019 s home revealed that Ms . Miller owned cyanide and a bolo knife . \u2022 Search of Ms . Moore \u2019 s home revealed that Ms . Moore owned a switchblade . \u2022 Search of Ms . Jones \u2019 s home revealed that Ms . Jones owned a bolo knife . \u2022 Search of Mr . Brown \u2019 s home revealed that Mr . Brown owned a bolo knife . \u2022 Search of Mr . Davis \u2019 s home revealed that Mr . Davis owned a bolo knife . \u2022 Search of Mr . Davis \u2019 s home revealed that Mr . Davis owned a 9 - mm handgun . \u2022 Search of Mr . Lewis \u2019 s home revealed that Mr . Lewis owned a 9 - mm handgun . \u2022 Search of Mr . Brown \u2019 s home revealed that Mr . Brown owned a 9 - mm handgun . \u2022 There is no evidence that Ms . Taylor owned any poten - tial murder weapons . \u2022 The victim , Mr . Hall , had many enemies . From the participants \u2019 points of view , other than receiv - ing messages , there was no obvious division of the trial period into phases . At the \ufb01 ve - minute mark , all partici - pants received the following messages : \u2022 Mr . Hall \u2019 s body was discovered at the Harbor Hotel in Honolulu , after the major project launch . \u2022 Security camera footage shows that Mr . Hall went directly to the Harbor Hotel after the major project launch and never left . \u2022 Initial examination indicates that the time of death was 8 pm . After the \ufb01 ve - minute mark , it is possible to \ufb01 nd clues that provide alibis to all suspects ( including Mr . Brown , the leading suspect from phase 1 ) other than the true cul - prit . Three suspects can be ruled out immediately from the shared information that the murder occurred in Hono - lulu ( Davis , Miller , and Taylor were all in other locations ) , and three additional suspects have alibis for the time of the murder ( Moore , Jones , and Brown were at the busi - ness meeting at the time of death ) . Thus , by process of elimination , it is possible to \ufb01 nd the correct answer dur - ing this phase . Clues that could be found by search after the \ufb01 ve - minute mark were the following : \u2022 Ms . Jones , Mr . Brown , and Mr . Lewis were present at the major project launch , according to a witness . \u2022 Ms . Miller was present at the political fundraiser . \u2022 Ms . Taylor did not attend the conference , according to an alibi witness . \u2022 Mr . Davis attended the conference . \u2022 Ms . Moore did not attend the major project launch , according to an alibi witness . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2089 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . \u2022 The Panoid Group \u2019 s business meeting occurred at the Grand Hotel , which is a 35 - minute drive from the Harbor Hotel . \u2022 The Panoid Group \u2019 s business meeting began at 6 and ended around 9 pm . \u2022 Ms . Moore , Ms . Jones , and Mr . Brown attended The Panoid Group \u2019 s business meeting after the major project launch . \u2022 Mr . Brown and Mr . Lewis attended The Supil Company \u2019 s cocktail hour after the major project launch . \u2022 The Supil Company \u2019 s cocktail hour began at 10 and ended around 12 pm . \u2022 The Supil Company \u2019 s cocktail hour occurred at the Gar - den Hotel , which is a 45 - minute drive from the Harbor Hotel . \u2022 Mr . Brown traveled directly between The Panoid Group \u2019 s business meeting and The Supil Company \u2019 s cocktail hour . \u2022 Mr . Brown was seen speaking to a friend at 7 : 50 pm at The Panoid Group \u2019 s business meeting . \u2022 The Panoid Group often hosted VIP parties . \u2022 Mr . Lewis was hoping to land a big contract with The Supil Company . Finally , at the 10 - minute mark , all participants received the following clue : \u2022 Autopsy results show that the murder weapon was a 9 - mm handgun . This \ufb01 nal clue focuses attention on three suspects who were in possession of a matching murder weapon . Two of those three can be logically ruledout , leavingonlythetrue culprit . Appendix B . 4 . Payment In designing our payment scheme , we worked to bal - ance the need to incentivize attentive participation to gather valid data against the desire to pay participants a fair wage for their time . In addition to using the pipeline to \ufb01 lter out inattentive participants , our solution was to have a three - part pay - ment structure . We provided a guaranteed but modest ( $ 0 . 40 ) base payment ; we paid $ 0 . 14 per minute that they had the correct answer to a simple attention check ques - tion registered . Finally , we paid $ 0 . 14 per minute that they had the correct answer to the murder mystery entered . The attention check ( which simply asked the name of the murder victim ) could be answered based on the \ufb01 rst \u201c message from headquarters \u201d clue provided to everyone at the start of the trial , and thus it accounted for the majority of subject payment , regardless of whether they eventually solved the mystery . If participants took 30 seconds to enter the answer to the attention check and did not solve the mystery , they would earn approximately $ 8 . 58 per hour , assuming that they spent 15 minutes on the trial itself plus 2 minutes reading the consent form and waiting for others to be ready to join the trial . In practice , our median hourly wage for participants who completed a trial and provided an answer to the attention check problem was $ 8 . 47 , assuming a total of 17 minutes spent on the task . That said , participants varied , and the 10th and 90th percentile wages were $ 6 . 52 and $ 10 . 41 , respectively . Prior to the trial starting , they received an email stating the base payment and bonus payment rates , and in the trial instructions and consent page , participants were given as accurate of an assessment of payment as we could generate given pilot testing data . The longer you have correct answers entered in the computer , the higher the payment . The base payment is 40 cents , but the majority of the compensation will be bonus payments for correct answers . Participants who are paying attention earn on average approxi - mately $ 8 per hour , and most participants earn between $ 5 and $ 12 . 25 per hour . Following Mason and Suri ( 2012 ) , we used a waiting room to facilitate synchronous start to multiperson trials . Prior to each trial beginning , we assigned more partici - pants than necessary to a trial , with the understanding that some workers either do not consent to a trial or have other reasons for not continuing , such as disinterest or inattention . If more than the required number of partici - pants consented to a trial and clicked the button indicating they are ready to begin , we randomly assigned the required number to begin the actual trial . All participants were paid $ 0 . 40 regardless of whether they consented and indicated they were ready to join ( an estimated $ 8 \u2013 $ 12 per hour equivalent ) . We paid an additional bonus of $ 0 . 40 to those who indicated they were ready to join but were not assigned to an active trial ( $ 16 \u2013 $ 24 per hour equivalent ) to encourage them to reenter the queue and eventually join a full trial . Appendix C . Results on Full Trials Only In the main results , we presented results \ufb01 tted on the full data without regard for whether participants dropped out of any of the trials . Here , we present regression results \ufb01 t - ted to only those data in which no participants dropped out of trials . Table C . 1 shows that results on this subset closely match the full data . Appendix D . Sample , Power , and Models Randomized , controlled experiments on grouped data present additional challenges with respect to sample size , power , and model selection . We document our decision making on these issues here . Appendix D . 1 . Sample Size Because we analyze outcomes at the collective ( nine - per - son network ) level , we had to obtain nine experimental participants for each independent observation . Accord - ingly , our biggest concern was having a suf \ufb01 cient sample size to draw reliable conclusions , and we decided to col - lect the largest balanced sample that we could given the limited amount of Mechanical Turk workers on the plat - form at any one time and the need to conduct synchro - nous experiments with nine workers at a time . Practically speaking , this meant that as we approached our \ufb01 nal sam - ple size , we were able to run fewer trials per day and had to tolerate an increasing rate of trials not starting because of the lack of nine quali \ufb01 ed participants at the trial \u2019 s start time . Our \ufb01 nal sample size was 30 trials per condition ; we collected a 31st trial for the hub - and - spoke and locally Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2090 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . clustered networks ; however , we made the determination that collecting further samples was no longer feasible , and we stopped data collection . Appendix D . 2 . Data Exclusions Results reported in the main text are based on an analy - sis of the \ufb01 rst 30 trials for each condition ( omitting the 31st observation for the hub - and - spoke and locally clus - tered networks ) to obtain a balanced sample . Among these \ufb01 rst 30 observations per condition , all data were analyzed with no exclusions . Appendix C provides a com - plementary analysis of only those trials that did not expe - rience any dropout . Prior to collecting the data analyzed in this paper , we did conduct earlier experimental trials that were too easy to serve as a test of adaptability to a shifting environment ; the vast majority of all participants immediately switched to the correct answer because the key information pre - sented at the beginning of phase 3 essentially gave the answer away without the need for any further thought . Thus , this \ufb01 rst round of trials failed to adequately model the experimental scenario we were intending to create ( the need to move from one solution to the other in the context of an uncertain and shifting information environment ) , and we have excluded data from these trials without analysis . Appendix D . 3 . Power and Model Selection As noted , our approach was to collect as many inde - pendent observations as we could given the availability of Mechanical Turk workers . Thus , our sample size was not determined by an a priori analysis of power . Using G * Power , we obtain post hoc \u201c achieved powers \u201d of 0 . 84 , 0 . 83 , and 0 . 60 for the differences between the hub - and - spoke network and the isolates , locally clustered network , and complete clique , respectively ( assuming Poisson regression with binomially distributed predictors and effective sample size of 60 ) . Similarly , we obtain achieved powers of 0 . 65 , 0 . 64 , and 0 . 37 for the differences Table C . 1 . Number of People Correct at End of Trial : Trials with No Dropout Only Dependent variable 25th percentile Median Mean 75th percentile 25th percentile Median Mean 75th percentile ( 1 ) ( 2 ) ( 3 ) ( 4 ) ( 5 ) ( 6 ) ( 7 ) ( 8 ) Hub - and - spoke \u2212 0 . 019 0 . 038 \u2212 0 . 000 \u2212 0 . 032 ( 0 . 487 ) ( 0 . 340 ) ( 0 . 167 ) ( 0 . 181 ) Core - periphery \u2212 0 . 173 0 . 105 0 . 000 0 . 116 ( 0 . 558 ) ( 0 . 344 ) ( 0 . 167 ) ( 0 . 161 ) Isolates 0 . 077 \u2212 0 . 163 \u2212 0 . 299 * * \u2212 0 . 403 * * * 0 . 096 \u2212 0 . 201 \u2212 0 . 299 * * \u2212 0 . 472 * * * ( 0 . 187 ) ( 0 . 318 ) ( 0 . 146 ) ( 0 . 142 ) ( 0 . 471 ) ( 0 . 156 ) ( 0 . 146 ) ( 0 . 125 ) Locally clustered \u2212 0 . 602 * * \u2212 0 . 295 \u2212 0 . 333 * \u2212 0 . 271 * \u2212 0 . 589 \u2212 0 . 333 \u2212 0 . 333 * \u2212 0 . 340 * * ( 0 . 285 ) ( 0 . 353 ) ( 0 . 181 ) ( 0 . 150 ) ( 0 . 518 ) ( 0 . 217 ) ( 0 . 181 ) ( 0 . 136 ) Complete clique \u2212 1 . 040 * * * \u2212 1 . 120 * * * \u2212 0 . 379 * * \u2212 0 . 018 \u2212 1 . 040 * \u2212 1 . 150 * * * \u2212 0 . 379 * * \u2212 0 . 040 ( 0 . 330 ) ( 0 . 367 ) ( 0 . 185 ) ( 0 . 311 ) ( 0 . 542 ) ( 0 . 240 ) ( 0 . 185 ) ( 0 . 197 ) Out core - periphery \u2212 0 . 495 \u2212 0 . 425 \u2212 0 . 395 * * \u2212 0 . 448 * * * \u2212 0 . 485 \u2212 0 . 433 \u2212 0 . 395 * * \u2212 0 . 520 * * * ( 0 . 318 ) ( 0 . 506 ) ( 0 . 180 ) ( 0 . 145 ) ( 0 . 537 ) ( 0 . 459 ) ( 0 . 180 ) ( 0 . 130 ) In core - periphery 0 . 311 0 . 088 \u2212 0 . 041 \u2212 0 . 204 0 . 323 0 . 056 \u2212 0 . 041 \u2212 0 . 268 * ( 0 . 192 ) ( 0 . 330 ) ( 0 . 171 ) ( 0 . 145 ) ( 0 . 473 ) ( 0 . 179 ) ( 0 . 171 ) ( 0 . 144 ) Constant 0 . 879 * * * 1 . 350 * * * 1 . 520 * * * 1 . 830 * * * 0 . 860 * 1 . 390 * * * 1 . 520 * * * 1 . 900 * * * ( 0 . 158 ) ( 0 . 312 ) ( 0 . 118 ) ( 0 . 126 ) ( 0 . 460 ) ( 0 . 144 ) ( 0 . 118 ) ( 0 . 108 ) Observations 216 216 216 216 216 216 216 216 * p < 0 . 1 ; * * p < 0 . 05 ; * * * p < 0 . 01 . Table D . 1 . Estimated Number of People Correct Under Different Levels of Social In \ufb02 uence Model No social influence Medium social influence Max social influence Ground truth 2 . 986 3 . 029 3 . 005 Binomial 2 . 986 3 . 029 3 . 005 Logit , clustered standard errors 2 . 986 3 . 029 3 . 005 Logit , random effects 2 . 986 2 . 191 0 . 333 Quasibinomial 2 . 986 3 . 029 3 . 005 Quasi - Poisson 2 . 986 3 . 029 3 . 005 Beta binomial 3 . 012 2 . 753 2 . 381 Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2091 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . between the core - periphery and the isolates , locally clus - tered network , and complete clique , respectively ( again assuming Poisson regression with binomially distributed predictors and effective sample size of 60 ) . However , these power estimates are hard to interpret in terms of whether this research will be reproducible because they are simultaneously overly optimistic and overly pessimistic in different ways . They are overly opti - mistic because standard power calculations do not account for the interference within trials and the resulting changes to the distribution of the dependent variable that are caused by different experimental treatments . They are overly pessimistic because when it comes to reproducibil - ity , we really care most not about power or sensitivity ( the ability to detect true statistical differences if they exist ) but speci \ufb01 city ( the ability to avoid concluding that a difference is real when it is not ) . In other words , we care less about detecting true positives than avoiding false pos - itives . To address this and the intertwined issue of model selection , we conduct simulation analysis to evaluate power , speci \ufb01 city , and model bias , tailored to our setting of network experimentation . Our experimental data on the number of people correct at the end of the trial do not conform neatly to the assumptions of standard statistical models . Individual suc - cesses are dichotomous , but they are nested within trials , within which social in \ufb02 uence occurs . Thus , at a trial level , the data could be considered the results of a binomial process ( measured as the number of successes and failures of nine possible people per trial ) but with overdispersion such that the success or failure of different individuals in the same trial is correlated . Alternatively , the trial - level data could be considered a Poisson - like process but with the caveat that the data are bounded above with a maxi - mum number of possible successes equal to nine . A fur - ther complication is that the amount and pattern of social in \ufb02 uence vary systematically ( but in a way unknown to the experimenters a priori ) by treatment condition . Thus , there is no single statistical model for our data that is obviously best , and we turn to a simulation - based approach to evaluate the merits of different approaches . We simulate 30 trials of nine people for each of three syn - thetic conditions intended to mimic different types of social in \ufb02 uence in networks , such as we observe in our experimental data . Each condition has the same global mean number of individuals correct , but the number of correct individuals per trial is drawn from different distri - butions . The \ufb01 rst is akin to an \u201c isolates \u201d condition , in which each individual is independently and identically a Bernoulli random variable ( each trial is a binomial draw ) . The second is akin to the \u201c complete clique \u201d condition , in which all nine people in a single trial have the same answer , such that the trial as a whole is determined by a single draw of a Bernoulli random variable . The third con - dition is akin to a more moderate level of social in \ufb02 uence , and the number of successes per trial is drawn from the beta - binomial distribution ( i . e . , an overdispersed bino - mial ) . For each condition , the grand mean was set such that , on average , three of nine people were correct ( chosen to approximate the average for the isolates condition in the experimental data ) . We create 1 , 000 simulated data sets and \ufb01 t several stat - istical models in order to assess bias , the false - positive rate ( speci \ufb01 city ) , and the false - negative rate ( power / sensi - tivity ) . We considered variations on standard models for dichotomous or count data . Speci \ufb01 cally , we \ufb01 t 1 . a standard binomial regression ( unit of observation is the trial , data are number of successes and failures per trial ) , 2 . a quasibinomial regression , 3 . a beta - binomial regression , 4 . a logistic regression with standard errors clustered by trial ( unit of observation is the individual ) , 5 . a random effects logistic regression with separate ran - dom variances for the two social in \ufb02 uence conditions , and 6 . a quasi - Poisson regression ( unit of observation is the trial ) . Fixed effects models could not be \ufb01 t on these data because of the perfect separation of the dependent varia - ble in the individual trials in the maximum social in \ufb02 u - ence condition . Our primary concerns in choosing a regression frame - work for analyzing these data were obtaining unbiased estimates of the mean number of people correct at the end of the trial and avoiding false - positive \ufb01 ndings . 7 There - fore , we assessed coef \ufb01 cient bias and the false - positive Table D . 2 . False - Positive Rates for Networks with Same True Mean Model \u03b1 (cid:2) 0 : 10 \u03b1 (cid:2) 0 : 05 \u03b1 (cid:2) 0 : 01 Binomial 0 . 255 0 . 198 0 . 118 Logit , clustered standard errors 0 . 068 0 . 036 0 . 009 Logit , random effects 0 . 424 0 . 391 0 . 358 Quasibinomial 0 . 044 0 . 021 0 . 004 Quasi - Poisson 0 . 044 0 . 022 0 . 004 Beta binomial 0 . 112 0 . 076 0 . 026 Table D . 3 . Estimated Number of People Correct Under Different Levels of Social In \ufb02 uence Model No social influence Medium social influence Max social influence Ground truth 3 . 007 4 . 742 3 . 015 Binomial 3 . 007 4 . 742 3 . 015 Logit , clustered standard errors 3 . 007 4 . 742 3 . 015 Logit , random effects 3 . 007 4 . 895 0 . 311 Quasibinomial 3 . 007 4 . 742 3 . 015 Quasi - Poisson 3 . 007 4 . 742 3 . 015 Beta binomial 3 . 031 4 . 499 2 . 372 Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2092 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . rate on data for which we know there is no true differ - ence in means by construction . Table D . 1 gives the coef \ufb01 cients estimated by each model . Most models give an unbiased estimate of the ground truth quantity of number of people correct , but the estimates from random effects logistic regression and the beta - binomial regression were extremely biased . Although these two models are intended to model endog - enous correlation , they performed poorly , and we rejected them for use in analyzing this experiment . The proportion of data sets for which a false - positive \ufb01 nding of statistical signi \ufb01 cance occurred ( for different decision thresholds , \u03b1 ) is in Table D . 2 . The standard bino - mial model does not account for overdispersion from social in \ufb02 uence within trials , and unsurprisingly , it has an unacceptably high false - positive rate ( almost 20 % at \u03b1 (cid:2) 0 : 05 ) . The remaining models \u2014 logistic regression with clustered standard errors , quasibinomial regression , and quasi - Poisson regression \u2014 had more reasonable false - positive rates . Among these three , logistic regression with clustered standard errors produced more false positives ( approximately 60 % more at \u03b1 (cid:2) 0 : 05 ) , so we reject it in favor of one of the remaining two . To assess power , or sensitivity , we also created simu - lated data in which the moderate social in \ufb02 uence condi - tion had a higher true mean , similar to the mean of the hub - and - spoke network in the real results . Tables D . 3 and D . 4 give mean coef \ufb01 cients and false - negative rates when there is a true difference in means . Results of this second simulated data set are similar to the \ufb01 rst one . Of particu - lar note is that the false - negative rate for the quasibino - mial and quasi - Poisson regressions is substantial : around 81 % for \u03b1 (cid:2) 0 : 05 . Thus , we have reason to believe that in the context of network interference , our power is much lower than standard power calculations ( such as can be obtained from software , such as G * Power ) would suggest . Nevertheless , as we explain , this does not mean that our results are not reproducible . With low false - positive rates and high false - negative rates , these regression methods are conservative tests of our hypotheses . We cannot say for certain what the prob - ability is that there is a true result conditional on a \ufb01 nding of statistical signi \ufb01 cance without making assumptions about the underlying distribution of effects in the world . Nevertheless , for the sake of illustration , assume that this present experiment is drawn from an abstract \u201c population \u201d of potential experiments that the same authors might con - duct . This population has a true rate of positive \ufb01 ndings ( we will call it \u03c4 for \u201c truth \u201d ) , which we do not know but can make arbitrary assumptions about . Given that this experiment \ufb01 nds a difference between treatment and con - trols , we would like to know what the probability that our \ufb01 nding is a true positive is , given \u03c4 , as well as our simu - lated rates for false positives and false negatives ( which are themselves calculated from the experimentally obtained effect sizes ) . Plugging in to Bayes \u2019 rule , we see P ( True | StatSig ) (cid:2) ( 1 \u2212 FalseNegativeRate ) ( \u03c4 ) ( ( 1 \u2212 FalseNegativeRate ) ( \u03c4 ) ) + ( ( FalsePositveRate ) ( 1 \u2212 \u03c4 ) ) : Thus , assuming that half of the experiments we run have a true difference between treatment and control \u2014 that is , that \u03c4 (cid:2) 0 : 5 \u2014 a signi \ufb01 cant \ufb01 nding corresponds to a true \ufb01 nding 89 . 4 % of the time . If \u03c4 (cid:2) 1 = 3 , a signi \ufb01 cant \ufb01 nding corresponds to a true \ufb01 nding 80 . 9 % of the time . If we only pick experiments with true effects one - quarter of the time , making \u03c4 (cid:2) 0 : 25 , a signi \ufb01 cant \ufb01 nding still corre - sponds to a true \ufb01 nding 73 . 8 % of the time because of the low false - positive rate and high false - negative rate . Appendix D . 4 . Conclusions : Model Choice and Reproducibility Both quasibinomial and quasi - Poisson regression pro - vided unbiased estimates and conservative hypothesis tests . In the tables of results , we opted for quasi - Poisson regres - sion so that the coef \ufb01 cients would be on the same scale as the models of conditional quantiles for count data . The true value of \u03c4 is unknown and cannot be known , but our hypotheses build directly on the results of prior experiments and thus are consistent with a relatively high \u03c4 . Therefore , given the low false - positive rate and high false - negative rate , we believe that it is likely that our stat - istical \ufb01 nding of difference corresponds to a true difference between centralized and noncentralized networks . Appendix E . Additional Details on the Design of Network Treatments In the abstract , core - periphery networks are de \ufb01 ned as \u201c blockmodels \u201d ( White et al . 1976 ) . In the idealized block - model of core - periphery networks , nodes in the adjacency matrix representation of the network are permuted such that high - degree ( core ) nodes have low node indices and low - degree ( peripheral ) nodes have high node indices ( where an index is simply the order that a node appears in the permutation ) . In a simple blockmodel , all of the entries in a given quadrant of the adjacency matrix ( see Figure E . 1 ) take the same value \u2014 either ones or zeros . In an idealized version of an undirected core - periphery network , all of the entries in quadrants I , II , and III are ones , and all of the entries in quadrant IV are zeroes . Core nodes are connected to each other , and peripheral nodes are connected to core nodes ; however , peripheral nodes are not connected to each other . In the idealized blockmodels of the directed ver - sions of the core - periphery network , either quadrant II or quadrant III is zeroes ( but not both ) ; in the out core - periphery network , quadrant III is all zeroes , and in the in core - periphery network , quadrant II is all zeroes . Table D . 4 . False - Negative Rates for Trials with Different True Means Model \u03b1 (cid:2) 0 : 10 \u03b1 (cid:2) 0 : 05 \u03b1 (cid:2) 0 : 01 Binomial 0 . 521 0 . 558 0 . 619 Logit , clustered standard errors 0 . 675 0 . 717 0 . 796 Logit , random effects 0 . 454 0 . 492 0 . 564 Quasibinomial 0 . 728 0 . 808 0 . 933 Quasi - Poisson 0 . 731 0 . 814 0 . 945 Beta binomial 0 . 735 0 . 784 0 . 874 Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2093 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . In the experimental treatments we used , the networks deviated from the idealized blockmodels by including three links ( ones in the adjacency matrix ) in a quadrant that is otherwise all zeroes . In the out core - periphery network , there are three ones in quadrant III , and in the in core - periphery network , there are three ones in quadrant II . We made this choice for two reasons . The \ufb01 rst reason is that adding those extra ties maintains a degree ( number of ties incident to each node ) of three for all nodes in the network . Speci \ufb01 cally , in the out core - periphery network , each node has three incoming ties , which hold availability of information constant within the network ; in the in core - periphery network , each node has three outgoing ties because this network is the exact transpose of the out core - periphery network . The second reason is that we believe it is more realistic to have some information com - ing from the periphery in the out core - periphery network rather than no information at all . Thus , for external valid - ity reasons , we also prefer including a few incoming ties . In addition to the directed core - periphery network , one additional network treatment has directed ties : the locally clustered network . The locally clustered network was cre - ated by creating three clusters ( complete cliques ) of three participants each and then adding ties between the cliques to create a connected network . We started by adding an undirected ( that is , two - way ) tie between each pair of clusters ; speci \ufb01 cally , we added the ties between node pairs ( 2 , 6 ) , ( 3 , 8 ) , and ( 5 , 9 ) . This left three nodes ( 1 , 4 , and 7 ) with only two incoming and two outgoing ties . Con - necting a single pair of these with an undirected tie would leave one additional node with a lesser degree . Therefore , to complete the network , we added directed ties from node 1 to node 7 , from node 7 to node 4 , and from node 4 to node 1 . Thus , our directed ties allow us to create a network in which the number of incoming ties , the number of outgoing ties , and the eigenvector centrality ( Bonacich 1987 ) are all equal ( on these meas - ures , the network \u2014 like the other controls \u2014 is completely decentralized ) . Endnotes 1 Communication network centralization , which has been measured as a structural characteristic of a communication network , can be the cause or consequence of status - or power - based differentiation and hierarchy ( Lin 1999 ) . Being central in a communication network can confer control and influence ; on the other hand , it can be an individual \u2019 s power that makes him or her central in the communi - cation structure . The term centralization has , therefore , been used to mean centralization of communication patterns , centralization of power , or both . In this experimental study , we focus specifically on centralization of communication patterns , disentangling its effects from the influence of power that derives from rank , status , or deci - sion rights . 2 Code is freely available at https : / / github . com / jessecshore / oceler . 3 This protocol was approved by the University Committee on the Use of Human Subjects under IRB17 - 1054 . 4 If dropout is affected by experimental treatment ( e . g . , if one treat - ment was more difficult and caused more dropout ) , then omitting trials risks biasing estimates of the causal effects of the treatment ( e . g . , if we omitted observations in which participants dropped out because of difficulty , our estimates would be positively biased because we ignored observations in which the treatment caused participants to have trouble ) . 5 Replication data are available at https : / / osf . io / 9d3tv / ? view _ only = ecc8f28edba749a0aa6d8ebc0b96de27 or https : / / doi . org / 10 . 7910 / DVN / X2HQXD . 6 Given the complexities of our analytical framework ( see Appendix D for more discussion ) and the natural interpretation for a decision maker who must decide how to structure a communication net - work , common language effect size is more straightforward than alternatives , such as Cohen \u2019 s d . 7 Note that our intention is to provide a conservative test to build theory . If our purpose was to decide which structure to implement in an organi - zation , we would also need to balance the costs of false negatives . References Afuah A , Tucci CL ( 2012 ) Crowdsourcing as a solution to distant search . Acad . Management Rev . 37 ( 3 ) : 355 \u2013 375 . Almaatouq A , Yin M , Watts DJ ( 2020 ) Collective problem - solving of groups across tasks of varying complexity . Working paper , https : / / doi . org / 10 . 31234 / osf . io / ra9qy . Aranda C , Arellano J , Davila A ( 2017 ) Organizational learning in target setting . Acad . Management J . 60 ( 3 ) : 1189 \u2013 1211 . Argote L , Ingram P ( 2000 ) Knowledge transfer : A basis for competi - tive advantage in \ufb01 rms . Organ . Behav . Human Decision Processes 82 ( 1 ) : 150 \u2013 169 . Argote L , Aven BL , Kush J ( 2018 ) The effects of communication net - works and turnover on transactive memory and group per - formance . Organ . Sci . 29 ( 2 ) : 191 \u2013 206 . Aven B ( 2010 ) Network structures of corrupt innovations : The case of Enron . PhD thesis , Stanford University , Stanford , CA . Bahrami B , Olsen K , Latham PE , Roepstorff A , Rees G , Frith CD ( 2010 ) Optimally interacting minds . Science 329 ( 5995 ) : 1081 \u2013 1085 . Banerjee AV ( 1992 ) A simple model of herd behavior . Quart . J . Econom . 107 ( 3 ) : 797 \u2013 817 . Bavelas A ( 1950 ) Communication patterns in task - oriented groups . J . Acoustical Soc . America 22 ( 6 ) : 725 \u2013 730 . Figure E . 1 . Illustration of the Four Quadrants ( Submatrices ) of a Blockmodel of an Adjacency Matrix Representation of a Core - Periphery Structure node j nod e i 2 4 6 8 2 4 6 8 I : core - > core links II : core - > periphery links IV : periphery - > periphery links III : periphery - > core links 3 1 Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2094 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . Becker J , Brackbill D , Centola D ( 2017 ) Network dynamics of social in \ufb02 uence in the wisdom of crowds . Proc . Natl . Acad . Sci . USA 114 ( 26 ) : 5070 \u2013 5076 . Berinsky AJ , Huber GA , Lenz GA ( 2012 ) Evaluating online labor markets for experimental research : Amazon . com \u2019 s Mechanical Turk . Political Anal . 20 ( 3 ) : 351 \u2013 368 . Bernstein ES ( 2012 ) The transparency paradox : A role for privacy in organizational learning and operational control . Admin . Sci . Quart . 57 ( 2 ) : 181 \u2013 216 . Bernstein ES ( 2022 ) Leveling the \u201c \ufb02 atter \u201d playing \ufb01 eld . J . Organ . Design , https : / / doi . org / 10 . 1007 / s41469 - 022 - 00111 - z . Bernstein E , Shore J , Lazer D ( 2018 ) How intermittent breaks in interaction improve collective intelligence . Proc . Natl . Acad . Sci . USA 115 ( 35 ) : 8734 \u2013 8739 . Bernstein E , Bunch J , Canner N , Lee M ( 2016 ) Beyond the holacracy hype . Harvard Bus . Rev . ( July / August ) , https : / / hbr . org / 2016 / 07 / beyond - the - holacracy - hype . Bigham J ( 2014 ) My mTurk ( half ) workday . Accessed January 19 , 2022 , http : / / www . cs . cmu . edu / ~ jbigham / posts / 2014 / half - workday - as - turker . html . Bikhchandani S , Hirshleifer D , Welch I ( 1992 ) A theory of fads , fash - ion , custom , and cultural change as informational cascades . J . Political Econom . 100 ( 5 ) : 992 \u2013 1026 . Bloom N , Genakos C , Sadun R , Van Reenen J ( 2012 ) Management practices across \ufb01 rms and countries . Acad . Management J . Per - spect . 26 ( 1 ) : 12 \u2013 33 . Bonacich P ( 1987 ) Power and centrality : A family of measures . Amer . J . Sociol . 92 ( 5 ) : 1170 \u2013 1182 . Borgatti SP , Everett MG ( 2000 ) Models of core / periphery structures . Soc . Networks 21 ( 4 ) : 375 \u2013 395 . Boudreau KJ , Lacetera N , Lakhani KR ( 2011 ) Incentives and prob - lem uncertainty in innovation contests : An empirical analysis . Management Sci . 57 ( 5 ) : 843 \u2013 863 . Burt RS ( 2004 ) Structural holes and good ideas . Amer . J . Sociol . 110 ( 2 ) : 349 \u2013 399 . Butts CT , Petrescu - Prahova M , Remy Cross B ( 2007 ) Responder communication networks in the world trade center disaster : Implications for modeling of communication within emergency settings . J . Math . Sociol . 31 ( 2 ) : 121 \u2013 147 . Cattani G , Ferriani S ( 2008 ) A core / periphery perspective on individ - ual creative performance : Social networks and cinematic achieve - ments in the Hollywood \ufb01 lm industry . Organ . Sci . 19 ( 6 ) : 824 \u2013 844 . Centola D ( 2010 ) The spread of behavior in an online social network experiment . Science 329 ( 5996 ) : 1194 \u2013 1197 . Centola D , Macy M ( 2007 ) Complex contagions and the weakness of long ties . Amer . J . Sociol . 113 ( 3 ) : 702 \u2013 734 . Christensen CM ( 2013 ) The Innovator \u2019 s Dilemma : When New Technolo - gies Cause Great Firms to Fail ( Harvard Business Review Press , Cambridge , MA ) . Colizza V , Flammini A , Serrano MA , Vespignani A ( 2006 ) Detecting rich - club ordering in complex networks . Nature Phys . 2 ( 2 ) : 110 \u2013 115 . Connelly BL , Johnson JL , Tihanyi L , Ellstrand AE ( 2011 ) More than adopters : Competing in \ufb02 uences in the interlocking directorate . Organ . Sci . 22 ( 3 ) : 688 \u2013 703 . Cummings JN , Cross R ( 2003 ) Structural properties of work groups and their consequences for performance . Soc . Networks 25 ( 3 ) : 197 \u2013 210 . Dahlander L , Frederiksen L ( 2012 ) The core and cosmopolitans : A relational view of innovation in user communities . Organ . Sci . 23 ( 4 ) : 988 \u2013 1007 . Detert JR , Burris ER ( 2007 ) Leadership behavior and employee voice : Is the door really open ? Acad . Management J . 50 ( 4 ) : 869 \u2013 884 . Detert JR , Burris ER , Harrison DA , Martin SR ( 2013 ) Voice \ufb02 ows to and around leaders : Understanding when units are helped or hurt by employee voice . Admin . Sci . Quart . 58 ( 4 ) : 624 \u2013 668 . Engel D , Woolley AW , Jing LX , Chabris CF , Malone TW ( 2014 ) Reading the mind in the eyes or reading between the lines ? Theory of mind predicts collective intelligence equally well online and face - to - face . PLoS One 9 ( 12 ) : e115212 . Frahm KM , Shepelyansky DL ( 2012 ) Google matrix of Twitter . Eur . Physical J . B 85 ( 10 ) : 355 . Freeman LC ( 1978 ) Centrality in social networks conceptual clari \ufb01 - cation . Soc . Networks 1 ( 3 ) : 215 \u2013 239 . Galinsky AD , Chou EY , Halevy N , Van Kleef GA ( 2012 ) The far - reaching effects of power : At the individual , dyadic , and group levels . Neale MA , Mannix EA , eds . Looking Back , Moving For - ward : A Review of Group and Team - Based Research ( Emerald Group Publishing Limited , London ) , 81 \u2013 113 . Geraci M , Bottai M ( 2014 ) Linear quantile mixed models . Statist . Comput . 24 ( 3 ) : 461 \u2013 479 . Golub B , Jackson MO ( 2010 ) Naive learning in social networks and the wisdom of crowds . Econom . J . Microeconomics 2 ( 1 ) : 112 \u2013 149 . Grant RM ( 1996 ) Toward a knowledge - based theory of the \ufb01 rm . Strategic Management J . 17 ( S2 ) : 109 \u2013 122 . Gray ML , Suri S , Ali SS , Kulkarni D ( 2016 ) The crowd is a collabora - tive network . Proc . 19th ACM Conf . Comput . - Supported Coopera - tive Work Soc . Comput . ( ACM , New York ) , 134 \u2013 147 . Greer LL , de Jong BA , Schouten ME , Dannals JE ( 2018 ) Why and when hierarchy impacts team effectiveness : A meta - analytic integration . J . Appl . Psych . 103 ( 6 ) : 591 \u2013 613 . Gruenfeld DH , Tiedens LZ ( 2010 ) Organizational preferences and their consequences . Fiske ST , Gilbert ST , Lindzey G , eds . Hand - book of Social Psychology ( John Wiley & Sons , Inc . , New York ) , 1252 \u2013 1287 . Gulati R ( 2007 ) Silo busting . Harvard Bus . Rev . 85 ( 5 ) : 98 \u2013 108 . Gulati R , Puranam P ( 2009 ) Renewal through reorganization : The value of inconsistencies between formal and informal organiza - tion . Organ . Sci . 20 ( 2 ) : 422 \u2013 440 . Hackman JR ( 2011 ) Collaborative Intelligence : Using Teams to Solve Hard Problems ( Berrett - Koehler Publishers , San Francisco ) . Hansen MT ( 1999 ) The search - transfer problem : The role of weak ties in sharing knowledge across organization subunits . Admin . Sci . Quart . 44 ( 1 ) : 82 \u2013 111 . Hong L , Page SE ( 2004 ) Groups of diverse problem solvers can out - perform groups of high - ability problem solvers . Proc . Natl . Acad . Sci . USA 101 ( 46 ) : 16385 \u2013 16389 . Horton JJ , Rand DG , Zeckhauser RJ ( 2011 ) The online laboratory : Conducting experiments in a real labor market . Experiment Econom . 14 ( 3 ) : 399 \u2013 425 . Huang S , Cummings JN ( 2011 ) When critical knowledge is most critical : Centralization in knowledge - intensive teams . Small Group Res . 42 ( 6 ) : 669 \u2013 699 . Immelt JR ( 2017 ) How I remade GE and what I learned along the way . Harvard Bus . Rev . 95 ( 5 ) : 42 \u2013 51 . Kane GC , Alavi M ( 2007 ) Information technology and organiza - tional learning : An investigation of exploration and exploitation processes . Organ . Sci . 18 ( 5 ) : 796 \u2013 812 . Lavie D , Stettner U , Tushman ML ( 2010 ) Exploration and ex - ploitation within and across organizations . Acad . Management Ann . 4 ( 1 ) : 109 \u2013 155 . Lazer D , Friedman A ( 2007 ) The network structure of exploration and exploitation . Admin . Sci . Quart . 52 ( 4 ) : 667 \u2013 694 . Lee MY , Edmondson AC ( 2017 ) Self - managing organizations : Exploring the limits of less - hierarchical organizing . Res . Organ . Behav . 37 ( 2017 ) : 35 \u2013 58 . Leonardi PM ( 2014 ) Social media , knowledge sharing , and innova - tion : Toward a theory of communication visibility . Inform . Sys - tems Res . 25 ( 4 ) : 796 \u2013 816 . Levay KE , Freese J , Druckman JN ( 2016 ) The demographic and political composition of Mechanical Turk samples . Sage Open 6 ( 1 ) : 2158244016636433 . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) 2095 D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d . Levine SS , Reypens C , Riedl C ( 2018 ) Should I stay or should I go ? The cognition of exploration and exploitation . Acad . Manage - ment Proc . ( Academy of Management , Briarcliff Manor , NY ) , 10248 . Levine SS , Apfelbaum EP , Bernard M , Bartelt VL , Zajac EJ , Stark D ( 2014 ) Ethnic diversity de \ufb02 ates price bubbles . Proc . Natl . Acad . Sci . USA 111 ( 52 ) : 18524 \u2013 18529 . Levinthal DA ( 1997 ) Adaptation on rugged landscapes . Management Sci . 43 ( 7 ) : 934 \u2013 950 . Lin N ( 1999 ) Social networks and status attainment . Annual Rev . Sociol . 25 ( 1 ) : 467 \u2013 487 . Machado JAF , Silva JS ( 2005 ) Quantiles for counts . J . Amer . Statist . Assoc . 100 ( 472 ) : 1226 \u2013 1237 . Madirolas G , De Polavieja G ( 2014 ) Wisdom of the con \ufb01 dent : Using social interactions to eliminate the bias in wisdom of the crowds . Proc . Collective Intelligence Conf . , 10 \u2013 12 . Majchrzak A , Cherbakov L , Ives B ( 2009 ) Harnessing the power of the crowds with corporate social networking tools : How IBM does it . MIS Quart . Executive 8 ( 2 ) : 103 \u2013 108 . Malone TW ( 2018 ) Superminds : The Surprising Power of People and Com - putersThinkingTogether ( Little , Brown and Company , New York ) . March JG ( 1991 ) Exploration and exploitation in organizational learning . Organ . Sci . 2 ( 1 ) : 71 \u2013 87 . Mason W , Suri S ( 2012 ) Conducting behavioral research on ama - zon \u2019 s mechanical Turk . Behav . Res . Methods 44 ( 1 ) : 1 \u2013 23 . Mason W , Watts DJ ( 2012 ) Collaborative learning in networks . Proc . Natl . Acad . Sci . USA 109 ( 3 ) : 764 \u2013 769 . Mason WA , Jones A , Goldstone RL ( 2008 ) Propagation of innova - tions in networked groups . J . Experiment . Psych . General 137 ( 3 ) : 422 \u2013 433 . McCredie MN , Morey LC ( 2019 ) Who are the Turkers ? A character - ization of mturk workers using the personality assessment inventory . Assessment 26 ( 5 ) : 759 \u2013 766 . Montgomery JM , Nyhan B , Torres M ( 2018 ) How conditioning on posttreatment variables can ruin your experiment and what to do about it . Amer . J . Political Sci . 62 ( 3 ) : 760 \u2013 775 . Morrison EW ( 2011 ) Employee voice behavior : Integration and direc - tions for future research . Acad . Management Ann . 5 ( 1 ) : 373 \u2013 412 . Neeley T ( 2021 ) Remote Work Revolution : Succeeding from Anywhere ( Harvard Business Review Press , Cambridge , MA ) . Nickerson JA , Zenger TR ( 2004 ) A knowledge - based theory of the \ufb01 rm - the problem - solving perspective . Organ . Sci . 15 ( 6 ) : 617 \u2013 632 . Prennushi G , Shaw KL , Ichniowski C ( 1997 ) The effects of human resource management practices on productivity : A study of steel \ufb01 nishing lines . Amer . Econom . Rev . 87 ( 3 ) : 291 \u2013 313 . Puranam P ( 2018 ) The Microstructure of Organizations ( Oxford Uni - versity Press , Cambridge , UK ) . Reagans R , McEvily B ( 2003 ) Network structure and knowledge transfer : The effects of cohesion and range . Admin . Sci . Quart . 48 ( 2 ) : 240 \u2013 267 . Reeves M , Faeste L , Whitaker K , Hassan F ( 2018 ) The truth about corporate transformation . MIT Sloan Management Rev . 59 ( 3 ) : 1 \u2013 7 . Reitzig M ( 2022 ) Get Better at Flatter : A Guide to Shaping and Leading Organizations with Less Hierarchy ( Springer , Cham , Switzerland ) . Schilke O , Hu S , Helfat CE ( 2018 ) Quo vadis , dynamic capabilities ? A content - analytic review of the current state of knowledge and recommendations for future research . Acad . Management Ann . 12 ( 1 ) : 390 \u2013 439 . Schilling MA , Phelps CC ( 2007 ) Inter \ufb01 rm collaboration networks : The impact of large - scale network structure on \ufb01 rm innovation . Management Sci . 53 ( 7 ) : 1113 \u2013 1126 . Shaw ME ( 1954 ) Group structure and the behavior of individuals in small groups . J . Psych . 38 ( 1 ) : 139 \u2013 149 . Shore J , Bernstein E , Lazer D ( 2015 ) Facts and \ufb01 guring : An experi - mental investigation of network structure and performance in information and solution spaces . Organ . Sci . 26 ( 5 ) : 1432 \u2013 1446 . Stasser G , Davis JH ( 1981 ) Group decision making and social in \ufb02 uence : Asocialinteractionsequencemodel . Psych . Rev . 88 ( 6 ) : 523 \u2013 551 . Stasser G , Stewart D ( 1992 ) Discovery of hidden pro \ufb01 les by decision - making groups : Solving a problem vs . making a judg - ment . J . Personality Soc . Psych . 63 ( 3 ) : 426 \u2013 434 . Stasser G , Titus W ( 2003 ) Hidden pro \ufb01 les : A brief history . Psych . Inquiry 14 ( 3 - 4 ) : 304 \u2013 313 . Ter Wal AL , Alexy O , Block J , Sandner PG ( 2016 ) The best of both worlds : The bene \ufb01 ts of open - specialized and closed - diverse syndication networks for new ventures \u2019 success . Admin . Sci . Quart . 61 ( 3 ) : 393 \u2013 432 . Tiedens LZ , Fragale AR ( 2003 ) Power moves : Complementarity in dominant and submissive nonverbal behavior . J . Personality Soc . Psych . 84 ( 3 ) : 558 \u2013 568 . Tortoriello M ( 2015 ) The social underpinnings of absorptive capacity : The moderating effects of structural holes on innova - tion generation based on external knowledge . Strateg . Manage - ment J . 36 ( 4 ) : 586 \u2013 597 . Tushman ML ( 1979 ) Work characteristics and subunit communication structure : A contingency analysis . Admin . Sci . Quart . 24 ( 1 ) : 82 \u2013 98 . Ugander J , Backstrom L , Marlow C , Kleinberg J ( 2012 ) Structural diver - sity in social contagion . Proc . Natl . Acad . Sci . USA 109 ( 16 ) : 5962 \u2013 5966 . Uzzi B ( 1997 ) Social structure and competition in inter \ufb01 rm networks : The paradox of embeddedness . Admin . Sci . Quart . 42 ( 1 ) : 35 \u2013 67 . Valentine MA ( 2018 ) Renegotiating spheres of obligation : The role of hier - archy in organizational learning . Admin . Sci . Quart . 63 ( 3 ) : 570 \u2013 606 . Van der Vegt GS , De Jong SB , Bunderson JS , Molleman E ( 2010 ) Power asymmetry and learning in teams : The moderating role of performance feedback . Organ . Sci . 21 ( 2 ) : 347 \u2013 361 . White HC , Boorman SA , Breiger RL ( 1976 ) Social structure from multiple networks . I . Blockmodels of roles and positions . Amer . J . Sociol . 81 ( 4 ) : 730 \u2013 780 . Woolley AW , Chabris CF , Pentland A , Hashmi N , Malone TW ( 2010 ) Evidence for a collective intelligence factor in the performance of human groups . Science 330 ( 6004 ) : 686 \u2013 688 . Wuchty S , Jones BF , Uzzi B ( 2007 ) The increasing dominance of teams in production of knowledge . Science 316 ( 5827 ) : 1036 \u2013 1039 . Yan B , Jian L , Ren R , Fulk J , Sidnam - Mauch E , Monge P ( 2021 ) The paradox of interaction : Communication network centralization , shared task experience , and the wisdom of crowds in online crowdsourcing communities . Comm . Res . 48 ( 6 ) : 796 \u2013 818 . Ethan S . Bernstein is the Edward W . Conard Associate Professor of Business Administration in the Organizational Behavior Unit at Harvard Business School . He received his doctorate from Harvard University . His research investigates the systems and structures that are being introduced by organizations \u2014 in particular , systems to increase workplace transparency and novel collaboration structures \u2014 to enhance employee effectiveness in an era when the nature of work is changing dramatically . Jesse C . Shore is a research scientist at Meta Platforms , Inc . ; this work was done while he was a professor at Boston University and is unrelated to his work at Meta . He received his doctorate from the Massachusetts Institute of Technology . His primary research focus is collective decision making and how it is shaped by social networks and digital technologies . Alice J . Jang is an assistant professor in the Department of Business Information Technology at Virginia Tech . She received her doctorate from Boston University . Her research focuses on understanding the dynamics of interaction in social networks and the effect of social dynamics on the performance and sustainability of organizations and online communities . Bernstein , Shore , and Jang : Network Centralization and Collective Adaptability 2096 Organization Science , 2023 , vol . 34 , no . 6 , pp . 2064 \u2013 2096 , \u00a9 2022 The Author ( s ) D o w n l o a d e d fr o m i n f o r m s . o r g by [ 129 . 2 . 89 . 115 ] on 03 M a y 2024 , a t 15 : 27 . F o r p e r s on a l u s e on l y , a ll r i gh t s r e s e r v e d .",
    "fuExpertRepresentationDesign2013": "Expert representation of design repository space : A comparison to and validation of algorithmic output Katherine Fu , Department of Mechanical Engineering , Carnegie Mellon University , 5000 Forbes Ave . , Pittsburgh , PA 15217 , USA Joel Chan and Christian Schunn , Department of Psychology , University of Pittsburgh , Pittsburgh , PA , USA Jonathan Cagan , Department of Mechanical Engineering , Carnegie Mellon University , Pittsburgh , PA , USA Kenneth Kotovsky , Department of Psychology , Carnegie Mellon University , Pittsburgh , PA , USA Development of design - by - analogy tools is a promising design innovation research avenue . Previously , a method for computationally structuring patent databases as a basis for an automated design - by - analogy tool was introduced . To demonstrate its strengths and weaknesses , a computationally - generated structure is compared to four expert designers\u2019 mental models of the domain . Results indicate that , compared to experts , the computationally - generated structure is sensible in clustering of patents and organization of clusters . The computationally - generated structure represents a space in which experts can \ufb01nd common ground / consensus e making it promising to be intuitive / accessible to broad cohorts of designers . The computational method o\ufb00ers a resource - e\ufb03cient way of usefully conceptualizing the space that is sensible to expert designers , while maintaining an element of unexpected representation of the space . (cid:1) 2013 Elsevier Ltd . All rights reserved . Keywords : computer supported design , design by analogy , design methods , en - gineering design D esign by analogy , in which designers draw inspiration from cross - domain design solutions , is a promising methodology for design prac - tice ( e . g . , Bhatta & Goel , 1996 ; Davies , Goel , & Nersessian , 2009 ; Goel , 1997 ; Herstatt & Kalogerakis , 2004 ; Vattam , Helms , & Goel , 2010 ) . Not all analogies are equally useful ( Casakin & Goldschmidt , 1999 ; Chan et al . , 2011 ; Dunbar , 1997 ; Fu et al . , in press - b ; Gick & Holyoak , 1980 ; Weis - berg , 2009 ) , raising the question of how to organize the very large space of possible analogies to any given problem . A number of methods exist for orga - nizing spaces of possible design analogies ( e . g . , Gentner et al . , 1997 ; Gentner & Markman , 1997 ; Linsey , Markman , & Wood , 2008 ; Mcadams & Wood , 2000 ; Corresponding author : Katherine Fu Katherine . Fu @ gmail . com www . elsevier . com / locate / destud 0142 - 694X $ - see front matter Design Studies - - ( 2013 ) - - e - - http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 1 (cid:1) 2013 Elsevier Ltd . All rights reserved . Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 Murphy , 2011 ; Verhaegen , D\u2019hondt , Vandevenne , Dewulf , & Du\ufb02ou , 2011 ) that make use of varying computational and representational techniques . For example , the computational work presented in this paper considers a diverse set of possible structural types ( e . g . , trees , rings , chains ) and uses Bayesian methods for choosing the best type and the best instance within each type . Because design - by - analogy inherently has a human in the loop , it is useful to understand how meaningful expert designers \ufb01nd the algorithm - generated or - ganizations of possible inspirational examples . We hypothesize that a tech - nique previously found to capture psychologically meaningful structures in other contexts is especially likely to provide structures that align with the mental models that humans develop through the design process . Such align - ment could aid in both the e\ufb00ectiveness of the tool at stimulating designer\u2019s thinking and in easing the transition for its adoption . The study presented here attempts to validate the proposed computational methodology and its output through judgment of its accuracy and adequacy by four design experts and also a direct comparison of its output to the experts\u2019 structuring of the same patent space , which presumably re\ufb02ects their mental model of the space . 1 Background Analogy and external stimuli in engineering design have been studied in a num - ber of ways . Studies have been performed to understand how the introduction of analogies a\ufb00ects the ideation process and outcomes ( e . g . , ( Christensen & Schunn , 2005 ; Dahl & Moreau , 2002 ; Goldschmidt & Smolkov , 2006 ; Linsey et al . , 2008 ) , with some studies speci\ufb01cally examining how the introduction of analogies with di\ufb00erent levels of applicability to the design problem a\ufb00ects indi - vidual designers ( Christensen & Schunn , 2007 ; Tseng , Moss , Cagan , & Kotovsky , 2008 ) . For brevity , we do not dive deeply into the design by analogy literature , here . Most relevant to the current context , the literature \ufb01nds that the distance of analogies to the current design problem can have a large in\ufb02uence on the helpfulness of the analogy , with some distance being useful ( Chan et al . , 2011 ; Dahl & Moreau , 2002 ; Dyer , Gregersen , & Christensen , 2011 ; Fu et al . , in press - b ; Gentner & Markman , 1997 ; Wilson , Rosen , Nelson , & Yen , 2010 ) , but analogies too far away show no bene\ufb01t ( Fu et al . , in press - b ) . Thus , \ufb01nding a tool that organizes the space of possible analogs by distance to the current problem may be especially useful . 1 . 1 Computational design tools The development of computational tools to assist designers during the design process is an area of considerable research in engineering design . To facilitate functional modeling , Stone and Wood created a functional basis that can serve as a universal language in the ideation process ( Stone & Wood , 2000 ) . This work has been extended and adapted a great deal , including , for example , the creation of a biological functional basis ( Cheong , Shu , Stone , & Wood , 2008 ) . 2 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 The U . S . patent database conveniently catalogs existing technology and engi - neering design and , thus , has been used in the development of numerous compu - tational design aids . For example , TRIZ helps engineers overcome di\ufb03culties in functional reasoning by searching through patents based on heuristic rules ( such as use of opposites ) ( Altshuller & Shapiro , 1956 ) . One axiomatic conceptual design model relies on the combination of TRIZ and functional basis ( Zhang , Cha , & Lu , 2007 ) . Patent citation data has been used to identify the interrelated - ness of technologies , and to demonstrate the bene\ufb01ts of tapping into the technol - ogy knowledge base created by competitors within a design \ufb01eld ( Chakrabarti , Dror , & Nopphdol , 1993 ) . The syntactic similarity between patent claims has been investigated as an aid in patent infringement research ( Indukuri , Ambekar , & Sureka , 2007 ) . Businesses haveusedpatentrepositorytools andpat - ent mining to predict potential market trends , identify proli\ufb01c inventors , and for other purposes . Among the characteristics explored in patent mining have been the number of citations , number of claims , average number of words per claim , and number of classes that the patent spans ( Kasravi & Risov , 2007 ) . The poten - tial for using the Design Repository at Oregon State University as a design aid was demonstrated by Bohm et al . by performing a function - based search using Chi Matrix and Morphological Matrix techniques to \ufb01nd components that were present in concepts generated by hand ( Bohm , Vucovich , & Stone , 2005 ) . Morebroadly , designrepositoriesofCADmodelsofcomponentsandassemblies have been explored as resources for designers to streamline product design of complex engineering systems by reusing , revising or gaining insight into previous designs and models ( Szykman , Sriram , Bochenek , Racz , & Senfaute , 2000 ) . The computational work examined in the current study focuses on structuring design repositories , using more open - ended analogical transfer , and using the textual content of the patents , which may generate richer outcomes . In order to assist in avoiding patent infringement , Mukherjea et al . created a BioMedical Patent Semantic Web that discovered semantic associations be - tween important biological terms within biomedical patents , returning a ranked list of patent resources and a Semantic Web showing the relationships between the terms and between resources . Unlike our work , the Semantic Webs generated by Mukherjea et al . are fully connected graphs with no imposed structure , and the data includes only the abstract of the patents being examined . Further , a Bayesian inference approach was not use to generate the webs ( Mukherjea , Bhuvan , & Kankar , 2005 ) . Bayesian inference was used by Chakrabarti et al . , however , to train a model using a small data set of docu - ments , which was then used to categorize the remaining documents into \u201ctopics\u201d and create a hierarchical structure ( Chakrabarti , Dom , Agrawal , & Raghavan , 1998 ) . Structures other than hierarchies were not explored and the structures created were not used as input for analogical design work . PatViz , a tool developed by Koch et al . allows visual exploration of iterative and complex patent searches and queries using a variety of patent data , Expert representation of design repositoxy space 3 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 including full text . Users may create one graph view with PatViz through a guided process , rather than an algorithm . The tool includes three visualiza - tions of interest : Patent Graph , which creates a fully - connected web of patents ; 3D IPD Treemap , which creates a 3D tree structure of the patents based on a prede\ufb01ned classi\ufb01cation scheme ; and Aggregation Tree , which is another tree view that uses prede\ufb01ned adjustable hierarchies ( Koch , Bosch , Giereth , & Ertl , 2009 ) . Unlike the work of Koch et al . , our work does not rely on prede\ufb01ned or user - de\ufb01ned classi\ufb01cation schemes ; rather , it uses an exploration methodology to discover the best - \ufb01tting structures among multiple types of structures to describe the set of patents . The \ufb01t to the data being examined determines the form of the structure . While the U . S . patent database represents a potentially productive repository of analogical or cross - domain design solutions , its size and complexity make it di\ufb03cult for designers to use e\ufb00ectively . There have been many attempts to automate , aid , or streamline the search of the US patent database , including simple key word searches on the United States Patent and Trademark O\ufb03ce ( USPTO ) website or Google Patents , as well as theories like TRIZ and their resulting tools ( Altshuller & Shapiro , 1956 ; Duran - Novoa , Leon - Rovira , Aguayo - Tellez , & Said , 2011 ; Hernandez , Schmidt , & Okudan , 2012a , 2012b ; Houssin & Coulibaly , 2011 ; Krasnoslobodtsev & Langevin , 2005 ; Liang , Tan , & Ma , 2008 ; Nakagawa , 2012 ; Nix , Sherret , & Stone , 2011 ; Zhang et al . , 2007 ) . It is still di\ufb03cult , however , to recognize and appreciate the char - acteristics that might be relevant to a particular design problem within the \u2018space\u2019 of patents . A number of computational \u201cinnovation support tools\u201d have been developed ( CREAX ; Gold\ufb01re ) , as well as a number of research - driven design support tools , proposals for design support tools , and method - ologies ( Bhatta & Goel , 1996 ; Chiu & Shu , 2005 ; Goel , Bhatta , & Stroulia , 1997 ; Verhaegen et al . , 2011 ; Vincent , Bogatyreva , Bogatyreva , Bowyer , & Pahl , 2006 ; Linsey , Wood & Markman , 2008 ) . Many of these tools and methods , however , rely heavily on the users to generate the terms or analogies and comb through search results . Psychology literature suggests that the prob - ability of retrieving similarities is low when surface dissimilarities exist , because retrieval of far - \ufb01eld analogies is cognitively di\ufb03cult ( Gick & Holyoak , 1980 ) and the range of remindings tend to be limited by surface sim - ilarity ( Forbus , Gentner , & Law , 1994 ) . If designers have a means to extract the interrelatedness and interconnectedness of patents in the space , especially with respect to a speci\ufb01c design problem , they may be able to more strategi - cally choose which cross - domain designs to review or to traverse the space in a more intentional and meaningful exploratory way . 1 . 2 Computational basis The computational methodology for discovering structural forms in patent spaces is based on the work of Kemp and Tenenbaum ( 2008 ) . Kemp and Ten - enbaum devised an algorithm that can discover the underlying structure that is 4 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 inherent to a particular data set , enabling the extraction of psychologically meaningful relationships or information . The algorithm discovers 1 ) the best - \ufb01tting form for a given set of data from a space of 8 possible forms , which include a partition , order , chain , ring , tree , hierarchy , grid , and cylinder ; and 2 ) the best - \ufb01tting instantiation of that selected form , which is called a struc - ture . For example , given human similarity judgments for a set of sixteen colors , the structural form discovered by the algorithm that best described the data was a ring , matching the color wheel with which we are all familiar from our elementary school days . Given data regarding votes of Supreme Court justices , the best structural form discovered was a chain , organizing the justices from liberal to conservative ; again , a meaningful structuring . Given data regarding features that a set of animals have , a tree structure was discovered to be the best , similar in form type to the biological classi\ufb01ca - tion scheme proposed by Linnaeus . The form types explored are based on the psychology literature ( Jaynes , 2003 ) , and have been used in formal models in many contexts ( Anderson , 1991 ; Bradley & Terry , 1952 ; Carroll , 1976 ; Collins & Quillian , 1969 ; Fiske , 1992 ; Guttman , 1944 , 1954 ; Huelsenbeck & Ronquist , 2001 ; Inhelder & Piaget , 1969 ; Kohonen , 1997 ; Shepard , 1980 ; Sneath & Sokal , 1973 ; Wiggins , 1996 ) . In addition , the algorithm is based on Bayesian inference , a statistical basis that has a long history of successfully describing and modeling human cognition ( Gri\ufb03ths , Kemp , & Tenenbaum , 2008 ) . Jaynes describes human plausible reasoning as a calculation of the degree of plausibility of a particular hypothesis being true based on previous experience and common sense , and given the facts at hand , corresponding directly to the components that must be considered when calculating the posterior probabil - ity of a hypothesis being true given a set of data using Bayes Rule ( Jaynes , 2003 ) . This basis in human cognition is an important motivation for the choice to use the Kemp and Tenenbaum algorithm in the computation methodology being validated in this study , as it has the potential to be intuitive in the human mind and thus form the basis for an automated design inspiration tool . 2 The current study The goal of the current study was to gain a better understanding of the output structures of the algorithm through comparison to human judgment . Struc - turing algorithms such as Kemp and Tenenbaum\u2019s ( 2008 ) have customarily em - ployed comparisons to pre - existing \u201ccorrect\u201d structures as a method of validation . In a similar vein , we employed comparisons of the algorithm\u2019s output structures to similar structures generated by expert designers , and also solicited their qualitative reactions to particular aspects of the output structure . This method was judged as appropriate for validation for two reasons . First , to our knowledge , there exists no one \u201cpsychologically meaningful\u201d , \u201ccorrect\u201d , or \u201canalytical\u201d solution to the question of how to best structure design repository data or patents . Second , the \ufb01nal application of this structuring work is in - tended to be the basis for an automated design by analogy stimulation tool , making it crucial that the output is intuitive or sensible to human minds . Expert representation of design repositoxy space 5 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 Speci\ufb01cally , the study was performed to address two primary research questions : 1 . In what ways is the algorithm\u2019s patent structure similar to or di\ufb00erent from how an expert designer might structure a space of patents in prep - aration for inspiration through design by analogy with regard to a speci\ufb01c design problem ? 2 . If the structure is di\ufb00erent , can experts nevertheless make sense of it and perhaps see new connections between patents ? The reasoning chain in Figure 1 represents possible outcomes that feed into three main potential inferences from the study . First , suppose that the algo - rithm structures patents in the exact same way that experts do . In that case , we might infer then that ( 1 ) the algorithm does what an expert would be able to do , but more e\ufb03ciently . This would not be an insigni\ufb01cant outcome , as lim - itations of attention and working memory capacity ( Cowan , 2000 ) , and time and e\ufb00ort e\ufb00ectively preclude the possibility of structuring large databases of potential analogical stimuli by hand . Furthermore , supporting new ways of representing the patent space are not the only way to help inspire innova - tion . Experts\u2019 mental representations are often based on deep structural simi - larity between knowledge structures and domains ( Chi , Feltovich , & Glaser , 1981 ) , and experts are often able to draw deep structural analogies from mem - ory ( Ahmed & Christensen , 2009 ; Ball & Christensen , 2009 ; Christensen & Schunn , 2007 ; Holyoak & Thagard , 1996 ; Novick , 1988 ) . Accurate external - izations of experts\u2019 mental representations could potentially inspire innovation by supporting search for non - obvious analogous solutions to design problems that are hidden by surface dissimilarities ( e . g . , application domain , engineering discipline ) . Further , engineering design is inherently and increasingly interdis - ciplinary ( Reich & Shai , 2012 ) ; even experts with one disciplinary focus may be inspired by representations that include insights from other relevant Figure 1 Reasoning chain of possible outcomes of study and main possible inferences . 6 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 engineering disciplines . The possibility of being able to aggregate expert repre - sentations across domains could be useful in catering to the increasingly inter - disciplinary nature of engineering design . These aggregate representations could give access to solutions that might be hidden by surface dissimilarities , given that when one approaches a domain in which she is not expert , process - ing is easily dominated by surface features ( Chi et al . , 1981 ; Hmelo - Silver & Pfe\ufb00er , 2004 ; Novick , 1988 ) . Secondly , suppose that a ) the algorithm produces a structure that is completely di\ufb00erent from that of the experts , but the experts are nevertheless able to make sense of it , or b ) the algorithm produces a structure that bears both similarities and di\ufb00erences to that of the experts , and the di\ufb00erences are not so large as to render the structure incomprehensible . From both these outcomes , we could infer that ( 2 ) the algorithm may structure the analogical stimuli in a way that potentially allows for novel connections to be made between patents . The algo - rithm\u2019s grouping of patents into nodes based on functional similarity could highlight features and functional principles that would otherwise potentially be overlooked ( e . g . , via analogical comparison ; ( Gentner et al . , 1997 ; Kurtz , Miao , & Gentner , 2001 ) ) yielding fresh insights for creative ideation that could inspire innovation . In the cognitive psychology literature , it has been shown that enabling changes in representation of objects and / or ideas , for example by leading problem solvers to attend to previously ignored features , is an e\ufb00ec - tive way of dealing with \u201cfunctional \ufb01xedness\u201d , where problem solvers have di\ufb03culty seeing a potential creative use of an object with which they are familiar ( Kaplan & Simon , 1990 ; McCa\ufb00rey , 2012 ) . Finally , suppose that a ) the algorithm produces a structure that is completely di\ufb00erent from that of the experts , and it is incomprehensible to them , or b ) the algorithm produces a structure that bears both similarities and di\ufb00erences to that of the experts , and the di\ufb00erences are so large as to render the structure incomprehensible . From both these outcomes , we could infer that ( 3 ) the algo - rithm may need revision in order to be useful for supporting design - by - analogy . These outcomes and inferences comprise the set of potential contributions of the study to the research and development of our automated design - by - analogy support tool . 3 Methods To address our research questions , we asked a set of expert designers to 1 ) structure a set of patents , and 2 ) provide focused feedback on our algorithm\u2019s output structure . Structuring of the patents was conducted in the context of preparing to solve a particular design problem , given our interest in comparing the experts\u2019 judgments of patents\u2019 relevance to the design problem to the algo - rithm\u2019s ordering of patents\u2019 relevance as represented in the layout of the struc - ture . The similarity of our algorithm\u2019s output structure to the experts\u2019 structures was evaluated both quantitatively and qualitatively . In the Expert representation of design repositoxy space 7 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 following sections , we describe the details of our methods of data collection , and also the algorithm we used to produce the structure . 3 . 1 Participants This study was conducted with four design experts in Pittsburgh , PA , USA . All participants were males with at least 10 years of experience in the \ufb01eld of product design , and an educational background in engineering or industrial design . Three participants were employed at local product development \ufb01rms , and one was employed at a local biomedical apparatus design and manufac - ture \ufb01rm . The participants were highly experienced and comfortable in review - ing and researching patents . Each participant was compensated with $ 100 for the 2 h of the study . Though there were only a small number of participants , the information is considered valuable because it is based on expert data , and includes deep analysis of interaction between 45 di\ufb00erent pieces of information ( patents ) for each of the four experts . Further , signi\ufb01cant insights in the study of knowledge representations have often been obtained with only a few partic - ipants whose behaviors are examined in depth ( Chase & Simon , 1973 ; Chi et al . , 1981 ; Chi & Koeske , 1983 ; Gobet & Simon , 2000 ; Huth , Nishimoto , Vu , & Gallant , 2012 ; Morais , Olsson , & Schooler , 2013 ) , and a review of research on design revealed that published studies have often been conducted with qualitative analyses of the performance of a small number of experts ( Mehalik & Schunn , 2007 ) . 3 . 2 Materials The experts were supplied with the following materials : (cid:1) Consent form (cid:1) Design problem description ( see Section 3 . 3 ) , on a 8 . 5 in by 5 . 5 in piece of paper , laminated , with mounting putty on the back (cid:1) Title , abstract text and one key \ufb01gure from thirty - seven randomly selected mechanical patents and eight patents selected from previous studies pre - sented in Chan et al . ( 2011 ) and Fu et al . ( in press - b ) ( see Appendix A ) , each on a 8 . 5 in by 5 . 5 in piece of paper , laminated , with mounting putty on the back (cid:1) A 56 in by 107 in sheet of white craft paper , \ufb01xed to a wall (cid:1) Multiple colored magic markers , one black marker , and one pen (cid:1) Pad of sticky notes (cid:1) An 8 . 5 in by 11 in depiction of the algorithm\u2019s output structure for the same 45 patents (cid:1) Title , abstract text and one key \ufb01gure from patents in 3 di\ufb00erent clusters of the algorithm\u2019s output structure , each on a 8 . 5 in by 5 . 5 in piece of paper . 3 . 3 Design problem The design problem supplied to the experts is one that has been used throughout the development and analysis of this work on the structuring of 8 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 patents for use in design by analogy , as presented in Chan et al . ( 2011 ) and Fu et al . ( in press - b ) . The textual design problem description as supplied to the experts of this study is as follows : Design a device to collect energy from human motion for use in developing and impoverished rural communities in places like India and many African countries . Our goal is to build a low - cost , easy to manufacture device tar - geted at individuals and small households to provide energy to be stored in a rechargeable battery with approximately 80 % e\ufb03ciency . The energy is in - tended to be used by small , low power draw electrical devices , such as a radio or lighting device , hopefully leading to an increase in the quality of life of the communities by increasing productivity , connection to the outside world , etc . The target energy production is 1 kW - hour per day , roughly enough to power eight 25 Watt compact \ufb02orescent light bulbs for 5 hours each per day , or enough to power a CB radio for the entire day . For reference , an average adult human can output about 200 watts with full body physical activity for short periods of time , with a signi\ufb01cant reduction for sustained power output . 3 . 4 Patent selection An initial random subset of 10 , 000 patents was selected from the patent data - base . From this initial set , patents were chosen for the \ufb01nal set if they were classi\ufb01ed within the U . S . Patent classi\ufb01cation system as \u201cBody Treatment And Care , Heating And Cooling , Material Handling And Treatment , Me - chanical Manufacturing , Mechanical Power , Static , and Related Arts\u201d . In addition , for continuity with previous studies and an understanding of human judgment of these patents , eight additional patents were included from two previous studies performed by the authors ( Chan et al . , 2011 ; Fu et al . , in press - b ) . Four of these patents were the \u201cfar\u201d analogical stimuli from the study performed to understand how analogical distance , commonness , and modality a\ufb00ect ideation outcomes presented in Chan et al . ( 2011 ) , and the other four were the \u201cfar\u201d analogical stimuli from a study performed to test the use of structures of patents created using Kemp and Tenenbaum\u2019s discov - ery of structural form algorithm to choose \u201cnear\u201d and \u201cfar\u201d patents and mea - sure their e\ufb00ect on ideation outcomes , presented in Fu et al . ( in press - b ) . Both previous studies used the same design problem used here , described in Section 3 . 3 . In total , 45 patents were provided to the experts as described in Section 3 . 2 . 3 . 5 Algorithmic structure generation The method used to produce the structure of patents shown in Figure 2 has been presented in depth in Fu ( 2012 ) and Fu et al . ( in press - a ) . The computa - tional methodology for generating structures of patents has three main parts . Expert representation of design repositoxy space 9 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 First , the set of 45 full - text patents are preprocessed using Latent Semantic Analysis ( LSA ) , a computational text analysis tool that allows for the extrac - tion of contextual similarity and synonymy between words and documents ( Deerwester , Dumais , Furnas , & Landauer , 1990 ; Foltz , Kintsch , & Landauer , 1998 ; Landauer , Foltz , & Laham , 1998 ) . The output from this \ufb01rst step is a similarity matrix , assigning a numeric value between 0 and 1 , connoting semantic similarity , to all pairwise relationships between patents . This similarity matrix is the input to the second portion of the methodology , which is an algorithm for the discovery of structural form in data , devised by Kemp and Tenenbaum ( 2008 ) . The algorithm is a hierarchical Bayesian algo - rithm that discovers the structure and form type , chosen from a prede\ufb01ned set of form types , including the partition , order , chain , ring , hierarchy , tree , grid , and cylinder , that achieves the highest posterior probability value for the description of the data by a particular form type and structure . Within the al - gorithm , the form of the structure itself changes as the data being examined changes and has the potential of revealing deep structural relationships be - tween patents . Finally , LSA is used again to generate labels for the output structure , which describe the clusters of patents in the best structure , allowing a designer viewing the structure to understand the meaning of the clustering of the pat - ents , and characterize the space for navigation and exploration . Labels were not included in the structures used in this study , as the researchers were inter - ested in the alignment of expert - generated labels and computationally - generated labels . A further post - processing step is done to extract \u201cregions\u201d of similarity in the structure , which further facilitate e\ufb03cient understanding and exploration of the space of patents . These regions were also excluded 12 8 1 40 11 7 19 4 39 33 6 30 23 28 2621 24 34 22 36 41 43 14 9 31 10 44 35 42 13 37 32 27 16 18 5 17 25 38 29 45 15 3 20 2 Figure 2 Algorithm output structure of 45 patents . 10 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 from the computational output used in this study . The main goal of this work is a comparison of the judgment of experts to , and a validation of , the struc - ture ( s ) generated by this methodology . The full text of the 45 patents used in the methodology described above , enumerated in Appendix A , were used to create the grid structure in Figure 2 . 3 . 5 . 1 Cluster label generation The structure shown to the experts ( see Figure 2 ) did not include labels to indi - cate the similarity or meaning of the associations of patents within the clusters . Having the experts suggest labels , both for the clusters within the structure generated by the algorithm , as well as for the clusters within their own patent space or structure , was a way to collect data that could then be compared to labels that were generated computationally . The labels were generated by em - ploying LSA , as described in Fu et al . ( in press - a ) . Two di\ufb00erent methods for label generation were employed , the results of which are presented next . The \ufb01rst method , the Highest Average Rank Labeling method , averaged the ranks of each word in the LSA space for each patent within a cluster , and used the top \ufb01ve highest average ranked words for that cluster . The second method , the Highest Cosine Similarity Labeling method , gathered the top twenty words with the highest cosine similarity to each patent within a cluster , pooled and sorted the words in descending cosine similarity value order , and used the top \ufb01ve words with the highest cosine similarity values in the pool of words for the patents within that cluster . The results yielded by Highest Average Rank Labeling method are shown in Figure 3 . The results yielded by the High - est Cosine Similarity Labeling method are shown in Figure 4 . The results of both methods are discussed for the three nodes in particular in Table 6 as compared to expert generated labels . Because four of the patents included in the space from the previous work presented in Chan et al . ( 2011 ) were not 12 8 1 40 11 7 19 4 39 33 6 30 23 28 2621 24 34 22 36 41 43 14 9 31 10 44 35 42 13 37 32 27 16 18 5 17 25 38 29 45 15 3 20 2 mean , show , operate , present , cause ratchet , pawl , guide , travel , house staple , stick , arm , push , magazine pot , support , plant , rod , port part , rotate , shaft , drive , mechanism blood , pump , needle , rotor , hypodermic valve , open , provide , pressure , connect position , provide , engage , support , extend provide , embodiment , form , describe , position insert , chip , hold , port , cut electrode , prosthesis , shaft , core , key provide , port , form , prefer , position Figure 3 Algorithm output structure of 45 patents with cluster labels , highest average rank method . Expert representation of design repositoxy space 11 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 part - of - speech tagged , many unimportant words were included in the text , and thus the LSA space . These words were manually \ufb01ltered out to increase the quality of the labels . This problem would not arise if all patents in the space have been pre - processed with part - of - speech tagging . 3 . 6 Experimental procedure Each participant was allotted 2 h to complete the study , and the \ufb01ve sessions were run in three di\ufb00erent locations . All locations were similar , in that they were large conference rooms , in which experts had ample light , table space on which to work , seating , and a large wall space for building the structure . 3 . 6 . 1 Experimenter\u2019s instructions and overview The study began with one of the experimenters explaining the motivation of the study , the schedule of tasks , and the speci\ufb01c instructions for completing those tasks . The experts were told that the goal of the study was to understand how an expert , or professional product designer , might organize a collection of patents to facilitate searching for potentially relevant inspiration for design - by - analogy , and how this is similar or di\ufb00erent from the output of an 12 8 1 40 11 7 19 4 39 33 6 30 23 28 2621 24 34 22 36 41 43 14 9 31 10 44 35 42 13 37 32 27 16 18 5 17 25 38 29 45 15 3 20 2 coil , anchor , pallet , escape , arbor sheave , rope , ratchet , lever , latch staple , stick , arm , push , magazine pot , support , plant , rod , port product , toy , sort , screw , cam blood , pump , needle , rotor , hypodermic fuel , valve , liquid , diaphragm , combustion bale , steer , envelope , print , member umbrella , rib , towel , tent , fasten insert , chip , hold , port , cut electrode , prosthesis , shaft , core , key valve , ablate , bottle , probe , damp Figure 4 Algorithm output structure of 45 patents with cluster labels , highest cosine similarity method . Table 1 Correlations between participant and algorithm - generated structures of patents E1 E2 E3 E4 Algorithm E1 0 . 22 a 0 . 21 a 0 . 02 0 . 24 a E2 0 . 22 a 0 . 10 a 0 . 01 0 . 11 a E3 0 . 21 a 0 . 10 a 0 . 00 0 . 15 a E4 0 . 02 0 . 01 0 . 00 0 . 17 a Algorithm 0 . 24 a 0 . 11 a 0 . 15 a 0 . 17 a Note : Values are Phi contingency coe\ufb03cients . a denotes statistical signi\ufb01cance at the . 01 level ( 2 - tailed ) 12 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 algorithm we have that tries to do this automatically for large collections of patents . The experts were then tasked with reading , processing , and placing the design problem and the 45 mechanical patents onto a large sheet of paper on the wall , such that the physical proximity between the patents connoted functional similarity , and physical proximity of patents to the design problem connoted relevance to the solving of that design problem . In addition , it was requested that the experts attempt to build the patent space in a way that fa - cilitates searching through the collection of patents for potential inspiration for a design problem . It was emphasized that the space of patents should cap - ture functional similarity , and not necessarily problem or technological domain similarity . To illustrate this point , experts were given the example of placing a patent for a car suspension system near a patent for a running shoe based on the fact that shock absorbing is a function that is key in both patents . This phase of the procedure took approximately 5 min . 3 . 6 . 2 Participant patent space creation The experts were given the design problem description along with the \ufb01rst 5 patents ( in numerical order , see Appendix A ) . They were told to \ufb01rst read the design problem and think about it for a few min . They were explicitly asked to try not to solve the design problem per se , but to think about the design problem as contextualizing information with which to approach the building of the patent space . Table 2 Logistic regression of algorithm pairwise patent links on number of experts judging pairs as linked B S . E . Wald df p Exp ( B ) 95 % C . I . for Exp ( B ) Lower Upper Constant (cid:3) 2 . 59 . 14 339 . 48 1 0 . 00 . 08 Expert links 0 . 93 . 11 68 . 37 1 0 . 00 2 . 53 2 . 03 3 . 15 Table 3 Algorithm structure nodes evaluated by experts during interview Node # of nodes away from design problem Patent # Patent name 1 0 27 Induction loop vehicle detector 16 Escapement mechanism for pendulum clocks 18 Earthquake isolation \ufb02oor 32 Accelerometer 2 1 5 Interactive toy 17 Single drive , multi - screw sorter with pusher means 25 Hanger tilt mechanism for hanging transportation apparatus 3 3 15 Fuel injection system for linear engines 29 Smoke generating apparatus 2 Self - cleaning pressure relief and bypass valve , dispensing apparatus and method 45 Animal waterer Expert representation of design repositoxy space 13 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 After this , experts were told to read each patent , look over the associated key patent \ufb01gure , and place a sticky note on it , writing down key functions or sub - functions of the patent as described . This functionally focused note was meant to serve as a quick reference for them as they built and iterated on the patent space , as well as to remind them to focus on the functionality of the patents as they are placing them in the space . After these six items were read and pro - cessed , the experts were videotaped placing the patents on the large piece of paper on the wall . They were given markers and told that they could annotate the structure as they progressed , and should feel free to modify the placement of the patents or the content of the sticky notes at any point in the patent space building process . After placing the design problem and \ufb01rst \ufb01ve patents in the space , the experts were given the remaining 40 patents in sets of 10 ( in study index number order as shown in Appendix A ) , with a total of four sets of 10 . Each set of 10 patents Table 4 Mean expert reactions to each of the algorithm nodes Participant Node 1 Node 2 Node 3 E1 1 . 00 1 . 00 1 . 00 E2 1 . 00 0 . 25 1 . 00 E3 1 . 00 0 . 00 1 . 00 E4 1 . 00 1 . 00 0 . 75 Mean \u201cSensibility\u201d 1 . 00 0 . 56 0 . 94 % Outright Yes 100 % 50 % 75 % 1 \u00bc Outright Yes , 0 . 75 \u00bc Yes , but there\u2019s an odd one out , 0 . 25 \u00bc Ok , if you insist , but I really wouldn\u2019t put them together unless I had no choice , and 0 \u00bc Outright No Table 5 Summary of expert reactions to algorithm\u2019s distance orderings Participant Summary of response Representative quotes E1 Makes sense The ordering of it makes sense to me , the order of how far away they are . E2 Overall relative ordering makes sense , although he would like to split node 1 into 2 and make part of it ( 16 and 18 ) a little further out , but still closer than node 2 . But I would actually almost do it like , these two , 32 and 27 , then these two 16 , 18 . And then those [ group 2 ] , your other group , and then those [ group 3 ] E3 Ordering of 1 and 3 makes sense ( doesn\u2019t like 2 , so doesn\u2019t want to talk about it ) Um , so yeah that makes sense , although this one [ group 2 ] , I , like I said . So like if this one [ group 1 ] were , if this were in the upper left hand corner , that\u2019s sort of your origin or your center . You know , I\u2019d probably put this [ group 3 ] , you know , closer to that than I would put this , \u2018 cause I just don\u2019t know what the heck this , this group [ group 2 ] means . E4 Makes sense Ok , I understand why that is 14 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 was read , processed , and placed on the wall before proceeding to the next set of 10 . The placement of the patents on the wall was videotaped for all sets . Once \ufb01nished , the experts were told to take a few min to ensure they were satis\ufb01ed with their placement of patents in the space , and to move them around if neces - sary until the placement was found to be satisfactory . The patents were pro - cessed in the same order across experts to eliminate any e\ufb00ect order of exposure or processing of the patents might have on their creation of the pat - ent space . Overall , this phase of the procedure took approximately 1 h and 15 min . 3 . 6 . 3 Participant explanation of structure On video , the experts were then asked to walk the experimenters through the structure , explaining any groups that emerged , the logic behind the organization of the space and regions of patents within the space , and the signi\ufb01cance of distance of patents from each other or from the design problem . They were asked to write down key words or labels on the pa - per itself describing their regions , or clusters of patents , as they talked through them . In addition , the experimenters asked the experts to draw lines indicating borders , clusters , and \ufb02ows , etc . where appropriate to clarify their patent space and the meaning behind it . This phase of the procedure varied between experts , but took approximately 10 min , on average . 3 . 6 . 4 Participant interview regarding algorithm output structure The experts were then asked to examine the structure that constituted the output of the algorithm ( see Figure 2 ) ; the method used to generate this struc - ture is described in Section 3 . 5 . This structure examination was performed sys - tematically for three di\ufb00erent nodes , and audio and video recordings of these interviews were made . The design problem is in the upper left corner , and indi - cated by the shaded circle . The nodes examined were the upper left corner , Table 6 Summary of expert proposed labels and LSA generated cluster labels Participant Node 1 Node 2 Node 3 E1 Motion dampening related to magnetic \ufb01elds Applying small forces to create larger movements Fluid movement and control E2 Energy transfer , energy capture Complex mechanical systems Hydraulic output control E3 Motion conversion N / A Fluid transfer E4 Oscillatory motion as input Rotation as input Dispensing \ufb02uid ( in a discrete amount ) HighestAverageRank Mean , show , operate , present , cause Part , rotate , shaft , drive , mechanism Valve , open , provide , pressure , connect HighestCosine Coil , anchor , pallet , escape , arbor Product , toy , sort , screw , cam Fuel , valve , liquid , diaphragm , combustion Expert representation of design repositoxy space 15 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 containing patents 32 , 18 , 16 , and 27 ; the left middle , containing patents 5 , 17 , and 25 ; and the bottom middle left , containing patents 29 , 2 , 45 , and 15 . Ex - perts were provided with half sheet descriptions of the patents in each of the three nodes , including the title , abstract , and one key \ufb01gure , and were asked the following questions for each of the three nodes , examining one node at a time : Node Grouping Evaluation Questions ( asked regarding three separate nodes ) The algorithm puts patents in the same node that are closely functionally related . Does it make sense to you that these patents are in the same node ? What key functional relationships do you think make them go together ? Are any of these relationships ones you hadn\u2019t thought of before ? Does the grouping of these patents yield any fresh insights or ideas that you think might be useful for the design problem ? Questions Regarding Reactions to Distance of Clusters from Design Problem The algorithm places nodes with respect to the design problem based on potential relevance . So , the patents in Node 1 ( distance 0 ) are supposed to be pretty similar / relevant to the design problem , followed by the patents in Node 2 ( distance 1 ) and then Node 3 ( distance 3 ) . Does this node place - ment make sense ? Why or why not ? General Reactions Question Do you have any overall impressions about this structure with respect to its usefulness for supporting creative design ideation ? If you were to use a structure like this for design by analogy ideation , would you \ufb01nd it useful to have labels describing the functional similarity of the patents within the clusters ? This phase of the procedure also varied between experts , but on average , the interview phase lasted 15 e 20 min . 4 Results The data collected in this study was very rich , allowing for a number of ways to examine the outcomes . The main analyses presented here include a quantita - tive comparison of the structures generated by all experts to one another and to the algorithm\u2019s output structure , a comparison of expert - suggested cluster labels to LSA - generated cluster labels , and a summary of key patterns in the audio and video data . The patent spaces generated by each expert are included in Appendix B . 16 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 4 . 1 Pairwise associations of patents The patent spaces or \u201cstructures\u201d that were created by the experts were diverse in arrangement , meaning , and format . The clear commonality between the participant structures and the algorithm\u2019s output structure is the clustering or grouping of patents . Thus , to examine the degree to which experts ( and the algorithm ) agreed on patent clustering , we examined correlations between structures at the pairwise - link level . Speci\ufb01cally , for all 990 possible pairwise links between the 45 patents , we noted whether each expert ( and the algorithm ) posited a link or not . Table 1 shows that experts 1 e 3 had statistically signi\ufb01 - cant similarities to each other at the pairwise - linking level ; however , correla - tions are small , in part re\ufb02ecting the correlation analysis\u2019s correction for chance agreement ( which is high in binary - to - binary correlations , and espe - cially high when the ratio of non - links to links is very large ) , and in part demonstrating substantive variation among experts\u2019 representations of the space . Some clarity on the reason for these variations is proposed in Section 4 . 5 . 2 , in which the analyses of video and audio data are presented . Notably , the algorithm correlated signi\ufb01cantly with all experts , including E4 . This is important to highlight , as the lack of correlation between E4 and the other 3 experts suggests that this expert was thinking di\ufb00erently about the inter - relationships between the patents and the design problem , and yet the algo - rithm was still able to incorporate some of that perspective into its structure . From the perspective of developing a tool that maps a space of patents and enables automated search for potentially relevant and inspiring patents , correctly not linking two patents that experts generally agree should not be linked is equally as important as correctly linking two patents that experts agree should be linked . A high rate of spurious links would render the tool practically useless , as any query for potentially inspiring patents within the structure would return a deluge of results that included a large number of irrel - evant patents . Furthermore , as there seems to exist nontrivial variation in terms of expert representations of patent spaces ( as indicated by the statisti - cally signi\ufb01cant but small inter - correlations ) , a useful tool should probabilis - tically track the degree of overlap between expert representations , such that patent links that at least one expert proposed should be proposed by the tool with a low but nonzero probability , and patent pairs that more experts agree are linked should be linked by the tool with a higher probability . To formally test whether our algorithm not only correctly proposed expert - agreed - upon links , but also correctly rejected expert - agreed - upon non - links , and whether the algorithm\u2019s link assignments probabilistically tracked the de - gree of agreement among experts for links , we \ufb01tted a logistic regression model predicting the probability of the algorithm linking any two patents , with expert links as the predictor , where expert links indexed the degree of agreement among experts on a given link , ranging from a minimum of 0 if no expert Expert representation of design repositoxy space 17 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 judged the pair as linked , and a maximum of 4 if every expert judged the pair as linked . The results of this analysis are shown in Table 2 . The overall model was statistically signi\ufb01cant , c 2 ( 1 ) \u00bc 70 . 75 , p < . 000 , and a Hosmer and Leme - show test of deviations from the logistic model turned out nonsigni\ufb01cant , c 2 ( 1 ) \u00bc 0 . 31 , p \u00bc . 58 , indicating a model with adequate \ufb01t . As shown in Figure 5 , for patent pairs for which no experts proposed a link , the algorithm proposed links for less than 10 % of those pairs , the proportion of algorithm - proposed links increased monotonically as a function of the number of expert - links , and the algorithm found 100 % of the links that all four experts agreed upon . This function is quanti\ufb01ed in the results of the logistic model , which estimated that , with each additional expert judging a patent pair as linked , the probability of the algorithm linking the pair more than doubles ( exp ( B ) \u00bc 2 . 53 ; see Table 2 ) . Taken together , the results of the pairwise patent link analysis point to the idea that the algorithm clustering of patents can be judged as \u201ccommon ground\u201d or consensus among experts within this study , both in terms of correctly grouping patents together that experts agree should be grouped , and correctly not grouping patents together that experts would agree should not be grouped . Importantly , the algorithm also appears to be able to incorporate an \u201coutlier\u201d perspective in its grouping of the patents , as evidenced by the algorithm\u2019s link correlations with E4 , despite a lack of signi\ufb01cant correlations between that expert and the other experts . 4 . 2 Expert reactions to algorithm output structure The interview of the experts\u2019 reactions to the algorithm output structure was comprised of two parts : 1 ) evaluating the sensibility of the clustering for each node , and 2 ) evaluating the sensibility of the ordering of node distances Figure 5 Proportion of patent pairs linked in algorithm output structure by number of experts judging pairs as linked . 18 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 from the design problem . The condensed \ufb01ndings for these two interview phases are presented next . 4 . 2 . 1 Expert reactions to sensibility of clustering The expert reactions to the algorithm\u2019s structure were obtained as they exam - ined three separate nodes in detail , at varying distances from the design prob - lem . Table 3 shows the contents of each node , along with their distances from the design problem . Table 4 shows the summary of expert reactions to the clus - tering of patents in the three nodes that were examined . The expert judgments were quanti\ufb01ed by the \ufb01rst two authors on a scale from 0 to 1 , where 1 \u00bc \u201cOutright Yes\u201d , 0 . 75 \u00bc \u201cYes , but there\u2019s an odd one out\u201d , 0 . 25 \u00bc \u201cOk , if you insist , but I really wouldn\u2019t put them together unless I had no choice\u201d , and 0 \u00bc \u201cOutright No\u201d . In addition , the proportion of experts that reported that the clustering \u201cmade sense\u201d or not was calculated , also presented in Table 4 . These results indicate that , overall , the experts were generally satis\ufb01ed with the clustering created by the algorithm output structure of patents . Node 2 was the only cluster of which the participants were not in majority agreement with the algorithm clustering . This may be due to di\ufb00erent conceptualiza - tions of the design problem space , or attending to di\ufb00erent features in the patents themselves when considering their clustering . Further discussion of these possible explanations is presented in Section 4 . 5 . Overall , however , the clustering of patents within the nodes examined was a\ufb03rmed by the ex - perts , and we emphasize that only one expert outright rejected the clustering in Node 2 as invalid . 4 . 2 . 2 Expert reactions to sensibility of node distances The second portion of gathering expert reactions to the algorithm output structure was gauging their feelings about the placement of the three clusters being examined in the structure with respect to one another and to the design problem location , indicated by the red circle in Figure 2 . As summarized in Table 5 , the experts generally found the node distances to make sense . Partic - ipant 3 , as seen in the previous section , did not agree with the clustering of Node 2 , and thus did not want to discuss its location in the structure . Partic - ipant 2 would have liked to see Node 1 further subdivided into two smaller clusters , grouping patents 32 and 27 in one , and 16 and 18 in the other . Overall , the locations of the nodes examined within the structure , both relative to one another and to the design problem were a\ufb03rmed as well . 4 . 3 Exploratory quantitative comparison of distance orderings To further explore the degree to which the algorithm\u2019s distance orderings were valid , we undertook an exploratory quantitative comparison of the distance orderings between the algorithm and the experts\u2019 structures . Unfortunately , 3 of the 4 experts did not appear to have an explicit overall similarity space Expert representation of design repositoxy space 19 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 that organized their structures above the cluster level , and also did not explic - itly represent distance of clusters to the design problem . E1 said he was not sure how any of the speci\ufb01c patents were related to the design problem , and did not talk about the relative relevance of each of his clusters to the design problem ; E3 did include a \u201cnot interesting\u201d category of patents ( i . e . , patents unlikely to be relevant to the design problem ) , but the overall spatial ordering of his structure was governed more in terms of a functional \ufb02ow structure than by relevance to the design problem ; E4 explicitly opposed any notion of ordering of relevance , contending that all clusters were potentially relevant . Only E2\u2019s structure bore evidence of spatial ordering of relevance to the design problem , both in the structure itself , and in the verbal content of his explana - tion of the structure . It is unclear to us why only E2 was willing to spatially order the patents by relevance e perhaps further studies with more experts might uncover whether this is due to timing ( e . g . , a general reluctance of expert designers to prune out stimuli as irrelevant in early stages of ideation ) , scale ( e . g . , greater willingness to order by relevance with more stimuli ) , or some other reason . Nevertheless , we thought it would be informative to examine the correlation between the distances of patents from the design problem in E2\u2019s structure as compared to the algorithm\u2019s structure , while bearing in mind that inferences made from this exploratory analyses should be made with care . To measure distances , the following procedure was used . First , an image of E2\u2019s structures was printed out on letter sized ( 8 . 5 (cid:4) 11 in ) paper . Next , dot markers were drawn by one of the authors at the centroid of each patent clus - ter . Finally , a standard metal ruler was used to measure the distance of each dot from the middle of the design problem . Measurements were in cm , in in - crements of 0 . 1 cm . Each patent in the cluster received the distance value of the centroid dot from the design problem . For example , if patents 1 and 4 were in cluster 1 , and the distance of the centroid of cluster 1 from the design problem was 6 cm , then patents 1 and 4 were regarded as being 6 cm from the design problem . In the algorithm\u2019s structure , each patent\u2019s distance from the design problem was the number of nodes ( shortest path ) from the patent\u2019s node location to the design problem . There is a pragmatic and a theoretical motivation for assigning distances at the cluster level rather than the patent level . Pragmatically , we had reason to believe that the experts did not pay much attention to distances at the patent level , due to constraints ( either perceived or real ) of the structuring procedure . For example , our instructions asked the experts to ( if possible ) avoid covering the titles / numbers of patents , which forces a minimum distance on the patents inside a cluster . Theoretically , this cluster - level way of measuring distance is closer to how the algorithm does so ( distance from design problem de\ufb01ned at the node level ) . 20 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 In this exploratory analysis , a statistically signi\ufb01cant and fairly sizable corre - lation was found between the distances of E2 and the algorithm ( Pearson r \u00bc 0 . 55 , p < 0 . 01 ) ( Figure 6 ) . This \ufb01nding provides a useful complement to the interview data , given the rela - tively large e\ufb00ect size and explicit examination of the agreement in distance or - derings for all 45 patents ( as opposed to the 11 in the qualitative reactions in the previous section ) , and lends strength to our conclusion that the algorithm achieves some success in organizing patents in the space in a way that is mean - ingful and sensible to experts . 4 . 4 Cluster label analysis The experts were also asked to propose functional labels for each node . Table 6 shows the labels they proposed , in addition to the labels generated by the two cluster label generation methods presented in Fu et al . ( in press - a ) , Highest Average Rank and Highest Cosine Similarity . In addition , the full cluster label set for both labeling methods are presented on the algorithm output structure in Figures 2 and 3 . On the surface , there seems to be a fair bit of overlap of the text within the labels suggested by the experts for Node 3 , to a slightly lesser extent with Node 1 , and much less with Node 2 . The experts tend to agree that Node 3 is a cluster of patents related to controlling or moving \ufb02uids . The Highest Average Rank method for cluster label generation yielded labels that indicate that \ufb02uid is the topic area common to the patents within Node 3 , but also the functionalities or components that are employed in the patents , like \u201cconnect\u201d and \u201cvalve . \u201d The Highest Cosine Similarity method of gener - ating cluster labels yielded words that are on topic with \ufb02uid applications , but much more speci\ufb01c in terms of actual processes or components , such as \u201ccombustion\u201d and \u201cdiaphragm . \u201d Experts suggested labels for Node 1 that mostly are related to motion , with one expert suggesting more energy - related labels . Similarly , the Highest Figure 6 Scatterplot of patent distances in E2\u2019s vs . algo - rithm\u2019s structures . Expert representation of design repositoxy space 21 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 Average Rank labels for Node 1 seem to be general functional terms , not spe - ci\ufb01c to motion or energy capture ; and the Highest Cosine Similarity labels , like those for Node 3 , are topical to the speci\ufb01c content of the patents within the node e though they reference mechanisms that would be useful when trying to capture energy from motion , like \u201cescape\u201d ( as in escapement ) and \u201ccoil\u201d ( as in induction ) . The expert - suggested labels for Node 2 are diverse in the speci\ufb01c words cho - sen , though they all relate to mechanical systems ( note , node 2 labels from participant 2 were erroneously not collected ) . Correspondingly , the Highest Average Rank labels make clear this common thread , with function and component labels that tie the expert labels together , like \u201cdrive\u201d and \u201cmecha - nism\u201d and \u201cshaft . \u201d The Highest Cosine Similarity labels again contain words that point more toward the speci\ufb01c patents in the cluster , like \u201ctoy . \u201d In general , as exhibited by the detailed node analysis and as can be observed in Figures 2 and 3 , the Highest Average Rank method tends to yield labels that aremoregeneral , indicatingthecommonfunctionalityorcomponentsexplaining the association of the patents within the cluster . In many cases , however , and particularly with nodes that have only one patent in them , the two methods of cluster label generation yield the same results . The Highest Cosine Similarity method , on the other hand , tends to yield labels that are very speci\ufb01c to the con - tent of the individual patents within the cluster , although sometimes containing words that may only pertain to one or two patents in the cluster . It should be noted that the patent text was not part - of - speech tagged for patents 16 , 18 , 27 , and 32 , as these were hand picked patents from the previous design by analogy study presented in Chan et al . ( 2011 ) , which may be a reason for the less descrip - tive label output from the Highest Average Rank labeling method for Node 1 . 4 . 5 Explaining inter - expert and expert - algorithm di\ufb00erences 4 . 5 . 1 Similarities in patent function labels Why did expert structures not completely agree with one another ? Did they agree what the categories of clusters should be but made somewhat di\ufb00erent decisions about which patents below inside each cluster ? An analysis of the overlap among experts between function labels given to each patent can pro - vide some insights . Because of synonyms and near synonym relationships among words , expert e expert similarity between labels should be determined along a semantic overlap dimension rather than exact match dimension . The cosines between words in the LSA vector word representation used by the computational algorithm can serve as an approximation of semantic overlap between words ( Landaurer & Dumais , 1997 ) . Figure 7 plots the mean cosines across patents for three comparisons : ( 1 ) be - tween experts\u2019 function labels for each patent ( 1st column ) , ( 2 ) between 22 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 experts\u2019 function labels and abstracts for each patent ( 2nd column ) , and ( 3 ) be - tween experts\u2019 labels and full - text for each patent . The mean cosine between abstract and full - text across patents is used as a baseline measure against which to compare the mean cosines . This choice of baseline re\ufb02ects our assumption that , in situations where a patent is transformed into a di\ufb00erent , potentially sparser representation ( e . g . , moving between di\ufb00erent experts\u2019 rep - resentations of a patent , summarizing patent in an abstract ) , the transforma - tion from patent full - text to abstract would be a sensible \u201cgold standard\u201d with respect to degree of similarity between transformations , to the extent that pat - ent writers attempt to create a reasonably comprehensive and accurate sum - mary of the patent in the abstract . The \ufb01rst comparison column ( vs . other ) gives the mean cosine of each expert against the other experts , averaged across experts for each patent , and then averaged across patents . If the experts in general were attending to similar as - pects of each patent , one would expect their mean cosine with each other for each patent to at least approximate the mean cosine between the abstract and full - text for that patent . The \ufb01gure clearly shows that this is manifestly not the case : on the contrary , the cosine values clearly indicate that the average simi - larity between experts is dramatically lower than the average abstract e full - text similarity . The second comparison column ( vs . abstract ) gives the mean cosine of each expert\u2019s functional representation of each patent with the patent\u2019s abstract , averaged across patents . The \ufb01gure clearly shows a range of similarities with the abstract , suggesting divergent approaches to representing each patent , perhaps with some experts attending to functional aspects of each patent in a more holistic , high - level manner ( E1 and E2 ) , and others attending to Figure 7 Mean semantic cosine similarity of expert functional patent labels compared to other experts , patent abstracts , and patent full text . Expert representation of design repositoxy space 23 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 much more focused , detailed aspects of each patent ( E3 and E4 ) . The third comparison column yields a similar insight . Overall , the authors interpret these data to suggest that the divergence between experts in the overall structures stems at least in part from their attending to di\ufb00erent aspects of each patent . In support of this interpretation , the inter - expert cosine for each patent was signi\ufb01cantly positively correlated with both average expert vs . abstract ( r \u00bc 0 . 49 , p < 0 . 01 ) and expert vs . full - text co - sines ( r \u00bc 0 . 35 , p < 0 . 05 ) , suggesting that the degree to which , on average , the experts were attending to the patents in a holistic manner ( i . e . , in accord with the full - text or abstract - level summary of the patent , rather than focusing on speci\ufb01c aspects of the patent apart from its core functionality ) was a key deter - minant of the average similarity between experts . Considering the average complexity of patents , this behavior of the experts is not unexpected . 4 . 5 . 2 Protocol video and audio interview analysis It is clear not only from qualitative analysis of the structures , but also from the correlations shown in Table 1 , that the structures created by the experts are diverse . The video and audio explanation of the structures both during the cre - ation and during the \u201cwalk through\u201d shed light on the reasons for this vari - ability . Some representative quotes from the interviews with the experts at di\ufb00erent stages in the experiment are listed as follows . The common thread among these expert quotes is that they all o\ufb00er some indication that the framing of the space prior to working with the patents , or contextualizing the space with the design problem itself , leads to the experts building the space , categorizing the patents , and attending to the content within the patents in di\ufb00erent and highly in\ufb02uenced ways . The quotes have been edited to remove speech dis\ufb02uencies . \u201cWell I guess in general I thought about people , people moving around , and people moving things around and those are things that , those are activities that , that happen , particularly , in the third world there\u2019s a lot of mechanic - , use of the body for labor . So , so when I thought about that , I thought about well how do you attach things to the body , and I thought about that . Let\u2019s see . And then also how do you attach things , how do you attach a cart , or connect a cart , and so you might be carrying water or something . And you might as well carry it on a cart , and it has wheels on it , and maybe those wheels spin , and there\u2019s a small amount of generation of electricity from that , that rotation of the wheels . And then , you\u2019re pushing a cart or you\u2019re pulling a cart . So these are all related to the collecting power from a cart\u201d ( E3 , while explaining structure ) \u201cIt depends what path you were going down . If you said I\u2019m doing something mechanical and you\u2019re describing the escapement mechanism , then that\u2019s \ufb01ne . But otherwise I\u2019m not sure . Again , I would go more with bicycle and 24 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 gears or bicycle and rotary motion . \u201d ( E2 , while evaluating node distances in algorithm output structure ) In addition , there was evidence that the experts experienced signi\ufb01cant cogni - tive load when performing the task , indicating that the automated generation of structures would be bene\ufb01cial in terms of removing the expense of energy and time from the designer during patent exploration and organization . Some representative quotes are below . \u201cSome of it , I think , some of it is wherever your starting \ufb01lter is , how you\u2019re going to sort these things , and what my groupings are . You\u2019ve already , once you start to choose those , you\u2019ve headed down a path , so if you forced a , had some kind of forced starting grouping methodology or some kind of grouping \ufb01lter\u201d ( E1 , while evaluating clustering in algorithm output structure ) \u201cIf I could see all the groups , maybe I\u2019d put it in a di\ufb00erent , di\ufb00erent kind of sort . \u201d ( E4 , while evaluating node distances in algorithm output structure ) 5 General discussion The research questions underlying this work were the following : 1 . How is the algorithm\u2019s patent structure similar to or di\ufb00erent from how an expert designer might structure a space of patents in preparation to be explored for inspiration through design by analogy with regard to a spe - ci\ufb01c design problem ? 2 . If the structure is di\ufb00erent , can experts nevertheless make sense of it and perhaps see new connections between patents ? Our central \ufb01nding was that the algorithm output structure of patents was a good representation of the consensus of experts ( to the extent that such consensus exists ) regarding the associations and clustering of patents in the space . Speci\ufb01cally , the pairwise links generated by the algorithm correlate signi\ufb01cantly with all experts\u2019 links , even with E4 , whose links did not correlate with the other experts , thereby capturing a diversity of expertise , which is important in the increasingly diverse and interdisciplinary nature of engineer - ing design ( Reich & Shai , 2012 ) . Additionally , our logistic regression analysis demonstrated that the algorithm was able to probabilistically track the degree of expert consensus on patent pairwise links , pairing patents experts agreed should be paired , not pairing patents experts agreed should not be paired , and assigning links with nonzero probabilities as a function of degree of expert agreement in areas of potential disagreement . From the verbal accounts given by the experts , both during their working with the patents to create their own patent space , and during the interview Expert representation of design repositoxy space 25 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 regarding the algorithm - generated structure , some additional key \ufb01ndings emerged . First , the general consensus with regard to the algorithm output structure clustering is that it makes sense , and in some cases , was both sensible and surprising e indicating that the output from the algorithm is intuitive to expert designers , and can potentially even help them conceptualize the space in new and useful ways , indicated by some interview data . Second , the node dis - tances with respect to the design problem and the interrelationships between the clusters examined were found to make sense to the experts . The explor - atory quantitative comparison of distance orderings in Section 4 . 3 converges with this \ufb01nding , showing a signi\ufb01cant correlation between the node distances created by E2 and the node distances generated by the algorithm . This strong overlap is of particular interest , as it indicates that the organization of the overall space itself can be meaningful , not just the clustering of patents e moti - vating the use of structures for exploring patents as opposed to simple catego - rization or grouping methods . Thus , the answer to the research questions posed is that the algorithm did pro - duce similar and di\ufb00erent patent spaces when compared to expert thinking , but in bene\ufb01cial ways for both outcomes . The algorithm - generated patent space was similar to expert thinking in that it was strongly correlated with their com - mon ground , and found by the experts to be logical and sensible in clustering , node distances , and node labels . The algorithm - generated patent space was di\ufb00erent from expert thinking in that the clustering of patents by functional similarity was not highly correlated with any individual expert , and found to be surprising at times ( based on interview data ) , while still being sensible to the experts . Most importantly , though , the experts did \ufb01nd the results meaningful . Returning to the potential outcomes outlined in Figure 1 , we interpret the re - sults of this work to be a combination of outcome 1 , the algorithm does what an expert would be able to do ( in this particular case , while further studies using larger expert pools are needed ) , but more e\ufb03ciently , and outcome 2 , the algo - rithm may structure the analogical stimuli in a way that potentially allows for novel connections to be made between patents . The algorithm - generated patent space was similar to the consensus of the experts\u2019 understanding and represen - tation of the patent space , making it much more e\ufb03cient than experts in creating a meaningful patent space . However , the algorithm - generated patent space was also meaningfully di\ufb00erent from each individual expert\u2019s represen - tation of the patent space , and found at times to be surprising to individual experts while still being sensible . This suggests that designers who traverse the algorithm\u2019s structures may be able to discover novel connections between patents . From a practical standpoint , the algorithm output structure o\ufb00ers a more objective conceptualization of the space , while saving the designer time 26 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 and energy . The experts reported that the task of organizing the patents into the space was one in which it was di\ufb03cult not to get locked into a represen - tation early on , and keeping track of all of the information and aspects was challenging . This indicates that the automated patent structuring achieved by the methodology presented in Fu et al . ( in press - a ) could provide signif - icant advantages to designers by presenting them with useful , sensible , and even sometimes surprising representations of patent spaces while saving them signi\ufb01cant mental e\ufb00ort and time . It took the algorithm less than 45 min to optimally generate 8 di\ufb00erent structures , running on a Windows 64 - bit dual core desktop computer , while it took the experts approximately 1 . 25 h to generate one representation of the patent space . As the number of patents grows , the time for the human would increase likely in a non - linear way due to fundamental limitations with human memory ( Cowan , 2000 ) . Perhaps more importantly , it is unlikely that a designer would take the time to structure large numbers of patents by hand , even though this work has shown it potentially valuable . The algorithmic structuring of pat - ents could be viewed as operating like another expert , providing a view that overlaps with that of other experts , while being somewhat di\ufb00erent as well . 6 Conclusion From an inspirational standpoint , the algorithm output structure o\ufb00ers a conceptualization of the space that is sensible to four expert designers , while still maintaining an element of surprise and unexpected representation of the space , making it a promising basis for a computational design by analogy inspiration tool . By allowing for more e\ufb03cient and insightful access to external analogical stimuli , designers have the potential to create more inno - vative design solutions . With a way to extract the interrelatedness and inter - connectedness of patents in the space , designers might be able to strategically choose which cross - domain designs to expose themselves to , or even traverse the space in a more intentional and meaningful exploratory way . Addition - ally , by making novel connections between patents , designers might be able to discover novel solutions via altered representations of the problem or design space . Explicit tests of such e\ufb00ects should be explored in future work . Acknowledgments The authors would like to thank Dr . Kristin Wood for his comments and in - sights on this work . This work was supported by the National Science Foun - dation , under grant CMMI0855326 . Expert representation of design repositoxy space 27 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 Appendix A 45 P atents used in algorithm generated structure Study index number US patent number Patent title 1 6 , 082 , 923 Converging Sphere Joint Assembly b 2 5 , 984 , 148 Self - Cleaning Pressure Relief and Bypass Valve , Dispensing Apparatus and Method 3 5 , 375 , 948 Cutting Insert for Cutting and Grooving Tools 4 6 , 367 , 521 Gravity Feed Fluid Dispensing Valve 5 6 , 497 , 607 Interactive Toy 6 4 , 853 , 977 Patient Garment 7 5 , 993 , 410 Adjustable Probe 8 4 , 223 , 996 Apparatus for Mixing Solid and Liquid Constituents of Mortar or The Like 9 4 , 589 , 668 Wheeled Cart with Removable Skis 10 3 , 962 , 735 Movable Bulkhead with Guiding and Overcanting Prevention Means 11 4 , 124 , 051 Shock Absorbing Wheel Hub 12 4 , 984 , 583 Air Bubbling Mats for Therapeutically Agitating Bath Water 13 7 , 175 , 212 Latch Having Releasable Cinching Mechanism 14 4 , 259 , 034 Bale Handling Apparatus 15 6 , 634 , 325 Fuel Injection System for Linear Engines 16 4 , 139 , 981 Escapement mechanism for pendulum clocks a 17 5 , 909 , 815 Single Drive , Multi - Screw Sorter with Pusher Means 18 4 , 402 , 483 Earthquake isolation \ufb02oor a 19 4 , 103 , 708 Ventilated Poppet Damper 20 3 , 964 , 473 Bone Prosthesis 21 4 , 705 , 064 Safety Seal for an Operating Lever 22 6 , 142 , 689 Envelope Leveler for Printer Feeder 23 5 , 273 , 173 Screw Top 24 5 , 438 , 724 Method for Using Plastic Fasteners for Shoe - Lasting Applications 25 5 , 234 , 096 Hanger Tilt Mechanism For Hanging Transportation Apparatus b 26 4 , 867 , 134 Fluid - Heating Solar Collector 27 4 , 568 , 937 Induction loop vehicle detector a 28 6 , 109 , 282 Self - Erecting Loop Structure 29 4 , 303 , 397 Smoke Generating Apparatus 30 5 , 899 , 571 Beach Towel , Tote Bag and Beach Umbrella System 31 6 , 234 , 452 Hand Operable Motorcycle Stand 32 4 , 335 , 611 Accelerometer a 33 5 , 085 , 240 Shelter Structure b 34 6 , 634 , 044 Compact Stretcher 35 4 , 270 , 310 Support Device for an Upstanding Plant Support Rod in a Plant Pot 36 5 , 423 , 097 Emergency Drop Fowler and Gatch 37 5 , 277 , 276 Compensating rope sheave tie down b 38 3 , 938 , 909 Single Needle Alternating Flow Blood Pump System 39 5 , 647 , 066 Safety Helmet Visor 40 6 , 119 , 041 Apparatus and Method for Linear Lesion Ablation 41 4 , 484 , 762 Ski Binding and Boot 42 4 , 762 , 262 Side - Fed Stapler 43 6 , 164 , 698 Steering Device for Automobiles 44 6 , 062 , 856 Dental Implant Hole Guide Extension 45 4 , 739 , 727 Animal Waterer a Far Analogical Stimuli from Chan et al . ( 2011 ) . b Far Analogical Stimuli from Fu et al . ( 2012 ) . 28 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 Appendix B Expert representation of design repositoxy space 29 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 References Ahmed , S . , & Christensen , B . T . ( 2009 ) . An in situ study of analogical reasoning in novice and experienced design engineers . Journal of Mechanical Design , 131 ( 11 ) , 111004 . Altshuller , G . S . , & Shapiro , R . B . ( 1956 ) . O P c j xo m o [ jj ji o b pe t a t e m : c l o [ o tc op y ec tc a . [ On the psychology of inventive creation ] . Bonpoc 9 P cuxo m o \u01a8 uu ( The Psychological Issues ) , 6 , 37 e 39 , ( in Russian ) . Anderson , J . R . ( 1991 ) . The adaptive nature of human categorization . Psycholog - ical Review , 98 ( 3 ) , 409 e 429 . Ball , L . J . , & Christensen , B . T . ( 2009 ) . Analogical reasoning and mental simula - tion in design : two strategies linked to uncertainty resolution . Design Studies , 30 ( 2 ) , 169 e 186 . Bhatta , S . , & Goel , A . ( 1996 ) . From design experiences to generic mechanisms : model - based learning in analogical design . AIEDAM , 10 ( 2 ) , 131 e 136 . Bohm , M . R . , Vucovich , J . P . , & Stone , R . B . ( 2005 ) . Capturing creativity : Using a design repository to drive concept innovation . Paper presented at the Proceed - ings of DETC2005 , DETC05 / CIE - 85105 , Long Beach , California . Bradley , R . A . , & Terry , M . E . ( 1952 ) . Rank analysis of incomplete block designs . 1 . The method of paired comparisons . Biometrika , 39 , 324 e 345 . Carroll , J . D . ( 1976 ) . Spatial , nonspatial and hybrid models for scaling . Psycho - metrika , 41 , 439 e 463 . Casakin , H . , & Goldschmidt , G . ( 1999 ) . Expertise and the use of visual analogy : implications for design education . Design Studies , 20 ( 2 ) , 13 e 175 . Chakrabarti , A . K . , Dror , I . , & Nopphdol , E . ( 1993 ) . Interorganizational transfer of knowledge : an analysis of patent citations of a defense \ufb01rm . IEEE Transac - tions on Engineering Management , 40 ( 1 ) , 91 e 94 . Chakrabarti , S . , Dom , B . , Agrawal , R . , & Raghavan , P . ( 1998 ) . Scalable feature selection , classi\ufb01cation and signature generation for organizing large text data - bases into hierarchical topic taxonomies . The VLDB Journal , 7 ( 3 ) , 163 e 178 . Chan , J . , Fu , K . , Schunn , C . , Cagan , J . , Wood , K . , & Kotovsky , K . ( 2011 ) . On the bene\ufb01ts and pitfalls of analogies for innovative design : ideation perfor - mance based on analogical distance , commonness , and modality of examples . Journal of Mechanical Design , 133 ( 8 ) . http : / / dx . doi . org / 10 . 1115 / 1 . 4004396 . Chase , W . G . , & Simon , H . A . ( 1973 ) . The mind\u2019s eye in chess . In W . G . Chase ( Ed . ) , Visual information processing ( pp . 215 e 281 ) . New York , NY : Academic Press . Cheong , H . , Shu , L . H . , Stone , R . , & Wood , K . L . ( 2008 ) . Translating terms of the functional basis into biologically meaningful keywords . Paper presented at the Proceedings of the ASME International Design Engineering Technical Confer - ence , New York , NY , USA . Chi , M . T . H . , Feltovich , P . J . , & Glaser , R . ( 1981 ) . Categorization and represen - tation of physics problems by experts and novices . Cognitive Science , 5 ( 2 ) , 121 e 152 . Chi , M . T . H . , & Koeske , R . D . ( 1983 ) . Network representation of a child\u2019s dino - saur knowledge . Developmental Psychology , 19 ( 1 ) , 29 e 39 . Chiu , I . , & Shu , L . H . ( 2005 ) . Bridging cross - domain terminology for biomimetic design . Paper presented at the ASME IDETC / CIE , Long Beach , CA , USA . Christensen , B . T . , & Schunn , C . D . ( 2005 ) . Spontaneous access and analogical incubation e\ufb00ects . Creativity Research Journal , 17 ( 2 e 3 ) , 207 e 220 . Christensen , B . T . , & Schunn , C . D . ( 2007 ) . The Relationship of analogical dis - tance to analogical function and preinventive structure : the case of engineering design . Memory & Cognition , 35 ( 1 ) , 29 e 38 . 30 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 Collins , A . M . , & Quillian , M . R . ( 1969 ) . Retrieval time from semantic memory . Journal of Verbal Learning and Verbal Behavior , 8 ( 2 ) , 240 e 247 . Cowan , N . ( 2000 ) . The magical number 4 in short - term memory : a reconsidera - tion of mental storage capacity . Behavioral and Brain Sciences , 24 , 87 e 185 . CREAX . CREAX : Creativity for innovation . Retrieved 07 . 09 . 12 , from http : / / www . creax . com . Dahl , D . W . , & Moreau , P . ( 2002 ) . The In\ufb02uence and value of analogical thinking during new product ideation . Journal of Marketing Research , 39 ( 1 ) , 47 e 60 . Davies , J . , Goel , A . , & Nersessian , N . ( 2009 ) . A computational model of visual analogies in design . Journal of Cognitive Systems Research , Special Issue on Analogies e Integrating Cognitive Abilities Cognitive Systems Research , 3 , 204 e 215 . Deerwester , S . , Dumais , S . T . , Furnas , G . W . , & Landauer , T . K . ( 1990 ) . Indexing by latent semantic analysis . Journal of the American Society for Information Science , 41 ( 6 ) , 391 e 407 . Dunbar , K . ( 1997 ) . How scientists think : on - line creativity and conceptual change in science . In T . B . Ward , S . M . Smith , & J . Vaid ( Eds . ) , Creative thought : An investigation of conceptual structures and processes . Washington , D . C . : Amer - ican Psychological Association . Duran - Novoa , R . , Leon - Rovira , N . , Aguayo - Tellez , H . , & Said , D . ( 2011 ) . Inven - tive problem solving based on dialectical negation , using evolutionary algo - rithms and TRIZ heuristics . Computers in Industry , 62 , 437 e 445 . Dyer , J . H . , Gregersen , H . B . , & Christensen , C . M . ( 2011 ) . The innovator\u2019s DNA : Mastering the \ufb01ve skills of disruptive innovators . Boston , MA : Harvard Busi - ness Review Press . Fiske , A . P . ( 1992 ) . The four elementary forms of sociality : framework for a uni - \ufb01ed theory of social relations . Psychological Review , 99 ( 4 ) , 689 e 723 . Foltz , P . W . , Kintsch , W . , & Landauer , T . K . ( 1998 ) . The measurement of textual coherence with latent semantic analysis . Discourse Processes , 25 ( 2 e 3 ) , 285 e 307 . Forbus , K . D . , Gentner , D . , & Law , K . ( 1994 ) . MAC / FAC : a model of similarity - based retrieval . Cognitive Science , 19 , 141 e 205 . Fu , K . ( 2012 ) . Discovering and exploring structure in design databases and its role in stimulating design by analogy . ( Ph . D . Dissertation ) , Carnegie Mellon Uni - versity , Pittsburgh , PA , USA . Fu , K . , Cagan , J . , Kotovsky , K . , & Wood , K . ( 2013 ) . Discovering structure in design databases through function and surface based mapping . Journal of me - chanical Design , 135 ( 3 ) , 031006 . Fu , K . , Chan , J . , Cagan , J . , Kotovsky , K . , Schunn , C . , & Wood , K . ( 2013 ) . The meaning of \u201cnear\u201d and \u201cfar\u201d : the impact of structuring design databases and the e\ufb00ect of distance of analogy on design output . ASME Journal of Mechan - ical Design , 135 ( 2 ) , 021007 . Gentner , D . , Brem , S . , Ferguson , R . W . , Wol\ufb00 , P . , Markman , A . B . , & Forbus , K . D . ( 1997 ) . Analogy and creativity in the works of Johannes Kepler . In T . B . Ward , S . M . Smith , & J . Vaid ( Eds . ) , Creative thought : An investiga - tion of conceptual structures and processes ( pp . 403 e 459 ) . Washington D . C . : American Psychological Association . Gentner , D . , & Markman , A . B . ( 1997 ) . Structure mapping in analogy and sim - ilarity . American Psychologist , 52 ( 1 ) , 45 e 56 . Gick , M . L . , & Holyoak , K . J . ( 1980 ) . Analogical problem solving . Cognitive Psy - chology , 12 ( 3 ) , 306 e 355 . Gobet , F . , & Simon , H . A . ( 2000 ) . Five seconds or sixty ? Presentation time in expert memory . Cognitive Science , 24 ( 4 ) , 651 e 682 . Expert representation of design repositoxy space 31 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 Goel , A . , Bhatta , S . , & Stroulia , E . ( 1997 ) . Kritik : an early case - based design sys - tem . In M . Maher , & P . Pu ( Eds . ) , Issues and applications of case - based reasoning in design ( pp . 87 e 132 ) . Mahwah , NJ : Erlbaum . Goel , A . K . ( 1997 ) . Design , analogy , and creativity . IEEE Expert , 12 ( 3 ) , 62 e 70 . Gold\ufb01re , I . M . Invention machine gold\ufb01re : Unleashing the power of research . Retrieved 19 . 02 . 12 , from http : / / inventionmachine . com / products - and - services / innovation - software / gold\ufb01re - Research / . Goldschmidt , G . , & Smolkov , M . ( 2006 ) . Variances in the impact of visual stimuli on design problem solving performance . Design Studies , 27 , 549 e 569 . Gri\ufb03ths , T . L . , Kemp , C . , & Tenenbaum , J . B . ( 2008 ) . Bayesian models of cogni - tion . In R . Sun ( Ed . ) , Cambridge handbook of computational psychology ( pp . 59 e 100 ) . New York , NY : Cambridge University Press . Guttman , L . ( 1944 ) . A basis for scaling qualitative data . American Sociological Review , 9 ( 9 ) , 139 e 150 . Guttman , L . ( 1954 ) . A new approach to factor analysis : The radex . In P . F . Lazarsfeld ( Ed . ) , Mathematical thinking in the social sciences ( pp . 258 e 348 ) . New York , NY : Free Press . Hernandez , N . V . , Schmidt , L . C . , & Okudan , G . E . ( 2012a ) . Experimental assess - ment of TRIZ e\ufb00ectiveness in idea generation . Paper presented at the ASEE Annual Conference , San Antonio , TX , USA . Hernandez , N . V . , Schmidt , L . C . , & Okudan , G . E . ( 2012b ) . Systematic ideation e\ufb00ectiveness study of TRIZ . Paper presented at the ASME IDETC / CIE , Chi - cago , IL , USA . Herstatt , C . , & Kalogerakis , K . ( 2004 ) . How to use analogies for breakthrough in - novations . Working Papers / Technologie - und Innovationsmanagement , 24 ( 11 April 2009 ) . Retrieved from . http : / / biblioteca . universia . net / html _ bura / \ufb01cha / params / id / 39800023 . html . Hmelo - Silver , C . E . , & Pfe\ufb00er , M . G . ( 2004 ) . Comparing expert and novice under - standing of a complex system from the perspective of structures , behaviors , and functions . Cognitive Science , 28 ( 1 ) , 127 e 138 . Holyoak , K . J . , & Thagard , P . ( 1996 ) . The analogical scientist . In K . J . Holyoak , & P . Thagard ( Eds . ) , Mental leaps : Analogy in creative thought ( pp . 185 e 209 ) . Cambridge , MA : MIT Press . Houssin , R . , & Coulibaly , A . ( 2011 ) . An approach to solve contradiction prob - lems for safety integration in innovative design process . Computers in Industry , 62 , 398 e 406 . Huelsenbeck , J . P . , & Ronquist , F . ( 2001 ) . MRBAYES : Bayesian inference of phylogenetic trees . Bioinformatics , 17 ( 8 ) , 754 e 755 . Huth , A . , Nishimoto , S . , Vu , A . , & Gallant , J . ( 2012 ) . A continuous semantic space describes the representation of thousands of object and action categories across the human brain . Neuron , 76 ( 6 ) , 1210 e 1224 . Indukuri , K . V . , Ambekar , A . A . , & Sureka , A . ( 2007 ) . Similarity analysis of patent claimsusingnaturallanguageprocessingtechniques . PaperpresentedattheInterna - tional Conference on Computational Intelligence and Multimedia Applications . Inhelder , B . , & Piaget , J . ( 1969 ) . The early growth of logic in the child . New York , NY : W W Norton & Company . Jaynes , E . T . ( 2003 ) . Chapter 1 : Plausible reasoning probability theory : The logic of science . Cambridge : Cambridge University Press . Kaplan , C . , & Simon , H . A . ( 1990 ) . In search of insight . Cognitive Psychology , 22 ( 3 ) , 374 e 419 . Kasravi , C . , & Risov , M . ( 2007 ) . Patent mining e Discovery of business value from patent repositories . Paper presented at the Proceedings of the 40th Hawaii In - ternational Conference on System Sciences . 32 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 Kemp , C . , & Tenenbaum , J . B . ( 2008 ) . The discovery of structural form . Proceed - ings of the National Academy of Sciences of the United States of America , 105 ( 31 ) , 10687 e 10692 . Koch , S . , Bosch , H . , Giereth , M . , & Ertl , T . ( 2009 ) . Iterative integration of visual insights during patent search and analysis . Paper presented at the IEEE Sympo - sium on Visual Analytics Science and Technology , Atlantic City , NJ , USA . Kohonen , T . ( 1997 ) . Self - organizing maps . New York , NY : Springer . Krasnoslobodtsev , V . , & Langevin , R . ( 2005 ) . TRIZ application in development of climbing robots . Paper presented at the First TRIZ Symposium , Japan . Kurtz , K . , Miao , C . , & Gentner , D . ( 2001 ) . Learning by analogical bootstrapping . Journal of the Learning Sciences , 10 ( 4 ) , 417 e 446 . Landauer , T . K . , & Dumais , S . T . ( 1997 ) . A solution to Plato\u2019s problem : the latent semantic analysis theory of acquisition , induction , and representation of knowledge . Psychological Review , 1 , 211 e 240 . Landauer , T . K . , Foltz , P . W . , & Laham , D . ( 1998 ) . An introduction to latent se - mantic analysis . Discourse Processes , 25 ( 2 e 3 ) , 259 e 284 . Liang , Y . , Tan , R . , & Ma , J . ( 2008 ) . Patent analysis with text mining for TRIZ . Paper presented at the IEEE ICMIT , Bangkok , Thailand . Linsey , J . , Markman , A . B . , & Wood , K . L . ( 2008 ) . WordTrees : A method for design - by - analogy . Paper presented at the ASEE Annual Conference . Linsey , J . S . , Wood , K . L . , & Markman , A . B . ( 2008 ) . Modality and representa - tion in analogy . Arti\ufb01cial Intelligence for Engineering Design , Analysis & Manufacturing , 22 , 85 e 100 . http : / / dx . doi . org / 10 . 1017 / S0890060408000061 . Mcadams , D . A . , & Wood , K . L . ( 2000 ) . Quantitative measures for design by anal - ogy . Paper presented at the ASME Design Engineering Technology Conference . McCa\ufb00rey , T . ( 2012 ) . Innovation relies on the obscure : a key to overcoming the classic problem of functional \ufb01xedness . Psychological Science , 23 ( 3 ) , 215 e 218 . Mehalik , M . , & Schunn , C . ( 2007 ) . What constitutes good design ? A review of empirical studies of design processes . International Journal of Engineering Ed - ucation , 22 ( 3 ) , 519 e 532 . Morais , A . S . , Olsson , H . , & Schooler , L . J . ( 2013 ) . Mapping the structure of se - mantic memory . Proceedings of Cognitive Science , 37 ( 1 ) , 125 e 145 . Mukherjea , S . , Bhuvan , B . , & Kankar , P . ( 2005 ) . Information retrieval and knowledge discovery utilizing a biomedical patent semantic web . IEEE Trans - actions on Knowledge and Data Engineering , 17 ( 8 ) , 1099 e 1110 . Murphy , J . T . ( 2011 ) . Patent - based analogy search tool for innovative concept gen - eration . ( Ph . D . Dissertation ) , The University of Texas , Austin , TX . Nakagawa , T . ( 2012 ) . Creative problem - solving methodologies TRIZ / USIT : overview of my 14 years in research , education , and promotion . The Bulletin of the Cultural and Natural Sciences in Osaka Gakuin University , 64 . Nix , A . A . , Sherret , B . , & Stone , R . B . ( 2011 ) . A function based approach to TRIZ . Paper presented at the ASME IDETC / CIE , Washington , D . C . , USA . Novick , L . R . ( 1988 ) . Analogical transfer , problem similarity , and expertise . Jour - nal of Experimental Psychology : Learning , Memory , and Cognition , 14 ( 3 ) , 510 e 520 . Reich , Y . , & Shai , O . ( 2012 ) . The interdisciplinary engineering knowledge genome . Research in Engineering Design , 23 ( 3 ) , 251 e 264 . Shepard , R . N . ( 1980 ) . Multidimensional scaling , tree - \ufb01tting , and clustering . Sci - ence , 210 ( 4468 ) , 390 e 398 . Sneath , P . H . , & Sokal , R . R . ( 1973 ) . Numerical taxonomy : The principles and practice of numerical classi\ufb01cation . San Francisco , CA : Freeman . Stone , R . , & Wood , K . L . ( 2000 ) . Development of a functional basis for design . Journal of Mechanical Design , 122 , 359 e 370 . Expert representation of design repositoxy space 33 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 Szykman , S . , Sriram , R . D . , Bochenek , C . , Racz , J . W . , & Senfaute , J . ( 2000 ) . Design repositories : engineering design\u2019s new knowledge base . IEEE Intelle - gent Systems , 1094 - 7176 / 00 , 48 e 55 . Tseng , I . , Moss , J . , Cagan , J . , & Kotovsky , K . ( 2008 ) . The role of timing and analogical similarity in the stimulation of idea generation in design . Design Studies , 29 , 203 e 221 . Vattam , S . , Helms , M . , & Goel , A . ( 2010 ) . A content account of creative analogies in biologically inspired design . AIEDAM , Special Issue on Biologically Inspired Design , 24 , 467 e 481 . http : / / dx . doi . org / 10 . 1017 / S089006041000034X . Verhaegen , P . , D\u2019hondt , J . , Vandevenne , D . , Dewulf , S . , & Du\ufb02ou , J . R . ( 2011 ) . Identifying candidates for design - by - analogy . Computers in Industry , 62 , 446 e 459 . Vincent , J . F . V . , Bogatyreva , O . A . , Bogatyreva , N . R . , Bowyer , A . , & Pahl , A . K . ( 2006 ) . Biomimetics : its practice and theory . Journal of the Royal Society Interface , 3 ( 9 ) , 471 e 482 . Weisberg , R . W . ( 2009 ) . On \u2018out - of - the - box\u2019 thinking in creativity . In K . W . A . Markman ( Ed . ) , Tools for innovation ( pp . 23 e 47 ) . New York , NY : Oxford University Press . Wiggins , J . S . ( 1996 ) . An informal history of the interpersonal circumplex tradi - tion . Journal of Personality Assessment , 66 ( 2 ) , 217 e 233 . Wilson , J . O . , Rosen , D . , Nelson , B . A . , & Yen , J . ( 2010 ) . The e\ufb00ects of biological examples in idea generation . Design Studies , 31 ( 2 ) , 169 e 186 . Zhang , R . , Cha , J . , & Lu , Y . ( 2007 ) . A conceptual design model using axiomatic design , functional basis and TRIZ . Paper presented at the Proceedings of the 2007 IEEE IEEM . 34 Design Studies Vol - - No . - - Month 2013 Please cite this article in press as : Fu , K . , et al . , Expert representation of design repository space : A comparison to and validation of algorithmic output , Design Studies ( 2013 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002",
    "roudot2023utrack3d": "Article u - track3D : Measuring , navigating , and validating dense particle trajectories in three dimensions Graphical abstract Highlights d u - track3D provides an improved algorithm and work\ufb02ow for particle tracking d The algorithm quanti\ufb01es molecular traf\ufb01cking , polymerization , and transcription d Dynamic regions of interest reveal collective events nested in the cellular volume d Trackability score enables the automatic localization of tracking errors Authors Philippe Roudot , Wesley R . Legant , Qiongjing Zou , . . . , Reto Fiolka , Eric Betzig , Gaudenz Danuser Correspondence philippe . roudot @ univ - amu . fr ( P . R . ) , gaudenz . danuser @ utsouthwestern . edu ( G . D . ) In brief 3D microscopy of entire cells presents an image complexity that challenges the computation and observation of measurements such as particle trajectories . Roudot et al . introduce u - track3D , a package that combines robust particle tracking with image navigation and track validation algorithms for the discovery and unbiased analysis of dynamic processes . Roudot et al . , 2023 , Cell Reports Methods 3 , 100655 December 18 , 2023 \u00aa 2023 The Authors . https : / / doi . org / 10 . 1016 / j . crmeth . 2023 . 100655 ll Article u - track3D : Measuring , navigating , and validating dense particle trajectories in three dimensions Philippe Roudot , 1 , 4 , 7 , * Wesley R . Legant , 2 , 3 Qiongjing Zou , 1 Kevin M . Dean , 1 Tadamoto Isogai , 1 Erik S . Welf , 1 Ana F . David , 5 Daniel W . Gerlich , 5 Reto Fiolka , 1 Eric Betzig , 6 and Gaudenz Danuser 1 , * 1 Lyda Hill Department of Bioinformatics , UT Southwestern Medical Center , Dallas , TX , USA 2 JointDepartmentofBiomedicalEngineering , UniversityofNorthCarolinaatChapelHill , NorthCarolinaStateUniversity , ChapelHill , NC , USA 3 Department of Pharmacology , University of North Carolina , Chapel Hill , NC , USA 4 Aix Marseille University , CNRS , Centrale Marseille , I2M , Turing Centre for Living Systems , Marseille , France 5 Institute of Molecular Biotechnology of the Austrian Academy of Sciences , Vienna BioCenter , Vienna , Austria 6 Department of Molecular & Cell Biology , University of California , Berkeley , Berkeley , CA , USA 7 Lead contact * Correspondence : philippe . roudot @ univ - amu . fr ( P . R . ) , gaudenz . danuser @ utsouthwestern . edu ( G . D . ) https : / / doi . org / 10 . 1016 / j . crmeth . 2023 . 100655 SUMMARY We describe u - track3D , a software package that extends the versatile u - track framework established in 2D to address the speci\ufb01c challenges of 3D particle tracking . First , we present the performance of the new package in quantifying a variety of intracellular dynamics imaged by multiple 3D microcopy platforms and on the stan - dard 3D test dataset of the particle tracking challenge . These analyses indicate that u - track3D presents a tracking solution that is competitive to both conventional and deep - learning - based approaches . We then present the concept of dynamic region of interest ( dynROI ) , which allows an experimenter to interact with dy - namic 3D processes in 2D views amenable to visual inspection . Third , we present an estimator of trackability that automatically de\ufb01nes a score for every trajectory , thereby overcoming the challenges of trajectory vali - dation by visual inspection . With these combined strategies , u - track3D provides a complete framework for unbiased studies of molecular processes in complex volumetric sequences . INTRODUCTION Light - sheet \ufb02uorescence microscopy ( LSFM ) achieves three - dimensional ( 3D ) imaging with minimal phototoxicity , fast sam - pling , and near - isotropic resolution , 1 , 2 allowing the study of dy - namic intracellular processes in the entire cellular volume . 1 \u2013 4 While computer vision techniques are well established for inter - rogating cell biological processes in 2D , 5 these tools do not translate to both the visualization of measurement results and their validation in 3D . A key challenge for image analysis in 3D is the user interaction with the data . The manipulation of time - lapse 3D image volumes is often cumbersome , and any of the projection mechanisms necessary to map the 3D volume into a 2D representation on a screen is prone to artifacts that may cause erroneous conclusions . 6 Thus , computational tools for 3D image analysis must be able to reveal the complexity of 3D cellular and sub - cellular processes while being as automated as possible to avoid selection and perception biases . The most elementary way to measure the behavior of intracel - lular processes is particle tracking . Particles can comprise sub - diffraction - sized objects that appear in the image volume as bona \ufb01de spots , objects of an extended size that appear as a rigid structure , and larger deformable objects . The more complex the object\u2019sshapeis , themoresophisticated the methods neededfor particledetection . Theproblemofparticletrackingisthende\ufb01ned as the reconstruction of plausible trajectories from the coordi - nates [ x ( t ) , y ( t ) , z ( t ) ] oftheidenti\ufb01edparticles . Because the number of hypothetical trajectories grows super - exponentially over time , MOTIVATION Particle tracking is a ubiquitous task in the study of dynamic molecular and cellular pro - cesses through microscopy . Light - sheet microscopy has opened a path to acquiring complete cell volumes for investigation in three dimensions . However , hypothesis formulation and quantitative analysis have re - mained dif\ufb01cult due to fundamental challenges in the visualization and the veri\ufb01cation of large and dense sets of three - dimensional ( 3D ) particle trajectories . New software tools are required that allow microsco - pists to automatically track diverse particle movements in 3D , inspect the resulting trajectories in an infor - mative manner , and receive unbiased assessments of the quality of trajectories . Cell Reports Methods 3 , 100655 , December 18 , 2023 \u00aa 2023 The Authors . 1 This is an open access article under the CC BY - NC - ND license ( http : / / creativecommons . org / licenses / by - nc - nd / 4 . 0 / ) . ll OPEN ACCESS many approaches have been proposed to approximate the best solution . 7 \u2013 9 They typically combine a modeling of intracellular dy - namics and transient disappearances ( or mis - detection ) , statisti - cal approaches , 10 \u2013 16 or recursive neural networks 17 \u2013 19 to esti - mate the likelihood of trajectory - to - measurement associations and discrete optimization 10 , 20 \u2013 22 to select the best set of associ - ations at each time point . Only a few of these methods have been implemented in 3D 7 , 11 , 23 , 24 and even fewer tackle the visualiza - tion and validation challenges in dense sets of trajectories . Several open - source 25 \u2013 29 and commercial 30 \u2013 32 software pack - ages have been released to make volumetric rendering technol - ogy amenable to time - resolved 3D bioimaging data . Together they have proposed advances in accelerating sequence rendering , 26 , 28 modularity and extensibility , 25 , 27 annotation tools , 33 , 34 or scalable raytracing for high - quality rendering . 29 , 35 Those approaches provide an outside view of the 3D data and a range of tools to investigate the volume manually ( e . g . , in focused regions of interest [ ROIs ] or slices ) . They also reveal many chal - lenges associated with visual occlusions , 3D perception on a \ufb02at screen , interactions , and time - consuming volume exploration . BigDataViewer 36 has tackled the occlusion and perception chal - lenges indirectly by focusing on the caching and rendering of 2D slices at arbitrary orientations . The primary effect is to accel - erate rendering at the expense of 3D context , but it also avoids distortion of pixel - level information . As such , the Mamut software has demonstrated that combining BigDataViewer and 3D rendering of vectorial information enables annotation in large and complex 3D trajectories . 37 The immersive quality of virtual re - ality ( VR ) headsets has also been exploited to better perceive and interact with complex 3D datasets . 38 \u2013 42 Finally , a complementary approach to full volume visualization is the presentation of soft - ware - de\ufb01ned ROIs . The embryoMiner package 43 implements this idea by building static ROIs from groups of trajectories Figure 1 . u - track3D is a complete pipeline for the measurement , visualization , and evalua - tion of large sets of 3D trajectories The pipeline is here illustrated on lattice light - sheet imagingofHeLacellsundergoingmitosislabeledwith eGFP - labeled EB3 ( marking microtubule plus - ends , rendered in gray ) and mCherry - labeled centromere protein A ( marking kinetochores , rendered in red ) . thatare thenusedto systematicallyvisualize the data in a series of 3D tiles . The ap - proaches implemented in u - track3D also belong to this category ; however , the ROIs adapt over time to the changing geometry of collective processes . Hence , u - track3D allows the inspection of speci\ufb01c particle dynamics in a crowded and dynamic surrounding . Considerably fewer works have been dedicated to trajectory validation . Previous work can be categorized into ground - truth - based approaches and error - inference - based approaches . Within the former category , a time - consuming but widely accessible approach consists in using manual annotations to build a ground truth to be compared to the measured data - set . 24 , 44 , 45 Already in 2D , these techniques are subject to selec - tion bias , as it is easier to annotate bright and well - separated ob - jects . In 3D , these challenges are compounded by the complexity of visualization and selection of particles . Another ground - truth - based approach relies on the simulation of image sequences that mimic the acquisition . 7 , 46 \u2013 48 In contrast to the annotation - based approach , simulation can automate the pre - diction of tracking performance for simpli\ufb01ed scenarios in which image formation and particle dynamics are known perfectly a priori . The second category of methods attempts to infer the like - lihood of tracking errors directly on the data . Early efforts proposed heuristics that combine motions and density measure - ments to identify error - prone areas . 2 , 49 Arguably , a more elegant approach would be a direct analysis of the optimality of the re - constructed trajectories . This idea has been proposed by Cardi - nale and colleagues 50 and applied to estimating the scale , inten - sity , position , and associated con\ufb01dence intervals of the spindle pole body . While this con\ufb01dence interval re\ufb02ects the uncertainty about positions and velocities , the correctness of trajectory - to - measurement matches is not evaluated . Our goal was to design an algorithm for tracking error inference that does not require annotation or speci\ufb01c simulations , does not rely on heuristics other than those considered by the tracking algorithm , can handle heterogeneous scenarios , scales with the computational complexity of the tracker itself , and provides an interpretable output for mis - matched , spurious , and missing trajectories com - parable to conventional benchmarks . 7 Building upon our previous particle tracking work , 2 , 10 , 13 we thus designed the software package u - track3D to enable the measurement , observation , and validation of dynamic pro - cesses in 3D ( Figure 1 ) . u - track3D can detect and track 2 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS morphologically and dynamically diverse cellular structures , including single molecules , adhesion complexes , and larger macromolecular structures such as growing microtubules . The software design is open , allowing users to import the coordinate \ufb01les from other detection routines and then apply the u - track3D framework only for trajectory reconstruction . We introduce a li - brary for visualization and mapping of dynamic ROIs ( dynROIs ) that move with the biological structure under evaluation and enable an intuitive visualization of particle behaviors . Finally , we present a scalable approach for automatic assessment of trackability of each particle throughout the image volume by evaluating the stability of trajectory - to - assignment associations . RESULTS Measuring , visualizing , and validating 3D trajectories with the u - track3D pipeline Multiple particle tracking To generate a 3D particle tracking package , we adopted and modi\ufb01ed features that were critical for accurate particle tracking in 2D . 10 This includes the breakdown of trajectory reconstruction into a frame - by - frame association of corresponding particles fol - lowed by an association of the resulting track segments into full - length trajectories . Both steps rely on the same solution for optimal one - to - one assignments of particle detections and track segments in a bipartite graph . 10 , 51 The two - step approach per - mits the closing of temporal gaps in the detection of a particle as well as the handling of particle merging and splitting events . U - track3D incorporates a Kalman \ufb01ltering approach to model on the \ufb02y the characteristics of a particle\u2019s Brownian , directed , and heterogeneous motion , which supports both the procedure for frame - by - frame particle association and the one of track segment association . To support the concurrent tracking of ob - jects of variable sizes , we implemented a multiscale particle de - tector equipped with a generalized adaptive thresholding approach ( see section \u2018\u2018Multiscale particle detector\u2019\u2019 in STAR Methods ) . DynROIs Moving from 2D to 3D images complicates the interaction of a human observer with both raw and derived data . Widely used global image projections , including maximum intensity projec - tion ( MIP ) , and other volume rendering techniques are limited by the overlap of many dynamic structures along the viewing axis . 6 However , detailed visualization of 3D images and trajec - tories in their local context is essential for a user to adjust soft - ware control parameters and to interpret the underlying biology . Projection approaches have to be tailored to emphasize a subset of selected voxel or aspects of highest interest . Such projections should not only bring the particle or group of particles of interest into focus but also continuously adapt as the particles move . To meet this requirement , u - track3D incorporates a framework for rendering particle - centric dynROIs , thereby allowing the user to follow the particle behavior throughout its lifetime in a visually comprehensible format . DynROIs are implemented in a hierar - chical object structure across molecular , macromolecular , and cellular scales ( see section \u2018\u2018dynamic region of interest estima - tion\u2019\u2019 in STAR Methods ) . First , u - track3D provides a variety of shapes ( rectangle cuboids , spheres , cones , tubes , and rounded tubes ) to de\ufb01ne an ROI encompassing one , two , or three trajec - tories . Second , to manage larger sets of tracks , dynROIs are built by estimating an af\ufb01ne transform between the associated point cloud in consecutive time points . Finally , the top - level dyn - ROI is de\ufb01ned for the cell . For example , cells embedded in a 3D environment are often randomly oriented , and their orientation changes over time . While image - based registration can be used to correct changes in cell orientation , it is computationally expensive , especially as the size of the volume and length of the sequence grow . To reduce the computational burden , we segment and transform the cell mask into a randomly down - sampled point cloud , which is then used to estimate an af\ufb01ne transform . Trackability score Validation of tracking results is crucial for proper parameter adjustment during image acquisition and analysis as well as the biological interpretation of integrated measurements . How - ever , it remains an extremely challenging task in 3D datasets , particularly when the particle density is high . Contrary to a sce - nario in 2D where a single \ufb01eld of view presents a wide range of trajectories for visual inspection , dynROIs in 3D tend to capture only a few trajectories and cannot represent the hetero - geneity of local image quality , particle density , and dynamic properties , which all affect the tracking accuracy . To solve this problem , we complemented u - track3D with an option to compute a local trackability score . We use Monte Carlo simula - tion to determine for every trajectory and every time point the con\ufb01dence by which the algorithm was able to assign the chosen particle to the trajectory ( see section \u2018\u2018stochastic programming for the evaluation of trackability\u2019\u2019 in STAR Methods ) . Speci\ufb01cally , we exploit the particle history , the detection accuracy , and the associated motion model ( s ) to derive a trackability metric that represents the likelihood of each of the chosen associations vis - a ` - vis the set of alternative associations with neighboring par - ticles . We demonstrate the performance of the resulting score and how it can be used to compare trackability across space , time , and the molecules under study . Measurement of the kinetics of endocytosis in 3D To assess the performance of u - track3D , we investigated the dy - namics of various cellular structures imaged by light - sheet mi - croscopy ( Figure 2 ) . As reported with u - track , 10 gap closing is a crucial step in 2D particle tracking because of frequent , tran - sient disappearances : particles might not be detected , particles move in and out of the microscope\u2019s in - focus volume , or parti - cles can temporarily overlap in space . While the latter two sour - ces of disappearance are largely eliminated by proper 3D imag - ing , the challenges of false or missing detections remain . To test the performance of u - track3D in closing gaps , we examined the lifetimes of clathrin - coated structures forming at the cell plasma membrane ( Figures 2A \u2013 2C ) . These structures represent mostly sub - diffraction objects ; i . e . , they appear in an imaging volume as 3D point - spread functions . We used high - resolution diago - nally swept light - sheet microscopy 2 to sample every second a full volume of puncta generated by the GFP - labeled AP2 subunit of the endocytic coat . U - track3D recovered the canonical life - time distributions of abortive and maturing clathrin - coated pits , 52 , 53 that is , an exponential decay for abortive pits and Cell Reports Methods 3 , 100655 , December 18 , 2023 3 Article ll OPEN ACCESS Figure 2 . u - track3D supports a variety of imaging and biological scenarios ( A ) Maximum intensity projections ( MIPs ) of a rat kidney cell layer imaged with diagonally scanned light - sheet microscopy ( diaSLM ) . Cells are expressing eGFP - labeledalphasubunitoftheAP - 2complex . Greenboxis160 3 40 3 12 m m . Insetshowstrajectoriesofclathrinaggregatesclassi\ufb01edasclathrin - coatedstructures or maturing pits . ( B ) Normalizedmaximumintensityofeachtrajectoryasafunctionoflifetimeplottedforsixcellularlayerscomposedofmultiplecellseach . Thegreenlinedenotes the median of the cumulated distribution ( value T ) . ( legend continued on next page ) 4 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS Rayleigh - like distribution with maximal probability around 20 s for maturing pits ( Figure 2C ; Data S1 ; see section \u2018\u2018clathrin - medi - ated endocytosis study on a glass coverslip\u2019\u2019 in STAR Methods ) . While in 2D the identi\ufb01cation of those two populations relied on extensive trajectory analysis to discount incomplete trajec - tories , 53 our u - track3D software achieves accurate trajectory classi\ufb01cation directly by thresholding the maximum intensity of trajectories in 3D ( Figures 2B and 2C ) . Importantly , the distinc - tion of two lifetime distributions can only be obtained with the support of gap closing ( Figure 2C ) , suggesting that gaps remain a hurdle for accurate tracking in 3D . Measurement of instability in microtubule dynamics across dense 3D mitotic spindles With limited sampling frequency in volumetric imaging , particle tracking can be improved by dynamic motion models through Kalman \ufb01ltering . To assess the performance of a 3D implemen - tation of previously published motion models for 2D tracking of microtubule polymerization dynamics , 54 , 55 we imaged and tracked microtubule dynamics in HeLa cells by following GFP fusions of the microtubule plus - end tracking protein EB1 sampled at 1 Hz by lattice light - sheet imaging . 1 We quanti\ufb01ed metrics such as growth rate , growth lifetime , and pause fre - quency ( see section \u2018\u2018microtubule instability measurement\u2019\u2019 in STAR Methods ) . The latter is a measure for the probability that a stalled or shrinking microtubule , which is accompanied by disappearance of the EB1 particle in the video , is rescued to renewed growth ( see Figure S1 and Data S1 ) . Consistent with our previous observations in 2D , 54 u - track3D faithfully de - tected a dose - dependent decrease in all three metrics upon treatment of cells with the microtubule - destabilizing drug noco - dazole ( Figures 2D \u2013 2E ) . We also investigated the destabilizing effect of nocodazole on the number and duration of pauses or shrinkages ( Figure 2E ) . We then extended our analyses to mitotic cells , where the density of EB1 particles is much higher in central regions of the mitotic spindle ( see Data S1 ) . Both sce - narios show a strong response in nocodazole concentration , indicating that u - track3D properly captures the drug - induced variation of growth rate and lifetime ( Figures 2F and 2G ) , despite strong variations in particle density . Measurement of interaction between transcription factors and chromatin We then sought to investigate the impact of the depth information onthemeasurementofbiologicalquantitieswhencomparedto2D particle tracking . We employed a lattice light - sheet microscope to image the interactions between transcription factors ( TFs ) and chromatin in embryonic stem cells . In a study using the same bio - logical system but performed with 2D imaging , Chen et al . 56 had shown that TFs alternate between short - lived binding events at non - speci\ufb01cchromatinsites ( residencetime (cid:1) 0 . 75s ) , 3Ddiffusion ( average duration (cid:1) 3 s ) , and longer lived transcription events where the TF is bound at speci\ufb01c chromatin sites ( residence time (cid:1) 12 s ) . We performed the same analysis , now applying 3D tracking , and contrasted the results to the tracking of 2D projec - tions of the same 3D volumes ( Figures 2H \u2013 2K and section \u2018\u2018single molecule dynamics study with lattice light - sheet microscopy\u2019\u2019 in STARMethods ) . Ananalysis on2D projections reproduced the re - sultsoftheoriginalstudy . However , with3Danalysis , theresidence timeofspeci\ufb01cbindingeventswasreducedbyone - third ( (cid:1) 7 . 8sin 3D vs . 11 . 9 s in 2D ) . Interestingly , the shorter binding time ex - tracted from 3D trajectories is consistent with measurements per - formedinnuclearreceptorsstudies . 57 , 58 Together , thesedatasug - gest that the overlap caused by axial projections for 2D tracking may bias kinetic measurements . U - track3D leads the \ufb01eld of algorithms evaluated in the particle tracking challenge We evaluated u - track3D\u2019s competitiveness using the standard 3D test dataset of the particle tracking challenge . 7 We compared u - track3D against the top \ufb01ve approaches competing in the orig - inal benchmark as well as two tracking approaches that make use of recent advances in recursive neural networks ( RNNs ) 17 , 18 ( see Figures3andS2 ) . Thedatasetisdesignedtomimicviraldynamics alternating between con\ufb01ned and directed displacements that are large enough to create signi\ufb01cant ambiguities in the densest sce - narios ( Figure 3A ) . The dataset includes 12 sequences represent - ing four signal - to - noise ratios and three density levels ( see Fig - ure S2 ) . The challenge prescribes four metrics to compare the precision and accuracy of each approach . The \ufb01rst precision metric , alpha , relates to the Euclidean distance between the real trajectories and measured trajectories , while beta weighs this by the rate of spurious tracks . The accuracy metrics ignore distance and instead count the points within 5 pixels of a real detection . The Jaccard similarity coef\ufb01cient ( JSC ) or Jaccard index ( JI ) is computed as ( TP / N + FP ) where N is the number of real particles , TP is the number of true - positive matches , and FP is the number of false - positive matches . The JSCt describes the same metric using the count of trajectories that match at least 50 % of the seg - ments ; the rest is considered FP . Ourresultscanbesummarizedasfollows : \ufb01rst , u - track3Dranks \ufb01rst on all metrics and densities for signal - to - noise ratio ( SNR ) 4 ( C ) Probability density of lifetime for the set of trajectories above and below the threshold value T , with and without gap closing ( n = 6 cellular layers , pooled trajectories lifetimes ) . ( D ) MIP of HeLa cells in interphase imaged with lattice light - sheet microscopy ( LLSM ) expressing eGFP - labeled EB1 ( orange area is 30 3 32 3 7 m m ) . Overlay highlights EB1 trajectories . ( E ) Average microtubule lifetimes , microtubule growth rate , as well as average number and duration of pause and shrinkage events per trajectory for increasing concentrations of nocodazole ( n = 5 per conditions ; center line , median ; box limits , 25 and 75 percentiles ; whiskers , extremum ) . ( F ) MIP of HeLa cells in metaphase imaged with LLSM along with 45 (cid:3) rotation around the vertical axis . Overlay highlights EB1 trajectories . ( G ) Same as ( E ) measured for cells in metaphase ( n = 5 per conditions ) . ( H ) MIP of mouse embryonic stem ( ES ) cell nucleus imaged with LLSM expressing GFP - labeled TFs . Green box is 13 3 13 3 3 m m . Overlay highlights SOX2 trajectories . ( I ) MIP of ES cell nucleus imaged with LLSM expressing GFP - labeled TFs . Overlay highlights SOX2 trajectories tracked after MIP transformation . ( J ) Probability density of SOX2 binding time measured in LLSM overlaid with a two - component decay \ufb01t ( n = 1 cell ) . ( K ) Probability density of SOX2 binding time measured in projected LLSM data overlaid with a two - component decay \ufb01t ( n = 1 cell ) . Cell Reports Methods 3 , 100655 , December 18 , 2023 5 Article ll OPEN ACCESS and SNR 7 except for one data point where it ranks second ( Figures 3B and S2 ) . Also , u - track3D ranks \ufb01rst or among the top three approaches for lower SNR data . Strikingly , deep - learning approaches are outperformed by conventional methods , except at the level SNR = 1 . Of note , this scenario represents a breakpoint dataset that is not representative of the typical image quality targeted for a meaningful analysis of trajectory counts and lifetimes . At this low SNR , RNN - based approaches provide the best performance ; however , they are nonetheless insuf\ufb01 - ciently robust to generate accurate trajectory estimates . While a precise ablation study is out of the scope of this paper , ananalysisofthedesignofcompetingapproachescanhelpindis - cussing these performances . First , the most competitive ap - proaches in the original challenge 20 , 23 , 59 present differences in their design but they all rely on similar strategies : object detection is based on some type of adaptive thresholding , they use motion modeling for displacement prediction , they use a discrete optimization approach to assign predictions to measurements , and they employ various techniques to detect and correct for transient object disappearances . The u - track framework pursues the same strategies but incorporates piecewise - stationary motion modeling and gap closing to enhance tracking robustness during transitionsbetweendynamicregimes . Thecompetingapproaches assume stationarity in motion type ( mostly Brownian 20 ) or smooth transitions . 23 , 59 Second , the main difference between u - track3D and RNN - based tracking lies in the method that evaluates the cost of each trajectory - to - measurement association and the detection of transient disappearances . Both approaches employ a conventional detector and temporally greedy assignments . The techniques based on RNNs also use a locally adaptive detector and a discrete combinatorial optimization approach to assign pre - dictiontomeasurement . 51 Assuch , thedifferenceinperformances must lie in the method for motion and gap prediction . In a nutshell , the \ufb01rst RNN technique uses a long / short - term memory ( thus , in Figure 3 it is referred to as model - free LSTM ) network to \ufb01lter and predict particleposition . The second , morerecent , RNN tech - nique by the same authors uses gated recurrent units ( referred to as model - based GRU ) and reduces the parameter space needed for prediction by learning the moment and the covariance of a normal distribution associated with the state of each trajectory . Both approaches are trained on a simulated dataset presenting a mixture of Brownian , directed , and heterogeneous motions . In the original paper , 18 the model - based approach performs better thanboththeconventionalandthemodel - freeRNNonthe\u2018\u2018vesicle scenario , \u2019\u2019 which simulates free diffusion . However , the difference is not as clear in the \u2018\u2018virus scenario , \u2019\u2019 where the model - free approach often outperforms the latter . This suggests that the per - formanceofRNNscan improve uponthe state ofthe artwhentak - ing a priori knowledge into account to constrain the parameter space . However , asdemonstratedinourpreviouswork , 13 thesud - den transition between con\ufb01ned and directed displacement are poorly approximated by a normal distribution and better modeled byapiecewise - stationaryconstraint . Thisqualitativeanalysisdoes not imply that deep - learning approaches cannot be competitive but that further improvement requires the consideration of a priori knowledge on molecular dynamics in the design of the network architecture . Our performance analysis shows that , with the motion models implemented in the package , u - track3D Figure 3 . u - track3D performance in comparison to existing methods evaluated on a standard 3D test dataset with high particle density ( A ) MIP of the simulated virus dynamics overlayed with trajectories reconstructed by u - track3D . Trajectories are colored following a random colormap . ( B ) Performance metrics for precision ( alpha and beta ) and accuracy ( JSC and JSCt ) to compare different tracking pipelines . 6 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS remains a state - of - the - art tracker , despite the advances deep learning and other pipelines deploy . Visualization of adhesion formation in 3D We illustrate the application of a whole - cell dynROI with the study of spatial interactions between cell - matrix adhesions and \ufb02uores - cently labeled 3D collagen \ufb01bers in osteosarcoma cells imaged by axially swept light - sheet microscopy 3 ( Figure 4A ; Data S2 ) . ThedynROI allowed us tovisualize the relationshipbetweenadhe - sion shapes and its proximity to collagen \ufb01brils , showing two pop - ulationsofglobularand elongatedadhesions ( Figures4B \u2013 4D ) . The most elongated adhesions are located predominantly at the tip of pseudopodial extensions and align with the protrusive direction , while globular adhesions concentrate in the quiescent part of the membrane . Our measurements show that this elongation distribu - tion can be decomposed further ( Figure 4E ) . We found a unimodal distribution of mostly globular adhesions in close contact with collagen \ufb01bers ( see section \u2018\u2018adhesions and collagen interaction imaging and analysis\u2019\u2019 in STAR Methods ) . In contrast , adhesions withalesserdegreeofcollagencontactdisplayabimodaldistribu - tion of globular and elongated adhesions . These data suggest\u2014 quite unexpectedly from what is known in 2D\u2014that the most me - chanically engaged adhesions may be the least elongated . 60 We conjecture that adhesion elongation in 3D may be less driven by a zippering of an integrin - mediated plaque along a collagen \ufb01ber but rather dictated by the organization of cell - cortical actin \ufb01bers Figure 4 . DynROIs reveal the behavior of mo - lecular adhesions in 3D environments ( A ) Dual - colored orthogonal MIP of osteocarci - noma cells expressing eGFP - labeled paxillin and embedded in collagen labeled with Alexa Fluor 568 . Overlay highlights dynROI . ( B ) View of the dynROI . ( C and D ) Detection of adhesions colored as a function of the degree of collagen contact and elongation . ( E ) Probability density of elongation for adhesions with high and low degree of contact with collagen \ufb01bers ( n = 1 cell ) . or the local collagen architecture . Indeed , this behavior becomes apparent by replay of time - lapse sequences of the proximity and elongation parameters in the spatially stabilized dynROI ( Data S2 ) . DynROIs are thus a powerful way to assess the spatial distribution and heterogeneity of molecular interactions in highly dynamic cells . DynROI applied to mitotic spindle dynamics Many cellular processes involve a large - scale reorganization of macromolecular structures , which challenges 3D analysis . A goodexampleisthe vertebrate mitoticspin - dle . 61 Thousands of microtubules form a dense bipolar array while the two spindle poles move apart and rotate back and forth . Concurrently , spindle microtubules establish contacts with chro - mosomes at kinetochores and subsequently move chromosomes toward poles or the spindle center . This process is virtually impos - sible to understand by mere visual inspection of volume render - ings . We therefore assessed how u - track3D and dynROIs may facilitate the analysis . The image dataset comprises dual - channel time - lapse sequences of GFP - labeled microtubule plus ends and mCherry - labeled centromeres protein A marking kinetochores in mitotic HeLa cells acquired at 0 . 1 Hz by lattice light - sheet micro - scopy 62 from prometaphase to metaphase . Microtubule plus ends , kinetochores , andspindlepoleswerelocalizedbymultiscale particle detection . Pole trajectories can then be used to de\ufb01ne a dynROIthatfollowsthespindlemotion ( Figure5Aandsection \u2018\u2018dy - namic region of interest estimation\u2019\u2019 in STAR Methods ) . An embedded second dynROI follows the point cloud formed by the kinetochore trajectories ( Figure 5B and section \u2018\u2018dynamic region ofinterestestimation\u2019\u2019inSTARMethods ) . Basedonthepairofdyn - ROIs , we further constructa planardynROI withanorientation that is de\ufb01ned by the interpolar axis and a vector following the kineto - chore - associated dynROI motion ( Figures 5C and 5D ; Data S3 ; section\u2018\u2018dynamicregion ofinterestestimation\u2019\u2019inSTARMethods ) . Our framework for dynROI estimation thus enables the visualiza - tion of mesoscale structures composed of different molecular assemblies . In previous work , 62 using volume acquired at 1 Hz , we showed withspindle - widestatisticsandindirectmodelingthatkinetochore Cell Reports Methods 3 , 100655 , December 18 , 2023 7 Article ll OPEN ACCESS \ufb01berformation isacceleratedbyanaugmin - dependentnucleation and directional growth along the \ufb01ber toward kinetochores . We now use dynROIs todirectly visualize the dynamic space between spindle poles and kinetochores ( Figures 5E \u2013 5G ; Video S1 ) . We de\ufb01ne a kinetochore \ufb01ber assembly dynROI by a cone whose medialaxisconnectsspindlepoleandtargetkinetochore ( seesec - tion \u2018\u2018dynamic region of interest estimation\u2019\u2019 in STAR Methods ) . Using such dynROIs , we noted a directional bias in microtubule polymerization toward kinetochores , consistent with previous ob - servations . 62 However , wealsoobservedmicrotubulepolymeriza - tion branching off a kinetochore \ufb01ber and polymerizing toward another kinetochore ( circled in red in Figure 5G , time 53 \u2013 72 s ) . The branching was followed by rapid poleward movement of the targetedkinetochoreandanincreaseofplus - endcountinthedyn - ROI ( Figure 5G and 5H , time 93 \u2013 102 s ) suggesting that the target kinetochore was captured , generating a new avenue for microtu - bule ampli\ufb01cation . The example underscores how the dynROI li - braryimplementedinu - track3Denablesthevisualdiscoveryofdy - namic processes that are obscured in 3D image volumes . Trackability score to detect tracking ambiguities We developed a pipeline to assign a trackability score to every trajectory , based on the ambiguity of trajectory - to - measure - ment associations ( Figure 6 and section \u2018\u2018stochastic program - ming for the evaluation of trackability\u2019\u2019 in STAR Methods ) . Fig - ure 6A shows an ambiguous association between time t (cid:4) 1 and t where two hypotheses for the assignment of new detec - tions to track heads have a similar association cost . The bipar - tite graph matching identi\ufb01es a single optimal solution ( see Fig - ure 6A . i ) . To determine the level of ambiguity in the solution , we resample all track head predictions N times and test the stabil - ity of the original assignment ( one resampling example is shown in Figure 6A . ii ) . The approach is illustrated in Figures 6B \u2013 6D based on the tracking of a TF in previously published multifocus microscopy data . 63 Each dot indicates a resampled prediction of the particle location at t , and blue versus red de\ufb01nes whether the newly computed local assignment matches or differs from the original solution . The trackability score is de\ufb01ned as the fraction of matching samples . Hence , the score accounts for the local competition among detections for track head associ - ations and the uncertainty of motion prediction for each track head . We evaluated the capacity of our score to predict tracking quality in several scenarios . We simulated trajectory sets of increasing stochasticity along with a noisy detector and applied u - track3D to trace the particle movements ( parameters are described in Tables S1 and S2 ) . Using the ground truth , we then classi\ufb01ed each link of the extracted traces as a TP or FP . Figure 5 . DynROIs drive the visualization of chromosome capture by microtubules and reveal possible interactions between neighboring kinetochore \ufb01bers ( A \u2013 D ) Dual - colored orthogonal MIP of HeLa cells undergoing mitosis labeled with eGFP - labeled EB3 ( marking microtubule plus - ends rendered in gray ) and mCherry - labeledcentromereproteinA ( markingkinetochoresrenderedinred ) . Overlayshighlight ( A ) adynROIbuiltaroundcentrosometrajectories , ( B ) adynROI built around kinetochores trajectories , and ( C ) a plane built to visualize the dynamics of chromosomes relative to the spindle location . ( D ) View of the dynROI following description in ( H ) . ( E ) De\ufb01nition of a conical dynROI between a centrosome and a kinetochore . ( F ) Dual - colored orthogonal MIP of HeLa cells during prometaphase . Overlay highlights the motion of the dynROI . ( G ) Cumulative overlays of the detected microtubule plus - end position for three periods of 10 s between 53 and 102 s post nucleus envelope breakage . ( H ) Plus - ends count function of time and distance from the pole ( n = 1 dynROI ) . 8 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS This classi\ufb01cation allows us to compute for each simulated set the true JI JI = TP TP + FP + FN = TPFP + L , with L denoting the number of simulated links . Figure 6F presents simulated trajectories with increasing diffusion coef\ufb01cients ; the display is truncated to \ufb01ve consecutive time points to improve visibility . With a detection density \ufb01xed to 0 . 1 m m (cid:4) 3 and increasing speed of diffusion , the tracking performance rapidly deteriorates to completely inaccurate links and trajectory lifetime distributions ( see Figures 6G and 6H ) . The trackability score follows the decrease in the JI until it plateaus at 0 . 5 for very challenging conditions ( Figure 6I ) . The initially close match between track - ability score and JI is expected as larger diffusion speeds in - crease ambiguities in parallel to FPs and false negatives . How - ever , beyond a diffusion of 0 . 6 m m 2 / s , the prediction of the particle location in the next frame is less likely centered on the correct detection . As such , during the resampling of the ex - pected particle location , the rate of samples in agreement versus disagreement with the original link is de\ufb01ned by chance ; hence , the score plateaus at 0 . 5 . We also simulated a scenario in which the particle density increases at a diffusion \ufb01xed to 0 . 3 m m 2 / s . Analogous to the increase in diffusion , the trackabil - ity score follows the JI up to a density of 0 . 25 m m (cid:4) 3 where the two performance measurements start to diverge ( see Fig - ure S3 ) . In the case of directed displacements and a given density of 0 . 1 m m (cid:4) 3 , our trackability score also follows the true JI up to a critical velocity of 1 . 8 m m / s , which is more Figure 6 . The trackability score relies on the stochastic footprintofeachtrajectorytoinfer tracking accuracy ( A ) Example of a tracking ambiguity due to three trajectories in close proximity ( orange , blue , and red ) . Dashed lines represent the true motion be - tween track heads at time t (cid:4) 1 and detections at time t , represented by gray dots . Colored gradients represent the likelihood of each expected particle location at time t , estimated using the history of positions up to time t (cid:4) 1 and considering multiple motion model hypotheses . The optimal assignment between the expected and detected particle posi - tions at time t in this case yields an erroneous assignment fromtheorangetrack headtodetection 2 and from theblue track head to detection 3 ( graph A . i ) . Resampling of the expected locations results in a new assignment ( graph A . ii ) , this time without error . ( B ) Orthogonal MIP of ES cells expressing eGFP - labeled Sox2 molecules imaged by multifocus mi - croscopy . Overlaid boxes highlight the ROI enlarged in ( C ) \u2013 ( E ) . ( C ) Orthogonal MIP of ROI . Overlay shows a tra - jectory where two close detections create assign - ment ambiguity . ( D ) Overlay illustrates the stochastic resampling of the predicted particle positions at this time point ; blue circles , assignments in agreement with the original solution ; red circles , assignments that differ from the original solution . ( E ) Overlay shows trajectory segments colored ac - cording to estimated trackability scores . ( G ) Examples ofsimulated trajectories withdiffusion coef\ufb01cients ranging from 0 . 1 to 1 m m 2 / s with a \ufb01xed particle density of 0 . 1 m m (cid:4) 3 . Visualization is limited to \ufb01ve consecutive frames to reduce clutter . ( F ) Lifetime of simulated trajectories ( the change in distribution is due to trajectories leaving the \ufb01eld of view as the diffusion coef\ufb01cient increases ) . ( H ) Lifetime distribution measured through tracking shows a loss of the original distributions when the diffusion coef\ufb01cient exceeds 0 . 2 m m 2 / s . ( I ) Accuracy measured through the Jaccard index ( JI , blue ) ; the trackability score ( orange , dashed ) , which is derived without external ground truth , closely follows the JI up to a diffusion coef\ufb01cient 0 . 6 m m 2 / s beyond which tracking is random . Cell Reports Methods 3 , 100655 , December 18 , 2023 9 Article ll OPEN ACCESS than twice the average distance between a particle and its closest neighbor ( see Figure S4 ) . Finally , we sought to test our approach in a scenario in which trajectories undergo sud - den transitions between diffusive and directed motion ( see Fig - ure S5 ) . Of note , the densities , diffusion coef\ufb01cients , and veloc - ities are \ufb01xed in this scenario and the only parameter that varies is the transition rate , ranging from 0 ( no transition ) to 0 . 5 ( on average one transition every two frames ) . Our results show that the trackability score correctly predicts the reduction in tracking accuracy as increasing transition rates render tracking more ambiguous . A quasi - plateau is reached due to the high frequency of dynamic transitions . In conclusion , the proposed trackability score is able to detect changes in tracking quality in a variety of scenarios . Trackability score to compare tracking quality across time , space , and \ufb02uorescent channels To test the trackability score in real - world tracking , we \ufb01rst analyzed the spatiotemporal variation in the tracking quality of en - docytic pits ( see section \u2018\u2018endosome trackability on cell cultured on top of collagen\u2019\u2019 in STAR Methods ) associated with quiescent and protrusive parts of a cell membrane ( Figure 7A ) . We manually selected dynROIs to capture a quiescent area , a slow \ufb02uctuating protrusion - retraction cycle , and an abrupt and rapid protrusion . These dynROIs were selected within a larger dynROI compen - sating whole - cell movement ( see Video S2 ) . Trackability scores were consistently high in the quiescent dynROI , cyclically decreased in the \ufb02uctuating protrusion , and showed a sharp decrease where and when the fastest protrusion was located Figure 7 . Demonstration of trackability score on experimental data ( A ) Orthogonal MIP of breast cancer cells imaged with diaSLM expressing eGFP - labeled alpha subunit of the AP - 2 complex . Boxes show ROIs with quiescent ( blue ROI ) and slow / fast protrusion - retraction activity ( orange and yellow ROIs ) . Dot overlays show local level of ambiguity . ( B ) Number of track segments over time for the three ROIs ( n = 1 cell ) . ( C ) Trackability score over time for the three ROIs ( n = 1 cell ) . ( D ) Cumulative distribution of the average trackability score of trajectories for both EB3 and kinetochore channels sampling the dynamics of the mitotic spindle shown in Figure 5 . ( E ) Four ROIs ( two for each channel ) showing trajectories colored according to their mean trackability score . Trajectories were selected near the 10th and 90th percentiles of the cumulative distribution . Yellow dots show surrounding detections . 10 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS ( Figures 7B and 7C ) . The score thus accurately re\ufb02ects variations in particle trackability across space and time and detects time points of high ambiguity due to rapid movement . In a second experiment , we analyzed the spindle assembly dataset shown in Figures 5F \u2013 5H . The cumulative distributions of trajectory - averaged trackability scores showed that kineto - chore trajectories overall are more reliably reconstructed compared to microtubule plus - end trajectories ( Figure 7D ) . The score also enables trackability analysis on a per - trajectory basis . Trajectories with a score near the 90th percentile of the cumula - tive distribution appear to be error free for both plus end and kinetochore channels ( Figure 7E ) . In contrast , a plus - end trajec - tory with a score near the 10 th percentile shows a likely erro - neous path in an area of dense , crisscrossing microtubules . Because of the overall much higher trackability of the kineto - chore channel , a trajectory near the 10 th percentile shows only one likely wrong link caused by FP particle detection ( see arrow in Figure 7E detail of ROI 2 ) . Hence , the trackability score is a faithful reporter of the overall accuracy of tracking results , and it assists the selection of correctly tracked objects in dense tra - jectory populations . DISCUSSION We describe a new version of the popular tracking framework u - track , which now enables the study of particle dynamics in 3D live microscopy and tackles key challenges in the exploration and analysis of those complex datasets . The u - track3D software is implemented in MATLAB and distributed with a user - friendly graphical user interface ( GUI ) and tutorial scripts . The GUI is designed for testing the soft - ware and for the interactive visualization of the particle detec - tions , trajectories , and dynROI locations overlaid onto the raw data . In particular , both raw voxels and measurements can be observed using either slice - by - slice visualization or MIPs in the frame of reference of the laboratory or in a frame of refer - ence of a dynROI . The scripts are primarily used for batch pro - cessing and analysis at scale , and they enable the systematic visualization of tracking results across a full dataset . Our rendering engine is designed for automated and parallelized visualization of raw data and overlaid measurements , taking advantage of the asynchronous nature of processing jobs . Montages of raw and overlaid images can be easily speci\ufb01ed and saved in a variety of formats ( png , avi , and gifs ) . The script interface also provides a \ufb01ner control of the shape of dynROIs than the GUI ( cone , tube , rounded tubes , etc . ) . Finally , both detection and tracking can be limited to a dynROI , enabling the rapid adjustment of algorithm parameters before process - ing a full dataset . Two datasets are provided to test the soft - ware , one extracted from the endocytosis imaging introduced in Figure 7A and the other extracted from the mitosis imaging experiment introduced in Figure 5 . Computation time and memory usage for the complete pipeline , including detection , tracking with trackability inference , and dynROI de\ufb01nition , has been benchmarked using the endocytosis dataset on a stan - dard computing workstation and of\ufb01ce laptop ( see \u2018\u2018Bench - marking computational time and memory usage\u2019\u2019 in STAR Methods ) . Limitations of the study While the robustness and applicability of the software have been tested in several studies , 62 , 64 challenges remain toward a generic approach for automated exploration of 3D sequences . A chief bottleneck comes with the multiple sources of motions occurring across scales . While a given frame rate may be suf\ufb01 - cient to sample and track the motion of particles on a static sub - strate , the object may not be trackable when the particle - embedding volume moves rapidly . u - track3D addresses this problem with the estimation of dynROIs , which allow the pre - alignment of particle groups associated with an entire cell or sub - cellular structure . However , the automated estimation of the scale , type , clusters , and magnitude of those displacements remains an open problem for heterogeneous groups of objects . New developments in stochastic \ufb01ltering approaches for multi - scale displacements are thus necessary . Another key challenge in the analysis of dynamic 3D data is the quanti\ufb01cation of the motion of diffuse signaling molecules or macromolecular structures that do not present a well - de\ufb01ned particle in the imaged volume . These motions can be estimated coarsely using 3D optic \ufb02ow approaches , for which a few prom - ising methods tailored to \ufb02uorescence imaging have been proposed . 65 \u2013 67 Finally , the visualization and interaction with large multidi - mensional data remain dif\ufb01cult . While we believe the proposed dynROIs add a powerful tool for exploration of 3D sequences , the underlying rendering engine is limited to MIPs or slide - by - slide visualization . Community efforts are currently underway to provide a generic and versatile graphic library along with GUI interface such as Napari 28 and Sciview . 27 They could complete the capabilities of our renderer with more advanced volumetric rendering ( alpha , ray casting ) as well as surface rendering . We thus introduce u - track3D as feature - complete software for the quanti\ufb01cation and analysis of particle - like trajectories in 3Ds but also as a stepping - stone toward automated and unbiased exploration of any type of dynamic datasets . As we deliver the software to the community , we are continuously improving the software by \ufb01xing bugs and evaluating suggestions for improve - ments made by the community . STAR + METHODS Detailed methods are provided in the online version of this paper and include the following : d KEY RESOURCES TABLE d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND STUDY PARTICIPANT DE - TAILS B Clathrin - mediated endocytosis in an epithelial cell layer B Mitosis in HeLa cells B Transcription factors activity in embryonic stem cells B Adhesions in collagen - embedded osteosarcoma cells Cell Reports Methods 3 , 100655 , December 18 , 2023 11 Article ll OPEN ACCESS B Trackability of endocytic pits in collagen - embedded breast cancer cells d METHOD DETAILS B Multiscale particle detector B Dynamic region of interest estimation B Stochastic programming for the evaluation of track - ability B Clathrin - mediated endocytosis study on a glass cover - slip B Microtubule instability measurement B Single molecule dynamics study with lattice light - sheet microscopy B Adhesions and collagen interaction imaging and anal - ysis B Endosome trackability on cell cultured on top of collagen B Benchmarking computational time and memory usage d QUANTIFICATION AND STATISTICAL ANALYSIS SUPPLEMENTAL INFORMATION Supplemental information can be found online at https : / / doi . org / 10 . 1016 / j . crmeth . 2023 . 100655 . ACKNOWLEDGMENTS The authors are grateful to Yuko Mimori - Kyosue at the RIKEN Institute for the gift of the HeLa cells expressing eGFP - EB1 and fruitful conversations . We are also grateful to Zhe Liu at Janelia Research Campus for giving the ES cells . P . R . was funded by fellowship LT000954 / 2015 from the Human Frontiers in Science Program and the Investissements d\u2019Avenir French government pro - gram managed by the French National Research Agency ( ANR - 16 - CONV - 0001 ) and from Excellence Initiative of Aix - Marseille University - A * MIDEX . Work in the Danuser lab was funded by grant R35GM136428 . Maintenance and dissemination of the software was supported by grant RM1GM145399 . W . R . L . acknowledges support from the Searle Scholars Program , the Beck - man Young Investigator Program , an NIH New Innovator Award ( DP2GM136653 ) , and the Packard Fellows Program . Research in the labora - tory of D . W . G . was supported by the Vienna Science and Technology Fund ( WWTF ; project number LS14 - 009 ) and by the Austrian Science Fund ( FWF special research program SFB Chromosome Dynamics ; project number SFB F34 - 06 ) . R . F . was supported by the Cancer Prevention Research Institute of Texas ( RR160057 ) and the NIH ( R33CA235254 and R35GM133522 ) . Lattice light - sheet imaging of the mitotic spindle data was produced in collaboration with the Advanced Imaging Center , a facility jointly supported by the Gordon and Betty Moore Foundation and Howard Hughes Medical Institute at the Ja - nelia Research Campus . AUTHOR CONTRIBUTIONS P . R . and G . D . designed the research . P . R . wrote the tracking and rendering software and performed data analysis . Q . Z . developed the software\u2019s graph - ical user interface . W . R . L . , K . M . D . , A . D . , T . I . , and E . S . W . performed the biochemical and imaging experiments . K . M . D . , R . F . , and E . B . provided imag - ing resources . P . R . and G . D . wrote the manuscript . All authors read and pro - vided feedback on the \ufb01nal manuscript . DECLARATION OF INTERESTS The authors declare no competing interests . INCLUSION AND DIVERSITY We support inclusive , diverse , and equitable conduct of research . Received : March 12 , 2023 Revised : August 10 , 2023 Accepted : November 9 , 2023 Published : December 1 , 2023 REFERENCES 1 . Chen , B . - C . , Legant , W . R . , Wang , K . , Shao , L . , Milkie , D . E . , Davidson , M . W . , Janetopoulos , C . , Wu , X . S . , Hammer , J . A . , Liu , Z . , et al . ( 2014 ) . Lat - ticelight - sheetmicroscopy : Imagingmoleculestoembryosathighspatio - temporal resolution . Science 346 , 1257998 . 2 . Dean , K . M . , Roudot , P . , Reis , C . R . , Welf , E . S . , Mettlen , M . , and Fiolka , R . ( 2016 ) . Diagonally Scanned Light - Sheet Microscopy for Fast Volumetric Imaging of Adherent Cells . Biophys . J . 110 , 1456 \u2013 1465 . 3 . Dean , K . M . , Roudot , P . , Welf , E . S . , Danuser , G . , and Fiolka , R . ( 2015 ) . De - convolution - free Subcellular Imaging with Axially Swept Light Sheet Mi - croscopy . Biophys . J . 108 , 2807 \u2013 2815 . 4 . Liu , T . - L . , Upadhyayula , S . , Milkie , D . E . , Singh , V . , Wang , K . , Swinburne , I . A . , Mosaliganti , K . R . , Collins , Z . M . , Hiscock , T . W . , Shea , J . , et al . ( 2018 ) . Observing the cell in its native state : Imaging subcellular dynamics in multicellular organisms . Science 360 , eaaq1392 . 5 . Kervrann , C . , Sanchez Sorzano , C . O . , Acton , S . T . , Olivo - Marin , J . C . , and Unser , M . ( 2016 ) . A Guided Tour of Selected Image Processing and Anal - ysisMethodsforFluorescenceandElectronMicroscopy . IEEEJ . Sel . Top . Signal Process . 10 , 6 \u2013 30 . 6 . Driscoll , M . K . , and Danuser , G . ( 2015 ) . Quantifying Modes of 3D Cell Migration . Trends Cell Biol . 25 , 749 \u2013 759 . 7 . Chenouard , N . , Smal , I . , de Chaumont , F . , Ma (cid:1) ska , M . , Sbalzarini , I . F . , Gong , Y . , Cardinale , J . , Carthel , C . , Coraluppi , S . , Winter , M . , et al . ( 2014 ) . Objective comparison of particle tracking methods . Nat . Methods 11 , 281 \u2013 289 . 8 . Smal , I . , and Meijering , E . ( 2015 ) . Quantitative comparison of multiframe data association techniques for particle tracking in time - lapse \ufb02uores - cence microscopy . Med . Image Anal . 24 , 163 \u2013 189 . 9 . Manzo , C . , and Garcia - Parajo , M . F . ( 2015 ) . A review of progress in single particle tracking : from methods to biophysical insights . Rep . Prog . Phys . 78 , 124601 . 10 . Jaqaman , K . , Loerke , D . , Mettlen , M . , Kuwata , H . , Grinstein , S . , Schmid , S . L . , and Danuser , G . ( 2008 ) . Robust single - particle tracking in live - cell time - lapse sequences . Nat . Methods 5 , 695 \u2013 702 . 11 . Chenouard , N . , Bloch , I . , andOlivo - Marin , J . C . ( 2013 ) . MultipleHypothesis Tracking for Cluttered Biological Image Sequences . IEEE Trans . Pattern Anal . Mach . Intell . 35 , 2736 \u2013 3750 . 12 . Genovesio , A . , Liedl , T . , Emiliani , V . , Parak , W . J . , Coppey - Moisan , M . , and Olivo - Marin , J . - C . ( 2006 ) . Multiple particle tracking in 3 - D + t microscopy : method and application to the tracking of endocytosed quantum dots . IEEE Trans . Image Process . 15 , 1062 \u2013 1070 . 13 . Roudot , P . , Ding , L . , Jaqaman , K . , Kervrann , C . , and Danuser , G . ( 2017 ) . Piecewise - Stationary Motion Modeling and Iterative Smoothing to Track Heterogeneous Particle Motions in Dense Environments . IEEE Trans . Im - age Process . 26 , 5395 \u2013 5410 . 14 . Godinez , W . J . , Lampe , M . , Wo\u00a8rz , S . , M \u20ac uller , B . , Eils , R . , and Rohr , K . ( 2009 ) . Deterministic and probabilistic approaches for tracking virus parti - cles in time - lapse \ufb02uorescence microscopy image sequences . Med . Im - age Anal . 13 , 325 \u2013 342 . 15 . Ritter , C . , Wollmann , T . , Lee , J . - Y . , Imle , A . , M \u20ac uller , B . , Fackler , O . T . , Bar - tenschlager , R . , andRohr , K . ( 2021 ) . DataFusionandSmoothingforProb - abilistic Tracking of Viral Structures in Fluorescence Microscopy Images . Med . Image Anal . 73 , 102168 . 16 . Smal , I . , Draegestein , K . , Galjart , N . , Niessen , W . , and Meijering , E . ( 2008 ) . Particle \ufb01ltering for multiple object tracking in dynamic \ufb02uorescence mi - croscopy images : Application tomicrotubule growth analysis . IEEE Trans . Med . Imaging 27 , 789 \u2013 804 . 12 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS 17 . Spilger , R . , Imle , A . , Lee , J . - Y . , M \u20ac uller , B . , Fackler , O . T . , Bartenschlager , R . , and Rohr , K . ( 2020 ) . A Recurrent Neural Network for Particle Tracking in Microscopy Images Using Future Information , Track Hypotheses , and Multiple Detections . IEEE Trans . Image Process . 29 , 3681 \u2013 3694 . 18 . Spilger , R . , Lee , J . - Y . , Chagin , V . O . , Schermelleh , L . , Cardoso , M . C . , Bar - tenschlager , R . , and Rohr , K . ( 2021 ) . Deep probabilistic tracking of parti - cles in \ufb02uorescence microscopy images . Med . Image Anal . 72 , 102128 . 19 . Yao , Y . , Smal , I . , Grigoriev , I . , Akhmanova , A . , and Meijering , E . ( 2020 ) . Deep - learning method for data association in particle tracking . Bioinfor - matics 36 , 4935 \u2013 4941 . 20 . Sbalzarini , I . F . , and Koumoutsakos , P . ( 2005 ) . Feature point tracking and trajectory analysis for video imaging in cell biology . J . Struct . Biol . 151 , 182 \u2013 195 . 21 . Racine , V . , Sachse , M . , Salamero , J . , Fraisier , V . , Trubuil , A . , and Sibarita , J . - B . ( 2007 ) . Visualization and quanti\ufb01cation of vesicle traf\ufb01cking on a three - dimensional cytoskeleton network in living cells . J . Microsc . 225 , 214 \u2013 228 . 22 . Liang , L . , Shen , H . , De Camilli , P . , and Duncan , J . S . ( 2014 ) . A Novel Mul - tiple Hypothesis Based Particle Tracking Method for Clathrin Mediated Endocytosis Analysis Using Fluorescence Microscopy . IEEE Trans . Image Process . 23 , 1844 \u2013 1857 . 23 . Godinez , W . J . , and Rohr , K . ( 2015 ) . Tracking Multiple Particles in Fluores - cence Time - Lapse Microscopy Images via Probabilistic Data Association . IEEE Trans . Med . Imaging 34 , 415 \u2013 432 . 24 . Tinevez , J . - Y . , Perry , N . , Schindelin , J . , Hoopes , G . M . , Reynolds , G . D . , Laplantine , E . , Bednarek , S . Y . , Shorte , S . L . , and Eliceiri , K . W . ( 2017 ) . TrackMate : an open and extensible platform for single - particle tracking . Methods 115 , 80 \u2013 90 . 25 . Peng , H . , Bria , A . , Zhou , Z . , Iannello , G . , and Long , F . ( 2014 ) . Extensible visualization and analysis for multidimensional images using Vaa3D . Nat . Protoc . 9 , 193 \u2013 208 . 26 . Royer , L . A . , Weigert , M . , G \u20ac unther , U . , Maghelli , N . , Jug , F . , Sbalzarini , I . F . , and Myers , E . W . ( 2015 ) . ClearVolume : open - source live 3D visualization for light - sheet microscopy . Nat . Methods 12 , 480 \u2013 481 . 27 . G \u20ac unther , U . , Pietzsch , T . , Gupta , A . , Harrington , K . I . S . , Tomancak , P . , Gumhold , S . , and Sbalzarini , I . F . ( 2019 ) . Scenery : Flexible Virtual Reality Visualization on the Java VM . In 2019 IEEE Visualization Conference ( VIS ) , pp . 1 \u2013 5 . 28 . Sofroniew , N . , Lambert , T . , Evans , K . , Nunez - Iglesias , J . , Bokota , G . , Win - ston , P . , Pen\u02dca - Castellanos , G . , Yamauchi , K . , Bussonnier , M . , Pop , D . D . , et al . ( 2021 ) . Napari / Napari : 0 . 4 . 11 ( Zenodo ) . 29 . Pettersen , E . F . , Goddard , T . D . , Huang , C . C . , Meng , E . C . , Couch , G . S . , Croll , T . I . , Morris , J . H . , and Ferrin , T . E . ( 2021 ) . UCSF ChimeraX : Structure visualization for researchers , educators , and developers . Protein Sci . 30 , 70 \u2013 82 . 30 . ThermoFisherScienti\ufb01c . AmiraSoftware . https : / / www . thermo\ufb01sher . com / fr / fr / home / electron - microscopy / products / software - em - 3d - vis / amira - software . html . 31 . Arivis . ZEISS arivis Pro . https : / / www . arivis . com / products / pro . 32 . Oxford Instruments . Imaris AI Microscopy Image Analysis Software . https : / / imaris . oxinst . com / . 33 . Peng , H . , Tang , J . , Xiao , H . , Bria , A . , Zhou , J . , Butler , V . , Zhou , Z . , Gonza - lez - Bellido , P . T . , Oh , S . W . , Chen , J . , et al . ( 2014 ) . Virtual \ufb01nger boosts three - dimensional imaging and microsurgery as well as terabyte volume image visualization and analysis . Nat . Commun . 5 , 4342 . 34 . Wan , Y . , Otsuna , H . , Holman , H . A . , Bagley , B . , Ito , M . , Lewis , A . K . , Cola - santo , M . , Kardon , G . , Ito , K . , and Hansen , C . ( 2017 ) . FluoRender : joint freehand segmentation and visualization for many - channel \ufb02uorescence data analysis . BMC Bioinf . 18 , 280 . 35 . Jo\u00a8nsson , D . , Steneteg , P . , Sunde\u00b4n , E . , Englund , R . , Kottravel , S . , Falk , M . , Ynnerman , A . , Hotz , I . , and Ropinski , T . ( 2020 ) . Inviwo \u2014 A Visualization System with Usage Abstraction Levels . IEEE Trans . Vis . Comput . Graph . 26 , 3241 \u2013 3254 . 36 . Pietzsch , T . , Saalfeld , S . , Preibisch , S . , and Tomancak , P . ( 2015 ) . BigDa - taViewer : visualization and processing for large image data sets . Nat . Methods 12 , 481 \u2013 483 . 37 . Wolff , C . , Tinevez , J . - Y . , Pietzsch , T . , Stamataki , E . , Harich , B . , Guignard , L . , Preibisch , S . , Shorte , S . , Keller , P . J . , Tomancak , P . , and Pavlopoulos , A . ( 2018 ) . Multi - view light - sheet imaging and tracking with the MaMuT software reveals the cell lineage of a direct developing arthropod limb . El - ife 7 , e34410 . 38 . Usher , W . , Klacansky , P . , Federer , F . , Bremer , P . - T . , Knoll , A . , Yarch , J . , Angelucci , A . , and Pascucci , V . ( 2018 ) . A Virtual Reality Visualization Tool for Neuron Tracing . IEEE Trans . Vis . Comput . Graph . 24 , 994 \u2013 1003 . 39 . Wang , Y . , Li , Q . , Liu , L . , Zhou , Z . , Ruan , Z . , Kong , L . , Li , Y . , Wang , Y . , Zhong , N . , Chai , R . , etal . ( 2019 ) . TeraVRempowersprecisereconstruction of complete 3 - D neuronal morphology in the whole brain . Nat . Commun . 10 , 3474 . 40 . El Beheiry , M . , Doutreligne , S . , Caporal , C . , Ostertag , C . , Dahan , M . , and Masson , J . - B . ( 2019 ) . Virtual Reality : Beyond Visualization . J . Mol . Biol . 431 , 1315 \u2013 1321 . 41 . G \u20ac unther , U . , Harrington , K . I . S . , Dachselt , R . , and Sbalzarini , I . F . ( 2020 ) . Bi - onic Tracking : Using Eye Tracking to Track Biological Cells in Virtual Re - ality . Preprint at arXiv . 42 . Fouche\u00b4 , G . , Argelaguet , F . , Faure , E . , and Kervrann , C . ( 2023 ) . Immersive andinteractivevisualizationof3Dspatio - temporaldatausingaspacetime hypercube : Applicationtocelldivisionandmorphogenesisanalysis . Front . Bioinform . 3 . 43 . Schott , B . , Traub , M . , Schlagenhauf , C . , Takamiya , M . , Antritter , T . , Bart - schat , A . , Lo\u00a8f\ufb02er , K . , Blessing , D . , Otte , J . C . , Kobitski , A . Y . , et al . ( 2018 ) . EmbryoMiner : A new framework for interactive knowledge discovery in large - scale cell tracking data of developing embryos . PLoS Comput . Biol . 14 , e1006128 . 44 . ImageJ . Manual Tracking with TrackMate . https : / / imagej . github . io / plugins / trackmate / tutorials / manual - tracking . 45 . Lee , B . H . , and Park , H . Y . ( 2018 ) . HybTrack : A hybrid single particle tracking software using manual and automatic detection of dim signals . Sci . Rep . 8 , 212 . 46 . Rezato\ufb01ghi , S . H . , Pitkeathly , W . T . E . , Gould , S . , Hartley , R . , Mele , K . , Hughes , W . E . , and Burch\ufb01eld , J . G . ( 2013 ) . A framework for generating realistic synthetic sequences of total internal re\ufb02ection \ufb02uorescence mi - croscopyimages . In2013IEEE10thInternational Symposium on Biomed - ical Imaging , pp . 157 \u2013 160 . 47 . Rigano , A . , Galli , V . , Gonciarz , K . , Sbalzarini , I . F . , and Caterina , S . - D . - C . ( 2018 ) . An algorithm - centric Monte Carlo method to empirically quantify motion type estimation uncertainty in single - particle tracking . Preprint at bioRxiv . 48 . Balsollier , L . , Lavancier , F . , Salamero , J . , andKervrann , C . ( 2023 ) . Agener - ative model to synthetize spatio - temporal dynamics of biomolecules in cells . Preprint at arXiv . 49 . Kuhn , T . , Hettich , J . , Davtyan , R . , and Gebhardt , J . C . M . ( 2021 ) . Single molecule tracking and analysis framework including theory - predicted parameter settings . Sci . Rep . 11 , 9465 . 50 . Cardinale , J . , Rauch , A . , Barral , Y . , Szekely , G . , and Sbalzarini , I . F . ( 2009 ) . Bayesian image analysis with on - line con\ufb01dence estimates and its appli - cation to microtubule tracking . In 2009 IEEE International Symposium on Biomedical Imaging : From Nano to Macro , pp . 1091 \u2013 1094 . 51 . Jonker , R . , and Volgenant , A . ( 1987 ) . A shortest augmenting path algo - rithm for dense and sparse linear assignment problems . Computing 38 , 325 \u2013 340 . 52 . Loerke , D . , Mettlen , M . , Yarar , D . , Jaqaman , K . , Jaqaman , H . , Danuser , G . , andSchmid , S . L . ( 2009 ) . CargoandDynaminRegulateClathrin - CoatedPit Maturation . PLoS Biol . 7 , e1000057 . 53 . Aguet , F . , Antonescu , C . N . , Mettlen , M . , Schmid , S . L . , and Danuser , G . ( 2013 ) . AdvancesinAnalysisofLowSignal - to - NoiseImagesLinkDynamin Cell Reports Methods 3 , 100655 , December 18 , 2023 13 Article ll OPEN ACCESS and AP2 to the Functions of an Endocytic Checkpoint . Dev . Cell 26 , 279 \u2013 291 . 54 . Matov , A . , Applegate , K . , Kumar , P . , Thoma , C . , Krek , W . , Danuser , G . , and Wittmann , T . ( 2010 ) . Analysis of microtubule dynamic instability using a plus - end growth marker . Nat . Methods 7 , 761 \u2013 768 . 55 . Applegate , K . T . , Besson , S . , Matov , A . , Bagonis , M . H . , Jaqaman , K . , and Danuser , G . ( 2011 ) . plusTipTracker : Quantitative image analysis software for the measurement of microtubule dynamics . J . Struct . Biol . 176 , 168 \u2013 184 . 56 . Chen , J . , Zhang , Z . , Li , L . , Chen , B . - C . , Revyakin , A . , Hajj , B . , Legant , W . , Dahan , M . , Lionnet , T . , Betzig , E . , et al . ( 2014 ) . Single - Molecule Dynamics of Enhanceosome Assembly in Embryonic Stem Cells . Cell 156 , 1274 \u2013 1285 . 57 . Paakinaho , V . , Presman , D . M . , Ball , D . A . , Johnson , T . A . , Schiltz , R . L . , Lev - itt , P . , Mazza , D . , Morisaki , T . , Karpova , T . S . , and Hager , G . L . ( 2017 ) . Sin - gle - moleculeanalysisofsteroidreceptorandcofactoractioninlivingcells . Nat . Commun . 8 , 15896 \u2013 15914 . 58 . Voss , T . C . , Schiltz , R . L . , Sung , M . - H . , Yen , P . M . , Stamatoyannopoulos , J . A . , Biddie , S . C . , Johnson , T . A . , Miranda , T . B . , John , S . , and Hager , G . L . ( 2011 ) . Dynamic Exchange at Regulatory Elements during Chromatin Remodeling Underlies Assisted Loading Mechanism . Cell 146 , 544 \u2013 554 . 59 . Coraluppi , S . , and Carthel , C . ( 2004 ) . Recursive track fusion for multi - sensor surveillance . Inf . Fusion 5 , 23 \u2013 33 . 60 . Gardel , M . L . , Sabass , B . , Ji , L . , Danuser , G . , Schwarz , U . S . , and Waterman , C . M . ( 2008 ) . Traction stress in focal adhesions correlates bi - phasically with actin retrograde \ufb02ow speed . J . Cell Biol . 183 , 999 \u2013 1005 . 61 . Heald , R . , and Khodjakov , A . ( 2015 ) . Thirty years of search and capture : The complex simplicity of mitotic spindle assembly . J . Cell Biol . 211 , 1103 \u2013 1111 . 62 . David , A . F . , Roudot , P . , Legant , W . R . , Betzig , E . , Danuser , G . , andGerlich , D . W . ( 2019 ) . Augmin accumulation on long - lived microtubules drives ampli\ufb01cation and kinetochore - directed growth . J . Cell Biol . 218 , 2150 \u2013 2168 . 63 . Grimm , J . B . , English , B . P . , Choi , H . , Muthusamy , A . K . , Mehl , B . P . , Dong , P . , Brown , T . A . , Lippincott - Schwartz , J . , Liu , Z . , Lionnet , T . , and Lavis , L . D . ( 2016 ) . Bright photoactivatable \ufb02uorophores for single - molecule im - aging . Nat . Methods 13 , 985 \u2013 988 . 64 . Isogai , T . , Dean , K . M . , Roudot , P . , Shao , Q . , Cillay , J . D . , Welf , E . S . , Dris - coll , M . K . , Royer , S . P . , Mittal , N . , Chang , B . - J . , etal . ( 2019 ) . Direct Arp2 / 3 - vinculin binding is essential for cell spreading , but only on compliant sub - strates and in 3D . Preprint at bioRxiv . 65 . Boquet - Pujadas , A . , Lecomte , T . , Manich , M . , Thibeaux , R . , Labruye ` re , E . , Guille\u00b4n , N . , Olivo - Marin , J . - C . , and Dufour , A . C . ( 2017 ) . BioFlow : a non - invasive , image - based method tomeasure speed , pressure and forces in - side living cells . Sci . Rep . 7 , 9178 . 66 . Manandhar , S . , Bouthemy , P . , Welf , E . , Roudot , P . , and Kervrann , C . ( 2018 ) . A sparse - to - dense method for 3D optical \ufb02ow estimation in 3D light - microscopy image sequences . In IEEE International Symposium on Biomedical Imaging , pp . 952 \u2013 956 . 67 . Manandhar , S . , Bouthemy , P . , Welf , E . , Danuser , G . , Roudot , P . , and Ker - vrann , C . ( 2020 ) . 3D \ufb02ow \ufb01eld estimation and assessment for live cell \ufb02uo - rescence microscopy . Bioinformatics 36 , 1317 \u2013 1325 . 68 . Gibbs , Z . A . , Reza , L . C . , Cheng , C . - C . , Westcott , J . M . , McGlynn , K . , and Whitehurst , A . W . ( 2020 ) . The testis protein ZNF165 is a SMAD3 cofactor that coordinates oncogenic TGF b signaling in triple - negative breast can - cer . Elife 9 , e57679 . 69 . Westcott , J . M . , Prechtl , A . M . , Maine , E . A . , Dang , T . T . , Esparza , M . A . , Sun , H . , Zhou , Y . , Xie , Y . , and Pearson , G . W . ( 2015 ) . An epigenetically distinct breast cancer cell subpopulation promotes collective invasion . J . Clin . Invest . 125 , 1927 \u2013 1943 . 70 . Basset , A . , Boulanger , J . , Salamero , J . , Bouthemy , P . , and Kervrann , C . ( 2015 ) . Adaptive Spot Detection With Optimal Scale Selection in Fluores - cence Microscopy Images . IEEE Trans . Image Process . 24 , 4512 \u2013 4527 . 71 . Olivo - Marin , J . - C . ( 2002 ) . Extraction of spots in biological images using multiscale products . Pattern Recognit 35 , 1989 \u2013 1996 . 72 . Lindeberg , T . ( 1998 ) . Feature detection withautomatic scaleselection . Int . J . Comput . Vis . 30 , 79 \u2013 116 . 73 . Besl , P . J . , and McKay , N . D . ( 1992 ) . A method for registration of 3 - D shapes . IEEE Trans . Pattern Anal . Mach . Intell . 14 , 239 \u2013 256 . 74 . Burkard , R . E . , and C\u00b8ela , E . ( 1999 ) . Linear Assignment Problems and Ex - tensions . In Handbook of Combinatorial Optimization : Supplement Vol - ume A , D . - Z . Du and P . M . Pardalos , eds . ( Springer US ) ) , pp . 75 \u2013 149 . 75 . Jones , W . , Chawdhary , A . , and King , A . ( 2015 ) . Revisiting Volgenant - Jonker for approximating graph edit distance . In International Workshop on Graph - Based Representations in Pattern Recognition ( Springer ) , pp . 98 \u2013 107 . 76 . Shapiro , A . ( 2001 ) . Monte Carlo simulation approach to stochastic pro - gramming . In Proceeding of the 2001 Winter Simulation Conference ( Cat . No . 01CH37304 ) , pp . 428 \u2013 431 . 77 . Frangi , A . F . , Niessen , W . J . , Vincken , K . L . , andViergever , M . A . ( 1998 ) . Mul - tiscale vessel enhancement \ufb01ltering . In Medical Image Computing and Computer - AssistedIntervention\u2014MICCAI\u201998LectureNotesinComputer Science ( Springer Berlin Heidelberg ) ) , pp . 130 \u2013 137 . 78 . Aguet , F . , Jacob , M . , and Unser , M . ( 2005 ) . Three - dimensional feature detection using optimal steerable \ufb01lters . In IEEE International Conference on Image Processing , 2005 . ICIP 2005 , pp . 1158 \u2013 1161 . p . II - . 14 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS STAR + METHODS KEY RESOURCES TABLE REAGENT or RESOURCE SOURCE IDENTIFIER Chemicals , peptides , and recombinant proteins DMEM Thermo Fisher Scienti\ufb01c 12491015 DMEM without phenol red Invitrogen N / A Antibiotic / antimycotic for IMCD cells Thermo Fisher Scienti\ufb01c 15240062 DMEM HeLa Cells Institute of Molecular Biotechnology of the Austrian Academy of Sciences N / A penicillin \u2013 streptomycin Sigma - Aldrich P0781 GlutaMAX Thermo Fisher Scienti\ufb01c 35050061 rat tail - derived Collagen Type I Corning 354236 truncated CMV promoter Addgene 110718 mNeonGreen - Paxillin Allele Biotechnology N / A Alexa Fluor 568 Thermo Fisher Scienti\ufb01c A20003 Deposited data Raw test dataset This paper https : / / zenodo . org / record / 6881276 Experimental models : Cell lines Inner medulla collecting duct ( IMCD ) mouse epithelial cells ATCC ATCC : CRL - 2123 , RRID : CVCL _ 0429 U2OS female cells ATCC ATCC : HTB - 96 RRID : CVCL _ 0042 Sum159O breast cancer cells Whitehurst 68 , 69 N / A HeLa Kyoto line S . Narumiya RRID : CVCL _ 1922 Stable Embryonic Stem Cells Liu 56 N / A Software and algorithms u - track3D This paper https : / / doi . org / 10 . 5281 / zenodo . 10055024 MATLAB 2015a - 2023a Mathworks RRID : SCR _ 001622 Gmic D . Tschumperle\u00b4 / GREYC / CNRS https : / / gmic . eu / AMIRA v6 Thermo Fisher Scienti\ufb01c RRID : SCR _ 007353 LabView Emerson RRID : SCR _ 014325 Other FCS Thermo Fisher Scienti\ufb01c A5256701 FBS U2OS Sigma F0926 - 500ML 5 mm diameter coverslips Thomas Scienti\ufb01c 64 \u2013 0700 NA 0 . 71 water dipping illumination objective Special Optics 54 - 10 - 7 25X / NA 1 . 1 water dipping detection objective Nikon Instruments CFI75 Apo LWD 25XY Camera ORCA - Flash 4 . 0 Hamamatsu Photonics C11440 - 22C Bidirectional scan unit Cambridge Technology 6215 Remote focusing system Nikon Instruments CFI S Plan Fluor ELWD Achromatic doublet Edmund Optics 49 \u2013 396 Filters Chroma Technology Corporation ZET405 / 488 / 561 / 640 , ZT568rdc , ET525 / 50m , ET600 / 50m Piezzo Actuator Physik Instrumente P - 603 . 1S2 Piezzo controller Physik Instrumente E (cid:4) 709 . SRG Cell Reports Methods 3 , 100655 , December 18 , 2023 e1 Article ll OPEN ACCESS RESOURCE AVAILABILITY Lead contact Further information and requests for resources should be directed to and will be ful\ufb01lled by the lead contact , Philippe Roudot ( philippe . roudot @ univ - amu . fr ) . Materials availability This study did not generate new unique reagents . Data and code availability d Raw TIFF light - sheet imaging data used in Figures 5 and 7 have been deposited in https : / / zenodo . org / record / 6881276 and are publicly available as of the date of publication . DOIs are listed in the key resources table . They are automatically downloaded by the tutorial scripts . The full imaging datasets used in this paper represent tens of Terabyte of data and are too large to be made available on a server maintained for public access . However , this data can be made available through other means ( such as mail , or large \ufb01le transfer sev ) upon request to the lead contact author . d All original code has been deposited at https : / / zenodo . org / records / 10055024 and is publicly available as of the date of pub - lication . DOIs are listed in the key resources table . A user\u2019s guide for both GUI and scripts and test datasets are available within the same repository . The repository used for update and bug\ufb01xes is at https : / / github . com / DanuserLab / u - track3D . d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request . EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS Clathrin - mediated endocytosis in an epithelial cell layer Inner medulla collecting duct ( IMCD ) mouse epithelial cells stably expressing alpha - adaptin GFP 53 were cultured in in Dulbecco\u2019s modi\ufb01ed Eagle\u2019s medium ( DMEM ) supplemented with 10 % fetal calf serum ( FCS ) and 1 % antibiotic / antimycotic at 37 (cid:3) C . Mitosis in HeLa cells HeLa cell lines stably expressing \ufb02uorescent reporter proteins were derived from a HeLa Kyoto line obtained from S . Narumiya ( Kyoto University , Kyoto , Japan ) and cultured in DMEM ( produced in - house at Institute of Molecular Biotechnology of the Austrian Academy of Sciences ) supplemented with 10 % FCS ( Thermo Fisher Scienti\ufb01c ) , 1 % ( v / v ) penicillin \u2013 streptomycin ( Sigma - Aldrich ) , and GlutaMAX ( Thermo Fisher Scienti\ufb01c ) at 37 (cid:3) C with 5 % CO2 in a humidi\ufb01ed incubator . Transcription factors activity in embryonic stem cells Stable Embryonic Stem ( ES ) cell lines were generated and cultured as described in . 56 Brie\ufb02y , ES cell - imaging experiments were per - formed in Dulbecco\u2019s modi\ufb01ed Eagle\u2019s medium ( DMEM ) without phenol red ( Invitrogen ) , 15 % FBS , 1 mM GlutaMAX at 37 (cid:3) C . Adhesions in collagen - embedded osteosarcoma cells Mycoplasma - free U2OS female cells were cultured in DMEM with 10 % FBS ( Sigma ; F0926 - 500ML ) at 5 % CO2 and 37 (cid:3) C . Trackability of endocytic pits in collagen - embedded breast cancer cells Sum159O breast cancer cells stably expressing alpha - adaptin GFP are a derivative of Sum159 cells obtained from Angelique Whitehurst and prepared as in . 68 , 69 Cells were plated on a (cid:1) 2 mm thick bed of rat tail - derived Collagen Type I ( 354236 , Corning ) at 37 (cid:3) C . METHOD DETAILS Multiscale particle detector Three - dimensional microscopy imposes speci\ufb01c constraints on the design of a particle detector . First , the diversity of shapes and sizes of intracellular structures may not be visible to the naked eye in a volumetric rendering , we must thus design a detector that is responsive to those variations . Second , light scattering and variation in signal intensity create large changes in signal - to - noise ratio ( SNR ) across space that are also dif\ufb01cult to assess visually . Our detector must then be adapted to those changes from low to high SNR . Finally , the large dimension of 3D data sets requires the design of computationally ef\ufb01cient approaches . Following , we describe a multiscale detector equipped with an adaptive thresholding approach that tests multiple possible scales at each location through the implementation of multiple iterations of \ufb01ltering . This tool is similar in spirit to other multiscale detectors that combine the local - ization task with the scale selection task using either Gaussian kernels 70 or the wavelet transforms . 71 The main difference is the com - bination of multiple thresholding masks obtained through the evaluation of a statistical test for each voxel and for each of the eval - uated scales . e2 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS We \ufb01rst developed a multiscale adaptive thresholding approach inspired by our previous work focused on the sensitive detection limited to the case of diffraction - limited \ufb02uorescent structures . 53 Let us consider the following image model : M \u00f0 x ; A ; s ; m ; C \u00de = A \u00f0 x \u00de G s ; m \u00f0 x \u00de + C \u00f0 x \u00de + e \u00f0 x \u00de where A denotes the spot amplitude , x the 3D position , G s ; m \u00f0 x \u00de is a Gaussian function with standard deviation s and mean m , C is the background signal and e \u00f0 x \u00de is the additive noise following a Poisson - Gaussian stochastic footprint . The least - square formulation of our optimization problem as argmin A \u00f0 x \u00de ; C \u00f0 x \u00de X x \u02db W (cid:2) A \u00f0 x \u00de G s ; m \u00f0 x \u00de + C \u00f0 x \u00de (cid:4) I \u00f0 x \u00de (cid:3) 2 ; where I \u00f0 : \u00de denotes the image volume and W is a 3D box of size 8 s , can be simpli\ufb01ed to the resolution of a linear system that can be decomposed in multiple \ufb01ltering passes : A \u00f0 x 0 \u00de = \u00f0 I (cid:5) G s ; 0 \u00de\u00f0 x 0 \u00de (cid:4) \u00f0 G s ; 0 (cid:5) 1 w \u00de\u00f0 x 0 \u00de nG s ; 0 2 + nG s ; 0 2 and , C \u00f0 x 0 \u00de = \u00f0 I (cid:5) 1 w \u00de\u00f0 x 0 \u00de (cid:4) nG s ; 0 A \u00f0 x 0 \u00de n where x 0 is the \ufb01xed voxel position 1 w is a unitary convolution kernel along W , n is the number of voxels encompassed in W . The statistical analysis of the local residuals resulting from the \ufb01t r \u00f0 x \u00de = (cid:2) A \u00f0 x 0 \u00de G s ; m \u00f0 x \u00de + C \u00f0 x 0 \u00de (cid:4) I \u00f0 x \u00de (cid:3) ; with x \u02db W , provides a p - value - based threshold for testing for the hypothesis that A \u00f0 x 0 \u00de [ C \u00f0 x 0 \u00de as described in . 53 This approach yields a sensitive binary map H 0 ; s \u00f0 : \u00de for the detection of the voxel describing a \ufb02uorescence object at scale s . This approach avoids the \ufb01tting of an object template in order to reduce computation time . Next , we carry out this adaptive thresholding step at multiple scale to obtain a vote map . V \u00f0 x \u00de = X s \u02db U H 0 ; s \u00f0 x \u00de where U is the scale range , typically ranging between 0 . 120 and 1 m m . The resulting object mask V \u00f0 : \u00de thus summarizes the presence of particles at any scale at a given voxel ( see Figure S6 ) using only \ufb01ltering operations that can process each voxel in a parallelized fashion . In order to re\ufb01ne the localization of objects present in contiguous object masks , we implemented a multiscale Laplacian of a Gaussian \ufb01ltering framework 72 to estimate a map of scale response S \u00f0 : \u00de for each voxel de\ufb01ned as : S \u00f0 x \u00de = argmax s \u02db U s 2 V 2 \u00f0 I \u00f0 x \u00de (cid:5) G s ; 0 \u00f0 x \u00de\u00de where V 2 \u00f0 : \u00de denotes the Laplacian operator . The watershed algorithm is then applied to further segment this scale response map to detecttouchingobjects . Thecenterofobjectisdeterminedthroughtheweightedcentroidofthevoxelsbelongingtoasameobjectmask . Dynamic region of interest estimation In order to visualize and map the molecular processes nested in volumetric time lapse sequences , we propose a framework for the de\ufb01nition of dynamic regions of interest ( dynROI ) from point cloud sequences . Those dynROIs are described by dynamic bounding boxes ( or rectangular cuboids ) that are sized to \ufb01t the data optimally and oriented according to a moving frame of reference . In this note , we describe the general principles underpinning the estimation of those dynROIs from dynamic point clouds and their imple - mentation across scales : from cellular down to molecular dynROIs . Generic point cloud tracking principle We \ufb01rst de\ufb01ne an optimal frame of reference in the \ufb01rst time point of the sequence with an origin O 0 described by the average point cloud position and with unit vectors \u00f0 u 0 ; v 0 ; w 0 \u00de described by the eigenvectors of the covariance matrix of the point positions ( a . k . a . principal component analysis ) . The orientation of the dynROI box in the \ufb01rst frame is described by this frame of reference and its size is de\ufb01ned by the boundaries of the point cloud augmented by a tunable margin ( default is set to 5 voxel ) . The frame of reference at time t is then estimated through a rigid transform as : d \u00f0 O t ; R t \u00de = argmin O t ; R t X x t \u02db U t x 0 \u02db U 0 k x t (cid:4) \u00f0 R t x 0 + O t \u00dek using the Iterative closest point algorithm , 73 where U t denotes the set of points coordinate at time t . The unit vector \u00f0 u t ; v t ; w t \u00de are then estimated by applying the rigid transform to \u00f0 u 0 ; v 0 ; w 0 \u00de . At each time point the size of the box is adjusted to \ufb01t the extension of the Cell Reports Methods 3 , 100655 , December 18 , 2023 e3 Article ll OPEN ACCESS current point cloud in the current orientation with the additional margin . Multiple dynROI shapes have then been implemented to adjust to the local process ( box , sphere , tube , rounded tube , plane and cone ) . Dynamic region of interest estimation for the cell The cell is \ufb01rst segmented using the Otsu algorithm and the point cloud representing the cell mask is downsampled randomly to reduce its density by 90 % and speed up computations . The generic point cloud tracking principle described above is applied to the downsampled sequence with a margin set to 30 voxels and a box - shaped dynROI ( see Figure S7 ) . Dynamic region of interest for the spindle Spindle poles were detected using the multiscale detector with the default p value ( set to 0 . 005 ) and scales ranging from 0 . 4 to 0 . 8 ums . The motion of poles was modeled with a piecewise stationary Brownian and Directed motion model with a maximum instantaneous displacement set to 3 times the process noise estimated from a Kalman \ufb01ltering of the trajectory , a lower bound set at 0 . 5 m m and upper boundsetat0 . 8 m m . Failuretodetectthe verydynamicaggregate onnucleatingmicrotubulesishandledwithgapclosing , themaximum gap is set to 2 s ( or 2 frames ) with a minimum length of 2 s for the track segment . The resulting dynROI was built has a rounded tube center with a fringe of 9 microns around the segment formed by the two brightest objects present during the complete sequence . Dynamic region of interest for the chromosomes The kinetochores marking the center of chromosomes were detected using the multiscale detector with the default p value ( set to 0 . 005 ) and scales ranging from 0 . 15 to 0 . 25 voxels . Motion was modeled with a Brownian motion model with a maximum instanta - neous displacement set to 5 times the process noise estimated by Kalman \ufb01ltering of the trajectory , a lower bound set at 0 . 4 m m and upper bound set at 0 . 6 m m . Variation in SNR were managed with a maximum gap set to 4 s ( or frames ) with a minimum length of 2 s for the track segment . The dynROI was estimated using the generic point cloud tracking principles described above using all the trajec - tories detected inside the spindle dynROI with a box - shaped dynROI and a margin of 0 . 1 m m . Dynamic region of interest estimation for the interpolar region Let \u00f0 O st ; u st ; v st ; w st \u00de and \u00f0 O kt ; u kt ; v kt ; w kt \u00de denote the frames of reference estimated for the spindle and the chromosome respectively . We want to build a frame of reference \u00f0 O it ; u it ; v it ; w it \u00de that follows an interpolar plane showing how microtubule nucleation events inside the spindle are orchestrated to capture chromosomes ef\ufb01ciently . We \ufb01rst set the origin to O it = O st and w it = w st so that one axis is following the spindle at all time . For the plane to describe the motion of the chromosome population , the second unit vector follows a slice of the kinetochore - associated dynROI v kt 0 = cos \u00f0 q \u00de u kt + sin \u00f0 q \u00de w kt projected to ensure orthogonality as v it = v kt 0 : \u00f0 1 (cid:4) w it T w it \u00de . Finally the last unit vector is set as u it = v it 3 w it . The dynROI type is a plane with a lateral fringe of 50 voxels , a height of 4 voxels and an angle q set to p 2 . Dynamic region of interest estimation for the kinetochore \ufb01bers Assuming K - \ufb01bers to span the region between poles and kinetochores as a straight polymer , its associated microtubule dynamics was observed using a conical dynROIs with an angle of p 12 . Stochastic programming for the evaluation of trackability The association of particle detections with trajectory heads is performed in a temporally greedy fashion , i . e . , particles detected at time t are linked to the heads of track segments de\ufb01ned up to time t (cid:4) 1 without consideration of the track segments beyond t and only indirect consideration of track segment before t - 1 . Therefore , our de\ufb01nition of trackability relates to the level of ambiguity in assigning particles detected in time point t to track segment heads in t \u2013 1 . The optimal association is obtained by linear assignment of heads to particles in a bipartite graph : argmin f a ij g X i \u02db U ; j \u02db D t c ij a ij s : t : X i \u02db U a ij = 1 and X j \u02db D t a ij = 1 ; where U is the set of track segment heads , D t is the set of detections measured at time t , a ij \u02db f 0 ; 1 g denotes the assignement of the i th track segment to the j th particle and c ij \u02db R is the cost associated to making that association . The association cost c ij typically re\ufb02ects the distance between the predicted location of the i th track segment at t and the j th detection at this same time point . This assign - ment problem is convex , hence with a guaranteed unique solution , and can be solved using a variety of linear programming algo - rithms . 51 , 74 , 75 However , a key challenge in our framework is the deterministic aspect of this solution . There is no measure of uncer - tainty attributed to the \ufb01nal graph of associations ( see Figure 5A ) . While several algorithms have been proposed to estimate the uncertainty related to the total optimal cost of a linear programming problem , a . k . a . stochastic programming , 76 they do not focus on the detection of local changes in association made in the bipartite graph . In this Section , we will \ufb01rst detail how we consider the randomness present in the history of each track to estimate the probability distribution associated to all assignment costs c ij . We will then describe how these uncertainties can then be exploited to detect local ambiguities in the assignment problem , which subsequently de\ufb01ne a score of trackability . Stochastic \ufb01ltering approaches are routinely used to estimate the parameter describing the dynamic properties of tracked particles from their position history . They enable the prediction of particle location from one frame to the next to re\ufb01ne the cost used for linear assignment . Those temporally recursive algorithms also provide inferences of track segment prediction uncertainty from t - 1 to t . Brie\ufb02y , let x t be a variable describing the state of the track segment . For a particle moving in a directed fashion , it is de\ufb01ned as : x t = \u00f0 x ; y ; z ; dx ; dy ; dz \u00de : e4 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS The associated probability p \u00f0 x t j z 1 : t \u00de can be estimated recursively thanks to the Bayes rule : p \u00f0 x t j z 1 : t \u00de f p \u00f0 z t j x t \u00de Z p \u00f0 x t j x t (cid:4) 1 \u00de p \u00f0 x t (cid:4) 1 j z 1 : t (cid:4) 1 \u00de d x t (cid:4) 1 ; where z 1 : t represents the past measured positions assigned to a particular track . Kalman \ufb01ltering is a scalable and \ufb02exible way to model the motions of thousands of particles in parallel , and as such is used in the majority of tracking approaches , 7 including u - track . In this framework , the relationships between random variables are assumed to be linear and described as follows : x t = Fx t (cid:4) 1 + w t z t = Hx t + v t where F is the state transition matrix between consecutive time points , H is the observation matrix , and w t and v t are the model and measurement noise respectively , bothassumed to be Gaussian with covariance matrices Q t and R t . The Gaussian andlinear assump - tion provides an analytical solution with a computationally ef\ufb01cient implementation to estimate p \u00f0 x t j z 1 : t \u00de(cid:1) N \u00f0 b x t ; b P t \u00de ( see our previous work 13 for a detailed review ) . Before optimal assignment between a track segment at t (cid:4) 1 and the object detected on frame t , the probability distribution of the predicted particle positon at time t is then described by p \u00f0 x t j z 1 : t (cid:4) 1 \u00de(cid:1) N \u00f0 F b x t (cid:4) 1 ; F b P t (cid:4) 1 F T + Q t \u00de . As such the variation of the cost to associate the i th track segment to the j th measurement can be expressed , without loss of generality as : c ij (cid:1) k Hx (cid:4) z jt k s : t : x (cid:1) p (cid:2) x it (cid:4)(cid:4) z i 1 : t (cid:4) 1 (cid:3) : This expression provides us with a direct way to explore the space of possible combination of cost values through Monte Carlo simulations . u - track 3D implements several types of stochastic \ufb01ltering approaches such as unimodal and multimodal Kalman \ufb01ltering as well as piecewise stationary motion \ufb01ltering or smoothing approaches , where the same principles can be straightforwardly applied . The principle underlying the use of our predicted probability distribution to evaluate assignment stability is described graphically in Figure 6 . Our local trackability score is de\ufb01ned as : T it = 1 N X N n = 1 (cid:5) a ij (cid:5) = a ij n (cid:6) where a ij (cid:5) is the initial assignment found for the i th trajectory , a ij n is a newly computed assignment resulting from the n th out of a total of N simultaneous resampling rounds of all costs c ij and [ . ] denotes the Iverson bracket . Each new assignment result , or vote , is considered different if the track segment is assigned to another detection , or determined to be a track termination . As such , a lower score T it re\ufb02ect a larger instability in the optimal assignment , hence a higher ambiguity and lower trackability . In our experiments , the number of resampling rounds is set to N = 20 . Clathrin - mediated endocytosis study on a glass coverslip Cell preparation and imaging Inner medulla collecting duct ( IMCD ) mouse epithelial cells were plated on 5 mm diameter coverslips ( 64 \u2013 0700 , Thomas Scienti\ufb01c ) and mounted to a custom machined holder for imaging with a high - NA version of diagonally scanned light - sheet microscopy . 2 This microscope is equipped with an NA 0 . 71 water dipping illumination objective ( 54 - 10 - 7 , Special Optics ) , and a 25X / NA 1 . 1 water dip - ping detection objective ( CFI75 Apo LWD 25XY , Nikon Instruments ) , and a Hamamatsu Flash 4 . 0 sCMOS camera . Brie\ufb02y , 500 time points were acquired with 30 m W of 488 nm illumination ( measured at the back pupil of the illumination objective ) and a 15 ms camera exposure . Each image stack was 106 . 5 3 39 . 9 3 23 . 1 m m , with a lateral and axial voxel size of 104 and 350 nm , respectively , resulting in a 1 . 008 Hz volumetric image acquisition rate . Clathrin structure trajectory estimation and post - processing Clathrin structure aggregates , labelled by alpha - adaptin GFP were detected using a multiscale particle detector with a p value set to 0 . 05 and scales ranging from 0 . 15 to 0 . 5 m m . For tracking , the motion of particles was modeled with a Brownian motion model with a maximum instantaneous displacement set as three times the process noise estimated by Kalman \ufb01ltering of the trajectory , a lower bound set at 0 . 1 m m and upper bound set at 0 . 3 m m . When detection gaps are enabled , the maximum gap length is set to 3 s ( or frames ) with a minimum length of 3 s for any track segment allowable to be connected by the gap closing algorithm . 10 The median of the maximum intensity reached per track was then used to discriminate between abortive and maturing CCPs . To account for the variation of \ufb02uorescence signal across acquisitions , the maximum intensities were scaled such that the empirical cu - mulative distribution function ( cdf ) of maximum intensities computed for each acquisition matched the median cdf of all acquisitions , as previously described in ref . 53 . Cell Reports Methods 3 , 100655 , December 18 , 2023 e5 Article ll OPEN ACCESS Microtubule instability measurement Cell preparation and imaging HeLa cells stably expressing EB3 - EGFP alone or along with mCherry - CENPA were plated on coverslips mounted on the microscope in CO2 - independent L15 medium containing 10 % FBS , without phenol red , and maintained at 37 (cid:3) C for the duration of the experi - ment . All images were recorded with an Orca Flash 4 . 0 v2 sCMOS camera ( C11440 - 22C ; Hamamatsu ) . Images were acquired in sample - scan imaging mode with a lateral translation of 0 . 4 m m and subsequently deskewed in postprocessing . Final voxel dimen - sions for all lattice light - sheet image datasets were 104 nm 3 104 nm 3 210 \u2013 217 nm . The microscope was controlled by custom - made software . Acquisition frequency was adapted for each experiment and is speci\ufb01ed in the main text . Plus - ends trajectory estimation Plus - ends , labelled through GFP tagging of EB1 , were detected using a multiscale detector with the default p value ( set to 0 . 005 ) and scales ranging from 0 . 15 m m to 0 . 25 m m . The polymerization of microtubule was modeled with a directed displacement estimated through a Kalman \ufb01ltering of the trajectory , similar to , 55 but now in 3D . The random component of this displacement was estimated as 3 times the process noise of the Kalman \ufb01lter with a lower bound of 0 . 3 m m and an upper bound of 0 . 6 m m . The shrinkages and pauses detection framework proposed in 55 has also been translated to 3D plus - ends trajectories . The detec - tion of both shrinkages and pauses is carried out by closing gaps between track segments , which implicitly delineate phases of microtubule growth . In our experiment , the minimum growth duration to consider gap closing was set to 4 s ( or frames ) and the maximum gap duration was set to 8 s . In order to detect pauses , the maximum angle between a speed vector estimated immediately prior and posterior to the pause event was set to 30 (cid:3) , the maximum positional \ufb02uctuation in the plus - ends location during a pause is set to 0 . 5 m m . To detect a shrinkage event between two segments , we \ufb01rst measure the distance D between the termination point of the earlier segment ( which is equivalent to the potential locus of a catastrophe event ) and the initiation point of the later segment ( which is equivalent to the potential locus of a rescue point ) along the path of the earlier segment . The two segments are connected by the gap closer if the distance between the initiation point to the closest point along the trajectory of the \ufb01rst segment does not exceeds Dsin \u00f0 q \u00de with q set to 20 (cid:3) in our experiment . Single molecule dynamics study with lattice light - sheet microscopy Cell preparation and imaging Fluorescently labelled Sox2 transcription factors were imaged over 100 time points with Lattice light - sheet microscopy imaging . Nine planes spaced 500nm apart were acquired at 50 ms of camera exposure , resulting in a 2 Hz volumetric image acquisition rate . Each image stack was 50 3 50 3 5 m m , then cropped around the nucleus , with a lateral and axial voxel size of 100 and 500 nm , respectively . Estimation of transcription factor binding times Transcription factor single molecules were detected using a multiscale detector using a p value of 0 . 01 and a scale ranging from 0 . 15 to 0 . 3 m m . Transcription factor motion was modeled using a Brownian motion model with a maximum instantaneous displacement estimated as 6 times the process noise estimated by Kalman \ufb01ltering of trajectory to account for speed variations during long periods of con\ufb01ned diffusion , a lower bound set at 0 . 3 m m and upper bound set at 0 . 5 m m . The maximum gap is set to 4 s ( or frames ) with a minimum length of 2 s for any track segment allowable to be connected by the gap closing algorithm . We assume that if the single molecule is detectable it immobilized at the DNA . Accordingly , characteristic binding times t are estimated by a double exponential \ufb01t to the lifetime distribution . Adhesions and collagen interaction imaging and analysis Cell preparation and imaging U2OS Cells were lentivirally transduced with a truncated CMV promoter ( Addgene # 110718 ) driving the expression of mNeonGreen - Paxillin ( Allele Biotechnology ) . Cells were seeded into a pH - neutralized collagen solution ( (cid:1) 2 mg / mL ) that , when polymerized , fully embedded cells in a three - dimensional extracellular matrix environment . For visualization of the extracellular matrix , a small concen - tration of the collagen was \ufb02uorescently conjugated with Alexa Fluor 568 NHS Ester ( A20003 , ThermoFisher ) . Samples were imaged with a high - NA variant of Axially Swept Light - Sheet Microscopy using 488 nm and 561 nm lasers for illumination ( OBIS LX , Coherent , Inc . ) . The details of this microscope will be published elsewhere . Brie\ufb02y , lasers are combined , spatially \ufb01ltered , expanded , and shaped into a light - sheet with a cylindrical lens . This light - sheet was relayed to a bidirectional scan unit ( 6215 , Cambridge Technol - ogy ) , a remote focusing system ( CFI S Plan Fluor ELWD , Nikon Instruments ) , and eventually to the illumination objective ( 54 - 10 - 7 , Special Optics ) . Fluorescence was detected in a wide\ufb01eld format with a water - dipping objective ( CFI75 Apo LWD 25SW , Nikon In - struments ) and imaged onto two sCMOS cameras ( ORCA - Flash4 . 0 , Hamamatsu Photonics ) with a 500 mm achromatic doublet ( 49 \u2013 396 , Edmund Optics ) , laser line \ufb01lter , a dichroic , and bandpass \ufb01lters ( ZET405 / 488 / 561 / 640 , ZT568rdc , ET525 / 50m , and ET600 / 50m , Chroma Technology Corporation ) . The laser laterally dithered for shadow reduction and scanned synchronously with the detection objective ( P - 603 . 1S2 and E (cid:4) 709 . SRG , Physik Instrumente ) to acquire a three - dimensional stack of images . All equipment was controlled with custom LabVIEW software , which is available from UTSW upon completion of a material transfer agreement . Adhesion detection and elongation analysis Paxillin aggregates as a surrogate for adhesions were detected using the multiscale detector described in Section \u2018\u2018Multiscale par - ticle detector\u2019\u2019 based on a p value of 0 . 001 and a scale ranging from 0 . 3 to 0 . 5 m m . The elongation of each detected adhesion is computed through a tubularity metric evaluated for each voxel and averaged across all the voxels associated to a single adhesion . e6 Cell Reports Methods 3 , 100655 , December 18 , 2023 Article ll OPEN ACCESS Similar to the classic vesselness estimator by Frangi and colleagues , 77 our tubularity metric is based on the eigen values of the Hes - sian matrix to describe local curvature , . Let \u00f0 l 1 < l 2 < l 3 \u00de be the three eigenvalues of the Hessian matrix computed at each voxel , the tubularity metric T = 1 (cid:4)j l 1 = l 2 j yields a value between 0 and 1 increasing with the elongation of the adhesions . As such , a noteworthy difference between the classic score described in Frangi\u2019s approach is the use of the two lowest eigen - values ( associated with the two axis of lowest curvature direction ) to discriminate between \ufb02at and elongated adhesions . Collagen detection and distance analysis Collagen was detected using the 3D implementation of steerable \ufb01ltering as described in . 78 An adhesion was considered in close contact with collagen if the center of at least one voxel belonging to its mask was less than 100 nm away from any voxels of the collagen mask . Endosome trackability on cell cultured on top of collagen Cell preparation and imaging Sum159O breast cancer cells 69 stably expressing alpha - adaptin GFP were imaged similarly to the one platted on glass coverslip , with the exception that they were plated on a (cid:1) 2 mm thick bed of rat tail - derived Collagen Type I ( 354236 , Corning ) . Clathrin structure trajectory estimation Clathrin structure aggregates , labelled by alpha - adaptin GFP were detected using a multiscale particle detector with a p value set to 0 . 01 and scales ranging from 0 . 125 to 0 . 5 m m . For tracking , the motion of particles was modeled with a Brownian motion model . In order to follow the erratic displacements caused by large protrusive motions , the maximum instantaneous displacement was set to 5 times the process noise estimated by Kalman \ufb01ltering of the trajectory , a lower bound set at 0 . 3 m m and upper bound set at 0 . 6 m m . The maximum gap length is set to 3 s ( or frames ) with a minimum length of 3 s for any track segment allowable to be connected by the gap closing algorithm . 10 Benchmarking computational time and memory usage Computational time and memory consumption have been tested on a computing workstation and of\ufb01ce laptop ( see speci\ufb01cation below ) . The test includes detection , tracking with trackability and dynROI de\ufb01nition on the endocytosis test dataset ( 400 Mb ) pro - vided along with this manuscript . Runtime takes an average of 17 . 3 s ( s . d . 0 . 2 s ) seconds on the workstation and 2 min 29 s ( s . d . 6 s ) on the laptop . The maximum memory used by the pipeline was 12Gb on the workstation and 0 . 5 Gb on the of\ufb01ce laptop . Of\ufb01ce laptop . d Model : X1 Carbon 6 th gen d Year : 2018 d CPU : Intel i7 - 8550U @ 1 . 80GHz , 4 Cores d RAM : 16 Gb d GPU : integrated d Disk : 1 Tb d OS : Windows 11 d MATLAB version : 2023a Computing Workstation . d Model : Colfax SX6300 d Year : 2020 d CPU : Intel Xeon 6242R @ 3 . 10GHz , 80 cores d GPU : Nvidia A6000 d Disk : 16 Tb d OS : Ubuntu 20 . 04 . 3 LTS d MATLAB version : 2021a QUANTIFICATION AND STATISTICAL ANALYSIS Statistical analysis was performed by considering either intracellular processes or cells as independent observations as speci\ufb01ed in the manuscript . Each statistic considering intracellular processes ( either experimental or simulated data ) includes more than a thou - sand events . Statistical details of all experiments at the cellular level can be found in the \ufb01gure legends including exact value of n , clear descriptions of what n represents and box - plot representations . Cell Reports Methods 3 , 100655 , December 18 , 2023 e7 Article ll OPEN ACCESS",
    "gilonAnalogyMiningSpecific2018": "Analogy Mining for Speci\ufb01c Design Needs Karni Gilon The Hebrew University of Jerusalem Jerusalem , Israel karni . gilon @ mail . huji . ac . il Joel Chan Carnegie Mellon University Pittsburgh , PA , United States joelchuc @ cs . cmu . edu Felicia Y Ng Carnegie Mellon University Pittsburgh , PA , United States fng @ cs . cmu . edu Hila Lifshitz - Assaf New York University New York , NY , USA h @ nyu . edu Aniket Kittur Carnegie Mellon University Pittsburgh , PA , United States nkittur @ cs . cmu . edu Dafna Shahaf The Hebrew University of Jerusalem Jerusalem , Israel dshahaf @ cs . huji . ac . il ABSTRACT Finding analogical inspirations in distant domains is a pow - erful way of solving problems . However , as the number of inspirations that could be matched and the dimensions on which that matching could occur grow , it becomes challenging for designers to \ufb01nd inspirations relevant to their needs . Fur - thermore , designers are often interested in exploring speci\ufb01c aspects of a product \u2013 for example , one designer might be inter - ested in improving the brewing capability of an outdoor coffee maker , while another might wish to optimize for portability . In this paper we introduce a novel system for targeting analogical search for speci\ufb01c needs . Speci\ufb01cally , we contribute an ana - logical search engine for expressing and abstracting speci\ufb01c design needs that returns more distant yet relevant inspirations than alternate approaches . Author Keywords Computational analogy , innovation , inspiration , creativity , product dimensions , abstraction , focus , text embedding ACM Classi\ufb01cation Keywords H . 5 . 3 Group and Organization Interfaces INTRODUCTION Analogy is a powerful strategy for designing new innovations . Thomas Edison invented the kinetoscope ( the precursor to motion picture projectors that are used in theaters today ) by working out how to do \u201cfor the eye what the phonograph does for the ear\u201d [ 22 ] . The Wright brothers solved a crucial aspect of how to keep their invented aircraft stable during \ufb02ight by analogy to maintaining balance while riding a bicycle [ 1 ] . More recently , a car mechanic created an innovative new way Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for pro\ufb01t or commercial advantage and that copies bear this notice and the full citation on the \ufb01rst page . Copyrights for components of this work owned by others than ACM mustbehonored . Abstractingwithcreditispermitted . Tocopyotherwise , orrepublish , to post on servers or to redistribute to lists , requires prior speci\ufb01c permission and / or a fee . Request permissions from permissions @ acm . org . CHI 2018 , April 21 \u2013 26 , 2018 , Montreal , QC , Canada \u00a9 2018 ACM . ISBN 978 - 1 - 4503 - 5620 - 6 / 18 / 04 . . . $ 15 . 00 DOI : https : / / doi . org / 10 . 1145 / 3173574 . 3173695 to assist with dif\ufb01cult childbirth , drawing on an analogy to a party trick for removing a cork stuck in a wine bottle [ 26 ] . Search engines that could support automatic retrieval of rele - vant analogies for design problems could signi\ufb01cantly increase the rate of innovation and problem solving today . The rise of knowledge databases and repositories on the Internet ( e . g . , the US Patent Database , Google Scholar , Amazon products , etc . ) provides a virtual treasure trove of ideas that could inspire so - lutions across a variety of domains . Research on creativity and innovation suggests that building on analogous inspirations that are not from the same domain as the source problem is a powerful strategy for generating creative ideas [ 6 , 12 , 28 ] . However , \ufb01nding useful distant analogies in large databases of textual documents remains challenging for existing machine learning models of document similarity [ 7 , 4 , 20 , 23 ] , which are largely dependent on surface features like word overlap . An additional challenge is that in real - world contexts with complex problems , designers are often interested in exploring and abstracting speci\ufb01c aspects of a problem rather than con - sidering the problem as a whole . To illustrate , consider the example of the Wright brothers inventing an airplane . Instead of trying to \ufb01nd an analogy for the entire plane , they regularly found analogies for partial problems they needed to solve , such as steering the wings , or controlling balance during \ufb02ight [ 19 ] . For each identi\ufb01ed problem , they then needed to abstract key properties of the problem in order to \ufb01nd useful analogs in other domains . In the example of steering the wings , they needed to look beyond some aspects of the wings \u2013 such as the color or particular material \u2013 while keeping in mind other aspects \u2013 such as the semi - rigid frame and need for the wing material to remain taut on the frame . Doing so may have led them to avoid overly general abstractions of \u201csteering\u201d that ended up less useful ( such as the wings of a bird or the rudder of a ship ) and towards more targeted analogical inspirations including the twisting of a cardboard box which drove their \ufb01nal design of warping the wings to steer [ 1 ] . There are two critical parts of the above example : focusing and targeted abstraction . By focusing we mean identifying CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 1 a particular framing or relation for which we would like to \ufb01nd analogies ; here , steering the wings or stabilizing the plane during \ufb02ight . Importantly , analogies speci\ufb01c to one focus may not be relevant to another ; for example , the problem of keeping the aircraft stable in turbulent air led to a different analogy of riding a bike , suggesting that small unconscious adjustments of the driver could address shifts in air turbulence [ 14 ] . By targeted abstraction we mean choosing the key properties of objects that are important to the core problem ( e . g . , semi - rigid , thin and \ufb02at ) while dropping out other less important properties ( e . g . , color , size ) . For example , in the steering wings problem , the desired abstraction might be something like \u201csteer < something that is semi - rigid , thin , and \ufb02at > \u201d . Targeting the abstraction is necessary in order to avoid \ufb01nding too many irrelevant matches ; for example , previous work has shown that abstracting all domain - speci\ufb01c features of the core relational structure of a problem yields less relevant analogies than retaining some [ 29 ] . Many real world solutions similarly require multiple problem framings that would bene\ufb01t from focusing and targeted ab - straction ; for example , a coffee maker taken camping may bene\ufb01t from distant inspirations that make it more lightweight , resistant to weather , a better grinder , or allow the camper to know when to pull it off the \ufb01re . Together , focus and targeted abstraction make it possible to \ufb01nd inspirations that are ana - logically relevant to very speci\ufb01c design needs , without being restricted to inspirations from the same / similar domains . To address the challenge of targeting analogical search for speci\ufb01c design needs , we present a system in which a de - signer can specify a focus for a given product description , and then abstract that focus beyond its surface features in a tar - geted manner by specifying the key properties of the relations and entities involved that are crucial for understanding the core relational structure . To facilitate expressing this query in a machine - readable way , we leverage a large database of commonsense knowledge ( Cyc ) to provide a set of controlled terms that humans can use to express key properties . Our sys - tem then uses this focus - abstracted query to computationally search a corpus of potential inspirations for analogically rele - vant matches tuned to the designer\u2019s speci\ufb01c design need . We compare this process to previous state - of - the - art approaches for \ufb01nding analogies among product descriptions , and \ufb01nd that using our Focus - Abstracted queries returns inspirations that are high on both relevance ( the results meet the needs of the query ) and domain distance ( the results are from different domains ) ; in contrast , state of the art approaches that operate on the whole document or only on speci\ufb01c keywords , either sacri\ufb01ce relevance or distance . These results have promising implications for creativity - support tools that aim to support designers in solving complex problems through analogy . RELATED WORK Computational Analogy The problem of computational analogy has a long history in arti\ufb01cial intelligence research . Early work focused on devising algorithms for reasoning about analogies between manually created knowledge representations that were rich in relational structure ( e . g . , predicate calculus representations ) [ 8 , 10 ] . While these algorithms achieved impressive human - like accu - racy for analogical reasoning , their reliance on well - crafted representations critically limited their applicability to mining analogies amongst databases of free - text documents . At the same time , much work in machine learning and infor - mation retrieval has devised methods for \ufb01nding documents that are relevant to some query from a user . These methods do not focus on analogy in particular ( and certainly not on far analogies ) : while they differ in the speci\ufb01cs of their methods ( e . g . , using singular value decomposition , or , more recently , neural networks ) , in general , they attempt to learn semantic representations of words based on the way that words are sta - tistically distributed across word contexts in a large corpus of documents ; notable examples include vector - space models like Latent Semantic Indexing [ 7 ] , probabilistic topic model - ing approaches like Latent Dirichlet Allocation [ 4 ] , and word embedding models like Word2Vec [ 20 ] and GloVe [ 23 ] . The semantic representations produced by these methods are quite useful for \ufb01nding very speci\ufb01cally relevant documents / results for a query , but are limited in their ability to \ufb01nd matches that are analogically related to a query ( especially if they do not share domain - speci\ufb01c keywords ) . Recent work by Hope et al [ 15 ] proposes to \ufb01nd analogies among free - text consumer product descriptions by learning to predict an overall representation of a product\u2019s purpose ( what it is good for ) and mechanism ( how it works ) . It uses annotators to mark words related to the purpose / mechanism of the product , and weighs the Glove [ 23 ] values of those words to assign an overall purpose / mechanism representation for each document . It then uses an arti\ufb01cial neural network model ( speci\ufb01cally a bidirectional recurrent neural network , or RNN [ 2 ] ) to learn the mapping between the product de - scription\u2019s word sequence in GloVe representation and the overall - purpose / mechanism representation captures by the pur - pose / mechanism annotations . Hope et al showed that they could use these representations to \ufb01nd analogies at a signi\ufb01 - cantly higher rate than comparison state - of - the - art approaches like TF - IDF - weighted GloVe vectors of the documents . While promising , as noted above , this approach is designed to \ufb01nd analogies for the overall purpose of a given product , and may therefore miss analogies for speci\ufb01c aspects of the product ( the speci\ufb01c focus of our paper ) . Abstraction during Analogy - Finding The essence of analogy is matching a seed document with other documents that share its core relational structure [ 10 ] ; when the analogous documents also have many other details that are very different from the seed document , they are known as far or domain - distant analogies . To \ufb01nd far analogies , it is essential to abstract away these irrelevant details from one\u2019s representation of the seed document ( and possibly also other documents in the corpus ) . Some research has explored how to enable people to construct problem representations that abstract away the surface details of the problem . For example , the WordTree method [ 18 ] has people use the WordNet [ 21 ] lexical ontology to systematically CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 2 \u201cwalk up\u201d levels of abstraction for describing the core desired functionality , leading to the possibility of discovering anal - ogous functions in other domains . For example , a designer who wanted to invent a device to fold laundry for students with very limited \ufb01ne motor skills might abstract the core function of \u201cfolding\u201d to \u201cchange surface\u201d , which could lead to analogous inspirations like \u201cree\ufb01ng\u201d ( rolling up a portion of a sail in order to reduce its area ) . Yu et al [ 27 ] explored how to systematically train crowd workers to convert problem descriptions into an abstracted form that ignored irrelevant surface details . Importantly , these abstraction methods do not blindly abstract all aspects of a problem description . In many cases , humans exert their judgment to select appropriate levels of abstraction , and also do extensive screening of possible matches based on whether they overlap with key properties / constraints in the original problem context . This is important because analogies that are \u201ctoo far\u201d away can actually lead to less creative ideas [ 5 , 9 , 13 ] . Yu et al [ 29 ] recently showed that describing the problem context in abstract terms , but retaining a domain - speci\ufb01c description of its key constraints , enabled crowd work - ers to \ufb01nd more useful far inspirations than a representation that is abstract on both the problem context and its constraints : for example , the description \u201cmake an object ( abstracted prob - lem context ) that does not tip over easily ( concrete constraint ) \u201d yields more useful inspirations for the problem of making a safe chair for kids , compared to \u201cmake an object ( abstracted problem context ) that is safe ( abstracted constraint ) \u201d ( which yields inspirations for safety that cannot be applied to chairs ) . The insight behind this recent innovation is that abstraction should be targeted : rather than completely abstracting away from all the properties of the objects involved in the core rela - tional structure ( e . g . , the wings in the steering problem for the Wright brothers ) , it is critical to retain the key properties of the objects that are important for the core relational structure . For example , in order to \ufb01nd inspirations that can suggest useful mechanisms for angling wings to steer a plane ( e . g . , twisting of a cardboard box ) , designers need to express to a search en - gine that they don\u2019t care about the color and size of wings , but they do care the fact that the wings are \ufb02at , physical objects , or even that they are composed of materials that respond to shear forces in a similar way to cardboard . This insight is consistent with classic cognitive models of analogy ( cited above , e . g . , [ 8 ] ) , which retain key properties of objects during analogical mapping that are essential to the core relational structure : for example , in the atom / solar - system analogy , the absolute size of the sun / planets vs . nucleus / electron doesn\u2019t matter , but the fact that they have mass does . We build on these insights to explore how we might create a focus - abstracted representation that enables a computational semantic model to \ufb01nd more relevant and distant inspirations . DATA SET We use a corpus of product descriptions from [ 15 ] . The prod - ucts in the corpus are from Quirky . com , an online crowd - sourced product innovation website . Quirky is useful for our study because it is large ( the corpus includes 8500 products ) and covers multiple domains , making cross - domain analogies Figure 1 . An example product description from our dataset : Soapy slider possible . Quirky users submit their ideas in free , unstruc - tured ( and often colloquial ) natural language . The ideas are written by non - experts and contain non - technical terms for which abstractions are easy to \ufb01nd . Figure 1 shows an example submission ( \u201cSoapy slider\u201d ) , demonstrating typical language . METHOD When approaching a product different designers may wish to focus on different parts of it . For example , consider the \u201cSoapy slider\u201d product in Figure 1 . One designer ( designer A ) may wish to explore ways of adjusting to different soap bar sizes while another ( designer B ) may be interested in removing soapy water from soap bars . To satisfy their needs , a straightforward approach is for the designers to search the corpus for keyword matches , for example \u201cchange soap size\u201d for one designer , and \u201cremove soap water\u201d for another . We propose a process that enables the designer to focus on a key need , and then abstract the description to include only the properties of that need that are actually important . For example , designer A ( originally interested in changing the size of soap bars ) can indicate that the \u201csoap\u201d domain is not important and can be replaced by the more general property of being a \u201cpersonal product\u201d . The system then \ufb01nds matches in the corpus using a method based on the state - of - the - art purpose representation engine from [ 15 ] . The matches will be analogous products that adjust to different sizes of some personal product , which could hopefully inspire the designer . From a system standpoint , the process can be divided into two phases : 1 ) expressing a focus - abstracted query , and 2 ) using the query to \ufb01nd analogies in a corpus . Below we describe our process in more detail for each phase . Expressing the focus - abstracted query Figure 2 shows a worked example of the overall process . We describe each step in turn . Step 1 : Focus on important sentences We assume the designer begins with an existing complete prod - uct description . Figure 2 shows one such example from the Quirky corpus ( the \u201cSoapy slider\u201d product ) . The designer se - lects the sentences most relevant to their need , thus identifying which aspect in the product they wish to further explore . In the \u201cSoapy slider\u201d example ( Figure 1 ) , designer A ( focusing on product size adjustments ) will choose the sentence \u201cextend - able for different sizes of soap bars\u201d ( see Step 1 in Figure 2 ) . CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 3 Figure 2 . Illustration of the process of expressing focus - abstracted queries in our system . Designers express a focus - abstracted query by ( 1 ) selecting important sentences in a product description that relate to an intended focus , ( 2 ) ignoring irrelevant terms , and ( 3 ) replacing im - portant terms ( where desired ) with appropriate abstracted properties , yielding a sequence of important terms and abstracted properties . Designer B ( interested in removing liquid from things ) will choose the sentence \u201cit removes soapy water away from the bar of soap keeping it dryer to last longer\u201d . Step 2 : Focus on important terms Important sentences from Step 1 may still contain terms or domain details that are irrelevant to the intended purpose of the designer . Ignoring irrelevant terms increases the focus of the query on the intended purpose of the designer . It also achieves a form of abstraction ( e . g . , ignoring irrelevant domain details ) . To achieve this function , the interface allows the designer to take any noun , verb , or adjective from the important sentences and mark them with an \u201cIGNORE\u201d \ufb02ag if they are irrelevant to her speci\ufb01c purpose . For example , designer A ( who is not interested in bars speci\ufb01cally ) , might choose to IGNORE the term \u201cbars\u201d ( see Step 2 in Figure 2 ) . Designer B ( who is interested speci\ufb01cally in how to remove water from a bar of soap ) may IGNORE the term \u201clast\u201d , which describes the ultimate purpose of keeping the bar of soap dry , but may not be shared by other products that also separate water from objects . Step 3 : Abstract important terms After Step 2 , the designer\u2019s needs are still expressed in the original domain ( e . g . , soap ) . In order to \ufb01nd domain - distant analogies it is necessary to replace key terms with their appro - priate abstractions . The designer can abstract a term by clicking on it and selecting the appropriate abstractions from a list . The list is grouped into semantic sets from which the designer chooses the most relevant one . The most useful set is often obvious ( e . g . , Soap - Personal is more relevant than SoapOpera ) , making it easier to narrow the list down . For example , designer A might not be interested in the fact that soap is a ToiletrySubstance , but rather that it is more generally a PersonalProduct , and select that property to abstract the term \u201csoap\u201d ( see Step 3 in Figure 2 ) . In designing this component of our system , we faced and dealt with several design challenges : \u2022 Choosing an appropriate knowledge base . To \ufb01nd abstrac - tions to show the designers , we explored several knowledge bases . WordNet [ 21 ] is a large English lexical database , including relations like synonym , hypernym and hyponym . WordNet is lexical and not focused on facts about the world , rendering it less useful for our purpose . In addition , the number of relations it supports is very small . Another alter - native we considered is ConceptNet [ 25 ] . ConceptNet in - cludes knowledge from crowdsourcing and other resources , rendering it very noisy . We ended up choosing Cyc [ 16 , 17 ] as our main Knowl - edge Base . Cyc is a very large , logic - based knowledge base representing commonsense knowledge . Cyc contains over \ufb01ve hundred thousand terms , seventeen thousand types of relations , and over seven million assertions , i . e . , sim - ple facts about the world . Examples of assertions : \u201c # $ isa # $ DonaldTrump # $ UnitedStatesPresident\u201d ( Donald Trump is the US president ) and \u201c # $ genls # $ Water # $ LiquidTangi - bleThing\u201d ( liquid is a generalization of water ) . If a term is missing from Cyc , we resort to WordNet . Crucially for our purposes , Cyc contains supersets of terms , which are useful for abstraction . For example , \u201cDomes - ticatedAnimal\u201d and \u201cCanisGenus\u201d are supersets of \u201cdog\u201d . \u201csoap\u201d may abstract to its supersets \u201cToiletrySubstance\u201d , \u201cWaterSolubleStuff\u201d , \u201cPersonalProduct\u201d , and many others . Another way of looking at it is that soap has the properties of being a water soluble toiletry substance and a personal product . Thus we use the terms Abstractions and Properties interchangeably . The level of abstraction controls the distance from the do - main , thus allowing the designers to choose far or near analogies . Importantly , the abstractions also give designers control over the direction of the abstraction ( e . g . , ignore all things that are about cleaning , but make sure they share the property of being personal products , or water soluble ) . \u2022 Dealing with natural language descriptions . Quirky prod - uct descriptions are written in unstructured natural language . To obtain and display appropriate abstractions for the de - signers to select from , we \ufb01rst preprocess the corpus and perform part of speech ( POS ) tagging . We then apply NLTK WordNet Morphy [ 3 ] to get the canonical form of each term ( according to its POS ) , and use this form to query Cyc for its associated properties . For example , we change \u201csizes\u201d to \u201csize\u201d before KB lookup . \u2022 Presenting a manageable set of abstractions to choose from . A \ufb01nal design consideration here is the number of abstractions to present to the designer for consideration , CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 4 Figure 3 . Illustration of the process of abstracting the corpus based on a given focus - abstracted query . All terms in each document that match the designer - selected abstracted properties ( shown in monospace font ) are replaced by the matching properties . This brings documents that might be different in domain ( e . g . , about \u201c\u2019knives\u201d ) but are nevertheless similar at the desired level of abstraction ( e . g . , PersonalProduct ) closer to the focus - abstracted query . since words in Cyc are often associated with many potential abstractions ; for example , the term \u201cdog\u201d has over 100 dif - ferent abstractions . To limit the number , we experimented with \ufb01ltering by level of abstraction ; we found that three abstraction levels appeared to be an optimal cutoff , above which the terms become to general and uninteresting . For example , \u201csoap\u201d may be abstracted to \u201cthing\u201d , which is common to all objects and therefore provides little infor - mation and can be replaced by the IGNORE option . The abstractions are sorted from speci\ufb01c to general . We considered sorting within the abstraction levels by mea - suring property prevalence . If a vast number of items share a certain property then it is probably too general ( e . g . , \u201cPhys - ical Entity\u201d ) , and will not be useful . If there are too few items , then maybe the property is too speci\ufb01c and less in - teresting ( e . g . \u201cGoldColor\u201d ) . However , since the relevant abstractions typically appear among the \ufb01rst \ufb01ve to ten abstractions we decided not to implement further ordering . Once the designer selects appropriate abstractions , the expres - sion phase is complete : the end result is a focus - abstracted query derived from the designer\u2019s operations on the origi - nal product description : unchecked sentences are omitted , words in the IGNORE list are omitted , and the words that were abstracted are replaced by their abstractions . For exam - ple , designer A\u2019s selections would yield the following focus - abstracted query : [ extendable , different , SpatialQuantity , Per - sonalProduct ] ( see Figure 2 ) . Finding analogies for the focus - abstracted query Now that the designer expressed their desired focus and ab - straction , we use our analogy engine to \ufb01nd analogies from a corpus of potential matches that are tuned for that particular focus ( while preserving abstraction ) . The most important step is to re - represent the corpus with the designer\u2019s chosen abstractions . Concretely , for each document , we \ufb01nd all terms in it that share the same abstracted properties ( in Cyc ) as those contained in the focus - abstracted query . For example , if the designer abstracted \u201csoap\u201d to \u201cPersonalProduct\u201d ( indicating that it is not the soap they care about , but rather being a personal product ) , the engine looks for other terms in the corpus which share the property of being \u201cPersonalProduct\u201d ( e . g . \u201cknife\u201d ) and abstracts them to \u201cPersonalProduct\u201d as well ( see Figure 3 ) . The goal of this abstraction step is to ensure that products from different domains that nevertheless share key relations or properties with the focus - abstracted query at the right level of abstraction can be seen as close to the query . Next , our engine \ufb01nds matching products in the abstracted corpus . We considered several options for the matching . We build on the work of [ 15 ] , which was shown to \ufb01nd good analo - gies on the same dataset . In short , [ 15 ] takes natural language product descriptions and uses deep learning ( speci\ufb01cally , a bidirectional RNN [ 2 ] ) to learn vector representations for pur - pose ( what is this product good for ? ) and mechanism ( how does it work ? ) . Given the vector representations , the algo - rithm of [ 15 ] \ufb01nds products with similar purpose but different mechanisms , that might serve as analogies . We use a similar algorithm for searching for products ; how - ever , in our case , since we focus on relevance for a speci\ufb01c abstracted need , we change the algorithm to focus only on \ufb01nding similar products with respect to purpose . We do this by computing a similarity score between the purpose represen - tation vector for the focus - abstracted query , and the purpose representations for all documents in the abstracted corpus , and selecting the 10 documents with the highest similarity score . In the case of designer A , who wanted to adjust to different soap bar sizes , the system suggested a knife rolodex , allowing storing knives of different sizes in the same holder ( Figure 3 ) . EVALUATION Our core hypothesis is that our system is able to \ufb01nd analogies for focused queries , while still retaining the ability to \ufb01nd analogies from distant domains . We evaluated this hypothesis across a range of focused query scenarios from seeds sampled from the larger corpus of prod - ucts . The general scenario is that of a designer looking for novel ways to redesign some speci\ufb01c aspect of an existing product . This is a common design task that requires creativity , CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 5 and may especially bene\ufb01t from distant analogies ( in order to maximize novelty ) , since the existence of many domain features may make it dif\ufb01cult for the designer to think of alternative mechanisms or domains . Preliminary Study We \ufb01rst conducted a preliminary usability study to assess the expressiveness of the interface . Four product designers and one non - designer were asked to examine three product descrip - tions randomly selected from the corpus and identify aspects in the product to redesign or generalize ( e . g . , generalize \u201cwater\u201d to \u201cliquid\u201d ) , and then to subsequently express those aspects using our interface . Prior to starting the task the users were given several examples ( using seeds not from the user - study set ) and a short Interface training session . For each aspect the users initiated a new interface session , checked the rele - vant sentences , set to IGNORE the unimportant terms ( e . g . , \u201cblack\u201d or \u201cin the mall\u201d ) , and abstracted terms according to their needs . Users reported they were mostly able to express their needs and in general the users thought the tool was bene\ufb01cial and easy to use . One of the product designers remarked : \u201cFor my own innovation I did not perform such an analogy search . I wish search engines had this Interface\u201d . One interesting \ufb01nding was that the interface helped users identify additional potential aspects and abstractions they had not thought of , and \ufb01nd products in distant domains . For example , one user used \u201cSeparationEvent\u201d ( of liquid ) as an abstraction of \u201cRemoving\u201d , which he had not previously considered ; this abstraction led to a product using hydrophobic coating . Users also cited challenges they encountered in using the interface which included words missing from the description ( e . g . , the option to add \u201clightweight\u201d was not available ) , and bigrams ( e . g . , \u201ccoffee maker\u201d ) that were not abstracted as one unit . One designer suggested marking some words as \u201cnice to have\u201d . These limitations could be addressed by simple extensions of our system ( i . e . , incorporating user - supplied keywords and looking up n - grams in the KB ) . For further discussion see Future Work . Search Scenarios After our usability study , we turned to evaluate the results of our engine . We randomly selected 5 seed products from the corpus , using the following screening criteria : \u2022 Is understandable ( e . g . , grammatical errors and typos do not overly obscure meaning , can actually visualize what it is ) \u2022 Has at least two well - speci\ufb01ed functions / aspects ( so we can de\ufb01ne distinct focus query scenarios ) \u2022 Judged to have corresponding inspirations in the corpus Two members of our research team then identi\ufb01ed two re - design scenarios for each seed . For example , in the \u201cSoapy slider\u201d product ( which was one of the selected seeds ) , the two scenarios were \u201cmake the dish compatible with different sizes of soap bars\u201d , and \u201ckeep excess soapy water away from the Seed product Scenario 1 Scenario 2 Soapy slider . Unique 2 piece horizontal soap dish with a slide that keeps excess soapy water away from the bar of soap . Make the dish com - patible with differ - ent sizes of soap bars Keep excess water away from the bar of soap Camp brew coffee maker . Light weight all in one cof - fee grinder and maker for camping and hiking . Tell when some - thing is done cooking Make food and drink outdoors Laundry folding table . Table that folds down out of the laundry room wall and provides a surface for folding laundry Make compact a pile of \ufb02exible , fold - able garments Make compact a piece of furniture On / off velcro pocket shoe . Attached / detached pocket for any shoe . Attach a small pocket to shoe / ankle comfortably & durably Make the attached pocket inconspicu - ous The restsack . Backpack that doubles as an outdoor chair stool . Carry items on the go Provide a portable seat Table 1 . Seed product descriptions and associated redesign scenarios used for our evaluation experiment . Descriptions shortened . bar of soap\u201d . We therefore have a total of 10 search scenarios ( see Table 1 ) . Constructing the queries The research team members who made the scenarios then used our system to create focus - abstracted queries for each of the scenarios . Figure 2 includes one example scenario and focus - abstracted query . Another example ( for the \u201cSoapy slider\u201d example ) is RemovingSomething , LiquidTangibleThing , SolidTangibleThing for the scenario need \u201ckeep excess soapy water away from the bar of soap\u201d . Measures Our goal is to \ufb01nd relevant analogies for a speci\ufb01c aspect of a seed product without being constrained to the same domain as the seed product ( i . e . , retaining the ability to \ufb01nd domain distant yet relevant analogies for a focused need ) . Therefore , we evaluate the relevance and domain distance of each match for its target query . Both measures were obtained by human judgments of the matches . All ratings were performed blind to the method that produced the match : that is , for each scenario , shared matches were combined across the methods , and the method that produced the match was not shown . Relevance We operationalized relevance as the degree to which the match meets the needs expressed in the query . Judgment of relevance took into account three factors : 1 ) the degree to which it shared the key purpose ( s ) expressed in the query ( e . g . , make compact , adjust ) , 2 ) the degree the objects related to the purpose shared the key properties of the key objects in the query ( e . g . , physical size of soap bars ) , and 3 ) the degree to which the purpose ( and associated mechanism ) was explicitly stated ( since some products state a function as a desirable property of the product , rather than a function it aims to achieve ) . This last factor is included because it is easier for a people to notice and CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 6 use analogies if the mapping to their problem is explicitly described / highlighted [ 24 ] . The judgment scale was a 5 - point Likert - like scale , with the following anchors ( developed after multiple rounds of training and piloting with a separate set of data ) : 1 = Matches none of the key functions and object properties in the query 2 = Implicitly matches a few of the key purpose ( s ) , but none of the key object properties in the query 3 = Implicitly or explicitly matches a few of the key pur - pose ( s ) AND a few of the key object properties in the query 4 = Implicitly or explicitly matches most of the key pur - pose ( s ) AND a most of the key object properties in the query 5 = Explicitly matches most / all of the key purpose ( s ) and key object properties in the query Two members of the research team ( who did not directly de - velop the system ) were trained to use the judgments on a separate dataset until they reached good inter - rater agreement , Cronbach\u2019s alpha = . 82 . They then each evaluated half of the matches independently . Examples of low and high relevance - scored matches are shown in Figure 4 . Domain distance We operationalized relevance as the degree to which the match shared domain features with the query\u2019s seed product . Note that this measure ignores the scenario and instead compares each match with the whole seed product . The judgment scale was also a 5 - point Likert - like scale , rang - ing from 1 ( very similar ) to 5 ( very different ) . One issue with this judgment is that many products could be thought of as being in a number of different domains : for example , the \u201ccamp brew coffee maker\u201d is about food / drink / coffee and camping / outdoors . Different raters might weight each feature differently as core / peripheral to the \u201cdomain\u201d of the prod - uct . For this reason , rather than utilizing single ratings from independent raters , each match received a rating from 2 in - dependent judges ( members of the research team ) , and the average of their ratings yielded the distance score for each match . The inter - rater reliability for this measure was good , Cronbach\u2019s alpha = . 79 ) . Examples of low and high distance - scored matches are shown in Figure 4 . Experiment Design and Hypotheses We compare our method against three other approaches : 1 . OverallPurpMech . This is the purpose - mechanism method from [ 15 ] . It estimates and matches on overall purpose and mechanism vectors . More speci\ufb01cally , we repli - cate the method from their evaluation experiment , which \ufb01nds matches for a given seed based on similarity of pur - pose , and aims to diversify by mechanism . We use this method to \ufb01nd a set of purpose - similar , mechanism - diverse matches for each scenario . The purpose of comparing our system to this method is to determine to what extent we have advanced the state - of - the - art in analogy - \ufb01nding . 2 . OverallGloVe baseline . [ 23 ] This method approximates the status quo for information - retrieval systems , which tend to operate on the whole document . Each document in the corpus is represented by the average of GloVe vectors for all words ( excluding stopwords ) . We use Glove pre - trained on the Common Crawl dataset ( 840B tokens , 300d vectors ) 1 . We then normalize each document vector , and calculate cosine similarity ( which is the same as Euclidean distance in this case ) between the resulting vectors for each seed and all other documents , and choose the 10 nearest as matches . 3 . FocusOnly baseline . This baseline helps tease apart the impact of focusing only versus focusing and abstracting ( with the knowledge base ) . For each scenario , we form a fo - cus query as a bag - of - words containing only the words that were abstracted during the process of making the focused - diverse query with our method , i . e . , the words themselves instead of their abstractions ( stopping at Step 2 in Figure 2 ) . We then calculate the average of the GloVe word vectors for all terms in the query and compare it with the averaged GloVe vectors of all other products . We again use cosine similarity \ufb01nd the 10 nearest products . We therefore have 4 different methods ( FocusAbstracted , Over - allPurpMech , OverallGloVe , and FocusOnly ) of obtaining 10 matches each for 10 different search scenarios . Figure 4 shows illustrative matches from each of these methods . We hypothesize that OverallPurpMech will do relatively poorly on relevance ( since it is tuned to capture \u201cthe\u201d overall purpose / mechanism of each document , which might miss the intended focus of a redesign scenario ) , but well on distance ( as it did in [ 15 ] ) . We hypothesize that OverallGlove will do poorly on relevance and distance ( since it is neither tuned for focus nor abstraction ) , and FocusOnly will do well on rele - vance , but poorly for distance ( since it is tuned for focus , but not abstraction ) . Finally , we hypothesize that our FocusAb - stracted method will do well on both relevance and distance ( since it is tuned for both focus and abstraction ) . Results As a \ufb01rst - pass analysis , we note that the methods return almost completely non - overlapping sets of matches for each scenario , giving us 394 unique matches out of 400 total possible unique matches . This initial result suggests that the methods behave quite differently . Indeed , as Figure 4 illustrates , OverallPurp - Mech appears to return domain - distant inspirations that are analogous on some purpose of the seed product ( though not necessarily the speci\ufb01ed purpose ) , and OverallGlove and Fo - cusOnly appear to return highly relevant inspirations from the same / similar domains , while FocusAbstracted matches appear to be both highly relevant and from distant domains . FocusAbstracted matches more relevant than OverallPurp - Mech , and as relevant as OverallGloVe and FocusOnly We now turn to formal quantitative tests of our hypothe - ses . Figure 5 ( left panel ) shows relevance scores by method , collapsed across scenarios . Using a one - way Analysis of 1 Available here : https : / / nlp . stanford . edu / projects / glove / CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 7 Figure 4 . Illustrative matches from each method for the scenario : \u201cmake the dish compatible with different sizes of soap bars\u201d . The abstraction of \u201csoap\u201d seems to allow the FocusAbstracted method to ignore the domain difference of knives vs . soap . OverallPurpMech \ufb01nds a match from a different domain that is analogous in terms of its overall purpose of keeping something clean / dry , but misses the core purpose of adapting to different sizes . In contrast , both OverallGloVe and FocusOnly \ufb01nd a highly relevant match from the same domain . Figure 5 . FocusAbstracted matches achieve comparable relevance to FocusOnly and GloVe baselines , and more relevance than OverallPurpMech ( left panel ) while being more domain distant than FocusOnly and GloVe baseline matches , and equivalently domain distant as OverallPurpMech matches ( right panel Variance ( ANOVA ) model with method as the sole between - observations factor , and matches as the unit of observation , we \ufb01nd signi\ufb01cant differences on the mean relevance score across the methods , F ( 3 , 396 ) = 14 . 1 , p < . 01 . A follow - up Tukey Honestly Signi\ufb01cant Difference ( HSD ) post - hoc test ( to correct for increased chance of false positives due to multiple comparisons ) shows that only OverallPurpMech has signi\ufb01 - cantly lower relevance compared to the other methods , p < . 01 vs . OverallGloVe , FocusOnly , and FocusAbstracted . FocusAbstracted matches more domain distant than FocusOnly and OverallGloVe , and as distant as OverallPurpMech Figure 5 ( right panel ) shows distance scores by method , col - lapsed across scenarios . Again using a one - way ANOVA model with method as the between - observations factor , we \ufb01nd signi\ufb01cant differences across the methods on mean dis - tance , F ( 3 , 396 ) = 14 . 1 , p < . 01 . A follow - up Tukey HSD post - hoc test shows that only FocusAbstracted has signi\ufb01 - cantly higher domain distance compared to OverallGloVe ( p < . 05 ) and FocusOnly ( p < . 05 ) . Despite being numerically more domain distant than OverallGloVe and FocusOnly , the OverallPurpMech method\u2019s matches are not signi\ufb01cantly more domain distant after the Tukey corrections for multiple com - parisons . Distance of FocusAbstracted matches uncorrelated with rele - vance , in contrast to OverallPurpMech and FocusOnly Finally , we explore the relationship between relevance and domain distance might vary across the methods . Since Over - allGloVe tends to match primarily based on surface features , we expect a strong negative correlation between relevance and distance , such that relevant matches tend to be less domain distant . We expect a similar relationship for FocusOnly ( since it operates in a very similar way to OverallGloVe ) , albeit pos - sibly weaker since it ignores other domain details that were in ignored sentences / terms , but no such relationship for Focus - Abstracted and OverallPurpMech ( since they are designed to abstract away from domain details . Indeed , across all matches from all methods for all scenarios , there is a signi\ufb01cant negative correlation between relevance and distance : on average , the more relevant a match is , the closer it is to the domain of the seed product , r = \u2013 0 . 19 , 95 % CI = [ \u2013 0 . 28 , \u2013 0 . 09 ] , p < . 01 . However , the relationship between CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 8 relevance and distance varies by method . As expected , the relationship is strongest for OverallGloVe matches , r = \u2013 0 . 36 [ \u2013 0 . 52 , \u2013 0 . 18 ] , followed by FocusOnly , r = \u2013 0 . 22 [ \u2013 0 . 40 , \u2013 0 . 03 ] , p < . 05 . In contrast , there is no signi\ufb01cant correlation between relevance and distance for either FocusAbstracted , r = \u2013 0 . 09 [ \u2013 0 . 28 , 0 . 11 ] , p = 0 . 38 , or OverallPurpMech , r = \u2013 0 . 02 [ \u2013 0 . 22 , 0 . 18 ] , p = 0 . 38 . Case Study To give an intuition for what might be driving these quantita - tive difference , we return to examine 4 illustrative matches for the scenario \u201cmake the dish compatible with different sizes of soap bars\u201d ( shown also in Figure 4 ) . OverallPurpMech returns cross - domain matches like a \u201cyoga mat wash stack machine\u201d , which includes drying and cleaning functions for the yoga mats , which match the overall main purpose of the \u201cSoapy slider\u201d product ( i . e . , keeping the bar of soap dry ; in fact , this yoga mat inspiration is relevant for the other \u201cSoapy slider\u201d scenario that focuses on this purpose ) . This illustrates how OverallPurpMech can return interestingly distant but ul - timately irrelevant matches if the designer wants to focus on an aspect of a seed product that is different from its main purpose . On the other extreme , OverallGlove and FocusOnly both return many relevant but near matches , like a \u201csoap saver\u201d device that fuses small used bars of soap together so they don\u2019t slip through the cracks , or a \u201ctouchless soap dispensing unit\u201d with a winding inner tube that expands to reach inside any size bottle . In contrast to both of these extremes , our FocusAbstracted method is able to return matches that are both relevant to the focus need and domain distant , like a \u201cknife rolodex\u201d product that includes multiple slots for different sized knives , or a \u201cmaximizing phone tablet\u201d ( not shown in Figure 4 ) , which uses a telescopic frame to adjust to different - sized phones . In both of these cases , our FocusAbstracted method is able to zero in on the idea of adjusting to different \u201cspatial quantities\u201d , while ignoring differences in the kind of \u201cpersonal product\u201d ( e . g . , knives , phones ) being adjusted to , due to the replacing of the domain - speci\ufb01c terms like knife and phone with abstracted properties that match those of the soap bar . DISCUSSION Summary and Implications of Contributions In this paper , we sought to design a system that can tune computational analogical search to \ufb01nd relevant and distant inspirations for speci\ufb01c design needs . We presented a system that allows designers to focus on a speci\ufb01c aspect of a product description by selecting key terms to form a query , and create a targeted abstraction of those terms by selecting properties from a knowledge base that are important for understanding the core relational structure of the design need . We demon - strated that this focus - abstracted approach led to the retrieval of inspirations that were both relevant and distant , in contrast to alternative state - of - the - art approaches that either sacri\ufb01ced relevance for distance , or vice versa . Thus , we contribute a promising new method \ufb01nding distant analogical inspirations for speci\ufb01c design needs . One speci\ufb01c \ufb01nding that deserves further discussion is the high performance of the OverallGlove condition in terms of relevance . Our initial prediction was that this condition would perform poorly on relevance , since , like the OverallPurpMech method from [ 15 ] , it operates on the whole product description as opposed to a speci\ufb01c focus query . Cognitive theories of analogy suggest one possible explanation . In particular , some researchers point to the \u201ckind world hypothesis\u201d to explain how humans learn abstract concepts : salient surface features ( which tend to be shared by things in the same domain ) tend to be strongly correlated with structural features . As Gentner [ 11 ] notes , \u201cif something looks like a tiger , it is probably a tiger\u201d . One implication of this is that things that are in the same domain likely share many relational features , including purposes . Thus , since OverallGlove is tuned to match based on surface features , it is possible that it found many relevant matches for the speci\ufb01c need simply by \ufb01nding things in the same domain . Limitations and Future Work Supporting more expressive queries We have shown how helpful a focus - abstraction interface can be for a designer wishing to re - design an aspect of a product . However , in our pilot tests of the interface , we noticed that some information needs are still hard to express . An interesting direction is to explore more expressive queries ( by adding more mechanisms to the interface , for example allowing designers to manually add important terms or prop - erties ) . This would also allow designers to explicitly express trade - offs . Improving expressiveness might be especially im - portant for domains with highly technical concepts / terms with very speci\ufb01c meanings ( e . g . , regularization in machine learn - ing ) that have poor coverage in existing knowledge bases like Cyc . Automatically identifying different purposes in a document Our analogy engine calculates a representation of the overall purpose of the focus - abstracted query and all the documents in the corpus . For the abstracted focus description this is a good \ufb01t , as the overall purpose is identical to the speci\ufb01c need . For the rest of the corpus the overall purpose comprises several purposes . We expect that automatically dividing each docu - ment to sub - purposes prior to searching for purpose matches would signi\ufb01cantly improve them . The usage patterns of our tool can serve as annotated set for learning how to segment the documents . Automatically suggesting queries Another interesting future direction might be to use informa - tion obtained from usage of the tool to learn common focus and abstraction patterns , and suggest focus - abstractions au - tomatically to designers . For example , we might learn that soap dishes , phone cases , and cake - cutters often have a focus - abstracted problem of < expanding to \ufb01t objects of different physical sizes > , and suggest these focus - abstractions to other designers creating queries for these ( and similar ) products . Extending to other datasets and domains While we have tested our method on Quirky innovations , it would be useful to explore its utility on other corpora . In CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 9 particular , it would be interesting to test our ideas on a corpus of products from manufacturing companies , which are con - stantly looking for innovations for improving their products , or even to corpora of research papers and patents . The tar - geted abstraction approach could be particularly powerful for \ufb01nding analogies for \ufb01elds of study where the properties of the objects are critical for determining what makes for useful analogies : for example , as we noticed from our experts in mechanical engineering and materials science , someone work - ing on ways to deform stretchable polymers would not likely bene\ufb01t from analogies to deformation techniques for concrete , since polymers ( by virtue of their material properties ) react very differently to physical stresses . Note , further research is required to understand how well the method generalizes to such corpora . As noted above , Cyc does not contain many technical terms , and we may need a different source for their abstractions . CONCLUSION In this paper , we contribute a novel system for tuning analogi - cal search for speci\ufb01c design needs , consisting of an interface for designers to express their speci\ufb01c needs in abstract terms , and an analogy search engine that uses this focus - abstracted query to \ufb01nd inspirations from a corpus that are both relevant and domain - distant . This work contributes a novel path for - ward to computational support for mining large databases of potential inspirations on the Web to improve design work . ACKNOWLEDGEMENTS The authors thank the anonymous reviewers for their helpful feedback , and Amir Shapira for his helpful insights into the design process . Dafna Shahaf is a Harry & Abe Sherman assistant professor . This work was supported by NSF grants CHS - 1526665 , IIS - 1149797 , IIS - 1217559 , Carnegie Mellon\u2019s Web2020 initiative , Bosch , Google , ISF grant 1764 / 15 , Alon grant , and the HUJI Cyber Security Research Center in con - junction with the Israel National Cyber Bureau in the Prime Minister\u2019s Of\ufb01ce . REFERENCES 1 . Smithsonian National Air and Space Museum . 2017 . Inventing a Flying Machine : The Breakthrough Concept . https : / / airandspace . si . edu / exhibitions / wright - % brothers / online / fly / 1899 / breakthrough . cfm . ( 2017 ) . Accessed : 2017 - 09 - 15 . 2 . Dzmitry Bahdanau , Kyunghyun Cho , and Yoshua Bengio . 2014 . Neural machine translation by jointly learning to align and translate . arXiv preprint arXiv : 1409 . 0473 ( 2014 ) . 3 . Edward Loper Bird , Steven and Ewan Klein . 2009 . Natural Language Processing with Python . O\u2019Reilly Media Inc . ( 2009 ) . 4 . David M . Blei , Andrew Y . Ng , Michael I . Jordan , and John Lafferty . 2003 . Latent Dirichlet Allocation . Journal of Machine Learning Research ( 2003 ) , 993 \u2013 1022 . 5 . Joel Chan , Steven P . Dow , and Christian D . Schunn . 2015 . Do The Best Design Ideas ( Really ) Come From Conceptually Distant Sources Of Inspiration ? Design Studies 36 ( 2015 ) , 31 \u2013 58 . DOI : http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 6 . Joel Chan , Katherine Fu , Christian . D . Schunn , Jonathan Cagan , Kristin L . Wood , and Kenneth Kotovsky . 2011 . On the bene\ufb01ts and pitfalls of analogies for innovative design : Ideation performance based on analogical distance , commonness , and modality of examples . Journal of Mechanical Design 133 ( 2011 ) , 081004 . DOI : http : / / dx . doi . org / 10 . 1115 / 1 . 4004396 7 . Scott Deerwester , Susan T . Dumais , Geroge W . Furnas , and Thomas K . Landauer . 1990 . Indexing by Latent Semantic Analysis . JASIST 41 , 6 ( 1990 ) , 1990 . 8 . Brian Falkenhainer , Kenneth D Forbus , and Dedre Gentner . 1989 . The structure - mapping engine : Algorithm and examples . Arti\ufb01cial intelligence 41 , 1 ( 1989 ) , 1 \u2013 63 . 9 . Katherine Fu , Joel Chan , Jonathan Cagan , Kenneth Kotovsky , Christian Schunn , and Kristin Wood . 2013 . The Meaning of Near and Far : The Impact of Structuring Design Databases and the Effect of Distance of Analogy on Design Output . JMD 135 , 2 ( 2013 ) , 021007 . DOI : http : / / dx . doi . org / 10 . 1115 / 1 . 4023158 10 . Dedre Gentner . 1983 . Structure - Mapping : A Theoretical Framework for Analogy * . Cognitive science 7 , 2 ( 1983 ) , 155 \u2013 170 . 11 . Dedre Gentner . 1989 . The mechanisms of analogical learning . In Similarity and Analogical Reasoning , Stella Vosniadou and Andrew Ortony ( Eds . ) . Cambridge University Press , New York , NY , USA , 197 \u2013 241 . http : / / dl . acm . org / citation . cfm ? id = 107328 . 107335 12 . Dedre Gentner and Arthur B Markman . 1997 . Structure mapping in analogy and similarity . American psychologist 52 , 1 ( 1997 ) , 45 . 13 . Milene Goncalves , Carlos Cardoso , and Petra Badke - Schaub . 2013 . Inspiration peak : exploring the semantic distance between design problem and textual inspirational stimuli . International Journal of Design Creativity and Innovation 1 , 4 ( 2013 ) , 215 \u2013 232 . 14 . Margaret Guroff and Margaret Guroff . 2016 . The Untold Story Of How Bicycle Design Led To The Invention Of The Airplane . ( July 2016 ) . https : / / www . fastcodesign . com / 3061592 / the - untold - story - of - how - bicycle - \\ design - led - to - the - invention - of - the - airplane 15 . Tom Hope , Joel Chan , Aniket Kittur , and Dafna Shahaf . 2017 . Accelerating Innovation Through Analogy Mining . In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . ACM , 235 \u2013 243 . 16 . Douglas B . Lenat . 1995 . CYC : a large - scale investment in knowledge infrastructure . In Communications of the ACM Volume 38 Issue 11 , Nov . 1995 . ACM , 33 \u2013 38 . 17 . Douglas B . Lenat , V . Guha , R . , K . Pittman , D . Pratt , and M . Shepherd . 1990 . CYC : towards programs with common sense . . In Communications of the ACM Volume 33 ( 8 ) Issue 11 , Aug . 1990 . ACM , 30 \u2013 49 . CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 10 18 . J . S . Linsey , A . B . Markman , and K . L . Wood . 2012 . Design by Analogy : A Study of the WordTree Method for Problem Re - Representation . Journal of Mechanical Design 134 , 4 ( April 2012 ) , 041009 \u2013 041009 . DOI : http : / / dx . doi . org / 10 . 1115 / 1 . 4006145 19 . Johnson Laird Phillip M . 2005 . Flying bicycles : How the Wright brothers invented the airplane . Mind & Society 4 , 1 ( 2005 ) , 27 \u2013 48 . 20 . Tomas Mikolov , Kai Chen , Greg Corrado , and Jeffrey Dean . 2013 . Ef\ufb01cient Estimation of Word Representations in Vector Space . arXiv : 1301 . 3781 [ cs ] ( Jan . 2013 ) . http : / / arxiv . org / abs / 1301 . 3781 arXiv : 1301 . 3781 . 21 . George A Miller . 1995 . WordNet : a lexical database for English . Commun . ACM 38 , 11 ( 1995 ) , 39 \u2013 41 . 22 . Library of Congress . 2017 . History of Edison Motion Pictures . https : / / www . loc . gov / collections / edison - company - motion - pictures - and - sound - recordings / articles - and - essays / history - of - edison - motion - pictures / . ( 2017 ) . Accessed : 2017 - 09 - 15 . 23 . Jeffrey Pennington , Richard Socher , and Christopher D Manning . 2014 . Glove : Global Vectors for Word Representation . In EMNLP , Vol . 14 . 1532 \u2013 43 . 24 . L . E . Richland , O . Zur , and K . J . Holyoak . 2007 . Cognitive supports for analogies in the mathematics classroom . Science 316 , 5828 ( 2007 ) , 1128 \u2013 1129 . DOI : http : / / dx . doi . org / 10 . 1126 / science . 1142103 25 . Robert Speer and Catherine Havasi . 2012 . Representing General Relational Knowledge in ConceptNet 5 . LREC . http : / / ai2 - s2 - pdfs . s3 . amazonaws . com 26 . Vibeke Venema . 2013 . The car mechanic who uncorked a childbirth revolution . BBC News ( Dec . 2013 ) . http : / / www . bbc . com / news / magazine - 25137800 27 . Lixiu Yu , Aniket Kittur , and Robert E Kraut . 2014a . Searching for analogical ideas with crowds . In Proceedings of the 32nd annual ACM conference on Human factors in computing systems . ACM , 1225 \u2013 1234 . 28 . L Yu , B Kraut , and A Kittur . 2014b . Distributed analogical idea generation : innovating with crowds . In CHI\u201914 . 29 . Lixiu Yu , Robert E . Kraut , and Aniket Kittur . 2016 . Distributed Analogical Idea Generation with Multiple Constraints . In Proceedings of the 19th ACM Conference on Computer - Supported Cooperative Work & Social Computing ( CSCW \u201916 ) . ACM , New York , NY , USA , 1236 \u2013 1245 . DOI : http : / / dx . doi . org / 10 . 1145 / 2818048 . 2835201 CHI 2018 Paper CHI 2018 , April 21 \u2013 26 , 2018 , Montr\u00e9al , QC , Canada Paper 121 Page 11",
    "laneEngineeringSerendipityWhen": "R E S E A R C H A R T I C L E Engineering serendipity : When does knowledge sharing lead to knowledge production ? Jacqueline N . Lane 1 | Ina Ganguli 2 | Patrick Gaule 3 | Eva Guinan 4 | Karim R . Lakhani 1 1 Laboratory for Innovation Science at Harvard , Harvard Business School , Boston , Massachusetts 2 Department of Economics , University of Massachusetts Amherst , Amherst , Massachusetts 3 Department of Economics , University of Bath , Bath , UK 4 Dana - Farber Cancer Institute , Boston , Massachusetts Correspondence Jacqueline N . Lane , Laboratory for Innovation Science at Harvard , Harvard Business School , 175 N . Harvard Street , Boston , MA 02134 . Email : jnlane @ hbs . edu Funding information Harvard Catalyst ; Laboratory for Innovation Science at Harvard University ; NIH Clinical Center , Grant / Award Numbers : UL1RR025758 - 02S4 , UL1TR000170 , UL1TR001102 , UL1TR002541 ; Harvard Business School Division of Research and Faculty Development Abstract Research Summary : We investigate how knowledge similarity between two individuals is systematically related to the likelihood that a serendipitous encoun - ter results in knowledge production . We conduct a field experiment at a medical research symposium , where we exogenously varied opportunities for face - to - face encounters among 15 , 817 scientist - pairs . Our data include direct observations of interaction pat - terns collected using sociometric badges , and detailed , longitudinal data of the scientists ' pos - tsymposium publication records over 6 years . We find that interacting scientists acquire more knowl - edge and coauthor 1 . 2 more papers when they share some overlapping interests , but cite each other ' s work between three and seven times less when they are from the same field . Our findings reveal both col - laborative and competitive effects of knowledge simi - larity on knowledge production outcomes . Managerial Summary : Managers often try to stimulate innovation by encouraging serendipitous interactions between employees , for example by using office space redesigns , conferences and similar events . Are such inter - ventions effective ? This article proposes that an effective encounter depends on the degree of common knowledge shared by the individuals . We find that scientists who Received : 4 February 2020 Revised : 27 October 2020 Accepted : 10 November 2020 DOI : 10 . 1002 / smj . 3256 This is an open access article under the terms of the Creative Commons Attribution License , which permits use , distribution and reproduction in any medium , provided the original work is properly cited . \u00a9 2020 The Authors . Strategic Management Journal published by John Wiley & Sons Ltd . Strat Mgmt J . 2020 ; 1 \u2013 30 . wileyonlinelibrary . com / journal / smj 1 attend the same conference are more likely to learn from each other and collaborate effectively when they have some common interests , but may view each other com - petitively when they work in the same field . Hence , when designing opportunities for face - to - face interactions , man - agers should consider knowledge similarity as a criteria for fostering more productive exchanges . K E Y W O R D S innovation , knowledge production , knowledge sharing , knowledge similarity , natural field experiment 1 | INTRODUCTION In 2013 , cell biologist William Earnshaw of the University of Edinburgh happened to attend the same academic conference as systems biologist Job Dekker of the University of Massachusetts Medical School and computational biologist Leonid Mirny from the Massachusetts Institute of Technology . By chance , Earnshaw attended a presentation by Dekker and Mirny about their joint work on mitotic chromosomes , during which he became convinced that his lab could pro - vide bench methods to improve Dekker and Mirny ' s computational models . Earnshaw approached Dekker and Mirny after the talk ; their conversation evolved into a three - lab collab - oration and a 2018 publication in Science ( Pain , 2018 ) . This anecdotal example suggests that serendipitous encounters can play a role in inno - vation , perhaps by exposing individuals to people that they would not otherwise have a chance to meet , and to unfamiliar sources of information that can be combined with their own knowledge stock and lead to new discoveries ( Fleming , Mingo , & Chen , 2007 ; Uzzi , Mukherjee , Stringer , & Jones , 2013 ) . However , the likelihood that serendipitous encounters lead to successful knowledge production remains poorly understood often because the pro - cess is uncertain , complex and relatively understudied . In particular , even if an encounter eventually produces new knowledge , this process may occur slowly and only emerge after a number of years ( Catalini , 2018 ; Fleming , 2001 ) . A significant degree of knowledge within organizations is tacit and is not easily transferred ( Hansen , 1999 ; Nonaka , 1994 ) . Knowl - edge creation via new collaborations can be costly ( Boudreau et al . , 2017 ; Dahlander & McFarland , 2013 ) , requiring coordinated effort , alignment of incentives , establishment of trust , generation of creative synergy , and \u201c matching \u201d criteria , such as personality and scheduling compatibility ( Azoulay , Ding , & Stuart , 2007 ; Catalini , 2018 ) . Even if knowledge transfer or creation is successful , knowledge tends to diffuse slowly through interpersonal networks ( Fleming et al . , 2007 ; Singh , 2005 ) . The question thus arises of when serendipi - tous encounters are more likely to facilitate useful idea exchange and knowledge produc - tion . Previous literature tends to focus on the factors that may increase the amount or intensity of short - term communication and interaction between individuals , rather than the implications of serendipitous encounters on knowledge production ( Allen , 1977 ; Kleinbaum , Stuart , & Tushman , 2013 ) . Given that geographic proximity increases the 2 LANE ET AL . likelihood of serendipitous encounters ( Catalini , 2018 ) , recent research has begun to exam - ine the effects of organizational redesign on knowledge production ( Catalini , 2018 ; Fang , Lee , & Schilling , 2010 ; Lee , 2019 ) . Although it is possible that organizational redesign and greater geographic proximity between individuals can increase knowledge production , most studies do not directly address how knowledge sharing between two individuals affects the path - dependent nature of the knowledge production process . Since prior work tends to focus on a single knowledge production outcome ( Phelps , Heidl , & Wadhwa , 2012 ) , often on the factors that impede knowledge transfer ( Carlile , 2004 ; Hansen , 1999 ; Szulanski , 1996 ; Uzzi , 1997 ) , it is unclear how knowledge sharing may also affect knowledge creation and diffusion . Hence , we know relatively little about whether serendipitous encounters actually lead to meaningful idea exchange that can be applied to one ' s own tasks , or whether these interactions can spark new collaborations and broader diffusion of ideas over time . The goal of our work is to address this gap in literature by examining to what extent a sys - tematic relationship exists between serendipitous encounters and knowledge production outcomes . We posit that an effective encounter hinges not just on the serendipity of the interac - tion but also on the existence of common knowledge that individuals use to make sense of each other ' s specialized knowledge and connect it to what they already know . Common knowledge enhances the ability for new knowledge to be assimilated into the concepts , objects and patterns that are already present in people ' s cognitive structures ( Bower & Hilgard , 1981 ; Cohen & Levinthal , 1990 ) . When new information is related to prior knowl - edge constructs , it is more readily absorbed , integrated , and applied in new settings ( Carlile , 2004 ; Kogut & Zander , 1992 ) without sufficient prior knowledge , individuals may have difficulty integrating new knowledge ( Rosenkopf & Almeida , 2003 ) . In this article , we argue that there are two essential characteristics to people ' s common knowledge : field similarity in their educational backgrounds and training , and intellectual similarity in their interests , passions and pursuits . Serendipitous encounters are more likely to be effec - tive when people share some knowledge similarity in both their field and intellectual interests . To systematically investigate the relationships between serendipitous encounters and knowledge production requires a rich longitudinal dataset that directly observes exogenous face - to - face knowledge sharing between two individuals and subsequent knowledge production outcomes . To draw causal inferences , we designed and executed a field experiment at a research symposium on advanced imaging in early 2012 at a leading academic medical center . We chose this setting due to the centrality of knowledge production to their organizational and individual performance goals and the prevalence of knowledge sharing norms ( Dahlander & McFarland , 2013 ) . We created exogenous variation in opportunities for information - rich , face - to - face encounters between cross - disciplinary scientists interested in applying for an internal grant program promoting the development of advanced imaging solutions to address an unmet clinical need . Using electronic sociometric badges ( Kim , McFee , Olguin , Waber , & Pentland , 2012 ) , we collected a unique dataset of fine - grained , live interactions between pairs of scientists who were randomly assigned to be in the same symposium room for knowledge shar - ing . We then combined this with a rich longitudinal dataset on long - run knowledge production outcomes from the scientists ' written publications over a six - year period , where we linked scien - tists ' pairwise keywords , collaborations and forward citations to examine the extent that the encounters led to knowledge transfer , creation and diffusion , respectively . This design enabled us to causally and systematically identify the relationships between knowledge sharing and LANE ET AL . 3 knowledge production to explain why some serendipitous encounters resulted in knowledge production while others did not . We find both cooperative and competitive effects of serendipitous knowledge sharing encounters on knowledge production . On the one hand , knowledge sharing led to greater knowledge transfer and knowledge creation when people had some overlapping intellec - tual and field interests , respectively . Although knowledge sharing had limited impact on knowledge transfer among pairs with either few or many overlapping interests , we observe a twofold increase in knowledge transfer ( of Medical Subject Headings or MeSH key - words ) for moderately similar pairs , suggesting an inverted U - shaped relationship between knowledge overlap and knowledge transfer . Similarly , we find that knowledge sharing between pairs with some common interests led to 1 . 2 more coauthored peer - reviewed publications , compared to a 0 . 98 decrease in copublications among pairs in the same field , which suggests that the knowledge sharing intervention reduced the search costs associated with forming synergistic collaborations . By comparing the knowledge drivers of near - term collaborations and long - term collaborations , we find that common knowledge is more critical for the former . On the other hand , knowledge sharing reduces knowledge diffusion between people from the same field , with interacting pairs citing each other between 2 . 8 and 6 . 7 times less than noninteracting pairs from more distant fields . We aim to make several contributions to the literature . First , we contribute to the lit - erature on search costs and information frictions associated with the knowledge produc - tion process ( Agrawal & Goldfarb , 2008 ; Boudreau et al . , 2017 ; Catalini , 2018 ; Hansen , 1999 ; Hansen & Haas , 2001 ) , by focusing on the extent that \u201c engineered seren - dipity \u201d can mitigate some of these costs . We emphasize the role of knowledge similarity in shaping different components of knowledge production , and over different time hori - zons . Most prior research has predominantly focused on one component or a single time frame ( Adams , Black , Clemmons , & Stephan , 2005 ; Boudreau et al . , 2017 ; Dahlander & McFarland , 2013 ; Szulanski , 1996 ) . This work is unique because it extends the findings from Boudreau et al . ( 2017 ) on search costs and short - term collaborations using the same field experiment , to show how engineered serendipity may alter the role of knowl - edge similarity in knowledge production over time . Our findings suggest that the magni - tude of these frictions is likely to be more difficult to overcome in the short - run compared to the long - run . Second , we make contributions to the literature on temporary colocation , and the value of conferences , symposia and similar events on the direction of inventive activity on the knowl - edge frontier and the emergent patterns of collaborative activity ( Biancani , McFarland , & Dahlander , 2014 ; Boudreau et al . , 2017 ; Campos , Leon , & McQuillin , 2018 ; Catalini , Fons - Rosen , & Gaul\u00e9 , 2020 ; Chai & Freeman , 2019 ) . Prior research has been limited in its ability to observe actual acts of knowledge sharing between pairs , and its implications on knowledge production . Our study indicates that even brief , information - rich encounters at conferences can benefit knowledge production , with the potential to alter people ' s research directions and collaboration networks . Third , we make methodological contributions by highlighting the benefits of long - term studies that amalgamate multiple forms and uses of data . Prospective experiments can support multiple lines of investigation involving both near - term and long - term outcomes that may not be possible in retrospective , archival studies and suggests the use of multiple sources of data for unpacking the dynamics of knowledge production . 4 LANE ET AL . 2 | THEORY AND HYPOTHESES 2 . 1 | Knowledge production and knowledge sharing According to the knowledge - based view of the firm , knowledge production is a multicomponent process that involves the transfer , creation and diffusion of knowledge ( Grant , 1996 ; Kogut & Zander , 1992 ; Nonaka , 1994 ) . Knowledge transfer involves the movement of facts , relationships , and insights from one setting to another , and becomes evident when the experience acquired by an individual , group or organization in one setting is applied in another setting ( Argote , 2012 ; Hansen , 1999 ; Szulanski , 1996 ) . Knowledge creation refers to the generation of facts , relationships and insights to solve new problems ( Nonaka , 1994 ) . The emphasis on new knowledge is what distinguishes the process of knowledge transfer from creation . Because new knowledge is not held by anyone prior to its creation , it cannot be transferred and applied directly ( McFadyen & Cannella Jr , 2004 ) . Conversely , knowledge diffusion refers to the process through which transferred or newly created knowledge is disseminated and used by other indi - viduals , groups and organizations ( Fleming et al . , 2007 ; Singh , 2005 ) . Knowledge diffusion is a critical component of the knowledge production process because it reduces duplication of effort and promotes efficiency by demarcating what is known from what is yet to be explored on the knowledge frontier ( Boudreau & Lakhani , 2015 ) . Critical to the efficiency of knowledge production is the process of knowledge sharing between individuals possessing different types of knowledge ( Grant , 1996 ; Kogut & Zander , 1992 ; Nonaka , 1994 ) , and their ability to learn from these interactions ( Levinthal & March , 1993 ; Simon , 1991 ) . Common knowledge between partners is important because it enables individuals to absorb the aspects of their knowledge sets that they do not hold in com - mon , or that are unique to each individual ( Carlile , 2004 ; Cohen & Levinthal , 1990 ; Maurer & Ebers , 2006 ) . Building on prior work , we propose that there are at least two separate pathways through which people develop their knowledge bases : their field and intellectual specialties . Because of the path - dependent nature of knowledge ( Hargadon & Sutton , 1997 ) , the overlap in two individuals ' field and intellectual specialties may have different implications on knowledge production . Table 1 provides a summary of our knowledge constructs . 2 . 2 | Two knowledge bases of similarity : Field specialty and intellectual specialty Field specialties are developed through an individual ' s educational background , skills and train - ing ( Bechky , 2011 ; Bourdieu , 1984 ; Haas & Park , 2010 ) . In biomedical research , which is our focal context of study , there are multiple medical field specialties , such as neurology , radiology or oncology that focus on investigations of the biological process and the causes of specific dis - eases . Organizations often structure their departmental and divisional memberships around field specialties ( Biancani et al . , 2014 ) . In universities and academia , this means that field spe - cialties are often the primary channels through which resources , such as salaries , funding , ten - ure , offices and laboratory space , are apportioned to faculty ( Biancani , Dahlander , McFarland , & Smith , 2018 ) . For these reasons , field specialties provide professional reference groups ( Haas & Park , 2010 ) for members to identify appropriate behavior , rigor of scholarship , career goals and aspirations ( Abbott , 2001 ) . In contrast , intellectual specialties are closely aligned to personal interests and passions that evolve over time . Individuals can have LANE ET AL . 5 memberships in multiple intellectual specialties depending on their current pursuits and priori - ties ( Caza et al . , 2018 ) . For example , in biomedical research , intellectual specialties include research topics such as aging , addictions , Alzheimer ' s disease , brain injury , sleep , and stem cells ; each of these intellectual specialties is associated with memberships in professional refer - ence groups ( Haas & Park , 2010 ) for people motivated by common problems , methods , trends , and processes . These intellectual interests can be specific to a field ' s research topics , such as Alzheimer ' s disease , which is typically studied by neurologists , or span different fields , such as stem cell research . In short , field and intellectual specialties shape people ' s identities , language , tastes and affil - iations through professional memberships ( Haas & Park , 2010 ) . Individuals will likely adopt multiple professional identities depending on their field and intellectual memberships : this means that people from different field specialties can share overlapping intellectual interests , while individuals from the same field specialty can have divergent intellectual interests , because field and intellectual interests are two distinct dimensions of individuals ' knowledge constructs . In distinguishing how field and intellectual overlap shape knowledge production , knowl - edge similarity describes the distance between the field specialties and intellectual specialties of two knowledge sharing partners . Individuals are more likely to recognize and absorb new ideas when they already have some existing expertise and find it more difficult when the ideas are outside their realm of expertise ( Carlile , 2004 ; Cohen & Levinthal , 1990 ) . Put differently , even though distant knowledge tends to be novel and valuable ( Jeppesen & Lakhani , 2010 ; Leahey TABLE 1 Summary of knowledge constructs Knowledgeconstruct Definition Citations Knowledgeproduction A multicomponent process that involves knowledge transfer , creation , and diffusion . Grant , 1996 ; Kogut & Zander , 1992 ; Nonaka , 1994 ; Spender , 1996 Knowledge transfer The movement of facts , relationships , and insights from one setting to another Argote , 2012 ; Hansen , 1999 ; Szulanski , 1996 Knowledgecreation The generation of facts , relationships , and insights to solve problems that are new to the knowledge frontier Arrow 1962 ; Boudreau & Lakhani , 2015 ; Nonaka , 1994 Knowledgediffusion The dissemination of facts , relationships , and insights that are subsequently used by others Fleming & Singh , 2010 ; Singh , 2005 Knowledgesimilarity The degree of common knowledge between two individuals ' field of study and intellectual interests Cohen & Levinthal , 1990 ; Carlile , 2004 ; Dahlander & McFarland , 2013 ; Leahey , Beckman , & Stanko , 2017 Fieldsimilarity The dimension of knowledge similarity that emerges from the overlap between two individual ' s educational background , skills , and training Bechky , 2011 ; Carlile , 2004 ; Bourdieu , 1984 ; Haas & Park , 2010 Intellectualsimilarity The dimension of knowledge similarity that emerges from the overlap between two individual ' s personal interests and passions Caza , Moss , & Vough , 2018 ; Dahlander & McFarland , 2013 6 LANE ET AL . et al . , 2017 ) , people are more likely to experience challenges communicating with one another , and may not recognize the value of external information ( Cohen & Levinthal , 1990 ; Grant , 1996 ) . For example , two neurologists are less likely to experience challenges communicating with each other than a neurologist and oncologist pair because they share more similar educational backgrounds , training and clinical expertise . However , there may be greater opportunities for knowledge produc - tion between the neurologist and oncologist pair because their field specialties are more distant . Similarly , although two sleep researchers may communicate and comprehend each other with rela - tive ease because they read and conduct similar research ( Dahlander & McFarland , 2013 ) , there is greater potential for knowledge production between an aging and a sleep researcher , who may offer greater utility and value to one another to solve problems related to both sleep and aging processes \u2014 provided that they have sufficient common knowledge ( Bechky , 2011 ; Grant , 1996 ) . This suggests that there is an inherent trade - off between the greater ease of conversing and absorb - ing ideas locally from individuals with many overlapping interests and the greater potential oppor - tunities for novel ideas and knowledge transfer , creation , and diffusion from sharing and assessing the knowledge of those with more distant interests . 2 . 3 | Knowledge transfer and knowledge similarity Knowledge transfer is a two - sided process because it depends on the efforts of a source to share knowledge with a recipient and the recipient ' s efforts and capacity to acquire , absorb , and learn it ( Argote , 2012 ) . Because of greater common knowledge , local knowledge found within groups of similar individuals tends to be more easily transferred than distant knowledge spanning group boundaries ( Carlile , 2004 ; Kogut & Zander , 1992 ; Rosenkopf & Almeida , 2003 ) . That said , more distant knowledge may present nonredundant ideas that benefit learning and acquiring new concepts . Consequently , during serendipitous idea exchanges , we would expect that part - ners from dissimilar field and intellectual specialties offer each other more novel ideas and opportunities to learn ( Leahey et al . , 2017 ; Lee , 2019 ) . Divergent interests create a wider pool of knowledge to draw upon , which allows for multiple perspectives and problem - solving approaches that increase the likelihood of new discoveries ( Boudreau , Guinan , Lakhani , & Riedl , 2016 ) \u2014 provided that the partners share sufficient common knowledge to make sense of each other ' s knowledge dissimilarities . Based on these reasons , we expect that pairs with less field or intellectual overlap have a greater potential pool of ideas for knowledge transfer , up until a threshold of dissimilarity , after which they will lack sufficient common interests to bene - fit from knowledge transfer . Hypothesis ( H1a ) Field similarity has an inverted U - shaped effect on the relationship between knowledge sharing and knowledge transfer . Hypothesis ( H1b ) Intellectual similarity has an inverted U - shaped effect on the relationship between knowledge sharing and knowledge transfer . 2 . 4 | Knowledge creation and knowledge similarity There is a growing view among innovation scholars that specialization has created a \u201c knowl - edge burden \u201d hypothesis that makes collaborative work combining the increasingly narrow LANE ET AL . 7 niches of specialization imperative to moving the knowledge frontier forward ( Jones , 2009 ; Uzzi et al . , 2013 ) . Scientific discovery is a process that combines individually focused tasks \u2014 such as reading , experimentation , and writing \u2014 with social interactions through joint sense - making with others that can spark new discoveries ( Boudreau et al . , 2016 ; Latour & Woolgar , 2013 ) . That said , collaborations tend to be constrained by social processes , such as preferences for others with shared attributes , as well as search costs that stifle attempts to identify suitable col - laborators . First , principles of homophily suggest that people are attracted to others who hold similar values because their interactions are more rewarding and less uncertain ( McPherson & Smith - Lovin , 1987 ) , as a result of the benefits of security and mutual attraction ( Dahlander & McFarland , 2013 ) . Moreover , joint knowledge creation is a long - term investment in a multiplex relationship ( Uzzi , 1997 ) , where partners engage in joint problem - solving and spend time together discussing , reflecting and interacting to achieve mutual benefit ( McFadyen & Cannella Jr , 2004 ) . Due to these reasons , collaborations are more likely to form among individuals who share overlapping field and intellectual interests ( Biancani et al . , 2014 ; Dahlander & McFarland , 2013 ) . Second , there is a growing view that search costs and resulting information frictions tend to constrain people ' s collaboration patterns ( Boudreau et al . , 2017 ; Catalini , 2018 ) . Engineered , serendipitous interactions can mitigate some of these frictions by providing people the opportu - nity to exchange information and develop intellectual links with others they would not other - wise interact with . Such interactions may enable diverse partners to discover the potential complementarities they offer each other , leading to new collaborations . However , to the extent that serendipitous knowledge sharing leads to joint knowledge creation will also likely require that potential partners share some field or intellectual overlap ; otherwise their interests may be too disparate to establish synergies in incentives ( e . g . , publication or tenure requirements ) or research interests . Accordingly , we expect that increasing field and intellectual similarity should benefit knowledge creation up to a threshold , over which there are decreasing marginal returns on greater levels of knowledge similarity . Hypothesis ( H2a ) Field similarity has an inverted U - shaped effect on the relationship between knowledge sharing and knowledge creation . Hypothesis ( H2b ) Intellectual similarity has an inverted U - shaped effect on the relationship between knowledge sharing and knowledge creation . Another pathway that may lead to joint knowledge creation relies on prior collaborations and tie persistence ( Hasan & Koning , 2019 ; Ingram & Morris , 2007 ; Zhang & Guler , 2020 ) . We examine a specific type of tie persistence , namely the likelihood that an elemental collaboration persists into a complete knowledge product . In science , grant coapplications and copublications represent two essential but opposing ends of knowledge creation , from idea - generation to final output ( McFadyen & Cannella Jr , 2004 ) . Both are also essential to scientists ' research productiv - ity and often used as evaluative criteria for important decisions , such as promotion , tenure , awards and recognition ( Dahlander & McFarland , 2013 ; Stephan , 1996 ) . The knowledge simi - larity requirements that improve the likelihood that elemental collaborations persist into copublications may differ from the processes that shape long - term collaborations on peer - reviewed research publications . In addition to being at the open end of research inquiry , grant applications tend to have finite submission deadlines and these can impose resource , attention , coordination and other constraints on potential partners ( Dahlander , O ' Mahony , & 8 LANE ET AL . Gann , 2016 ) . Although serendipitous knowledge exchanges can reduce information search costs and expose people to promising new contacts , multiplex relationships require repeated interac - tions ( Ingram & Morris , 2007 ) . In the short - term , potential partners may turn to other individ - uals with whom they share significant knowledge overlap , due to fewer frictions and relational uncertainties ( e . g . , priorities , personalities , scheduling constraints ) associated with similar part - ners . Due to these reasons , common knowledge may be even more critical to shaping elemental collaborations . We thereby expect that elemental collaborations are more likely to persist into a final product of knowledge creation when knowledge sharing partners have significant knowl - edge overlap in terms of their field and intellectual interests . Hypothesis ( H2c ) An elemental collaboration is more likely to result in knowledge creation as the field similarity between two individuals increases . Hypothesis ( H2d ) An elemental collaboration is more likely to result in knowledge creation as the intellectual similarity between two individuals increases . 2 . 5 | Knowledge diffusion and knowledge similarity Among research scientists , forward citations to others ' publications are a primary means for dif - fusing knowledge . Beyond knowledge diffusion , forward citations constitute a critical means of social recognition for acknowledging the contributions of predecessors ( Merton , 1973 ) and trac - ing the path of scientific discovery and diffusion ( Stephan , 1996 ) . Some scholars argue that the number of citations a publication has received is perhaps the most common way to measure the importance of an individual ' s contribution to science ( Stephan , 1996 ) . Consistent with this view , citations are a critical currency of scientific credit ( Latour & Woolgar , 2013 ) , both driving research impact and constituting the basis of reward systems in science , including promotion , status , funding , peer esteem , honors , and awards ( Boudreau & Lakhani , 2015 ) . There are generally two views of how knowledge is diffused in science : openness and secrecy . According to the Mertonian norms of communalism or \u201c openness , \u201d publication enables scientists to establish priority of discovery and allows them to be the first to communi - cate an advance in knowledge and allow others to freely use it ( Merton , 1973 ) . Thus , publica - tion promotes the open diffusion of scientific knowledge , as long as scientists ' own internal agents ( i . e . , other scientists ) appropriately recognize and diffuse their work ( Boudreau & Lakhani , 2015 ) . On the other hand , social recognition is a discretionary act among scientists , and strong evidence points to the existence of counter - norms promoting secrecy , competition , and information withholding ( Haas & Park , 2010 ; Haeussler , Jiang , Thursby , & Thursby , 2014 ) . Scientists often compete for similar resources , funding , and recognition . Given limited resources , scientists need to be assured that they will be appropriately remunerated for openly diffusing others ' knowledge ( Hagstrom , 1974 ; Murray & O ' Mahony , 2007 ; Reschke , Azoulay , & Stuart , 2018 ) . On balance , the evidence suggests that scientists do not unequivocally diffuse each other ' s publications . Among scientists with greater knowledge overlap , serendipitous idea sharing may heighten the competition effect , particularly as information exchange can introduce people to novel work they cannot discover on their own via publications or other publicly available sources . These new projects may be similar to their own undertakings , risking duplication of efforts . Because scientists compete for priority ( Merton , 1973 ) , recognition ( Hagstrom , 1974 ; LANE ET AL . 9 Reskin , 1977 ) , and funding ( Stephan , 1996 ) , partners with highly overlapping interests may dis - cover a need to differentiate themselves from one another . Social recognition of highly similar peers would not only highlight redundancies but may also detract attention and resources away from one ' s own work ( Campanario & Acedo , 2007 ) . We expect that partners sharing high field or intellectual similarity may refrain from citing each other ' s research to outperform their peers . Hypothesis ( H3a ) Knowledge sharing is less likely to lead to knowledge diffusion as the field similarity between individuals increases . Hypothesis ( H3b ) Knowledge sharing is less likely to lead to knowledge diffusion as the intellec - tual similarity between individuals increases . 3 | EXPERIMENTAL METHODS In an ideal setting , all of the prior interactions and efforts that go into knowledge production would be fully observable to scholars to theorize and validate through empirical observations . The reality , however , is that the vast majority of prior scholarly work concerned with how knowledge is produced has primarily focused on observed outcomes \u2014 e . g . , papers , funding , pat - ents , citations , team structure , and so forth \u2014 to draw inferences about the mechanisms underly - ing the knowledge production process ( Dahlander & McFarland , 2013 ; Fleming et al . , 2007 ; Staw & Epstein , 2000 ) . For example , research on scientists will include the papers they publish , the collaborators they have worked with , and the knowledge that they have developed and have diffused through citations . However , a concern with relying on published trace data is that it masks all of the work and activities that occurred prior to knowledge production , such as the entire risk set of alters an individual may have interacted with prior to settling on a particular team . The empirical shortcomings of not being able to directly observe knowledge sharing and the drivers of knowledge production may insert biases in our inferences , such as self - selection and survivor bias . Some work aims to address these concerns by extending the risk set to unsuc - cessful collaborations ( e . g . , non - awarded grant applications ) to examine quality - adjusted research output of published ( i . e . , visible ) trace data ( Arora & Gambardella , 2005 ; Ayoubi , Pezzoni , & Visentin , 2019 ; Ganguli , 2017 ) . That said , unobserved heterogeneity between the two groups remains a persistent challenge . A feasible alternative is to design a field experiment that enables the capture of data around interactions between scientists and overcomes concerns around endogeneity of affiliation , team formation , and knowledge exposure by randomizing the encounters that the scientists have with each other . The benefit of this approach is that it can provide causal explanations about the factors that impact knowledge production , weighed against the challenges of drawing infer - ences with smaller sample sizes in an experiment as opposed to relying on all observed data of many more scientists . 3 . 1 | Setting and research design We carried out our study in the context of a medical symposium for research on advanced imag - ing , which is used to detect diseases and other health conditions early , to allow health care 10 LANE ET AL . practitioners to direct patients to the health care services that they need . We collaborated with the administrators of a large U . S . medical school to layer the medical symposium onto the University ' s Clinical and Translational Science Center pilot grant program , which provides seed funding in the form of pilot grants to support nascent research efforts that are awarded compet - itively to faculty within the University . 1 The purpose of the grant opportunity was to solicit pro - posals to improve methods for using advanced imaging technologies to address unmet clinical needs , and offered $ 50 , 000 per award for up to 15 pilot grants . A major challenge in the field of advanced imaging is that furthering the knowledge frontier requires expertise in the latest imaging tools and technologies and a deep understanding of the health problems to which they could be applied . These different types of knowledge are typically held by people from distinct disciplinary backgrounds , which makes advanced imaging an ideal setting for our research . In November 2011 , we invited all life sciences faculty and researchers at the university to a medical research symposium for the unique grant funding opportunity . Our field experiment involved faculty and researchers at the University and its affiliated hospitals and institutions , which are independently owned and managed , with each appearing as a separate entity in hospital rankings and lists of National Institutes of Health ( NIH ) grant recipients . We communicated to applicants that eligibility to submit a grant application was conditional on attending the symposium . In the first stage , investigators interested in applying for the grants submitted a statement of interest in which they briefly described a specific medical problem that advanced imaging techniques could potentially address . We collected basic bio - graphical information ( e . g . , degree , institution , department appointment ) at this stage . 3 . 2 | Participants and randomization of knowledge - sharing partners The symposia were held on January 31 , February 1 , and February 2 , 2012 , at one of the univer - sity ' s innovation labs . Figure 1 summarizes the key details of the randomization : 402 unique participants were randomly assigned to one of three nights of the symposium , one of four breakout rooms , one of two groups , as well as a poster location to stand next to around the perimeter of the room . 2 Participants were provided an electronic device , called a \u201c sociometric badge \u201d to automatically record their face - to - face interactions during the symposium ( Kim et al . , 2012 ) . Each of the three nights featured a 30 - min welcome address , followed by two 45 - min poster sessions in breakout rooms , with a 15 - min social break in between the poster sessions . Within each breakout room , scientists were randomly assigned to a poster location and either the first or second poster group . The scientists presented and exchanged ideas with one another during the poster sessions . The grant administrators prepared the posters to be a standard size and in a standard format , based on each participant ' s state - ments of interest . Thus , treatment pairs assigned to the same breakout rooms had more opportunities for knowledge sharing than control pairs . Shortly after the symposia , all participants received an e - mail invitation to submit applications for the pilot grants or concept awards by the deadline of March 8 , 2012 . At this time , they also 1 The same field experiment was used in Boudreau et al . ( 2017 ) to investigate the effect of search costs on collaborations in the immediate aftermath of the experiment . It did not include long - run knowledge production outcomes or leverage the sociometric badge data , described in Section 3 . 3 . 2 Compared to the entire Harvard Medical School ( HMS ) population ( N = 22 , 625 ) , participants were more likely to be PhDs , had more prior publications on average and were more likely to be instructors , assistant or associate professors relative to the overall distribution at HMS . These differences are consistent with the academic research setting of the symposium and the pilot grant funding opportunity . See Table A1 for summary statistics between participants and the HMS population . LANE ET AL . 11 received PDF booklets with the contact information and posters of all participants so that all researchers had identical information apart from the knowledge acquired in the breakout rooms . 3 . 3 | Data collection and variables We tested our hypotheses using data from a variety of resources from the advanced imaging symposium and the 6 years of publication records from 2013 to 2018 on the attendees . In the analyses , we did not include the year 2012 to remove potential research topics or ideas that were in progress prior to the sym - posium . We used data from the scientists ' registration form for the symposium , which contained infor - mation about their institution , department , academic position , self - identity as an imager or clinician , and statements of interest . The sociometric badges automatically recorded their face - to - face encounters when two badges were facing each other with a direct line of sight within a 30 (cid:1) cone of 1 m . We verified that all recorded interactions were within 10 m using Bluetooth proximity data ( Kim et al . , 2012 ) and required that two badges be in contact with each other over a span of at least 1 min ( Ingram & Morris , 2007 ) . We collected face - to - face interaction data for 306 ( 74 % ) scientists who attended the sym - posium , and the subsequent analyses are based on these scientists . 3 After the symposium , we collected information on the coapplicants and the awardees of the advanced imaging grants and used the Scopus database to collect scientists ' publication records . FIGURE 1 Randomization of participants by night , room , group , and poster location 3 Badges were randomly assigned to all symposium attendees . 26 % of badges malfunctioned and did not record face - to - face interaction data ; participants were not aware of whether they were assigned a working or faulty badge . For each symposium night , the sociometric badges recorded interactions for 64 . 4 , 68 . 2 , and 90 . 4 % of the participants ( Table A2 ) . We conducted balance checks on all available scientist covariates between the complete and observed sample and found no differences between the samples ( Table A3 ) . As robustness , we also perform the main OLS regression analyses for each knowledge production outcome on the full sample of participants to verify that the results are directionally consistent across the full sample and the observed badge sample . 12 LANE ET AL . 3 . 3 . 1 | Dependent variables Knowledge transfer We base our knowledge transfer measure on the MeSH lexicon . In the life sciences , the U . S . National Library of Medicine ( NLM ) uses a controlled MeSH taxonomy of keywords to index biomedical and health - related information for articles appearing in MEDLINE / PubMed , the NLM Catalog , and other NLM databases . Each article is associated with a set of MeSH key - words that describe the content of the citation . MeSH keywords are assigned by professional sci - ence librarians and computer algorithms to ensure global and consistent assignment of keywords across the life sciences ( Coletti & Bleich , 2001 ) . We extracted the unique MeSH key - words associated with each scientist ' s publications to create two vectors of MeSH keywords : the first with all MeSH keywords prior to the advanced imaging symposium ( i . e . , pre - 2012 ) , and the second with all MeSH keywords after the symposium ( i . e . , 2013 \u2013 2018 ) . For each scientist - pair { i , j } , we then counted the number of MeSH terms that j transferred to i and the number of MeSH terms that i transferred to j by taking the intersection of MeSH keywords between i ' s pre - 2012 vector and j ' s post - 2012 vector and between j ' s pre - 2012 vector and i ' s post - 2012 vector , excluding any MeSH terms that were common to both i and j in their pre - 2012 MeSH vectors . Our resulting knowledge transfer measure , MeSH keyword transfer is the count of \u201c transferred \u201d MeSH keywords between pair { i , j } , normalized by the total number of MeSH keywords in scien - tists i and j ' s post - 2012 MeSH vectors , expressed as a percentage . The resulting measure ranges from 0 to 100 % to and is interpreted as the percentage of postsymposium MeSH keywords that were transferred between scientist - pair { i , j } , with 0 % representing no transferred MeSH key - words and 100 % representing complete transfer . Knowledge creation We measure knowledge creation as the postsymposium ( 2013 \u2013 2018 ) count of Copublications between scientist - pair { i , j } in peer - reviewed journals , conference proceedings , and book chapters . Knowledge diffusion We measure knowledge diffusion as the postsymposium ( 2013 \u2013 2018 ) count of noncoauthored Forward citations between scientist - pair { i , j } in peer - reviewed journals , conference proceedings , and book chapters . 3 . 3 . 2 | Independent variables Knowledge sharing We use two alternative variables to capture knowledge sharing . First , we use the dummy vari - able , Same room , to measure whether scientist - pair { i , j } was randomly assigned to the same room ( treatment pairs ) or different breakout rooms ( control pairs ) . Second , we use the dummy variable , F2F ( face - to - face ) communication to measure whether scientist - pair { i , j } engaged in least 1 min of interaction , recorded using sociometric badges . Knowledge similarity Knowledge similarity is comprised of field and intellectual similarity . We measure Field similar - ity using clinical areas ( third - party coded from the scientists ' statements of interest ) , which LANE ET AL . 13 pertain to the primary area of responsibility for \u201c bedside \u201d patient care . There were a total of 24 unique clinical areas ( e . g . , oncology , neurology , immunology ) , and some statements of inter - ests ( 4 . 24 % ) spanned two clinical areas , such as neurology / endocrinology . We then use a cate - gorical variable to indicate whether scientist - pair { i , j } shared Low , Moderate or High field similarity depending on whether their clinical areas shared no ( 85 . 5 % ) , partial ( 1 . 9 % ) , or com - plete ( 12 . 6 % ) overlap , respectively . We measure Intellectual similarity using the count of common MeSH keywords shared by scientist - pair { i , j } prior to 2012 and dichotomize the distribution into three equal - sized groups , to indicate whether scientist - pair { i , j } had low , moderate , or a high number of common MeSH keywords prior to the 2012 symposium . Low intellectual similarity corresponds to 0 \u2013 2 common keywords , Moderate intellectual similarity corresponds to 3 \u2013 11 common keywords , and High intellectual similarity corresponds to more than 11 common keywords . Elemental collaboration We use Grant coapplicant to measure an elemental collaboration , which captured whether scientist - pair { i , j } coapplied on the grant following the symposium . 3 . 3 . 3 | Other variables The analysis strategy relies most critically on the research design ' s randomization . We use dummy variables for each symposium night and room ( i . e . , fixed effects ) to control for unobserved night and room characteristics . To test Hypotheses ( H2c ) and ( H2d ) , we use the dummy variable , Grant awardee to capture whether scientist - pair { i , j } included a grant recipi - ent . Among the 306 scientists , 13 pilot grant proposals were awarded funding , comprising 6 . 54 % of pairs with grant awardees . 3 . 4 | Estimation approach We wish to estimate the effects of field and intellectual similarity between knowledge sharing partners on knowledge production outcomes . The unit of analysis is the scientist - pair { i , j } , and pairs are considered to be \u201c at risk \u201d if they attended the same night of the symposium \u2014 a total of 15 , 817 pairs . First , we analyze the effect of being in the same ( treatment ) versus different ( control ) breakout rooms on knowledge production outcomes using ordinary least squares ( OLS ) regression . We then interact Same room with Field similarity , and Same room with Intel - lectual similarity to examine knowledge similarity effects . Second , we analyze the effect of face - to - face communication on knowledge production outcomes using instrumental variable ( IV ) regression . We use an IV approach to account for the common endogeneity issue in pairwise interaction data . We exploit exogenous variation in the likelihood of interaction between scientists who are assigned to the same versus different breakout rooms by using Same room as an instrument for F2F communication in the first stage , and the estimates for F2F com - munication in the second stage . We then interact F2F communication with Field similarity , and F2F communication with Intellectual similarity to examine knowledge similarity effects . We address the nonindependence , common - person problem of dyadic regressions by esti - mating robust SE s that are simultaneously clustered on both members of the dyad , using 14 LANE ET AL . multiway clustering , developed theoretically by Cameron , Gelbach , and Miller ( 2011 ) and implemented for Stata in clus _ nway . ado ( Kleinbaum et al . , 2013 ) . 4 | RESULTS In this section , we begin by presenting descriptives of the main variables . Table 2 presents the means , standard deviations , and correlations of the main variables . We also note that the sum - mary statistics of the covariates for same room versus different room pairs indicate that the ran - domization achieved balance across covariates ( see Table A4 , Supporting Information ) . Figure 2 shows the yearly trend in copublication ( left ) and forward citation ( right ) rates between different room ( N = 11 , 611 ) and same room ( N = 4 , 206 ) pairs by year , in the 6 years to and since the symposium , with 95 % CIs . The plots show that there were no differences between the same room and different room pairs before the knowledge sharing intervention . After the intervention , we observe an increase in copublication and citation rates for the same room pairs . Next , we present our main regression results in subsections , beginning with knowledge transfer , then creation , and finally , diffusion . We note that the F - statistics for the IV regressions are all above the threshold of 10 for strong instruments ( Table A5 , Supporting Information ) , and that our results are robust to the inclusion of all scientist - pair covariates ( Table A6 , Supporting Information ) and the reduced form ( OLS ) models for the full sample of participants ( Tables A7 \u2013 A9 , Supporting Information ) . 4 . 1 | Knowledge transfer results We present the OLS and IV regression knowledge transfer results in Table 3 . The dependent variable is the percentage of scientist - pair { i , j } ' s MeSH postsymposium keywords that were transferred between i and j , with 0 % being no transfer , and 100 % being complete transfer . TABLE 2 Correlation between main variables ( N = 15 , 817 ) Variable Mean SD 1 2 3 4 5 6 7 1 MeSH keyword transfer 3 . 693 3 . 709 2 Copublications 0 . 031 1 . 004 0 . 010 3 Forward citations 0 . 265 2 . 497 0 . 019 0 . 479 4 Grant coapplicant 0 . 002 0 . 044 0 . 004 0 . 097 0 . 027 5 Same room 0 . 266 0 . 442 0 . 024 \u2212 0 . 004 \u2212 0 . 001 0 . 006 6 F2F communication 0 . 078 0 . 268 0 . 003 0 . 047 0 . 039 0 . 067 0 . 198 7 Intellectual similarity 1 . 076 0 . 810 \u2212 0 . 005 0 . 037 0 . 043 0 . 037 \u2212 0 . 002 0 . 047 8 Field similarity 0 . 272 0 . 673 0 . 437 0 . 029 0 . 074 0 . 017 \u2212 0 . 008 0 . 021 0 . 026 Note : Intellectual and field similarity are categorical variables with the following distributions : intellectual ( low : 32 . 4 % , moderate : 32 . 7 % , high : 35 . 0 % ) and field ( low : 85 . 5 , 1 . 9 , and 12 . 6 % ) . LANE ET AL . 15 Models 1 and 2 present the baseline models with main effects only ; Models 3 and 4 add the field similarity interaction terms ; Models 5 and 6 add the intellectual similarity interaction terms ; and Models 7 and 8 show the full results with both interaction terms . Hypotheses ( H1a ) and ( H1b ) suggest that field and intellectual similarity have an inverted U - shaped effect on the relationship between knowledge sharing and knowledge transfer , respectively . Models 3 and 4 indicate that there is no meaningful association between moderate field similarity and knowledge transfer among same room pairs ( 0 . 712 , p = . 132 ) and communi - cating pairs ( 4 . 088 , p = . 197 ) . Models 5 and 6 show that compared to pairs with high intellectual overlap , scientists with moderate intellectual similarity transferred a higher percentage of MeSH keywords for both same room pairs ( 0 . 362 , p = . 042 ) and communicating pairs ( 3 . 348 , p = . 032 ) , respectively . These results are consistent in Models 7 and 8 , which include both inter - action terms . Figure 3a shows the change in knowledge transfer among communicating versus non - communicating pairs and intellectual similarity from Model 4 , with 95 % CIs , and illustrates the inverted U - shaped relationship . While communication did not meaningfully impact knowledge transfer among pairs with low field overlap , it led to an increase of 3 . 581 % among pairs with moderate intellectual overlap from 4 . 085 % [ 3 . 795 % , 4 . 376 % ] to 7 . 666 % [ 4 . 995 % , 10 . 338 % ] per - cent , a nearly twofold increase , compared to a 0 . 233 % [ \u2212 1 . 239 % , 1 . 705 % ] increase for high intel - lectual overlap pairs . The percentage increase among moderately overlapping pairs corresponds to about six new MeSH keywords per scientist , roughly equivalent to the average number of MeSH keywords in one publication . 4 This result supports Hypothesis ( H1b ) by showing cooper - ative effects of knowledge transfer for pairs with moderate intellectual similarity . 4 . 2 | Knowledge creation We present the OLS and IV knowledge creation regression results in Table 4 . The dependent variable is the number of copublications between scientist - pair { i , j } . Models 1 and 2 present the baseline models with main effects only ; Models 3 and 4 add the field similarity interaction FIGURE 2 Yearly trend in copublication ( left ) and forward citation ( right ) rates between same room ( treatment ) and different room ( control ) pairs with 95 % CIs 4 In the postsymposium period , each scientist has M = 172 . 57 ( SD = 142 . 61 ) MeSH keywords and M = 6 . 85 ( SD = 4 . 28 ) MeSH keywords per publication . 16 LANE ET AL . T A B L E 3 R e g r e ss i o n m o d e l s o f k n o w l e d g e t r a n s f e r \u2014 % o f M e S H k e y w o r d s t r a n s f e rr e d b e t w ee n s c i e n t i s t - p a i r { i , j } ; N = 15 , 817 V a r i a b l e s M o d e l 1 O L S M o d e l 2 I V M o d e l 3 O L S M o d e l 4 I V M o d e l 5 O L S M o d e l 6 I V M o d e l 7 O L S M o d e l 8 I V S a m e r oo m 0 . 107 ( 0 . 0979 ) 0 . 0344 ( 0 . 144 ) 0 . 0336 ( 0 . 138 ) \u2212 0 . 0379 ( 0 . 172 ) F 2 F c o mm u n i c a t i o n 0 . 838 ( 0 . 763 ) 0 . 214 ( 1 . 108 ) 0 . 233 ( 1 . 025 ) \u2212 0 . 512 ( 1 . 352 ) S a m e r oo m \u00d7 l o w f i e l d s i m il a r i t y 0 . 0707 ( 0 . 150 ) 0 . 0686 ( 0 . 150 ) S a m e r oo m \u00d7 m o d e r a t e f i e l d s i m il a r i t y 0 . 712 ( 0 . 471 ) 0 . 681 ( 0 . 468 ) S a m e r oo m \u00d7 l o w i n t e ll e c t u a l s i m il a r i t y \u2212 0 . 175 ( 0 . 218 ) \u2212 0 . 171 ( 0 . 218 ) S a m e r oo m \u00d7 m o d e r a t e i n t e ll e c t u a l s i m il a r i t y 0 . 362 ( 0 . 177 ) 0 . 361 ( 0 . 177 ) F 2 F \u00d7 l o w f i e l d s i m il a r i t y 0 . 621 ( 1 . 204 ) 0 . 804 ( 1 . 239 ) F 2 F \u00d7 m o d e r a t e f i e l d s i m il a r i t y 4 . 088 ( 3 . 169 ) 4 . 059 ( 3 . 379 ) F 2 F \u00d7 l o w i n t e ll e c t u a l s i m il a r i t y \u2212 1 . 459 ( 1 . 753 ) \u2212 1 . 531 ( 1 . 869 ) F 2 F \u00d7 m o d e r a t e i n t e ll e c t u a l s i m il a r i t y 3 . 348 ( 1 . 564 ) 3 . 305 ( 1 . 644 ) L o w f i e l d s i m il a r i t y 0 . 140 ( 0 . 135 ) 0 . 171 ( 0 . 135 ) 0 . 121 ( 0 . 143 ) 0 . 104 ( 0 . 183 ) 0 . 141 ( 0 . 135 ) 0 . 176 ( 0 . 137 ) 0 . 123 ( 0 . 143 ) 0 . 0900 ( 0 . 189 ) M o d e r a t e f i e l d s i m il a r i t y \u2212 0 . 130 ( 0 . 274 ) \u2212 0 . 136 ( 0 . 264 ) \u2212 0 . 316 ( 0 . 269 ) \u2212 0 . 606 ( 0 . 381 ) \u2212 0 . 136 ( 0 . 272 ) \u2212 0 . 174 ( 0 . 274 ) \u2212 0 . 315 ( 0 . 268 ) \u2212 0 . 638 ( 0 . 398 ) L o w i n t e ll e c t u a l s i m il a r i t y \u2212 4 . 130 ( 0 . 238 ) \u2212 4 . 120 ( 0 . 237 ) \u2212 4 . 130 ( 0 . 238 ) \u2212 4 . 123 ( 0 . 238 ) \u2212 4 . 081 ( 0 . 246 ) \u2212 4 . 015 ( 0 . 282 ) \u2212 4 . 082 ( 0 . 246 ) \u2212 4 . 013 ( 0 . 289 ) LANE ET AL . 17 T A B L E 3 ( C o n t i n u e d ) V a r i a b l e s M o d e l 1 O L S M o d e l 2 I V M o d e l 3 O L S M o d e l 4 I V M o d e l 5 O L S M o d e l 6 I V M o d e l 7 O L S M o d e l 8 I V M o d e r a t e i n t e ll e c t u a l s i m il a r i t y \u2212 0 . 880 ( 0 . 153 ) \u2212 0 . 862 ( 0 . 152 ) \u2212 0 . 880 ( 0 . 153 ) \u2212 0 . 866 ( 0 . 152 ) \u2212 0 . 974 ( 0 . 163 ) \u2212 1 . 095 ( 0 . 196 ) \u2212 0 . 974 ( 0 . 163 ) \u2212 1 . 096 ( 0 . 204 ) N i g h t F E Y Y Y Y Y Y Y Y R oo m F E Y Y Y Y Y Y Y Y R - s q u a r e d . 224 . 219 . 224 . 218 . 224 . 201 . 224 . 216 N o t e : M u l t i w ay , r o b u s t S E s i n p a r e n t h e s e s . S i g n i f i c a n c e s t a r s ( * ) a r e o m i tt e d . 18 LANE ET AL . terms ; Models 5 and 6 add the intellectual similarity interaction terms ; Models 7 and 8 show the full results with both interaction terms ; Model 9 \u2013 11 adds Grant coapplicant , followed by the field and intellectual similarity interaction terms , as well as both interaction terms , and controls for Grant awardee . Hypotheses ( H2a ) and ( H2b ) predict that field and intellectual similarity have an inverted U - shaped effect on the relationship between knowledge sharing and knowledge creation , respectively , while Hypotheses ( H2c ) and ( H2d ) predict that knowledge sharing partners who start an elemental collaboration are more likely to copublish when they share greater field and intellectual overlap , respectively . Models 3 and 4 show that compared to same field pairs , mod - erate field similarity has a strong positive relationship with knowledge creation among same room pairs ( 0 . 329 , p = . 047 ) and communicating pairs ( 2 . 159 , p = . 036 ) . There is also a smaller , positive association between low field similarity and knowledge creation among same room pairs ( 0 . 133 , p = . 092 ) and communicating pairs ( 1 . 040 , p = . 101 ) . Turning to intellectual simi - larity , Models 5 and 6 show there is no meaningful association between moderate intellectual similarity and knowledge creation among same room pairs ( 0 . 0409 , p = . 257 ) and communicat - ing pairs ( 0 . 333 , p = . 234 ) . The results are consistent in Models 7 and 8 , which includes both interaction terms . Figure 3b plots the change in copublications between communicating and noncommunicating pairs and field similarity , with 95 % CIs from Model 3 , and illustrates the estimated inverted U - shaped relationship . We observe that while communication did not meaningfully benefit pairs with low field overlap , communication led to an increase of 1 . 174 [ \u2212 0 . 207 , 2 . 556 ] copublications among pairs with moderate field overlap , and \u2212 0 . 984 [ \u2212 1 . 731 , \u2212 0 . 237 ] fewer copublications among pairs with high field overlap . The patterns suggest that communication reduced the search costs of identi - fying synergistic collaborations with scientists sharing some field interests , resulting in a reallocation of resources away from pairs with high field overlap . We also examined the extent that the knowledge sharing intervention impacted the scien - tists ' overall research portfolios . Turning to the scientists ' postsymposium publications , each sci - entist had an average of M = 32 . 394 ( SD = 35 . 412 ) peer - reviewed research articles ; the increase of 1 . 174 publications among pairs with moderate field overlap corresponds to about 7 % of a sci - entist ' s publications , which is of considerable magnitude from a 90 min intervention . Turning FIGURE 3 Change in ( a ) knowledge transfer , ( b ) knowledge creation , and ( c ) knowledge diffusion and knowledge similarity for communicating and noncommunicating pairs LANE ET AL . 19 T A B L E 4 R e g r e ss i o n m o d e l s o f k n o w l e d g e c r e a t i o n \u2014 # o f c o p u b li c a t i o n s b e t w ee n s c i e n t i s t - p a i r { i , j } ; N = 15 , 817 V a r i a b l e s M o d e l 1 O L S M o d e l 2 I V M o d e l 3 O L S M o d e l 4 I V M o d e l 5 O L S M o d e l 6 I V M o d e l 7 O L S M o d e l 8 I V M o d e l 9 O L S M o d e l 10 O L S M o d e l 11 O L S S a m e r oo m \u2212 0 . 00683 ( 0 . 0144 ) \u2212 0 . 127 ( 0 . 076 9 ) \u2212 0 . 0294 ( 0 . 0360 ) \u2212 0 . 148 ( 0 . 0931 ) \u2212 0 . 00784 ( 0 . 0148 ) \u2212 0 . 00822 ( 0 . 0148 ) \u2212 0 . 00799 ( 0 . 0147 ) F 2 F c o mm u n i c a t i o n \u2212 0 . 0535 ( 0 . 111 ) \u2212 0 . 984 ( 0 . 616 ) \u2212 0 . 221 ( 0 . 268 ) \u2212 1 . 147 ( 0 . 746 ) S a m e r oo m \u00d7 l o w f i e l d s i m . 0 . 133 ( 0 . 0788 ) 0 . 132 ( 0 . 0783 ) S a m e r oo m \u00d7 m o d . f i e l d s i m . 0 . 329 ( 0 . 165 ) 0 . 330 ( 0 . 164 ) S a m e r oo m \u00d7 l o w i n t e ll e c t u a l s i m . 0 . 0306 ( 0 . 0378 ) 0 . 0307 ( 0 . 0372 ) S a m e r oo m \u00d7 m o d . i n t e ll e c t u a l s i m . 0 . 0409 ( 0 . 0361 ) 0 . 0397 ( 0 . 0355 ) F 2 F \u00d7 l o w f i e l d s i m . 1 . 040 ( 0 . 634 ) 1 . 048 ( 0 . 637 ) F 2 F \u00d7 m o d . f i e l d s i m . 2 . 159 ( 1 . 029 ) 2 . 110 ( 1 . 022 ) F 2 F \u00d7 l o w i n t e l l e c t u a l s i m . 0 . 240 ( 0 . 297 ) 0 . 205 ( 0 . 306 ) F 2 F \u00d7 m o d . i n t e l l e c t u a l s i m . 0 . 333 ( 0 . 280 ) 0 . 332 ( 0 . 290 ) G r a n t c o a pp l i c a n t 3 . 708 ( 1 . 583 ) 2 . 482 ( 1 . 211 ) 3 . 810 ( 1 . 614 ) C o a pp l i c a n t \u00d7 l o w f i e l d s i m . \u2212 2 . 427 ( 1 . 845 ) \u2212 2 . 299 ( 2 . 109 ) C o a pp l i c a n t \u00d7 m o d . f i e l d S i m . \u2212 3 . 303 ( 1 . 621 ) \u2212 3 . 405 ( 1 . 659 ) C o a pp l i c a n t \u00d7 l o w i n t e l l e c t u a l s i m . \u2212 0 . 639 ( 2 . 085 ) \u2212 0 . 324 ( 2 . 236 ) 20 LANE ET AL . T A B L E 4 ( C o n t i n u e d ) V a r i a b l e s M o d e l 1 O L S M o d e l 2 I V M o d e l 3 O L S M o d e l 4 I V M o d e l 5 O L S M o d e l 6 I V M o d e l 7 O L S M o d e l 8 I V M o d e l 9 O L S M o d e l 10 O L S M o d e l 11 O L S C o a pp l i c a n t \u00d7 m o d . i n t e l l e c t u a l s i m . \u2212 1 . 083 ( 1 . 482 ) \u2212 0 . 571 ( 1 . 632 ) A w a r d ee 0 . 000295 ( 0 . 0140 ) \u2212 0 . 00167 ( 0 . 0137 ) \u2212 2 . 47 e - 05 ( 0 . 0140 ) L o w f i e l d s i m . \u2212 0 . 109 ( 0 . 0540 ) \u2212 0 . 111 ( 0 . 0566 ) \u2212 0 . 14 ( 0 . 073 5 ) \u2212 0 . 220 ( 0 . 120 ) \u2212 0 . 109 ( 0 . 0539 ) \u2212 0 . 112 ( 0 . 0579 ) \u2212 0 . 144 ( 0 . 0733 ) \u2212 0 . 222 ( 0 . 121 ) 0 . 0552 ( 0 . 0422 ) 0 . 0405 ( 0 . 0441 ) 0 . 0552 ( 0 . 0422 ) M o d e r a t e f i e l d s i m . \u2212 0 . 0529 ( 0 . 0659 ) \u2212 0 . 0525 ( 0 . 0655 ) \u2212 0 . 139 ( 0 . 070 1 ) \u2212 0 . 297 ( 0 . 132 ) \u2212 0 . 0534 ( 0 . 0660 ) \u2212 0 . 0567 ( 0 . 0675 ) \u2212 0 . 140 ( 0 . 0706 ) \u2212 0 . 296 ( 0 . 130 ) 0 . 0887 ( 0 . 0536 ) 0 . 0981 ( 0 . 0539 ) 0 . 0887 ( 0 . 0536 ) L o w i n t e l l e c t u a l s i m . \u2212 0 . 0696 ( 0 . 0222 ) \u2212 0 . 0702 ( 0 . 0226 ) \u2212 0 . 0695 ( 0 . 022 1 ) \u2212 0 . 0738 ( 0 . 0240 ) \u2212 0 . 0777 ( 0 . 0297 ) \u2212 0 . 0908 ( 0 . 0437 ) \u2212 0 . 0777 ( 0 . 0294 ) \u2212 0 . 0916 ( 0 . 0455 ) \u2212 0 . 0651 ( 0 . 0222 ) \u2212 0 . 0643 ( 0 . 0224 ) \u2212 0 . 0645 ( 0 . 0224 ) M o d e r a t e i n t e ll e c t u a l s i m . \u2212 0 . 0657 ( 0 . 0209 ) \u2212 0 . 0669 ( 0 . 0219 ) \u2212 0 . 0656 ( 0 . 0207 ) \u2212 0 . 0699 ( 0 . 0231 ) \u2212 0 . 0765 ( 0 . 0272 ) \u2212 0 . 0928 ( 0 . 0399 ) \u2212 0 . 0760 ( 0 . 0268 ) \u2212 0 . 0954 ( 0 . 0418 ) \u2212 0 . 0602 ( 0 . 0209 ) \u2212 0 . 0591 ( 0 . 0209 ) \u2212 0 . 0593 ( 0 . 0210 ) N i g h t F E Y Y Y Y Y Y Y Y Y Y Y R o o m F E Y Y Y Y Y Y Y Y Y Y Y R - s q u a r e d . 003 . 002 . 004 \u2014 . 003 \u2014 . 004 . 012 . 015 . 012 . 015 N o t e : M u l t i w ay r o b u s t S E s i n p a r e n t h e s e s . S i g n i f i c a n c e s t a r s ( * ) a r e o m i tt e d . LANE ET AL . 21 to changes in research direction , we also note that same room pairs were more likely to publish in advanced imaging journals ( e . g . , Magnetic Resonance in Medicine , NMR in Biomedicine ) : 21 % for same room versus 6 % for different room pairs ( t = 2 . 302 , p = . 0249 ) . 5 This suggests that the knowledge sharing treatment not only reduced search costs of finding collaborators but also potentially reshaped the pairs ' research trajectories . Turning to Hypothesis ( H2c ) , Model 9 shows that controlling for Grant awardee , grant coapplicants are more likely to copublish when they are from the same field ( low : \u2212 2 . 427 , p = . 189 ; moderate : \u2212 3 . 303 , p = . 042 ) . Finally , turning to H2d , Model 10 indicates that intel - lectual similarity is not a meaningful predictor of elemental tie persistence ( low : \u2212 0 . 639 , p = . 759 ; moderate : \u2212 1 . 083 , p = . 465 ) . The results remain consistent in Model 11 , which includes both interaction terms . This suggests that although strong field overlap is a catalyz - ing force for initiating early - stage collaborations , these requirements may dissipate over the long - run as these ties persist into copublications . Thus , our results show support for Hypotheses ( H2a ) and ( H2c ) . 4 . 3 | Knowledge diffusion We present the OLS and IV knowledge diffusion regression results in Table 5 . The dependent variable is the number of forward citations between scientist - pair { i , j } . Models 1 and 2 present the baseline models with main effects only ; Models 3 and 4 add the field similarity interaction terms ; Models 5 and 6 add the intellectual similarity interaction terms ; Models 7 and 8 show the full results with both interaction terms . Hypotheses ( H3a ) and ( H3b ) predict that knowledge sharing partners are less likely to dif - fuse knowledge as their field and intellectual similarity increase , respectively . Models 3 and 4 show that increasing field similarity reduced knowledge diffusion among same room pairs ( low : 0 . 366 , p = . 024 ; moderate : 1 . 003 , p = . 098 ) and communicating pairs ( low : 2 . 823 , p = . 033 ; moderate : 6 . 603 , p = . 050 ) . Models 5 and 6 show that there is no evidence that high intellectual similarity lowered knowledge diffusion among same room ( low : \u2212 0 . 0487 , p = . 739 ; moderate : \u2212 0 . 003 , p = . 986 ) or communicating pairs ( low : \u2212 0 . 426 , p = . 712 ; moderate : \u2212 0 . 091 , p = . 936 ) . The results are consistent in Models 7 and 8 , which include both knowledge similarity interac - tion terms . Figure 3c plots the change in knowledge diffusion ( forward citations ) between communicat - ing and noncommunicating pairs from Model 4 , with 95 % CIs . We observe that while commu - nication did alter forward citation trends for pairs with low field overlap , and led to a small increase of 3 . 521 [ \u2212 1 . 465 , 8 . 508 ] citations for pairs with moderate field overlap , we observe a large decrease from 0 . 874 [ 0 . 347 , 1 . 402 ] to \u2212 2 . 207 [ \u2212 4 . 326 , \u2212 0 . 088 ] for same field pairs , corresponding to a reduction of about 3 . 082 citations per pair . To generate greater insight into the types of publications being cited among the low and moderate field overlap pairs , we compared the publication titles of cited papers for pairs assigned to the same versus different rooms . We observe that the same room pairs were more likely to cite papers using advanced imaging technologies ( e . g . , sample paper titles included : \u201c Massively parallel MRI detector arrays \u201d ; \u201c An fMRI study of facial emotion processing patients 5 A journal was coded as an advanced imaging if the journal title included any of the three technologies featured at the symposium ( physiological magnetic resonance , positron emission tomography , and optical imaging ) . There were 48 unique journals for same room versus 169 unique journals for the different room pairs . 22 LANE ET AL . T A B L E 5 R e g r e ss i o n m o d e l s o f k n o w l e d g e d i ff u s i o n \u2014 # o f f o r w a r d c i t a t i o n s b e t w ee n s c i e n t i s t - p a i r { i , j } ; N = 15 , 817 V a r i a b l e s M o d e l 1 O L S M o d e l 2 I V M o d e l 3 O L S M o d e l 4 I V M o d e l 5 O L S M o d e l 6 I V M o d e l 7 O L S M o d e l 8 I V S a m e r oo m \u2212 0 . 0687 ( 0 . 0710 ) \u2212 0 . 399 ( 0 . 165 ) \u2212 0 . 0540 ( 0 . 150 ) \u2212 0 . 384 ( 0 . 211 ) F 2 F c o mm u n i c a t i o n \u2212 0 . 537 ( 0 . 551 ) \u2212 3 . 082 ( 1 . 331 ) \u2212 0 . 391 ( 1 . 115 ) \u2212 2 . 910 ( 1 . 656 ) S a m e r oo m \u00d7 l o w f i e l d s i m il a r i t y 0 . 366 ( 0 . 162 ) 0 . 367 ( 0 . 162 ) S a m e r oo m \u00d7 m o d e r a t e f i e l d s i m il a r i t y 1 . 003 ( 0 . 605 ) 0 . 999 ( 0 . 605 ) S a m e r oo m \u00d7 l o w i n t e ll e c t u a l s i m il a r i t y \u2212 0 . 0487 ( 0 . 146 ) \u2212 0 . 0476 ( 0 . 146 ) S a m e r oo m \u00d7 l o w i n t e ll e c t u a l s i m il a r i t y \u2212 0 . 00251 ( 0 . 145 ) \u2212 0 . 00592 ( 0 . 145 ) F 2 F \u00d7 l o w f i e l d s i m il a r i t y 2 . 823 ( 1 . 324 ) 2 . 835 ( 1 . 332 ) F 2 F \u00d7 m o d e r a t e f i e l d s i m il a r i t y 6 . 603 ( 3 . 364 ) 6 . 679 ( 3 . 334 ) F 2 F \u00d7 l o w i n t e ll e c t u a l s i m il a r i t y \u2212 0 . 426 ( 1 . 154 ) \u2212 0 . 537 ( 1 . 162 ) F 2 F \u00d7 m o d e r a t e i n t e ll e c t u a l s i m il a r i t y \u2212 0 . 0908 ( 1 . 138 ) \u2212 0 . 112 ( 1 . 135 ) L o w f i e l d s i m il a r i t y \u2212 0 . 316 ( 0 . 129 ) \u2212 0 . 336 ( 0 . 137 ) \u2212 0 . 413 ( 0 . 165 ) \u2212 0 . 633 ( 0 . 256 ) \u2212 0 . 316 ( 0 . 129 ) \u2212 0 . 335 ( 0 . 140 ) \u2212 0 . 413 ( 0 . 165 ) \u2212 0 . 632 ( 0 . 258 ) M o d e r a t e f i e l d s i m il a r i t y \u2212 0 . 209 ( 0 . 208 ) \u2212 0 . 205 ( 0 . 214 ) \u2212 0 . 471 ( 0 . 185 ) \u2212 0 . 955 ( 0 . 360 ) \u2212 0 . 209 ( 0 . 208 ) \u2212 0 . 204 ( 0 . 215 ) \u2212 0 . 471 ( 0 . 186 ) \u2212 0 . 962 ( 0 . 358 ) L o w i n t e ll e c t u a l s i m il a r i t y \u2212 0 . 446 ( 0 . 0920 ) \u2212 0 . 452 ( 0 . 0926 ) \u2212 0 . 446 ( 0 . 0918 ) \u2212 0 . 462 ( 0 . 0950 ) \u2212 0 . 433 ( 0 . 102 ) \u2212 0 . 417 ( 0 . 141 ) \u2212 0 . 433 ( 0 . 102 ) \u2212 0 . 418 ( 0 . 143 ) M o d e r a t e i n t e ll e c t u a l s i m il a r i t y \u2212 0 . 353 ( 0 . 0762 ) \u2212 0 . 364 ( 0 . 0791 ) \u2212 0 . 352 ( 0 . 0760 ) \u2212 0 . 373 ( 0 . 0810 ) \u2212 0 . 352 ( 0 . 0866 ) \u2212 0 . 354 ( 0 . 128 ) \u2212 0 . 351 ( 0 . 0862 ) \u2212 0 . 361 ( 0 . 130 ) LANE ET AL . 23 T A B L E 5 ( C o n t i n u e d ) V a r i a b l e s M o d e l 1 O L S M o d e l 2 I V M o d e l 3 O L S M o d e l 4 I V M o d e l 5 O L S M o d e l 6 I V M o d e l 7 O L S M o d e l 8 I V N i g h t F E Y Y Y Y Y Y Y Y R oo m F E Y Y Y Y Y Y Y Y R - s q u a r e d . 011 . 004 . 011 \u2014 . 011 . 004 . 011 \u2014 N o t e : M u l t i w ay , r o b u s t S E s i n p a r e n t h e s e s . S i g n i f i c a n c e s t a r s ( * ) a r e o m i tt e d . 24 LANE ET AL . with schizophrenia \u201d ) . Overall , there was a greater use of advanced imaging technologies in the publication titles of same room pairs : 42 . 6 % for same room pairs versus 31 . 3 % for different room pairs ( t = 2 . 406 , p = . 017 ) . 6 This suggests that the knowledge sharing intervention enabled pairs from different fields to learn about \u201c outside \u201d work , which they subsequently integrated into their own future research . In contrast , the knowledge sharing intervention may have led to a crowding out effect among highly similar pairs . In summary , we find evidence for competitive effects of field similarity on knowledge diffusion , which confirms Hypothesis ( H3a ) . 5 | DISCUSSION The ability to produce and manage knowledge is a critical source of competitive advantage for organizations of all types ( Argote , 2012 ; Hargadon & Sutton , 1997 ) . Yet the knowledge produc - tion process is both time - intensive and uncertain ( Maggitti , Smith , & Katila , 2013 ) . To address this issue , the premise of this article is based on the notion that \u201c engineering serendipity \u201d can promote greater knowledge sharing and more efficient knowledge production . We systemati - cally examine how the degree of field and intellectual similarity between knowledge sharing partners affects the likelihood that engineered encounters affects three knowledge production outcomes : knowledge transfer , creation , and diffusion . We find both cooperative and competitive effects of serendipitous knowledge sharing on knowledge production . On the one hand , knowledge sharing leads to greater knowledge trans - fer and creation when people share some overlapping research interests : knowledge sharing partners are more likely to transfer and acquire MeSH keywords from one another when they already have some intellectual overlap and more likely to copublish when they share field over - lap , with field similarity being more critical to collaborations in the near - term than over longer horizons . On the other hand , knowledge sharing appears to have little benefit for pairs with low knowledge similarity and reduces knowledge diffusion between people from the same field : we observe that pairs from the same field are less likely to cite each other after interacting . These findings make a number of contributions to the literature . First , we contribute to the literature on search costs and information frictions associated with knowledge production ( Agrawal & Goldfarb , 2008 ; Boudreau et al . , 2017 ; Catalini , 2018 ; Hansen , 1999 ; Szulanski , 1996 ) , by focusing on the extent that \u201c engineered serendipity \u201d can mitigate some of these barriers . Engineered , serendipitous encounters differ from traditional knowledge produc - tion mechanisms because they give people a common time and space to meet , thereby removing some of the search costs associated with finding suitable knowledge sharing partners . Our work suggests that engineered serendipity creates opportunities for synergistic collaborations over the long - run that have the potential to broaden collaboration networks and reshape research trajec - tories . Consistent with the information frictions explanation , knowledge overlap is a greater constraint on near - term than long - term knowledge creation . Knowledge similarity can be a cat - alyzing force for near - term research activities because there are fewer start - up costs and uncer - tainties associated with initiating collaborations with similar others , such as people working in the same field . In Boudreau et al . ( 2017 ) , we found that the knowledge sharing intervention ' s effect on grant coapplications was most beneficial for pairs from the same field . This suggests 6 There were a total of 149 ( same room ) and 415 ( different room ) forward citations for pairs with low / moderate field overlap , respectively . A publication title was coded as advanced imaging if it included any of the three technologies featured at the symposium ( physiological magnetic resonance , positron emission tomography , and optical imaging ) . LANE ET AL . 25 that in the short - term , opportunity and discretion remain important factors driving knowledge production ( Dahlander & McFarland , 2013 ) . We expand on these findings to show that while field overlap remains a difficult hurdle to overcome in the near - term , over time , scientists place smaller premiums on knowledge similarity , instead emphasizing synergistic collaborations where partners have both common knowledge and divergent expertise . This shift may be due to the time and repeated interactions needed to develop an initial encounter into a multiplex rela - tionship ( Ingram & Morris , 2007 ; Uzzi , 1997 ) , or the evolving expectations from collaborators as projects move from exploratory research to joint execution of research ideas . They also suggest that homophilous collaborations ( Dahlander & McFarland , 2013 ) can be attributed in part to search costs and information costs . Second , this study advances understanding of how opportunities for temporary colocation , offered by conferences , symposia and similar events potentially impact the direction and quality of scientific knowledge production . Although prior research suggests that temporary colocation is critical to knowledge production ( Campos et al . , 2018 ; Catalini , 2018 ; Catalini et al . , 2020 ; Chai & Freeman , 2019 ) , it has typically inferred face - to - face interaction from observational data , rather than actual acts of knowledge sharing . We suggest that one attractive aspect of con - ferences over informal serendipitous encounters ( e . g . , in hallways , watercoolers ) is that they have a focusing effect on the topic of knowledge sharing and idea exchange . This may be why even brief , highly structured , information - rich interventions can have long - term consequences on knowledge production , most notably by promoting research activities on the conference ' s topics of inquiry . It is important to note that the effects of these events go beyond monetary incentives , which may be an early catalyst to attract participants but the effects are more wide - spread and longer lasting . Third , we make empirical and methodological contributions by designing a prospective study that amalgamates multiple sources and uses of data , namely a natural field experiment , direct observations of face - to - face communications and archival publication data to study long - run knowledge production outcomes . This careful combination of data sources enables us to investigate the causal relationships between knowledge sharing and knowledge production over different time horizons , which are not possible with archival data ( Fleming et al . , 2007 ) . Given the rise in field experiments in innovation research ( Chatterji , Delecourt , Hasan , & Koning , 2019 ) , research designs that at the outset plan for short - and long - term time horizons may yield deeper insights that can help to offset the costs associated with multiple forms of data collection . In addition , our research design can be used as a template for future experiments with multiterm horizons , in terms of learning from its features and its potential shortcomings . In this research , we have made a thorough effort to analyze how knowledge sharing leads to knowledge production in science . However , our study has some empirical limitations . First , our study focused on an interdisciplinary setting in a large , highly selective university . Therefore , it may be interpreted as a best - case scenario in fostering knowledge production , as we must draw a boundary around the network and incentives under consideration . The symposium was highly structured to facilitate knowledge sharing , while knowledge production is a core activity for academics . Second , the geographic proximity of scientists may have facilitated scheduled or serendipitous encounters with greater ease ( Catalini , 2018 ) . Third , our focus on the life sciences also draws upon a specific population , where intragroup competition is normative ( Haas & Park , 2010 ) , in part due to the \u201c winner - takes - all \u201d model of rewards and recognition ( Stephan , 1996 ) . We recognize that other settings may feature greater geographic dispersion or promote more openness and cooperation ; a promising future direction would be to extend this research to different settings . 26 LANE ET AL . We also focused most heavily on archival data ( e . g . , public trace data from publications ) to examine long - term knowledge production outcomes , which provided greater insights into the scientists ' behaviors rather than the reasons behind the observed patterns . Although we made efforts to qualify these findings by examining the content of the knowledge transfer , publica - tions , and citations , future research can seek to supplement archival data with alternative methods ( e . g . , surveys , interviews ) that continue to track the scientists ' interaction patterns , as well as the invisible track of \u201c failed \u201d knowledge production . Such complementary data sources will ultimately provide deeper insights on how engineered serendipitous encounters between people can be cultivated into productive relationships over time . This future work would facili - tate a deeper performative assessment of how knowledge sharing can be \u201c serendipitously engineered \u201d to shape the quality of knowledge outcomes . Overall , this study takes a critical step towards identifying the processes that explain when serendipitous encounters shape knowledge production outcomes among innovating individuals . We show that brief , information - rich interactions between people with some overlapping knowledge interests can have a productive effect on knowledge transfer , creation and diffusion . ACKNOWLEDGEMENTS This work was conducted with support from Harvard Catalyst / The Harvard Clinical and Trans - lational Science Center ( National Center for Advancing Translational Sciences , National Insti - tutes of Health Awards UL1TR001102 , UL1TR000170 , UL1RR025758 - 02S4 , and UL1TR002541 ) , Laboratory for Innovation Science at Harvard University , Harvard Business School Division of Research and Faculty Development , and financial contributions from Harvard University and its affiliated academic health care centers . The authors also acknowledge Harvard Catalyst infrastructure for support and cooperation in implementing the experiment and providing new data for this analysis . The authors are thankful to Charles Ayoubi , Ethan Bernstein , Linus Dahlander , Giada Di Stefano , Adam Kleinbaum , Rory McDonald , Siobhan O ' Mahony , Ryan Raffaelli , Misha Teplitskiy , and the review team at SMJ for comments on earlier drafts . The authors would also like to thank Ruihan Wang for help with data collection . ORCID Jacqueline N . Lane https : / / orcid . org / 0000 - 0002 - 3744 - 9149 Ina Ganguli https : / / orcid . org / 0000 - 0002 - 0691 - 5709 Patrick Gaule https : / / orcid . org / 0000 - 0002 - 8712 - 0405 Eva Guinan https : / / orcid . org / 0000 - 0003 - 3874 - 0007 Karim R . Lakhani https : / / orcid . org / 0000 - 0002 - 5535 - 8304 REFERENCES Abbott , A . ( 2001 ) . Time matters : On theory and method , Chicago , IL : University of Chicago Press . Adams , J . D . , Black , G . C . , Clemmons , J . R . , & Stephan , P . E . ( 2005 ) . Scientific teams and institutional collabora - tions : Evidence from US universities , 1981 \u2013 1999 . Research Policy , 34 ( 3 ) , 259 \u2013 285 . Agrawal , A . , & Goldfarb , A . ( 2008 ) . Restructuring research : Communication costs and the democratization of university innovation . American Economic Review , 98 ( 4 ) , 1578 \u2013 1590 . Allen , T . J . ( 1977 ) . Managing the flow of technology : Technology transfer and the dissemination of technological information within the R & D organisation . Cambridge , MA : The Massachusetts Institute of Technology . Argote , L . ( 2012 ) . Organizational learning : Creating , retaining and transferring knowledge . New York , NY : Springer Science & Business Media . Arora , A . , & Gambardella , A . ( 2005 ) . The impact of NSF support for basic research in economics . Annales d ' Economie et de Statistique , 79 - 80 , 91 \u2013 117 . LANE ET AL . 27 Ayoubi , C . , Pezzoni , M . , & Visentin , F . ( 2019 ) . The important thing is not to win , it is to take part : What if scien - tists benefit from participating in research grant competitions ? Research Policy , 48 ( 1 ) , 84 \u2013 97 . Azoulay , P . , Ding , W . , & Stuart , T . ( 2007 ) . The determinants of faculty patenting behavior : Demographics or opportunities ? Journal of Economic Behavior & Organization , 63 ( 4 ) , 599 \u2013 623 . Bechky , B . A . ( 2011 ) . Making organizational theory work : Institutions , occupations , and negotiated orders . Orga - nization Science , 22 ( 5 ) , 1157 \u2013 1167 . Biancani , S . , Dahlander , L . , McFarland , D . A . , & Smith , S . ( 2018 ) . Superstars in the making ? The broad effects of interdisciplinary centers . Research Policy , 47 ( 3 ) , 543 \u2013 557 . Biancani , S . , McFarland , D . A . , & Dahlander , L . ( 2014 ) . The semiformal organization . Organization Science , 25 ( 5 ) , 1306 \u2013 1324 . Boudreau , K . J . , Guinan , E . C . , Lakhani , K . R . , & Riedl , C . ( 2016 ) . Looking across and looking beyond the knowl - edge frontier : Intellectual distance , novelty , and resource allocation in science . Management Science , 62 ( 10 ) , 2765 \u2013 2783 . Boudreau , K . J . , & Lakhani , K . R . ( 2015 ) . \u201c Open \u201d disclosure of innovations , incentives and follow - on reuse : The - ory on processes of cumulative innovation and a field experiment in computational biology . Research Policy , 44 ( 1 ) , 4 \u2013 19 . Boudreau , K . J . , Brady , T . , Ganguli , I . , Gaule , P . , Guinan , E . , Hollenberg , A . , & Lakhani , K . ( 2017 ) . A field exper - iment on search costs and the formation of scientific collaborations . Review of Economics and Statistics , 99 ( 4 ) , 565 \u2013 576 . Bourdieu , P . ( 1984 ) . Distinction : A social critique of the judgement of taste . Cambridge , MA : Harvard University Press . Bower , G . H . , & Hilgard , E . R . ( 1981 ) . Theories of learning . Englewood Cliffs , NJ : Prentice - Hall . Cameron , A . C . , Gelbach , J . B . , & Miller , D . L . ( 2011 ) . Robust inference with multiway clustering . Journal of Business & Economic Statistics , 29 ( 2 ) , 238 \u2013 249 . Campanario , J . M . , & Acedo , E . ( 2007 ) . Rejecting highly cited papers : The views of scientists who encounter resistance to their discoveries from other scientists . Journal of the American Society for Information Science and Technology , 58 ( 5 ) , 734 \u2013 743 . Campos , R . , Leon , F . , & McQuillin , B . ( 2018 ) . Lost in the storm : The academic collaborations that went missing in hurricane ISSAC . The Economic Journal , 128 ( 610 ) , 995 \u2013 1018 . Carlile , P . R . ( 2004 ) . Transferring , translating , and transforming : An integrative framework for managing knowl - edge across boundaries . Organization science , 15 ( 5 ) , 555 \u2013 568 . Catalini , C . ( 2018 ) . Microgeography and the direction of inventive activity . Management Science , 64 ( 9 ) , 4348 \u2013 4364 . Catalini , C . , Fons - Rosen , C . , & Gaul\u00e9 , P . ( 2020 ) . How do travel costs shape collaboration ? Management Science , 66 ( 8 ) , 3340 \u2013 3360 . Caza , B . B . , Moss , S . , & Vough , H . ( 2018 ) . From synchronizing to harmonizing : The process of authenticating multiple work identities . Administrative Science Quarterly , 63 ( 4 ) , 703 \u2013 745 . Chai , S . , & Freeman , R . B . ( 2019 ) . Temporary colocation and collaborative discovery : Who confers at confer - ences . Strategic Management Journal , 40 ( 13 ) , 2138 \u2013 2164 . Chatterji , A . , Delecourt , S . , Hasan , S . , & Koning , R . ( 2019 ) . When does advice impact startup performance ? Stra - tegic Management Journal , 40 ( 3 ) , 331 \u2013 356 . Cohen , W . M . , & Levinthal , D . A . ( 1990 ) . Absorptive capacity : A new perspective on learning and innovation . Administrative Science Quarterly , 35 ( 1 ) , 128 \u2013 152 . Coletti , M . H . , & Bleich , H . L . ( 2001 ) . Medical subject headings used to search the biomedical literature . Journal of the American Medical Informatics Association , 8 ( 4 ) , 317 \u2013 323 . Dahlander , L . , & McFarland , D . A . ( 2013 ) . Ties that last : Tie formation and persistence in research collaborations over time . Administrative Science Quarterly , 58 ( 1 ) , 69 \u2013 110 . Dahlander , L . , O ' Mahony , S . , & Gann , D . M . ( 2016 ) . One foot in , one foot out : How does individuals ' external search breadth affect innovation outcomes ? Strategic Management Journal , 37 ( 2 ) , 280 \u2013 302 . Fang , C . , Lee , J . , & Schilling , M . A . ( 2010 ) . Balancing exploration and exploitation through structural design : The isolation of subgroups and organizational learning . Organization Science , 21 ( 3 ) , 625 \u2013 642 . Fleming , L . ( 2001 ) . Recombinant uncertainty in technological search . Management Science , 47 ( 1 ) , 117 \u2013 132 . 28 LANE ET AL . Fleming , L . , Mingo , S . , & Chen , D . ( 2007 ) . Collaborative brokerage , generative creativity , and creative success . Administrative Science Quarterly , 52 ( 3 ) , 443 \u2013 475 . Ganguli , I . ( 2017 ) . Saving soviet science : The impact of grants when government R & D funding disappears . Amer - ican Economic Journal : Applied Economics , 9 ( 2 ) , 165 \u2013 201 . Grant , R . M . ( 1996 ) . Toward a knowledge - based theory of the firm . Strategic Management Journal , 17 ( S2 ) , 109 \u2013 122 . Haas , M . R . , & Park , S . ( 2010 ) . To share or not to share ? Professional norms , reference groups , and information withholding among life scientists . Organization Science , 21 ( 4 ) , 873 \u2013 891 . Haeussler , C . , Jiang , L . , Thursby , J . , & Thursby , M . ( 2014 ) . Specific and general information sharing among com - peting academic researchers . Research Policy , 43 ( 3 ) , 465 \u2013 475 . Hagstrom , W . O . ( 1974 ) . Competition in science . American Sociological Review , 39 ( 1 ) , 1 \u2013 18 . Hansen , M . T . ( 1999 ) . The search - transfer problem : The role of weak ties in sharing knowledge across organiza - tion subunits . Administrative Science Quarterly , 44 ( 1 ) , 82 \u2013 111 . Hansen , M . T . , & Haas , M . R . ( 2001 ) . Competing for attention in knowledge markets : Electronic document dis - semination in a management consulting company . Administrative Science Quarterly , 46 ( 1 ) , 1 \u2013 28 . Hargadon , A . , & Sutton , R . I . ( 1997 ) . Technology brokering and innovation in a product development firm . Administrative Science Quarterly , 42 , 716 \u2013 749 . Hasan , S . , & Koning , R . ( 2019 ) . Prior ties and the limits of peer effects on startup team performance . Strategic Management Journal , 40 ( 9 ) , 1394 \u2013 1416 . Ingram , P . , & Morris , M . W . ( 2007 ) . Do people mix at mixers ? Structure , homophily , and the \u201c life of the party \u201d . Administrative Science Quarterly , 52 ( 4 ) , 558 \u2013 585 . Jeppesen , L . B . , & Lakhani , K . R . ( 2010 ) . Marginality and problem - solving effectiveness in broadcast search . Organization Science , 21 ( 5 ) , 1016 \u2013 1033 . Jones , B . F . ( 2009 ) . The burden of knowledge and the \u201c death of the renaissance man \u201d : Is innovation getting har - der ? The Review of Economic Studies , 76 ( 1 ) , 283 \u2013 317 . Kim , T . , McFee , E . , Olguin , D . O . , Waber , B . , & Pentland , A . S . ( 2012 ) . Sociometric badges : Using sensor technol - ogy to capture new forms of collaboration . Journal of Organizational Behavior , 33 ( 3 ) , 412 \u2013 427 . Kleinbaum , A . M . , Stuart , T . E . , & Tushman , M . L . ( 2013 ) . Discretion within constraint : Homophily and struc - ture in a formal organization . Organization Science , 24 ( 5 ) , 1316 \u2013 1336 . Kogut , B . , & Zander , U . ( 1992 ) . Knowledge of the firm , combinative capabilities , and the replication of technol - ogy . Organization science , 3 ( 3 ) , 383 \u2013 397 . Latour , B . , & Woolgar , S . ( 2013 ) . Laboratory life : The construction of scientific facts . Princeton , NJ : Princeton Uni - versity Press . Leahey , E . , Beckman , C . M . , & Stanko , T . L . ( 2017 ) . Prominent but less productive : The impact of interdisciplin - arity on scientists ' research . Administrative Science Quarterly , 62 ( 1 ) , 105 \u2013 139 . Lee , S . ( 2019 ) . Learning - by - moving : Can reconfiguring spatial proximity between organizational members pro - mote individual - level exploration ? Organization Science , 30 ( 3 ) , 467 \u2013 488 . Levinthal , D . A . , & March , J . G . ( 1993 ) . The myopia of learning . Strategic Management Journal , 14 ( S2 ) , 95 \u2013 112 . Maggitti , P . G . , Smith , K . G . , & Katila , R . ( 2013 ) . The complex search process of invention . Research Policy , 42 ( 1 ) , 90 \u2013 100 . Maurer , I . , & Ebers , M . ( 2006 ) . Dynamics of social capital and their performance implications : Lessons from bio - technology start - ups . Administrative Science Quarterly , 51 ( 2 ) , 262 \u2013 292 . McFadyen , M . A . , & Cannella , A . A . , Jr . ( 2004 ) . Social capital and knowledge creation : Diminishing returns of the number and strength of exchange relationships . Academy of management Journal , 47 ( 5 ) , 735 \u2013 746 . McPherson , J . M . , & Smith - Lovin , L . ( 1987 ) . Homophily in voluntary organizations : Status distance and the com - position of face - to - face groups . American Sociological Review , 52 , 370 \u2013 379 . Merton , R . K . ( 1973 ) . The sociology of science : Theoretical and empirical investigations . Chicago , IL : University of Chicago Press . Murray , F . , & O ' Mahony , S . ( 2007 ) . Exploring the foundations of cumulative innovation : Implications for organi - zation science . Organization Science , 18 ( 6 ) , 1006 \u2013 1021 . Nonaka , I . ( 1994 ) . A dynamic theory of organizational knowledge creation . Organization Science , 5 ( 1 ) , 14 \u2013 37 . Pain , E . ( 2018 ) . Collaborating for the win . Science . http : / / dx . doi . org / 10 . 1126 / science . caredit . aat4606 . LANE ET AL . 29 Phelps , C . , Heidl , R . , & Wadhwa , A . ( 2012 ) . Knowledge , networks , and knowledge networks : A review and research agenda . Journal of Management , 38 ( 4 ) , 1115 \u2013 1166 . Reschke , B . P . , Azoulay , P . , & Stuart , T . E . ( 2018 ) . Status spillovers : The effect of status - conferring prizes on the allocation of attention . Administrative Science Quarterly , 63 ( 4 ) , 819 \u2013 847 . Reskin , B . F . ( 1977 ) . Scientific productivity and the reward structure of science . American Sociological Review , 42 , 491 \u2013 504 . Rosenkopf , L . , & Almeida , P . ( 2003 ) . Overcoming local search through alliances and mobility . Management sci - ence , 49 ( 6 ) , 751 \u2013 766 . Simon , H . A . ( 1991 ) . Bounded rationality and organizational learning . Organization Science , 2 ( 1 ) , 125 \u2013 134 . Singh , J . ( 2005 ) . Collaborative networks as determinants of knowledge diffusion patterns . Management Science , 51 ( 5 ) , 756 \u2013 770 . Singh , J . , & Fleming , L . ( 2010 ) . Lone inventors as sources of breakthroughs : Myth or reality ? Management Science , 56 ( 1 ) , 41 \u2013 56 . Staw , B . M . , & Epstein , L . D . ( 2000 ) . What bandwagons bring : Effects of popular management techniques on cor - porate performance , reputation , and CEO pay . Administrative Science Quarterly , 45 ( 3 ) , 523 \u2013 556 . Stephan , P . E . ( 1996 ) . The economics of science . Journal of Economic literature , 34 ( 3 ) , 1199 \u2013 1235 . Spender , J . C . ( 1996 ) . Making knowledge the basis of a dynamic theory of the firm . Strategic Management Journal , 17 ( S2 ) , 45 \u2013 62 . Szulanski , G . ( 1996 ) . Exploring internal stickiness : Impediments to the transfer of best practice within the firm . Strategic Management Journal , 17 ( S2 ) , 27 \u2013 43 . Uzzi , B . ( 1997 ) . Social structure and competition in interfirm networks : The paradox of embeddedness . Adminis - trative Science Quarterly , 42 , 35 \u2013 67 . Uzzi , B . , Mukherjee , S . , Stringer , M . , & Jones , B . ( 2013 ) . Atypical combinations and scientific impact . Science , 342 ( 6157 ) , 468 \u2013 472 . Zhang , L . , & Guler , I . ( 2020 ) . How to join the club : Patterns of embeddedness and the addition of new members to interorganizational collaborations . Administrative Science Quarterly , 65 ( 1 ) , 112 \u2013 150 . SUPPORTING INFORMATION Additional supporting information may be found online in the Supporting Information section at the end of this article . How to cite this article : Lane JN , Ganguli I , Gaule P , Guinan E , Lakhani KR . Engineering serendipity : When does knowledge sharing lead to knowledge production ? Strat Mgmt J . 2020 ; 1 \u2013 30 . https : / / doi . org / 10 . 1002 / smj . 3256 30 LANE ET AL .",
    "mwangangi2021structure": "Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 1 of 10 BIOCHEMISTRY The structure of the actin filament uncapping complex mediated by twinfilin Dennis M . Mwangangi 1 , 2 , Edward Manser 1 , 2 , Robert C . Robinson 1 , 3 , 4 * Uncapping of actin filaments is essential for driving polymerization and depolymerization dynamics from cap - ping protein \u2013 associated filaments ; however , the mechanisms of uncapping leading to rapid disassembly are un - known . Here , we elucidated the x - ray crystal structure of the actin / twinfilin / capping protein complex to address the mechanisms of twinfilin uncapping of actin filaments . The twinfilin / capping protein complex binds to two G - actin subunits in an orientation that resembles the actin filament barbed end . This suggests an unanticipated mechanism by which twinfilin disrupts the stable capping of actin filaments by inducing a G - actin conformation in the two terminal actin subunits . Furthermore , twinfilin disorders critical actin - capping protein interactions , which will assist in the dissociation of capping protein , and may promote filament uncapping through a second mechanism involving V - 1 competition for an actin - binding surface on capping protein . The extensive interactions with capping protein indicate that the evolutionary conserved role of twinfilin is to uncap actin filaments . INTRODUCTION Numerous cellular processes , such as morphogenesis , migration , cytokinesis , endocytosis , and memory , rely on rapid reorganization of actin cytoskeletal networks ( 1 \u2013 3 ) . Coordinated local assembly and disassembly of actin filaments generate force and structure , which is harnessed to drive these specific functions . A number of actin - regulating proteins control the assembly , disassembly , and organization of the filament networks ( 4 , 5 ) . Among the key evolu - tionarily conserved actin regulators are capping protein ( CP ) and the actin depolymerization factor homology ( ADF - H ) domain family of proteins , which include ADF / cofilins and twinfilin ( 6 \u2013 10 ) . While primitive , functional ADF - H domain proteins are found in Asgard archaea ( 7 , 9 ) , CP and twinfilin have only been found , and are ubiq - uitous , in eukaryotes ( 9 , 10 ) . Thus , CP and twinfilin likely arose during eukaryogenesis and were present in the last eukaryotic common ancestor ( LECA ) . The architecture of twinfilin is unique , composed of two ADF - H domains connected by a short linker and followed by a conserved C - terminal tail ( 11 ) . ADF - H proteins generally regulate cytoskeletal reorganization by accelerating the disassembly of actin filaments ( 6 , 12 , 13 ) . Twinfilin - 1 is ubiquitously expressed in almost all tissue types in mammals , where it regulates actin dynamics through mechanisms involving interactions with actin monomers , actin filaments , and CP ( 11 , 14 ) . The biological outputs from twin - filin regulation of actin dynamics include cell motility and synaptic endocytosis ( 15 ) . The reported in vitro roles of twinfilin in actin dynamics are nu - merous , diverse , and somewhat contradictory . Twinfilin binds and sequesters adenosine diphosphate ( ADP ) \u2013 actin monomers with high affinity , inhibiting nucleotide exchange and preventing assem - bly into filaments ( 11 , 14 , 16 ) . Twinfilin also interacts directly with actin filament barbed ends , blocking filament elongation , suggestive of a capping activity ( 17 \u2013 19 ) . Recent studies have also demonstrated that twinfilin accelerates depolymerization of actin filament barbed ends containing ADP - actin subunits ( 20 , 21 ) , and at low pH , twin - filin can sever filaments ( 22 ) . In addition , twinfilin interacts strongly with heterodimeric CP through interactions that include those in - volving the conserved twinfilin C - terminal tail ( 23 ) . CP is a hetero - dimer that binds to the barbed ends of actin filaments to prevent actin subunit exchange ( 24 \u2013 26 ) . Although twinfilin binds to CP with high affinity , its exact biological role in promoting CP capping or uncapping is debated ( 27 , 28 ) . X - ray structural studies of twinfi - lin have been limited to single ADF - H domains , which show high structural conservation , and both domains bind actin monomers ( 19 ) . CP\u2019s interactions with the actin filaments or dynactin fila - ments are resolved to 23 - and 3 . 4 - \u00c5 resolution , respectively , via cryo \u2013 electron microscopy ( cryo - EM ) ( 29 , 30 ) . However , the molec - ular mechanism by which twinfilin interacts with CP at the actin filament barbed ends is unknown . Here , we address the role of twinfilin in uncapping of actin filaments by elucidating the x - ray structure of the twinfilin / CP / actin complex . RESULTS The crystal structure of the twinfilin / CP / actin complex Previous biochemical data have shown that twinfilin\u2019s interaction with CP - capped actin filaments protects CP from displacement by CARMIL , suggesting a stable interaction between twinfilin , CP , and barbed - end actin subunits ( 27 ) . We therefore used purified twinfilin - 1 ( human ) , heterodimeric CP ( mouse CapZ \uf061 1 / \uf062 2 , henceforth CP ) , and skeletal muscle actin ( rabbit ) to reconstitute the complex be - tween twinfilin , CP , and actin monomers . The twinfilin , CP , and actin complex was highly stable in gel filtration chromatography , and this complex was used to prepare protein crystals suitable for structure determination by x - ray crystallography at 3 . 2 - \u00c5 resolu - tion ( fig . S1 ) . The complex consists of two ADP - bound actin subunits , one subunit each of the heterodimeric CP ( CP \uf061 1 and CP \uf062 2 ) and one full - length twinfilin - 1 ( Fig . 1 , A and B ) . In the structure , CP adopts its canonical mushroom - shaped architecture consisting of a cap and stalk and interacts with the barbed - end faces of two G - actin 1 Institute of Molecular and Cell Biology , A * STAR ( Agency for Science , Technology and Research ) , Biopolis , Singapore 138673 , Singapore . 2 Department of Pharmacol - ogy , Yong Loo Lin School of Medicine , National University of Singapore , Singapore 117597 , Singapore . 3 School of Biomolecular Science and Engineering ( BSE ) , Vidya - sirimedhi Institute of Science and Technology ( VISTEC ) , Rayong 21210 , Thailand . 4 Research Institute for Interdisciplinary Science ( RIIS ) , Okayama University , Okayama 700 - 8530 , Japan . * Corresponding author . Email : br . okayama . u @ gmail . com Copyright \u00a9 2021 The Authors , some rights reserved ; exclusive licensee American Association for the Advancement of Science . No claim to original U . S . Government Works . Distributed under a Creative Commons Attribution NonCommercial License 4 . 0 ( CC BY - NC ) . CORRECTED 1 APRIL 2021 ; SEE FULL TEXT D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024 Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 2 of 10 subunits via the top surface of the mushroom cap ( Fig . 1 ) . The two actin subunits are in structurally similar conformations and adopt a typical G - actin fold consisting of four subdomains . The actin sub - units do not show the subunit flattening and twisting associated with the G - to - F actin transition ( 31 ) . The deoxyribonuclease I bind - ing loops in the two actin subunits are disordered ( fig . S2 , A and B ) . The relative orientation of the two actin subunits resembles that of barbed - end actin subunits from the F - actin cryo - EM structure ( fig . S2 , C to E ) ( 32 ) . The actin subunits are arranged by the short pitch helical filament relationship , similar to two actin subunits across a filament . They do not adopt the relative positioning of two longitu - dinally related subunits in a single strand . Twinfilin adopts an elon - gated architecture in which the two ADF - H domains each bind one actin subunit , with ADF - H domain 2 ( D2 ) binding to the terminal actin subunit , relative to a filament barbed end . The linker connect - ing the ADF - H domains , which includes an \uf061 - helix , extends across the upper surface ( cap ) of the mushroom - shaped CP and also contacts both actin subunits ( Fig . 1 , A and B ) . The twinfilin C - terminal tail extends from D2 and wraps around the stalk of the CP \uf062 - subunit , which is located opposite to the actin - binding interface , below the CP cap ( Fig . 1 , C and D ) . Binary interactions in the twinfilin / CP / actin complex Each of the two twinfilin ADF - H domains binds to an actin subunit at analogous interfaces , providing basis for monomer sequestration ( Fig . 2A ) . The individual ADF - H domains adopt similar architec - tures except for a difference in the conformation of the \uf062 - sheets . The \uf062 - 3 and \uf062 - 4 strands in D2 form a protrusive extension relative to that in ADF - H domain 1 ( D1 ) ( fig . S3 , A to C ) . The key structural elements of the two ADF - H domains are highly conserved in mouse twinfilin and human twinfilin - 2 ( 33 ) . Despite twinfilin D2 having 10 times higher affinity for G - actin ( 16 ) , its actin - binding interface is markedly similar to that of D1 . Thus , the precise selection of res - idues in the two binding interfaces is likely to explain the differences in actin - binding affinity . Further , the \uf061 - helix in the linker between D1 and D2 loosely associates with the D2 - bound actin subunit , and this likely strengthens D2 interaction with G - actin ( Fig . 2A and fig . S4 , A and C ) . Fig . 1 . The twinfilin / CP / actin complex . ( A ) Front view of the pentameric complex in cartoon representation . CP consists of two subunits , \uf061 - subunit ( CP \uf061 ) and \uf062 - subunit ( CP \uf062 ) . Twinfilin comprises two ADF - H domains ( D1 and D2 ) , a linker between D1 and D2 that includes a helix ( residues 151 to 165 ) , and a C - terminal tail ( Tail ; residues 316 to 342 , the last eight amino acids are disordered ) . Actin subunit 1 is bound to twinfilin D1 , and subunit 2 to twinfilin D2 . The twinfilin secondary structure elements are colored differentially , helices in red , strands in cyan , and loops and extended regions in orange , with the Tail in lime green . ( B ) Front view [ same as in ( A ) ] , in which the actin subunits and twinfilin are represented as surfaces . Twinfilin is shown entirely in orange . ( C and D ) Back and side views , respectively . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024 Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 3 of 10 The CP actin - binding interface is located at the top surface of the mushroom cap ( Fig . 2B ) . In the absence of twinfilin , CP interacts with the barbed - end protomers of a filament via C - terminal exten - sions to the \uf061 - and \uf062 - subunits , the \uf061 - and \uf062 - tentacles ( 24 , 26 , 29 ) . In the twinfilin - bound structure , the CP \uf062 - tentacle is mostly disor - dered in the structure and , hence , has no direct contact with either of the actin subunits ( fig . S4 , D and E ) . The \uf061 - tentacle is partially ordered and forms an interaction with actin 1 ( fig . S4F ) . Twinfilin binds CP through its C - terminal tail . This tail wraps around the stalk of CP \uf062 - subunit , a common binding site for filament uncap - ping proteins with CP interaction ( CPI ) motifs ( Fig . 2C ) ( 34 ) . The ordered portion of the tail of twinfilin ( residues 315 to 342 ) includes a basic stretch that interacts with the negatively charged groove be - tween the CP \uf062 - subunit stalk and the underside of the mushroom - shaped cap ( fig . S5 , A to C ) . Furthermore , the complex reveals additional interactions between twinfilin and CP , beyond the C - terminal tail . First , the \uf061 - helix in the linker between D1 and D2 forms an interac - tion with CP that also involves the start of the \uf062 - tentacle ( Fig . 2C and fig . S5 , D to F ) . Second , twinfilin D1 forms a direct interface with the CP \uf062 - subunit ( Fig . 2C and fig . S5G ) . Comparison of the conformations of twinfilin tail \u2013 bound and CARMIL CPI \u2013 bound CP The twinfilin tail - binding site on CP is distant from the actin - binding site , which is centered on the \uf061 - tentacle ( Fig . 3 , A and B ) . Comparison of the tail - binding site with the uncapping CPI motif from CARMIL shows an overlapping interaction on the underside of the CP \uf062 - subunit ( 34 ) ; however , the N termini of the two pep - tides take divergent paths around the CP stalk ( Fig . 3A ) . The CARMIL CPI motif half encircles the CP stalk , with its N terminus making contact with the CP \uf061 - subunit stalk . By contrast , the N ter - minus of the twinfilin tail ( residues 316 to 322 ) follows a straight path and does not form contacts with the CP \uf061 - subunit . This region ( residues 316 to 322 ) is elongated yet ordered with clear electron density ( Fig . 2C and fig . S5K ) , despite not being stabilized by inter - actions , suggesting that it may be under tension to extend the poly - peptide chain into an ordered conformation . The overlapping interface provides a structural basis for the competition between the twinfilin tail and CARMIL CPI for CP binding , and twinfilin\u2019s attenuation of CARMIL - mediated dissociation of CP from filament barbed ends ( 27 ) . Fig . 2 . Binary interactions in the twinfilin / CP / actin complex . ( A ) Front and back views of actin bound to twinfilin D1 or twinfilin D2 . ( B ) Two views of the CP interaction with actin , in which actin 1 and actin 2 are shown in similar orientations . ( C ) Three orientations of the CP / twinfilin interaction . Examples of the electron density of key features are shown in fig . S5 ( H to K ) . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024 Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 4 of 10 CP can adopt two known conformations , which are likely to be in dynamic equilibrium in solution . The uncapping CPI motifs of CARMIL , CD2AP , and CKIP stabilize one conformation , actin - free conformation of CP ( 34 ) . Superimposition of the complex with CARMIL - bound CP and unbound CP shows key structural changes that CP subunits undergo in adopting the twinfilin / CP / actin com - plex conformation ( Fig . 3 , C and D , and fig . S6 ) . First , the CP \uf062 - subunit mushroom cap in this complex moves upward relative to the stalk , while the \uf061 - tentacle repositions to adopt the actin - bound con - formation . Second , the \uf061 - subunit adjusts upward flattening the mushroom cap and adopts a similar conformation to that stabilized by V - 1 or Arp1 ( from the dynactin complex ) ( Fig . 3E and fig . S6 ) ( 30 , 35 ) . The binding site of V - 1 , a steric CP inhibitor , overlaps with that of actin in this complex ( fig . S6J ) ( 35 ) . Therefore , the CP struc - ture in the twinfilin / CP / actin complex represents the mushroom cap \u2013 bound conformational state , to either actin , Arp1 , or V - 1 , while the CPI - bound structure of CP represents a stabilized mush - room cap \u2013 unbound conformation . Binding of the CARMIL CPI motif around the stem of CP locks CP in the mushroom cap \u2013 unbound conformation , which is less compatible with actin barbed - end in - teraction , leading to uncapping of the filament ( 34 , 36 ) . The in - ability of the twinfilin tail to induce a change in the CP conformation from mushroom cap \u2013 bound to unbound state suggests that , in the complex , actin binding dominates the CP conformation or the twinfilin tail does not stabilize the mushroom cap \u2013 unbound con - formation . To distinguish between these possibilities , we tested for uncapping activity within the twinfilin tail in isolation . We per - formed pyrene - actin polymerization assays in which an increase in pyrene fluorescence reports on the efficiency of actin assembly from CP precapped actin filament seeds . The CARMIL CPI peptide , a positive control , displayed potent uncapping activity leading to polymer - ization from the CP precapped filament seeds , indicated by the increase in fluorescence ( fig . S7 ) . By contrast , the maltose - binding protein ( MBP ) \u2013 tagged twinfilin tail ( residues Gln 321 - Asp 350 ) had no detect - able uncapping effect , showing a similar polymerization profile to the CP precapped filament seeds alone ( fig . S7 ) . Thus , in this assay , the twinfilin tail alone does not uncap CP - capped filaments , suggesting that it does not strongly stabilize the CP mushroom cap \u2013 unbound conformation . Twinfilin influences CP interaction with actin barbed ends The structure reveals additional interactions between twinfilin and CP that might influence the CP - binding mode to the actin filament barbed ends ( fig . S5 , D to G ) . The dynactin filament ( consisting of Fig . 3 . Conformations of CP . ( A ) Comparison of the twinfilin tail ( yellow ) binding site with CARMIL ( cyan ) on CP . Both CP - binding peptides run in the same direction , and N terminus of the twinfilin tail is labeled N . ( B ) The actin - binding site on the CP \uf061 - tentacle is distant from the twinfilin tail - binding site . ( C and D ) Structural superimposition of the \uf062 - subunits of CP reveals that the conformation of CP in the twinfilin / CP / actin complex is different to the CARMIL - bound and unbound conformations of CP . ( E ) Superimposition reveals that the conformation of CP in the twinfilin / CP / actin complex is similar to the V - 1 \u2013 bound conformation of CP . \uf061 - and \uf062 - subunits of CP are colored light green and dark green , respectively , for the CARMIL , unbound , and V - 1 complexes . Black arrows indicate conformational changes in CP in adopting the twinfilin / CP / actin complex structure . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024 Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 5 of 10 Arp1 instead of actin ) is structurally similar to the actin filament and is capped at its barbed end by CP ( 30 ) . Comparison of the CP - binding modes to actin in the twinfilin / CP / actin complex and to Arp1 subunits reveals similar overall geometries ( Fig . 4 and fig . S8 ) . However , there are considerable differences in the binding of the CP \uf061 - and \uf062 - tentacles to their respective actin / Arp1 subunits . These differences arise from the presence of twinfilin , because the binding modes of the \uf061 - and \uf062 - tentacles are similar in the dynactin complex and in the 23 - \u00c5 cryo - EM structure of the CP / actin fila - ment ( 29 , 30 ) . The \uf062 - tentacle , which is bound to the terminal Arp1 subunit in the dynactin complex , is disordered in the twinfilin / CP / actin complex ( Fig . 4A and fig . S8A ) . Structural superimposition indicates that the twinfilin linker obscures the \uf062 - tentacle binding site on actin in the actin / twinfilin / CP complex ( Fig . 4A ) . In the dy - nactin complex , the \uf061 - tentacle is fully ordered and bound to Arp1 . By contrast , the \uf061 - tentacle in the twinfilin / CP / actin complex only forms a partial interaction with actin ( Fig . 4B and fig . S8 , B and C ) . Superimposition reveals that the \uf061 - tentacle \u2013 binding site on actin is partially obstructed by twinfilin D1 ( Fig . 4B ) . The binding of the tentacles to the actin protomers will also be influenced by the actin protomer conformation . A twinfilin - induced shift from an F - actin to G - actin conformation may aid dissociation of the tentacles . In the absence of twinfilin , the CP \uf061 - tentacle is a critical interac - tion with barbed end of a filament , while the \uf062 - tentacle offers a sec - ond important actin - binding interface to stabilize capping ( 36 ) . The obstruction of the CP tentacles by twinfilin in the actin / twinfilin / CP complex will destabilize the CP interaction , leading to weakened CP affinity for actin filament barbed ends . Thus , the CP - binding mode in the dynactin complex represents fully bound CP , conferring strong and stable capping activity , while the CP - binding mode in the actin / twinfilin / CP complex represents a weak and unstable capping activity . This unstable interaction state may be targeted by other regulatory factors , such as the CP - sequestering protein V - 1 . We used the pyrene - actin polymerization assay to test whether V - 1 can enhance CP uncapping in the presence of twinfilin ( 28 ) . Addition of twinfilin into CP precapped actin filament seeds mixed with pyrene - labeled monomers and profilin did not accelerate polymerization , indicat - ing either a lack of uncapping or uncapping followed by recapping ( Fig . 5 ) . Addition of V - 1 alone displayed partial polymerization attributable to CP sequestration ( Fig . 5 ) . However , the presence of both twinfilin and V - 1 accelerated polymerization to near the level of uncapped actin filament seeds , indicating that the cooperative activities of twinfilin and V - 1 can induce dissociation of CP from barbed ends and prevent recapping by CP ( Fig . 5 ) . This indicates that twinfilin is able to remove CP from filaments ; however , the CP - sequestering protein V - 1 is required to prevent recapping of fil - aments . We propose that the high concentration of profilin ( 2 . 8 \uf06d M ) , relative to twinfilin ( 1 \uf06d M ) , used in this assay was sufficient to com - petitively remove actin from the CP / twinfilin complex and allow recapping in the absence of V - 1 . The requirement of both the un - capping agent twinfilin and CP - sequestering protein V - 1 to observe robust uncapping in this in vitro assay partially explains some of the disparities in the reported activities of twinfilin . Uncapping has been difficult to observe in many assays in which recapping has not been excluded . DISCUSSION This analysis of the actin / twinfilin / CP complex reveals the structural basis by which twinfilin interacts with actin and with CP , and pro - vides insight into molecular mechanisms for the regulation of CP in actin filament barbed - end dynamics . The unanticipated geometry Fig . 4 . Comparison of the CP - binding mode to Arp1 in the dynactin complex ( Arp1 / CP ) with the binding mode to actin in actin / twinfilin / CP complex . ( A and B ) Focus on the \uf062 - and \uf061 - tentacles , respectively , which are highlighted by black circles . The \uf062 - tentacle is disordered and unbound in the actin / twinfilin / CP complex , and the \uf061 - tentacle is partially bound and ordered , relative to the dynactin complex . Enlargements of the tentacle regions show the superimpositions of the CP ( \uf061 - and \uf062 - subunits colored red and blue , respectively ) from the dynactin complex on to actin ( green and teal ) and twinfilin ( yellow ) from the actin / twinfilin / CP complex . ( A ) Enlargement , the twinfilin linker ( yellow ) binds to the \uf062 - tentacle \u2013 binding site on actin 2 ( green ) . ( B ) Enlargement , the N terminus of twinfilin D1 ( yellow ) occupies half of the \uf061 - tentacle \u2013 binding site on actin 1 ( teal ) . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024 Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 6 of 10 in which twinfilin binds to two actin subunits in a pseudo - filament barbed - end orientation has major implications for its possible mech - anisms of action . Previously proposed mechanisms for twinfilin\u2019s activities assumed that twinfilin contacts two longitudinally related actin subunits at the barbed end of an actin filament ( 17 , 18 , 21 , 27 , 28 ) . However , the structure presented here unambiguously identifies that the two actin subunits are laterally related . Twinfilin binds across the two strands of an actin filament , rather than down the side of a single strand in the filament . We discuss the ramifications of this binding orientation for actin filament regulation . The primary biological role of twinfilin has not been settled . Mammalian twinfilin - 1 was shown to preferentially localize to re - gions of the cell enriched for F - actin ( 14 ) . Loss of twinfilin - 1 in B16 cells leads to their inability to generate lamellipodia ( 27 ) . In vitro and cell - based experiments have led to multiple proposed activities . Twinfilin\u2019s actin - related functions have been variously described as follows : an actin monomer - sequestering protein that forms a 1 : 1 complex with actin that prevents actin polymerization and inhibits nucleotide exchange in actin ( 11 , 14 ) ; a CP - interacting and phos - phatidylinositol 4 , 5 - bisphosphate ( PIP 2 ) \u2013 interacting protein ( 37 ) ; an actin monomer - shuttling protein between the pointed and barbed ends of filaments ( 38 ) ; a protein that does not affect CP - capping activity , nor does CP affect twinfilin\u2019s monomer binding function ( rather , twinfilin localizes actin monomers to sites of assembly through interaction with CP ) ( 23 ) ; a barbed - end actin filament CP ( 17 ) ; an actin filament - severing protein ( 22 ) ; a protein that speeds up actin depolymerization at both ends of the actin filament in con - junction with cyclase - associated protein ( CAP ) ( 20 ) ; a protein that enhances CP actin filament capping , which competes with CARMIL for CP binding ( 27 ) ; and an actin filament CP - uncapping protein ( 28 ) . The actin / twinfilin / CP complex provides the structural basis to reassess these activities . In the complex , twinfilin forms extensive interactions with CP and with two actin protomers , which adopt a pseudo - actin filament barbed - end orientation relative to each other . These multiple inter - actions indicate that twinfilin\u2019s primary and evolutionary conserved role , throughout eukaryotes , is to regulate CP capping at the barbed end of actin filaments . In yeast , the CP : twinfilin ratio is estimated to be 2 . 5 : 1 ( 37 , 39 ) , which indicates that the actin - related cellular activities of twinfilin will be dominated by the CP : twinfilin com - plex , rather than twinfilin acting alone on filaments . Similarly , the molar abundance of mammalian CP : twinfilin - 1 ratio was deter - mined as 2 : 1 ( 27 ) . Twinfilin has been shown to strongly interact with actin filament barbed ends and with CP , dissociation constants ( K d ) of 13 and 50 nM , respectively ( 17 , 27 ) . However , these affinities are at least one order of magnitude weaker than the CP affinity for actin filament barbed ends [ K d = 0 . 1 to 1 . 0 nM ( 40 ) ] . The actin / twinfilin / CP complex structure demonstrates that twinfilin disrupts CP - actin interfaces , sterically competing with the CP \uf061 - and \uf062 - tentacles in binding to actin , indicating that twinfilin\u2019s principal role is to destabilize CP capping ( Fig . 6A ) . We discuss the possible mechanisms that lead to uncapping . The ADP - bound actin protomer conformations in the actin / twinfilin / CP complex are very similar to that in the cofilin - decorated actin filament ( fig . S2B ) ( 41 ) . Cofilin , which consists of a single ADF - H domain , severs actin filaments toward the pointed end of a section of filament decorated with cofilin , by inducing a G - actin \u2013 like conformation in the actin protomers , which are not sta - ble as a filament ( 13 , 41 ) . We hypothesize that the orientation of the two ADF - H twinfilin domains , in binding across the filament , allows twinfilin to induce a G - actin \u2013 like conformation in the final two barbed - end actin protomers . This conformation will destabilize these two terminal actin subunits , leading to them being \u201csevered\u201d from the end of the filament , by an analogous mechanism to cofilin severing ( 13 , 41 ) . Thus , we propose that the two twinfilin ADF domains induce severing at the boundary between ADF - H \u2013 bound and ADF - H \u2013 free portions of F - actin , dissociating CP , twinfilin , and the two terminal actin subunits as a complex ( Figs . 1A and 6 , A and B ) . Immediate reassociation of the actin / twinfilin / CP complex with the filament would be unfavorable because the actin subunits in the complex are held in the ADP - bound G - actin state , and nucleotide exchange is inhibited by the twinfilin ADF - H domains ( 11 ) . We propose that the actin / twinfilin / CP complex will then un - dergo a process of recycling . Twinfilin strongly binds to ADP - bound actin monomers ( K d ~ 40 to 50 nM ) , and this affinity is unaffected by the presence of CP ( 16 , 23 ) . For comparison , CAP , thymosin - \uf062 4 , profilin , and cofilin are characterized in their affinities for ADP - bound actin monomers by K d of 20 nM , 80 to 100 nM , 0 . 17 \uf06d M , and 0 . 4 \uf06d M , respectively [ reviewed in ( 42 ) ] . The two ADP - bound actin subunits , from the actin / twinfilin / CP complex , may then be con - verted to adenosine triphosphate ( ATP ) \u2013 bound actin by the actions of CAP , due to the high affinity of CAP for ADP - bound actin ( 20 , 21 , 33 , 43 ) . Subsequently , the ATP - bound actin monomers will be released to profilin , and possibly to thymosin - \uf062 4 , due to their superior affinities for ATP - bound actin relative to twinfilin , CAP , and cofilin , 0 . 1 , 0 . 1 to 4 . 0 , 0 . 5 , 1 . 9 , and 6 \uf06d M ( 16 , 23 , 42 ) , respectively . This will replenish the polymerization - competent pool of actin monomers ( 42 ) . Because CAP also binds to the pointed ends of actin filaments ( 33 ) , association of the actin / twinfilin / CP complex with the pointed end bound CAP may provide a mechanism to spa - tially separate the twinfilin / CP complex away from elongating actin filament barbed ends close to the membrane . Fig . 5 . V - 1 uncaps filament barbed ends in the presence of twinfilin . Pyrene - actin polymerization assay showing uncapping of F - actin seeds by V - 1 in the presence of twinfilin . Unlabeled F - actin seeds ( 0 . 5 \uf06d M ) were capped with 100 nM CP and subse - quently polymerized in either 2 \uf06d M actin ( 10 % pyrene ) / 2 . 8 \uf06d M profilin ( red ) or 2 \uf06d M actin ( 10 % pyrene ) / 2 . 8 \uf06d M profilin supplemented with either 1 \uf06d M twinfilin ( yellow ) , 5 \uf06d M V - 1 ( purple ) , or a mixture of 1 \uf06d M twinfilin and 5 \uf06d M V - 1 ( blue ) . As a control , F - actin seeds were polymerized in the absence of CP ( black ) . As shown in the pro - file , the presence of both twinfilin and V - 1 induces accelerated polymerization of pre - capped F - actin seeds ( blue ) to almost the same level as the positive control ( black ) . By contrast , twinfilin alone ( yellow ) induces very low level of polymerization , similar to CP alone ( red ) , while V - 1 alone ( purple ) only induces minimal polymerization . AU , arbitrary units . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024 Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 7 of 10 The fate of the twinfilin / CP complex may then be several - fold . First , the twinfilin / CP complex may encounter a free barbed end of a filament and undergo a further round of transient binding followed by dissociation of the terminal actin protomers . Thus , the twinfilin / CP complex may depolymerize the filament in two subunit cycles , maintaining transient capping while depolymerizing the filament . Second , twinfilin may competitively dissociate from CP in favor of CP - capped filaments due to the superior affinity for filament ends relative to CP alone , K d of 13 and 50 nM , respectively ( 17 , 27 ) . This mechanism assumes that twinfilin affinity for the CP - capped fila - ment is also high ; however , the value is unknown and possibly diffi - cult to measure due to the uncapping mechanism . Last , the twinfilin / CP complex may be dissociated at the membrane by PIP 2 or by competition with CPI proteins , such as CARMIL ( 27 , 37 ) . The structure of the actin / twinfilin / CP complex suggests that a second possible mechanism of filament uncapping may operate in the presence of V - 1 . In this mechanism , the twinfilin / CP complex \u201cwobbles\u201d ( 44 ) at the barbed end of a filament , remaining attached through the high - affinity actin - binding site on twinfilin D2 , while twinfilin D1 dissociates from actin ( Fig . 6 , A and C ) ( 16 ) . The ex - tended portion of the twinfilin tail ( residues 316 to 322 , between the end of D2 and the first tail residue with significant contact with CP ) , which we hypothesize is under tension , may aid the wobble process . There is homology between twinfilin and CARMIL in this region ( 317 - HAHKQSFAKP - 326 in twinfilin - 1 and 981 - KLEHFTKLRP - 990 in CARMIL1 ) ( 27 ) , which may allow the twinfilin tail to associate more intimately with CP in a manner similar to the CARMIL CPI interaction with CP ( Fig . 3A ) ( 34 ) . Association of this region of twinfilin with CP would hold CP away from the actin subunit , to which it was previously bound via the \uf061 - tentacle , thus freeing the actin - binding site on the \uf061 - tentacle to bind to V - 1 , and preventing the \uf061 - tentacle from reassociating with actin . The partially bound complex would then dissociate as twinfilin / CP / V - 1 bound to a sin - gle actin , with the high - affinity twinfilin D2 inducing a G - actin \u2013 like conformation in the actin subunit to destabilize its association with the filament ( 16 ) . Flow - cell total internal reflection fluorescence microscopy observations of CP uncapping by twinfilin detected en - hanced uncapping in the presence of V - 1 ( 28 ) . This suggests that V - 1 may play an active role in twinfilin - aided uncapping as hypoth - esized in this mechanism ( Fig . 6 , A and C ) . In summary , the actin / twinfilin / CP complex structure clarifies the possible mechanisms by which twinfilin operates as a diffusing uncapping agent . The majority of the in vitro observations outlined above can be rationalized in terms of twinfilin\u2019s proposed role in transforming CP from a strong capping agent into a transiently cap - ping depolymerization complex , which aids in the recycling of actin monomers and is sensitive to the presence of V - 1 . In the bulk cyto - plasm , the outcome of twinfilin\u2019s uncapping activities is likely to be actin depolymerization while maintaining transient capping . This role contrasts starkly to the CPI - containing proteins , which are target - bound uncapping agents that are hypothesized to engender actin polymerization for membrane remodeling ( 45 , 46 ) . The location of uncapping , in either actin assembly - or disassembly - rich regions of a cell , will determine whether filament polymerization or depolym - erization will be the product of the uncapping . Any twinfilin - induced uncapping at a membrane may therefore result in the opposite ac - tivity , actin polymerization . Because twinfilin / CP - bound filaments , relative to CP - bound filaments , have enhanced barbed - end fila - ment dynamics , we propose that cells are able to differentially regu - late these actin filament barbed - end binding states to dictate the Fig . 6 . Models of the filament uncapping . ( A ) Cartoon comparison of CP and CP / twinfilin association at the barbed end of a filament . ( B ) Twinfilin - aided uncapping is a result of twinfilin ADF - H domains inducing G - actin \u2013 like conformations in the terminal two actin protomers , weakening actin : actin interactions in the filament , leading to the dissociation of the complex . ( C ) Twinfilin / V - 1 \u2013 aided uncapping requires space for V - 1 to reach its binding site on CP via a wobble state of the twinfilin - bound complex . Once V - 1 is bound , CP is unable to reassociate with actin . In vitro effects of twinfilin alone on actin filaments and comparisons of uncapping models in the absence of twinfilin are shown in fig . S9 . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024 Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 8 of 10 lifetimes of individual actin filaments . The proposed mechanism of depolymerization while maintaining transient filament capping adds to the evidence that actin filament ends are highly controlled and are rarely free in mammalian cells . During depolymerization , filament barbed ends are controlled by twinfilin / CP and filament pointed ends by CAP ( 33 ) . During polymerization modes , the fila - ment barbed ends can be regulated by formins ( 47 ) , or alternatively by the VASP family proteins ( 48 ) , following either ARP2 / 3 nucle - ation or filament uncapping at the membrane ( 34 , 40 , 49 , 50 ) . In each case , the filament end is protected . MATERIALS AND METHODS Protein expression and purification The gene sequences encoding full - length human twinfilin - 1 and profilin - 1 were codon - optimized for Escherichia coli , synthesized ( GenScript ) , and cloned into a pSY5 vector that includes an N - terminal eight - histidine tag followed by a human rhinovirus 3C protease cleavage site ( 51 ) . The mouse full - length CP ( \uf061 1 / \uf062 2 ) con - struct was provided by P . Lappalainen ( University of Helsinki , Finland ) . The construct was in the pRSFDuet - 1 vector designed for coexpres - sion of two interacting target proteins and contains an N - terminal six - histidine tag on the \uf061 - subunit ( 52 ) . All constructs encoding proteins of interest were transformed and recombinantly expressed in phage - resistant E . coli strain BL21 Star ( DE3 ) ( New England Biolabs ) . Fresh LB medium supple - mented with either ampicillin ( pSY5 ) or kanamycin ( pRSFDuet - 1 ) was inoculated with respective overnight cultures and shaken at 37\u00b0C until the cell density reached OD 600 ( optical density at 600 nm ) ~ 0 . 6 . Cells were induced for protein expression with 0 . 25 mM isopropyl - \uf062 - d - thiogalactopyranoside at 16\u00b0C overnight . Cells were harvested via centrifugation at 4000 g at 4\u00b0C for 1 hour , and the pellets were resuspended in 50 ml of His - binding buffer [ 50 mM tris - HCl ( pH 7 . 5 ) , 500 mM NaCl , 20 mM imidazole , and one protease inhibitor tablet ] . The cells were lysed by sonication with a Vibra - Cell ultrasonic processor and clarified by centrifuga - tion at 19 , 000 rpm in SS - 34 rotor using an RC 5C Plus centrifuge ( Sorvall ) at 4\u00b0C for 1 hour followed by filtration through a 0 . 45 - \uf06d m Minisart syringe filter ( Sartorius ) . The proteins were purified on an \u00c4KTAxpress system ( GE Healthcare ) by affinity chromatography using 5 ml of HisTrap FF column with or without ( CP ) on - column cleavage of the His - tag with human rhinovirus 3C protease . Proteins purified with the His - tag were eluted in buffer containing 50 mM tris - HCl ( pH 7 . 5 ) , 500 mM NaCl , and 500 mM imidazole . The eluted proteins were concentrated to 5 ml and subjected to size exclusion chromatography ( SEC ) in a Superdex 75 pg column ( GE Healthcare ) preequilibrated with buffer containing 50 mM tris - HCl ( pH 7 . 5 ) and 150 mM NaCl . All purified proteins were verified by SDS \u2013 polyacrylamide gel electrophoresis ( PAGE ) before being snap - frozen and stored at \u221280\u00b0C . MBP - tagged twinfilin tail constituting residues Gln 321 - Asp 350 was polymerase chain reaction ( PCR ) \u2013 amplified from full - length human twinfilin - 1 expression plasmid and inserted into pSY7 vector , which incorporates a cleavable N - terminal histidine and MBP tag . MBP control was prepared by expressing pSY7 vector alone in BL21 ( DE3 ) E . coli . Both MBP - tagged twinfilin tail and MBP were purified by affinity chromatography using 1 ml of HisTrap FF column with on - column digestion with human rhinovirus 3C protease to remove the histidine tag . The proteins were further purified by gel filtration chromatography with a Superdex 75 pg column ( GE Healthcare ) equilibrated with 50 mM tris - HCl ( pH 9 ) and 150 mM NaCl . Human V - 1 ( pGEX - 6P - 3 vector ) was expressed as a glutathione S - transferase ( GST ) fusion E . coli BL21 Star ( DE3 ) and affinity - purified on an \u00c4KTAxpress system ( GE Healthcare ) by loading cleared lysate onto 1 ml of GSTrap FF Column ( GE Healthcare ) equilibrated with 50 mM tris - HCl ( pH 7 . 4 ) , 150 mM NaCl , and 1 mM dith - iothreitol ( DTT ) . GST tag was removed by on - column digestion with human rhinovirus 3C protease overnight followed by further purification by gel filtration chromatography . CARMIL CP \u2013 binding region ( CBR ) was expressed and purified as previously described ( 34 ) . Preparation of the actin , twinfilin , and CP complex Rabbit skeletal muscle actin was purified from skeletal muscle ace - tone powder ( Pel - Freez ) ( 53 , 54 ) . The protein was subjected to a final SEC using HiLoad Superdex 200 on an \u00c4KTA Prime system . The purity of G - actin was assessed by SDS - PAGE , and concentra - tion was determined by measuring OD at 290 nm . The protein complex was prepared by mixing human twinfilin - 1 , mouse CP \uf061 1 \uf062 2 , and rabbit actin in buffer A [ 2 mM tris - HCl ( pH 7 . 5 ) , 0 . 2 mM ATP , 0 . 5 mM DTT , 1 mM NaN 3 , and 0 . 1 mM CaCl 2 ] at a molar ratio of 1 : 1 : 2 . 5 . The mixture was incubated on ice for 10 min to allow com - plex formation and then purified by SEC ( HiLoad 16 / 60 Superdex 200 preequilibrated with buffer A ) on an \u00c4KTAxpress system ( GE Healthcare ) . Fractions corresponding to ultraviolet absorption peak were analyzed by SDS - PAGE gel to verify complex formation . Crystallization and crystal optimization The fractions corresponding to the peak from SEC were pooled to - gether , concentrated with Vivaspin 20 MWCO 10 , 000 concentrator ( Sartorius ) to approximately 20 mg / ml , and subjected to commer - cial crystallization screens . Crystallization screens were set up as sitting drops in three - drop Intelli - Plate 96 ( Hampton Research ) in 2 : 1 , 1 : 1 , and 1 : 2 ratios consisting of protein solution and precipitant and stored in room temperature ( 25\u00b0C ) . Crystal hits were observed in a number of conditions in the JBScreen Classic HTS I screen after 3 days . One condition [ 12 % polyethylene glycol ( PEG ) 8000 , 10 % glycerol , and 500 mM potassium chloride ] produced the best shaped crystals composed of thin rods . A seed bead kit ( Hampton Research ) was used to generate seed stocks of protein crystals for further opti - mization . Crystals were set up using the hanging - drop vapor diffu - sion method in 2 - \uf06d l drops of 1 : 1 ratio consisting of protein solution and seed stock in the crystallization condition ( 12 % PEG 8000 , 10 % glycerol , and 500 mM potassium chloride ) . Crystal data processing and structure determination Harvested crystals were soaked in 25 % glycerol before being care - fully fished with cryoloops and flash - frozen in liquid nitrogen for subsequent data collection at the National Synchrotron Radiation Research Center , Taiwan . Indexing , scaling , and merging of data - sets were carried out in HKL2000 ( 55 ) . The crystal structure twinfilin / CP / actin complex was solved at a resolution of 3 . 2 \u00c5 by molecular replacement using Phaser ( 56 ) , sequentially searching for one copy of mouse twinfilin - 1 D2 ( 3DAW ) ( 19 ) and one copy of chicken CP hetero - dimer ( 3AA7 ) ( 35 ) followed by a second copy of mouse twinfilin - 1 D2 ( 3DAW ) ( 19 ) . The resultant model was rebuilt by hand , and the structure was subjected to several rounds of refinement using Phenix ( 56 ) and further manual rebuilding in COOT ( 57 ) . The crystal data collection and structure refinement statistics are presented in table S1 . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024 Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 9 of 10 Pyrene - actin polymerization assays Pyrene - actin polymerization assays to monitor filament uncapping were performed in 100 - \uf06d l reactions containing 2 \uf06d M rabbit skeletal muscle G - actin ( 10 % pyrene \u2013 labeled ) , 0 . 5 \uf06d M F - actin seeds , 2 . 8 \uf06d M profilin ( to prevent pointed - end polymerization ) , 100 nM CP , and variable concentrations of the test proteins . In all pyrene - actin po - lymerization assays , the final concentrations of twinfilin , CP , CARMIL , V - 1 , MBP - tagged tail , and MBP were set at 1 \uf06d M , 100 nM , 250 nM , 5 \uf06d M , 10 \uf06d M , and 10 \uf06d M , respectively . The components were mixed in buffer A [ 2 mM tris - HCl ( pH 7 . 5 ) , 0 . 2 mM ATP , 0 . 5 mM DTT , 1 mM NaN 3 , and 0 . 1 mM CaCl 2 ] to a final volume of 70 \uf06d l in a black flat - bottom 96 - well plate ( Corning ) . To prepare actin filament seed stocks , 5 \uf06d M G - actin was polym - erized for 1 hour at room temperature , after which the filaments were mechanically sheared by repeatedly passing the F - actin solu - tion through a 0 . 7 - mm - diameter needle for 1 min . Thereafter , 10 \uf06d l of the F - actin seed stock was mixed with 10 \uf06d l of 1 \uf06d M CP and left for 5 min to allow barbed - end capping of the F - actin seeds . As a control , the F - actin seed stock was mixed with buffer A , without CP . Then , 20 \uf06d l of this precapped F - seed stock , or control , was add - ed to pyrene - actin polymerization mixture and actin polymeriza - tion initiated by addition of 10 \uf06d l of 10\u00d7 KMEI buffer ( 500 mM KCl , 10 mM MgCl 2 , 10 mM EGTA , and 100 mM imidazole , pH 7 . 4 ) in a total reaction volume of 100 \uf06d l . The pyrene fluorescence intensities were monitored immediately on a Safire 2 fluorimeter plate reader ( Tecan ) with excitation and emission wavelengths set at 365 and 407 nm , respectively . SUPPLEMENTARY MATERIALS Supplementary material for this article is available at http : / / advances . sciencemag . org / cgi / content / full / 7 / 5 / eabd5271 / DC1 REFERENCES AND NOTES 1 . O . L . Mooren , B . J . Galletta , J . A . Cooper , Roles for actin assembly in endocytosis . Annu . Rev . Biochem . 81 , 661 \u2013 686 ( 2012 ) . 2 . T . D . Pollard , J . A . Cooper , Actin , a central player in cell shape and movement . Science 326 , 1208 \u2013 1212 ( 2009 ) . 3 . J . W . Rudy , Actin dynamics and the evolution of the memory trace . Brain Res . 1621 , 17 \u2013 28 ( 2015 ) . 4 . T . D . Pollard , Actin and actin - binding proteins . Cold Spring Harb . Perspect . Biol . 8 , a018226 ( 2016 ) . 5 . A . D . Siripala , M . D . Welch , SnapShot : Actin regulators I . Cell 128 , 626 ( 2007 ) . 6 . M . Poukkula , E . Kremneva , M . Serlachius , P . Lappalainen , Actin - depolymerizing factor homology domain : A conserved fold performing diverse roles in cytoskeletal dynamics . Cytoskeleton ( Hoboken ) 68 , 471 \u2013 490 ( 2011 ) . 7 . C . Ak\u0131l , L . T . Tran , M . Orhant - Prioux , Y . Baskaran , E . Manser , L . Blanchoin , R . C . Robinson , Insights into the evolution of regulated actin dynamics via characterization of primitive gelsolin / cofilin proteins from Asgard archaea . Proc . Natl . Acad . Sci . U . S . A . 117 , 19904 \u2013 19913 ( 2020 ) . 8 . P . W . Gunning , U . Ghoshdastider , S . Whitaker , D . Popp , R . C . Robinson , The evolution of compositionally and functionally distinct actin filaments . J . Cell Sci . 128 , 2009 \u2013 2019 ( 2015 ) . 9 . C . Ak\u0131l , Y . Kitaoku , L . T . Tran , D . Liebl , H . Choe , D . Muengsaen , W . Suginta , A . Schulte , R . C . Robinson , Mythical origins of the actin cytoskeleton . Curr . Opin . Cell Biol . 68 , 55 \u2013 63 ( 2020 ) . 10 . F . Rivero , F . Cvrckov\u00e1 , Origins and evolution of the actin cytoskeleton . Adv . Exp . Med . Biol . 607 , 97 \u2013 110 ( 2007 ) . 11 . B . L . Goode , D . G . Drubin , P . Lappalainen , Regulation of the cortical actin cytoskeleton in budding yeast by twinfilin , a ubiquitous actin monomer - sequestering protein . J . Cell Biol . 142 , 723 \u2013 733 ( 1998 ) . 12 . S . Ono , Mechanism of depolymerization and severing of actin filaments and its significance in cytoskeletal dynamics . Int . Rev . Cytol . 258 , 1 \u2013 82 ( 2007 ) . 13 . W . A . Elam , H . Kang , E . M . De la Cruz , Biophysics of actin filament severing by cofilin . FEBS Lett . 587 , 1215 \u2013 1219 ( 2013 ) . 14 . M . Vartiainen , P . J . Ojala , P . Auvinen , J . Peranen , P . Lappalainen , Mouse A6 / twinfilin is an actin monomer - binding protein that localizes to the regions of rapid actin dynamics . Mol . Cell . Biol . 20 , 1772 \u2013 1783 ( 2000 ) . 15 . D . Wang , L . Zhang , G . Zhao , G . Wahlstrom , T . I . Heino , J . Chen , Y . Q . Zhang , Drosophila twinfilin is required for cell migration and synaptic endocytosis . J . Cell Sci . 123 , 1546 \u2013 1556 ( 2010 ) . 16 . P . J . Ojala , V . O . Paavilainen , M . K . Vartiainen , R . Tuma , A . G . Weeds , P . Lappalainen , The two ADF - H domains of twinfilin play functionally distinct roles in interactions with actin monomers . Mol . Biol . Cell 13 , 3811 \u2013 3821 ( 2002 ) . 17 . E . Helfer , E . M . Nevalainen , P . Naumanen , S . Romero , D . Didry , D . Pantaloni , P . Lappalainen , M . F . Carlier , Mammalian twinfilin sequesters ADP - G - actin and caps filament barbed ends : Implications in motility . EMBO J . 25 , 1184 \u2013 1195 ( 2006 ) . 18 . V . O . Paavilainen , M . Hellman , E . Helfer , M . Bovellan , A . Annila , M . F . Carlier , P . Permi , P . Lappalainen , Structural basis and evolutionary origin of actin filament capping by twinfilin . Proc . Natl . Acad . Sci . U . S . A . 104 , 3113 \u2013 3118 ( 2007 ) . 19 . V . O . Paavilainen , E . Oksanen , A . Goldman , P . Lappalainen , Structure of the actin - depolymerizing factor homology domain in complex with actin . J . Cell Biol . 182 , 51 \u2013 59 ( 2008 ) . 20 . A . B . Johnston , A . Collins , B . L . Goode , High - speed depolymerization at actin filament ends jointly catalysed by Twinfilin and Srv2 / CAP . Nat . Cell Biol . 17 , 1504 \u2013 1511 ( 2015 ) . 21 . D . M . Hilton , R . M . Aguilar , A . B . Johnston , B . L . Goode , Species - specific functions of twinfilin in actin filament depolymerization . J . Mol . Biol . 430 , 3323 \u2013 3336 ( 2018 ) . 22 . J . B . Moseley , K . Okada , H . I . Balcer , D . R . Kovar , T . D . Pollard , B . L . Goode , Twinfilin is an actin - filament - severing protein and promotes rapid turnover of actin structures in vivo . J . Cell Sci . 119 , 1547 \u2013 1557 ( 2006 ) . 23 . S . Falck , V . O . Paavilainen , M . A . Wear , J . G . Grossmann , J . A . Cooper , P . Lappalainen , Biological role and structural mechanism of twinfilin - capping protein interaction . EMBO J . 23 , 3010 \u2013 3019 ( 2004 ) . 24 . M . A . Wear , A . Yamashita , K . Kim , Y . Maeda , J . A . Cooper , How capping protein binds the barbed end of the actin filament . Curr . Biol . 13 , 1531 \u2013 1537 ( 2003 ) . 25 . J . A . Cooper , D . Sept , New insights into mechanism and regulation of actin capping protein . Int . Rev . Cell Mol . Biol . 267 , 183 \u2013 206 ( 2008 ) . 26 . A . Yamashita , K . Maeda , Y . Maeda , Crystal structure of CapZ : Structural basis for actin filament barbed end capping . EMBO J . 22 , 1529 \u2013 1538 ( 2003 ) . 27 . A . B . Johnston , D . M . Hilton , P . McConnell , B . Johnson , M . T . Harris , A . Simone , G . K . Amarasinghe , J . A . Cooper , B . L . Goode , A novel mode of capping protein - regulation by twinfilin . eLife 7 , e41313 ( 2018 ) . 28 . M . Hakala , H . Wioland , M . Tolonen , A . Jegou , G . Romet - Lemonne , P . Lappalainen , Twinfilin uncaps filament barbed ends to promote turnover of lamellipodial actin networks . bioRxiv 864769 [ Preprint ] . 4 December 2019 . https : / / doi . org / 10 . 1101 / 864769 . 29 . A . Narita , S . Takeda , A . Yamashita , Y . Maeda , Structural basis of actin filament capping at the barbed - end : A cryo - electron microscopy study . EMBO J . 25 , 5626 \u2013 5633 ( 2006 ) . 30 . L . Urnavicius , C . K . Lau , M . M . Elshenawy , E . Morales - Rios , C . Motz , A . Yildiz , A . P . Carter , Cryo - EM shows how dynactin recruits two dyneins for faster movement . Nature 554 , 202 \u2013 206 ( 2018 ) . 31 . T . Oda , M . Iwasa , T . Aihara , Y . Maeda , A . Narita , The nature of the globular - to fibrous - actin transition . Nature 457 , 441 \u2013 445 ( 2009 ) . 32 . S . Z . Chou , T . D . Pollard , Mechanism of actin polymerization revealed by cryo - EM structures of actin filaments with three different bound nucleotides . Proc . Natl . Acad . Sci . U . S . A . 116 , 4265 \u2013 4274 ( 2019 ) . 33 . T . Kotila , H . Wioland , G . Enkavi , K . Kogan , I . Vattulainen , A . J\u00e9gou , G . Romet - Lemonne , P . Lappalainen , Mechanism of synergistic actin filament pointed end depolymerization by cyclase - associated protein and cofilin . Nat . Commun . 10 , 5320 ( 2019 ) . 34 . M . Hernandez - Valladares , T . Kim , B . Kannan , A . Tung , A . H . Aguda , M . Larsson , J . A . Cooper , R . C . Robinson , Structural characterization of a capping protein interaction motif defines a family of actin filament regulators . Nat . Struct . Mol . Biol . 17 , 497 \u2013 503 ( 2010 ) . 35 . S . Takeda , S . Minakata , R . Koike , I . Kawahata , A . Narita , M . Kitazawa , M . Ota , T . Yamakuni , Y . Maeda , Y . Nitanai , Two distinct mechanisms for actin capping protein regulation\u2014 Steric and allosteric inhibition . PLOS Biol . 8 , e1000416 ( 2010 ) . 36 . T . Kim , G . E . Ravilious , D . Sept , J . A . Cooper , Mechanism for CARMIL protein inhibition of heterodimeric actin - capping protein . J . Biol . Chem . 287 , 15251 \u2013 15262 ( 2012 ) . 37 . S . Palmgren , P . J . Ojala , M . A . Wear , J . A . Cooper , P . Lappalainen , Interactions with PIP2 , ADP - actin monomers , and capping protein regulate the activity and localization of yeast twinfilin . J . Cell Biol . 155 , 251 \u2013 260 ( 2001 ) . 38 . S . Palmgren , M . Vartiainen , P . Lappalainen , Twinfilin , a molecular mailman for actin monomers . J . Cell Sci . 115 , 881 \u2013 886 ( 2002 ) . 39 . K . Kim , A . Yamashita , M . A . Wear , Y . Ma\u00e9da , J . A . Cooper , Capping protein binding to actin in yeast : Biochemical mechanism and physiological relevance . J . Cell Biol . 164 , 567 \u2013 580 ( 2004 ) . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024 Mwangangi et al . , Sci . Adv . 2021 ; 7 : eabd5271 27 January 2021 S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E 10 of 10 40 . D . A . Schafer , P . B . Jennings , J . A . Cooper , Dynamics of capping protein and actin assembly in vitro : Uncapping barbed ends by polyphosphoinositides . J . Cell Biol . 135 , 169 \u2013 179 ( 1996 ) . 41 . K . Tanaka , S . Takeda , K . Mitsuoka , T . Oda , C . Kimura - Sakiyama , Y . Maeda , A . Narita , Structural basis for cofilin binding and actin filament disassembly . Nat . Commun . 9 , 1860 ( 2018 ) . 42 . B . Xue , R . C . Robinson , Guardians of the actin monomer . Eur . J . Cell Biol . 92 , 316 \u2013 332 ( 2013 ) . 43 . T . Kotila , K . Kogan , G . Enkavi , S . Guo , I . Vattulainen , B . L . Goode , P . Lappalainen , Structural basis of actin monomer re - charging by cyclase - associated protein . Nat . Commun . 9 , 1892 ( 2018 ) . 44 . N . Bhattacharya , S . Ghosh , D . Sept , J . A . Cooper , Binding of myotrophin / V - 1 to actin - capping protein : Implications for how capping protein binds to the filament barbed end . J . Biol . Chem . 281 , 31021 \u2013 31030 ( 2006 ) . 45 . A . I . Fokin , V . David , K . Oguievetskaia , E . Derivery , C . E . Stone , L . Cao , N . Rocques , N . Molinie , V . Henriot , M . Aumont - Nicaise , M . - V . Hinckelmann , F . Saudou , C . Le Clainche , A . P . Carter , G . Romet - Lemonne , A . M . Gautreau , The Arp1 / 11 minifilament of dynactin primes the endosomal Arp2 / 3 complex . bioRxiv 2020 . 05 . 29 . 123372 [ Preprint ] . 29 May 2020 . https : / / doi . org / 10 . 1101 / 2020 . 05 . 29 . 123372 . 46 . M . H . Lanier , P . McConnell , J . A . Cooper , Cell migration and invadopodia formation require a membrane - binding domain of CARMIL2 . J . Biol . Chem . 291 , 1076 \u2013 1091 ( 2016 ) . 47 . S . H . Zigmond , M . Evangelista , C . Boone , C . Yang , A . C . Dar , F . Sicheri , J . Forkey , M . Pring , Formin leaky cap allows elongation in the presence of tight capping proteins . Curr . Biol . 13 , 1820 \u2013 1823 ( 2003 ) . 48 . S . D . Hansen , R . D . Mullins , VASP is a processive actin polymerase that requires monomeric actin for barbed end association . J . Cell Biol . 191 , 571 \u2013 584 ( 2010 ) . 49 . M . D . Welch , A . Iwamatsu , T . J . Mitchison , Actin polymerization is induced by Arp2 / 3 protein complex at the surface of Listeria monocytogenes . Nature 385 , 265 \u2013 269 ( 1997 ) . 50 . P . A . Janmey , C . Chaponnier , S . E . Lind , K . S . Zaner , T . P . Stossel , H . L . Yin , Interactions of gelsolin and gelsolin - actin complexes with actin . Effects of calcium on actin nucleation , filament severing , and end blocking . Biochemistry 24 , 3714 \u2013 3723 ( 1985 ) . 51 . S . Nag , Q . Ma , H . Wang , S . Chumnarnsilpa , W . L . Lee , M . Larsson , B . Kannan , M . Hernandez - Valladares , L . D . Burtnick , R . C . Robinson , Ca 2 + binding by domain 2 plays a critical role in the activation and stabilization of gelsolin . Proc . Natl . Acad . Sci . U . S . A . 106 , 13713 \u2013 13718 ( 2009 ) . 52 . Y . Soeno , H . Abe , S . Kimura , K . Maruyama , T . Obinata , Generation of functional beta - actinin ( CapZ ) in an E . coli expression system . J . Muscle Res . Cell Motil . 19 , 639 \u2013 646 ( 1998 ) . 53 . H . Wang , R . C . Robinson , L . D . Burtnick , The structure of native G - actin . Cytoskeleton ( Hoboken ) 67 , 456 \u2013 465 ( 2010 ) . 54 . J . A . Spudich , S . Watt , The regulation of rabbit skeletal muscle contraction . I . Biochemical studies of the interaction of the tropomyosin - troponin complex with actin and the proteolytic fragments of myosin . J . Biol . Chem . 246 , 4866 \u2013 4871 ( 1971 ) . 55 . Z . Otwinowski , W . Minor , Processing of X - ray diffraction data collected in oscillation mode . Methods Enzymol . 276 , 307 \u2013 326 ( 1997 ) . 56 . P . D . Adams , P . V . Afonine , G . Bunkoczi , V . B . Chen , I . W . Davis , N . Echols , J . J . Headd , L . W . Hung , G . J . Kapral , R . W . Grosse - Kunstleve , A . J . McCoy , N . W . Moriarty , R . Oeffner , R . J . Read , D . C . Richardson , J . S . Richardson , T . C . Terwilliger , P . H . Zwart , PHENIX : A comprehensive Python - based system for macromolecular structure solution . Acta Crystallogr . D Biol . Crystallogr . 66 , 213 \u2013 221 ( 2010 ) . 57 . P . Emsley , B . Lohkamp , W . G . Scott , K . Cowtan , Features and development of Coot . Acta Crystallogr . D Biol . Crystallogr . 66 , 486 \u2013 501 ( 2010 ) . 58 . L . Holm , L . M . Laakso , Dali server update . Nucleic Acids Res . 44 , W351 \u2013 W355 ( 2016 ) . 59 . S . Nag , M . Larsson , R . C . Robinson , L . D . Burtnick , Gelsolin : The tail of a molecular gymnast . Cytoskeleton ( Hoboken ) 70 , 360 \u2013 384 ( 2013 ) . Acknowledgments : We thank B . Xue for technical advice , S . Takeda and B . Xue for valuable discussions , P . Lappalainen ( University of Helsinki , Finland ) and J . Cooper ( Washington University School of Medicine , USA ) for constructs , and the experimental facility and the technical services provided by the Synchrotron Radiation Protein Crystallography Facility of the National Core Facility Program for Biotechnology , Ministry of Science and Technology and the National Synchrotron Radiation Research Center , a national user facility supported by the Ministry of Science and Technology , Taiwan . Funding : This work was supported by Biomedical Research Council of A * STAR under the Singapore International Graduate Award ( SINGA ) scholarship , by NMRC Grant OFIRG / 027 / 2016 , and by Human Frontiers Science Program award RGP0028 / 2018 . Author contributions : D . M . M . and R . C . R . conceived experiments , analyzed data , and wrote the paper . D . M . M . performed experiments . R . C . R . and E . M . supervised the work . Competing interests : The authors declare that they have no competing interests . Data and materials availability : The atomic coordinates and structure factors have been deposited in the Protein Data Bank ( PDB ) under the accession code 7CCC . All other data are available from the corresponding author upon reasonable request . Submitted 26 June 2020 Accepted 8 December 2020 Published 27 January 2021 10 . 1126 / sciadv . abd5271 Citation : D . M . Mwangangi , E . Manser , R . C . Robinson , The structure of the actin filament uncapping complex mediated by twinfilin . Sci . Adv . 7 , eabd5271 ( 2021 ) . D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g on A p r il 09 , 2024",
    "chanComparingDifferentSensemaking2016": "Compar ing Different Sensemaking Approaches for Large - Scale Ideation Joel Chan 1 Steven Dang 1 Steven P . Dow 2 1 Human - Computer Interaction Institute Carnegie Mellon University , Pittsburgh , PA USA { joelchuc , stevenda } @ cs . cmu . edu 2 Cognitive Science Department UC San Diego , La Jolla , CA , USA spdow @ ucsd . edu ABSTRACT Large - scale idea generation platforms often expose ideators to previous ideas . However , research suggests people generate better ideas if they see abstracted solution paths ( e . g . , descriptions of solution approaches generated through human sensemaking ) rather than being inundated with all prior ideas . Automated and semi - automated methods can also offer interpretations of earlier ideas . To benefit from sensemaking in practice with limited resources , ideation platform developers need to weigh the cost - quality tradeoffs of different methods for surfacing solution paths . To explore this , we conducted an online study where 245 participants generated ideas for two problems in one of five conditions : 1 ) no stimuli , 2 ) exposure to all prior ideas , or solution paths extracted from prior ideas using 3 ) a fully automated workflow , 4 ) a hybrid human - machine approach , and 5 ) a fully manual approach . Contrary to expectations , human - generated paths did not improve ideation ( as meas - ured by fluency and breadth of ideation ) over simply showing all ideas . Machine - generated paths sometimes significantly improved fluency and breadth of ideation over no ideas ( although at some cost to idea quality ) . These findings suggest that automated sensemaking can improve idea generation , but we need more research to understand the value of human sensemaking for crowd ideation . Author Keywords Creativity ; crowdsourcing ; brainstorming ; sensemaking ACM Classification Keywords Human - centered computing ~ Collaborative and social computing INTRODUCTION Large - scale ideation platforms show potential for solving difficult problems by leveraging the scale and diversity of online crowds [ 4 , 7 , 37 , 59 ] . Current approaches excel at collecting many solutions , especially in a competitive paradigm where crowd innovators work in isolation from each other . However , a significant concern with this ap - proach is that it often results in wasted effort as newcomers retrace obvious but suboptimal solution paths [ 37 ] . For example , in their recent 10 to the 100 th crowdsourced innovation project , Google had to recruit 3 , 000 of their employees to prune the 150 , 000 ideas received from the crowd , pushing the project nine months behind schedule . Cognitive research on creative ideation suggests that people can produce better ideas if they are able to learn from the efforts of others , recombining ideas into new ideas and iterating on new ideas to improve them [ 19 , 21 , 32 , 55 ] . However , at crowd scale , simply exposing all ideas to everyone may not be the most effective strategy . Ideators may not have sufficient time or cognitive resources to sift through potentially hundreds to thousands of other ideas . Consequently , they may only superficially process and build on ideas rather than leveraging them to generate new insights [ 35 , 36 ] . The presence of superficial details in raw ideas might also lead to cognitive fixation [ 34 , 44 , 56 ] . For these reasons , abstracted solution paths \u2014 which distill essential solution approaches from a number of different ideas while avoiding superficial details \u2014 may provide a better way for ideators to interact with prior ideas . Available sensemaking strategies for abstracting solution paths can be placed on a hypothesized cost - quality tradeoff continuum , where sensemaking quality ( especially its benefit for ideation ) is a function of the strategy\u2019s cost . At the high end of the continuum ( high cost , hypothesized high quality ) , there is a mature set of design synthesis strategies ( e . g . , affinity diagramming ) employed effectively by design teams to make sense of a solution space [ 5 , 29 , 38 ] ; these strategies require significant manual human effort , and therefore may be hard to scale to crowd ideation . However , these sensemaking outputs should hypothetically have high value for ideation . At the low end ( low cost , hypothesized low quality ) sit a number of automated approaches ( e . g . , unsupervised machine learning methods like Latent Seman - tic Indexing ( LSI ) [ 18 ] and K - means clustering ) that can extract semantic themes in text - based idea sets very quickly and efficiently . However , the intrinsic quality of these themes ( e . g . , as measured by correspondence with gold Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page . Copyrights for components of this work owned by others than the author ( s ) must be honored . Abstracting with credit is permitted . To copy otherwise , or republish , to post on servers or to redistribute to lists , requires prior specific permission and / or a fee . Request permissions from Permissions @ acm . org . CHI ' 16 , May 07 - 12 , 2016 , San Jose , CA , USA Copyright is held by the owner / author ( s ) . Publication rights licensed to ACM . ACM 978 - 1 - 4503 - 3362 - 7 / 16 / 05 \u2026 $ 15 . 00 DOI : http : / / dx . doi . org / 10 . 1145 / 2858036 . 2858178 Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2717 standard human clusters / categories ) is often relatively low ( certainly lower than human - produced themes ) . Thus , these sensemaking outputs should hypothetically have relatively low value for ideation . In between sit hybrid approaches ( medium cost , medium quality ) that combine human and machine effort [ 39 , 58 , 62 , 63 , 65 ] . Some of these approaches may require optimization of complex machine learning parameters to be successful , limiting their accessibility to large scale ideation platforms . But , relatively simple and accessible workflows are also possible , where humans label initial machine clusterings from algorithms like K - means . Ideation platform developers would benefit greatly from knowing the cost - quality tradeoffs of sensemaking ap - proaches and how to use them at scale with limited re - sources , but prior work has rarely compared these ap - proaches directly . To address this important gap , we ex - plore two main research questions : 1 ) Do abstracted solution paths inspire more and better ideas compared to seeing all raw ideas or no ideas at crowd scale ? 2 ) If so , how does the sensemaking strategy for abstract solution paths\u2014from fully automated to hybrid human - machine to fully manual\u2014affect the quantity and quali - ty of generated ideas ? In an online study , 245 Amazon Mechanical Turk ( mTurk ) participants generated ideas for two innovation problems ( generating ideas for a novel fabric technology and for improving mTurk\u2019s mobile experience ) in one of five conditions : 1 ) no stimulation ( control ) , 2 ) viewing all prior ideas for the problem ( all ideas ) , 3 ) with solution paths extracted from human - labeled paths generated for human - generated clusters of ideas ( human - human ) , 4 ) human - labeled paths generated for machine - generated clusters ( machine - human ) , or 5 ) machine - labeled paths generated for machine - generated clusters ( machine - machine ) . Our main finding is that an off - the - shelf machine approach ( as exemplified by our machine - machine condition ) can yield comparable benefits to a costly fully manual approach ( as exemplified by our human - human condition ) . Contrary to expectations , human - generated solution paths do not improve ideation over either simply showing all ideas or no ideas at all , as measured by fluency ( total number of ideas ) and breadth of ideation ( mean pairwise distance in LSI representation of the solution space ) . However , machine - generated solution paths sometimes improve fluency and breadth of ideation over no ideas ( although at a slight cost to idea quality ) . These findings suggest that large - scale ideation could benefit from automated sensemaking , but more research is needed to better understand the value of human - generated sensemaking in a crowd setting . This paper contributes : 1 ) Empirical findings on the relative value of different sensemaking approaches for ideation 2 ) Evidence that a simple , easy to implement approach to automated sensemaking ( using LSI and k - means clus - tering ) can improve fluency and breadth of ideation . RELATED WORK Effective Collaborative Ideation There is consensus in the literature on creative cognition that seeing what others have thought and / or are currently thinking about the problem can increase people\u2019s ability to generate creative ideas ( i . e . , ideas that are both novel and of high quality ) [ 19 , 21 , 32 , 55 ] . For example , people can generate more creative ideas when they have access to example designs [ 41 , 43 , 53 \u2013 55 ] , draw analogies to past experiences [ 11 , 17 , 33 , 64 ] , or see some ideas from others who are working on the same problem [ 8 , 19 ] . Some studies have also shown that building on others\u2019 ideas not only improves individual creativity , but also maximizes the community\u2019s ability to reach an optimal solution [ 8 , 61 ] . However , these benefits of collaborative ideation can be challenging to realize at crowd scale ( typically hundreds to thousands of ideas ) . At this scale , interesting but statistical - ly rare ideas may be less likely to be noticed and built upon , leading the crowd to focus on common ideas . From a cognitive perspective , effective collaborative ideation depends on being able to attend to and deeply process other ideas [ 35 , 55 ] . But people are limited in their capacity to process information [ 16 , 47 ] , and the number of ideas produced at crowd scale certainly exceeds this capacity . When given too many ideas as potential inspiration , people may stop attending to them , or only build on them in superficial ways [ 35 , 55 ] . Ideally , we would like crowd ideators to focus on the primary task of generating ideas , rather than expending most of their effort making sense of a large volume of ideas in order to extract useful inspiration . Another issue is that the presentation of the ideas in their \u201craw form\u201d , which includes many superficial details , can constrain ideators\u2019 search to ideas closely related to the raw idea [ 34 , 44 , 56 ] . Therefore , we hypothesize that simply exposing all ideas to all contributors , while simple to implement , is not likely to be an effective solution . Higher - level abstracted solution paths that distill essential solution approaches shared by a number of different raw ideas may be a good alternative to raw ideas . A classic example of a solution path in the cognitive science litera - ture is the convergence solution schema ( successfully attack a single target by converging from multiple points if a single , focused attack is not feasible ) , abstracted from a variety of situations ( e . g . , generals attacking a fortress via mined bridges radiating from the fortress , a doctor destroy - ing a tumor with radiation rays without also destroying healthy tissue ) [ 25 ] . Abstractions such as categories can be a powerful way to compress large volumes of information into more manageable chunks [ 12 , 46 ] , greatly reducing the number of \u201cbits\u201d that ideators need to process . This should increase the probability that ideators will actually be able to attend to and benefit from them . Also , by representing Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2718 primarily the \u201cessential\u201d information that helps to organize ideas , such as in schemas [ 25 , 64 ] , abstractions could help reduce fixation on superficial details . Some studies have shown that providing one or a few manually generated and selected abstractions can lead to better creative performance compared to showing ideas in their raw form [ 45 , 60 , 64 ] . However , this manual selection may not always be feasible at crowd scale . In this paper , we extend this prior work by examining the value of viewing many ( ~ 15 - 20 ) abstractions rather than just one or two . Automated Sensemaking Strategies Numerous automated sensemaking methods exist , ranging from relatively mature ( e . g . , well - studied , shown to be robust across a range of settings ) and simple ( i . e . , produces reasonable results without requiring significant tuning of many complex model parameters ) vector - space models like Term - Frequency Inverse - Document Frequency ( TF - IDF ) and LSI [ 18 ] models and clustering algorithms like K - means and agglomerative hierarchical clustering , to more sophisticated methods like probabilistic topic models [ 6 ] . Research on these methods has largely focused on improv - ing correspondence with \u201cgold standard\u201d human models of the items being structured ( e . g . , gold standard human - generated categories [ 9 ] ) , or matching human performance on various benchmark tasks ( e . g . , simple A : B , C : D word analogies [ 48 ] ) . Some work has examined the value of automated sensemaking outputs for complex tasks like ideation [ 20 , 24 ] , intelligence analysis [ 57 ] , and conducting scientific literature reviews [ 13 ] . However , relatively little work has compared the value of automated sensemaking outputs to human - produced outputs for complex tasks in general . One notable exception is a study by Andr\u00e9 et al [ 3 ] , who found that simple TF - IDF sensemaking over academic papers could provide suggestions for papers to attendees at an academic conference that were rated at comparable levels of relevance as suggestions provided from a human - generated sensemaking model ( via partial clustering ) . We are not aware of any related studies in the context of idea - tion . This paper explores the cost - quality tradeoffs involved in selecting automated vs . manual sensemaking approaches . SOLUTION PATH EXTRACTION Our goal is to explore the cost - quality curve for current \u201coff - the - shelf\u201d methods that would be readily available to crowd innovation platforms . Here we describe in more detail each of the three points on the cost - quality curve that we select for comparison : 1 ) manual human - human ( highest cost ) , 2 ) machine - human ( medium cost ) , and 3 ) machine - machine ( lowest cost ) . For comparability , all approaches follow the same general process : raw ideas form initial inputs into sensemaking , which includes a cluster phase \u2014 where similar ideas are grouped together \u2014 followed by a label phase to produce \" solution paths \" \u2013 descriptions of key features shared by multiple ideas in a cluster ( often with fewer concrete details than raw ideas ) . Each of the three approaches is defined by whether humans or machines complete each phase , with the intuition that least human involvement ( machines complete both phases ) would be least costly , while maximal human involvement ( in both phases ) would be most costly . Datasets We extracted solution paths for ideas for two problems : 1 ) new product ideas for a novel \u201cfabric display\u201d ( fabric display problem ) , and 2 ) ideas for improving workers\u2019 experience performing HITs on mTurk on mobile devices ( improve mTurk problem ) . Both idea datasets were assem - bled from idea datasets collected in prior studies ( [ 54 ] for the Fabric Display problem , and [ 40 ] for the Improve Turk problem ) . The ideas in those datasets were collected from mTurk workers , and the authors shared the data with us . We randomly sampled 120 ideas for each problem . We chose 120 ideas because it is small enough to apply best practice manual approaches ( e . g . , affinity diagramming ) that would be difficult for 1000s of ideas , yet large enough to motivate the need for sensemaking and to yield intuitions for scala - bility comparisons . Examples of solution paths extracted by each method are shown in Table 1 . Sensemaking Methods Machine - Machine Solution Path Extraction For this method of extracting solution paths , machines complete both the cluster and label phases of solution extraction . Our goal is to design an automated workflow that closely approximates realistic crowd ideation scenarios . In realistic scenarios , one very rarely has \u201cgold standard\u201d data available to assist with evaluation and tuning of complex model parameters . Therefore , we would like to choose relatively mature models with well - known proper - Method Fabric Display Improve Turk Machine - Machine advertisements billboards create shirts / pants cuffs companies time timer site assigned loading notification Machine - Human use for new locations for advertising send timer alerts Human - Human advanced display of advertisements provide notifications for time - sensitive activity Table 1 . Example solution paths extract by each method for both problems . For comparability , we select examples that are close in semantic meaning . Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2719 ties that can work relatively well \u201cout - of - the - box\u201d without significant parametric tuning . To meet these constraints , we chose to do unsupervised feature identification using a combination of TF - IDF ( widely used in information re - trieval ; no parameter settings required ) and LSI ( widely used in information retrieval ; only choose number of dimensions ) , and clustering using the standard K - means algorithm ( widely applicable ; only choose number of clusters ) . In our workflow , we first identify semantic features in the set of ideas using TF - IDF and LSI , and then use these features to achieve unsupervised clustering of the ideas with K - means . We then automatically obtain labels for the clusters by finding the most informative keywords within each cluster . In the cluster phase , ideas are first tokenized , removing uninformative ( stop ) words , and then weighted using TF - IDF . Each idea then is re - represented as a vector of w TF - IDF weights ( where w is the number words in the corpus of ideas ) . We then use LSI to reduce the dimensionality of the vectors . Based on prior experience clustering brainstorming ideas , our intuition is that somewhere between 15 to 25 patterns is often sufficient to adequately describe emerging solution patterns . Therefore , we set our LSI model to retain 15 , 20 and 25 dimensions . For each of these parameter settings , we then identify solution paths by finding the same number of clusters of ideas ( i . e . , 15 , 20 or 25 ) with the K - means clustering algorithm . We use the LSI dimension weights as features for the ideas . Rather than report the running time ( which would vary by machine , dataset , and implementation of the algorithms ) , we leave it to the reader to examine the well - studied computational complexity properties of LSI and K - means and extrapolate accordingly to their desired settings . As a ballpark estimate , all our clustering runs for the 120 ideas took less than 30 seconds . For the label phase , we again select a method that would not require significant parameter tuning . To do this , we treat each cluster as a document , and compute TF - IDF weights for all words . We then choose the top n words with the highest TF - IDF weights within each cluster . N is deter - mined by the average length of human labels ( minus stopwords ) . The intuition behind this method is to identify words that distinguish between the clusters . Machine - Human Solution Path Extraction For this method , the machines complete the cluster phase , but humans are recruited for the label phase . The cluster phase is the same as the machine - machine method ( i . e . , clusters generated by combination of LSI and K - means ) . In the label phase , the labelers received each of the M clusters of ideas identified in the machine - clustering phase . The labeling task was completed in a Google Document , which contained 1 ) instructions for labeling , 2 ) the problem description , and 2 ) each of the M clusters as unlabeled lists with blank headers . Labelers described clusters by typing a description in the blank headers ( see Fig . 1 , left panel ) . Labelers received the following instructions : \u201cHere is a set of 120 ideas for the problem , divided into clusters of ideas . Each cluster of ideas is meant to represent a solution pattern for the problem . Recall that we would like to provide these patterns as inspiration for future brainstorm - ers who will also work on this problem . Your goal is to write a meaningful description for each cluster . Good descriptions capture the essential shared theme of a group of ideas in an appropriately specific fashion ( i . e . , contain enough detail to be useful to future brainstormers , but are not simple restatements of each idea ) . Ideally , they capture a theme shared by all ideas in their cluster ; in some ( rare ) cases , you may find it difficult to discern the common theme : in this case , please do your best to describe a pattern shared by at least 2 - 3 ideas in the cluster . As a good test , good descriptions can fit well into the template \u2018How can we _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ to solve the problem ? \u2019 . \u201d A set of three research assistants \u2014advanced HCI design students \u2014 served as labelers . Labelers on average took approximately 28 minutes to complete labeling for each problem . This approach represents a midpoint on the cost - quality tradeoff curve , and also allows us to potentially tease apart the value provided by clustering ( which influences the semantics of the labels produced ) and labeling ( which vary in such features as coherence , specificity , and phrasing ) provided by machines and humans . Human - Human Solution Path Extraction For this method , humans performed both cluster and label phases in a single task . The clusterer - labelers received all 120 ideas in a single stack . Each idea was printed out on a slip of paper . Clusterer - labelers then sorted ideas into clusters on a 45\u201d by 45\u201d table , and labeled clusters by writing descriptions on Post - It notes . Figure 1 ( right panel ) shows the physical setup of the task . Clusterer - labelers received a description of the problem as well as the following instructions : \u201cHere is a set of 120 ideas for the problem . Please identify solution patterns in the set of ideas . We would like to provide these patterns as inspiration for future brainstormers who will also work on Figure 1 . Screenshot of human labeling task ( left ) physical setup of human cluster - label task ( right ) Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2720 this problem . Do this by grouping ideas on the whiteboard into clusters that define patterns and writing descriptions of those patterns . Good descriptions capture the essential shared theme of a group of ideas in an appropriately specific fashion ( i . e . , contain enough detail to be useful to future brainstormers , but are not simple restatements of each idea ) . As a good test , good descriptions can fit well into the template \u2018How can we _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ to solve the problem ? \u2019 You may identify clusters as large or as small as you like ( even singleton clusters ) . Please do not create miscellaneous clusters or sort ideas by quality . \u201d A different set of three research assistants \u2014 advanced HCI design students \u2014 performed this task . On average , cluster - ing and labeling a single problem took approximately 52 minutes , significantly longer than just labeling . An average of 22 solution paths were identified for the Fabric Display problem , and 16 for the Improve Turk problem . Note that this task could be performed using digital tools ; however , we chose to implement these steps on paper to have as close a match as possible to existing \u201coff - the - shelf\u201d human sensemaking approaches ( e . g . , affinity diagramming ) . IDEATION EXPERIMENT Overview With extracted solution paths , we then conducted an online ideation experiment to compare the relative ideation value of the different sensemaking approaches , compared to simple exposure to all prior ideas , or no stimulation . Method Participants We recruited 139 mTurk workers ( 42 % female , mean age = 32 . 6 years , SD = 9 . 6 ) for this study . To ensure quality data , all participants had to have approval rates of at least 95 % with at least 100 completed HITs . Participants were paid $ 2 . 00 for about 20 minutes of participation ( on average ) , yielding an average hourly rate of $ 6 / hour . Study Design Participants were randomly assigned to one of 5 conditions : 1 ) unconstrained ( control , N = 21 ) , 2 ) viewing all 120 prior ideas for the problem ( all - ideas , N = 24 ) , viewing 3 ) ma - chine - labeled solution paths generated for machine - generated clusters ( machine - machine : least costly , N = 32 ) , 4 ) human - labeled paths generated for machine - generated clusters ( machine - human : somewhat costly , N = 34 ) , or 5 ) with human - labeled paths generated for human - generated clusterings of ideas ( human - human : most costly , N = 28 ) . Within the path conditions , participants were randomly assigned one of the 3 path sets generated using that solution path extraction method . Brainstorming Task Participants generated ideas for both the Fabric Display and Improve Turk problems . Brainstorming Interface Participants generated ideas using a simple ideation inter - face . Inspirations ( whether ideas or solution paths ) were provided to participants in an \u201cinspiration feed\u201d in the right panel of their interface ( see Figure 2 ) . Participants could \u201cbookmark\u201d particular inspirations that they found helpful . No sorting and filtering options were provided . The limited sorting / filtering available in the all - ideas condition might be considered primitive , but it is actually similar to many existing platforms that might provide rudimentary filtering by very broad , pre - defined categories ( that may be reused across problem ) . Participants in the control condition generated ideas with a simpler ideation interface that removed the inspiration feed . Procedure After providing informed consent , participants experienced a brief tutorial to familiarize themselves with the interface . Embedded within the tutorial was an alternative uses task ( where participants were asked to think of as many alterna - tive uses as possible for a bowling pin ) . Participants then Figure 2 . Screenshot of ideation interface . Actual inspirations from the machine - machine condition are shown . Participants enter ideas on the left , and view inspirations on the right . Participants can star inspirations they find useful . Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2721 generated ideas for both the Fabric Display and Improve Turk problems . Participants were given 8 minutes to work on each problem , and the order of the problems was ran - domized across participants . Participants in the inspiration conditions received the following instructions regarding the inspirations : \u201cBelow are some inspirations to boost your creativity . Feel free to create variations on them , elaborate on them , recombine them into new ideas , or simply use them to stimulate your thinking . If you find an inspiration to be helpful for your thinking , please let us know by clicking on the star button ! This will help us provide better inspira - tions to future brainstormers . \u201d After completing both problems , participants then completed a brief survey with questions about demographics and the participants\u2019 experi - ences during the task . Measures Conceptually , we distinguish process ( fluency and breadth of search ) from outcome ( novelty , quality ) measures for exploring the impact of sensemaking on ideation . This distinction follows previous design ideation research [ 52 ] . Theoretically , our process measures also map to Guilford\u2019s [ 28 ] notion of divergent thinking as a process basis for creativity , while our outcome measures map to standard theoretical notions of creative products as a combination of novelty and usefulness / practicality [ 1 , 31 , 50 , 51 ] . Fluency : Number of Ideas Fluency was operationalized as the number of ideas gener - ated by a participant for a given problem . Breadth of Search in Solution Space We used LSI to characterize the nature of participants\u2019 search through the solution space . While LSI can be more accurate with longer texts ( our raw ideas are 8 - 10 words , on average ) , researchers have successfully used LSI to model creativity with similar - length texts [ 23 , 30 ] . To maximize the accuracy of the model , for each problem we enriched the model with the full set of ideas previously collected from mTurk workers for each of the two problems ( 1 , 354 for Fabric Display , and 2 , 287 for Improve Turk ) . Thus , the total sizes of the training corpora for the LSI models were 2 , 354 ideas for Fabric Display ( prior ideas plus 1 , 000 ideas from our experiment ) and 3 , 403 for Improve Turk ( prior ideas plus 1 , 116 ideas from our experiment ) . We used the same procedure to build the LSI space as with the first step of machine clustering ( i . e . , weight words with TF - IDF before estimating LSI ) , except that we retained 200 dimensions ( in keeping with prior rules of thumb for larger datasets [ 42 ] ) . Breadth was operationalized as the mean pairwise distance between a given participant\u2019s ideas . Higher mean pairwise distance indicates that participants\u2019 ideas are sampled from very diverse regions of the solution space . Distances were calculated by subtracting pairwise cosines from 1 , yielding distance scores between 0 ( semantically identical ) and 1 ( semantically very different ) . Novelty and Quality of Ideas To explore impact on novelty and quality , we obtained ratings of novelty and quality for ideas for the Improve Turk problem only , due to concerns about mTurk workers\u2019 ability to provide valid ratings for the Fabric Display problem . MTurk workers are suitable judges for the Im - prove Turk problem , since they have first - hand expertise in the domain of worker experience on mTurk , and are also potential users of mTurk mobile products . 414 workers from mTurk rated the ideas for novelty and quality . Novelty was operationalized as the degree to which an idea was novel , ranging from a scale of 1 ( Extremely Obvious ) to 7 ( Extremely Novel ) . Quality was operational - ized as the degree to which an idea would be useful for solving the problem , assuming it was implemented ( i . e . , separating out concerns over feasibility ) , ranging from a scale of 1 ( Not Useful at All ) to 7 ( Extremely Useful ) . Each worker rated a random sample of approximately 20 ideas . While the raw inter - rater agreement was relatively low ( Krippendorff\u2019s alpha = . 23 and . 25 , respectively ) , the overall aggregate measure had acceptable correspondence with each judges\u2019 intuitions . Computing correlations between each judges\u2019 ratings and the overall aggregate score yielded average correlations of . 52 for novelty and . 58 for quality . To deal with potential differences in usage of High Quality Low Quality High novelty ( 1 . 04 , 1 . 51 ) Allow demographic info to be filled out automatically so it doesn ' t have to be done each time ( 1 . 52 , 1 . 01 ) Virtual keyboard that allows you to have a row of buttons for only the keys that are used in the task , instead of finding them on the normal keyboard ( 1 . 22 , - 1 . 09 ) Fingerprint Captchas are possible . ( 1 . 10 , - 1 . 74 ) music to focus worker Low novelty ( - 0 . 87 , 1 . 04 ) Grey out HITS that have been done before and cannot be repeated ( - 1 . 00 , 1 . 16 ) make it easy to find good hits ( - 1 . 11 , - 1 . 38 ) Be able to complete more hits ( - 1 . 58 , - 1 . 83 ) Have direct contact information such as a phone number clearly listed for MTurk workers . Table 2 . Example ideas at each combination of low and high novelty and quality for the Improve Turk problem . Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2722 the rating scale across raters ( e . g . , some might only use the upper end of the scale ) , we normalized scores within raters ( i . e . , difference between rating and mean rating provide by rater , divided by the standard deviation of the raters\u2019 ratings ) . Table 2 shows examples of low and high novelty and quality ideas in the data . Inspiration Use : Number of Bookmarked Inspirations We also measured the degree to which participants found inspirations to be useful by counting the number of inspira - tions that were bookmarked by each participant . Control Measures Our primary control measure is participants\u2019 performance for the baseline fluency task ( i . e . , number of bowling pin alternative uses generated ) . This measure captures aspects of participants\u2019 base level of creative fluency ( as a proxy for individual creativity [ 27 ] ) , as well as aspects of motiva - tion and conscientiousness ; all of these factors are expected to influence creative performance , and are therefore ac - counted for in our analyses by including baseline fluency as a covariate predictor in our statistical models . Results Participants generated a total of 2 , 116 ideas across the two problems , across conditions ( 1 , 000 ideas for Fabric Display and 1 , 116 for Improve Turk ) . Machine - Generated Paths Stimulate More Ideas Table 3 shows the means and standard errors for each of the conditions . We first estimated separate ANCOVAs with condition as a main effect and baseline fluency as a control covariate for each of the two problems . For Fabric Display , there was no statistically significant overall effect of condi - tion , F ( 4 , 133 ) = 1 . 11 , p = 0 . 35 . However , planned contrasts suggest that both the all - ideas ( model beta = 1 . 9 , p = 0 . 08 , Cohen\u2019s d = 0 . 49 ) and machine - machine conditions ( be - ta = 1 . 8 , p = 0 . 08 , d = 0 . 42 ) trend towards more ideas than control ( see Table 3 , left column ) . That is , adjusting for baseline fluency , participants who saw all prior ideas or machine - generated solution paths generated about 2 more ideas than participants who received no stimulation . In terms of standardized effect sizes , the Cohen\u2019s d of about 0 . 5 corresponds to a medium - sized effect according to Cohen\u2019s [ 15 ] classification of effect sizes in the behavioral sciences . For Improve Turk , there was no significant overall effect of condition , F ( 4 , 133 ) = 1 . 40 , p = 0 . 23 . However , the machine - machine condition did display a significant trend towards more ideas than control , beta = 2 . 5 , p = 0 . 04 , d = 0 . 41 ( see Table 3 , right column ) . Similar to the Fabric Display problem , adjusting for baseline fluency , participants who saw machine - generated paths generated about 2 more ideas than participants who received no stimulation . Machine - Generated Paths Increase Breadth of Search Table 4 shows the means and standard errors for each of the conditions . Baseline fluency was not significantly correlat - ed with breadth of search in either problem ( r = . 06 , p = . 47 for Fabric Display , and r = . 03 , p = . 76 for Improve Turk ) . Therefore , we estimated separate ANOVAs with condition as the only factor for each of the two problems . For the fabric display problem , there was a marginally significant overall effect of condition , F ( 4 , 133 ) = 2 . 12 , p = 0 . 08 . Planned contrasts suggest that both the machine - machine ( be - ta = 0 . 06 , p = . 01 , d = 0 . 65 ) and machine - human paths ( be - ta = 0 . 02 , p = . 02 , d = 0 . 57 ) lead to significantly more breadth of search than control , while human - human paths are marginally significantly better than control ( beta = 0 . 05 , p = . 05 , d = 0 . 52 ) . All effect sizes are medium - sized , by Cohen\u2019s classification . For the Improve Turk problem , the overall main effect of condition is not statistically signifi - cant , F ( 4 , 139 ) = 1 . 53 , p = 0 . 20 . However , planned contrasts suggest that both all - ideas ( beta = 0 . 07 , p = . 04 , d = 0 . 72 ) and machine - machine conditions ( beta = 0 . 06 , p = . 04 , d = 0 . 57 ) are better than control . The all - ideas effect is close to \u201clarge\u201d , by Cohen\u2019s classification ( the suggested cut - off is 0 . 8 ) , while the machine - machine effect is medium - sized . Breadth of Search Fabric display Improve Turk Condition M ( SE ) M ( SE ) Control 0 . 89 ( 0 . 02 ) 0 . 81 ( 0 . 03 ) All - Ideas 0 . 93 ( 0 . 02 ) 0 . 89 ( 0 . 03 ) * Machine - Machine 0 . 95 ( 0 . 02 ) * * 0 . 89 ( 0 . 02 ) * Machine - Human 0 . 95 ( 0 . 02 ) * 0 . 86 ( 0 . 02 ) Human - Human 0 . 94 ( 0 . 02 ) m 0 . 86 ( 0 . 03 ) * p < . 05 vs . control , m p < . 10 vs control Table 4 . External stimulation increases breadth of search by relative to control condition , for both problems . Number of ideas Fabric display Improve Turk Condition M ( SE ) M ( SE ) Control 6 . 46 ( 0 . 79 ) 7 . 13 ( 0 . 93 ) All - Ideas 8 . 40 ( 0 . 77 ) m 8 . 32 ( 0 . 91 ) Machine - Machine 8 . 29 ( 0 . 67 ) m 9 . 59 ( 0 . 78 ) * Machine - Human 7 . 56 ( 0 . 63 ) 7 . 43 ( 0 . 75 ) Human - Human 7 . 30 ( 0 . 68 ) 8 . 13 ( 0 . 81 ) * p < . 05 vs . control , m p < . 10 vs control Table 3 . Machine - generated paths increase number of ideas relative to control condition . Means are adjusted for baseline fluency . Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2723 Solution Paths Do not Impact Novelty of Ideas Table 5 shows the means and standard errors for each of the conditions . Baseline fluency was not significantly correlat - ed with mean novelty . Therefore , we estimated an ANOVA with condition as main effect for mean novelty of ideas . There was no main effect of condition on mean novelty of ideas , F ( 4 , 131 ) = 0 . 79 , p = 0 . 53 . Machine - Generated Paths Reduce Quality of Ideas Table 5 shows the means and standard errors for each of the conditions . Baseline fluency was not significantly correlat - ed with mean quality . Therefore , we estimated an ANOVA with condition as main effect for mean quality of ideas . There was a marginally significant main effect of condition on mean quality of ideas , F ( 4 , 129 ) = 2 . 15 , p = . 08 . Planned contrasts suggested that the mean quality of ideas in the machine - machine ( M = \u2013 0 . 08 , SE = 0 . 06 ) and machine - human conditions ( M = \u2013 0 . 02 , SE = 0 . 05 ) were of signifi - cantly lower quality than in the control condition ( M = 0 . 17 , SE = . 07 ) , beta = - . 25 , p = . 01 , d = 0 . 69 , and beta = \u2013 . 18 , p = . 04 , d = 0 . 54 . Ideas in the human - human condition were of marginally lower quality ( M = 0 . 01 , SE = 0 . 06 ) than in the control condition ( beta = - . 16 , p = . 09 , d = 0 . 47 ) . All effects are approximately medium - sized , by Cohen\u2019s classification . Equal Number of Good Ideas Across Conditions Given the reduction in quality associated with machine labels , we wondered about the impact of those paths on aggregate creativity ( a combination of both novelty and quality ) . Ideation platforms might accept a slight drop in mean quality as long as overall creativity does not suffer . We follow Reinig et al [ 49 ] to measure creative output as the number of good ideas , where a good idea is defined as an idea with novelty and quality scores that are both above the mean ( see Table 2 , upper right quadrant , for examples ) . An ANCOVA controlling for baseline fluency showed no main effect of condition , F ( 4 , 128 ) = 0 . 98 , p = . 42 ; further , the trends approximately track the statistical patterns for total number of ideas , suggesting that participants across condi - tions are generating good ideas at a fairly constant rate . Indeed , when examining the proportion of good ideas as a dependent measure , an ANCOVA again shows no signifi - cant differences across conditions , F ( 4 , 128 ) = . 61 , p = . 66 . DISCUSSION In this research , we examine the relative value of different approaches to inspiring crowd ideators with prior ideas . We empirically test whether abstracted solution paths ( generat - ed by a range of sensemaking methods ) can enable ideators to better benefit from prior ideas ( as measured by impact on fluency , breadth , and novelty and quality of ideas ) . We also empirically explore how the ideation value of solution paths varies with the cost of the sensemaking strategy that pro - duced them . Our study yields two main sets of findings , which we discuss in turn . No Consistent Benefit of Solution Paths The first main set of findings is that solution paths do not consistently improve ideation more than simply showing all raw ideas to ideators . With respect to fluency , for the Improve Turk problem , machine - machine paths ( but not machine - human , human - human , or all - ideas ) improved fluency over control ; however , for the Fabric Display problem , all - ideas and machine - machine paths both improved fluency . With respect to breadth of search , solution paths ( regardless of source ) improved breadth of search over control for the Fabric Display problem ; howev - er , for the Improve Turk problem , only all - ideas and machine - machine paths improved breadth over control . None of the solution paths conditions improved novelty over control , and some ( human - machine and machine - machine ) even reduced quality relative to control , while all - ideas did not . There is a range of possible explanations for why solution paths did not improve ideation in our experiment . First , perhaps the solution paths reduced quality because they were potentially derived from both bad and good ideas ( recall that we randomly sampled the initial set of 120 ideas from the prior datasets ) . However , it seems unlikely to have been the main driving factor . While one participant who saw all ideas did complain in the open - ended portion of our post - task surveys that there were many bad ideas , nobody voiced this complaint in the paths conditions . Further , one Mean Novelty Mean Quality Condition M ( SE ) M ( SE ) Control 0 . 04 ( 0 . 07 ) 0 . 17 ( 0 . 07 ) All - Ideas 0 . 02 ( 0 . 07 ) 0 . 07 ( 0 . 07 ) Machine - Machine 0 . 06 ( 0 . 06 ) - 0 . 09 ( 0 . 06 ) * * Machine - Human 0 . 03 ( 0 . 06 ) - 0 . 02 ( 0 . 06 ) * Human - Human - 0 . 09 ( 0 . 07 ) 0 . 01 ( 0 . 07 ) m * p < . 05 vs . control , m p < . 10 vs control Table 5 . External stimulation does not impact novelty of ideas , but machine - generated paths reduces quality relative to control condition . Figure 3 . Participants generate similar numbers of good ideas across conditions , adjusted for baseline fluency . Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2724 participant ( in the all - ideas condition ) specifically noted that , \u201ca lot of people had some good ideas about different features for a Mturk app . \u201d Indeed , ideas that were more similar to the initial 120 raw ideas were considered to be of slightly higher quality : we found a small but significantly positive correlation between LSI similarity of ideas vs . initial 120 raw ideas and quality ratings , r = . 07 , p = . 01 . Nevertheless , it is possible that the presence of even one or two bad ideas might be sufficient to contribute to fixation . Second , it could be that that there were too many inspira - tions to sift through . However , 15 - 25 inspirations seems like a manageable enough number of items to scan through relatively quickly and find interesting ideas to build on . We also specifically designed the system to be able to fit all of the inspirations on one screen in the paths conditions ( so that there would not need to be much scrolling ) . Indeed , a number of participants in the paths conditions appreciated that they could just glance over to see inspirations , while many participants in the all - ideas condition complained that there were too many inspirations to consider at once . A more interesting alternative explanation might be that , rather than providing bad solutions , the stimuli were steer - ing participants away from good solutions , due to a desire to not repeat ideas ( despite our instructions saying it was ok to build on or recombine inspirations ) . Indeed , some partic - ipants said in the survey that they were influenced by the inspirations in this way . For example , one participant ( in the machine - human condition ) said , \u201cI could just read them and see if an idea I had had already been provided and if it was I would move on to the next idea . \u201d One partic - ipant ( in the same condition ) even said , \u201cI would come up with an idea and then notice it was already listed on the inspiration list , so I felt discouraged about that . \u201d Another participant ( in the machine - machine condition ) said that it was challenging \u201ctrying not to repeat the inspirations . The first inspiration was about a movie theater , which coinci - dentally , was also similar to my first idea . \u201d If this happened often enough , participants might have searched more broadly , but away from any good ideas that happened to be represented in the solution paths . In fact , perhaps the relative ease of gaining a broad sense of the solution space with solution paths\u2014rather than with the 120 raw ideas\u2014 could explain why quality was reduced only in the paths conditions . Relatedly , participants might have felt over - whelmed by the prospect of adding something new to the existing solution space . One participant in the machine - human condition said , \u201cThere were so many inspirations for the second brainstorming task that it was hard to come up with anything new to add . \u201d Overall , these findings suggest that it may not be straight - forward to translate the benefits of abstracted solution paths from individual and group settings into a crowd setting . For example , if providing tens of abstracted solution paths at once as potential inspirations could be overwhelming , alternative delivery mechanisms , such as on - demand delivery of individual inspirations [ 10 , 54 ] , or focusing ideators on a single theme [ 22 , 45 ] . Automated Sensemaking Can Improve Ideation The second main set of findings concerns the value of machine - generated paths . In our data , the only form of stimulation that provided consistent benefits was machine - generated solution paths , which improved fluency and breadth across both problems . However , this increase in fluency and breadth came at the cost of reduced quality of ideas . We did also find , however , that the reduction in quality did not hamper the production of good ideas ( i . e . , ideas that are both novel and of high quality ) . Thus , our findings suggest that machine - generated paths can improve divergent thinking ( generating many possible ideas ) , but not necessarily creativity ( generating high - quality , novel ideas ) . One explanation for why these solution paths ( though hypothesized to be of low intrinsic quality ) , might have benefited ideation is that participants were essentially leveraging them as keyword clouds . For example , one participant in the Machine - Machine condition said , \u201cThe inspirations seemed nonsensical when read as a whole , but one or two words would catch my attention and spark an idea . For example , I think one of the ones for the fabric said \" bathroom \" or \" towel \" or something like that . That gave me an image of a shower curtain and having video play on the shower curtain . The inspirations were jumping off points . \u201d Another participant said , \u201cI would list as many ideas as possible and then browse quickly through key words and run with whatever came to mind . Often I would combine words and that would help me think of an original idea . \u201d In theory , the keywords generated using our ma - chine - machine workflow should identify words that can best discriminate between major patterns of solutions in the set of ideas . It would be valuable to test whether these paths benefit ideators more than simpler ways of generating keyword clouds ( e . g . , frequency - based ) . This set of findings suggests that the relationship between cost and quality ( in terms of value for ideation ) might not be simply linear , monotonic , or even positive . From a practical standpoint , large - scale ideation platforms could gain value from employing very simple automated sense - making methods ( such as the workflow we used in this study ) . These methods might provide the most value in the early stages of the innovation process , where the focus is on quickly exploring as much of the space as possible before focusing in on more promising solution approaches , and some reduction in quality might be acceptable in exchange . Our results suggest that , accounting for our fluency and breadth effects , platforms that use machine paths may need fewer people to achieve comparable coverage of the solu - tion space ( vs . not providing any exposure to prior ideas ) . FUTURE WORK Our finding that solution paths did not consistently benefit ideators ( over providing all ideas or no ideas ) runs counter to theory , and suggests caution when using solution paths in Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2725 a crowd setting . It will be important for future research to test the potential explanations we explored for the lack of ( and even slightly negative ) effect of solution paths , partic - ularly since many of these factors are likely to be present in collective innovation settings ( e . g . , large number of poten - tial inspirations , presence of some bad ideas , pressure to come up with original ideas rather than build on ideas ) . It would also likely be useful to investigate task designs and workflows that could overcome some of those potential challenges . For example , could rephrasing inspirations as questions ( e . g . , \u201cHow can we\u2026 ? \u201d ) mitigate participants\u2019 tendency to move away from solution paths ? Could we also intelligently divide up solution paths to different portions of the crowd such that they can explore a smaller set of solution path in depth and would this help mitigate potential cognitive load issues from seeing many inspirations ? Would it be useful to integrate evaluation into the sense - making process to more quickly weed out bad ideas ? Future efforts to enable more collaborative large - scale ideation could benefit from examination of these questions . Also , in this study we used a fully manual approach to human sensemaking . But , methods exist for sensemaking using distributed human computation [ 2 , 3 , 14 , 26 ] . Future work might fruitfully explore how the value of human sensemaking for crowd ideation might not only depend on task design and workflow factors , but whether such distrib - uted workflows could provide similar benefits . Finally , future work can explore the generality of these findings . Our findings were similar across the two problems we examined ; thus , our results seem robust to domain expertise effects . Also , because our participants were paid mTurkers , our results might map on to other reward - motivated situations ( e . g . , contests ) . However , we are uncertain if they generalize to volunteer communities . Participant motivation may be an important moderating effect in these settings , particularly since the different sensemaking approaches impose varying degrees of cogni - tive load on ideators ( e . g . , less cognitive load for machine - generated paths compared to human - generated paths ) . CONCLUSION In this paper , we examined the ideation value of different approaches to sensemaking over prior ideas , using an online ideation experiment that compares ideation under no stimulation , exposure to all ideas , or abstracted solution paths from fully automated , machine - human hybrid , or fully manual sensemaking approaches . Our results suggest that simple automated sensemaking methods can provide some value ( e . g . , increased fluency , breadth of search ) to large - scale ideation platforms . Our results also motivate further research on how to best enable crowd ideators to benefit from ( human ) sensemaking outputs . ACKNOWLEDGEMENTS We thank all participants , especially Amazon Mechanical Turk workers who provided feedback on our experimental protocol , our volunteer clusterers and labelers , and Michelle Tai for help with data analysis . This research was supported by NSF grants # 1208382 and # 1122206 and a Google Faculty Award . REFERENCES 1 . Teresa M . Amabile . 1983 . The social psychology of creativity : A componential conceptualization . Journal of personality and social psychology 45 , 2 : 357 . 2 . Paul Andr\u00e9 , Aniket Kittur , and Steven P . Dow . 2014 . Crowd Synthesis : Extracting Categories and Clusters from Complex Data . Proceedings of the 17th ACM Conference on Computer Supported Cooperative Work & Social Computing , ACM , 989 \u2013 998 . http : / / doi . org / 10 . 1145 / 2531602 . 2531653 3 . Paul Andr\u00e9 , Haoqi Zhang , Juho Kim , Lydia Chilton , Steven P . Dow , and Robert C . Miller . 2013 . Communi - ty clustering : Leveraging an academic crowd to form coherent conference sessions . First AAAI Conference on Human Computation and Crowdsourcing . 4 . Brian P . Bailey and Eric Horvitz . 2010 . What\u2019s Your Idea ? : A Case Study of a Grassroots Innovation Pipe - line Within a Large Software Company . Proceedings of the SIGCHI Conference on Human Factors in Com - puting Systems , ACM , 2065 \u2013 2074 . http : / / doi . org / 10 . 1145 / 1753326 . 1753641 5 . Hugh Beyer and Karen Holtzblatt . 1997 . Contextual design : defining customer - centered systems . Elsevier . 6 . David M . Blei . 2012 . Probabilistic topic models . Com - munications of the ACM 55 , 4 : 77 \u2013 84 . http : / / doi . org / 10 . 1145 / 2133806 . 2133826 7 . Kevin J . Boudreau and Karim R . Lakhani . 2013 . Using the crowd as an innovation partner . Harvard Business Review 91 , 4 : 60 \u2013 69 . 8 . Kevin J . Boudreau and Karim R . Lakhani . 2015 . Open disclosure of innovations , incentives and follow - on re - use : Theory on processes of cumulative innovation and a field experiment in computational biology . Research Policy 44 , 1 : 4 \u2013 19 . http : / / doi . org / 10 . 1016 / j . respol . 2014 . 08 . 001 9 . Jonathan Chang , Sean Gerrish , Chong Wang , Jordan L . Boyd - graber , and David M . Blei . 2009 . Reading tea leaves : How humans interpret topic models . Advances in neural information processing systems , 288 \u2013 296 . 10 . Joel Chan , Steven C . Dang , and Steven P . Dow . 2016 . Improving crowd innovation with expert facilitation . Proceedings of the ACM Conference on Computer - Supported Cooperative Work & Social Computing . http : / / doi . org / 10 . 1145 / 2818048 . 2820023 11 . Joel Chan and Christian Schunn . 2015 . The impact of analogies on creative concept generation : Lessons from an in vivo study in engineering design . Cognitive Science 39 , 1 : 126 \u2013 155 . http : / / doi . org / 10 . 1111 / cogs . 12127 12 . William G . Chase and Herbert A . Simon . 1973 . The mind\u2019s eye in chess . In Visual Information Processing , William G . Chase ( ed . ) . New York , NY , 215 \u2013 281 . Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2726 13 . Duen Horng Chau , Aniket Kittur , Jason I . Hong , and Christos Faloutsos . 2011 . Apolo : Making Sense of Large Network Data by Combining Rich User Interac - tion and Machine Learning . Proceedings of the SIGCHI Conference on Human Factors in Computing Systems , ACM , 167 \u2013 176 . http : / / doi . org / 10 . 1145 / 1978942 . 1978967 14 . Lydia B . Chilton , Greg Little , Darren Edge , Daniel S . Weld , and James A . Landay . 2013 . Cascade : Crowdsourcing taxonomy creation . Proceedings of the SIGCHI Conference on Human Factors in Computing Systems , ACM , 1999 \u2013 2008 . http : / / doi . org / 10 . 1145 / 2470654 . 2466265 15 . Jacob Cohen . 1988 . Statistical power analysis for the behavioral sciences . Psychology Press , Hillsdale , N . J . 16 . Nelson Cowan . 2000 . The magical number 4 in short - term memory : A reconsideration of mental storage ca - pacity . Behavioral and Brain Sciences 24 : 87 \u2013 185 . 17 . Darren W . Dahl and Page Moreau . 2002 . The Influence and value of analogical thinking during new product ideation . Journal of Marketing Research 39 , 1 : 47 \u2013 60 . 18 . Scott Deerwester , Susan T . Dumais , Geroge W . Furnas , and Thomas K . Landauer . 1990 . Indexing by Latent Semantic Analysis . Journal of the American Society for Information Science 41 , 6 : 1990 . 19 . Darleen M . DeRosa , Carter L . Smith , and Donald A . Hantula . 2007 . The medium matters : Mining the long - promised merit of group interaction in creative idea generation tasks in a meta - analysis of the electronic group brainstorming literature . Computers in Human Behavior 23 , 3 : 1549 \u2013 1581 . http : / / doi . org / 10 . 1016 / j . chb . 2005 . 07 . 003 20 . J . R . Duflou and P . Verhaegen . 2011 . Systematic inno - vation through patent based product aspect analysis . CIRP Annals - Manufacturing Technology 60 , 1 : 203 \u2013 206 . 21 . C . Eckert and M . Stacey . 1998 . Fortune Favours Only the Prepared Mind : Why Sources of Inspiration are Es - sential for Continuing Creativity . Creativity and Inno - vation Management 7 , 1 : 1 \u2013 12 . 22 . Antonio Ferreira , Valeria Herskovic , and Pedro An - tunes . 2008 . Attention - Based Management of Infor - mation Flows in Synchronous Electronic Brainstorm - ing . In Groupware : Design , Implementation , and Use , Robert O . Briggs , Gert - Jan - . J . Vreede and Aaron S . Read ( eds . ) . Springer - Verlag , Berlin , Heidelberg , 1 \u2013 16 . Retrieved from http : / / dx . doi . org / 10 . 1007 / 978 - 3 - 540 - 92831 - 7 _ 1 23 . Eve Forster and Kevin N . Dunbar . 2009 . Creativity evaluation through Latent Semantic Analysis . Pro - ceedings of the 31st Annual Conference of the Cogni - tive Science Society . 24 . Katherine Fu , Joel Chan , Jonathan Cagan , Kenneth Kotovsky , Christian Schunn , and Kristin Wood . 2013 . The Meaning of Near and Far : The Impact of Structur - ing Design Databases and the Effect of Distance of Analogy on Design Output . Journal of Mechanical Design 135 , 2 : 021007 . http : / / doi . org / 10 . 1115 / 1 . 4023158 25 . M . L . Gick and K . J . Holyoak . 1983 . Schema induction and analogical transfer . Cognitive Psychology 15 , 1 : 1 \u2013 38 . 26 . Ryan G . Gomes , Peter Welinder , Andreas Krause , and Pietro Perona . 2011 . Crowdclustering . Advances in Neural Information Processing Systems 24 . 27 . Joy P . Guilford . 1950 . Creativity . American Psycholo - gist 5 : 444 \u2013 454 . 28 . Joy P . Guilford . 1956 . The structure of intellect . Psycho - logical Bulletin 53 , 4 : 267 \u2013 293 . http : / / doi . org / 10 . 1037 / h0040755 29 . Raja Gumienny , Steven Dow , Matthias Wenzel , Lutz Gericke , and Christoph Meinel . 2015 . Tagging User Research Data : How to Support the Synthesis of In - formation in Design Teams . In Design Thinking Re - search , Hasso Plattner , Christoph Meinel and Larry Leifer ( eds . ) . Springer International Publishing , 169 \u2013 191 . 30 . J Isaiah Harbison and Henk Haarmann . 2014 . Automat - ed scoring of originality using semantic representa - tions . Proceedings of the 36th Annual Conference of the Cognitive Science Society . 31 . Beth A . Hennessey and Teresa M . Amabile . 2010 . Creativity . Annual Review of Psychology 61 : 569 \u2013 98 . http : / / doi . org / 10 . 1146 / annurev . psych . 093008 . 100416 32 . Scarlett R . Herring , Chia - Chen Chang , Jesse Krantzler , and Brian P . Bailey . 2009 . Getting inspired ! : under - standing how and why examples are used in creative design practice . Proceedings of the 27th international conference on Human factors in computing systems , ACM , 87 \u2013 96 . http : / / doi . org / 10 . 1145 / 1518701 . 1518717 33 . Keith J . Holyoak and Paul Thagard . 1996 . Mental leaps : Analogy in creative thought . Cambridge , MA . 34 . David G . Jansson and Steven M . Smith . 1991 . Design fixation . Design Studies 12 , 1 : 3 \u2013 11 . 35 . Elahe Javadi and Wai - Tat - . T . Fu . 2011 . Idea Visibility , Information Diversity , and Idea Integration in Elec - tronic Brainstorming . Proceedings of the 6th Interna - tional Conference on Foundations of Augmented Cog - nition : Directing the Future of Adaptive Systems , Springer - Verlag , 517 \u2013 524 . 36 . Elahe Javadi , Joseph Mahoney , and Judith Gebauer . 2013 . The impact of user interface design on idea inte - gration in electronic brainstorming : an attention - based view . Journal of the Association for Information Sys - tems 14 , 1 : 1 \u2013 21 . 37 . Mark Klein and Gregorio Convertino . 2015 . A Roadmap for Open Innovation Systems . Journal of Social Media for Organizations 2 , 1 : 1 . 38 . Jon Kolko . 2011 . Exposing the magic of design : A practitioner\u2019s guide to the methods and theory of syn - thesis . Oxford University Press . 39 . Adriana Kovashka and Kristen Grauman . 2014 . Discov - ering Shades of Attribute Meaning with the Crowd . Third International Workshop on Parts and Attributes . Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2727 40 . Filip Krynicki . 2014 . Methods and models for quantita - tive analysis of crowd brainstorming . Master ' s thesis . University of Waterloo , Waterloo , Ontario , Canada . 41 . Chinmay Kulkarni , Steven P . Dow , and Scott R . Klem - mer . 2012 . Early and repeated exposure to examples improves creative work . Proceedings of the 34th An - nual Meeting of the Cognitive Science Society . 42 . Thomas K . Landauer , Peter W . Foltz , and Darrell Laham . 1998 . An introduction to latent semantic anal - ysis . Discourse Processes 25 , 2 : 259 \u2013 284 . 43 . Brian Lee , Savil Srivastava , Ranjitha Kumar , Ronen Brafman , and Scott R . Klemmer . 2010 . Designing with Interactive Example Galleries . Proceedings of the SIGCHI Conference on Human Factors in Computing Systems , ACM , 2257 \u2013 2266 . http : / / doi . org / 10 . 1145 / 1753326 . 1753667 44 . Julie S . Linsey , Ian Tseng , Katherine Fu , Jonathan Cagan , Kristin L . Wood , and Christian D . Schunn . 2010 . A Study of Design Fixation , Its Mitigation and Perception in Engineering Design Faculty . Journal of Mechanical Design 132 , 4 : 041003 . 45 . Lan Luo and Olivier Toubia . 2015 . Improving Online Idea Generation Platforms and Customizing the Task Structure Based on Consumers\u2019 Domain Specific Knowledge . Journal of Marketing . http : / / doi . org / 10 . 1509 / jm . 13 . 0212 46 . Douglas L . Medin . 1989 . Concepts and conceptual structure . American Psychologist 44 , 12 : 1469 \u2013 1481 . 47 . George A . Miller . 1956 . The magical number seven , plus or minus two : some limits on our capacity for processing information . Psychological Review 63 , 2 : 81 \u2013 97 . http : / / doi . org / 10 . 1037 / h0043158 48 . Jeffrey Pennington , Richard Socher , and Christopher D Manning . 2014 . Glove : Global vectors for word repre - sentation . Proceedings of the Empiricial Methods in Natural Language Processing ( EMNLP 2014 ) 12 : 1532 \u2013 1543 . 49 . Bruce A . Reinig , Robert O . Briggs , and Jay F . Nunamaker . 2007 . On the Measurement of Ideation Quality . Journal of Management Information Systems 23 , 4 : 143 \u2013 161 . http : / / doi . org / 10 . 2753 / MIS0742 - 1222230407 50 . Mark A . Runco . 2004 . Creativity . Annual Review of Psychology 55 : 657 \u2013 687 . 51 . R . Keith Sawyer . 2012 . Explaining creativity : the science of human innovation . Oxford University Press , New York . 52 . Jamie J . Shah , Noe Vargas - Hernandez , and Steven M . Smith . 2003 . Metrics for measuring ideation effective - ness . Design Studies 24 , 2 : 111 \u2013 134 . 53 . Pao Siangliulue , Kenneth C . Arnold , Krzysztof Z . Gajos , and Steven P . Dow . 2015 . Toward Collabora - tive Ideation at Scale : Leveraging Ideas from Others to Generate More Creative and Diverse Ideas . Proceed - ings of the ACM Conference on Computer Supported Cooperative Work & Social Computing . http : / / doi . org / 10 . 1145 / 2675133 . 2675239 54 . Pao Siangliulue , Joel Chan , Kzryzstof Gajos , and Steven P . Dow . 2015 . Providing timely examples improves the quantity and quality of generated ideas . Proceed - ings of the ACM Conference on Creativity and Cogni - tion . http : / / doi . org / 10 . 1145 / 2757226 . 2757230 55 . Ut Na Sio , Kenneth Kotovsky , and Jonathan Cagan . 2015 . Fixation or inspiration ? A meta - analytic review of the role of examples on design processes . Design Studies 39 : 70 \u2013 99 . http : / / doi . org / 10 . 1016 / j . destud . 2015 . 04 . 004 56 . Steven M . Smith , Thomas B . Ward , and Jay S . Schu - macher . 1993 . Constraining effects of examples in a creative generation task . Memory & Cognition 21 , 6 : 837 \u2013 45 . 57 . John Stasko , Carsten G\u00f6rg , and Zhicheng Liu . 2008 . Jigsaw : Supporting Investigative Analysis through In - teractive Visualization . Information Visualization 7 , 2 : 118 \u2013 132 . http : / / doi . org / 10 . 1057 / palgrave . ivs . 9500180 58 . Omer Tamuz , Ce Liu , Serge Belongie , Ohad Shamir , and Adam Tauman Kalai . 2011 . Adaptively learning the crowd kernel . arXiv preprint arXiv : 1105 . 1033 . 59 . Christian Terwiesch and Yi Xu . 2008 . Innovation contests , open innovation , and multiagent problem solving . Management science 54 , 9 : 1529 \u2013 1543 . 60 . Thomas B . Ward , Merryl J . Patterson , and Cynthia M . Sifonis . 2004 . The Role of Specificity and Abstraction in Creative Idea Generation . Creativity Research Jour - nal 16 , 1 : 1 \u2013 9 . 61 . Thomas N . Wisdom and Robert L . Goldstone . 2011 . Innovation , Imitation , and Problem Solving in a Net - worked Group . Nonlinear Dynamics - Psychology and Life Sciences 15 , 2 : 229 . 62 . Jinfeng Yi , Rong Jin , Shaili Jain , Tianbao Yang , and Anil K . Jain . 2012 . Semi - crowdsourced clustering : Generalizing crowd labeling by robust distance metric learning . Advances in Neural Information Processing Systems , 1772 \u2013 1780 . 63 . Yisong Yue , Chong Wang , Khalid El - Arini , and Carlos Guestrin . 2014 . Personalized Collaborative Clustering . Proceedings of the 23rd International Conference on World Wide Web , ACM , 75 \u2013 84 . http : / / doi . org / 10 . 1145 / 2566486 . 2567991 64 . Lixiu Yu , Aniket Kittur , and Robert E . Kraut . 2014 . Distributed Analogical Idea Generation : Inventing with Crowds . Proceedings of the SIGCHI Conference on Human Factors in Computing Systems , ACM , 1245 \u2013 1254 . http : / / doi . org / 10 . 1145 / 2556288 . 2557371 65 . James Zou , Kamalika Chaudhuri , and Adam Kalai . 2015 . Crowdsourcing Feature Discovery via Adaptive - ly Chosen Comparisons . Third AAAI Conference on Human Computation and Crowdsourcing . Crowdsourcing and Creation : Large - scale Ideas and Content Production # chi4good , CHI 2016 , San Jose , CA , USA 2728",
    "kamrahHowDiverseInitial2023": "Eesh Kamrah Department of Mechanical Engineering , University of Maryland , College Park , MD 20742 e - mail : kamrah @ umd . edu Seyede Fatemeh Ghoreishi Department of Civil and Environmental Engineering & Khoury College of Computer Sciences , Northeastern University , Boston , MA 02115 e - mail : f . ghoreishi @ northeastern . edu Zijian \u201c Jason \u201d Ding College of Information Studies , University of Maryland , College Park , MD 20742 e - mail : ding @ umd . edu Joel Chan College of Information Studies , University of Maryland , College Park , MD 20742 e - mail : joelchan @ umd . edu Mark Fuge 1 Department of Mechanical Engineering , University of Maryland , College Park , MD 20742 e - mail : fuge @ umd . edu How Diverse Initial Samples Help and Hurt Bayesian Optimizers Design researchers have struggled to produce quantitative predictions for exactly why and when diversity might help or hinder design search efforts . This paper addresses that problem by studying one ubiquitously used search strategy \u2014 Bayesian optimization ( BO ) \u2014 on a 2D test problem with modi \ufb01 able convexity and dif \ufb01 culty . Speci \ufb01 cally , we test how providing diverse versus non - diverse initial samples to BO affects its performance during search and introduce a fast ranked - determinantal point process method for comput - ing diverse sets , which we need to detect sets of highly diverse or non - diverse initial samples . We initially found , to our surprise , that diversity did not appear to affect BO , neither helping nor hurting the optimizer \u2019 s convergence . However , follow - on experiments illuminated a key trade - off . Non - diverse initial samples hastened posterior convergence for the underlying model hyper - parameters \u2014 a model building advantage . In contrast , diverse initial samples accelerated exploring the function itself \u2014 a space exploration advan - tage . Both advantages help BO , but in different ways , and the initial sample diversity directly modulates how BO trades those advantages . Indeed , we show that \ufb01 xing the BO hyper - parameters removes the model building advantage , causing diverse initial samples to always outperform models trained with non - diverse samples . These \ufb01 ndings shed light on why , at least for BO - type optimizers , the use of diversity has mixed effects and cautions against the ubiquitous use of space - \ufb01 lling initializations in BO . To the extent that humans use explore - exploit search strategies similar to BO , our results provide a testable conjecture for why and when diversity may affect human - subject or design team experiments . [ DOI : 10 . 1115 / 1 . 4063006 ] Keywords : design of experiments , design optimization , machine learning , metamodeling 1 Introduction and Related Work One open question within design research is when or under what conditions providing diverse stimuli or starting solutions to either humans or algorithms can improve their designs \u2019 \ufb01 nal performance . Researchers have struggled to produce quantitative predictions or explanations for exactly why and when diversity might help or hinder design search efforts . In studies of human designers or teams , there have been numerous empirical results on the effect of diverse stimuli or sets of stimuli on designers , typically referred to under the topic of design \ufb01 xation ( for recent reviews , see Refs . [ 1 , 2 ] ) . In general , available empirical results are mixed and it is dif - \ufb01 cult to quantitatively predict , for a new problem or person , whether or not diversity in problem stimuli will or will not help . For instance , there are a number of empirical demonstrations of positive effects of example diversity on novelty and diversity of ideas [ 3 \u2013 5 ] , but substantially more mixed results on the effects of diversity on solution quality , with some observations of positive effects [ 6 \u2013 8 ] , some null or contingent effects [ 4 , 9 \u2013 14 ] , and even some negative effects on solution quality [ 15 , 16 ] . Likewise , in research focused purely on optimization , common academic and industrial practice initializes search algorithms with different strategies like Latin hypercube sampling ( LHS ) [ 17 ] and others in an attempt to \u201c \ufb01 ll \u201d or \u201c cover \u201d a space as uniformly as pos - sible [ 18 ] or via quasi - random methods [ 19 \u2013 21 ] . Some methods build diversity - encouraging loss functions directly into their core search algorithms , such as in common meta - heuristic optimizers [ 22 ] such as particle swarm optimization , simulated annealing , and genetic algorithms , with one of the most well - known diversity - inducing ones being the non - dominated sorting Genetic Algorithm - II ( NSGA - II ) [ 23 ] . For Bayesian optimization ( BO ) spe - ci \ufb01 cally , a common strategy is to build diversity directly into the acquisition function used in sampling new points from the Gaussian process ( GP ) posterior [ 24 ] . As with human - subjects experiments , the precise effect of diversity on optimization performance is often problem dependent [ 22 ] and dif \ufb01 cult to predict a priori . Nev - ertheless , optimization practitioners take these steps to improve initial sample diversity with the hope that the optimizer will con - verge faster or \ufb01 nd better global optima . But is encouraging initial diversity in this way always a good idea ? If so , when and why is it good ? Are there any times or con - ditions when diversity might hurt rather than help our search for good designs ? 1 Corresponding author . Contributed by the Design Automation Committee of ASME for publication in the J OURNAL OF M ECHANICAL D ESIGN . Manuscript received August 31 , 2022 ; \ufb01 nal manuscript received July 13 , 2023 ; published online August 29 , 2023 . Assoc . Editor : Douglas Allaire . Journal of Mechanical Design NOVEMBER 2023 , Vol . 145 / 111703 - 1 Copyright \u00a9 2023 by ASME D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 ( Spoiler Alert : Yes , it can \u2014 see Sec . 4 for how and Sec . 6 for why . ) To address the above questions , this paper studies one type of commonly used search strategy \u2014 BO \u2014 and how the diversity of its initialization points affects its performance on a search task . We uncover a fascinating dance that occurs between two competing advantages that initial samples endow upon BO \u2014 a model building versus space exploration advantage that we de \ufb01 ne later \u2014 and how the initial samples \u2019 diversity directs the choreography . While the fundamental reason for this interplay will later appear straightfor - ward ( and perhaps even discernible through thought experiments rather than numerical experiments ) , it nevertheless \ufb02 ies in the face of how most practitioners initialize their BO routines or conduct optimal experimental design studies . It also posits a testable prediction about how to induce greater effects of diversity on novice human designers or the conditions under which there may be mixed or even negative effects ( see Sec . 6 ) . Before describing our particular experiment and results , we will \ufb01 rst review why BO is a meaningful and generalizable class of search algorithm to use , as well as past work that has tried to under - stand how diversity affects search processes such as optimization . Why model design search as Bayesian optimization ? While this paper addresses only BO , this is an important algorithm in that it plays an out - sized role within the design research and optimization community . For example , BO underlies a vast number of industri - ally relevant gradient - free surrogate modeling approaches imple - mented in major design or analysis packages , where it is referred to under a variety of names , including Kriging methods or meta - modeling [ 25 , 26 ] . Its use in applications of computationally expen - sive multidisciplinary optimization problems is , while not unilateral [ 27 ] , quite widespread . Likewise , researchers studying human designers often use BO as a proxy model [ 28 ] to understand human search , due to the interplay between exploration and exploi - tation that lies at the heart of most BO acquisition functions like expected improvement ( EI ) . More generally , there is a robust history of fruitful research in cognitive science modeling human cognition as Bayesian processing [ 29 ] , such as concept learning in cognitive development [ 30 ] , causal learning [ 31 ] , and analogical reasoning [ 32 ] . While the bulk of BO - related papers focus on new algorithms or acquisition functions , few papers focus on how BO is initialized , preferring instead the general use of space - \ufb01 lling initializations that have a long history in the \ufb01 eld of optimal experiment design [ 27 ] . In contrast , this paper shows that in certain situations that faith in space - \ufb01 lling designs might be misplaced , particularly when the BO kernel hyper - parameters are adjusted or \ufb01 t during search . What does it even mean for samples to be diverse ? As a practical matter , if we wish to study how diverse samples impact BO , we face a subtle but surprisingly non - trivial problem : how exactly do you quantify whether one set of samples is more or less diverse than another ? This is a set - based ( i . e . , combinatorially large ) problem with its own rich history too large to cover extensively here , however our past work on diversity measurement [ 33 \u2013 35 ] , compu - tation [ 36 ] , and optimization [ 37 ] provides further pointers for inter - ested readers , and in particular the thesis of Ahmed provides a good starting point for the broader literature and background in this area [ 38 ] . For the purposes of understanding how this paper relates to exist - ing approaches , it suf \ufb01 ces to know the following regarding common approaches to quantifying diversity : ( 1 ) most diversity measure - ment approaches focus on some variant of a hyper - volume objective spanned by the set of selected points ; ( 2 ) since this measure depends on a set rather than individual points , it becomes combinatorially expensive , necessitating fast polynomial - time approximation , one common tool for which is a determinantal point process ( DPP ) [ 39 ] ; however , ( 3 ) while sampling the most diverse set via DPPs is easy , sampling percentile sets from the DPP distribution to get the top 5 % , median , or lowest 5 % of diverse sets becomes exceed - ingly slow for a large sample pool . In contrast , for this paper , we created a faster DPP - type sampling method to extract different percentiles of the distribution without actually needing to observe the entire DPP distribution and whose sampling error we can bound using concentration inequalities . Section 2 provides further mathematical background , including information on DPP hyper - parameters and how to select them intel - ligently , and the Supplementary Material on the ASME Digital Collection provides further algorithmic details . With an understand - ing of diversity distribution measures in hand , we can now address diversity \u2019 s speci \ufb01 c effects on optimization more generally . How does diversity in initial inputs affect optimizers ? While there are a number of papers that propose either different initialization strategies or benchmarking of existing strategies for optimization , there is limited prior work addressing the direct effect of initial sample diversity . For general reviews and benchmarking on how to initialize opti - mizers and the effects of different strategies , papers such as Refs . [ 20 , 22 ] compare initialization strategies for particular optimizers and quantify performance differences . An overall observation across these contributions is the inability of a single initialization method to improve performance across functions of varying com - plexity . These studies also do not directly measure or address the role of sample diversity directly , only noting such behavior as it cor - relates indirectly with the sampling strategy . A second body of work tries to customize initialization strategies on a per - problem basis , often achieving faster convergence on domain - speci \ufb01 c problems [ 18 , 19 , 40 \u2013 42 ] . While useful in their designed domain , these studies do not directly address the role of diversity either . In contrast , this paper addresses diversity directly using properties of BO that are suf \ufb01 ciently general to apply across multiple domains and applications . Lastly , how to initialize optimizers has garnered new interest from the machine learning community , for example , in the initial settings of weights and biases in a neural network and the down - stream effects on network performance [ 43 , 44 ] . There is also general interest in how to collect diverse samples during learning , either in an active learning [ 45 ] or reinforcement learning context [ 46 , 47 ] ; however , those lines of work address only diversity throughout data collection , rather than the impact of initial samples considered in this paper . What does this paper contribute beyond past work ? This paper \u2019 s speci \ufb01 c contributions are : ( 1 ) To compute diversity : we describe a fast DPP - based diversity scoring method for selecting diverse initial examples with a \ufb01 xed size k . Any set of size k with these initial examples can be then used to approximate the percentile of diversity that the set belongs to . This method requires selecting a hyper - parameter relating to the DPP measure . We describe a principled method for selecting this parameter in Sec . 2 . 1 and provide numerical evidence of the improved sampling performance in the Supplementary Material . Compared to prior work , this makes percentile sampling of DPP distribu - tions computationally tractable . ( 2 ) To study effects on BO : we empirically evaluate how diverse initial samples affect the convergence rate of a Bayesian opti - mizer . Section 4 \ufb01 nds that low diversity samples provide a model building advantage to BO while diverse samples provide a space exploration advantage that helps BO con - verge faster . Section 5 shows that removing the model build - ing advantage makes having diverse initial samples uniformly better than non - diverse samples . 2 We will next describe our overall experimental approach and common procedures used across all three of our main experiments . 2 For grammatical simplicity and narrative \ufb02 ow , we will use the phrase \u201c non - diverse \u201d throughout the paper to refer to cases where samples are taken from the 5th percentile of diverse sets \u2014 these aretechnically \u201c low - diversity \u201d rather thanbeing abso - lutely \u201c non - diverse \u201d which would occur when all points in the set are identical , but we trust that readers can keep this minor semantic distinction in mind . 111703 - 2 / Vol . 145 , NOVEMBER 2023 Transactions of the ASME D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 We will introduce individual experiment - speci \ufb01 c methods only when relevant in each separate experiment section . To fully repro - duce the below results , the link in the footnote provides the code associated with this paper . 3 2 Overall Experimental Approach This section will \ufb01 rst describe how we compute diverse initial samples , including how we set a key hyper - parameter that controls the DPP kernel needed to measure sample set diversity . It then brie \ufb02 y describes the controllable 2D test problem that we use in our experiments . It ends with a description of how we setup the BO search process and the hyper - parameters that we study more deeply in each individual experiment . 2 . 1 Measuring and Sampling From Diverse Sets Using Determinantal Point Processes . As mentioned above , we measure diversity of a set of points using DPP , which get their name from the fact that they compute the determinant of a matrix referred to as an L - ensemble ( as seen in Eq . ( 1 ) ) that correlates with the volume spanned by a collection or set of samples ( Y ) taken from all possible sets ( Y ) given a diversity / similarity ( feature ) metric . P ( L Y ) \u221d det ( K ( L Y ) ) ( 1 ) Here , L is the ensemble de \ufb01 ned by any positive semi - de \ufb01 nite matrix [ 39 ] , and K is the kernel matrix . For sampling diverse examples , this positive semi - de \ufb01 nite matrix is typically chosen as a kernel matrix ( K ) that de \ufb01 nes the similarity measure between pairs of data points . For this paper , we use a standard and commonly used similarity measure de \ufb01 ned using a radial basis function ( RBF ) kernel matrix [ 48 ] . Speci \ufb01 cally , each entry in L Y for two data points with index i and j is [ L Y ] i , j = exp \u2212 \u03b3 \u00b7 (cid:2) x i \u2212 x j (cid:2) 2 (cid:1) (cid:3) ( 2 ) The hyper - parameter \u03b3 in the DPP kernel can be set in the interval ( 0 , \u221e ) and will turn out to be quite important in how well we can measure diversity . The next section explores this choice in more depth , but to provide some initial intuition : set \u03b3 too high and any selection of points looks equally diverse compared to any other set , essentially destroying the discriminative power of the DPP , while setting \u03b3 too low causes the determinant of L to collapse to zero for any set of cardinality greater than the feature - length of x . With L in hand , we can now turn Eq . ( 1 ) into an equality by using the fact that (cid:4) Y \u2282 Y det ( L Y ) = det ( L + I ) , where I is an identity matrix of the same shape as the ensemble matrix L . Then , using Theorem 2 . 2 from Ref . [ 39 ] , we can write the P ( Y \u2208 Y ) as follows : P ( Y ) = det ( L Y ) det ( L + I ) ( 3 ) This is the probability that a given set of points ( Y ) is highly diverse compared to other possible sets ( Y ) \u2014 that is , the higher P ( Y ) the more diverse the set . The popularity of DPP - type measures is due to their ability to ef \ufb01 ciently sample diverse samples of \ufb01 xed size k . Sampling a set of k samples from a DPP is done using a con - ditional DPP called k - DPP [ 49 ] . k - DPP are able to compute mar - ginal and conditional probabilities with polynomial complexity , in turn allowing sampling from the DPP in polynomial complexity . k - DPPs are also well researched and there exists several different methods to speed up the sampling process using a k - DPP [ 50 , 51 ] . Our approach allows sampling in constant complexity ; however , there is a trade - off in complexity in generating the DPP distribution . The complexity for generating traditional DPP distributions is inde - pendent of \u201c k , \u201d while our approach has linear dependence on \u201c k . \u201d Since existing k - DPP approaches lack the ability to ef \ufb01 ciently sample from different percentiles of diversity and thus make it com - putationally expensive to regenerate the distribution to alternatively sample from different percentiles . To tackle this problem , our approach is designed to sample ef \ufb01 - ciently from different percentiles of diversity . This is made possible by creating an absolute diversity score . This score is generated by taking a log determinant of the kernel matrix de \ufb01 ned over the set Y . Randomly sampling the k - DPP space allows us to bound errors in generating this absolute score through the use of concentration inequalities . The details about how to sample from this distribution and calculate the score are mentioned in the Supplementary Material , so as not to disrupt the paper \u2019 s main narrative . Addition - ally , the Supplementary Material provides empirical results to support our earlier claims regarding ef \ufb01 cient sampling from our approach versus the traditional k - DPP approach , as well as the trade - off in complexity when generating the DPP distribution . Figure 2 shows example sets of \ufb01 ve points and their corresponding DPP score , where the diversity score is monotonic and a positive score corresponds to a more diverse subset . 2 . 1 . 1 Selecting the Hyper - parameter for the Determinantal Point Process Kernel . As mentioned above , the choice of \u03b3 impacts the accuracy of the DPP score , and when we initially \ufb01 xed \u03b3 to | Y i | / 10 , where Y i is the set of data points over which the RBF kernel is calculating the DPP score as suggested by Ref . [ 52 ] , the DPP seemed to be producing largely random scores . To select an appropriate \u03b3 , we designed a kernel - independent diagnostic method for assessing the DPP kernel with four steps . First , we randomly generate M samples of size k sets ( think of these as random k - sized samples from Y ) . Second , we compute their DPP scores for different possible \u03b3 values and then sort those M sets by that score . Third , we compute the rank correlation among these sets for different pairs of \u03b3 \u2014 intuitively , if the rank cor - relation is high ( toward 1 ) , then either choice of \u03b3 would produce the same rank orders of which points were considered diverse , meaning the ( relative ) DPP scores are insensitive to \u03b3 . In contrast , if the rank correlation is 0 , then the two \u03b3 values produce essentially random orderings . This rank correlation between two different \u03b3 settings is the color / value shown in each cell of the matrix in Fig . 1 . Large ranges of \u03b3 with high - rank correlation mean that the rankings of DPP scores are stable or robust to small perturbations in \u03b3 . Lastly , Fig . 1 Correlation matrix showing the relative correlation between two gammas by comparing the way our DPP approach ranks 10 , 000 sampled sets of cardinality , k = 10 . The gamma values in both axes here are logarithmic values with base 10 . 3 https : / / github . com / IDEALLab / JMD - Diversity - in - Bayesian - Optimization Journal of Mechanical Design NOVEMBER 2023 , Vol . 145 / 111703 - 3 D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 we use this \u201c robust \u03b3 \u201d region by choosing the largest range of \u03b3 values that have a relative correlation index of 0 . 95 or higher . We compute the mean of this range and use that as our selected \u03b3 in our later experiments . We should note that the functional range of \u03b3 is dependent on sample size ( k ) , and so this \u201c robust \u03b3 \u201d needs to be recomputed for different initialization sizes . The detailed settings for the results as seen in Fig . 1 are as follows : M = 10 , 000 ; k = 10 ; and \u03b3 \u2208 [ e \u2212 7 , e \u2212 2 ] . The correlation matrix shows a range of \u03b3 with strongly correlating relative ordering of the test sets . All \u03b3 within this range provide a consistent ranking . 2 . 2 A Test Function With Tunable Complexity . A problem that is common across the study of initialization methods is their inconsistency across problems of varying dif \ufb01 culty , motivating the need to test BO \u2019 s search behavior on a problem class with var - iable complexity . Synthetic objective functions are often used to test the ef \ufb01 ciency of different optimizers and there are several libraries online to choose these functions from Ref . [ 53 ] , though these func - tions are largely static , in the sense that there is only a single test function de \ufb01 nition . There has been research into developing objec - tive function generators ; for example , in Ref . [ 54 ] , the author uses a mixture of four features to generate synthetic objective functions . These have been well categorized and the relative performance of different optimizers is documented on each landscape . Similar to this , Ref . [ 55 ] looks at using a mixture of different sinusoidal func - tions to create a noisy 1D function . Both the generators discussed are capable of generating complicated landscapes , but the complex - ity arises from mixing different randomly generated sinusoids and thus are unable to control or quantify a measure of complexity of the generated landscapes . To address this controllable complexity problem directly , we created a simple 2D test function generator with tunable complexity parameters that allow us to instantiate multiple random surfaces of similar optimization dif \ufb01 culty . We modi \ufb01 ed this function from the one used in Ref . [ 56 ] where it was referred to as \u201c Wildcat Wells , \u201d though the landscape is functionally just a normal distribu - tion with additive noise of different frequency spectra . We used four factors to control the synthetic objective functions : ( 1 ) the number of peaks , ( 2 ) noise amplitude , ( 3 ) smoothness , and ( 4 ) distance between peaks and a seed . The number of peaks control the number of layers of multivariate normal with single peaks . The noise amplitude in the range of [ 0 , 1 ] controls the relative height of the noise compared to the height of the peaks . Setting this to 1 would essentially make the noise in the function as tall as the peaks and give the function in \ufb01 nite peaks . Smoothness in the range of [ 0 , 1 ] controls the weighted contribution of the smooth Gaussian function compared to the rugged noise to the wildcat wells landscape . Setting this to 1 would remove the noise from the function because then the normal distribution completely con - trols and dominates the function . The last parameter , the distance between peaks , can be tuned in the range of [ 0 , 1 ] . This parameter prevents overlap of peaks when the function is generated with more than one peak . Some of these parameters overlap in their effects . For example , N controls the number of peaks , and ruggedness amplitude controls the height of the noise in the function , so increasing the noise auto - matically increases the peaks in the function thus we will only look at varying the ruggedness amplitude . Apart from this , ruggedness frequency ( the distance between peaks ) plays the same role as smoothness ( radius of in \ufb02 uence of each individual on its neighbor ) . Thus , for the numerical experiments presented in Secs . 3 \u2013 5 , only the ruggedness amplitude and smoothness will be varied between [ 0 . 2 , 0 . 8 ] with increments of 0 . 2 . To provide some visual examples of the effect of these parameters on the generated functions , Fig . 3 visualizes an example random surface generated with different smoothness and ruggedness amplitude parameters . 2 . 3 Bayesian Optimization . BO has emerged as a popular sample - ef \ufb01 cient approach for optimization of these expensive black - box functions . BO models the black - box function using a sur - rogate model , typically a GP . The next design to evaluate is then selected according to an acquisition function . The acquisition func - tion uses the GP posterior and makes the next recommendation for function evaluation by balancing between exploration and exploita - tion . It allows exploration of regions with high uncertainty in the objective function and exploitation of regions where the mean of the objective function is optimum . At each iteration , the GP gets updated according to the selected sample , and this process contin - ues iteratively according to the available budget . Each data point in the context of Bayesian optimization is extremely expensive ; thus , there is a need for selection of an infor - mative set of initial samples for the optimization process . Toward this , this paper investigates the effect of the level of initial diverse Fig . 2 Scatter plots showing randomly chosen sets with k = 5 high - diversity and low - diversity samples with their diversity score on top of each of the chosen set 111703 - 4 / Vol . 145 , NOVEMBER 2023 Transactions of the ASME D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 coverage of the input space on convergence of Bayesian optimiza - tion policies . For the purpose of numerical experiments , the optimizer used is from the BOTorch Library [ 57 ] . The optimizer uses a single task GP model with expected improvement ; the kernel used is a Mat\u00e9rn kernel . A GP is speci \ufb01 ed by its mean and covariance functions as f ( x ) \u223c GP \u03bc ( x ) , k ( x , x ) (cid:1) (cid:3) ( 4 ) where \u03bc ( . ) and k ( . , . ) are the mean function and a real - valued kernel function encoding the prior belief on the correlation among the samples in the design space . In Gaussian process regression , the kernel function dictates the structure of the surrogate model we can \ufb01 t . An important kernel for Bayesian optimization is the Mat\u00e9rn kernel , which incorporates a smoothness parameter \u03bd to permit greater \ufb02 exibility in modeling functions : k M at \u00b4 ern ( x 1 , x 2 ) = 2 1 \u2212 \u03bd \u0393 ( \u03bd ) (cid:5)(cid:5)(cid:5) 2 \u03bd \u221a (cid:2) x 1 \u2212 x 2 (cid:2) \u03b8 (cid:6) (cid:7) \u03bd H \u03bd (cid:5)(cid:5)(cid:5) 2 \u03bd \u221a (cid:2) x 1 \u2212 x 2 (cid:2) \u03b8 (cid:6) (cid:7) ( 5 ) where \u0393 ( \u00b7 ) and H \u03bd ( \u00b7 ) are the Gamma function and the Bessel func - tion of order \u03bd , and \u03b8 is the length scale hyper - parameter which denotes the correlation between the points within each dimension and speci \ufb01 es the distance that the points in the design space in \ufb02 u - ence one another . Here , we use a constant mean for the mean func - tion . The model building advantage that we refer to in this paper corresponds to learning these hyper - parameters . The hyper - parameters of the Gaussian process , namely , the parameters of the kernel function and the mean function are : Lengthscale of the Mat\u00e9rn Kernel : In Eq . ( 5 ) , \u03b8 is the lengthscale parameter of the kernel . This parameter controls the ruggedness expected by the Bayesian optimizer in the black box function being studied . The effects of the parameter are similar to \u03bd , but \u03bd is not learned during the optimization process while lengthscale is . So , \u03bd is not studied as a parameter that in \ufb02 uences the modeling behavior but rather studied as an additional parameter for sensitivity . Output Scale of Scale Kernel : Output scale is used to control how the Mat\u00e9rn kernel is scaled for each batch . Since our Bayesian opti - mizer uses a single task GP , we do not use batch optimization . Thus , this parameter is unique for us and the way it is implemented using BoTorch can be seen in Eq . ( 6 ) . K scaled = \u03b8 scale K orig ( 6 ) Noise for Likelihood Calculations : The noise parameter is used to model measurement error or noise in the data . So , as the Gaussian process gets more data , the noise term decreases . So , ideally , this term should converge to 0 when the Bayesian optimizer has found an optimal value since our test functions did not have any added noise . Constant for Mean Module : This constant is used as the mean for the normal distribution that forms the prior of the Gaussian process as shown in Eq . ( 4 ) . Further studies and results regarding the effects of the hyper - parameters are available in the Supplementary Material . We now describe the \ufb01 rst experiment where we explore the effects of diversity of initial samples on the convergence of Bayesian optimizers . 3 Experiment 1 : Does Diversity Affect Optimization Convergence ? 3 . 1 Methods . To test the effects of diversity of initial samples on optimizer convergence , we \ufb01 rst generated a set of initial training samples of size ( k ) 10 either from low ( 5th percentile of diversity ) or high - diversity ( 95th percentile of diversity ) using our procedure in Sec . 2 . 1 . Next , we created 100 different instances of the wildcat wells function with different randomly generated seeds for each cell in a 4 \u00d7 4 factor grid of four values each of the smoothness and ruggedness amplitude parameters of the wildcat wells function ( ranging from 0 . 2 to 0 . 8 , in steps of 0 . 2 ) . For simplicity here , we Fig . 3 A grid plot showing how the landscape of wildcat wells changes with smoothness and ruggedness amplitude Journal of Mechanical Design NOVEMBER 2023 , Vol . 145 / 111703 - 5 D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 refer to these combinations as families of the wildcat wells function . This resulted in 1600 function instances . Our experiment consisted of 200 runs of the Bayesian optimizer within each of the smoothness \u2013 ruggedness function families , where each run consisted of 100 iterations , and half of the runs were ini - tialized with a low - diversity training sample , and half were initial - ized with a high - diversity training sample . We then compared the cumulative optimality gap ( COG ) across the iterations for the runs with low - diverse initializations and high - diverse initializations within each smoothness \u2013 ruggedness combination family . We did this by computing bootstrapped mean and con \ufb01 dence intervals within each low - diverse and high - diverse sets of runs within each family . Given the full convergence data , we compute a COG which is just the area under the optimality gap curve for both the 5th and 95th diversity curves . Intuitively , a larger COG corresponds to a worse overall performance by the opti - mizer . Using these COG values , we can numerically calculate the improvement of the optimizer in the 95th percentile . The net improvement of COG value while comparing the 5th and 95th per - centile is also presented as text in each subplot in Fig . 4 . 3 . 2 Results . As Fig . 4 shows , the cumulative optimality gap does not seem to have a consistent effect across the grid . Diversity produces a positive convergence effect for some cells , but is negative in others . Moreover , there are wide empirical con \ufb01 dence bounds on the mean effect overall , indicating that should an effect exist at all , it likely does not have a large effect size . Changing the function rug - gedness or smoothness did not signi \ufb01 cantly modulate the overall effect . As expected , given suf \ufb01 cient samples ( far right on the x - axis ) , both diverse and non - diverse initializations have the same optimality gap , since at that point the initial samples have been crowded out by the new samples gathered by BO during its search . 3 . 3 Discussion . Overall , the results from Fig . 4 seem to indi - cate that diversity helps in some cases and hurts in others , and regardless has a limited impact one way or the other . This seems counter to the widespread practice of diversely sampling the initial input space using techniques like LHS . Figure 4 shows that it has little effect . Why would this be ? Given decades of research into initialization schemes for BO and optimal experiment design , we expected diver - sity to have at least some ( perhaps small but at least consistent ) pos - itive effect on convergence rates , and not the mixed bag that we see in Fig . 4 . How were the non - diverse samples gaining such an upper hand when the diverse samples had a head start on exploring the space \u2014 what we call a space exploration advantage ? The next section details an experiment we conducted to test a hypothesis regarding a potential implicit advantage that non - diverse samples might endow to BO that would impact the convergence of BO \u2019 s hyper - parameter posteriors . As we will see next , this acceler - ated hyper - parameter posterior convergence caused by non - diverse initialization is the Achilles \u2019 heel of diversely initialized BO that allows the non - diverse samples to keep pace and even exceed diverse BO . 4 Experiment 2 : Do Lower Diversity Samples Improve Hyper - parameter Posterior Convergence ? After reviewing the results from Fig . 4 , we tried to determine why the space exploration advantage of diversity was not helping BO as we thought it should . We considered as a thought experiment the one instance where a poorly initialized BO model with the same acquisition function might outperform another : if one model \u2019 s kernel hyper - parameter settings were so grossly incorrect that the Fig . 4 Experiment 1 : optimality gap grid plot showing the difference in current optimality gap between optimizers initialized with 5th versus 95th percentile diverse sample ( y - axis ) as a function of optimization iteration ( x - axis ) . The different factors in the factor grid plot the effects of diversity as the noise amplitude and smoothness are varied in the range [ 0 . 2 , 0 . 8 ] . Each plot also has text indicating the net cumulative optimality gap ( NCOG ) , a positive value corresponds to a better performance by high - diversity samples compared to the low - diversity samples . The plot shows that BO bene \ufb01 ts from diversity in some cases but not others . There is no obvious trends in how the NCOG values change in the grid . The results are further discussed in Sec . 3 . 111703 - 6 / Vol . 145 , NOVEMBER 2023 Transactions of the ASME D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 model would waste many samples exploring areas that it did not need to if it had the correct hyper - parameters . Could this misstep be happening in the diversely sampled BO but not in the non - diverse case ? If so , this might explain how non - diverse BO was able to keep pace : while diverse samples might give BO a head start , it might be unintentionally blindfolding BO to the true function posteriors , making it run ragged in proverbial directions that it need not . If this hypothesis was true , then we would see this re \ufb02 ected in the comparative accuracy of the kernel hyper - parameters learned by the diverse versus non - diverse BO samples . This experiment set out to test that hypothesis . 4 . 1 Methods . The key difference from experiment 1 is that rather than comparing the overall optimization convergence , we instead focus on how the initial samples \u2019 diversity affects BO \u2019 s hyper - parameter posterior convergence , and compare how far each is from the \u201c ground truth \u201d optimal hyper - parameters . As with experiment 1 , we used the same smoothness and rugged - ness amplitude families of the wildcat wells function . To then gen - erate the data for each instance in one of these families , we sampled 20 sets of initial samples . Half of the sampled 20 sets were low ( 5th percentile of diversity ) and the other half from high - diversity ( 95th percentile of diversity ) percentiles . For each initial sample , we then maximized the GP \u2019 s kernel mar - ginal log likelihood ( via BOTorch \u2019 s GP \ufb01 t method ) . We then recorded the hyper - parameters obtained for all 20 initial samples . The mean of the ten samples from low diversity was then used as one point in the box plot \u2019 s low - diversity distribution as seen in Fig . 5 . We then repeated this process for the high - diversity initial samples . Each point in the box plot can then be understood as the mean hyper - parameter learned by BOTorch given just the initial sample of size ( k ) 10 points . To get the full box plot distribution for each family , the above process is repeated over 100 seeds and Fig . 5 provides the resulting box plot for both diverse and non - diverse initial samples for all the 16 families of wildcat wells func - tion as described in experiment 1 . To provide a ground truth for the true hyper - parameter settings , we ran a binary search to \ufb01 nd the size of the sample ( k optimal ) for which BO \u2019 s kernel hyper - parameters converged for all families . The hyper - parameter found by providing k optimal amount of points for each instance in the family was then plotted as a horizontal line in each box plot . An interesting observation is that some fam - ilies have non - overlapping horizontal lines . This is because for Fig . 5 Experiment 2 : box plot showing distribution of \u201c lengthscale \u201d hyper - parameter learned by BO when initiated with diverse ( right - side ) and non - diverse samples ( left - side ) for 16 different families of wildcat wells functions of the same parameters but 100 different seeds . The optimal hyper - parameter for each of the 100 wildcat wells instances from each family is also plotted as hor - izontal lines \u2014 in many but not all cases these overlap . Each cell in the plot also has the 95th percentile con \ufb01 dence bound on MAE for both diverse and non - diverse samples . The results show that MAE con \ufb01 dence bounds for non - diverse samples are smaller compared to diverse samples for all the families of wildcat wells function . Thus , indicating a presence of model building advan - tage for non - diverse initial samples . The results of this \ufb01 gure are further discussed in Sec . 4 . Journal of Mechanical Design NOVEMBER 2023 , Vol . 145 / 111703 - 7 D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 some families there are more than one modes of \u201c optimal hyper - parameters . \u201d The mode chosen as the \u201c optimal hyper - parameter \u201d is the more observed mode . The process for \ufb01 nding the \u201c optimal hyper - parameter \u201d and which mode is chosen as the optimal hyper - parameter have been described in the Supplementary Material . If an initial sample provides a good initial estimate of the kernel hyper - parameter posterior , then the box plot should align well or close to the horizontal lines of the true posterior . Figure 5 only shows the results for the Mat\u00e9rn kernel \u2019 s lengthscale parameter , given its out - sized importance in controlling the GP function poste - riors compared to the other hyper - parameters ( e . g . , output scale , noise , etc . ) , which we do not plot here for space reasons . We provide further details and plots for all hyper - parameters in the Supplementary Material for interested readers . To quantify the average distance between the learned and true hyper - parameters , we also plot in Fig . 5 the mean absolute error ( MAE ) for both highly diverse ( 95th ) and less diverse ( 5th ) points . The MAE is the sum of the absolute distance of each pre - dicted hyper - parameter from the optimal hyper - parameter for the particular surface of each wildcat wells function . The range as seen in each cell in Fig . 5 corresponds to a 95th percentile con \ufb01 - dence bound on the mean absolute error across all the 100 runs . 4 . 2 Results and Discussion . The results in Fig . 5 show that the MAE values for low diversity samples are always lower compared to the MAE for high - diversity samples . This general behavior is also qualitatively visible in the box plot . This means that after only the initial samples , the non - diverse samples provided much more accurate estimates of the kernel hyper - parameters compared to diverse samples . Moreover , BO systematically underestimates the correct lengthscale with diverse samples \u2014 this corresponds to the diverse BO modeling function posteriors that have higher fre - quency components than the true function actually does ( as shown via the pedagogical examples in the Supplementary Material ) . This provides evidence for the model building advantage of non - diverse samples that we de \ufb01 ned in Sec . 2 . 3 . It also con \ufb01 rms our pre - vious conjecture from the thought experiment that diverse samples might be impacting BO by causing slower or less accurate conver - gence to the right BO hyper - parameters . The space exploration advantage of the diverse samples helps it compensate somewhat for its poor hyper - parameters , but BO trained with non - diverse samples can leverage the better hyper - parameters to make more judicious choices about what points to select next . We did not see major differences in the other three kernel hyper - parameters such as output scale , noise , or the mean function ( see Supplementary Material ) ; however , this is not surprising , since BO is not highly sensitive to any of these parameters and the lengthscale parameter dominates large changes in BO behavior . Comparing the different smoothness and ruggedness settings , when the function is more complex ( the top right of the grid at low smoothness and high ruggedness amplitude values ) , the function \u2019 s lengthscale is lower and closer to the value learned by the diverse samples . Looking at the low diversity MAE values ( \u201c MAE 5 \u201d ) , we can see they are much closer to those of the high - diversity samples ( \u201c MAE 95 \u201d ) , in contrast to when the function is less complex ( bottom left side of the grid ) . Under such conditions , low - diversity samples lose some of the relative model building advantage they have over high - diversity samples . This conjecture aligns with exper - iment 1 ( Fig . 4 ) where the COG values on the top right part are positive while those on the bottom left are negative . Figure 5 demonstrated our hypothesized model building advan - tage that non - diverse initial samples confer to BO . But how do we know that this is the actual causal factor that accelerates BO conver - gence , and not just correlated with some other effect ? If correct , our conjecture posits a natural testable hypothesis : if we \ufb01 x the values of the hyper - parameter posteriors to identical values between the non - diverse and diverse samples and do not allow the BOto update or optimize them , then this should effectively eliminate the model building advantage , and diverse samples should always outperform non - diverse samples . Metaphorically , if we were to take away the arrow that Paris used against Achilles , would the Battle of Troy have ended differently ? Our next experiment \ufb01 nds this out . 5 Experiment 3 : Does Diversity Affect Optimization Convergence If Hyper - parameters Are Fixed to Optimal Values ? 5 . 1 Methods . This experiment is identical to experiment 1 , with two key differences : ( 1 ) we now \ufb01 x the kernel hyper - parameters to the \u201c optimal hyper - parameter \u201d values we found in experiment 2 for all the instances in each family of the wildcat wells function ( 2 ) and we do not allow either BO model to further optimize the kernel hyper - parameters . This should remove the hypothesized model building advantage of non - diverse samples without altering any other aspects of experiment 1 and the results in Fig . 4 . 5 . 2 Results and Discussion . Figure 6 shows that once the kernel hyper - parameters are \ufb01 xed \u2014 removing the model building advantage of non - diverse samples \u2014 diverse samples consistently and robustly outperform non - diverse initial samples . This holds for both the initial optimality gap at the beginning of the search as well as the cumulative optimality gap and is not qualitatively affected by the function smoothness or roughness amplitude . Unlike in experiment 1 where diversity could either help or hurt the optimizer , once we remove the model building advantage , diver - sity only helps . 6 General Discussion and Conclusions 6 . 1 Summary and Interpretation of Findings . This paper \u2019 s original goal was to investigate how and when diverse initial samples help or hurt Bayesian optimizers . Overall , we found that the initial diversity of the provided samples created two competing effects . First , experiment 2 showed that non - diverse samples improved BO \u2019 s abilities to quickly converge to optimal hyper - parameters \u2014 we called this a model building advantage . Second , experiment 3 showed that conditioned on the same \ufb01 xed hyper - parameters diverse samples improved BO \u2019 s convergence to the optima through faster exploration of the space \u2014 we called this a space exploration advantage . In experiment 1 , diversity had mixed - to - negligible effects since both of these advantages were in play and competed with one another . This interaction provides insight for academic or industrial BO users since common practice recommends initializing BO with space - \ufb01 lling samples ( to take advantage of the space exploration advantage ) , and ignores the model building advantage of non - diverse samples . Beyond our main empirical result , our improvements to existing diverse sampling approaches ( Sec . 2 . 1 ) provide new methods for studying how different percentile diversity sets affect phenomena . Researchers may \ufb01 nd this contribution of separate technical and scienti \ufb01 c interest for related studies that investigate the impact of diversity . 6 . 2 Implications and Future Work . Beyond the individual results we observed and summarized in each experiment , there are some overall implications and limitations that may guide future work or interpretation of our results more broadly , which we address below . Where does this model building advantage induced by non - diverse samples come from ? As we conjectured in experiment 2 ( Sec . 4 ) and con \ufb01 rmed in experiment 3 ( Sec . 5 ) , the key advantage of using non - diverse initial samples lies in their ability to induce faster and more accurate posterior convergence when inferring the optimal kernel hyper - parameters , such as length scale and others . This allowed the BO to make more judicious and aggressive 111703 - 8 / Vol . 145 , NOVEMBER 2023 Transactions of the ASME D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 choices about what points to sample next , so while the diversely ini - tialized models might get a head start on exploring the space , non - diversely initialized models needed to explore less of the space overall , owing to tighter posteriors of possible functions under the Gaussian process . While we do not have space to include it in the main paper , the Supplementary Material document \u2019 s Sec . 5 shows how this model building advantage occurs as we provide BO with a greater number of initial samples . Brie \ufb02 y , there are three \u201c regimes \u201d : ( 1 ) sample - de \ufb01 cient , where there are too few samples to induce a mod - eling advantage regardless of how diversely we sample the initial points ; ( 2 ) the \u201c modeling advantage \u201d region , where low - diversity samples induce better hyper - parameter convergence than high - diversity samples ; and ( 3 ) sample - saturated , where there are enough initial samples to induce accurate hyper - parameter posteri - ors regardless of how diversely we sample initial points . We direct interested readers to Sec . 5 of the Supplementary Material for a deeper discussion on this . What this behavior implies more broadly is that non - diverse samples , whether given to an algorithm or a person , have a unique and perhaps underrated value in cases where we have high entropy priors over the Gaussian process hyper - parameters or kernel . In such cases , sacri \ufb01 cing a few initial non - diverse points to better infer key length scales in the GP model may well be a worthwhile trade . We also saw that in cases where the BO hyper - parameters were not further optimized ( as in experiment 3 where hyper - parameters were \ufb01 xed to optimal values ) , using diverse points only helped BO . Researchers or practitioners using BO would bene \ufb01 t from care - fully reviewing what kernel optimization strategy their library or implementation of choice actually does since that will affect whether or not the model building advantage of non - diverse samples is actually in play . What if hyper - parameters are \ufb01 xed to non - optimal values ? We showed in experiment 3 that \ufb01 xing BO hyper - parameters to their optimal values ahead of time using an oracle allowed diverse initial samples to unilaterally outperform non - diverse samples . An interesting avenue of future work that we did not explore here for scope reasons would be to see if this holds when hyper - parameters are \ufb01 xed to non - optimal values . In practical problems , we will not often know the optimal hyper - parameters ahead of time as we did in experiment 3 which caused diversity \u2019 s unilateral advantage , so we do not have evidence to generalize beyond this . However , our explanation of the model building advantage would predict that , so long as the hyper - parameters remain \ufb01 xed ( to any value ) , BO would not have a practical mechanism to bene \ufb01 t much from non - diverse samples , on average . What are theimplications for how we currently initialize BO ? One of our result \u2019 s most striking implications is how it might in \ufb02 uence BO initialization procedures that are often considered standard prac - tice . For example , it is common to initialize a BO procedure with a small number of initial space - \ufb01 lling designs , using techniques like LHS before allowing BO to optimize its acquisition function for future samples . In cases where the BO hyper - parameters will remain \ufb01 xed , experiment 3 implies that this standard practice is excellent advice and far better than non - diverse samples . However , in cases where you plan to optimize the BO kernel during search , using something like LHS becomes more suspect . In principle , from experiment 1 , we see that diverse samples may help or hurt BO , depending on how much leverage the model build - ing advantage of the non - diverse samples can provide . For example , in the upper right of Fig . 4 , the function is effectively random noise , and so there is not a strong model building advantage to be gained . In contrast , in the lower left , the smooth and well - behaved functions allowed non - diverse initialization to gain an upper hand . Our results propose a perhaps now obvious initialization strategy : if you plan on optimizing the BO hyper - parameters , use some non - diverse samples to strategically provide an early model building advantage , while leveraging the rest of the samples to diversely cover the space . Fig . 6 Experiment 3 : optimality gap plot showing effects of diversity when the optimizer is not allowed to \ufb01 t the hyper - parameters for the Gaussian process and the hyper - parameters are instead \ufb01 xed to the values found in experiment 2 . The results from this plot show positive NCOG values for all families of wildcat wells function , showing that once the model building advantage is taken away , the diverse samples outperform non - diverse samples . Further dis - cussion on this plot can be read in Sec . 5 . Journal of Mechanical Design NOVEMBER 2023 , Vol . 145 / 111703 - 9 D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 How might other acquisition functions modulate diversity \u2019 s effect ? While we have been referring to BO as though it is a single method throughout this paper , individual BO implementations can vary , both in terms of their kernel structure and their choice of acqui - sition function \u2014 or how BO uses information about the underlying \ufb01 tted Gaussian process to select subsequent points . In this paper \u2019 s experiments , we used EI since it is one of the most widespread choices and behaves qualitatively like other common improvement - based measures like probability of improvement , posterior mean , and upper con \ufb01 dence bound functions . Indeed , we hypothesize that part of the reason non - diverse initial samples are able to gain a model building advantage over diverse samples due to a faster col - lapse in the posterior distribution of possible GP functions which serves as strong input to EI methods and related variants . Yet EI and its cousins are only one class of acquisition function ; would our results hold if we were to pick an acquisition function that directly attacked the GP \u2019 s posterior variance ? For example , either entropy - based or active learning - based acquisition functions ? This paper did not test this and it would be a logical and valuable future study . Our experimental results and proposed explanation would predict the following : the model building advantage seen by non - diverse samples should reduce or disappear in cases where the acquisition function explicitly samples new points to minimize the posterior over GP function classes since in such cases BO itself would try to select samples that reduced overall GP variance , reducing its dependence on what the initial samples provide . To what extent should we expect these results to generalize to other types of problems ? We selected a simple 2D function with controllable complexity in this paper to aid in experimental simpli - city , speed , replicability , and ease of visualization ; however , this does raise the question of whether or not these results would truly transfer to more complex problems of engineering interest . While future work would have to address more complex problems , we performed two additional experiments studying how the above phenomena change as we ( 1 ) increased the wildcat wells function from two to three dimensions and ( 2 ) how this behavior changes for other types of common optimization test functions \u2014 speci \ufb01 - cally , we chose the N - dimensional sphere , Rastrigin , and Rosen - brock functions from two to \ufb01 ve dimensions . While the existing paper length limits did not allow us to include all of these addi - tional results in the paper \u2019 s main body , we direct interested readers to Secs . 6 and 7 of the Supplementary Material document . Brie \ufb02 y , our results align overall with what we described above for the 2D wildcat wells function , and we do not believe that the phe - nomena we observed are restricted to only our chosen test function or dimension , although clearly future research would need to conduct further tests on other problems to say this with any cer - tainty . Beyond these supplemental results , we can also look at a few critical problem - speci \ufb01 c factors and ask what our proposed explanatory model would predict . For higher dimensional problems , standard GP kernel choices like RBF or Mat\u00e9rn begin to face exponential cost increases due to how hyper - volumes expand . In such cases , having strong con - straints ( via hyper - parameter priors or posteriors ) over possible GP functions becomes increasingly important for fast BO conver - gence . Our results would posit that any model building advantages from non - diverse sampling would become increasingly important or impactful in cases where it helped BO rapidly collapse the hyper - parameter posteriors . For discontinuous functions ( or GP kernels that assumed as much ) , the model building advantage of non - diverse samples would decrease since large sudden jumps in the GP posterior mean and variance would make it harder for BO to exploit a model building advantage . However , in discontinuous cases where there were still common global smoothness parameters that governed the continuous portions , the model building advantage would still accelerate advantages for BO convergence . How might theresultsguide human - subject experimentsor under - standing of human designers ? One possible implication of our results for human designers is that the effects of example diversity on design outcomes may vary as a function of the designer \u2019 s prior knowledge of the design problem . More speci \ufb01 cally , the model building advantage observed in experiment 2 ( and subsequent removal in experiment 3 ) suggests that when designers have prior knowledge of how quickly the function changes in a local area of thedesignspace , theycanmorereliablybene \ufb01 tfromthespaceexplo - ration advantage of diverse examples . This leads to a potentially counter - intuitive prediction that domain experts may bene \ufb01 t more from diverse examples compared to domain novices since domain experts would tend to have prior knowledge of the nature of the design problem ( a model building advantage ) . Additionally , perhaps under conditions of uncertainty about the nature of the design problem , it would be useful to combine the strengths of diverse and non - diverse examples ; this could be accomplished with a cluster - sampling approach , where we sample diverse points of the design space , but include local non - diverse clusters of exam - ples that are nearby , to facilitate learning of the shape of the design function . While these implications might be counter - intuitive in that common guidance suggests that the most informative method is to only diversely sample initial points , the crux of our paper \u2019 s argu - ment is that non - diverse points can , surprisingly , be informative to Bayesian optimization due to their ability to quickly concentrate the posterior distribution of the kernel hyper - parameters , and thus accelerate later optimization . Given this tension , a natural question is \u201c how many non - diverse samples do I really need to take advan - tage of the modeling advantage without giving up the space explo - ration advantage ? \u201d If I have , say , a budget of ten experiments , should I spend only one low - diversity sample ? Or do I need two ? Half of my budget ? We did not explore these practical questions in this work , due to space constraints , but we think this would be an excellent avenue for continued study . Acknowledgment This research was supported by funding from the National Science Foundation through award # 1826083 . Con \ufb02 ict of Interest There are no con \ufb02 icts of interest . Data Availability Statement The data and information that support the \ufb01 ndings of this article are freely available . 4 References [ 1 ] Sio , U . N . , Kotovsky , K . , and Cagan , J . , 2015 , \u201c Fixation or Inspiration ? A Meta - Analytic Review of the Role of Examples on Design Processes , \u201d Des . Stud . , 39 , pp . 70 \u2013 99 . [ 2 ] Crilly , N . , and Cardoso , C . , 2017 , \u201c Where Next for Research on Fixation , Inspiration and Creativity in Design ? , \u201d Des . Stud . , 50 , pp . 1 \u2013 38 . [ 3 ] Baruah , J . , and Paulus , P . B . , 2011 , \u201c Category Assignment and Relatedness in the Group Ideation Process , \u201d J . Exp . Social Psychol . , 47 ( 6 ) , pp . 1070 \u2013 1077 . [ 4 ] Siangliulue , P . , Arnold , K . C . , Gajos , K . Z . , and Dow , S . P . , 2015 , \u201c Toward Collaborative Ideation at Scale : Leveraging Ideas From Others to Generate More Creative and Diverse Ideas , \u201d Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing , CSCW \u2019 15 , Vancouver , BC , Canada , Mar . 14 \u2013 18 , ACM , pp . 937 \u2013 945 . [ 5 ] Nijstad , B . A . , Stroebe , W . , and Lodewijkx , H . F . M . , 2002 , \u201c Cognitive Stimulation and Interference in Groups : Exposure Effects in an Idea Generation Task , \u201d J . Exp . Social Psychol . , 38 ( 6 ) , pp . 535 \u2013 544 . [ 6 ] Taylor , A . , and Greve , H . R . , 2006 , \u201c Superman or the Fantastic Four ? Knowledge Combination and Experience in Innovative Teams , \u201d Acad . Manage . J . , 49 ( 4 ) , pp . 723 \u2013 740 . 4 See Note 3 . 111703 - 10 / Vol . 145 , NOVEMBER 2023 Transactions of the ASME D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024 [ 7 ] Zeng , L . , Proctor , R . W . , andSalvendy , G . , 2011 , \u201c FosteringCreativityinProduct andService Development : ValidationintheDomain of InformationTechnology , \u201d Human Factors , 53 ( 3 ) , pp . 245 \u2013 270 . [ 8 ] Howard - Jones , P . A . , Blakemore , S . - J . J . , Samuel , E . A . , Summers , I . R . , and Claxton , G . , 2005 , \u201c Semantic Divergence and Creative Story Generation : An FMRI Investigation , \u201d Cognit . Brain Res . , 25 ( 1 ) , pp . 240 \u2013 250 . [ 9 ] Chan , J . , and Schunn , C . D . , 2015 , \u201c The Importance of Iteration in Creative Conceptual Combination , \u201d Cognition , 145 , pp . 104 \u2013 115 . [ 10 ] Gielnik , M . M . , Frese , M . , Graf , J . M . , and Kampschulte , A . , 2011 , \u201c Creativity in the Opportunity Identi \ufb01 cation Process and the Moderating Effect of Diversity of Information , \u201d J . Bus . Ventur . , 27 ( 5 ) , pp . 559 \u2013 576 . [ 11 ] Althuizen , N . , and Wierenga , B . , 2014 , \u201c Supporting Creative Problem Solving WithaCase - BasedReasoningSystem , \u201d J . Manage . Inf . Syst . , 31 ( 1 ) , pp . 309 \u2013 340 . [ 12 ] Yuan , H . , Lu , K . , Jing , M . , Yang , C . , and Hao , N . , 2022 , \u201c Examples in Creative Exhaustion : The Role of Example Features and Individual Differences in Creativity , \u201d Personality Individual Diff . , 189 , p . 111473 . [ 13 ] Doboli , A . , Umbarkar , A . , Subramanian , V . , and Doboli , S . , 2014 , \u201c Two Experimental Studies on Creative Concept Combinations in Modular Design of Electronic Embedded Systems , \u201d Des . Stud . , 35 ( 1 ) , pp . 80 \u2013 109 . [ 14 ] Jang , S . , 2014 , \u201c The Effect of Image Stimulus on Conceptual Combination in the Design Idea Generation Process , \u201d Arch . Des . Res . , 112 ( 4 ) , p . 59 . [ 15 ] Mobley , M . I . , Doares , L . M . , and Mumford , M . D . , 1992 , \u201c Process Analytic Models of Creative Capacities : Evidence for the Combination and Reorganization Process , \u201d Creativity Res . J . , 5 ( 2 ) , pp . 125 \u2013 155 . [ 16 ] Baughman , W . A . , and Mumford , M . D . , 1995 , \u201c Process - Analytic Models of Creative Capacities : Operations In \ufb02 uencing the Combination - and - Reorganization Process , \u201d Creativity Res . J . , 8 ( 1 ) , pp . 37 \u2013 62 . [ 17 ] Kamath , C . , 2021 , \u201c Intelligent Sampling for Surrogate Modeling , Hyperparameter Optimization , and Data Analysis , \u201d Tech . Rep . LLNL - TR - 829837 , Lawrence Livermore National Lab . ( LLNL ) , Livermore , CA , Dec . [ 18 ] Yang , X . - S . , 2014 , \u201c Swarm Intelligence Based Algorithms : A Critical Analysis , \u201d Evol . Intel . , 7 ( 1 ) , pp . 17 \u2013 28 . [ 19 ] Ma , Z . , and Vandenbosch , G . A . E . , 2012 , \u201c Impact of Random Number Generators on the Performance of Particle Swarm Optimization in Antenna Design , \u201d 2012 6th European Conference on Antennas and Propagation ( EUCAP ) , Prague , Czech Republic , Mar . 26 \u2013 30 , IEEE , pp . 925 \u2013 929 . [ 20 ] Kazimipour , B . , Li , X . , and Qin , A . K . , 2014 , \u201c Effects of Population Initialization on Differential Evolution for Large Scale Optimization , \u201d 2014 IEEE Congress on Evolutionary Computation ( CEC ) , Beijing , China , July 6 \u2013 11 , pp . 2404 \u2013 2411 . [ 21 ] Maaranen , H . , Miettinen , K . , and M\u00e4kel\u00e4 , M . M . , 2004 , \u201c Quasi - random Initial Population for Genetic Algorithms , \u201d Comput . Math . Appl . , 47 ( 12 ) , pp . 1885 \u2013 1895 . [ 22 ] Li , Q . , Liu , S . - Y . , and Yang , X . - S . , 2020 , \u201c In \ufb02 uence of Initialization on the Performance of Metaheuristic Optimizers , \u201d Appl . Soft Comput . , 91 , p . 106193 . [ 23 ] Deb , K . , Pratap , A . , Agarwal , S . , and Meyarivan , T . , 2002 , \u201c A Fast and Elitist Multiobjective Genetic Algorithm : NSGA - II , \u201d IEEE Trans . Evol . Comput . , 6 ( 2 ) , pp . 182 \u2013 197 . [ 24 ] Shu , L . , Jiang , P . , Shao , X . , and Wang , Y . , 2020 , \u201c A New Multi - Objective Bayesian Optimization Formulation With the Acquisition Function for Convergence and Diversity , \u201d ASME J . Mech . Des . , 142 ( 9 ) , p . 091703 . [ 25 ] Simpson , T . , Poplinski , J . , Koch , P . N . , and Allen , J . , 2001 , \u201c Metamodels for Computer - Based Engineering Design : Survey and Recommendations , \u201d Eng . Comput . , 17 ( 2 ) , pp . 129 \u2013 150 . [ 26 ] Queipo , N . V . , Haftka , R . T . , Shyy , W . , Goel , T . , Vaidyanathan , R . , and Kevin Tucker , P . , 2005 , \u201c Surrogate - Based Analysis and Optimization , \u201d Prog . Aerosp . Sci . , 41 ( 1 ) , pp . 1 \u2013 28 . [ 27 ] Jin , R . , Chen , W . , and Simpson , T . , 2001 , \u201c Comparative Studies of Metamodelling Techniques Under Multiple Modelling Criteria , \u201d Struct . Multidiscipl . Optim . , 23 ( 1 ) , pp . 1 \u2013 13 . [ 28 ] Sexton , T . , and Ren , M . Y . , 2017 , \u201c Learning an Optimization Algorithm Through Human Design Iterations , \u201d ASME J . Mech . Des . , 139 ( 10 ) , p . 101404 . [ 29 ] Tauber , S . , Navarro , D . J . , Perfors , A . , and Steyvers , M . , 2017 , \u201c Bayesian Models of Cognition Revisited : Setting Optimality Aside and Letting Data Drive Psychological Theory , \u201d Psychol . Rev . , 124 ( 4 ) , pp . 410 \u2013 441 . [ 30 ] Kemp , C . , and Tenenbaum , J . B . , 2008 , \u201c The Discovery of Structural Form , \u201d Proc . Natl . Acad . Sci . U . S . A . , 105 ( 31 ) , pp . 10687 \u2013 10692 . [ 31 ] Lu , H . , Yuille , A . L . , Liljeholm , M . , Cheng , P . W . , and Holyoak , K . J . , 2008 , \u201c Bayesian Generic Priors for Causal Learning , \u201d Psychol . Rev . , 115 ( 4 ) , pp . 955 \u2013 984 . [ 32 ] Lu , H . , Chen , D . , and Holyoak , K . J . , 2012 , \u201c Bayesian Analogy With Relational Transformations , \u201d Psychol . Rev . , 119 ( 3 ) , pp . 617 \u2013 648 . [ 33 ] Fuge , M . , Stroud , J . , and Agogino , A . , 2013 , \u201c Automatically Inferring Metrics for Design Creativity , \u201d IDETC - CIE2013 , Volume 5 : 25th International Conference on Design Theory and Methodology , Portland , OR , Aug 4 \u2013 7 , American Society of Mechanical Engineers , p . V005T06A010 . [ 34 ] Ahmed , F . , Ramachandran , S . K . , Fuge , M . , Hunter , S . , and Miller , S . , 2021 , \u201c Design Variety Measurement Using Sharma \u2013 Mittal Entropy , \u201d ASME J . Mech . Des . , 143 ( 6 ) , p . 061702 . [ 35 ] Miller , S . R . , Hunter , S . T . , Starkey , E . , Ramachandran , S . , Ahmed , F . , and Fuge , M . , 2021 , \u201c How Should We Measure Creativity in Engineering Design ? A Comparison Between Social Science and Engineering Approaches , \u201d ASME J . Mech . Des . , 143 ( 3 ) , p . 031404 . [ 36 ] Ahmed , F . , Ramachandran , S . K . , Fuge , M . , Hunter , S . , and Miller , S . , 2019 , \u201c Interpreting Idea Maps : Pairwise Comparisons Reveal What Makes Ideas Novel , \u201d ASME J . Mech . Des . , 141 ( 2 ) , p . 021102 . [ 37 ] Ahmed , F . , and Fuge , M . , 2018 , \u201c Ranking Ideas for Diversity and Quality , \u201d ASME J . Mech . Des . , 140 ( 1 ) , p . 011101 . [ 38 ] Ahmed , F . , 2019 , \u201c Diversity and Novelty : Measurement , Learning and Optimization , \u201d Ph . D . thesis , University of Maryland , College Park , MD . [ 39 ] Kulesza , A . , and Taskar , B . , 2012 , \u201c Determinantal Point Processes for Machine Learning , \u201d Found . Trends\u00ae Mach . Learn . , 5 ( 2 \u2013 3 ) , pp . 123 \u2013 286 . [ 40 ] Li , C . , Chu , X . , Chen , Y . , and Xing , L . , 2015 , \u201c A Knowledge - Based Initialization Technique of Genetic Algorithm for the Travelling Salesman Problem , \u201d 2015 11th International Conference on Natural Computation ( ICNC ) , Zhangjiajie , China , Aug . 15 \u2013 17 , pp . 188 \u2013 193 . [ 41 ] Dong , N . , Wu , C . - H . , Ip , W . - H . , Chen , Z . - Q . , Chan , C . - Y . , and Yung , K . - L . , 2012 , \u201c An Opposition - Based Chaotic GA / PSO Hybrid Algorithm and Its Application in Circle Detection , \u201d Comput . Math . Appl . , 64 ( 6 ) , pp . 1886 \u2013 1902 . [ 42 ] Eskandar , H . , Sadollah , A . , Bahreininejad , A . , and Hamdi , M . , 2012 , \u201c Water Cycle Algorithm \u2014 A Novel Metaheuristic Optimization Method for Solving Constrained Engineering Optimization Problems , \u201d Comput . Struct . , 110 \u2013 111 , pp . 151 \u2013 166 . [ 43 ] Mishkin , D . , and Matas , J . , 2016 , \u201c All YouNeed Is aGood Init , \u201d 4thInternational Conference on Learning Representations , San Juan , Puerto Rico , May 2 \u2013 4 . [ 44 ] Yuan , W . , Han , Y . , Guan , D . , Lee , S . , andLee , Y . - K . , 2011 , \u201c InitialTrainingData Selection for Active Learning , \u201d Proceedings of the 5th International Conference on Ubiquitous Information Management and Communication , ICUIMC \u2019 11 , Seoul , South Korea , Feb . 21 \u2013 23 , Association for Computing Machinery , pp . 1 \u2013 7 . [ 45 ] Settles , B . , 2012 , Active Learning ( Synthesis Lectures on Arti \ufb01 cial Intelligence and Machine Learning ) , Springer International Publishing , Cham . [ 46 ] Yoon , J . , Arik , S . , and P \ufb01 ster , T . , 2020 , \u201c Data Valuation Using Reinforcement Learning , \u201d Proceedings of the 37th International Conference on Machine Learning , Vienna , Austria , July 13 \u2013 18 , PMLR , pp . 10842 \u2013 10851 . [ 47 ] Eysenbach , B . , Gupta , A . , Ibarz , J . , and Levine , S . , 2019 , \u201c Diversity Is All You Need : Learning Skills Without a Reward Function , \u201d Seventh International Conference on Learning Representations , New Orleans , LA , May 6 \u2013 9 . [ 48 ] Sch\u00f6lkopf , B . , and Smola , A . J . , 2018 , Learning with Kernels : Support Vector Machines , Regularization , Optimization , and Beyond , The MIT Press , Cambridge , MA . [ 49 ] Kulesza , A . , and Taskar , B . , 2011 , \u201c k - DPPs : Fixed - Size Determinantal Point Processes , \u201d Proceedings of the 28th International Conference on International Conference on Machine Learning , ICML \u2019 11 , Bellevue , WA , June 28 \u2013 July 2 , pp . 1193 \u2013 1200 . [ 50 ] Calandriello , D . , Derezinski , M . , and Valko , M . , 2020 , \u201c Sampling From a K - DPP Without Looking at All Items , \u201d 34th Conference on Neural Information Processing Systems , Vancouver , Canada , Dec . 6 \u2013 12 , Curran Associates , Inc . , pp . 6889 \u2013 6899 . [ 51 ] Li , C . , Jegelka , S . , and Sra , S . , 2016 , \u201c Ef \ufb01 cient Sampling for K - Determinantal Point Processes , \u201d Proceedings of the 19th International Conference on Arti \ufb01 cial Intelligence and Statistics , Cadiz , Spain , May 9 \u2013 11 . [ 52 ] Mariet , Z . E . , 2016 , \u201c Learning and Enforcing Diversity With Determinantal Point Processes , \u201d Master \u2019 s thesis , Massachusetts Institute of Technology , Boston , MA . [ 53 ] Hansen , N . , Finck , S . , Ros , R . , and Auger , A . , 2009 , \u201c Real - Parameter Black - Box Optimization Benchmarking 2009 : Noisy Functions De \ufb01 nitions , \u201d Research Report RR - 6869 , INRIA . [ 54 ] R\u00f6nkk\u00f6nen , J . , Li , X . , Kyrki , V . , and Lampinen , J . , 2011 , \u201c A Framework for Generating Tunable Test Functions for Multimodal Optimization , \u201d Soft Comput . , 15 ( 9 ) , pp . 1689 \u2013 1706 . [ 55 ] Mo , H . , Li , Z . , and Zhu , C . , 2017 , \u201c Epistasis - Tunable Test Functions With Known Maximum Constructed With Sinusoidal Bases , \u201d 2017 12th International Conference on Intelligent Systems and Knowledge Engineering ( ISKE ) , Jiangsu , China , Nov . 24 \u2013 26 , pp . 1 \u2013 6 . [ 56 ] Mason , W . , and Watts , D . J . , 2012 , \u201c Collaborative Learning in Networks , \u201d Proc . Natl . Acad . Sci . U . S . A . , 109 ( 3 ) , pp . 764 \u2013 769 . [ 57 ] Balandat , M . , Karrer , B . , Jiang , D . , Daulton , S . , Letham , B . , Wilson , A . G . , and Bakshy , E . , 2020 , \u201c BoTorch : A Framework for Ef \ufb01 cient Monte - Carlo Bayesian Optimization , \u201d 34th Conference on Neural Information Processing Systems , Vancouver , Canada , Dec . 6 \u2013 12 , Vol . 33 , Curran Associates , Inc . , pp . 21524 \u2013 21538 . Journal of Mechanical Design NOVEMBER 2023 , Vol . 145 / 111703 - 11 D o w n l oaded f r o m h tt p : / / a s m ed i g i t a l c o ll e c t i on . a s m e . o r g / m e c han i c a l de s i gn / a r t i c l e - pd f / 145 / 11 / 111703 / 7037097 / m d _ 145 _ 11 _ 111703 . pd f b y U n i v e r s i t y o f M a r y l and L i b r a r i e s u s e r on 03 M a y 2024",
    "chanFormulatingFixatingEffects2024": "Formulating or Fixating : Effects of Examples on Problem Solving Vary as a Function of Example Presentation Interface Design JOEL CHAN , College of Information Studies , University of Maryland , USA ZIJIAN DING , College of Information Studies , University of Maryland , USA EESH KAMRAH , Department of Mechanical Engineering , University of Maryland , USA MARK FUGE , Department of Mechanical Engineering , University of Maryland , USA Interactive systems that facilitate exposure to examples can augment problem solving performance . However designers of such systems are often faced with many practical design decisions about how users will interact with examples , with little clear theoretical guidance . To understand how example interaction design choices affect whether / how people benefit from examples , we conducted an experiment where 182 participants worked on a controlled analog to an exploratory creativity task , with access to examples of varying diversity and presentation interfaces . Task performance was worse when examples were presented in a list , compared to contextualized in the exploration space or shown in a dropdown list . Example lists were associated with more fixation , whereas contextualized examples were associated with using examples to formulate a model of the problem space to guide exploration . We discuss implications of these results for a theoretical framework that maps design choices to fundamental psychological mechanisms of creative inspiration from examples . Additional Key Words and Phrases : Creativity , Examples , Interface , Problem Solving ACM Reference Format : Joel Chan , Zijian Ding , Eesh Kamrah , and Mark Fuge . 2023 . Formulating or Fixating : Effects of Examples on Problem Solving Vary as a Function of Example Presentation Interface Design . 1 , 1 ( January 2023 ) , 25 pages . https : / / doi . org / 10 . 1145 / 1122445 . 1122456 1 INTRODUCTION Examples \u2014 descriptions or representations of possible solutions ( or parts thereof ) for the same or related problems [ 8 , 18 , 35 , 83 , 84 ] \u2014 are an integral part of the creative problem solving process . Examples can take many forms , such as previous physical prototypes brought to a brainstorm [ 33 ] , search results for patents for related problems [ 25 ] , spoken ideas from collaborators [ 66 ] , UI designs in an accessible gallery [ 50 ] ) , or references from memory to earth animals or science fiction creatures when inventing new fictional alien creatures [ 89 ] ) . Importantly , examples can substantially shape what ideas come to mind [ 84 , 89 ] . This \" structuring of imagination \" [ 89 ] is sometimes helpful \u201c inspiration \" that leads to more creative ideas [ 19 , 23 , 33 , 80 ] . But examples can also have harmful \u201c fixation \u201d effects that constrain novelty and innovation [ 40 , 54 , 84 , 89 ] ( for recent reviews , see [ 83 ] and [ 18 ] ) . Importantly , examples can influence problem solving without conscious effort or recognition [ 54 , 60 , 61 ] , and persist in spite of creators\u2019 explicit intentions not to be influenced by them [ 40 , 84 , 85 ] . Perhaps in recognition of these facts , effective creators take an active role Authors\u2019 addresses : Joel Chan , College of Information Studies , University of Maryland , USA ; Zijian Ding , , College of Information Studies , University of Maryland , USA ; Eesh Kamrah , Department of Mechanical Engineering , University of Maryland , USA ; Mark Fuge , Department of Mechanical Engineering , University of Maryland , USA . Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page . Copyrights for components of this work owned by others than ACM must be honored . Abstracting with credit is permitted . To copy otherwise , or republish , to post on servers or to redistribute to lists , requires prior specific permission and / or a fee . Request permissions from permissions @ acm . org . \u00a9 2023 Association for Computing Machinery . Manuscript submitted to ACM Manuscript submitted to ACM 1 a r X i v : 2401 . 11022v2 [ c s . H C ] 23 J a n 2024 2 Chan , et al . in finding , structuring and interacting with examples [ 23 , 35 , 79 ] using a variety of analog and digitally mediated systems and practices , such as search engines [ 35 , 68 , 69 , 97 ] , design workbooks and commonplace notebooks [ 26 ] , online whiteboards [ 58 , 86 ] , mood boards [ 55 ] , and wider interactions with their community of practice , such as trade publications and conventions [ 23 , 33 ] . An important area of HCI research on creativity support tools therefore studies the design of interactive systems that can assist creators in discovering examples [ 9 , 39 , 47 , 48 , 74 , 80 , 82 , 87 , 88 ] , structuring , analyzing , and exploring collections of examples [ 15 , 16 , 53 , 57 , 81 , 93 ] , and adapting and using examples [ 39 , 48 , 50 ] . Designers of such systems need to grapple with an array of very practical interaction design decisions . For example , how should we support interaction with examples over different screen sizes ? Should examples be delivered via recommendation ( in small sets ) , a feed , or some other interaction paradigm ? What information should be presented alongside an example ? We would like to have a consistent theory to draw from to make these decisions . Beyond considerations of usability , we conjecture that such a theory would need to map design decisions ( or classes thereof ) to creativity - relevant behaviors and outcomes , ideally with a nuanced specification of the precise benefits and costs of each design decision for these behaviors and outcomes . A theory of human - example interaction like this could help us design better systems with sensible defaults , prioritize and negotiate design requirements , and guide evaluation . As a step towards developing such a theory , we conducted an experiment with 182 participants solving a controlled analog to an exploratory creativity task [ 7 ] . We varied both the diversity of examples and the types of presentations : overlaid on the search environment ( the \u201c In - Context \" condition ) , presented in a list ( the \u201c List \" condition ) , or in a dropdown selectable menu ( the \u201c Dropdown \" condition ) . The \u201c In - Context \" design was inspired by an emerging pattern of contextualizing examples in the creator\u2019s workspace or problem in HCI systems for example - based creativity [ 39 , 47 , 78 , 91 , 93 ] on the one hand , and theoretical descriptions of the use of examples to ( re ) formulate problems [ 34 , 35 , 43 , 59 , 67 , 79 ] ; the latter \u201c List \" and \u201c Dropdown \" conditions were designed to be representative of common interfaces for interacting with examples ( in search results lists and pages of recommendations ) . Our primary results were threefold : 1 ) \u201c List \" presentation harmed solution quality compared with \u201c In - Context \" or \u201c Dropdown \" presentation ; 2 ) each interface condition was associated with distinct self - reported example usage strategies ( notably , more usage of examples to \u201cmodel \" the problem space to guide exploration in the In - Context vs . List or Dropdown conditions , and more usage of examples to \u201cstimulate \" a specific direction of exploration in the List condition ) ; and 3 ) the List condition\u2019s propensity for stimulation - based strategies was corroborated by an increased usage of \u201chill - climbing \" strategies early on , as evidenced by analyses of Dropdown distance between participants\u2019 moves . We discuss how these results , in conversation with the literature on example - based creativity support systems as well as psychological mechanisms of creativity with examples , could contribute to a theoretical framework for designing interactive systems for creative problem solving with examples . 2 RELATED WORK 2 . 1 Sources of empirical variability in effects of examples Prior work has examined how the consequences of examples for creative problem solving outcomes are related to characteristics of the examples , such as their novelty [ 1 , 6 , 11 , 80 ] ( generally positive effects ) , conceptual distance from the problem domain [ 10 , 19 , 25 , 31 , 90 ] ( mixed or curvilinear effects ) , and example diversity [ 12 , 21 , 29 , 38 , 80 , 94 , 96 ] ( generally positive or contingently positive effects ) . Our work contributes additional empirical results on the relative contributions of ( and potential interactions , in the statistical sense , between ) example characteristics and example Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 3 interface characteristics . In particular , we explore how the example characteristic of diversity might interact with example interface characteristics , such as whether the examples are presented as a list vs . in context of a representation of the design space . To do this , we need to also consider the cognitive mechanisms of inspiration or fixation from examples ( or varying characteristics ) , which might be more or less afforded by example interfaces . We discuss this body of literature in the next section . 2 . 2 Theoretical insights into human - example interaction A number of detailed in - situ studies of creators have documented a range of strategies for working with examples , ranging from simpler , more source - driven strategies like direct source adaptation [ 24 ] , to more complex and reflective strategies associated with more radical transformation of examples , such as source analysis and schema - driven source selection [ 24 ] , analogical reasoning [ 3 , 27 , 28 , 37 ] , and generating novel emergent features that can connect disparate attributes across examples [ 92 ] . These \u201cprocessing strategies \" can be described by a variety of theoretical frames from the psychological literature on creativity . We believe this theoretical level of description could facilitate our goal of synthesizing mappings between interface characteristics and effects of examples on creative problem solving outcomes . Some notable examples include basic memory mechanisms such as priming [ 63 ] and spreading activation [ 17 , 73 ] , and higher level cognitive processes such as conceptual abstraction and analogical transfer [ 20 , 28 ] , conceptual combination [ 92 ] , and problem reformulation based on examples [ 22 , 34 , 57 ] . Of particular interest in our study is a contrast between priming and spreading activation mechanisms on the one hand , which is associated with lower - level conceptual influences , and problem ( re ) formulation , which is associated with more complex , higher - order processing of examples . In this study , we extend this literature by exploring how two specific mechanisms of processing examples ( stimulation , and ( re ) formulation ) might be helped or hindered by different example interaction interfaces . To set the context for our results , we briefly review the literature on each mechanism here . 2 . 2 . 1 Using examples to stimulate ideation . Spreading activation has been invoked to explain the impact of external stimuli on ideation . For instance , the \u201csearch for ideas in associative memory \" ( SIAM ) model [ 66 ] proposes that when ideas come to mind , whether from memory , or through discussion with others or exposure to examples , they also raise the activation level of other associated concepts and features in memory , which can stimulate or inhibit ideation by making certain sets of ideas more or less likely to be generated based on the current network of associations in memory . For example , an example idea \u201cuse as paperweight \" ( for a design prompt to generate alternative use for a brick ) may activate related concepts such as \u201coffice \" , or \u201cis heavy \" , along with their associated concepts ; subsequent ideas such as \u201cconstruct a table \" , or \u201cprop up a bookshelf \" may then be more likely to come to mind , compared to ideas like \u201cuse as a weapon \" or \u201cmakeshift goalposts for soccer \" . In this way , exposure to examples can shape the trajectory of ideation , and the corresponding floor or ceiling of creativity [ 6 , 13 ] . In this paper , we discuss this set of mechanisms under the label \u201c stimulation \" , to capture the intuition of examples stimulating ideation along a particular direction . 2 . 2 . 2 Using examples to ( re ) formulate problems . Past research has documented how people can use examples to construct , refine , and even reformulate their understanding of the creative problem they are trying to solve [ 34 , 35 , 43 , 59 , 67 , 79 ] , through processes such as intentional free - form curation of examples [ 56 ] or on mood boards [ 55 ] . For instance , Okada et al . documented how two artists used individual artworks to shape not just their ideas , but used a process of \u201canalogical modification \" to search for and modify higher - order concepts , and their creative vision over the course of years [ 67 ] . This process of example - influenced problem formulation is related to computational and neurobiological models of cognitive search [ 36 ] which study the range of strategies that intelligent agents can use to Manuscript submitted to ACM 4 Chan , et al . structure their search processes : here , there is an important distinction between \u201cmodel - free \" search , where external feedback from the world on the agents\u2019 actions guide search in a simpler , more local , stimulus - response manner , and \u201cmodel - based \" search , where the agent constructs a model or representation of the task and environment ( in the case of creative problem - solving , this would be the problem and / or design space [ 22 , 30 , 65 ] ) partly on the basis of reflection its own actions and possibly observation of others\u2019 actions , and uses that model to decide where and when to explore in the task environment vs . continue to sample locally . Additionally , insight problem solving research has documented how people not only construct models , but also substantially revise them in radical ways , to solve difficult creative problems [ 43 , 45 ] ; this process is a key source of difficulty for creative problems , where one\u2019s initial problem formulation ( e . g . , key constraints or requirements ) , may be unhelpful [ 45 ] . In this paper , we discuss this set of mechanisms under the label \u201c ( re ) formulation \" , to capture the intuition of creators leveraging examples to ( re ) formulate their understanding of the design space . 2 . 3 Interactive systems for interacting with examples In this study , we are interested in understanding how the impact of examples on problem solving varies as a function of interaction design decisions for how creators will interact with examples . To set the context , we briefly review here some emerging interaction design patterns in HCI systems research into how ( vs . when , as in recommendation systems ) participants interact with sets of examples ( vs . understand or modify a single example ) . One higher - level design pattern involves explorable overviews of examples . For example , MetaMap [ 42 ] supports exploration of examples through keywords and colors and offering a playground to curate examples ; RecipeScape [ 16 ] uses a map UI to present recipe examples ; Sifter [ 70 ] presents large collections of image manipulation tutorials in a faceted view based on their command - level structure ; the Adaptive Ideas Web tool [ 53 ] enables designers to explore and structure collections of web design examples ; the Freed system [ 64 ] empowers design students to spatially organize their digital collection of examples , define relations and reflect on their interrelationships ; and Cabinet [ 44 ] supports collecting and organizing of visual examples for inspiration and reference . Another emerging design pattern can be described as contextualizing examples in the creator\u2019s workspace or problem , enabling designers to curate and reflect on the examples to build an understanding of their design space . For example , ReflectionSpace [ 78 ] interactively contextualizes design artifacts in project timelines ( and associated comments and reflections ) to promote reflection and learning ; MoodCubes [ 39 ] enables designers to curate , compare , and explore suggested 3D design elements in the context of an overall 3D \u201ccube \" room layout ; IdeateRelate [ 93 ] locates design examples in coordinates of similarities corresponding to the users\u2019 original ideas ; the IdeaMache system provides an environment for free - form visual curation and sensemaking of creative materials in the context of a project canvas [ 91 ] ; and ImageSense embeds the process of searching for , exploring , and integrating examples into both individual and shared work spaces [ 47 ] . We build on this work by directly testing how the emerging pattern of contextualizing examples might impact the effects of examples on problem solving . To facilitate downstream theoretical development , we go beyond formulations of problem solving effects and outcomes that are task - specific \u2014 such as writing code [ 9 ] , designing websites [ 50 ] , or designing room layouts [ 39 ] \u2014 and / or removed from creativity - specific mechanisms , such as browsing and searching and exploring , to more theoretically grounded descriptions of psychological mechanisms such as fixation and problem reformulation . Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 5 3 METHODS 3 . 1 The WildCat Wells Task as a Controlled Analog to Exploratory Creative Problem Solving We experimentally investigated our research questions using a controlled analog to exploratory creativity , a term introduced by Margaret Boden\u2019s influential model of creativity [ 7 ] to describe a subset of creative problem solving processes that involve exploration within a conceptual space that is often large and complex . This conception of exploratory creative problem solving as search in a space has deep roots in research on search landscapes and innovation in organization and management science [ 5 ] , as well as psychological models of problem solving [ 65 ] and creativity [ 71 ] ( as reviewed in our discussion of model - based mechanisms for using examples in 2 . 2 . 2 ) . A key insight from this literature is that local search and hill - climbing are insufficient in more rugged and complex search landscapes , because they can trap searchers in local optima ; to overcome this , searchers need to find ways to explore or \u201cjump \" to new regions of the landscape [ 5 ] , such as guiding search through ( re ) modeling of the search space [ 36 ] . This contrast between local and distant exploration is often described in terms of the shift between exploitation and exploration [ 5 ] , where the latter search dynamics are more associated with innovation and creativity [ 71 ] . Note that the notion of exploratory creativity is distinct from another important class of creative processes that involve what Boden [ 7 ] calls transformational creativity : in this form of creativity , creators search for or construct alternative problem spaces ( as discussed in the related work ) [ 22 , 43 ] , rather than search within an existing problem space as given . Our controlled analog is the WildCat Wells task . The name of the task takes inspiration from the real - world task of wildcat drilling 1 , a form of exploratory drilling for oil and gas in an unfamiliar environment where the distribution of resource - rich locations is unknown . Accordingly , in this task , participants can \u201cdrill\u201d for \u201cresources \" in a 2D grid by clicking on locations in the grid ; clicking on a grid location then uncovers a score amount , analogous to the amount of oil / gas uncovered at a drilling site . Like its real - world counterpart , the distribution of resources in this task is unknown ; in our particular instantation , participants\u2019 goal is to uncover the most resource - rich drilling location ( i . e . , the grid location with the highest score ) . Following our conceptualization of examples as descriptions / representations of possible solutions to the same / similar problem , in this task , we operationalized examples as possible grid locations and their associated scores . We chose this task for several reason . First , we had a high degree of parameter control over the properties of the task and examples , which allowed us to precisely control the ruggedness of the task structure and also employ a within - subjects design while mitigating learning effects by constructing and sampling from a set of Wildcat Wells\u2019 tasks with isomorphic ruggedness / complexity properties ( see Section 3 . 1 . 1 ) . The task structure also gave us granular and precise measures of process and outcome dynamics . Finally , the simplicity of the task allowed us to minimize the impact of prior knowledge because the task does not require specialized domain expertise . While the specific task structure in terms of distribution of rewards over the search space is unknown to participants , the generic task structure of searching a space for rewards is probably not unfamiliar to most people . The Wildcat Wells task and its operationalization of examples is also conceptually similar to other instances of exploratory creativity that may draw on example solutions from a very similar problem : for instance , when searching for effective parameter settings for wing airfoil designs , other airfoil designs \u2014 which , like our grid location , are also combinations of parameters \u2014 may serve as relevant examples ; when designing effective ads for a vaccine persuasion campaign , other vaccine persuasion ads \u2014 which are also combinations of design features \u2014 may serve as relevant examples ; and when designing effective UI elements , other websites and their UI elements \u2014 which are also compositions of UI features \u2014 may serve as relevant examples . We also 1 https : / / en . wikipedia . org / wiki / Wildcatter Manuscript submitted to ACM 6 Chan , et al . Fig . 1 . Example Wildcat Wells search environment with color coding of points to indicate their scores ( 0 - 50 : dark blue to light blue , 50 - 100 : light red to dark red ) ( A ) , and example sets of high and low diversity sets of points in this search environment , which are given as examples ( B ) . adapted the Wildcat Wells task specifically from a prior study [ 62 ] of the dynamics of exploration and exploitation in collaborative problem solving . However , because the WildCat wells task is only analogous to exploratory creativity , our results here can only speak to effects of examples on exploratory , but not transformational , creative problem solving . 3 . 1 . 1 Search Environments . Our WildCat Wells search environments consisted of a 100x100 grid of points ( with corresponding scores controlled by a synthetic objective function that determines the distribution of scores ; see Algorithm 1 in Appendix A and our source code for generating search environments ) . Figure 1 ( A ) shows a representative search environment we used in our experiment . Our goal was to more closely match the difficult search spaces that the creativity theorist David Perkins calls \u201cKlondike spaces \" [ 71 ] , which are environments where simple \u201chill - climbing \" exploration strategies are insufficient , and likely outperformed by other creative exploration strategies such as a mix of exploration and exploitation [ 5 , 71 ] . We describe the specific parameter settings and algorithm we used to generate these task environments in Appendix A ( and share the code generating the environments in the Supplementary Material ) ; here , we note that we set the parameters to yield a search environment that was fairly rugged ( adding more false \u201cpeaks \" to incorrectly intuit as the location of the maximum score ) and locally noisy ( reducing the local correlation between scores in the grid , such that searchers would often be surprised by the score of nearby regions in the grid ) . The parameters to generate search environments were determined by a series of rubrics ( e . g . more than one area with scores higher than 80 ) , and pilots for a qualitative sense of difficulty based on the topology of the solution space . To reduce the likelihood that our results were tied to a specific formulation of the search environment , we generated ten search environments with the same synthetic objective function parameters but different random seeds . The resulting search environments were qualitatively similar to Mason & Duncan\u2019s work [ 62 ] . 3 . 1 . 2 Examples . To prepare sets of initial examples with High Diversity ( HD ) or Low Diversity ( LD ) , we randomly generated 10 , 000 sets of 10 examples each ( recall that each example is a \" drilling location \" point in the 100x100 Wildcat Wells search environment , with a corresponding score ) for each of our 10 search environments , and ranked the diversity of each example set with a close variant of the Determinantal Point Process ( DPP ) approach [ 49 ] described in Algorithm 2 in Appendix B proposed by Eesh et al . [ 41 ] ; intuitively , this approach measured the \u201chyper - volume \" spanned by a Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 7 selected set of points , such that larger volumes corresponded to higher levels of diversity , since these points spanned a larger set of the space of possible moves [ 49 ] . We then randomly picked three HD examples sets that were greater than the 99th percentile of the distribution of diversity across the example sets , and three LD examples sets with diversity lower than 1st percentile of the diversity distribution . To ensure that examples would not directly reveal the location of the peak , or provide a high enough score that participants might simply stop after seeing the example instead of searching , we discarded example sets that had any point with a score over 80 ( scores in the search environment ranged from 0 to 100 ) , and resampled example sets as necessary \u2014 subject to the same low / high diversity sampling criteria ) to construct our final example sets for each search environment . Figure 1 ( B ) shows an LD and HD example set used in our experiment . 3 . 2 Experiment Design We conducted a mixed design experiment . Example interface was a between - subjects factor , with three conditions : 1 ) \u201cparallel examples with context \" interface : all 10 examples were shown in the 100x100 space with color coding to denote the score associated with each point , referred to as the \u201c In - Context \" interface 2 ) \u201cparallel examples without context \" interface ( shown in Figure 2 ) : all 10 examples were shown in a list , also with color coding to indicate example score , referred to as the \u201c List \" interface 3 ) \u201cserial examples without context \" interface : only one example was shown at a time and the participant needed to use a dropdown button to see other examples , referred to as the \u201c Dropdown \" interface . Figure 2 shows the experimental interface of the List condition as an example . The In - Context interface was inspired by design patterns of example interfaces that contextualized examples in the creator\u2019s workspace or problem ( e . g . , [ 39 , 47 , 78 , 91 ] The List interface was inspired by the familiar design pattern of a \u201clist \" of examples , often in the context of a search interface ( as search results ) , or list of recommendations in a recommender interface . The Dropdown interface was designed to approximate more constrained interfaces for interacting with examples , such as through chat - based or recommendation systems ( e . g . , popping up one or two examples at a time ) . The three example interfaces were shown in the context of the WildCat Wells task in Figure 3 . We conjectured that interfaces that allowed for comparison between examples ( whether in the context of a task environment , as in the In - Context interface , or just with attributes shown for comparison , as in the List interface ) might facilitate more model - based usage of examples ( what we called a \u201c ( re ) modeling \" mechanism in 2 . 2 ) . Since we designed our Wildcat Wells task to be unsuitable for simpler hill - climbing ( e . g . , \u201cstimulation - based \" mechanism as described in 2 . 2 ) , we also expected that these interfaces might also lead to better performance on the task , through , for example , model - based exploration strategies . Example diversity was a within - subjects factor : each participant attempted the WildCat Wells task twice , once with a set of HD examples , and once with LD examples . Recall that we generated 10 variant search environments , each with their own set of HD and LD example sets . To approximate counterbalancing of our within - subjects factor , we created 2 \u201crun \" variants for each search environment , with each variant having an HD or LD example set as the first trial . Participants were randomly assigned first to an example interface condition , and then randomly assigned to one of the 20 potential \u201cruns \" in each interface condition ( but constraining assignment such that participants would not see the same search environment twice ) . Based on prior research on example diversity , we expected that participants would perform better when given high vs . low diversity example sets . Manuscript submitted to ACM 8 Chan , et al . Fig . 2 . Screenshot of experimental interface , shown for the List condition : the 100x100 grid , which constituted the search environment for the task , was shown on the left panel : participants explored the space by clicking anywhere on the 100x100 grid . The 10 initial examples , moves remaining , the score of current move , the current max score and score legend were shown on the right panel . In the Dropdown condition , the dropdown menu as seen in Figure 3 was shown in the same position as the list of examples in the List condition . In the In - Context condition , examples were instead overlaid as points , with corresponding values , on the search grid , as shown in Figure 3 . 3 . 3 Participants We recruited participants from the Amazon MTurk platform , limiting participants to U . S . residents with more than 500 HITs with at least 99 % approval rate . Each participant was paid US $ 1 . 3 for their participation , which was an effective rate of $ 10 per hour , given the average task completion time of 8 minutes . We aimed for a total sample size of 195 ( 65 per each of the three conditions ) , to achieve target statistical power of over 0 . 80 to detect medium - sized statistical effects in a mixed between - within design experiment analysis . After rejecting invalid work of 42 participants for irrelevant responses ( e . g . , \u201cnice \" ) to the closing survey question about how they used examples ) , we obtained data from 182 participants ( 63 females , 118 males , 1 other ; 65 in context , 56 List , and 61 Dropdown ) in total , yielding an effective statistical power of 0 . 86 for medium - sized effects . 3 . 4 Experimental Procedures Participants experienced the WildCat Wells task as a 100x100 space ( see Figure 2 ) . Their task was to find the square with the highest score . Participants explored squares by clicking on them to reveal their underlying score , shown in color coding , similar to the examples . To simulate the constrained nature of real creative tasks ( which often have some time / budget pressure ) and reduce the likelihood of ceiling effects , participants had a total budget of 60 moves for exploring squares . This budget was estimated from our pilot studies , where on average , most participants found the highest scoring square within 50 moves . We also provided incentives to encourage participants\u2019 exploration : there was a $ 0 . 25 bonus for achieving a highest score greater than 95 , and a $ 0 . 50 bonus for achieving the maximum score of 100 . The information panel on the right side of the experimental interface ( see Figure 2 showed moves remaining , the score of the current exploration and the maximum score the participant had achieved in the current round . Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 9 Fig . 3 . Three conditions of presenting examples : \u201c In - Context \" ( directly on the search environment grid ) , \u201c List \" ( in a list ) and \u201c Dropdown \" ( in a clickable dropdown selector ) . Since the WildCat Wells task does not interact strongly with prior knowledge in any particular domain , we addressed potential pre - existing differences in ability by measuring participants\u2019 baseline divergent thinking ability , a correlate of creative ability [ 75 ] . Before the study , we asked participants to generate as many alternative uses of coffee cup as they could in 2 minutes ( an instance of the commonly used Alternative Uses task [ 32 ] for measuring divergent thinking [ 75 ] ) . Participants were then given one trial round through the WildCat Wells task ( without examples ) to familiarize them with the interface and task . After that , participants completed two formal rounds of the WildCat Wells task , which constituted the main experimental trials in our study . Finally , participants completed a post - study questionnaire , with three free - response questions : 1 ) What strategy did you use for hunting ? 2 ) How did you use initial examples ( the values of ten points given to you ) ? 3 ) What differences did you notice between initial examples given in those two rounds ? Which did you find helpful ? We obtained institutional IRB approval for the whole project prior to the study . 4 RESULTS : PLANNED ANALYSES 4 . 1 No significant differences in baseline divergent thinking ability across interface conditions We first report the results of our check for random assignment with respect to divergent thinking ability and baseline performance on our task . We observed no statistically significant difference in the number of generated alternative uses across three conditions ( \" In - Context \" participants : \ud835\udc40 = 6 . 52 , \ud835\udc46\ud835\udc37 = 3 . 02 ; \u201c List \" participants : \ud835\udc40 = 5 . 75 , \ud835\udc46\ud835\udc37 = 3 . 58 ; \u201c Dropdown \" participants : \ud835\udc40 = 6 . 46 , \ud835\udc46\ud835\udc37 = 3 . 87 , Kruskal - Wallis \ud835\udc3b = 2 . 40 , \ud835\udc5d = 0 . 30 ) . Similarly , we observed no statistically significant difference in participants\u2019 best score on the trial run of the Wildcat Wells task across the conditions ( \" In - Context \" participants : \ud835\udc40 = 90 . 97 , \ud835\udc46\ud835\udc37 = 7 . 52 ; \u201c List \" participants : \ud835\udc40 = 90 . 91 , \ud835\udc46\ud835\udc37 = 7 . 13 ; \u201c Dropdown \" participants : \ud835\udc40 = 91 . 18 , \ud835\udc46\ud835\udc37 = 7 . 37 , Kruskal - Wallis \ud835\udc3b = 0 . 19 , \ud835\udc5d = 0 . 91 ) . This suggests that participants across the interface conditions were comparable in terms of baseline divergent thinking ability as well as baseline task performance . 4 . 2 List presentation of examples and low diversity example sets associated with lower best scores The List condition had slightly lower scores on average compared to the other conditions ( regardless of example diversity ; Fig . 4 , top right ) . There was also an overall slight advantage of HD examples over LD examples ( Fig . 4 , bottom right ) . A linear mixed effects model with best score as the dependent variable , interface condition and example diversity as factors , and random intercepts for participants ( estimated in the \u2018lme4\u2018 package in \u2018R\u2018 ) , showed a significant main Manuscript submitted to ACM 10 Chan , et al . Fig . 4 . Distribution of best scores by interface and example diversity conditions . Participants in the List interface condition had lower best scores than participants in the other interface conditions regardless of example diversity ( top right ) . Best scores were also lower when participants were given low vs . high diversity examples ( bottom right ) . effect of interface condition , F ( 2 , 179 ) = 5 . 92 , \ud835\udc5d < . 01 , \ud835\udc52\ud835\udc61\ud835\udc4e 2 = 0 . 06 . Pairwise post - hoc comparisons with Bonferroni corrections showed that participants in the List interface condition had significantly lower best scores ( est . marginal mean = 90 . 4 , SE = 0 . 65 ) compared to both the In - Context ( est . marginal mean = 92 . 7 , SE = 0 . 61 , contrast t ratio = 2 . 54 , \ud835\udc5d < . 05 ) and Dropdown interface conditions ( est . marginal mean = 93 . 4 , SE = 0 . 63 , contrast t ratio = 3 . 31 , \ud835\udc5d < . 01 ) . There was also a significant main effect of example diversity , F ( 1 , 181 ) = 4 . 59 , \ud835\udc5d < . 05 , \ud835\udc52\ud835\udc61\ud835\udc4e 2 = 0 . 02 ; post - hoc comparisons with Bonferroni corrections showed that participants had higher best scores when they received HD ( est . marginal mean = 92 . 7 , SE = 0 . 46 ) vs . LD examples ( est . marginal mean = 91 . 5 , SE = 0 . 46 , contrast t ratio = 2 . 14 , \ud835\udc5d < . 05 ) . 5 RESULTS : EXPLORATORY ANALYSES We conducted a set of exploratory analyses to better understand the results of our main planned analyses , focusing on understanding process effects of interface conditions that might plausibly explain performance differences . 5 . 1 In - Context presentation of examples associated with early performance advantages , and List presentation of examples with early and persistent performance disadvantages First , for a more granular view of performance , we examined how the participants\u2019 best score changed as a function of their move sequence . This analysis confirmed a cumulative disadvantage for participants in the List condition , but also showed an early advantage for the In - Context interface , particularly with LD examples ( see Figure 5 ) . Using a Kruskal - Wallis H - test on the current max score from the 1st to the 30th move , we observed statistically significant differences from the 1st move to the 6th move and the 8th move except the 7th move ( see Table 1 ) . 5 . 2 Variations in example presentation interfaces associated with different self - reported example usage strategies Next , to understand how participants used the initial examples in their exploration , two researchers coded participants\u2019 responses to the question \u201cHow did you use initial examples ( the values of ten points given to you ) ? \" with three codes : Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 11 Fig . 5 . Maximum score at n - th move for each participant : left ) HD ; right ) LD . We observe ( 1 ) cumulative disadvantages for the List condition , as well as ( 2 ) early advantages for the In - Context interface , especially with LD examples . n - th move In - Context ( \ud835\udc40 ) List ( \ud835\udc40 ) Dropdown ( \ud835\udc40 ) Statistic p - value 1 60 . 06 52 . 30 50 . 64 9 . 13 0 . 01 2 67 . 65 60 . 45 61 . 39 6 . 93 0 . 03 3 74 . 00 66 . 95 67 . 89 9 . 58 0 . 008 4 76 . 65 70 . 54 70 . 84 9 . 06 0 . 01 5 78 . 20 71 . 93 73 . 66 10 . 13 0 . 006 6 79 . 15 73 . 07 74 . 59 9 . 59 0 . 008 7 79 . 98 75 . 07 76 . 87 5 . 93 0 . 06 8 81 . 46 76 . 86 78 . 21 6 . 91 0 . 03 Table 1 . Means ( \ud835\udc40 ) and results of Kruskal - Wallis H - test on the current best score for three interface conditions with LD examples . There were statistically significant differences between the conditions from the 1st to 8th moves ( p < 0 . 05 except the 7th move ) . not using , stimulation - based and model - based . This classification was guided and refined by our initial theoretical interest in the contrast between stimulation - based and ( re ) modeling - based use of examples , as discussed in 2 . 2 . Examples of responses coded as \u201cnot using \" include \" I did not give much thought to it \" , and \" Not much to be honest \" ; examples of \u201cstimulation - based \" responses included \" Start at the reddest one and explore its surroundings \" , and \" I looked around the higher values for boxes that were darker \" ; examples of \u201cmodel - based \" responses include \" To get an overview on which squares would be best \" , and \" They gave a vague idea of whether or not there might be \u201chot \" or \u201ccool \" zones around those points \" . When we could not infer how the participants used the initial examples , the answers were coded as \u201cunclear \" . The researchers were blinded to condition during coding . Inter - rater reliability was substantial , at Cohen\u2019s \ud835\udf05 = 0 . 725 [ 51 ] ; all disagreements were resolved by discussion . The In - Context condition had the largest portion ( 30 . 8 % ) of participants who self - reported using the initial examples to model the space , compared to the List condition ( 10 . 7 % ) and the Dropdown condition ( 11 . 5 % ) ( see Figure 6 ) . In contrast , for self - reported use of initial examples to stimulate their exploration , the List condition had the highest percentage ( 42 . 9 % ) followed by the Dropdown condition ( 29 . 5 % ) and the in context condition ( 29 . 2 % ) . Finally , 17 / 61 ( 27 . 9 % ) Dropdown participants self - reported that they were not using examples , which was higher than participants in the other two conditions . Our log data were consistent with this observation : 37 / 61 ( 60 . 7 % ) participants in the Dropdown Manuscript submitted to ACM 12 Chan , et al . Fig . 6 . Raw proportion of participants expressed \u201cnot using ( examples ) \" , \u201cstimulation - based \" or \u201cmodel - based \" in their answer to \u201cHow did you use initial examples ( the values of ten points given to you ) ? \" . Error bars are standard error of proportion . More participants self - reported using a model - based strategy in the In - Context condition compared to other conditions . condition never clicked the dropdown button to see other examples in both HD and LD conditions . Of the remaining participants who did click on the dropdown button , we infer \u2014 assuming that each subsequent click corresponds to an example view \u2014 that the mean number of examples viewed was M = 7 . 37 ( SD = 3 . 40 ) for HD , and M = 7 . 05 ( SD = 3 . 61 ) for LD . 5 . 2 . 1 In - Context participants more likely than other interface conditions to self - report model - based example usage , and Dropdown participants more likely than other conditions to self - report not - using examples . We first statistically tested these patterns with a series of logistic regressions , one for each example strategy ( not - using , stimulation - based , and model - based ) ( see Table 2 ) . We ran separate logistic regressions rather than a single multinomial regression given our interest at this step in the relative likelihood across interface conditions of self - reporting a particular example strategy , rather than relative differences across strategies within each condition ( best answered by a multinomial logistic regression ) . We first observe that participants in the List and Dropdown conditions were less likely to self - report using a model - based strategy compared to the In - Context condition ( \ud835\udc35 = - 1 . 31 , 95 % CI = [ - 2 . 31 , - 0 . 31 ] , z = - 2 . 57 , \ud835\udc5d < . 05 for List vs . In - Context , and \ud835\udc35 = - 1 . 23 , 95 % CI = [ - 2 . 18 , - 0 . 29 ] , z = - 2 . 55 , \ud835\udc5d < . 01 for Dropdown vs . In - Context ) . In more intuitive terms , In - Context participants were approximately 3x more likely to self - report using a model - based example usage strategy compared to List or Dropdown participants ( Odds Ratio = 3 . 7 and 3 . 4 for In - Context vs . List , and In - Context vs . Dropdown ) . The overall model fit was better than a null model ( \ud835\udc3f\ud835\udc3f \ud835\udc5a\ud835\udc5c\ud835\udc51\ud835\udc52\ud835\udc59 = - 80 . 93 vs . \ud835\udc3f\ud835\udc3f \ud835\udc5b\ud835\udc62\ud835\udc59\ud835\udc59 = - 86 . 16 ) , Likelihood Ratio \ud835\udf12 2 ( 2 ) = 5 . 23 , \ud835\udc5d < . 01 . Next , we observe that participants in the In - Context condition were less likely to self - report not using examples compared to the Dropdown condition ( \ud835\udc35 = - 1 . 01 , 95 % CI = [ - 1 . 94 , - 0 . 09 ] , z = 2 . 14 , \ud835\udc5d < . 05 ) ; in more intuitive odds ratio terms , Dropdown participants were 2 . 7x more likely than In - Context participants to self - report a \u201cnot using \" strategy ( Odds Ratio = 2 . 75 ) . The overall model fit , though better than a null model ( \ud835\udc3f\ud835\udc3f \ud835\udc5a\ud835\udc5c\ud835\udc51\ud835\udc52\ud835\udc59 = - 86 . 62 vs . \ud835\udc3f\ud835\udc3f \ud835\udc5b\ud835\udc62\ud835\udc59\ud835\udc59 = - 89 . 10 ) , was marginally significant , Likelihood Ratio \ud835\udf12 2 ( 2 ) = 5 . 14 , \ud835\udc5d = . 08 . Finally , we observe that there were no significant differences across conditions in the likelihood of self - reporting a stimulation - based strategy . Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 13 Not using Stimulation - based Model - based Intercept - 0 . 951 [ - 1 . 511 , - 0 . 391 ] - 0 . 288 [ - 0 . 817 , 0 . 242 ] - 0 . 811 [ - 1 . 338 , - 0 . 284 ] In - Context - 1 . 013 [ - 1 . 940 , - 0 . 085 ] * - 0 . 597 [ - 1 . 349 , 0 . 156 ] List - 0 . 575 [ - 1 . 459 , 0 . 309 ] - 1 . 309 [ - 2 . 307 , - 0 . 312 ] * * Dropdown - 0 . 583 [ - 1 . 347 , 0 . 180 ] - 1 . 232 [ - 2 . 179 , - 0 . 285 ] * Table 2 . Coefficient estimates from multinomial logistic regressions of probability of self - reported example usage strategy ( 1 each for not - using , stimulation - based , model - based ) on interface condition . Statistics reported as \" coefficient , 95 % CI ( [ lower , upper ] ) \" . When the cell for a given interface condition is blank , that condition was used as the reference class in the regression . * p < . 05 ; * * p < . 01 ; * * * p < . 001 . Indeed , the overall model fit , though nominally better than the null model ( \ud835\udc3f\ud835\udc3f \ud835\udc5a\ud835\udc5c\ud835\udc51\ud835\udc52\ud835\udc59 = \u2212 114 . 52 vs . \ud835\udc3f\ud835\udc3f \ud835\udc5b\ud835\udc62\ud835\udc59\ud835\udc59 = \u2212 116 . 08 was not statistically significant , Likelihood Ratio \ud835\udf12 2 ( 2 ) = 1 . 56 , \ud835\udc5d = . 21 . 5 . 2 . 2 List participants more likely to self - report a stimulation - based example usage strategy compared to not - using or model - based example usage . Next , we focus on statistically evaluating the apparent predominance of a stimulation - based self - reported example usage strategy for List participants . We fitted a multinomial logistic regression , with model - based usage as the reference outcome class . Participants in the List condition were significantly more likely to self - report using a stimulation - based strategy compared to a model - based one , \ud835\udc35 = \u2212 1 . 44 , 95 % CI = [ - 0 . 34 , - 2 . 53 ] , z = - 2 . 58 , \ud835\udc5d < . 01 . In odds ratio terms , participants in the List interface condition were 4x more likely to self - report a stimulation - based vs . model - based strategy ( Odds Ratio = 4 . 21 ) . The overall model fit was statistically significantly better than a null model , Likelihood Ratio \ud835\udf12 2 ( 4 ) = 14 . 39 , \ud835\udc5d < . 01 . 5 . 3 List presentation of examples associated with more local initial exploration of the solution space Finally , we explored how log data might be consistent ( or not ) with participants\u2019 self - reported example usage strategies . We wanted to study how initial examples would affect participants\u2019 exploration behaviors , especially at the beginning of exploration when the examples provided were a major source of information . To explore this , we first constructed an exploration graph for the first 30 moves of each participant trial by computing Euclidean distances between each successive move ; the intuition was that long sequences of low distances between moves would suggest \u201chill - climbing \" , and large distances would suggest \u201cjumps \" . We conjectured that a \u201chill - climbing \" - like exploration graph would be consistent with a stimulation - based strategy , rather than a model - based strategy . Two coders independently coded all 364 exploration graphs ( each of the 182 participants had an HD plot and a LD plot ) , coding whether the exploration behaviors were hillclimbing ( h ) or not ( n ) for sequences of 10 moves ( the 0th - 10th , the 10th - 20th , the 20 - 30th ) . Two examples of coding are shown in Figure 7 . We coded 1092 10 - move instances ( 3 10 - move instances per round x 2 rounds per participant x 182 participants ) with substantial inter - rater reliability , Cohen\u2019s \ud835\udf05 = 0 . 78 . For all trials , a higher proportion of participants in the List interface condition used an exclusively hill - climbing strat - egy for the first 30 moves ( proportion = 0 . 161 , lower bound = 0 . 112 , upper bound = 0 . 210 ) compared to the In - Context ( pro - portion = 0 . 046 , lower bound = 0 . 020 , upper bound = 0 . 072 ) and Dropdown condition ( proportion = 0 . 082 , lower bound = 0 . 047 , upper bound = 0 . 117 ; see Figure 8aa ) . Similarly , for LD trials , the proportion of participants in using an exclusively hill - climbing strategy for the first 30 moves was higher for the List condition ( 34 % ) compared to the In - Context ( 18 % ) and Dropdown condition ( 18 % ) ( Fig . 8ab , left ) . In contrast , for HD trials , the proportion of participants using an Manuscript submitted to ACM 14 Chan , et al . Fig . 7 . Example exploration graphs used for coding hill - climbing strategies : not \u201call hill - climbing \" ( left ) and \u201call hill - climbing \" ( right ) , where each transition between participant moves is plotted on the x - axis , and the Euclidean distance between each move and its immediately preceding move is plotted on the y - axis . In the not \u201call hill - climbing \" example , the first 10 moves ( the 0th - 10th moves ) , where there are substantial variations in Dropdown move distances across the sequences , would be coded as non - hillclimbing ( N ) , while the 10th - 20th moves and the 20th - 30th moves , where Dropdown move distances are consistently low , would be coded as hillclimbing ( H ) . In the \u201call hill - climbing \" example , all first 30 moves would be coded as hillclimbing ( H ) . exclusively hill - climbing strategy for the first 30 moves was similar for the In - Context ( 18 % ) , List ( 25 % ) , and Dropdown conditions ( 23 % ) ( Fig . 8ab , left ) . To statistically test these observations , we fitted a series of logistic regressions , estimated with maximum likelihood , predicting \ud835\udc5d ( \ud835\udc4e\ud835\udc59\ud835\udc59 _ \u210e\ud835\udc56\ud835\udc59\ud835\udc59 ) , the probability of being all hill - climbing in the first 30 moves as a function of \ud835\udc56\ud835\udc5b\ud835\udc61\ud835\udc52\ud835\udc5f\ud835\udc53 \ud835\udc4e\ud835\udc50\ud835\udc52 . Prior work suggests that choice of exploration vs . exploitation is influenced by the \u201cgoodness \" of the current region of the search space ( better scores makes hill - climbing more likely ) [ 5 , 36 ] . Our data confirmed this pattern : the average score of the first move in each of the first three 10 - move blocks was positively correlated with the likelihood of being all hill - climbing in the first 30 moves , Kendall\u2019s \ud835\udf0f = . 17 , \ud835\udc5d < . 01 . Thus , we conditioned our logistic regression models on the average score at the beginning of each 10 - move block . We first analyzed \ud835\udc5d ( \ud835\udc4e\ud835\udc59\ud835\udc59 _ \u210e\ud835\udc56\ud835\udc59\ud835\udc59 ) aggregated across both HD and LD trials ( value would be 1 if both LD and HD trials were 1 ) , and then HD and LD trials separately . Table 3 shows the coefficient estimates for each of these models . For all trials , the coefficient for the contrast between the List and In - Context conditions was \ud835\udc35 = - 1 . 79 , 95 % CI = [ - 3 . 40 , - 0 . 45 ] , z = - 2 . 45 , \ud835\udc5d < . 05 ; in odds ratio terms , participants in the List condition were 6x more likely to use an exclusively hill - climbing strategy for the first 30 moves , compared to participants in the In - Context condition ( Odds Ratio = 5 . 99 ) . Note that this effect was independent of the significant positive coefficient for the average first score in the block . The overall model fit was statistically significantly better than a null model ( \ud835\udc3f\ud835\udc3f \ud835\udc5a\ud835\udc5c\ud835\udc51\ud835\udc52\ud835\udc59 = - 48 . 42 vs . \ud835\udc3f\ud835\udc3f \ud835\udc5b\ud835\udc62\ud835\udc59\ud835\udc59 = - 56 . 48 ) , Likelihood Ratio test \ud835\udf12 2 ( 3 ) = 16 . 12 , \ud835\udc5d < . 01 . Similarly , for LD trials , there was a statistically significant coefficient in the logistic regression model for the contrast between the In - Context and List conditions , \ud835\udc35 = - 0 . 95 , 95 % CI = [ - 1 . 83 , - 0 . 10 ] , z = - 2 . 16 , \ud835\udc5d < . 05 . In odds ratio terms , participants in the List condition were 2 . 5x more likely to use an exclusively hill - climbing strategy for the first 30 moves , compared to participants in the In - Context condition ( Odds Ratio = 2 . 58 ) . There was a similar contrast between the Dropdown and List conditions , \ud835\udc35 = - 0 . 95 , 95 % CI = [ - 1 . 85 , - 0 . 10 ] , z = - 2 . 12 , \ud835\udc5d < . 05 . As with the all trials model , this effect was independent of the significant positive coefficient for the average first score in the block . The overall model fit was statistically significantly better than a null model ( \ud835\udc3f\ud835\udc3f \ud835\udc5a\ud835\udc5c\ud835\udc51\ud835\udc52\ud835\udc59 = 92 . 98 vs . \ud835\udc3f\ud835\udc3f \ud835\udc5b\ud835\udc62\ud835\udc59\ud835\udc59 = \u2212 98 . 32 ) , Likelihood Ratio test \ud835\udf12 2 ( 3 ) = 10 . 67 , \ud835\udc5d < . 05 . In contrast , there were no statistically significant contrasts Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 15 All trials HD trials LD trials Intercept - 7 . 322 [ - 11 . 73 , - 3 . 739 ] - 1 . 734 [ - 3 . 522 , - 0 . 052 ] - 2 . 551 [ - 4 . 409 , - 0 . 870 ] In - Context vs . List - 1 . 789 [ - 3 . 400 , - 0 . 449 ] * - 0 . 431 [ - 1 . 325 , 0 . 448 ] - 0 . 950 [ - 1 . 833 , - 0 . 103 ] * Dropdown vs . List - 0 . 926 [ - 2 . 220 , 0 . 260 ] - 0 . 113 [ - 1 . 325 , 0 . 448 ] - 0 . 946 [ - 1 . 850 , - 0 . 0858 ] * * Avg . first score in block 0 . 089 [ 0 . 035 , 0 . 152 ] * * 0 . 010 [ - 0 . 016 , 0 . 037 ] 0 . 031 [ 0 . 005 , 0 . 059 ] * Table 3 . Coefficient estimates from logistic regressions of probability of predominantly hill - climbing strategy in the first 30 moves on interface condition and average first score in block , across all , HD , and LD trials . Statistics reported as \" coefficient , 95 % CI ( [ lower , upper ] ) \" . * p < . 05 ; * * p < . 01 ; * * * p < . 001 . ( a ) ( b ) Fig . 8 . Raw proportion of participants with a predominantly hillclimbing strategy in the first 30 moves , across interface conditions ( a ) with both HD and LD examples ( b ) . More \u201c List \" participants did hillclimbing in the first 30 moves with both high and low diverse example sets than \u201c In - Context \" participants . More \u201c List \u201d participants did hillclimbing for the first 30 moves with LD examples than other combinations of the presentation and example sets . between conditions in the HD trials , though the numerical pattern of results were similar to the other models ( generally negative coefficients for In - Context and Dropdown vs . List conditions ) . The overall model fit , though nominally better than a null model ( \ud835\udc3f\ud835\udc3f \ud835\udc5a\ud835\udc5c\ud835\udc51\ud835\udc52\ud835\udc59 = \u2212 95 . 14 vs . \ud835\udc3f\ud835\udc3f \ud835\udc5b\ud835\udc62\ud835\udc59\ud835\udc59 = \u2212 95 . 85 , was not statistically significant , Likelihood Ratio \ud835\udf12 2 ( 3 ) = 1 . 42 , \ud835\udc5d = . 70 . Because we were concerned this pattern of differences might be driven by pre - existing individual differences in propensity to hill - climbing , rather than a shift due to the interface condition , we repeated our coding procedure for exploration graphs generated from the initial trial round of exploration , which did not include examples ( described in 3 . 4 . The proportion of participants who displayed a predominant hill - climbing strategy , as described above , was distributed across conditions as follows : In - Context = 0 . 05 ( SE = . 03 ) , List = . 07 ( SE = . 03 ) , and Dropdown = 0 . 15 ( SE = . 05 ) . A logistic regression predicting the probability of a predominant hill - climbing strategy ( yes or no ) as a function of interface did not improve fit over a null model with no predictors , \ud835\udf12 2 ( 2 ) = 4 . 17 , \ud835\udc5d = . 12 ; note , however that the overall frequency of hill - climbing strategies was lower than in the main trials , and the List condition was not the condition with highest frequency ( in contrast to the main trials ) . There was also no significant correlation between the likelihood of hill - climbing predominance and either the number of alternative uses task responses , r = . 04 , \ud835\udc5d = . 61 , or the likelihood of hill - climbing predominance in the main trials , r = . 03 , \ud835\udc5d = . 65 . Altogether , these results are inconsistent with the alternative explanation that List participants were simply more likely ( due to individual differences ) to choose a predominantly hill - climbing strategy overall ; instead , taken together with the survey results , we believe this set of Manuscript submitted to ACM 16 Chan , et al . results suggest a shift in strategy towards hill - climbing ( or , as we describe in this paper , a stimulation - based strategy for using examples ) . 6 DISCUSSION AND CONCLUSION 6 . 1 Summary and Interpretation of Results In this paper , we aimed to contribute to a theory of human - example interaction to guide the design of example - based creativity support tools . Towards this goal , we conducted an experiment with a controlled analog to an exploratory creativity [ 7 ] task to investigate how example presentation interface variations influence whether / how people benefit from examples . We found evidence that List presentation of examples might harm the quality of final solutions . For example we found variations across interface conditions in mean best score obtained at the end of trials across example interface and diversity conditions ( List participants had worse best scores compared to In - Context and Dropdown participants ; Section 4 . 2 ) and cumulative performance differences ( with an early and persistent disadvantage of the List condition compared to the other conditions , with an especially pronounced early disadvantage relative to the In - Context condition ; Section 5 . 1 ) . This result is conceptually significant because the \u201c List \u201d participants received more information ( seeing all 10 examples at the same time ) than the \u201c Dropdown \u201d participants ( only seeing 1 example at a time . and over 60 % of them never checked other examples ) , and approximately equivalent information but different presentation compared with the \u201c In - Context \u201d condition . This suggests that seemingly unimportant , low - level interaction design decisions with respect to presentation of examples can have measurable consequences for creative problem solving performance . Separately , we also observed beneficial effects of diversity for final solution quality , in line with some previous work [ 2 , 4 , 29 , 38 , 80 , 96 ] ; importantly , this effect was similar in magnitude to the example presentation effects , suggesting that example presentation considerations may be just as important to consider as example characteristics when designing example - based creativity support systems . Second , our exploratory analyses suggest that \" In - Context \" and \u201c List \" presentation of examples may lead to distinct patterns of example usage . \u201c In - Context \u201d presentation of initial examples was associated with a greater likelihood of a \u201cmodel - based \" strategy for using examples , where participants self - reported using the examples to gain an overall understanding of the distribution of score in the search environment to guide their exploration , compared to List or Dropdown presentations . Conversely , the \u201c List \u201d presentation of initial examples seemed to encourage a predominant \u201cstimulation - based \" example usage strategy , where participants selected promising examples as starting points for their exploration . Importantly this self - report data was consistent with patterns in our log data : we observed that List participants were more likely to use a predominantly \u201chill - climbing \" strategy ( with low Dropdown distance between their moves ) early in their exploration , relative to the In - Context and Dropdown participants ; this association was independent of the relationship between hill - climbing behavior and the \u201cgoodness \" of initial moves ( hill - climbing in a given block of moves was more likely when the initial move was higher - scoring , consistent with prior empirical work on exploration / exploitation decisions [ 5 , 36 ] ) . Considering these results alongside the performance results suggests that List participants were being fixated [ 40 ] by the examples . A fruitful direction for further research would be to investigate the mechanisms that drive fixation in the List condition . One reason might be the upper limit in scores ( no more than 80 / 100 ) on the examples presented to the participants ; if taken as starting points to begin hill - climbing , those relatively low quality examples could be misleading , and block access to high - quality solutions . Another reason might be the increased effort needed to connect examples of Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 17 the List condition to the search space , which would be consistent with past research on the cognitive load benefits of integrating diagrams and text ( similar to integrating examples and the search space ) in instructional design [ 14 ] . We view the difficulty of transferring from the text modality of lists of examples to the visuospatial modality of the In - Context solution space as a potential mechanism by which example presentation variations might shape their impact on ideation : future work could investigate in more detail how different example presentation designs might shift the cost structure of different processing strategies , in a similar way that variations in environment or interface structure have been shown to shape sensemaking by changing the cost structure of various crucial actions , such as skimming / previewing , moving documents , applying schemas to documents , or adjusting schemas [ 72 , 76 , 77 ] . Separately , we observed that the \u201c Dropdown \u201d presentation was associated with limited usage of the examples : many Dropdown participants self - reported not using examples ( moreso than In - Context participants , for example ) , and this was also corroborated in their log data ( via a lack of interaction with the example interface ) . We do not think that this lack of example usage is indicative of a lack of engagement : recall , for instance , that performance in this condition was on par with the In - Context condition ( i . e . , higher than in the List condition ) . Post - survey comments indicating enjoyment and engagement ( e . g . , \" Fun game . Thank you ! \" ) were also seen across conditions at similar rates , and there were no statistically significant mean differences across the conditions in the trial run of the task . For these reasons , we believe that \u2014 possibly due to the interaction affordances \u2014 the Dropdown condition appeared to act similarly to a \" no - examples \" control condition , where participants used a wider mix of strategies vs . a particular set of example - based strategies tied to an experimental intervention . In light of this , the overall strong performance of the Dropdown condition is akin to past observations of strong performance by control \u201cno - intervention \" conditions in ideation experiments ( see , e . g . , [ 11 , 80 ] ; we thus add to a growing body of evidence that it may be easier to harm rather than help creative ideation by intervening ( as in the List condition ) . Overall , our results suggest that interaction design considerations for human - example interaction go beyond usability : there is indeed a space of mappings to explore between design affordances and fundamental psychological mechanisms of creative inspiration from examples . From a practical standpoint , our empirical results suggest the limitations of only showing examples without the problem space as context , especially if the problem space is large ( a common feature of real world problems ) and there exist some potential solutions far away from the initial examples . This implication is significant since the \u201c List \u201d view of examples - examples presented in a list - is commonly used in current creativity support tools , such as search engines and recommendation systems , yet was associated with substantial negative effects on the usage of examples and task performance relative to In - Context presentation of examples . 6 . 2 Limitations The WildCat Wells task we used in our experiment is simpler than most real - world exploratory creativity tasks\u2014 such as airfoil design , ad design , or UI design \u2014 of which it is an analog . For instance , the task did not require any specialized domain knowledge , and the generic task structure of searching a space for rewards is probably familiar to most people : indeed , one participant in our study noted that in the post - survey that the task was \" very fun and somewhat similar to minesweeper . \" Additionally , although we carefully constructed our Wildcat Wells task surfaces to be rugged , with multiple peaks of good solutions , our task technically has a single best solution ; in contrast , many real - world creative problems \u2014 such as policy design \u2014 lack a single best solution , due to task factors such as intrinsic tradeoffs between different problem requirements ; in these cases , creators often search for and construct \u201cgood enough \" solutions under high uncertainty ( though this might sometimes be a function of feasibility constraints rather than intrinsic properties of the task ) . It is unsurprising , then , that participants performed relatively well as a whole , and \u201c Dropdown \u201d participants Manuscript submitted to ACM 18 Chan , et al . also had competitive performance even though they did not interact or use the examples . We note , however , that performance was not quite at ceiling : only 16 % of participants reached the global max in either trial ( and 42 % reached the threshold score of 95 for the first bonus . Still , caution is warranted when generalizing to other more complex instances of exploratory creativity ; for instance , it may be that the effects of examples , and the corresponding effects of variations in their presentation interactions , will become more pronounced in more sophisticated tasks . Relatedly , the Wildcat Wells task captures aspects of search dynamics ( exploration and exploitation ) in exploratory creative problem solving quite well , but does not enable observation of more sophisticated psychological mechanisms for working with examples . For instance , it is unclear what it might mean to \u201ccombine \" different problem solving moves for this task . Additionally , while participants engaged in modeling of the problem space , they were not able to make larger changes to the problem space , such as questioning assumptions or relaxing constraints [ 45 ] , or even changing the goal / problem altogether [ 43 ] , mechanisms that are common in real - world creative problem solving tasks , such as design [ 22 ] . Thus , we reiterate that our results cannot speak to how example presentation design decisions might influence example usage for transformational creativity [ 7 ] tasks . Thus , more work is needed to extend our exploration of patterns in example interaction design choices to more complex settings : for example , what might it mean to design a \u201ccontextualized \" presentation of examples for UI elements , more complex airfoil designs , ad persuasion campaigns , research papers , or policy ideas ? We are keen to build on existing design patterns similar to this in previous systems such as ReflectionSpace [ 78 ] , MoodCubes [ 39 ] , and ImageSense [ 47 ] , as discussed in Section 2 . 3 . Our implementation of the Dropdown condition may also be quite different from other Dropdown presentations , such as forward / backward interfaces ( e . g . , image suggestions [ 46 ] ) . Future studies can explore the consequences of these differences . For now , we note that our main results on the contrast between In - Context and List conditions are independent of this limitation , and recommend caution in generalizing the results from the Dropdown condition around non - use of examples . Finally , we did not measure demographic information that may have been correlated with task performance or example usage and / or exploration patterns \u2013 for example , personality traits such as disagreeableness or extraversion may be correlated with real - world creative achievement [ 95 ] ; and gender might interact with potential differences in visuospatial reasoning demands between example interfaces , given some existing research on gender differences in spatial ability [ 52 ] . 6 . 3 Towards an interaction - oriented theory of creative inspiration from examples Returning to our higher - level goal of constructing an interaction - oriented theory of human - example interaction , we now reflect on how the empirical results from our study , in conversation with theoretical mechanisms and design patterns from prior work , could contribute to an overall theory that bridges design patterns to psychological mechanisms . We conjecture that a useful theory of human - example interaction could be conceptualized as paths through multiple coordinated spaces of example interaction patterns , example - ideation psychological mechanisms , ideation characteristics and creative outcomes . Paths through this overall set of coordinated spaces could then represent a set of principled design hypotheses about how to best support creative work with examples . For instance , bringing our empirical results in conversation with the literature we reviewed in sections 2 . 2 and 2 . 3 , we could hypothesize that , given a particular exploratory creativity task environment like our instantiation of the WildCat wells task , where the key creative outcome of solution quality is determined at least in part by the ideation characteristic of diversity of search , which is in turn positively influenced by the psychological mechanism of ( re ) modeling , and negatively influenced by the mechanism of stimulation , it may be advantageous to choose example interaction patterns like Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 19 contextualizing examples in the problem space ( which is positively mapped to ( re ) modeling mechanisms ) , over patterns like List viewing of examples ( which is positively mapped to stimulation mechanisms ) . Multiple other hypothesized paths could be generated and refined to map other example interaction patterns from prior work , such as faceted search systems , or example dissection / analysis , to other psychological mechanisms , such as conceptual combination , or analogical abstraction ; each of these mechanisms might then in turn be contextually important for certain kinds of creative problems , such as policymaking or room layout design . We believe that fleshing out these paths through these coordinated spaces towards a theory of human - example interaction can both make contributions to fundamental HCI theory \u2014 by enhancing synthesis of design knowledge about how to best support creative inspiration from examples \u2014 and practice \u2014 by providing a principled framework that is sufficiently granular and directly connected to design decisions , to guide effective design decisions when building example - based creativity support systems , and to practicing creators who wish to more effectively leverage examples in their creative process . We invite the rest of the creativity support systems community to join us in these efforts . REFERENCES [ 1 ] Marine Agogu\u00e9 , Akin Kazak\u00e7i , Armand Hatchuel , Pascal Le Masson , Benoit Weil , Nicolas Poirel , and Mathieu Cassotti . 2014 . The Impact of Type of Examples on Originality : Explaining Fixation and Stimulation Effects . The Journal of Creative Behavior 48 , 1 ( 2014 ) , 1 \u2013 12 . https : / / doi . org / 10 . 1002 / jocb . 37 00000 _ eprint : https : / / onlinelibrary . wiley . com / doi / pdf / 10 . 1002 / jocb . 37 . [ 2 ] Niek Althuizen and Berend Wierenga . 2014 . Supporting Creative Problem Solving with a Case - Based Reasoning System . Journal of Management Information Systems 31 , 1 ( 2014 ) , 309 \u2013 340 . https : / / doi . org / 10 . 2753 / MIS0742 - 1222310112 [ 3 ] L . J . Ball , T . C . Ormerod , and N . J . Morley . 2004 . Spontaneous analogising in engineering design : a comparative analysis of experts and novices . Design Studies 25 , 5 ( 2004 ) , 495 \u2013 508 . [ 4 ] Jonali Baruah and Paul B . Paulus . 2011 . Category assignment and relatedness in the group ideation process . Journal of Experimental Social Psychology 47 , 6 ( 2011 ) , 1070 \u2013 1077 . 00000 . [ 5 ] Oliver Baumann , Jens Schmidt , and Nils Stieglitz . 2019 . Effective Search in Rugged Performance Landscapes : A Review and Outlook . Journal of Management 45 , 1 ( Jan . 2019 ) , 285 \u2013 318 . https : / / doi . org / 10 . 1177 / 0149206318808594 Publisher : SAGE Publications Inc . [ 6 ] Justin M . Berg . 2014 . The primal mark : How the beginning shapes the end in the development of creative ideas . Organizational Behavior and Human Decision Processes 125 , 1 ( 2014 ) , 1 \u2013 17 . https : / / doi . org / 10 . 1016 / j . obhdp . 2014 . 06 . 001 00056 . [ 7 ] M . A . Boden . 2004 . The creative mind : Myths and mechanisms . New York . 00000 . [ 8 ] Joel Brandt , Mira Dontcheva , Marcos Weskamp , and Scott R . Klemmer . 2010 . Example - centric Programming : Integrating Web Search into the Development Environment . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201910 ) . ACM , New York , NY , USA , 513 \u2013 522 . https : / / doi . org / 10 . 1145 / 1753326 . 1753402 [ 9 ] Joel Brandt , Mira Dontcheva , Marcos Weskamp , and Scott R . Klemmer . 2010 . Example - centric Programming : Integrating Web Search into the Development Environment . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201910 ) . ACM , New York , NY , USA , 513 \u2013 522 . https : / / doi . org / 10 . 1145 / 1753326 . 1753402 [ 10 ] JoelChan , StevenP . Dow , andChristianD . Schunn . 2015 . DoTheBestDesignIdeas ( Really ) ComeFromConceptuallyDistantSourcesOfInspiration ? Design Studies 36 ( 2015 ) , 31 \u2013 58 . https : / / doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 00105 . [ 11 ] Joel Chan , Katherine Fu , Christian . D . Schunn , Jonathan Cagan , Kristin L . Wood , and Kenneth Kotovsky . 2011 . On the benefits and pitfalls of analogies for innovative design : Ideation performance based on analogical distance , commonness , and modality of examples . Journal of Mechanical Design 133 ( 2011 ) , 081004 . https : / / doi . org / 10 . 1115 / 1 . 4004396 00304 . [ 12 ] Joel Chan and Christian D . Schunn . 2015 . The importance of iteration in creative conceptual combination . Cognition 145 ( Dec . 2015 ) , 104 \u2013 115 . https : / / doi . org / 10 . 1016 / j . cognition . 2015 . 08 . 008 [ 13 ] Joel Chan , Pao Siangliulue , Denisa Qori McDonald , Ruixue Liu , Reza Moradinezhad , Safa Aman , Erin T . Solovey , Krzysztof Z . Gajos , and Steven P . Dow . 2017 . Semantically Far Inspirations Considered Harmful ? : Accounting for Cognitive States in Collaborative Ideation . In Proceedings of the 2017 ACM SIGCHI Conference on Creativity and Cognition ( C & C \u201917 ) . ACM , New York , NY , USA , 93 \u2013 105 . https : / / doi . org / 10 . 1145 / 3059454 . 3059455 [ 14 ] Paul Chandler and John Sweller . 1992 . The Split - Attention Effect as a Factor in the Design of Instruction . British Journal of Educational Psy - chology 62 , 2 ( 1992 ) , 233 \u2013 246 . https : / / doi . org / 10 . 1111 / j . 2044 - 8279 . 1992 . tb01017 . x _ eprint : https : / / onlinelibrary . wiley . com / doi / pdf / 10 . 1111 / j . 2044 - 8279 . 1992 . tb01017 . x . [ 15 ] Kerry Shih - Ping Chang and Brad A . Myers . 2012 . WebCrystal : understanding and reusing examples in web authoring . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . ACM , Austin Texas USA , 3205 \u2013 3214 . https : / / doi . org / 10 . 1145 / 2207676 . 2208740 [ 16 ] Minsuk Chang , Leonore V . Guillain , Hyeungshik Jung , Vivian M . Hare , Juho Kim , and Maneesh Agrawala . 2018 . RecipeScape : An Interactive Tool for Analyzing Cooking Instructions at Scale . In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems . ACM , Montreal QC Manuscript submitted to ACM 20 Chan , et al . Canada , 1 \u2013 12 . https : / / doi . org / 10 . 1145 / 3173574 . 3174025 [ 17 ] Allan M . Collins and Elizabeth F . Loftus . 1975 . A spreading - activation theory of semantic processing . Psychological Review 82 , 6 ( 1975 ) , 407 \u2013 428 . [ 18 ] Nathan Crilly and Carlos Cardoso . 2017 . Where next for research on fixation , inspiration and creativity in design ? Design Studies 50 ( May 2017 ) , 1 \u2013 38 . https : / / doi . org / 10 . 1016 / j . destud . 2017 . 02 . 001 [ 19 ] Darren W . Dahl and Page Moreau . 2002 . The Influence and value of analogical thinking during new product ideation . Journal of Marketing Research 39 , 1 ( 2002 ) , 47 \u2013 60 . https : / / doi . org / 10 . 1509 / jmkr . 39 . 1 . 47 . 18930 00714 . [ 20 ] Zijian Ding , Arvind Srinivasan , Stephen Macneil , and Joel Chan . 2023 . Fluid Transformers and Creative Analogies : Exploring Large Language Models\u2019 Capacity for Augmenting Cross - Domain Analogical Creativity . In Proceedings of the 15th Conference on Creativity and Cognition ( C & amp ; C \u201923 ) . Association for Computing Machinery , New York , NY , USA , 489 \u2013 505 . https : / / doi . org / 10 . 1145 / 3591196 . 3593516 [ 21 ] Alex Doboli , Anurag Umbarkar , Varun Subramanian , and Simona Doboli . 2014 . Two experimental studies on creative concept combinations in modular design of electronic embedded systems . Design Studies 35 , 1 ( 2014 ) , 80 \u2013 109 . https : / / doi . org / 10 . 1016 / j . destud . 2013 . 10 . 002 [ 22 ] Kees Dorst and Nigel Cross . 2001 . Creativity in the design process : co - evolution of problem \u2013 solution . Design studies 22 , 5 ( 2001 ) , 425 \u2013 437 . 02587 . [ 23 ] Claudia Eckert and Martin Stacey . 1998 . Fortune favours only the prepared mind : Why sources of inspiration are essential for continuing creativity . Creativity and Innovation Management 7 , 1 ( 1998 ) , 1 \u2013 12 . 00056 . [ 24 ] C . Eckert and M . Stacey . 2003 . Adaptation of Sources of Inspiration in Knitwear Design . Creativity Research Journal 15 , 4 ( 2003 ) , 355 \u2013 384 . 00000 . [ 25 ] Katherine Fu , Joel Chan , Jonathan Cagan , Kenneth Kotovsky , Christian Schunn , and Kristin Wood . 2013 . The Meaning of Near and Far : The Impact of Structuring Design Databases and the Effect of Distance of Analogy on Design Output . Journal of Mechanical Design 135 , 2 ( 2013 ) , 021007 . https : / / doi . org / 10 . 1115 / 1 . 4023158 [ 26 ] William Gaver . 2011 . Making spaces : how design workbooks work . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201911 ) . Association for Computing Machinery , New York , NY , USA , 1551 \u2013 1560 . https : / / doi . org / 10 . 1145 / 1978942 . 1979169 [ 27 ] D . Gentner and Arthur B . Markman . 1997 . Structure mapping in analogy and similarity . American Psychologist 52 , 1 ( 1997 ) , 45 \u2013 56 . [ 28 ] M . L . Gick and K . J . Holyoak . 1983 . Schema induction and analogical transfer . Cognitive Psychology 15 , 1 ( 1983 ) , 1 \u2013 38 . https : / / doi . org / 10 . 1016 / 0010 - 0285 ( 83 ) 90002 - 6 04045 . [ 29 ] M . M . Gielnik , M . Frese , J . M . Graf , and A . Kampschulte . 2011 . Creativity in the opportunity identification process and the moderating effect of diversity of information . Journal of Business Venturing 27 , 5 ( 2011 ) , 559 \u2013 576 . https : / / doi . org / 10 . 1016 / j . jbusvent . 2011 . 10 . 003 00000 . [ 30 ] Vinod Goel and Peter Pirolli . 1992 . The structure of design problem spaces . Cognitive Science 16 ( 1992 ) , 395 \u2013 429 . [ 31 ] Milene Gon\u00e7alves , Carlos Cardoso , and Petra Badke - Schaub . 2013 . Inspiration peak : exploring the semantic distance between design problem and textual inspirational stimuli . International Journal of Design Creativity and Innovation 1 , 4 ( 2013 ) , 215 \u2013 232 . [ 32 ] Joy P . Guilford . 1967 . The nature of human intelligence . McGraw - Hill , New York , NY . [ 33 ] Andrew Hargadon and Robert I . Sutton . 1997 . Technology Brokering and Innovation in a Product Development Firm . Administrative Science Quarterly 42 , 4 ( 1997 ) , 716 \u2013 749 . https : / / doi . org / 10 . 2307 / 2393655 03534 Publisher : [ Sage Publications , Inc . , Johnson Graduate School of Management , Cornell University ] . [ 34 ] M . E . Helms , S . Vattam , and A . K . Goel . 2008 . Compound analogical design , or how to make a surfboard disappear . In Proceedings of the 30th Annual Conference of the Cognitive Science Society . 00013 . [ 35 ] Scarlett R . Herring , Chia - Chen Chang , Jesse Krantzler , and Brian P . Bailey . 2009 . Getting inspired ! : understanding how and why examples are used in creative design practice . In Proceedings of the 27th international conference on Human factors in computing systems . ACM , 87 \u2013 96 . https : / / doi . org / 10 . 1145 / 1518701 . 1518717 [ 36 ] Thomas T . Hills , Peter M . Todd , David Lazer , A . David Redish , and Iain D . Couzin . 2015 . Exploration versus exploitation in space , mind , and society . Trends in Cognitive Sciences 19 , 1 ( Jan . 2015 ) , 46 \u2013 54 . https : / / doi . org / 10 . 1016 / j . tics . 2014 . 10 . 004 [ 37 ] Keith J . Holyoak and Paul Thagard . 1996 . Mental leaps : Analogy in creative thought . MIT Press , Cambridge , MA . 02398 . [ 38 ] Paul A . Howard - Jones , Sarah - Jayne . J . Blakemore , Elspeth A . Samuel , Ian R . Summers , and Guy Claxton . 2005 . Semantic divergence and creative story generation : An fMRI investigation . Cognitive Brain Research 25 , 1 ( 2005 ) , 240 \u2013 250 . [ 39 ] Alexander Ivanov , David Ledo , Tovi Grossman , George Fitzmaurice , and Fraser Anderson . 2022 . MoodCubes : Immersive Spaces for Collecting , Discovering and Envisioning Inspiration Materials . In Designing Interactive Systems Conference ( DIS \u201922 ) . Association for Computing Machinery , New York , NY , USA , 189 \u2013 203 . https : / / doi . org / 10 . 1145 / 3532106 . 3533565 [ 40 ] David G . Jansson and Steven M . Smith . 1991 . Design fixation . Design Studies 12 , 1 ( 1991 ) , 3 \u2013 11 . [ 41 ] Eesh Kamrah , Fatemeh Ghoreishi , Zijian Ding , Joel Chan , and Mark Fuge . 2023 . How Diverse Initial Samples Help and Hurt Bayesian Optimizers . Journal of Mechanical Design ( July 2023 ) , 1 \u2013 32 . https : / / doi . org / 10 . 1115 / 1 . 4063006 [ 42 ] Youwen Kang , Zhida Sun , Sitong Wang , Zeyu Huang , Ziming Wu , and Xiaojuan Ma . 2021 . MetaMap : Supporting Visual Metaphor Ideation through Multi - dimensional Example - based Exploration . In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems . ACM , Yokohama Japan , 1 \u2013 15 . https : / / doi . org / 10 . 1145 / 3411764 . 3445325 [ 43 ] Craig A . Kaplan and Herbert A . Simon . 1990 . In search of insight . Cognitive Psychology 22 , 3 ( 1990 ) , 374 \u2013 419 . https : / / doi . org / 10 . 1016 / 0010 - 0285 ( 90 ) 90008 - R [ 44 ] Ianus Keller , Froukje Sleeswijk Visser , Remko van der Lugt , and Pieter Jan Stappers . 2009 . Collecting with Cabinet : or how designers organise visual material , researched through an experiential prototype . Design Studies 30 , 1 ( Jan . 2009 ) , 69 \u2013 86 . https : / / doi . org / 10 . 1016 / j . destud . 2008 . 06 . 001 Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 21 [ 45 ] G . Knoblich , S . Ohlsson , H . Haider , and D . Rhenius . 1999 . Constraint relaxation and chunk decomposition in insight problem solving . Journal of Experimental Psychology : Learning , Memory , and Cognition 25 , 6 ( 1999 ) , 1534 \u2013 1555 . https : / / doi . org / 10 . 1037 / 0278 - 7393 . 25 . 6 . 1534 00691 . [ 46 ] Janin Koch , Andr\u00e9s Lucero , Lena Hegemann , and Antti Oulasvirta . 2019 . May AI ? : Design Ideation with Cooperative Contextual Bandits . In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems ( CHI \u201919 ) . ACM , New York , NY , USA , 633 : 1 \u2013 633 : 12 . https : / / doi . org / 10 . 1145 / 3290605 . 3300863 event - place : Glasgow , Scotland Uk . [ 47 ] Janin Koch , Nicolas Taffin , Michel Beaudouin - Lafon , Markku Laine , Andr\u00e9s Lucero , and Wendy E . Mackay . 2020 . ImageSense : An Intelligent Collaborative Ideation Tool to Support Diverse Human - Computer Partnerships . Proceedings of the ACM on Human - Computer Interaction 4 , CSCW1 ( May 2020 ) , 1 \u2013 27 . https : / / doi . org / 10 . 1145 / 3392850 00010 . [ 48 ] Janin Koch , Nicolas Taffin , Andr\u00e9s Lucero , and Wendy E . Mackay . 2020 . SemanticCollage : Enriching Digital Mood Board Design with Semantic Labels . In Proceedings of the 2020 ACM Designing Interactive Systems Conference . ACM , Eindhoven Netherlands , 407 \u2013 418 . https : / / doi . org / 10 . 1145 / 3357236 . 3395494 [ 49 ] Alex Kulesza and Ben Taskar . 2012 . Determinantal point processes for machine learning . Foundations and Trends\u00ae in Machine Learning 5 , 2 - 3 ( 2012 ) , 123 \u2013 286 . https : / / doi . org / 10 . 1561 / 2200000044 [ 50 ] Ranjitha Kumar , Jerry O . Talton , Salman Ahmad , and Scott R . Klemmer . 2011 . Bricolage : example - based retargeting for web design . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201911 ) . Association for Computing Machinery , New York , NY , USA , 2197 \u2013 2206 . https : / / doi . org / 10 . 1145 / 1978942 . 1979262 [ 51 ] J . Richard Landis and Gary G . Koch . 1977 . The Measurement of Observer Agreement for Categorical Data . Biometrics 33 , 1 ( March 1977 ) , 159 . https : / / doi . org / 10 . 2307 / 2529310 [ 52 ] Jillian E . Lauer , Eukyung Yhang , and Stella F . Lourenco . 2019 . The development of gender differences in spatial reasoning : A meta - analytic review . Psychological Bulletin 145 ( 2019 ) , 537 \u2013 565 . https : / / doi . org / 10 . 1037 / bul0000191 Place : US Publisher : American Psychological Association . [ 53 ] Brian Lee , Savil Srivastava , Ranjitha Kumar , Ronen Brafman , and Scott R . Klemmer . 2010 . Designing with Interactive Example Galleries . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201910 ) . ACM , New York , NY , USA , 2257 \u2013 2266 . https : / / doi . org / 10 . 1145 / 1753326 . 1753667 [ 54 ] Julie S . Linsey , Ian Tseng , Katherine Fu , Jonathan Cagan , Kristin L . Wood , and Christian D . Schunn . 2010 . A Study of Design Fixation , Its Mitigation and Perception in Engineering Design Faculty . Journal of Mechanical Design 132 , 4 ( 2010 ) , 041003 . [ 55 ] Andr\u00e9s Lucero . 2012 . Framing , aligning , paradoxing , abstracting , and directing : how design mood boards work . In Proceedings of the Designing InteractiveSystemsConference ( DIS\u201912 ) . AssociationforComputingMachinery , NewYork , NY , USA , 438 \u2013 447 . https : / / doi . org / 10 . 1145 / 2317956 . 2318021 [ 56 ] Nic Lupfer , Andruid Kerne , Andrew M . Webb , and Rhema Linder . 2016 . Patterns of Free - form Curation : Visual Thinking with Web Content . In Proceedings of the 2016 ACM on Multimedia Conference ( MM \u201916 ) . ACM , New York , NY , USA , 12 \u2013 21 . https : / / doi . org / 10 . 1145 / 2964284 . 2964303 [ 57 ] Stephen MacNeil , Zijian Ding , Kexin Quan , Ziheng Huang , Kenneth Chen , and Steven P . Dow . 2021 . ProbMap : Automatically constructing design galleries through feature extraction and semantic clustering . In The Adjunct Publication of the 34th Annual ACM Symposium on User Interface Software and Technology . ACM , Virtual Event USA , 134 \u2013 136 . https : / / doi . org / 10 . 1145 / 3474349 . 3480203 00000 . [ 58 ] Stephen MacNeil , Ziheng Huang , Kenneth Chen , Zijian Ding , Alex Yu , Kendall Nakai , and Steven P . Dow . 2023 . Freeform Templates : Combining Freeform Curation with Structured Templates . In Creativity and Cognition . 478 \u2013 488 . https : / / doi . org / 10 . 1145 / 3591196 . 3593337 arXiv : 2305 . 00937 [ cs ] . [ 59 ] N . R . F . Maier . 1931 . Reasoning in humans . II . The solution of a problem and its appearance in consciousness . Journal of Comparative Psychology 12 , 2 ( 1931 ) , 181 \u2013 194 . https : / / doi . org / 10 . 1037 / h0071361 [ 60 ] R . L . Marsh and G . H . Bower . 1993 . Eliciting cryptomnesia : unconscious plagiarism in a puzzle task . Journal of Experimental Psychology : Learning , Memory , and Cognition 19 , 3 ( 1993 ) , 673 \u2013 88 . [ 61 ] R . L . Marsh , T . B . Ward , and J . D . Landau . 1999 . The inadvertent use of prior knowledge in a generative cognitive task . Memory & Cognition 27 , 1 ( 1999 ) , 94 \u2013 105 . https : / / doi . org / 10 . 3758 / bf03201216 [ 62 ] Winter Mason and Duncan J . Watts . 2012 . Collaborative learning in networks . Proceedings of the National Academy of Sciences 109 , 3 ( Jan . 2012 ) , 764 \u2013 769 . https : / / doi . org / 10 . 1073 / pnas . 1110069108 [ 63 ] T . P . McNamara . 1992 . Priming and Constraints It Places on Theories of Memory and Retrieval . Psychological Review 99 , 4 ( 1992 ) , 650 \u2013 662 . https : / / doi . org / 10 . 1037 / 0033 - 295x . 99 . 4 . 650 [ 64 ] Philip Mendels , Joep Frens , and Kees Overbeeke . 2011 . Freed : a system for creating multiple views of a digital collection during the design process . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . ACM , Vancouver BC Canada , 1481 \u2013 1490 . https : / / doi . org / 10 . 1145 / 1978942 . 1979160 [ 65 ] A . Newell and H . A . Simon . 1972 . Human problem solving . Englewood Cliffs , NJ . [ 66 ] Bernard A . Nijstad and Wolfgang Stroebe . 2006 . How the group affects the mind : a cognitive model of idea generation in groups . Personality and Social Psychology Review 10 , 3 ( 2006 ) , 186 \u2013 213 . https : / / doi . org / 10 . 1207 / s15327957pspr1003 _ 1 00643 . [ 67 ] T . Okada , S . Yokochi , K . Ishibashi , and K . Ueda . 2009 . Analogical modification in the creation of contemporary art . Cognitive Systems Research 10 , 3 ( 2009 ) , 189 \u2013 203 . 00032 . [ 68 ] Srishti Palani , Zijian Ding , Stephen MacNeil , and Steven P . Dow . 2021 . The \" Active Search \" Hypothesis : How Search Strategies Relate to Creative Learning . In Proceedings of the 2021 Conference on Human Information Interaction and Retrieval . ACM , Canberra ACT Australia , 325 \u2013 329 . https : / / doi . org / 10 . 1145 / 3406522 . 3446046 Manuscript submitted to ACM 22 Chan , et al . [ 69 ] Srishti Palani , Zijian Ding , Austin Nguyen , Andrew Chuang , Stephen MacNeil , and Steven P . Dow . 2021 . CoNotate : Suggesting Queries Based on Notes Promotes Knowledge Discovery . In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems ( CHI \u201921 ) . Association for Computing Machinery , New York , NY , USA , 1 \u2013 14 . https : / / doi . org / 10 . 1145 / 3411764 . 3445618 [ 70 ] Amy Pavel , Floraine Berthouzoz , and Bjorn Hartmann . [ n . d . ] . Browsing and Analyzing the Command - Level Structure of Large Collections of Image Manipulation Tutorials . ( [ n . d . ] ) , 12 . [ 71 ] David N Perkins , David N . Perkins , and Margaret A Boden . 1994 . Creativity : beyond the Darwinian paradigm . In Dimensions of creativity . MIT Press , Cambridge , MA , 119 \u2013 142 . [ 72 ] P . Pirolli and S . Card . 1999 . Information foraging . Psychological Review 106 , 4 ( 1999 ) , 643 \u2013 675 . https : / / doi . org / 10 . 1037 / 0033 - 295x . 106 . 4 . 643 [ 73 ] Jeroen G . Raaijmakers and Richard M . Shiffrin . 1981 . Search of associative memory . Psychological Review 88 , 2 ( 1981 ) , 93 \u2013 134 . [ 74 ] Samuel Rhys Cox , Yunlong Wang , Ashraf Abdul , Christian von der Weth , and Brian Y . Lim . 2021 . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation . In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems ( CHI \u201921 ) . Association for Computing Machinery , New York , NY , USA , 1 \u2013 35 . https : / / doi . org / 10 . 1145 / 3411764 . 3445782 [ 75 ] Mark A . Runco . 2010 . Divergent thinking , creativity , and ideation . In The Cambridge handbook of creativity . Cambridge University Press , New York , NY , US , 413 \u2013 446 . https : / / doi . org / 10 . 1017 / CBO9780511763205 . 026 [ 76 ] Daniel M Russell , Malcolm Slaney , Yan Qu , and Mave Houston . 2006 . Being literate with large document collections : Observational studies and cost structure tradeoffs . In System Sciences , 2006 . HICSS\u201906 . Proceedings of the 39th Annual Hawaii International Conference on , Vol . 3 . IEEE , 55 \u2013 55 . [ 77 ] Daniel M . Russell , Mark J . Stefik , Peter Pirolli , and Stuart K . Card . 1993 . The Cost Structure of Sensemaking . In Proceedings of the INTERACT \u201993 and CHI \u201993 Conference on Human Factors in Computing Systems ( CHI \u201993 ) . ACM , New York , NY , USA , 269 \u2013 276 . https : / / doi . org / 10 . 1145 / 169059 . 169209 [ 78 ] Moushumi Sharmin and Brian P . Bailey . 2013 . ReflectionSpace : an interactive visualization tool for supporting reflection - on - action in design . In Proceedings of the 9th ACM Conference on Creativity & Cognition ( C & C \u201913 ) . Association for Computing Machinery , New York , NY , USA , 83 \u2013 92 . https : / / doi . org / 10 . 1145 / 2466627 . 2466645 [ 79 ] Moushumi Sharmin , Brian P . Bailey , Cole Coats , and Kevin Hamilton . 2009 . Understanding knowledge management practices for early design activity and its implications for reuse . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . ACM , Boston MA USA , 2367 \u2013 2376 . https : / / doi . org / 10 . 1145 / 1518701 . 1519064 00047 . [ 80 ] Pao Siangliulue , Kenneth C Arnold , Krzysztof Z Gajos , and Steven P Dow . 2015 . Toward collaborative ideation at scale : Leveraging ideas from others to generate more creative and diverse ideas . In Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing . 937 \u2013 945 . [ 81 ] Pao Siangliulue , Joel Chan , Steven P . Dow , and Krzysztof Z . Gajos . 2016 . IdeaHound : Improving Large - scale Collaborative Ideation with Crowd - Powered Real - time Semantic Modeling . In Proceedings of the 29th Annual Symposium on User Interface Software and Technology ( UIST \u201916 ) . ACM , New York , NY , USA , 609 \u2013 624 . https : / / doi . org / 10 . 1145 / 2984511 . 2984578 [ 82 ] Pao Siangliulue , Joel Chan , Krzysztof Gajos , and Steven P . Dow . 2015 . Providing timely examples improves the quantity and quality of generated ideas . In Proceedings of the ACM Conference on Creativity and Cognition . https : / / doi . org / 10 . 1145 / 2757226 . 2757230 [ 83 ] Ut Na Sio , Kenneth Kotovsky , and Jonathan Cagan . 2015 . Fixation or inspiration ? A meta - analytic review of the role of examples on design processes . Design Studies 39 ( July 2015 ) , 70 \u2013 99 . https : / / doi . org / 10 . 1016 / j . destud . 2015 . 04 . 004 00174 . [ 84 ] Steven M . Smith , T . B . Ward , and Jay S . Schumacher . 1993 . Constraining effects of examples in a creative generation task . Memory & Cognition 21 , 6 ( 1993 ) , 837 \u2013 45 . [ 85 ] Luis A . Vasconcelos , Maria A . Neroni , and Nathan Crilly . 2018 . The effect of explicit instructions in idea generation studies . AI EDAM 32 , 3 ( Aug . 2018 ) , 308 \u2013 320 . https : / / doi . org / 10 . 1017 / S0890060417000658 00009 Publisher : Cambridge University Press . [ 86 ] Shaun Wallace , Brendan Le , Luis A . Leiva , Aman Haq , Ari Kintisch , Gabrielle Bufrem , Linda Chang , and Jeff Huang . 2020 . Sketchy : Drawing Inspiration from the Crowd . Proceedings of the ACM on Human - Computer Interaction 4 , CSCW2 ( Oct . 2020 ) , 1 \u2013 27 . https : / / doi . org / 10 . 1145 / 3415243 [ 87 ] Hao - Chuan Wang , Dan Cosley , and Susan R . Fussell . 2010 . Idea expander : supporting group brainstorming with conversationally triggered visual thinking stimuli . In Proceedings of the 2010 ACM conference on Computer supported cooperative work - CSCW \u201910 . ACM Press , Savannah , Georgia , USA , 103 . https : / / doi . org / 10 . 1145 / 1718918 . 1718938 [ 88 ] YunlongWang , PriyadarshiniVenkatesh , andBrianYLim . 2022 . InterpretableDirectedDiversity : LeveragingModelExplanationsforIterativeCrowd Ideation . In CHI Conference on Human Factors in Computing Systems . ACM , New Orleans LA USA , 1 \u2013 28 . https : / / doi . org / 10 . 1145 / 3491102 . 3517551 [ 89 ] T . B . Ward . 1994 . Structured imagination : The role of category structure in exemplar generation . Cognitive Psychology 27 , 1 ( 1994 ) , 1 \u2013 40 . 00000 . [ 90 ] T . B . Ward . 1998 . Analogical distance and purpose in creative thought : Mental leaps versus mental hops . In Advances in Analogy Research : Integration of Theory and Data from the Cognitive , Computational , and Neural Sciences , Keith J . Holyoak , Dedre Gentner , and B . Kokinov ( Eds . ) . Sofia , Bulgaria , 221 \u2013 230 . 00142 . [ 91 ] Andrew M . Webb , Andruid Kerne , Rhema Linder , Nic Lupfer , Yin Qu , Kade Keith , Matthew Carrasco , and Yvonne Chen . 2016 . A Free - Form Medium for Curating the Digital . In Curating the Digital . Springer , Cham , 73 \u2013 87 . https : / / doi . org / 10 . 1007 / 978 - 3 - 319 - 28722 - 5 _ 6 [ 92 ] M . J . Wilkenfeld and T . B . Ward . 2001 . Similarity and Emergence in Conceptual Combination . Journal of Memory and Language 45 , 1 ( 2001 ) , 21 \u2013 38 . https : / / doi . org / 10 . 1006 / jmla . 2000 . 2772 [ 93 ] Xiaotong ( Tone ) Xu , Rosaleen Xiong , Boyang Wang , David Min , and Steven P . Dow . 2021 . IdeateRelate : An Examples Gallery That Helps Creators Explore Ideas in Relation to Their Own . Proceedings of the ACM on Human - Computer Interaction 5 , CSCW2 ( Oct . 2021 ) , 1 \u2013 18 . https : / / doi . org / 10 . 1145 / 3479496 00000 . Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 23 [ 94 ] HuanYuan , KelongLu , MengsiJing , CuirongYang , andNingHao . 2022 . Examplesincreativeexhaustion : Theroleofexamplefeaturesandindividual differences in creativity . Personality and Individual Differences 189 ( April 2022 ) , 111473 . https : / / doi . org / 10 . 1016 / j . paid . 2021 . 111473 00000 . [ 95 ] Darya L . Zabelina , Elina Zaonegina , William Revelle , and David M . Condon . 2021 . Creative achievement and individual differences : Associations across and within the domains of creativity . Psychology of Aesthetics , Creativity , and the Arts ( 2021 ) , No Pagination Specified \u2013 No Pagination Specified . https : / / doi . org / 10 . 1037 / aca0000439 Place : US Publisher : Educational Publishing Foundation . [ 96 ] Liang Zeng , Robert W . Proctor , and Gavriel Salvendy . 2011 . Fostering creativity in product and service development : validation in the domain of information technology . Hum Factors 53 , 3 ( 2011 ) , 245 \u2013 70 . [ 97 ] Yinglong Zhang , Rob Capra , and Yuan Li . 2020 . An In - situ Study of Information Needs in Design - related Creative Projects . In Proceedings of the 2020 Conference on Human Information Interaction and Retrieval . ACM , Vancouver BC Canada , 113 \u2013 123 . https : / / doi . org / 10 . 1145 / 3343413 . 3377973 00003 . Manuscript submitted to ACM 24 Chan , et al . APPENDIX A PROCEDURE FOR GENERATING RUGGED WILDCAT WELLS SEARCH ENVIRONMENTS We used four factors to control the synthetic objective functions : ( 1 ) The ruggedness amplitude , in the range of [ 0 , 1 ] , controlled the relative \" height \" of noise added to the search environment , compared to the height of the peaks . Increasing this parameter made the task more difficult by adding more places to incorrectly intuit as the location of the maximum reward . Setting this parameter to 1 would essentially make the search environment have infinite peaks . ( 2 ) The smoothness , in the range of [ 0 , 1 ] , controlled the degree of local correlation of scores in the grid . Intuitively , if smoothness was high ( closer to 1 ) , then a hill - climbing or gradient - based strategy could be viable , as a searcher that saw successive points in increasing score could ( correctly ) intuit that further points down that search path were likely to be of higher score ; at lower values of smoothness , the search environment became very \" bumpy \" , such that searchers would often be surprised by the score of nearby regions in the grid . ( 3 ) The number of peaks controlled the number of \" maxima \" in the search environment ; intuitively , this specified the number of regions in the space where a searcher might ( correctly or otherwise ) intuit as the location of the treasure . Mathematically , this parameter controlled the number of layers of multivariate normal with single peaks . ( 4 ) The distance between peaks , in the range of [ 0 , 1 ] , which prevented overlap of peaks when the function was generated with more than 1 peak . All search environments in this study were generated with ruggedness amplitude set to 0 . 7 ( fairly noisy ) , smoothness level to 0 . 2 ( fairly rugged ) , and the number of peaks to 1 with a maximum of 100 ; distance between peaks was not relevant because we only used a single peak , using the following algorithm : Algorithm 1 Constructing the Wildcat Wells search environment with given ruggedness ( noise ) amplitude ( \ud835\udc45\ud835\udc62\ud835\udc54 \ud835\udc4e\ud835\udc5a\ud835\udc5d ) , smoothness ( \ud835\udc46\ud835\udc5a\ud835\udc61 ) , number of peaks ( \ud835\udc41 ) and distance between peaks ( \ud835\udc45\ud835\udc62\ud835\udc54 \ud835\udc53\ud835\udc5f\ud835\udc52\ud835\udc5e ) . 1 : for \ud835\udc45\ud835\udc62\ud835\udc54 \ud835\udc4e\ud835\udc5a\ud835\udc5d , \ud835\udc46\ud835\udc5a\ud835\udc61 , \ud835\udc41 , \ud835\udc45\ud835\udc62\ud835\udc54 \ud835\udc53\ud835\udc5f\ud835\udc52\ud835\udc5e do 2 : Get \ud835\udc4b \ud835\udc50\ud835\udc52\ud835\udc5b\ud835\udc61\ud835\udc52\ud835\udc5f\ud835\udc60 = \ud835\udc53 ( \ud835\udc41 , \ud835\udc45\ud835\udc62\ud835\udc54 \ud835\udc53\ud835\udc5f\ud835\udc52\ud835\udc5e ) 3 : Sample (cid:205) \ud835\udc41\ud835\udc56 \ud835\udc60\ud835\udc62\ud835\udc5f\ud835\udc53 \u223c N ( \ud835\udc4b \ud835\udc56 ) 4 : Sample \ud835\udc41\ud835\udc5c\ud835\udc56\ud835\udc60\ud835\udc52 \u223c OpenSimplex ( \ud835\udc46\ud835\udc5a\ud835\udc61 ) 5 : end for 6 : Return \ud835\udc60\ud835\udc62\ud835\udc5f\ud835\udc53 + \ud835\udc5b\ud835\udc5c\ud835\udc56\ud835\udc60\ud835\udc52 \u00d7 \ud835\udc54 ( \ud835\udc45\ud835\udc62\ud835\udc54 \ud835\udc4e\ud835\udc5a\ud835\udc5d ) . Manuscript submitted to ACM Interface Presentation Effects on Example - Based Problem Solving 25 B ALGORITHM FOR SAMPLING DIVERSE AND NON - DIVERSE EXAMPLE POINTS Algorithm 2 Generating a ranked distribution of example sets with Determinantal Point Process ( DPP ) approach [ 49 ] , \ud835\udc40 is batch size ( 10000 ) . \ud835\udc46 \ud835\udc58 is a combinatorial set defined on a finite set \ud835\udc4b \u2208 R 2 , where each element \ud835\udc46 \ud835\udc58\ud835\udc4c \ud835\udc56 \u2208 \ud835\udc46 \ud835\udc58 is k elements long . 1 : for \ud835\udc56 \u2208 \ud835\udc5f\ud835\udc4e\ud835\udc5b\ud835\udc54\ud835\udc52 ( \ud835\udc40 ) do 2 : Sample \ud835\udc46 \ud835\udc58\ud835\udc4c \ud835\udc56 \u223c IID ( \ud835\udc46 \ud835\udc58 ) [ identically sampling unordered sets without replacement ] 3 : Calculate \ud835\udc54 ( \ud835\udc46 \ud835\udc58\ud835\udc4c \ud835\udc56 ) = \ud835\udc54 \ud835\udc66 \ud835\udc56 and append this to \ud835\udc46\ud835\udc50\ud835\udc5c\ud835\udc5f\ud835\udc52\ud835\udc60 \ud835\udc46 \ud835\udc58 4 : end for 5 : Return DPP Score of sets of examples = \ud835\udc46\ud835\udc50\ud835\udc5c\ud835\udc5f\ud835\udc52 \ud835\udc46\ud835\udc58 \u2212 \ud835\udc5a\ud835\udc52\ud835\udc4e\ud835\udc5b ( \ud835\udc46\ud835\udc50\ud835\udc5c\ud835\udc5f\ud835\udc52 \ud835\udc46\ud835\udc58 ) \ud835\udc60 . \ud835\udc51 . ( \ud835\udc60\ud835\udc50\ud835\udc5c\ud835\udc5f\ud835\udc52 \ud835\udc46\ud835\udc58 ) . Manuscript submitted to ACM",
    "purcellDesignOtherTypes1996": "Design and Other Types of Fixation or Is Fixation Always Incompatible with Innovation ? A . T . Purcell and J . S . Gero Key Centre of Design Computing Department of Architectural and Design Science Sydney University , 2006 Australia { terry , john } @ arch . su . edu . au Abstract Design educators often comment on the difficulties that result from a premature commitment by students to a solution to a design problem . Similarly practitioners can find it difficult to move away from an idea they have developed or precedents in a field . In the psychology of problem solving this effect is called functional fixedness or fixation . This refers to the observation that people find it very difficult to see , for example , that objects with well known uses or functions can be employed in new or unusual ways to develop an innovative solution to a problem or to solve a problem that requires innovation . Given that fixation refers to situations where innovation is blocked , it is not surprising that these effects should occur in design problem solving . Design problems are essentially ill - defined problems . As such they inherently contain the opportunity for innovation . However , while these types of issues have been discussed in the context of design , there has been little systematic evidence available about whether or not and under what conditions design fixation does occur . The paper reviews the results of a series of recent experiments which begin to addresses these issues . The results of the experiments will then be examined in terms of what insights they provide into the design process and what implications they have for design education and their relevance for the role that computers may play in the design process . 1 . Introduction Through an unusual inversion of everyday ways of thinking , the Gestalt psychologists ( Maier , 1931 ; Wertheimer , 1982 ) sought to understand innovation and creativity in problem solving by studying problems that most people find it difficult to solve . The basic logic attached to this approach appears to have been as follows . If problems are chosen which require innovative solutions , then studying the conditions under which people fail can give insights into why people find it difficult to produce innovative solutions . One of the central concepts to emerge from this approach was the idea of fixation or functional fixedness . For example , people appear to be unable to see new ways of using objects which could lead to an innovative solution to a problem , because they are blocked or fixated on well learnt uses or properties of the object . A typical problem where fixation is exhibited requires that two spatially separated pieces of string hanging from the ceiling of a room be joined . The pieces of string are not long enough for one to be simply picked up and carried to the other . Present in the situation are a variety of everyday objects . Many of these objects are capable of being used as a weight which can be combined with one of the pieces of string to form a pendulum . If a pendulum created in this way is set in motion , it 2 is possible to pick up the end of the other piece of string , catch the end of the pendulum and tie the two together . However this solution is rarely spontaneously produced . The reason for this result , it is argued , lies in the way that the well established , everyday functions of objects prevent the problem solver from seeing this unusual and innovative use . This basic idea is appealing because it taps experiences common to many people and to problem solving in many domains . The area of design is no exception . Here a common and often commented on form of fixation is the premature commitment to a particular problem solution , observed in students and practitioners alike . As in other domains , the designer appears trapped by the characteristics of a possible solution that has been developed or an existing precedent solution . However , in the design domain , the majority of the discussion of this phenomena is essentially anecdotal and not based on either principled argument or the results of empirical research . Design would also appear to offer a particularly suitable domain for this approach to studying these aspects of problem solving . It is well recognised that design problems are inherently ill - defined ( Simon , 1981 ) . Consequently every design problem has the potential for innovative and creative solutions removing the necessity for specifically developing problems that have these characteristics . While all discussion of fixation in design has been largely anecdotal , recently experimental methods have been developed for examining both whether or not the phenomena exists and , perhaps of more interest , the basis for the effect . The aim of this paper is to review the available research results and then to draw out the implications of these results in three areas . First what do the results tell us about the design process ? If there are insights into the design process what are the implications for design education ? Finally there is the question of the implications of the design fixation effect , ( if it exists ) , for the role of computers in the design process . Broadly and briefly computers would appear to potentially have two types of roles in design . In one the aim is to have the computer produce the design ; in the other the computer is used to provide intelligent design support system for a human designer . The first role for computers appears to have been downgraded in the past few years , largely because of the consequences of the ill - defined nature of design problems , while the second has received increasing attention . A central facet of the second computational approach to design would appear to be the reuse in some way of existing designs as in case - based systems . In this approach , a large library of designs of different types and their associated attributes is established . In the context of a particular design problem this library is searched for designs which match , using some measure of similarity , to the requirements of a particular design problem with the human designer specifying what these requirements are . Fixation effects could occur in this situation in a number of ways . However , if the design fixation effect exists , it could be particularly relevant to the development and use of case - based systems . The available evidence indicates that design fixation may be associated with seeing pictorial representations of possible design solutions . It would also appear to be highly likely that the design cases would contain a number of forms of pictorial representation from pictures of the completed design to various types of diagrammatic representations . The form of representation used in such systems would as a result appear to establish the conditions for fixation to occur . However , before exploring this issue further , whether or not fixation occurs and the basis of the effect needs to be established . 2 . Design Fixation Jansson and Smith ( 1991 ) were the first to develop an experimental approach to the problem of fixation in design . They argued that showing designers a picture of a potential design solution to a problem prior to a design session should result in fixation . In effect 3 the picture would act as a precedent , blocking access to other ways of solving the problem . They also extended the argument about the basis of fixation . They suggested that the process of design involves operating on effectively two types of mental representation of the problem . One representation they refer to as the conceptual space which consists of abstract knowledge about principles , concepts and rules which can be used to solve the problem . The other representation takes the form of particular physical objects and elements which could form the physical realisation of a solution to the problem . This representation is referred to as the object space . Jansson and Smith argue that the location of the fixation induced by a pictorial representation is the object space and that innovation is prevented because the designer cannot move to the conceptual space which is where they consider that innovative changes can occur . In order to test this hypothesis Jansson and Smith used three different types of design problem and had advanced undergraduate and practicing mechanical engineers engage in solving the problem . The experimental design was quite straight forward . For each of the design problems two groups of designers were used . One group ( either a student or an expert group ) were simply given a statement of the problem and acted as a control group . Separate groups corresponding in level of expertise to the control groups were given the statement of the problem together with a picture of a possible solution . The example solutions were specifically designed so that aspects were included which were incorrect given the problem statement . A set of features which characterised each pictorial example was developed and the designs produced were scored for the presence or absence of the features . The design sessions lasted for one hour and the designers were allowed to produce as many sketch designs as they liked . It would be expected that , if fixation resulted from exposure to a pictorial representation of a possible design , more features of the example should occur with the groups shown the example . This was found to be the case . For each of the design problems and for student and expert designers , more features associated with the example , including incorrect features , were found with those who had been shown the pictorial representation . 3 . Exploring the Design Fixation Effect The potential significance of this effect is apparent because of the role of precedents , generally presented in the form of pictorial and other forms of visual representations , in the teaching and practice of design . A number of significant issues relating to the effect can be identified . Does the effect occur when pictorial representations of different possible designs are presented prior to a design session ? Is the effect associated specifically with pictorial representations or would similar effects occur if verbal descriptions of possible design solutions were made available ? Does the effect occur with other disciplines and levels of expertise ? In order to begin to explore the basis for the design fixation effect we ( Purcell and Gero , 1991 ) carried out an experiment using one of the Jansson and Smith problems - the design of a bicycle rack for a car . Five different design solutions were identified and these are shown in Figure 1 . Pictures of three of these designs were using as fixating examples ( a , b and c in the Figure ) and for two of these solutions a verbal description was developed ( the verbal description of one of the designs is presented in Appendix 1 ) . Novice students in architecture and industrial design participated in the experiment using the same experimental design as Jannson and Smith . Separate groups of students were shown the example design and were told that it was to illustrate what was meant by a sketch design with corresponding control groups being given only the statement of the problem . Lists of features of each design were developed and the designs produced by the experimental and control groups were scored for the presence or absence of the features . In addition the number of designs developed by each participant was recorded and the participants answered a number of questions at the completion of the session . 4 Differences in the frequency of design features were statistically analysed and part of the results are presented in Table 1 . Effectively no evidence of fixation was found with the exception of one of the pictorial examples for one feature . The fixation effect , on the basis of these results , does not appear to be associated with simply the presentation of a pictorial example nor does it occur with a verbal description of a possible design . Where evidence for fixation was found with one of the examples , an analysis of the results of one of the questions asked at the end of the design session indicated a somewhat different possible basis for the results obtained . Participants were shown the drawings of all of the examples shown in Figure 1 and asked how familiar they were with each . Tabulation of the responses to this question revealed that the most familiar example was also the example where some evidence of fixation was obtained ( see Table 1 ) . What appears to be a fixation effect could as a result simply reflect familiarity with the example . Subsequent enquires with bicycle retailers and a group representing bicyclists demonstrated that the most familiar example was also the type of bicycle rack that predominated in sales . Interestingly this type of design was also different to the rack design that had produced fixation in Jansson and Smith\u2019s work . There is however significant differences between our work and the original Jannson and Smith experiments . They had used advanced student and practicing mechanical engineering designers and we had used novice designers from two other disciples . A possible interpretation of the effect obtained could therefore be that novice designers from these other disciplines , lacking any domain specific knowledge , simply relied on their general , everyday knowledge which was triggered when shown the drawing of the example . This possible explanation is reinforced by the absence of any effects with the drawings of the other types of designs . This result also indicates that the effects obtained with the most familiar example do not simply involve copying as it would have been equally easy to copy each of the other types of design . 4 . The Role of Design Discipline and Expertise in the Design Fixation Effect Our next experiment attempted to examine some of these issues . We again used one of the Jansson and Smith problems - the design of a device to be used by the blind in measuring quantities for cooking . We chose this design problem because it would be unlikely that our designers would have either designed or seen an example of a design for such a device in contrast to the bicycle rack problem . In this way we were able to minimise the effects of familiarity with existing design solutions . Advanced student designers in mechanical engineering and industrial design in their final year participated in the experiment . This choice was made both because we wanted to use the designers from the same discipline as in Jansson and Smith\u2019s work and because we wanted to examine whether or not the effect occurred with different disciplines while using designers with a similar level of expertise to those in the original Jansson and Smith experiment . Industrial designers were chosen as a second discipline because the two disciplines can deal with similar types of problems . This removes a possible effect resulting from using a design discipline , such as architecture , which deals with qualitatively different types of design problems to the problem we had chosen . We also pursued the question of whether or not fixation would be found with other examples to that used by Jansson and Smith . In Australia there is a organisation which represents the blind and we approached them to determine if there were examples of such a device available . They market a device for this propose and the pictorial representations of the two examples are shown in Figure 2 . 5 Again the same experimental design was used as in previous design fixation work . We did however introduce a change in the way the designs were analysed . In previous work features characteristic of each design were identified and each design produced was scored in terms of these features . However , in reflecting on the results of these experiments , it appeared to us that features could be of two types . One type would represent the actual details of the features present in the design example . However it is possible that a design feature could represent the same concept as was present in the design example without having the same detailed characteristics . Consequently for each design two lists of features were developed to reflect this distinction ( for details of the features of each of the designs and the results of the analysis see Purcell , Williams , Gero and Colbron , 1993 ) . In brief the analysis demonstrated that fixation did not occur with the industrial design groups for either of the example designs . However fixation was apparent with the mechanical engineers but only with the example design used in the original Jannson and Smith experiments . This appears to indicate that fixation may occur only where the combination of the design discipline and the example design used are the same as in the original Jansson and Smith experiments suggesting that the effect may only have limited applicability . The presence of fixation in the mechanical engineers and not in the industrial designers could for example reflect differences in approaches to education in the two disciplines . While this would be an interesting result , it would diminish the general significance of the effect . However an examination of the two example designs in Figure 2 suggests another possible basis for this finding . The Royal Blind Society example is a particularly simple device that essentially uses an everyday cup with three dimensional markers identifying discrete quantities . By contrast the Jannson and Smith example is a more complex device which , significantly , appears to involve principles that could be thought of as typical of the mechanical engineering discipline . This suggests the possibility that the fixation effect is not simply discipline specific but involves the use of an example that embodies principles that are specific to a particular domain of knowledge . 5 . Domain Specific Knowledge and Design Fixation In order to test this hypothesis we sought the assistance of experts in mechanical engineering and industrial design . Our aim was to develop a new design problem which would again be the type of problem that could form a part of the practice of each discipline but where it would be unlikely that groups of advanced undergraduate students would have been involved in designing or have seen examples of solutions . We also wanted a problem where there were a number of existing solutions which both involved principles which would be typical of mechanical engineering and also examples which the experts considered unusual and innovative . The problem we identified was the design of a device for assisting the elderly into and out of a bath in a domestic setting . There are a number of existing solutions to this problem four of which are shown in Figure 3 . To initially examine our hypothesis about the role of domain specific knowledge being embodied in the pictorial representation of a design , we chose the example , the Autolift device , labelled as ( c ) in the Figure . We had asked the experts in mechanical engineering to assess whether or not the examples involved typical mechanical engineering principles . In their judgement this example involved the use of such principles and should therefore produce fixation . Our experiment followed the same design as previously . Advanced undergraduate students in their final year in mechanical engineering and industrial design participated in the experiment with two groups from each discipline . Both groups in each discipline received the verbal description of the problem with one group in each discipline also being shown the drawing of the example design . Lists of features related to the principles involved in the example design and the details of the design were again developed . However we 6 considered that measuring the frequency of occurrence of design features was not the only way of assessing design fixation . For example , it could be argued that design fixation should lead to fewer designs being produced because other ways of solving the problem would not be available to the designer . Similarly fixation would be expected to produce a restriction in the range of solution types that were produced . In order to assess these other aspects of fixation , the number of designs produced by each participant was again recorded together with the type of each design produced by each participant . Design type was assessed in the following way . Brief descriptions of approximately 20 % of all the designs were developed independently by two of the researchers working on the project . Similar descriptions were then grouped together to form categories which were then given type labels . This initial list of types was tested on a further sub - set of the designs before the full set of designs was coded . Designs which did not fit into any of the types were classified as \u201cother\u201d . At the completion of the coding the \u201cother\u201d category was reviewed and new categories developed with the \u201cother\u201d category finally referring to unique designs ; that is only one example of the type in the complete data set . The results of the analyses of these various measures are present in Tables 2 , 3 and 4 . Table 2 ( with proportions rather than raw frequencies being reported ) demonstrates quite clearly that fixation had occurred with the mechanical engineering group for both features that involved principles and features involving detail . Of the seven principle features , significant differences were found between the experimental and control groups for the mechanical engineers for five of these features but only for one of these features in the industrial design groups . For the twelve detail features , the numbers of significant differences were eight and three respectively for the two disciplines . Exposing a group of mechanical engineering designers to a pictorial representation of an example design which embodies typical principles of the discipline does , as a result , appear to induce a fixation effect . However this result is contradicted by the analyses of the other two measures of fixation . A comparison between the average number of designs produced showed no differences between the experimental and control conditions for each design discipline ( see Table 3 ) . If fixation was operating to constrain the types of solutions available to the designer such differences would have been expected for the mechanical engineering group . Similarly fewer types of design solutions should have been found where fixation occurred however this was not the case as is apparent from Table 3 . The results of the analysis of the various types of data in this experiment are therefore somewhat paradoxical . In terms of the usual , feature based measures there is quite clear evidence of fixation with the predicted experimental conditions for the mechanical engineers . However with the two other measures , which would appear to be equally as reasonable as the features approach in terms of demonstrating fixation , no evidence for fixation occurred . 6 . Design Fixation and Innovative Design Examples There are a number of other aspects of the results which however are also noteworthy even though they , at least superficially , do not appear to be specifically related to design fixation . Over all of the experiments using advanced student designers there has been no evidence of fixation with the industrial design groups . In this experiment it is apparent that there are large differences between the mechanical engineering groups and the industrial designers . From Table 3 it is clear that the industrial designers produce approximately twice as many designs and twice as many types of design solutions as the mechanical engineers . It is also apparent from Table 4 that , for the industrial designers , by far the most frequently 7 occurring type of design is \u201cother\u201d . That is this group frequently produced , whether under the control or experimental conditions , unique , one off designs and these types of designs were very infrequently found with the mechanical engineers . These differences between the two groups , because they are effectively independent of the experimental manipulations carried out , could be interpreted , for example , as resulting from differences in educational processes in the two disciplines . It could be that industrial design emphasises creativity and difference and associates this with the idea of trying out many ideas . The emphasis on innovation and trying many different ideas would clearly produce many more designs , many more types of design and many more unique designs . However , while these aspects of the results appear to be unrelated to design fixation , they suggest another perspective on what was occurring with the design fixation results . While industrial design may place different emphases on what is important to the outcome of a design process , it is also apparent that the areas of knowledge that make up industrial design are more diverse than those studied in mechanical engineering . It would also appear that many of these areas are associated with less well articulated bodies of knowledge than those that make up the knowledge base of mechanical engineering . For example , aesthetics plays a prominent role in industrial design education , often appearing as a separate subject , while it plays little formal role in mechanical engineering . Aesthetics also has an obviously less articulated and formal knowledge base than for example mechanics in mechanical engineering . It is possible therefore that the absence of highly articulated , formal knowledge bases in industrial design results in the absence of fixation that has been demonstrated in our experiments . While this is a plausible , but after the event , explanation for these results , there is an alternative . It could be that it is not the absence of a highly articulated knowledge base which results in no fixation being found , but that the example used does not emphasise aspects which are considered important in industrial design . If innovation is considered important then the example used in the previous experiments , which was judged by experts from both disciplines as not being innovative , would not emphasise issues which were considered to be important in industrial design . This argument leads to the hypothesis that it may be possible to produce fixation in industrial designers by using an innovative example . One of the designs that our panel of experts considered to be innovative is example b , the Hydrocushion , shown in Figure 3 . This essentially consists of a cushion which can be filled with water to the level of the bath edge to allow access , with the water then being emptied into the bath to lower the person with the process being repeated to exit the bath . If innovation is an important issue in the knowledge base of industrial designers , then this example should produce fixation . For mechanical engineers , the principles involved are not as typical , in the context of lifting weights , as are the principles involved in the Autolift example used in the preceding experiments ( the second example shown in Figure 3 ) . If fixation did not occur with the mechanical engineers it would indicate that the principles involved in an example need to be typical of the task to be performed , whereas if fixation does occur , it would indicate that activation of relevant , although not typical ( in the context of lifting weights ) mechanical engineering principles can produce fixation . We therefore repeated our basic experiment using this example . Table 5 presents the features which were scored for each design that was produced . These are again divided into principle and detail features with an additional group of features relating to aspects of the context that is depicted in the example drawing . In this design example we identified six principle features . However an examination of the list demonstrates a number of the features were not unique to this particular type of design with these features resulting from general constraints on the design eg a removable device or a device which can raise or lower the person in the bath . In effect what characterises this design and makes it innovative is the use a single basic principle - the use of water pressure which is already 8 present in the design context - to resolve many of the design issues . This contrasts with the lists of features associated with the Autolift example . Here the principle features actually identify a series of principles which solve various parts of the problem . The detail features are again those that directly reflect the specific ways in which the principled features are realised in the fixation example . The contextual features are those which are present in the fixation example but either represent expected parts of the design context eg tiles on the adjacent wall or are ancillary drawings which illustrate other views of the device and ways in which the device can be used . Table 5 also presents the proportions representing the occurrence of each of these features in the experimental and control groups for the two design disciplines . Because of the small pool of final year industrial design students that are available in Sydney , only thirteen individuals have participated in this part of the experiment to date with however twenty mechanical engineering students participating . As a result of the unequal numbers in the groups , statistical analyses have not been carried out yet . In order to allow a comparison between the results of the two experiments , the results are expressed in terms of proportions rather than raw frequencies with the potentially significant differences ( based on previous analyses ) being identified by asterisks . A comparison between Tables 3 and 5 identifies a number of interesting differences and similarities between the two experiments . With the Autolift device and the mechanical engineering group , there was clear evidence of what appears to be fixation in the significant differences between the experimental and control groups on the majority of both the principle and detail features . This pattern does not appear to occur with the innovative design example with this group . One principle feature - operates on water pressure - is clearly different between the experimental and control groups with no designs in the control group exhibiting this feature . The features relating to raising and lowering the individual and connecting the device to the existing water tap may demonstrate an effect . With the detail features only one - the presence of a rectangular platform as part of the design - may be significant . It is also apparent that many of these detail features , which show no evidence of differences , are very specifically related to this type of design and are not likely to be present in other types of designs . For example the last six of the detail features are quite specifically associated with a device of the type pictured . However these features are effectively missing from the designs that were produced . The results using the innovative example with this group therefore indicate that fixation is associated with the principles involved in the example design and that the designs produced do not reflect the details of the example . The pattern of results with the industrial design student group is however quite different . There may be a difference between two of the principle features - the device is used to raise and lower the individual from the edge of the bath and the device is removable and two of the detail features - the device has a platform seat with rounded corners . However it is apparent that these features are not specifically associated with the example device that uses water pressure and a water filled container . It therefore appears that , for this group , there is essentially no indication that any of the devices produced in either the experimental or control groups used this innovative approach to the design of the device because neither the principled or detailed features that are specifically associated with this type of design are present in the designs produced . The absence of any fixation effect with this group is therefore similar to the results in the previous experiment however the implications of the result in this case may be quite different . The results with the contextual features are also of interest . For the mechanical engineers differences appear to be associated with two of the types of features which would be consistent with the use of a water filled device - placing the device at the opposite end of the bath to the taps and showing a single bath tap . It is also apparent that many more potentially significant differences are associated with the contextual features than with 9 either the principle or detailed features . Further many of these differences are associated with what might be referred to as conventions associated with the development of a sketch design . For example , the set of features which involve using additional insets which illustrate how the device can be used are quite typical techniques that are used during design development . The reasons why more of these types of features appear in the results of those groups shown the design example across both design disciplines is not immediately apparent . 7 . Conclusions Design Fixation , Design Education and the Design Process The results of this final experiment have potentially important implications . The initial conception of design fixation was that it represented an impediment to innovative design . Designers shown an example demonstrated a lack of flexibility in their design process because they reproduced the characteristics of the example shown . Often this lack of flexibility also extended to reproducing faults that were present in the example design . While this effect resulted from being shown an example design , it could be extended to the more typical design situation . Designers often look at precedent designs and could become fixated in this way . Even more typically , designers , once they have produced a drawing of a potential solution , could become fixated on that solution - an effect often commented on by designers and by those who teach design . However the results of this series of experiments point to a different perspective on the fixation effect . Fixation appeared in our initial experiments to be associated , if it occurred at all , with the absence of domain specific knowledge and a reliance on everyday knowledge activated by exposure to a picture of a familiar example . When the familiarity effect is removed and the type of problem is matched to the discipline of the designer , the results indicated that fixation only occurred with one discipline - mechanical engineering - and only with an example design that embodied principles that formed a part of the knowledge base of that discipline . With the other discipline - industrial design - there was very little evidence of fixation . The role of principles specific to a discipline in producing fixation was demonstrated by using a problem that would not involve the familiarity effect combined with an example solution which was judged by experts to involve knowledge that would be typical of the mechanical engineering discipline . However , while this effect was found when features associated with the example were analysed , two other measures of fixation - the number of designs produced and the number of solution types - showed no fixation effects . Again no fixation effects were found with the industrial designers but there were notable differences between the two disciplines . Industrial designers produced many more designs , many more types of design and many more one off designs . These effects were independent of whether or not an example design was shown . While these differences between the two disciplines could reflect differences in educational processes , it was argued that the absence of the effect with the industrial designers could result from the use of example designs that did not embody aspects relevant to that discipline . The fact that the participants from this discipline produced many more designs , many more types of designs and many more one off designs suggests that there is a focus on innovation in this discipline . If this is correct an innovative example design should produce fixation for this discipline . Once again however no evidence was found for fixation with this discipline but evidence for fixation was again found with the mechanical engineers . 10 However consideration of the attributes of the innovative design and the results with both disciplines suggests that fixation as it has traditionally been conceived may not be the appropriate way to look at the results using this example . The example design used is innovative both because it uses a principle which is not typical of the general approach to lifting weights and because it uses a single principle to resolve a number of the design issues . The results with mechanical engineers demonstrated that what we have normally referred to as fixation was really associated with a concentration predominantly on the core innovative principle involved in the example design - the use of water pressure to raise and lower the individual - with many of the specific aspects of the example design not being found in the designs produced . While the industrial designers showed no evidence of fixation , they also failed to produce any designs which used this innovative approach in either the experimental or control groups . This way of looking at the results raises a number of issues . Is it that the use of typical examples produces the traditional fixation effect while the use of innovative examples focuses the designer on the principles involved ? Does this then result in designs that explore and develop the application of this innovative principle ? We are currently investigating aspects of these designs to determine if they exhibit these characteristics . If this is the case it suggests that , for example , at least part of the use of precedents in design education should involve the deliberate use of and focus on innovative example designs . While innovative designs are discussed and analysed in design education , it may be that this is not effective and that the use of such designs should take the form of showing them prior to a design session with the analysis coming after the session has been completed . The absence of either type of fixation effect with the industrial designers also raises a number of particularly interesting issues . The production of large numbers of designs of many different types would appear to indicate that these designers were attempting to be innovative . The absence of any fixation in the experiments where the \u201ctraditional\u201d fixation effect is found with the mechanical engineers would appear to indicate that this approach perhaps prevents the occurrence of fixation effects of this type . However it is possible that what occurs in fact is a search for difference rather than innovation . The absence of any examples of the innovative design approach represented by the example in the final experiment with this group could reflect just such a process of searching for difference . If this is the case and it represents the results of the educational processes used in this discipline , it indicates that the potential benefits available from innovative design will not be available to the discipline . The results of our research to date may be summarised in the following way . Fixation in the traditional sense may well be found where designers are forced to rely on everyday knowledge . Mechanical engineers become fixated in the traditional sense when the example they are shown embodies typical principles which are characteristic of the knowledge base of the discipline . When shown an innovative example , where the principle involved is unusual in the context and / or resolves many of the design issues using a single principle , mechanical engineers become \u201cfixated\u201d on the principle involved and appear to then explore ways of solving the problem using the principle . Industrial designers appear to show no evidence of fixation under any of the experimental conditions we have employed . However , while showing no evidence of \u201ctraditional\u201d fixation , the industrial designers showed no evidence of producing innovative designs using the principle involved in the innovative example . In a sense these groups may have become \u201cfixated\u201d on being different . \u201cFixation\u201d therefore appears to possibly exist in a number of forms and we as researchers need to be wary of becoming fixated on our conception of what fixation is . Design Fixation and Computational Approaches to Design 11 Given these results from our research it is possible to return to the issue of the relationship between fixation effects and computational approaches to design discussed at the end of the Introduction to this paper . A major finding of this research was that there are differences between the two design disciplines studied in our project . With the mechanical engineers , fixation in the traditional sense of reproducing the characteristics of a design , including incorrect features , occurred where the example shown embodied principles that were typical of the knowledge base of the discipline . If the cases included in a case - based mechanical engineering design support system both embodied principles typical of the discipline and either contained mistakes or features that were inappropriate in relation to the particular design problem being addressed , then fixation effects that are negative in terms of the design outcome for the particular problem are likely to occur . However , if the design examples that are presented are innovative , it appears likely that designers may identify the principle involved and then explore how this could be used in the particular design situation . This latter outcome points towards ways in which such systems might be used to both encourage innovative design and prevent design fixation effects . This could be done in at least two ways . The cases that were included in the system could be deliberately selected so as to include a range of examples that went from the typical to the atypical within any particular type of artefact . In addition the similarity metric which is used to identify relevant cases could be designed so that cases of other types that have relevant features could also be retrieved providing further possibilities for exploration of the implications of these ideas and increasing the possibility of innovation in design . These approaches however would not necessarily prevent the first type of fixation occurring . For example if the cases that are retrieved first all embody typical principles in the field then fixation of the first type could occur with the designer not proceeding to explore the further possibilities of the system . This suggests that it would also be necessary to carefully consider the issue of the order in which such a system retrieves cases and that the process of learning to use such systems should specifically be designed to make users aware of these issues . The results with the industrial designers have equally important but more general implications for the design of such systems . The evidence from our work is that presenting pictorial examples has no effect on the designs produced by the participants from this discipline . Taken at face value this would indicate that there would be little point in developing a case - based design support system for this discipline as it would be unlikely to have any effect . In fact our results would lead to the prediction that such a system would be unlikely to be used as a result of designers from this discipline being \u201cfixated\u201d on difference . This in turn raises some very basic issues about the education of industrial designers which cannot be addressed here but which point to the need for further research in the area of industrial design to establish the generality of the effects we have observed . References Jansson , D . G . and Smith , S . M . ( 1991 ) Design fixation . Design Studies , 12 , 3 - 11 . Maier , N . R . F . ( 1931 ) Reasoning in humans : II . The solution of a problem and its appearance in consciousness . Journal of Comparative Psychology , 12 , 181 - 194 . Purcell , A . T . and Gero , J . S . ( 1991 ) The effects of examples on the results of a design activity . In Artificial Intelligence in Design \u201891 Ed . J . S . Gero , Butterworth - Heinemann : Oxford , pp 525 - 542 . 12 Purcell , A . T . , Williams , P . , Gero , J . S . and Colbron , B . ( 1993 ) Fixation effects : Do they exist in design ? Environment and Planning B : Planning and Design , 20 , 333 - 345 . Purcell , A . T . , Gero , J . S . , Edwards , H . M . and Matka , E . ( 1994 ) Design Fixation and Intelligent Design Aids . In J . S . Gero and F . Sudweeks ( Eds ) Artificial Intelligence in Design \u201894 , 483 - 495 , Kluwer Academic Publishers , The Netherlands . Simon , H . A . ( 1981 ) The Sciences of the Artificial ( 2nd . edition ) M . I . T . Press : Cambridge , Massachusetts . Wertheimer , M . ( 1982 ) Productive Thinking . University of Chicago Press : Chicago . Acknowledgments The research reported in this paper was supported by Grant Number A89231572 from the Australian Research Council to A . T . Purcell and J . S . Gero . Part of the material presented in Section 3 of the paper was reported in Purcell and Gero ( 1991 ) ; part of the material in Section 4 was reported in Purcell , Williams , Gero and Colbron ( 1993 ) and part of the material in Section 5 was reported in Purcell , Gero , Edwards and Matka ( 1994 ) . 13 Figure 1 Illustrations of five bicycle rack designs . 14 Figure 2 Illustrations of the two example designs of devices for use by the blind in measuring quantities for cooking . 15 Figure 3 Illustrations of four devices designed to assist the elderly in entering and leaving a bath . Table 1 Probability of the difference in frequency of occurrence of four bicycle design features between experimental and control groups . 16 Group Location of rack Car attachment Bicycle support Bicycle attachment Control versus single - post picture 0 . 36 0 . 38 0 . 53 0 . 17 Control versus single post description 0 . 83 0 . 88 0 . 72 0 . 06 Control versus A - frame picture 0 . 79 0 . 64 0 . 21 0 . 02 Control versus A - frame description 0 . 3 0 . 13 0 . 61 0 . 12 Control versus boot picture 0 . 89 0 . 99 0 . 86 0 . 26 Table 2 Proportions of two types of features found for the experimental and control groups in two design disciplines for the bath access problem with the Autolift fixating example ( statistically significant differences ( . 05 ) are identified with an asterisk ) . Mechanical Engineers Industrial Designers ExperimentalControl Experimental Control Principles Fixed * . 72 . 42 . 52 . 34 Fixed to floor * . 25 . 06 * . 16 . 04 Column * . 36 . 06 . 14 . 05 Lifting mechanism within * . 23 . 03 . 06 . 04 Handle on column . 03 . 00 . 03 . 00 Boom * . 31 . 03 . 11 . 06 Seat . 67 . 44 . 63 . 55 Details Fixed with base plate * . 22 . 00 * . 09 . 01 Fixed with bolted base plate * . 14 . 00 . 05 . 00 Column - tripartite . 00 . 00 . 03 . 00 Bolts on column . 00 . 00 . 02 . 00 Winder with knob . 03 . 00 . 02 . 00 Rigid boom * . 28 . 00 . 08 . 05 Moulded seat with back * . 44 . 14 * . 38 . 10 Perforated seat * . 17 . 03 . 09 . 03 Arms on seat * . 31 . 08 . 09 . 05 Incorrect orientation to taps . 08 . 03 . 03 . 00 Orientation of bath as in example * . 44 . 11 * . 48 . 13 Tiles on bath as in example * . 25 . 00 . 05 . 00 Table 3 Number of designs and number of solution types produced by the experimental and control groups in two design disciplines for the bath access problem , Autolift fixating example . Industrial Design Mechanical Engineering Control Autolift Inflatable Cushion Control Autolift Inflatable cushion Average number of designs 3 . 8 3 . 4 3 . 1 1 . 8 1 . 7 2 . 1 Average number of solution types 3 . 2 2 . 9 2 . 8 1 . 7 1 . 5 2 . 0 17 Table 4 Frequencies of solution types for the bath access problem produced in the experimental and control groups in two design disciplines , Autolift fixating example . Solution Type Mechanical Engineers Industrial Designers ExperimentalControl Experimental Control Ledge 2 0 2 4 Static rim seat 0 0 1 3 Swivel seat 5 5 9 6 Slide 0 0 1 2 Inflatable cushion 0 2 1 9 Scissor lift 0 4 2 7 Overhead pulley 9 4 2 5 Boom seat with column 9 0 4 2 Door in side of bath 1 1 2 1 Other bath alteration 0 2 3 2 Step and rail 2 5 5 6 Tracks and racks 3 4 7 2 Grip rails 1 4 6 6 Other 4 5 22 28 Total 36 36 64 80 Table 5 Proportions of three types of features for the experimental and control groups in two design disciplines with the Hydrocushion fixating example ( possible significant differences are identified with an asterisk ) . Mechanical Engineers Industrial Designers ExperimentalControl ExperimentalControl Principles Operates on water pressure * . 29 . 00 . 00 . 03 Used to raise and lower person from edge of bath * . 44 . 28 * . 50 . 34 Device is removable . 51 . 50 * . 67 . 55 Device connects to existing faucet * . 15 . 00 . 00 . 01 Valve to direct water flow . 07 . 00 . 00 . 00 Provision made for overflow when device is raised . 02 . 00 . 00 . 00 Details Water filled container . 12 . 00 . 00 . 03 Inflatable cushion . 15 . 06 . 06 . 09 Platform seat , no sides or back . 27 . 19 * . 42 . 30 Rectangular platform * . 27 . 11 . 19 . 15 Rounded corners . 17 . 11 * . 22 . 07 Connection to faucet via an inflatable pressure collar . 02 . 00 . 00 . 00 Collar attached by a hand squeeze pump . 02 . 00 . 00 . 00 Cushion connects to faucet via number of pipes . 02 . 00 . 00 . 00 Fixed piping incorporating a rightangle . 02 . 00 . 00 . 00 18 Insert plug to fill cushion , release plug to drain water . 02 . 00 . 00 . 00 J shaped vertical pipe connected to faucet . 05 . 00 . 00 . 00 Context Unit in intended environment . 90 . 97 * . 92 . 75 Device or user at opposite end to taps * . 27 . 08 * . 19 . 01 Bath positioned in return . 12 . 19 . 14 . 05 Handrail located above bath fixed to tiled wall surface . 02 . 06 . 03 . 00 Tiles shown . 07 . 03 * . 17 . 00 Single bath tap shown * . 12 . 00 . 00 . 00 Insets or series of illustrations used . 73 . 64 * . 86 . 59 i ) User operating the unit * . 39 . 19 * . 44 . 24 ii ) Rear perspective of user seated on unit in bath . 02 . 00 . 03 . 00 iii ) User lowering device * . 22 . 00 * . 14 . 04 iv ) User raising the device . 10 . 00 . 06 . 01 v ) User on raised device ready to leave bath . 07 . 03 . 03 . 00 vi ) Side elevation shown * . 51 . 28 * . 36 . 24 vii ) Schematic representation plan or section view . 56 . 64 * . 61 . 51 19 Appendix 1 Instructions for verbal description fixation group The aim of the design exercise is to come up with a sketch design ( s ) for a bicycle rack for three bicycles for a car . The bicycles have to be held securely , and without damage to either the bicycle or the car . The bicycles must not extend beyond the overall width dimension of the car to avoid potential damage to people or vehicles in passing . There are a number of key issues to be considered in the design of a bicycle rack . The first is the way in which the rack is attached to the car . Then , there has to be a structural system that will support the bicycles . Third , there has to be a way of attaching the bicycles to the support system . Fourth , both the structural system and the way of attaching the bicycles have to have the correct relationship to fulfil the other more general conditions of safety etc . To illustrate , the bicycle rack can be attached to the car at a number of locations : for example at the rear of the car , to a tow bar fitting , or directly to the chassis of the car . The bicycles can be sup [ ported structurally in a number of ways : for example , two steel posts in the shape of an A joined across the bottom of the A can be attached to the fitting fixed to the car . This has to be of sufficient height to provide clearance above the ground . Similarly the bicycles can be attached to the rack in a number of ways : for example , they can be held by the top horizontal section of the bicycle frame in short half sections of pipe of a diameter that allows the bicycle frame section to fit into the pipe with the other half section closing over the bike frame . The bicycles then have to have the appropriate relationship to the car : for example , to the top of the A - frame a small steel section can be attached at rightangles in line with the length of the car . The fitting with the brackets attaching the bicycles can be placed at rightangles to the length of the car on this section of the post , with the bicycles being placed parallel to and clear of , the boot of the car . Detailed and accurate drawings are not required : simple , rough , outline sketches are all that is needed . In addition to the sketches , you can write comments on the drawings to illustrate what you mean . You will be allowed 45 minutes in which to complete the sketch design . If you wish you may complete more than one design . This paper is a copy of : Purcell , T . A . and Gero , J . S . ( 1996 ) Design and other types of fixation , Design Studies 17 ( 4 ) : 363 - 383",
    "ulrichs2023multicomponent": "Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Multicomponent regulation of actin barbed end assembly by twin \ufb01 lin , formin and capping protein Heidi Ulrichs 1 , 2 , 3 , Ignas Gaska 1 , 2 , 3 & Shashank Shekhar 1 , 2 Cells control actin assembly by regulating reactions at actin \ufb01 lament barbed ends . Formins accelerate elongation , capping protein ( CP ) arrests growth and twin \ufb01 lin promotes depolymerization at barbed ends . How these distinct activities get integrated within a shared cytoplasm is unclear . Using micro \ufb02 uidics - assisted TIRF microscopy , we \ufb01 nd that formin , CP and twin \ufb01 lin can simultaneously bind \ufb01 lament barbed ends . Three \u2011 color , single - molecule experiments reveal that twin \ufb01 lin cannot bind barbed ends occupied by formin unless CP is present . This trimeric complex is short - lived ( ~ 1 s ) , and results in dissociation of CP by twin \ufb01 lin , promoting formin - based elongation . Thus , the depolymerase twin \ufb01 lin acts as a pro - formin pro - polymerization factor when both CP and formin are present . While one twin \ufb01 lin binding event is suf \ufb01 cient to displace CP from the barbed - end trimeric complex , ~ 31 twin \ufb01 lin binding events are required to remove CP from a CP - capped barbed end . Our \ufb01 ndings establish a paradigm where polymerases , depolymerases and cappers toge - ther tune actin assembly . Cellular actin dynamics are essential in several key processes , such as cell migration , wound healing , cell division , and endocytosis 1 , 2 . Depending upon the requirements of speci \ufb01 c processes , cells dyna - mically tune the rate of assembly , size , and architecture of their \ufb01 la - mentous actin networks . For example , while formins assemble fast - elongatinglinearactinnetworkscomprisingoflongbundled \ufb01 laments , e . g . , in \ufb01 lopodia , stereocilia , and stress \ufb01 bers 3 \u2013 7 , the Arp2 / 3 complex assembles dendritic actin arrays made of short , relatively slowly growing branched actin \ufb01 laments , e . g . , in lamellipodia of motile cells and at sites of endocytosis 1 , 8 , 9 . How cells assemble actin networks with such diverse morphologies and dynamics in a shared cytoplasm remains an open question . Cellular actin assembly is primarily governed by reactions occur - ring at thebarbedend of actin \ufb01 laments ( sometimesalsoreferred toas the plus end ) 2 , 10 . Even though actin \ufb01 laments can elongate by sponta - neous addition of actin monomers , barbed end dynamics can be fur - ther tuned by three distinct classes of proteins that directly bind barbedends . Theseincludepolymerases , cappers , anddepolymerases . Polymerases , like formins and Ena / VASP , nucleate actin \ufb01 laments , and support \ufb01 lament elongation by remaining processively bound to \ufb01 la - ment barbed ends 11 , 12 . Formins can additionally accelerate the rate of barbed - end assembly by up to \ufb01 vefold in the presence of pro \ufb01 lin - bound actin monomers ( referred to as pro \ufb01 lin - actin monomers or PA henceforth ) 13 , 14 . Cappers , like capping protein ( CP ) and gelsolin , cause the complete arrest of \ufb01 lament growth by preventing monomer addition at free barbed ends 15 , 16 . Depolymerases , like twin \ufb01 lin , com - prise the third class of barbed - end binding proteins 17 , 18 . Initially dis - covered as actin monomer - sequestering protein 19 , twin \ufb01 lin induces barbed - end depolymerization in a nucleotide - speci \ufb01 c fashion 17 , 18 . While twin \ufb01 lin accelerates depolymerization of newly assembled ( ADP - P i ) \ufb01 laments , itslowsdowndepolymerizationofaged ( ADP ) actin \ufb01 laments 18 , 20 . Notably , twin \ufb01 lin \u2019 s depolymerization activities persist even in cytosol - mimicking conditions , i . e . , the presence of high con - centrations of actin monomers 17 , 18 , 20 . Althoughtheindividualeffectsofformin , CPandtwin \ufb01 linonactin assembly are relatively well - characterized , how they simultaneously act in multiprotein teams at barbed ends to in \ufb02 uence actin assembly remains unclear . This question is especially important as these factors Received : 22 December 2022 Accepted : 22 June 2023 Check for updates 1 Department of Physics , Emory University , Atlanta , GA 30322 , USA . 2 Department of Cell Biology , Emory University , Atlanta , GA 30322 , USA . 3 These authors contributed equally : Heidi Ulrichs , Ignas Gaska . e - mail : shekhar @ emory . edu Nature Communications | ( 2023 ) 14 : 3981 1 1 2 3 4 5 6 7 8 9 0 ( ) : , ; 1 2 3 4 5 6 7 8 9 0 ( ) : , ; are often found in the same cellular compartments , such as \ufb01 lopodia , lamellipodia and stereocilia 6 , 18 , 21 \u2013 24 . In addition to twin \ufb01 lin \u2019 s monomer sequestration and barbed end depolymerization functions , it can also directly bind CP via its C - terminal tail . The CP - twin \ufb01 lin interaction is requiredforproper intracellularlocalizationofCP but isdispensable for twin \ufb01 lin \u2019 s uncapping activity in vitro 20 , 25 . Moreover , in absence of direct visualization of CP \u2019 s uncapping by twin \ufb01 lin , the underlying mechanism remains unclear . While a number of recent studies have looked attwin \ufb01 lin - CP interaction , twin \ufb01 lin \u2019 s effects on formin have not yetbeenextensivelystudied . Wepreviouslyreportedthatformin \u2019 srate of barbed - end elongation is not affected by a high concentration of twin \ufb01 lin 18 . Nevertheless , twin \ufb01 lin \u2019 s effects on formin \u2019 s long - lived resi - dence time at the barbed end has yet to be investigated . CP and formin were initially thought to bind barbed ends in a mutually exclusive fashion 26 \u2013 28 . Contrary to this assumption , it was discovered that formin and CP simultaneously bind the same barbed end to form a so - called barbed - end decision complex ( henceforth interchangeably referred to as BFC , where B is the barbed end , F is formin , and C is CP ) 29 , 30 . Their concurrent presence at the \ufb01 lament end accelerates barbed - end dissociation of formin and CP byabout 50 - fold and 10 - fold , respectively 29 . While the mechanisms discussed above reducebarbed - endresidencetimesofforminandCPfrom120minand 30min to just a few minutes 29 , these time scales are still too slow to explain the rapid rates at which intracellular actin structures get assembled , arrested , and turned over in a few seconds 1 , 31 . This prompted us to take a fresh look at the multicomponent dynamics between CP , formin , and twin \ufb01 lin at \ufb01 lament barbed ends . Here we show that while twin \ufb01 lin alone has no direct effects on formin \u2019 s processivity , it greatly enhances formin \u2019 s processivity in presence of CP ( Fig . 1 ) . Using micro \ufb02 uidics - assisted total internal re \ufb02 ection \ufb02 uorescence ( mf - TIRF ) microscopy 32 , 33 , we \ufb01 nd that despite its depolymerization of free barbed ends , twin \ufb01 lin effectively promotes actin assembly when both formin and CP are present together ( Fig . 2 ) . We \ufb01 nd that formin , CP and twin \ufb01 lin can all simultaneously bind a \ufb01 lament barbed end in a trimeric com - plex ( Fig . 3 ) . The dynamics of this formin - CP - twin \ufb01 lin complex at the barbed end are visualized by multicolor single - molecule imaging . We discover that twin \ufb01 lin reduces the lifetime of the CP - formin decision complex at the barbed end by about 17 - fold . The trimeric complex formation leads to accelerated transitions between these proteins at the barbed end . We also visualize twin \ufb01 lin \u2019 s uncapping of CP - bound barbed ends and \ufb01 nd that twin \ufb01 lin displaces CP from the formin - CP complex much more ef \ufb01 ciently than from barbed ends bound only to CP ( Fig . 4 ) . While only a single twin \ufb01 lin binding event is suf \ufb01 cient to remove CP from formin - CP complexes , on average it takes about 31 twin \ufb01 lin binding events to displace CP from barbed ends bound only to CP . Using separation - of - function mutants , we \ufb01 nd that twin \ufb01 lin \u2019 s direct interaction with the actin \ufb01 lament is necessary for its ability to rescue formin \u2019 s processivity ( Fig . 5 ) . This provides direct evidence of a depolymerase , polymerase , and capper simulta - neously binding a growing \ufb01 lament end , and demonstrates the multicomponent mechanism by which they can regulate each other \u2019 s barbed end activities . Results Capping protein and twin \ufb01 lin differentially in \ufb02 uence formin \u2019 s processivity We studied how CP and twin \ufb01 lin in \ufb02 uence formin \u2019 s processivity at actin \ufb01 lament barbed ends . Fluorescent actin \ufb01 laments were initiated in a mf - TIRF chamber by exposing coverslip - anchored formins ( mDia1 FH1 - FH2 - C ) to a solution containing \ufb02 uorescent actin monomers and pro \ufb01 lin - actin ( PA ) ( Fig . 1a ) . Then , to con \ufb01 rm these \ufb01 laments were indeed elongating from formins , a solution containing pro \ufb01 lin and unlabeled actin monomers was \ufb02 owed . The use of unlabeled actin prevents any artifacts that might arise from the use of labeled actin . Since elongation occurs by insertion of unlabeled monomers at the formin - bound barbed end , the pre - existing \ufb02 uorescent segment of the \ufb01 lament appears to move in the direction of the \ufb02 ow ( Fig . 1a , b and Supplementary Movie 1 ) , away from the location of formin anchoring . Dissociation of \ufb01 laments from surface - anchored formins caused the immediate disappearance of \ufb01 laments from the \ufb01 eld of view ( BF \u2192 B + F , where B denotes the barbed end , F denotes the formin , and BF is the formin - bound barbed end ) . We recorded the disappearance of actin \ufb01 laments from the \ufb01 eld of view over time . Changes in the time - dependent survival fraction offormin - bound \ufb01 laments were then used to determine the dissociation rate of formin from the barbed end . Using unlabeled actin subunits eliminates the effects of labeled actin on formin \u2019 s processivity . To con \ufb01 rm that \ufb01 lament disappearance was due to detachment of \ufb01 laments from formin rather than due to the detachment of the entire formin - \ufb01 lament complex from the glass coverslip , we re - exposed the surface to a \ufb02 ow containing 1 \u00b5 M Alexa - 488 G - actin and 0 . 5 \u00b5 M pro \ufb01 lin . Consistent with previous studies , over 80 % of formins were able to renucleate new \ufb01 laments following detachment of initially nucleated \ufb01 laments 29 , 34 . First , we asked how CP in \ufb02 uences the processivity of formins at \ufb01 lament barbed ends . Formin - elongated \ufb02 uorescent \ufb01 laments were exposed to a solution containing either pro \ufb01 lin - actin alone or with a range of concentrations of CP . In control reactions , the \ufb02 uorescent segment of actin \ufb01 laments continued to move away from the attached formin , at a constant speed , in the direction of the \ufb02 ow , indicating processive elongation by formin ( Fig . 1b and Supplementary Movie 1 ) . Onlya smallfraction ( ~ 20 % ) of \ufb01 lamentsdissociatedfromforminsover the duration of the experiment ( 500 s ) . The average barbed - end dwell timeof formin mDia1 was ~ 35 min in control experiments ( Fig . 1f , dwell time = 1 / ( dissociation rate ) ) . Formin \u2019 s long barbed - end residence time measured here agrees with previous studies 13 , 14 , 34 . Interestingly , when cappingproteinwasintroduced ( inthepresenceofPA ) , actin \ufb01 laments rapidly dissociated from formins and disappeared from the \ufb01 eld of view ( Fig . 1c , f and Supplementary Movie 2 ) . The rate of dissociation of formins increased with CP concentration ( 5nM to 1 \u00b5 M ) . Compared to control , 1 \u00b5 M CP increased formin \u2019 s rate of dissociation from the bar - bed end by about 30 - fold ( Fig . 1g ) . While the cytoplasmic concentration of CP is around 1 \u00b5 M 35 , majority of it is thought to be sequestrated by V1 / myotrophin 36 . As a result , only about 10 \u2013 20nM free CP is expected to be available for binding barbed ends in cells 36 . We found thateven a low concentration of CP in this range ( ~ 50nM ) was suf \ufb01 cient to accelerate formin dis - sociation by about tenfold compared to the control ( Fig . 1g ) . We then asked if twin \ufb01 lin also in \ufb02 uenced formin \u2019 s dissociation from barbed ends . In contrast to CP , we found that the presence of up to 1 \u00b5 M twin \ufb01 lin ( mouse mTwf1 , referred to as \u201c twin \ufb01 lin \u201d hereafter ) did not signi \ufb01 cantly change formin \u2019 s barbed - end dwell time ( Fig . 1d , h , i ) . Consistent with our earlier study , twin \ufb01 lin also had no observable effect on formin \u2019 s rate of elongation 18 ( Supplementary Fig . 1 ) . Our results imply that unlike capping protein , formins fully protect barbed ends from twin \ufb01 lin , indicating that twin \ufb01 lin cannot associate with a formin - bound barbed end . We then asked how simultaneous presence of twin \ufb01 lin and CP might in \ufb02 uence formin \u2019 s processivity . When formin - nucleated \ufb01 la - mentsareexposedtoCPandtwin \ufb01 linatthesametime ( inthepresence of PA ) , formin dissociated at a rate intermediate between that of control and capping protein ( Fig . 1h , i and Supplementary Movie 3 ) . While there was a tenfold reduction in formin \u2019 s processivity in the presenceof 50 nMCP , thereductionwasjustthreefoldwhen 50nMCP was supplemented with 1 \u00b5 M twin \ufb01 lin . Our data suggest that twin \ufb01 lin \u2019 s presence led to a reduction in the adverse effects of CP on formin \u2019 s processivity . Consistently , actin \ufb01 laments also grew substantially longer prior to their detachment in the presence of twin \ufb01 lin and CP together ( Fig . 1e ) as compared to CP alone ( Fig . 1c ) . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 2 Twin \ufb01 lin accelerates dissociation of the formin \u2013 CP complex at barbed ends X - ray diffraction and EM studies suggest that both CP and twin \ufb01 lin interact directly with \ufb01 lament barbed ends 37 \u2013 40 . Moreover , twin \ufb01 lin destabilizes CP \u2019 s barbed - end localization and causes a sixfold reduc - tion in CP \u2019 s barbed - end lifetime 20 . We , therefore , wondered if twin \ufb01 lin \u2019 s rescue of formin \u2019 s processivity from CP might be due to its destabilizing effects on CP in the formin - CP - barbed end decision complex . To investigate this , we formed decision complexes by exposing formin - nucleated \ufb02 uorescent actin \ufb01 laments to a high con - centration of CP ( 1 \u00b5 M ) . This caused an almost immediate arrest of actin \ufb01 lament elongation and formation of formin - CP - barbed end 0 100 200 300 400 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 ControlmTwf1CPmTwf1 + CP n o i t c a r f l a v i v r u s n i m r o F Time ( s ) Control5 nM 50 nM 250 nM 500 nM 1 \u00b5 M 0 100 200 300 400 500 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 no i t c a r f l a v i v r u s n i m r o F Time ( s ) h b i a f 0 200 400 600 800 1000 0 . 000 0 . 005 0 . 010 0 . 015 s ( e t a r no i t a i c o ss i d n i m r o F - 1 ) [ Capping protein ] ( nM ) Control mTwf1 CP CP + mTwf1 0 . 000 0 . 002 0 . 004 0 . 006 0 . 008 s ( e t a r no i t a i c o ss i d n i m r o F - 1 ) Capping protein Formin Actin ( labeled ) Actin Profilin Flow and / or Twinfilin Coverslip c d e g PA 60s 10 \u03bc m PA * PA 10 \u03bc m 60s PA * PA PA + CP 10 \u03bc m 60s PA * PA PA + CP + mTwf1 10 \u03bc m 60s PA * PA PA + mTwf1 Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 3 decision complexes 29 , 30 . To study the effects of twin \ufb01 lin on BFC com - plexes , BFC complexes were then exposed to a \ufb02 ow containing either pro \ufb01 lin - actin only ( control ) or supplemented with a range of con - centrations of twin \ufb01 lin ( Fig . 2a ) . The BFC complexes can dissociate by two different routes \u2014 ( 1 ) via dissociation of CP ( BFC \u2192 BF + C , with CP departing from the BFC complex with a rate k \u2019 - C ) , which leads to resumption of elongation of formin - bound \ufb01 laments ( Fig . 2b , top ) or ( 2 ) via dissociation of formin ( BFC \u2192 BC + F , withformindetachingfromtheBFCcomplexwitharate k \u2019 - F ) , which leads to detachment and disappearance of the \ufb01 lament ( Fig . 2b , bottom ) . We found that twin \ufb01 lin dramatically accelerated disassembly of BFC complexes in a concentration - dependent manner ( Fig . 2c , d ) . Compared to the control , 1 \u00b5 M twin \ufb01 lin increased the rate of BFC dissociation ( k - BFC ) by about 18 - fold . While 1 \u03bc M twin \ufb01 lin increased CP \u2019 s dissociation rate from the BFC complex by ~ 11 - fold , formin \u2019 s rate of dissociation did not change ( Fig . 2e , f ) . The dissociation kinetics of BFC complexes into BC and BF wereexponentialwithobservedrateconstant k - BFC = k \u2019 - F + k \u2019 - C where k \u2019 - F is the observed rate constant of dissociation of formin from the BFC complex and k \u2019 - C is the observed rate constant of dissociation of CP from the BFC complex ( Fig . 2e and Supplementary Fig . 2 ) . The fraction of \ufb01 laments transitioning to BF and BC states upon BFC dissociation is givenbythe k \u2019 - C / ( k \u2019 - C + k \u2019 - F ) and k \u2019 - F / ( k \u2019 - C + k \u2019 - F ) ( seeMethods ) . Wederived the values k \u2019 - C and k \u2019 - F asa function of twin \ufb01 linconcentration ( Fig . 2e , f , and Supplementary Fig . 2 ) . We also analyzed the route by which the BFC complexes dissociated ( BFC \u2192 BF + C or BFC \u2192 BC + F ) . In absence oftwin \ufb01 lin , slightlymorethanhalfofallBFCcomplexestransitionedto BC ( ~ 54 % ) and the rest transitioned to BF ( ~ 46 % ) . In presence of twin - \ufb01 lin however , this ratio was skewed heavily towards BF ( ~ 90 % at 1 \u03bc M twin \ufb01 lin ) , i . e . , CP dissociated and left formin solely bound to the bar - bed end in majority of BFC complexes ( Fig . 2g ) . These \ufb01 laments immediately returned to rapid elongation , characteristic of formin \u2019 s presence at the barbed end . Taken together , our results indicate that although twin \ufb01 lin is a depolymerase of free barbed ends , it promotes \ufb01 lament assembly by formin when both CP and formin are present . Just as twin \ufb01 lin can uncap CP from free barbed ends 20 , our results indicate twin \ufb01 lin might also be able to destabilize CP from BFC complexes by forming a ternary complex , BFCT , with formin and CP at the barbed end ( T in BFCT denotes twin \ufb01 lin ) . However , due to our inability to directly visualize twin \ufb01 lin , it was not possible to ascertain whether twin \ufb01 lin \u2019 s effects were due to its barbed - end binding or its interactions with \ufb01 lament sides . Single - molecule visualization of twin \ufb01 lin \u2019 s effects on BFC dynamics To directly visualize the effects of twin \ufb01 lin on the dynamics of the BFC complex , we used three - color single - molecule TIRF imaging . SNAP - tagged constructs of CP ( SNAP - CP ) and formin ( SNAP - mDia1 ) were expressed and labeled with benzylguanine functionalized green - excitable ( 549 - CP ) and red - excitable ( 649 - mDia1 ) \ufb02 uorescent dyes . Photobleaching data con \ufb01 rmed that majority of 549 - CP molecules were labeled with only one dye molecule , consistent with a single SNAP - tag per heterodimeric CP molecule 30 ( Supplementary Fig . 3 ) . Photobleaching tests of 649 - mDia1 showed that majority of these molecules exhibited a single - or double - step bleaching pro \ufb01 le , con - sistent with the dimeric nature of formin mDia1 molecules 29 , 41 ( Sup - plementary Fig . 4 ) . Alexa - 488 labeled G - actin actin monomers were incubated with 649 - mDia1 in a non - micro \ufb02 uidic , conventional \ufb02 ow cell . The majority of newly nucleated \ufb01 laments displayed \ufb02 uorescent formins at their barbed ends ( Fig . 3a ) . All \ufb01 lament barbed ends with detectable 649 - mDia1 underwent rapid , continuous elongation with no noticeable pauses . Consistent with long barbed - end dwell times and low photobleaching , the majority of 649 - mDia1 molecules remained bound to elongating actin \ufb01 laments for the duration of the experiment . 649 - mDia1 bound barbed ends were then exposed to a pro \ufb01 lin - actin solution containing either 549 - CP alone or together with unla - beled twin \ufb01 lin . In the absence of twin \ufb01 lin , we routinely observed the following ( Fig . 3a , b ) : \ufb01 rst , the 649 - mDia1 was joined at the barbed end bya549 - CPmoleculetoformtheBFCcomplex ; second , uponarrivalof 549 - CP , the \ufb01 lament immediately stopped elongating ; third , following abriefintervaloftimeduringwhichbothmoleculeswerejointlybound to the barbed end , the 649 - mDia1 : 549 - CP complex dissociated , and one of these two molecules departed from the barbed end , leaving the other behind ( Supplementary Movie 4 ) . In approximately 58 % of the complexes , 649 - mDia1 dissociated ( N = 42 out of 72 complexes ) , leaving 549 - CP at the barbed end , and no further elongation was observed ( Fig . 3a , c ) . In the remaining complexes , 549 - CP departed ( N = 30 out of 72 complexes ) and the \ufb01 lament immediately switched from the paused state to the rapidly elongating state , characteristic of formin - bound barbed ends . The 649 - mDia1 : 549 - CP barbed end com - plex had an average lifetime of 149 \u00b112 . 4s ( mean\u00b1sem ) ( Fig . 3e , f ) . In contrast , when 649 - mDia1 bound \ufb01 laments were exposed to 549 - CP but in presenceof 20nM twin \ufb01 lin , a much shorter pause wasobserved duringwhichboth549 - CPand649 - mDia1wereboundandthe \ufb01 lament elongation was arrested ( Fig . 3d ) . The average complex lifetime reduced to 35 . 4\u00b15s ( mean\u00b1sem ) , a 75 % reduction over control ( Fig . 3e , f ) . In agreement with our mf - TIRF experiments , twin \ufb01 lin also altered the outcome of BFC complex resolution . In presence of 20 nM twin \ufb01 lin , about 73 % of complexes ( 36 out of 49 ) transitioned to the 649 - mDia1 BF state as compared to 42 % of complexes ( 30 out of 72 ) in control experiments . Our mf - TIRF and single - molecule experiments together con - clusively establish that twin \ufb01 lin in \ufb02 uences both the lifetime and eventual outcome ( BF or BC ) of the BFC complex . We then asked if these effects are caused by binding of twin \ufb01 lin along the \ufb01 la - ment length or by its interactions at the barbed end . To do this , we expressed mouse twin \ufb01 lin - 1 as a SNAP - tagged fusion protein and directly visualized its interactions with the BFC complex . Fig . 1 | Effect of twin \ufb01 lin and capping protein ( CP ) on processivity of formin . a Schematic of the experimental strategy . Actin \ufb01 laments were nucleated from coverslip - anchored formins by a \ufb02 ow containing 1 \u00b5 M G - actin ( 15 % Alexa - 488 labeled ) and0 . 5 \u00b5 Mpro \ufb01 lin . Filamentswerethenelongated inthepresenceof1 \u00b5 M unlabeled G - actin and 4 \u00b5 M pro \ufb01 lin to ensure insertional elongation between \ufb02 uorescentfragmentsandsurface - anchoredformins . Filamentswerethenexposed toa \ufb02 owcontaining0 . 2 \u00b5 MunlabeledG - actinand0 . 7 \u00b5 Mpro \ufb01 lin ( PA ) ( control ) with or without a range of concentrations of CP and / or mTwf1 ( alone or together ) . The survival fraction of formin - bound \ufb01 laments attached was monitored as a function of time . b Representative kymographs of a formin - anchored \ufb01 lament elongating from 0 . 2 \u00b5 M unlabeled G - actin and 0 . 7 \u00b5 M pro \ufb01 lin ( PA ) ( see Supplementary Movie 1 ) . c Same conditions as ( b ) but supplemented with 50nM CP ( see Supple - mentary Movie 2 ) . d Same conditions as ( b ) but supplemented with 1 \u00b5 M mTwf1 . e Same conditions as ( b ) but supplemented with 50nM CP and 1 \u00b5 M mTwf1 ( see Supplementary Movie 3 ) . f Survival fraction of formin - bound \ufb01 laments ( BF ) as a function of time in the presence of PA supplemented with a range of CP con - centrations . Experimental data ( symbols ) are \ufb01 tted to a single - exponential decay function ( lines ) todeterminetheformindissociationrate k - F ( BF \u2192 B + F ) . Numberof \ufb01 lamentsanalyzedforeachcondition ( 0to1 \u00b5 MCP ) : 70 , 36 , 75 , 56 , 26 , 26 . g Formin dissociationrate k - F asafunctionofCPconcentration , determinedfromdatain ( f ) . h Survival fraction of formin - bound \ufb01 laments ( BF ) as a function of time in the presence of PA alone ( black symbols , n = 51 \ufb01 laments ) or supplemented with 1 \u00b5 M mTwf1 ( magenta symbols , n = 50 \ufb01 laments ) , 50nM CP ( yellow symbols , n = 51 \ufb01 la - ments ) or 1 \u00b5 M mTwf1 , and 50nM CP together ( orange symbols , n = 50 \ufb01 laments ) . Experimental data ( symbols ) are \ufb01 tted to a single - exponential decay function ( lines ) . i Formin dissociation rate k - F as determined from data in ( h ) . Error bars in g , i indicate 65 % con \ufb01 dence intervals based on \ufb01 ts ( see Methods ) . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 4 SNAP - tagging did not alter mTwf1 \u2019 s ability to uncap CP from CP - bound barbed ends ( Supplementary Fig . 5 ) . Photobleaching records of surface adsorbed 549 - SNAP - mTwf1 showed that almost all molecules exhibited single - step photobleaching , con \ufb01 rming that labeled twin \ufb01 lin molecules were monomeric ( Supplemen - tary Fig . 6 ) . Using labeled twin \ufb01 lin , we further examined the mechanism by which twin \ufb01 lin accelerated dissociation of BFC complexes . We exposed 649 - mDia1 bound Alexa - 488 actin \ufb01 laments to a solution containing \ufb02 uorescently labeled 549 - mTwf1 and unlabeled CP . Expectedly , \ufb01 laments rapidly paused due to BFC complex formation between unlabeled CP and 649 - mDia1 , leading to an abrupt stop in the a c d f 46 % 73 % 89 % 54 % 27 % 11 % 0 nM 50 nM 1 \u03bcM 0 20 40 60 80 100 no i t a i c o ss i d nopu s e x e l p m o c C F B f o e t a F [ mTwf1 ] BC BF 0 50 1000 0 . 00 0 . 01 0 . 02 0 . 03 0 . 04 0 . 05 0 . 06 ( P C f o e t a r no i t a i c o ss i d d e v r e s b O k ' - C r o ) ( n i m r o F k ' - F s ( C F B m o r f ) - 1 ) [ mTwf1 ] ( nM ) k ' - C k ' - F 0 200 400 600 800 1000 0 . 00 0 . 03 0 . 06 0 . 09 0 . 12 0 . 15 s ( e t a r no i t a i c o ss i d C F B - 1 ) [ mTwf1 ] ( nM ) 0 100 200 300 400 500 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 Control15 nM mTwf1 50 nM mTwf1 100 nM mTwf1 250 nM mTwf1 1 \u00b5 M mTwf1 de l b m e ss a s i d s e x e l p m o c C F B f o no i t c a r F Time ( s ) 0 100 200 300 400 500 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 s t ne m a li f f o no i t c a r F Time ( s ) BFC BC BF 0 nM mTwf1 0 100 200 300 400 500 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 s t ne m a li f f o no i t c a r F Time ( s ) BFC BF BC 50 nM mTwf1 0 100 200 300 400 500 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 s t ne m a li f f o no i t c a r F Time ( s ) BFC BF BC 1 \u00b5 M mTwf1 Flow Capping protein Formin Actin ( labeled ) Actin Profilin Twinfilin Coverslip 30s 10 \u03bc m PA * CP PA ( + / - mTwf1 ) 30s 10 \u03bc m PA * CP PA ( + / - mTwf1 ) b e g Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 5 translocation of 649 - mDia1 molecules bound to actin \ufb01 laments . Toour surprise , we observed 549 - mTwf1 molecules were now able to tran - siently join 649 - mDia1 ( and unlabeled CP ) at the barbed end ( Fig . 3g , h ) . Association of 549 - mTwf1 molecules with 649 - mDia1 : unlabeled CP - bound barbed ends were very brief , with an average dwell time of 1 . 4 \u00b10 . 2 s ( mean\u00b1 sd , n = 8 events ) , translating to a dissociation rate of 0 . 71s \u2212 1 . The majority of 549 - mTwf1 : 649 - mDia1 : unlabeled CP BFCT complexes remained intact just for a single frame implying the 0 . 71 s \u2212 1 rate of dissociation of 549 - mTwf1 molecules fromBFCT complex is the lower bound for this rate . Importantly , the departure of 549 - mTwf1 from the BFCT complex led to an immediate transition of the barbed end from the arrested state to the fast formin - based elongation state with visible translocation of the 649 - mDia1 molecule . Notably , in absence of CP , we never observed colocalization of 649 - mDia1 and 549 - mTwf1 at barbed ends . This explains our mf - TIRF results that twin \ufb01 lin does not in \ufb02 uence the rate of elongation or processivity of formins ( Fig . 1h and Supplementary Fig . 1 ) . Together , these observa - tions indicate that the unlabeled CP molecule and twin \ufb01 lin departed the barbed end simultaneously , leaving formin behind . Notably , we never observed twin \ufb01 lin and CP simultaneously arrive at formin - occupied barbed ends . Single - molecule analysis uncovers twin \ufb01 lin \u2019 s uncapping mechanism Although twin \ufb01 lin accelerates CP \u2019 s dissociation 42 , in the absence of simultaneous visualization of these two proteins at barbed ends , the underlying mechanism has remained obscure . We therefore tran - siently exposed actin \ufb01 laments elongating from Alexa - 488 labeled monomers to 649 - CP in a non - micro \ufb02 uidic , conventional \ufb02 ow cell ( Fig . 4a ) . Expectedly , all \ufb01 laments with a visible 649 - CP signal at their barbed end immediately stopped growing following 649 - CP \u2019 s arrival . Upon exposure to 549 - mTwf1 , short - lived associations of twin \ufb01 lin at the 649 - CP - bound barbed ends were observed . To our surprise , unlike inthecaseofBFCcomplexeswhereasingletwin \ufb01 linbindingeventwas suf \ufb01 cient to cause CP \u2019 s departure , 649 - CP remained bound to the barbed end despite repeated arrivals and departures of 549 - mTwf1 molecules at the barbed end ( Fig . 4b ) . Each 549 - mTwf1 : 649 - CP colo - calization event lasted about a second . However , owing to the slow rateofuncappingbytwin \ufb01 lin ( itonlyacceleratestheuncappingrateby sixfold ) , capturing simultaneous departure of 649 - CP and 549 - mTwf1 required frequent , long - term imaging , which in turn caused excessive photobleaching of 649 - CP . As a result , we observed that the dis - appearance of 649 - CP from barbed ends was not always followed by \ufb01 lament depolymerization . These observations suggested that the disappearance of the 649 - CP signal from barbed ends was due to photobleaching and not because of CP \u2019 s dissociation . Owing to photobleaching of 649 - CP , we instead used unlabeled CP . In these experiments , we used the onset of depolymerization of barbed ends to detect the moment when 549 - mTwf1 dissociated unlabeled CP from barbed ends ( Fig . 4c ) . Successful CP dissociation required on average 30 . 9\u00b16 . 5 ( mean \u00b1sem ) successive 549 - mTwf1 binding events , each lasting for 1 . 9\u00b10 . 1 s ( mean\u00b1 sem ) ( Fig . 4d , e ) . Notably , 549 - mTwf1 intensity appeared and disappeared from the barbed end in a single step , suggesting only a single 549 - mTwf1 molecule colocalized with CP at a time at the \ufb01 lament barbed end ( Fig . 4d ) . Twin \ufb01 lin \u2019 s interaction with actin is essential for its effects on BFC dynamics As mentioned earlier , twin \ufb01 lin can directly bind CP as well as interact with terminal F - actin subunits of the barbed end 37 , 43 . Which of these interactions is responsible for twin \ufb01 lin \u2019 s effects on BFC complex dynamics ? To answer this question , we puri \ufb01 ed two twin \ufb01 lin mutants : ( 1 ) the \u201c ADF - domainmutant \u201d , which inhibits twin \ufb01 lin \u2019 s interaction with actin and ( 2 ) the \u201c tail mutant \u201d , which interferes with twin \ufb01 lin \u2019 s direct binding to CP 44 ( Fig . 5a ) . Pre - formed BFC complexes ( similar to the strategy used in Fig . 2a ) were exposed to a solution containing either pro \ufb01 lin - actin only or supplemented with wild - type or mutant twin \ufb01 lin ( Fig . 5b ) . Mutations in the twin \ufb01 lin tail led to BFC dissociation activity partway between the control and the wildtype ( Fig . 5c , d ) . However , the ADF - domain mutant didn \u2019 t cause any visible acceleration of BFC complex dissociation , implying that direct contact between twin \ufb01 lin and actin is essential for these activities . Consistently , an earlier study showedthattheADF - domainmutantisalsoincapableofuncappingCP from free barbed ends 20 . Taken together , our data suggests that while twin \ufb01 lin \u2019 s interaction with CP via its tail domain is bene \ufb01 cial , twin \ufb01 lin \u2019 s binding to actin via its ADF domains is necessary for it to be able to form the BFCT ternary complex at the barbed end and for its ability to rapidly dissociate BFC complexes . Discussion Living cells can rapidly tune actin assembly in response to external stimuli 1 . The actin \ufb01 lament barbed end is considered to be the primary site of regulation of actin assembly . Over the last few decades , several barbed - end interacting proteins , including polymerases ( e . g . , formin ) , cappers ( e . g . , CP ) , and depolymerases ( e . g . , twin \ufb01 lin ) have been dis - covered . Howthesedistinctprotein activities get integrated within the cytoplasm remains an open question . On its own , CP caps free barbed ends and promotes dissociation of formin from formin - bound barbed ends 29 , 30 . Twin \ufb01 lin promotes barbed - end depolymerization and accel - erates the dissociation of CP from barbed ends 17 , 18 , 20 . Here , we have uncovered a mechanism by which twin \ufb01 lin , formin , and CP simulta - neouslybinda \ufb01 lamentbarbedendtoformamulticomponenttrimeric protein complex . We also \ufb01 nd that twin \ufb01 lin does not bind actin \ufb01 la - ment ends occupied by formin mDia1 unless CP is present . Our single - molecule experiments bring valuable insights into twin \ufb01 lin \u2019 s uncapping mechanism ( Fig . 4 ) . We found that unlike co \ufb01 lin , which uncaps by decorating actin \ufb01 lament sides 45 , twin \ufb01 lin directly associates with CP - bound barbed ends to uncap actin \ufb01 laments . Twin \ufb01 lin on its own is , at best , a weak uncapper and only increases the Fig . 2 | Effect of twin \ufb01 lin on capping protein ( CP ) \u2014 formin decision complex . a Experimental strategy schematic . Actin \ufb01 laments were nucleated from coverslip - anchored formins by \ufb02 owing 1 \u00b5 M G - actin ( 15 % Alexa - 488 labeled ) and 0 . 5 \u00b5 M pro \ufb01 lin . Filaments were then exposed to \ufb02 ow containing 1 \u00b5 M unlabeled G - actin , 4 \u00b5 M pro \ufb01 lin , and 1 \u00b5 M CP for ~ 10s to convert formin - bound barbed ends ( BF ) to formin - CP - bound barbed ends ( BF + C \u2192 BFC ) . BFC complexes were then exposed to \ufb02 ow containing PA only or with a range of mTwf1 concentrations . b Representative \ufb01 lament kymographs are transitioning to the BFC state upon exposure to 1 \u00b5 M CP . Upon CP removal from solution and exposure to PA ( with or without mTwf1 ) , \ufb01 laments resume elongation following CP dissociation from BFC ( top , BFC \u2192 BF + C ) or detach from formin ( bottom , BFC \u2192 BC + F ) . White arrow - heads denote BFC complex dissociation . c Dissociation fraction of formin - CP - bound \ufb01 laments ( BFC ) as a function of time in the presence of PA with / without a range of mTwf1 concentrations . Experimental data ( symbols ) are \ufb01 tted to a single - exponential function ( lines ) to determine the rate of BFC disassembly . Number of \ufb01 lamentsanalyzedpercondition ( 0to1 \u00b5 MmTwf1 ) : 31 , 51 , 30 , 44 , 42 , and50 . d BFC dissociation rate into BC or BF as a function of mTwf1 concentration , determined from ( c ) . e Fraction of BFC ( black symbols ) \ufb01 laments transitioning to BF ( magenta symbols ) or BC ( yellow symbols ) . Experimental data ( symbols ) are \ufb01 tted to expo - nential \ufb01 ts ( lines ) , suchthat k - BFC = k \u2019 - F + k \u2019 - C , where k \u2019 - F istheformindissociationrate from BFC ( BFC \u2192 BC + F ) and k \u2019 - C is CP dissociation rate from BFC ( BFC \u2192 BF + C ) . Conditions \u2013 left ( 0nM mTwf1 , 50 \ufb01 laments ) , center ( 50nM mTwf1 , 37 \ufb01 laments ) , right ( 1 \u00b5 M mTwf1 , 30 \ufb01 laments ) . See Supplementary Fig . 2 for the range of mTwf1 concentrations . f Dissociation rate ofCP ( k \u2019 - C ) or Formin ( k \u2019 - F ) from BFC complexes g PercentagesofBFCcomplexesin ( e ) transitioningtoBC ( yellow ) orBF ( magenta ) at different mTwf1 concentrations . Error bars in ( d ) indicate 65 % con \ufb01 dence intervalsbasedon \ufb01 ts ( seemethods ) . SourcedataareprovidedasaSourceData \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 6 0 10 20 0 25 50 75 100 125 150 1a i D m - 94 6 : P C - 945 f o e m i t e f il nae M ) s ( s e x e l p m o c no i s i c ed [ mTwf1 ] ( nM ) 0 s 368 s 184 s mDia1 CP Merge 0 s 92 s 184 s 276 s 368 s 2 \u00b5 m 0 250 500 0 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 no i t c a r F F r a c t i on F r a c t i on 0 nM mTwf1 5 nM mTwf1 20 nM mTwf1 0 250 500 0 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 Residence time of 549 - CP & 649 - mDia1 complexes at barbed ends ( s ) 0 250 500 0 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 a b c d e f 0 1 649 - mDia1 549 - CP 488 - Actin 0 1 e c ne cs e r ou l F i n t en s i t y ( a r b . un i t s ) 0 100 200 300 400 0 4 8 12 t ne m a li F l eng t h ( \u00b5 m ) Time ( s ) 0 1 649 - mDia1 549 - CP 488 - Actin 0 1 e c ne cs e r ou l F y t i s ne t n i ) s t i nu . b r a ( 0 100 200 0 4 8 12 16 t ne m a li F l eng t h ( \u00b5 m ) Time ( s ) BF BFC BC BF BFC BF g 0 1 649 - mDia1 549 - mTwf1 488 - Actin 0 1 e c ne cs e r ou l F i n t en s i t y ) s t i nu . b r a ( 0 50 100 150 0 1 2 3 F il a m en t l eng t h ( \u00b5 m ) Time ( s ) BF BFC BF BFCT 138 s 128 s 133 s mDia1 mTwf1 Merge h Fig . 3 | Direct visualization of formin , CP , and mTwf1 at barbed ends . a Representative time - lapse images of a multicolor single - molecule TIRF experi - ment ( see Supplementary Movie 4 ) . Actin \ufb01 laments were initiated byincubationof 0 . 5 \u00b5 MG - actin ( 15 % Alexa - 488labeled , 0 . 5 % biotin - labeled ) , 1 \u00b5 Mpro \ufb01 lin , and50pM 649 - mDia1 ( magenta ) . 649 - mDia1 bound \ufb01 laments were then exposed to PA and 10nM 549 - CP ( yellow ) withorwithout mTwf1 . The white boxaround each \ufb01 lament end indicates the location of the barbed end . Insets show individual and merged channels localized at the barbed end . Scale bar , 2 \u00b5 m . b Fluorescence images of a 13 \u00d7 13 - pixel box around the barbed end of the \ufb01 lament from ( a ) show the for - mation and dissociation of a mDia1 \u2013 CP complex at the barbed end . The Formin channel is magenta and the CP channel is yellow . c Fluorescence intensity and length records of the \ufb01 lament in ( a ) with formation and resolution of the decision complex in the absence of mTwf1 . d Same as ( c ) but in the presence of 20nM mTwf1 . Gray shaded boxes indicate the period when both 549 - CP and 649 - mDia1 were simultaneously present at the barbed end , i . e . , the BFC complex . e Distributionoflifetimesof649 - mDia1 : 549 - CPdecisioncomplexesatbarbedends in presence of 0nM mTwf1 ( left , n = 110 BFC complexes ) , 5nM mTwf1 ( center , n = 60 BFC complexes ) , and 20nM mTwf1 ( right , n = 57 BFC complexes ) . f Mean lifetimes ( \u00b1sem ) of 649 - mDia1 : 549 - CP decision complexes at barbed ends as a function of mTwf1 concentration , determined from data in ( e ) . g Fluorescence intensity and length records of a \ufb01 lament with formation and resolution of the decision complex formed in the presence of 100pM 649 - mDia1 , 10nM unlabeled CP , and 40nM 549 - mTwf1 . The gray shaded box indicates the time duration when both 549 - mTwf1 , 649 - mDia1 , and unlabeledCPwere simultaneously present atthe barbed end ( BFCT ) . h Cropped \ufb02 uorescence images of a 13 \u00d7 13 - pixel box around thebarbedendofthe \ufb01 lamentin ( g ) showtheformationanddissociationofa649 - mDia1 \u2013 unlabeledCPcomplexinthepresenceof549 - mTwf1 . Theforminchannelis magenta and twin \ufb01 lin channel is yellow . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 7 rate of uncapping by about sixfold . A successful uncapping event , on average , takes about 31 separate twin \ufb01 lin association and disassocia - tion events at the barbed end . In contrast , CARMIL accelerates CP \u2019 s departure by about 180 - fold , bringing down CP \u2019 s barbed - end dwell time from 30 min to just about 10s 46 . As a result , although both CARMIL and twin \ufb01 linbind CP similarlyviatheir CPI motifs 47 , they exert vastly different control on CP . In addition to uncapping , twin \ufb01 lin is also an actin depolymerase 17 , 18 . How does it carry out these two distinct activities ? Our experiments with twin \ufb01 lin variants containing mutations in the actin - binding ADF domains suggest that twin \ufb01 lin \u2019 s interactions with the actin \ufb01 lament are essential for it to rescue formin \u2019 s processivity from CP . Mutations in twin \ufb01 lin \u2019 s tail domain interfere with direct twin \ufb01 lin - CP 20 binding . Although less strongly than the wildtype , we 0 2 4 6 8 10 0 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 no i t c a r F 549 - mTwf1 dwell times on CP - bound ends ( s ) 1 . 9 \u00b1 0 . 1 s b 15s 1 \u00b5 m 647 - CapZ 488 - Actin Merge 549 - mTwf1 e d c a 125 . 5 s 127 . 1 s mTwf1 0 1 0 50 100 150 0 1 e c ne cs e r ou l F i n t en s i t y ( a r b . un i t s ) Time ( s ) m T w f 1 E v en t s Twinfilin ( labeled ) CP ( unlabeled ) Actin ( labeled ) Twinfilin ( labeled ) CP ( labeled ) Actin ( labeled ) Fig . 4 | Visualization and characterization of twin \ufb01 lin \u2019 s interactions with CP - bound barbedends . a Schematicofthree - colorsingle - molecule experiments with labeledCPandlabeledmTwf1 . Actin \ufb01 lamentswereassembledfromof1 \u00b5 MG - actin ( 15 % Alexa - 488 labeled , 1 . 4 % biotin - labeled ) and then capped by 10nM 649 - CP . These \ufb01 lamentswerethen exposed to 20nM 549 - mTwf1 . The greenarrow denotes polymerization . b Binding of labeled twin \ufb01 lin on CP - bound \ufb01 lament barbed ends recorded at 1s time resolution . Kymographs show Alexa - 488 actin ( top ) , 649 - CP ( secondfromtop ) , 549 - mTwf1 ( thirdfromtop ) , andmerge ( bottom ) . Magentabars denote episodes in whicha 549 - mTwf1 moleculewas present at the 649 - CP - bound barbed end of the \ufb01 lament . c Schematic of the two - color single - molecule experi - ments with unlabeled CP and labeled mTwf1 . Actin \ufb01 laments were polymerized from 1 \u00b5 M G - actin ( 15 % Alexa - 488labeled , 1 . 4 % biotin - labeled ) and0 . 5 \u00b5 M pro \ufb01 ling , then capped by unlabeled CP . The capped \ufb01 laments were then exposed to 15nM 549 - mTwf1 . Arrival and departureof 549 - mTwf1 molecules at the barbed end were recorded until the \ufb01 lament started depolymerizing , i . e . , CP dissociated . The green arrow denotes polymerization , and the red arrow denotes depolymerization . d Distribution of residence times of 549 - mTwf1 on unlabeled CP - bound \ufb01 lament barbed ends ( n = 510 binding events across 16 \ufb01 laments ) . Mean dwell time = 1 . 9\u00b1 0 . 1s ( \u00b1sem ) . The histogram represents 99 % of all binding events ( remaining 1 % outliers are not shown ) . 100 % of binding events were included in calculations ofmeandwelltime . e Top : timerecordsof549 - mTwf1 \ufb02 uorescenceintensityatthe unlabeled CP - bound barbed end of an actin \ufb01 lament . Intensity is integrated over a 5 \u00d7 5 - pixel square centered around the barbed end of the \ufb01 lament . Bottom : pre - sence ( 1 ) or absence ( 0 ) of 549 - mTwf1 molecules at the CP - 649 - bound \ufb01 lament barbed end shown above . Inset : cropped \ufb02 uorescence images of a 10 \u00d7 10 - pixel box around the barbed end of the \ufb01 lament showing the arrival , presence , and departureofasingle549 - mTwf1moleculeattheCP - boundbarbedend . Sourcedata are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 8 \ufb01 nd that the tail mutant can still dissociate CP from the formin - CP barbed end complex and modestly rescue formin \u2019 s processivity from CP . Thiscanbeinterpretedintwoways . Firstly , fortwin \ufb01 lin \u2019 seffectson the BFC complex , its direct binding to CP is bene \ufb01 cial but not neces - sary . Interestingly , a previous study found that the tail mutant uncaps CP from free barbed ends much faster than the wildtype 20 . Alter - natively , thiscouldalsomeanthat thereareadditionalsitesintwin \ufb01 lin , apart from its tail , which participate in its direct binding to CP . Indeed , the linker region between the two ADF domains in twin \ufb01 lin has been proposed as an extra site of contact between CP and twin \ufb01 lin when both are bound to actin 37 . Importantly , mutations in ADF domains , which interfere with twin \ufb01 lin \u2019 s actin binding and extinguish twin \ufb01 lin \u2019 s uncapping activities at free barbed ends 20 , also extinguish twin \ufb01 lin \u2019 s effects on BFC complex dissociation . Together , these results indicate that in the ternary BFCT complex comprising of formin , CP , and twin \ufb01 lin at barbed ends , each factor is directly interacting with the actin \ufb01 lament and not via each other . How would this multicomponent mechanism be relevant in vivo ? Twin \ufb01 lin can depolymerize free barbed ends even in the presence of polymerizable actin monomers 17 , 18 , 20 . In such cytosol - mimicking con - ditions , the simultaneous presence of twin \ufb01 lin and CP would strongly favor \ufb01 lament capping and actin disassembly over actin assembly . How , then , would intracellular actin assembly occur at all ? Combining our observations with previous studies , we present a working model for how polymerases , depolymerases , and cappers can simultaneously regulate actin assembly ( Fig . 6 ) . Eventual barbed - end outcomes would depend upon whether the barbed end was bound to a formin or not . Free barbed ends would \ufb01 rst rapidly get capped by CP ( Fig . 6a ) . Fol - lowing uncapping by twin \ufb01 lin , barbed ends would undergo twin \ufb01 lin - mediated depolymerization 20 . Formin - bound barbed ends , on the other hand , would \ufb01 rst get paused by CP , leading to the formation of BFCdecisioncomplexes ( Fig . 6b ) 29 , 30 . Twin \ufb01 lin \u2019 ssubsequentbindingto BFC complexes to form BFCT complexes would cause CP \u2019 s dissocia - tion and resumption of formin - based \ufb01 lament elongation . As the \ufb01 la - ment ages , other actin - binding proteins with a preference for ADP - F - actin , e . g . , co \ufb01 lin and cyclase - associated protein ( CAP ) , would initiate \ufb01 lament severing and pointed - end depolymerization 45 , 48 \u2013 50 . As a result , \ufb01 lamentswithfreebarbedendswouldrapidlydepolymerizefromboth ends into monomers and formin - bound \ufb01 laments would continue polymerizing at their barbed ends while depolymerizing from their pointed ends , akin to treadmilling . Therefore , the depolymerase twin \ufb01 lin can promote polymerization in the presence of both CP and formin . Although surprising , twin \ufb01 lin \u2019 s ability to promote assembly is not unique . Twin \ufb01 lin is part of the ADF homology ( ADF - H ) family of proteins and contains two ADF domains 51 . Like twin \ufb01 lin , co \ufb01 lin also promotes uncapping and accelerates \ufb01 lament depolymerization ( on top of its severing activities ) 45 , 50 , which in turn promotes assembly by replenishing the pool of monomeric actin . Where in a cell would the mechanisms uncovered here be rele - vant ? Twin \ufb01 lin , CP , and formin operate simultaneously in a number of Control WT mTwf1 Tail mutant ADF mutant 0 . 00 0 . 02 0 . 04 0 . 06 0 . 08 0 . 10 0 . 12 s ( e t a r no i t a i c o ss i d C F B - 1 ) a ADF - H ADF - H ADF - H ADF - H ADF - H ADF - H 1 139 175 313 350 WT mTwf1 Tail mutant ADFdomainmutant b c 0 100 200 300 400 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 ControlmTwf1 WT mTwf1 tail mutant mTwf1 ADF mutant no i t c a r f l a v i v r u s ) C F B ( x e l p m o c no i s i c e D Time ( s ) Flow Formin Actin ( labeled ) Actin Profilin Twinfilin ( WT / mutants ) Coverslip d Capping protein Fig . 5 | Twin \ufb01 lin \u2019 s direct interaction with actin \ufb01 lament is essential for its effects on formin - CP decision complexes . a Domain diagram of wild - type and mutant mTwf1 constructs used here 20 . \u201c x \u201d denotes the location of mutations . b Schematic representation of the experimental strategy . Actin \ufb01 laments were nucleated from coverslip - anchored formins by introducing a \ufb02 ow containing 1 \u00b5 M G - actin ( 15 % Alexa - 488 labeled ) and 0 . 5 \u00b5 M pro \ufb01 lin . The \ufb01 laments were then exposed to a \ufb02 ow containing 1 \u00b5 M unlabeled G - actin , 4 \u00b5 M pro \ufb01 lin , and 500nM CP for about 10s to convert formin - bound barbed ends ( BF ) to formin - CP - bound barbed ends or decision complexes ( BF + C \u2192 BFC ) . These BFC complexes were then exposed to a \ufb02 ow containing PA only or supplemented with wildtype or mutant mTwf1 . c Survival fraction of formin - CP - bound \ufb01 laments ( BFC complexes ) as a function of time in the presence of PA only ( black symbols , 92 \ufb01 laments ) , or supplemented with 1 \u00b5 M wild - typemTwf1 ( black , 36 \ufb01 laments ) , mTwf1ADFmutant ( green , 77 \ufb01 laments ) , or mTwf1 tail mutant ( blue symbols , 22 \ufb01 laments ) . Experi - mental data ( symbols ) were \ufb01 tted to a single - exponential function ( lines ) to determine BFC dissociation rate k - BFC . Error bars indicate 65 % con \ufb01 dence intervals based on \ufb01 ts ( see methods ) . d BFC dissociation rate for wild - type and mutant mTwf1 , determined from data in ( c ) . Error bars in ( d ) indicate 65 % con \ufb01 dence intervalsbasedon \ufb01 ts ( seeMethods ) . SourcedataareprovidedasaSourceData \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 9 cellular compartments , including \ufb01 lopodia , lamellipodia , and stereocilia 6 , 18 , 21 \u2013 24 where size and elongation rates of actin \ufb01 laments are tightly controlled . Given the high barbed - end af \ufb01 nity and intracellular concentration of barbed - end binding proteins like formin , CP , twin \ufb01 - lin , and Ena / VASP , we expect that newly nucleated free barbed ends initiated by Arp2 / 3complex wouldbe rapidly captured by one of these factors . Althoughforminsaremainlythoughtofinthecontextoflinear actin structures , formins like FMNL2 have been shown to play a key role in branched actin networks at the leading edge 24 . In light of our results , we speculate that majority of growing barbed ends in cells might be protected by formins or Ena / VASP . In the future , it will be interesting to explore if other biochemical or mechanical factors which in \ufb02 uence activities of CP and / or formin , e . g . , CARMIL 16 , 36 , V1 / myotrophin 36 , Spire 52 , Bud14 53 , IQGAP 54 , 55 , or Hof1 56 , polyphosphoinositides 42 , 57 orforce 34 mightalsoin \ufb02 uenceBFCcomplex dynamics totune actin assembly . Itwillalsobe important toexamineif the interactions between twin \ufb01 lin , CP , and formin reported here might also play a role in human diseases and disorders , including cancer invasionandprogression 58 , 59 , hearingloss 60 aswellasneuropathiesand cardiac conditions 61 in which these proteins have been implicated . Methods Puri \ufb01 cation and labeling of actin Rabbit skeletal muscle actin was puri \ufb01 ed from acetone powder gen - erated from frozen ground hind leg muscle tissue of young rabbits ( PelFreez , USA ) . Lyophilized acetone powder stored at \u2212 80\u00b0C was mechanically sheared in a coffee grinder , resuspended in G - buffer ( 5 mM Tris - HCl pH 7 . 5 , 0 . 5mM Dithiothreitol ( DTT ) , 0 . 2mM ATP , and 0 . 1 mM CaCl 2 ) , and cleared by centrifugation for 20 min at 50 , 000 \u00d7 g . The supernatant was collected and further \ufb01 ltered with the Whatman paper . Actinwasthenpolymerizedovernightat4 \u00b0C , slowlystirring , by the addition of 2 mM MgCl 2 and 50mM NaCl to the \ufb01 ltrate . The next morning , NaClpowderwasadded toa \ufb01 nal concentration of 0 . 6M and stirring was continued for another 30min at 4 \u00b0C . Then , F - actin was pelleted by centrifugation for 150min at 280 , 000 \u00d7 g , the pellet was solubilized by Dounce homogenization and dialyzed against G - buffer for 48 h at 4\u00b0C . Monomeric actin was then precleared at 435 , 000 \u00d7 g and loaded onto a Sephacryl S - 200 16 / 60 gel - \ufb01 ltration column ( Cytiva , USA ) equilibrated in G - Buffer . Fractions containing actin were stored at 4 \u00b0C . To biotinylate actin , puri \ufb01 ed G - actin was \ufb01 rst dialyzed overnight at 4 \u00b0C against G - buffer lacking DTT . The monomeric actin was then polymerized by the addition of an equal volume of 2X labeling buffer ( 50mM imidazole pH 7 . 5 , 200mM KCl , 0 . 3 mM ATP , 4 mM MgCl 2 ) . After 5 min , the actin was mixed with a \ufb01 vefold molar excess of NHS - XX - Biotin ( Merck KGaA , Germany ) and incubated inthe darkfor 15 h at 4 \u00b0C . The F - actin was pelleted as above , and the pellet was rinsed with G - buffer , then homogenized with a Dounce and dialyzed against G - buffer for 48 h at 4 \u00b0C . Biotinylated monomeric actin was puri \ufb01 ed further on a Sephacryl S - 200 16 / 60 gel - \ufb01 ltration column as above . Aliquots of biotin - actin were snap frozen in liquid N 2 and stored at \u2212 80\u00b0C . To \ufb02 uorescently label actin , G - actin was polymerized by dialyzing overnight against a modi \ufb01 ed F - buffer ( 20mM PIPES pH 6 . 9 , 0 . 2mM CaCl 2 , 0 . 2mMATP , and 100mMKCl ) 32 . F - actin wasincubated for 2h at room temperature with a \ufb01 vefold molarexcessof Alexa - 488 NHS ester dye ( Thermo Fisher Scienti \ufb01 c , USA ) . F - actin was then pelleted by centrifugationat450 , 000\u00d7 g for 40minatroomtemperature , andthe pellet was resuspended in G - buffer , homogenized with a Dounce , and incubatedonicefor2 htodepolymerizethe \ufb01 laments . Themonomeric actin was then re - polymerized on ice for 1 h by the addition of 100 mM KCl and 1 mM MgCl 2 . F - actin was once again pelleted by centrifugation for 40min at 450 , 000 \u00d7 g at 4 \u00b0C . The pellet was homogenized with a Dounce and dialyzed overnight at 4 \u00b0C against 1 L of G - buffer . The solution was precleared by centrifugation at 450 , 000 \u00d7 g for 40 min at 4 \u00b0C . The supernatant was collected , and the concentration and labeling ef \ufb01 ciency of actin was determined . Puri \ufb01 cation and labeling of mTwf1 polypeptides Wildtype and mutant mouse mTwf1 plasmids were a gift from Pekka Lappalainen 20 . All of these proteins were expressed in E . coli BL21 ( pRare ) . Cells were grown in Terri \ufb01 c Broth to log phase at 37\u00b0C . Formin ( F ) ADP - Actin ATP - Actin Capping protein ( C ) ADP - P i - Actin Twinfilin ( T ) Cofilin Cyclase associated protein ( CAP ) b a BC BCT BT B BF BFC BFCT BF Formin - bound barbed ends Free barbed ends Fig . 6 | Working model for regulation of actin dynamics by twin \ufb01 lin , formin , andCP . Barbed - endoutcomeswoulddependuponwhetherbarbedendswerefree or formin - bound . a Free barbed ends ( B ) would rapidly get capped by CP ( C ) , followedbyCP \u2019 sdissociationbytwin \ufb01 lin ( T ) . Thiswouldleavetwin \ufb01 linaloneatthe \ufb01 lamentend , causingitsdepolymerization . Atthesametime , co \ufb01 linwouldbindthe sides of the aging \ufb01 lament and synergize with cyclase - associated protein ( CAP ) to initiate the \ufb01 lament \u2019 s pointed - end depolymerization . The simultaneous depoly - merizationatthetwoendswouldresultinthecompletedisassemblyofthe \ufb01 lament into monomers . b Formin ( F ) bound barbed ends would get paused by CP to form BFC complexes . Twin \ufb01 lin \u2019 s binding to BFC complexes would cause CP \u2019 s dissocia - tion and renewal of formin - based \ufb01 lament elongation . As a result , the \ufb01 lament wouldappeartoatreadmill , i . e . , continueelongatingatthebarbedendwhileatthe same time being disassembled at the pointed end by CAP - co \ufb01 lin synergy . \u201c B \u201d denotesbarbedend , \u201c F \u201d denotesformin , \u201c C \u201d denotesCP , and \u201c T \u201d denotestwin \ufb01 lin . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 10 Expression was induced overnight at 18 \u00b0C by the addition of 1 mM IPTG . Cells were harvested by centrifugation at 11 , 200 \u00d7 g for 15min and the cell pellets were stored at \u2212 80\u00b0C . For puri \ufb01 cation , frozen pellets were thawed and resuspended in 35mL lysis buffer ( 50 mM sodium phosphate buffer pH 8 , 20 mM imidazole , 300mMNaCl , 1 mM DTT , 1 mM PMSF , and protease inhibitors ( pepstatin A , antipain , leu - peptin , aprotinin , and chymostatin , 0 . 5 \u03bc M each ) ) . Cells were lysed using a tip sonicator while kept on ice . The cell lysate was then cen - trifuged at 120 , 000 \u00d7 g for 45min at 4\u00b0C . The supernatant was then incubated with 1 mL of Ni - NTA beads ( Qiagen , USA ) while rotating for 2 h at 4 \u00b0C . The beads were then washed three times with the wash buffer ( 50mM sodium phosphate buffer pH 8 , 300mM NaCl , 20 mM imidazole , and 1 mM DTT ) . The beads were then transferred to a dis - posable column ( Bio - Rad , USA ) . Protein was eluted using the elution buffer ( 50mM phosphate buffer pH 8 , 300 mM NaCl , 250mM imida - zole , and 1mM DTT ) . Fractions containing the protein were con - centrated and loaded onto a size exclusion Superdex 75 Increase 10 / 300 column ( Cytiva , USA ) pre - equilibrated with 20mM HEPES pH 7 . 5 , 1 mM EDTA , 50 mM KCl , and 1 mM DTT . Peak fractions were collected , concentrated , aliquoted , and \ufb02 ash - frozen in liquid N 2 and stored at \u2212 80\u00b0C . Mouse His - SNAP - mTwf1 plasmid was ordered from Twist Bios - ciences . SNAP - mTwf1 was puri \ufb01 ed using the same protocol as above . Puri \ufb01 ed SNAP - mTwf1 was incubated with 5x excess of SNAP - surface - 549dye ( NewEnglandBiolabs , Ipswich , MA ) overnightat4\u00b0C . Freedye was removed using a PD - 10 desalting column ( Cytiva , USA ) . Labeled protein was collected , concentrated , aliquoted , and \ufb02 ash - frozen in liquid N 2 and stored at \u2212 80\u00b0C . Puri \ufb01 cation , labeling , and biotinylation of formin mDia1 Mouse his - tagged mDia1 ( FH1 - FH2 - C ) formin was expressed in E . coli ; BL21 ( DE3 ) pLysScells . CellsweregrowninTerri \ufb01 cBrothtologphaseat 37\u00b0C . Expression was induced overnight at 18\u00b0C by the addition of 1 mM IPTG . Cells were harvested by centrifugation at 11 , 200 \u00d7 g for 15min and the cell pellets were stored at \u2212 80 \u00b0C . For puri \ufb01 cation , frozen pellets were thawed and resuspended in 35 mL lysis buffer ( 50mM sodium phosphate buffer pH 8 , 20mM imidazole , 300 mM NaCl , 1 mM DTT , 1 mM PMSF , and protease inhibitors ( 0 . 5 \u03bc M each of pepstatin A , antipain , leupeptin , aprotinin , and chymostatin ) ) . Cells were lysed using a tip sonicator while being kept on ice . The cell lysate wasthencentrifugedat120 , 000\u00d7 g for45minat4 \u00b0C . Thesupernatant was then incubated with 1 mL of Ni - NTA beads ( Qiagen , USA ) while rotating for 2 h at 4 \u00b0C . The beads were then washed three times with the wash buffer ( 50mMsodium phosphate buffer pH 8 , 300mMNaCl , 20mM imidazole , and 1 mM DTT ) and were then transferred to a dis - posable column ( Bio - Rad , USA ) . Protein was eluted using the elution buffer ( 50mM phosphate buffer pH 8 , 300 mM NaCl , 250mM imida - zole , and 1mM DTT ) . Fractions containing the protein were con - centrated and loaded onto a size exclusion Superdex 200 increase 10 / 300 GL column ( Cytiva , USA ) pre - equilibrated with 20 mM HEPES pH 7 . 5 , 150 mM KCl , 10 % glycerol , and 0 . 5mM DTT . Peak fractions were collected , concentrated , aliquoted , and \ufb02 ash - frozen in liquid N 2 and stored at \u2212 80\u00b0C . SNAP - mDia1 30 was expressed and puri \ufb01 ed using the protocol above . Puri \ufb01 ed SNAP - mDia1 was incubated with 5x excess of SNAP - surface - 649 dye ( New England Biolabs , USA ) overnight at 4 \u00b0C . Free dye was removed using a Superdex 200 increase 10 / 300 GL column ( Cytiva , USA ) . Labeled protein was collected , concentrated , aliquoted , and \ufb02 ash - frozen in liquid N 2 and stored at \u2212 80\u00b0C . Biotin - SNAP - mDia1 was prepared by incubating puri \ufb01 ed SNAP - mDia1 with Benzylguanine - Biotin ( New England Biolabs , USA ) accord - ing to the manufacturer \u2019 s instructions . Free biotin was removed using size - exclusion chromatography by loading the labeled protein on a Superose6gel - \ufb01 ltrationcolumn ( GEHealthcare , Pittsburgh , PA ) eluted with 20mM HEPES pH 7 . 5 , 150 mM KCl , and 0 . 5mM DTT . Puri \ufb01 cation and labeling of capping protein Mouse his - tagged capping protein was expressed in E . coli ; BL21 ( DE3 ) pLysS cells . Capping protein subunits \u03b1 1 and \u03b2 2 were expressed from the same plasmid with a single His - tag on the alpha subunit 29 . Cells were grown in Terri \ufb01 c Broth to log phase at 37\u00b0C . Expression was induced overnight at 18 \u00b0C by the addition of 1 mM IPTG . Cells were harvested by centrifugation at 11 , 200 \u00d7 g for 15min and the cell pellets were stored at \u2212 80\u00b0C . For puri \ufb01 cation , frozenpellets werethawed and resuspended in 35mL lysis buffer ( 50 mM sodium phosphate buffer pH 8 , 20 mM imidazole , 300mM NaCl , 1 mM DTT , 1mM PMSF , and protease inhibitors ( 0 . 5 \u03bc M each of pepstatin A , antipain , leupeptin , aprotinin , and chymostatin ) ) . Cells were lysed using a tip sonicator while being kept on ice . The cell lysate was then centrifuged at 120 , 000 \u00d7 g for 45 min at 4 \u00b0C . The supernatant was incubated with 1 mL of Ni - NTA beads ( Qiagen , USA ) while rotating for 2 h at 4 \u00b0C . The beads were then washed three times with the wash buffer ( 50mM sodium phosphate buffer pH 8 , 300mM NaCl , 20mM imidazole , and 1 mM DTT ) and transferred to a disposable column ( Bio - Rad , USA ) . Proteinwaselutedusingelutionbuffer ( 50mMphosphatebufferpH8 , 300mM NaCl , 250 mM Imidazole , and 1 mM DTT ) . Fractions contain - ing the protein were concentrated and loaded onto a size exclusion Superdex 75 Increase 10 / 300 column ( Cytiva , USA ) pre - equilibrated with 20 mM Tris - HCl , 50 mM KCl , and 1 mM DTT . Peak fractions were collected , concentrated , aliquoted , and \ufb02 ash - frozen in liquid N 2 and stored at \u2212 80\u00b0C . SNAP - CP wasexpressed from a singleplasmidcontaining His - and SNAP - tagged \u03b2 1 subunit and untagged \u03b1 1 subunit 30 . It was puri \ufb01 ed using the protocol above . Puri \ufb01 ed SNAP - CP was incubated with 5x excess of SNAP - surface - 549 dye or SNAP - surface - 649 dye ( New Eng - landBiolabs , USA ) overnightat4 \u00b0C . FreedyeswereremovedusingPD - 10 desalting columns ( Cytiva , USA ) . Labeled protein was collected , concentrated , aliquoted , and \ufb02 ash - frozen in liquid N 2 and stored at \u2212 80\u00b0C . Puri \ufb01 cation of pro \ufb01 lin Human pro \ufb01 lin - 1 was expressed in E . coli strain BL21 ( pRare ) to log phase in LB broth at 37 \u00b0C and induced with 1 mM IPTG for 3 h at 37\u00b0C . Cells were then harvested by centrifugation at 15 , 000 \u00d7 g at 4 \u00b0C and stored at \u2212 80\u00b0C . For puri \ufb01 cation , pellets were thawed and resus - pended in 30mL lysis buffer ( 50mM Tris - HCl pH 8 , 1mM DTT , 1 mM PMSF protease inhibitors ( 0 . 5 \u03bc M each of pepstatin A , antipain , leu - peptin , aprotinin , and chymostatin ) ) was added , and the solution was sonicated on ice by a tip sonicator . The lysate was centrifuged for 45min at 120 , 000 \u00d7 g at 4 \u00b0C . The supernatant was then passed over 20ml of Poly - L - proline conjugated beads in a disposable column ( Bio - Rad , USA ) . The beads were \ufb01 rst washed at room temperature in wash buffer ( 10 mMTrispH8 , 150 mMNaCl , 1 mMEDTA , and1 mMDTT ) and then washed again with two column volumes of 10mM Tris pH 8 , 150 mM NaCl , 1 mM EDTA , 1 mM DTT , and 3 M urea . Protein was then eluted with \ufb01 ve column volumes of 10mM Tris pH 8 , 150 mM NaCl , 1 mM EDTA , 1 mM DTT , and 8 M urea . Pooled and concentrated frac - tions were then dialyzed in 4 L of 2 mM Tris pH 8 , 0 . 2 mM EGTA , 1 mM DTT , and 0 . 01 % NaN 3 ( dialysis buffer ) for 4 h at 4\u00b0C . The dialysis buffer was replaced with fresh 4 L buffer and the dialysis was continued overnight at 4 \u00b0C . The protein was centrifuged for 45min at 450 , 000 \u00d7 g at 4 \u00b0C , concentrated , aliquoted , \ufb02 ash - frozen in liquid N 2 , and stored at \u2212 80 \u00b0C . Conventional TIRF microscopy for single - molecule imaging Glass coverslips ( 60 \u00d7 24 mm ; Thermo Fisher Scienti \ufb01 c , USA ) were \ufb01 rst cleaned by sonication in detergent for 20min , followed by suc - cessive sonications in 1 M KOH , 1M HCl , and ethanol for 20min each . Coverslips were then washed extensively with H 2 O and dried in an N 2 stream . The cleaned coverslips were coated with 2 mg / mL methoxy - polyethylene glycol ( mPEG ) - silane MW 2000 and 2 \u00b5 g / mL biotin - PEG - Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 11 silane MW 3400 ( Laysan Bio , USA ) in 80 % ethanol ( pH 2 . 0 ) and incu - bated overnight at 70 \u00b0C . Flow cells were assembled by rinsing PEG - coated coverslips with water , drying with N 2 , and adhering to \u03bc - Slide VI0 . 1 ( 0 . 1mm \u00d7 17 mm \u00d7 1 mm ) \ufb02 ow chambers ( Ibidi , Germany ) with double - sided tape ( 2 . 5cm \u00d7 2 mm \u00d7 120 \u03bc m ) and epoxy resin for 5min ( Devcon , USA ) . Before each reaction , the \ufb02 ow cell was sequen - tially incubated for 1 min each with 4 \u03bc g / ml streptavidin and 1 % BSA in 20mM HEPES pH 7 . 5 , and 50 mM KCl . The \ufb02 ow cell was then equili - brated with TIRF buffer ( 10mM imidazole , pH 7 . 4 , 50mM KCl , 1mM MgCl 2 , 1 mM EGTA , 0 . 2 mM ATP , 10mM DTT , 2 mM DABCO , and 0 . 5 % methylcellulose [ 4000cP ] ) . For 649 - mDia1 , 549 - CP , and 488 - actin experiments ( Fig . 3a \u2013 f ) , 0 . 5 \u00b5 M G - actin ( 15 % Alexa - 488 labeled , 0 . 5 % biotin - labeled ) , 1 \u00b5 M pro - \ufb01 lin , along with 50pM 649 - mDia1 in TIRF buffer were introduced into the \ufb02 ow cell and \ufb01 laments were allowed to grow for 2 to 3 min . The \ufb02 ow cell was then rinsed with TIRF buffer to remove free formins , and thesolution was replacedwithpro \ufb01 lin - actinand 10nM549 - CP ( with or without mTwf1 ) . In experiments with \ufb02 uorescently labeled twin \ufb01 lin ( Fig . 3g , h ) , 100pM mDia1 was introduced to the \ufb02 ow cell , followed by \ufb02 ows containing 10nM unlabeled CP along with 40nM 549 - mTw1 . Time - lapse images were acquired every 4 s . For three - color uncapping experiments with 549 - mTwf1 , 649 - CapZ , and 488 - actin ( Fig . 4a , b ) , actin \ufb01 laments were elongated from 1 \u00b5 MG - actin ( 15 % Alexa - 488labeledand1 . 4 % biotinylatedG - actin ) . Free actin monomers were removed by rinsing the \ufb02 ow cell with an excess of TIRF buffer and the \ufb01 laments wereexposed to a solution containing 10nM 649 - CP . Following capping , the chamber was once again rinsed with TIRF buffer to remove free 649 - CP and then 20 nM 549 - mTwf1 was introduced . Time - lapse images were acquired every 1s . For two - color experiments involving 549 - mTwf1 , unlabeled CapZ , and 488 - actin ( Fig . 4c \u2013 e ) , actin \ufb01 laments were elongated from 1 \u00b5 M G - actin ( 15 % Alexa - 488 labeled and 1 . 4 % biotinylated G - actin ) and 0 . 5 \u00b5 M pro \ufb01 lin . Free pro \ufb01 lin and actin monomers were removed by rinsing the \ufb02 ow cell with an excess of TIRF buffer and the \ufb01 laments were exposed to a solution containing 10 nM unlabeled CP and 1 \u00b5 M G - actin ( 15 % Alexa - 488 labeled and 1 . 4 % biotinylated G - actin ) with 0 . 5 \u00b5 M pro \ufb01 lin . Following capping , the chamber was once again rinsed withTIRF buffertoremovefreeunlabeledCP , actin , pro \ufb01 ling , andthen 15nM 549 - mTwf1 was introduced . Time - lapse images were acquired either every 200 ms or every 300ms . Micro \ufb02 uidics - assisted TIRF ( mf - TIRF ) imaging and analysis Actin \ufb01 laments were \ufb01 rstassembled inmicro \ufb02 uidics - assisted TIRF ( mf - TIRF ) \ufb02 ow cells 32 , 33 . Coverslips were \ufb01 rst cleaned by sonication in Micro90 detergent for 20min , followed by successive 20min sonica - tions in 1 M KOH , 1M HCl , and 200 - proof ethanol for 20min each . Washed coverslips were then stored in fresh 200 - proof ethanol . Cov - erslips were then washed extensively with H 2 O and dried in an N 2 stream . These dried coverslips were coated with 2 mg / mL methoxy - poly ( ethyleneglycol ) ( mPEG ) - silaneMW2000and2 \u00b5 g / mLbiotin - PEG - silane MW 3400 ( Laysan Bio , USA ) in 80 % ethanol ( pH 2 . 0 ) and incu - bated overnight at 70\u00b0C . A 40 \u00b5 m high PDMS mold with three inlets and one outlet was mechanically clamped onto a PEG - Silane - coated coverslip . The chamber was then connected to a Maes \ufb02 o micro \ufb02 uidic \ufb02 ow - control system ( Fluigent , France ) , rinsed with mf - TIRF buffer ( 10 mM imidazole , pH 7 . 4 , 50 mM KCl , 1 mM MgCl 2 , 1 mM EGTA , 0 . 2 mM ATP , 10mM DTT , and 1 mM DABCO ) and incubated with 1 % BSA and 10 \u00b5 g / mLstreptavidin in 20 mM HEPES pH 7 . 5 , and 50 mM KCl for 5min . About 100pM Biotin - SNAP - mDia1 molecules in TIRF buffer were then \ufb02 owed in and allowed to anchor on the glasscoverslip . Actin \ufb01 laments with barbed ends anchored to the formins were grown by \ufb02 owing in a solution containing 1 \u00b5 M G - actin ( 15 % Alexa - 488 labeled ) and 0 . 5 \u00b5 M pro \ufb01 lin in mf - TIRF buffer . All experiments were carried out atroomtemperatureinaTIRFbuffer . Eachexperimentwasrepeatedat least three times . Data from a single replicate is presented in the \ufb01 gures . Image acquisition and analysis Single - wavelength time - lapseTIRF imaging was performed on a Nikon - Ti2000 inverted microscope equipped with a 40 mW Argon laser , a 60X TIRF - objective with a numerical aperture of 1 . 49 ( Nikon Instru - ments Inc . , USA ) , and an IXON LIFE 888 EMCCD camera ( Andor Ixon , UK ) . One pixel was equivalent to 144 \u00d7 144 nm . Focus was maintained by the Perfect Focus system ( Nikon Instruments Inc . , Japan ) . Time - lapsed images were acquired every 2 or 5s using Nikon Elements imaging software ( Nikon Instruments Inc . , Japan ) . For three - color images , the sample was sequentially excited by 488 , 561 , and 640nm lasers . Imageswereacquiredeithercontinuouslyorwitha1or4sdelay between consecutive images . Images were analyzed in Fiji 62 . Background subtraction was con - ducted using the rolling ball background subtraction algorithm ( ball radius \ufb01 ve pixels ) . Time - lapse images of between 50 and 100 \ufb01 laments were acquired in a single \ufb01 eld of view for each condition and all of these \ufb01 laments were included to determine the cumulative distribu - tion functions ( CDFs ) showing the time - dependent survival fraction of various complexes . For mf - TIRF assays , the kymograph plugin was used to draw kymographs of individual \ufb01 laments . The kymographs were used toidentifythetimepoint of thedetachment of \ufb01 lamentsasa function of time . For three - color images , a 5 \u00d7 5 - pixel box was drawn at the location of the barbed end of the \ufb01 lament and the time - dependent integrated intensity values were recorded for the single - molecule channels . The integrated intensity values were background corrected by subtracting the integrated intensity of a 5 \u00d7 5 - pixel box drawn away from the \ufb01 lament . Data analysisand curve \ufb01 tting werecarried out in MicrocalOrigin . Determination of rates of CP or formin dissociation from BFC complexes ( k \u2019 - F and k \u2019 - C ) The time - dependent fraction of BFC complexes dissociating by either transitioning to BF ( re - elongation ) or BC ( detachment ) upon addition of twin \ufb01 lin to the \ufb02 ow were plotted versus time ( black symbols in Fig . 2e ) . The kinetics of dissociation of BFC complex and simultaneous appearance of BF ( magenta symbols ) and BC ( yellow symbols ) were analyzed using the approach described in our earlier study 29 . Brie \ufb02 y , BFC dissociation can occur via one of the following two routes : BFC (cid:1) ! k 0(cid:1) C BF + C \u00f0 1 \u00de BFC (cid:1) ! k 0 (cid:1) F BC + F \u00f0 2 \u00de These reactions can be described by the following differential equations : d \u00f0 BFC \u00de = dt = (cid:1) BFC \u00f0 k 0(cid:1) F + k 0(cid:1) C \u00de \u00f0 3 \u00de d \u00f0 BF \u00de = dt = k 0(cid:1) C \u00f0 BFC \u00de \u00f0 4 \u00de d \u00f0 BC \u00de = dt = k 0(cid:1) F \u00f0 BFC \u00de \u00f0 5 \u00de The number of \ufb01 laments transitioning out of BFC , and into BF and BC allvaryexponentiallywithrateconstant ( k \u2019 - F + k \u2019 - C ) . Thevalueof k \u2019 - F and k \u2019 - C was derived from the relative fraction of \ufb01 laments in BF and BC states as follows . BFC \u00f0 t \u00de = BFC 0 = e (cid:1)\u00f0 k 0(cid:1) C + k 0(cid:1) F \u00de t \u00f0 6 \u00de Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 12 BF \u00f0 t \u00de = BFC 0 = k 0(cid:1) C k 0(cid:1) C + k 0(cid:1) F (cid:1) (cid:3) * \u00f0 1 (cid:1) e (cid:1)\u00f0 k 0(cid:1) C + k 0(cid:1) F \u00de t \u00de \u00f0 7 \u00de BC \u00f0 t \u00de = BFC 0 = k 0(cid:1) F k 0(cid:1) C + k 0(cid:1) F (cid:1) (cid:3) * \u00f0 1 (cid:1) e (cid:1)\u00f0 k 0(cid:1) C + k 0(cid:1) F \u00de t \u00de \u00f0 8 \u00de BFC 0 is the total number of \ufb01 laments in the BFC state just prior to \ufb02 owing in twin \ufb01 lin . The ratio of the number of \ufb01 laments taking either of the two routes ( BF or BC ) is given by N BF = N BC = k \u2019 - C / k \u2019 - F . Note that k \u2019 - F + k \u2019 - C are observed rate constants as they depend upon the con - centration of twin \ufb01 lin as well as elongation rates of \ufb01 laments asso - ciated with a formin . Statistical analysis and error bars for dissociation rates The uncertainty in dissociation rates of formin bound to barbed ends ( BF \u2192 B + F ) or BFC complexes ( BFC \u2192 BF + C or BFC \u2192 BC + F ) were determined by bootstrapping strategy 34 . The dissociation rate was determined by \ufb01 tting the survival fraction ( or CDF ) data to a single - exponential function ( y = e - k t or y = 1 \u2013 e - k t ) . A custom - written MATLAB code was then used to simulate BF ( or BFC ) complex lifetimes for N \ufb01 laments ( where N is the number of \ufb01 laments in the particular experiment ) based on the rate k determined from the experimental data . The simulation was repeated 1000 times to generate 1000 indi - vidual survival fractions of N \ufb01 laments . Each dataset was then \ufb01 t to an exponential function and an observed rate constant k obs was deter - mined foreachof the 1000 simulated datasets . Thestandard deviation oftheseestimatedratesallowedustodeterminetheuncertaintyinour measured rates . Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article . Data availability Data supporting the \ufb01 ndings of this manuscript are available from the corresponding author upon request . Source data are provided with this paper Source data are provided with this paper . Code availability Code used in this manuscript is available from the corresponding author upon request . References 1 . Lappalainen , P . , Kotila , T . , Jegou , A . & Romet - Lemonne , G . Bio - chemical and mechanical regulation of actin dynamics . Nat . Rev . Mol . Cell Biol . 23 , 836 \u2013 852 ( 2022 ) . 2 . Carlier , M . F . & Shekhar , S . Global treadmilling coordinates actin turnover and controls the size of actin networks . Nat . Rev . Mol . Cell Biol . 18 , 389 \u2013 401 ( 2017 ) . 3 . Breitsprecher , D . et al . Clustering of VASP actively drives proces - sive , WH2 domain - mediated actin \ufb01 lament elongation . EMBO J . 27 , 2943 \u2013 2954 ( 2008 ) . 4 . Schirenbeck , A . et al . The bundling activity of vasodilator - stimulated phosphoprotein is required for \ufb01 lopodium formation . Proc . Natl Acad . Sci . USA 103 , 7694 \u2013 7699 ( 2006 ) . 5 . Barzik , M . , McClain , L . M . , Gupton , S . L . & Gertler , F . B . Ena / VASP regulatesmDia2 - initiated \ufb01 lopodiallength , dynamics , andfunction . Mol . Biol . Cell 25 , 2604 \u2013 2619 ( 2014 ) . 6 . Yang , C . et al . Novel roles of formin mDia2 in lamellipodia and \ufb01 lopodia formation in motile cells . PLoS Biol . 5 , e317 ( 2007 ) . 7 . Courtemanche , N . Mechanisms of formin - mediated actin assembly and dynamics . Biophys . Rev . 10 , 1553 \u2013 1569 ( 2018 ) . 8 . Rotty , J . D . , Wu , C . & Bear , J . E . New insights into the regulation and cellular functions of the ARP2 / 3 complex . Nat . Rev . Mol . Cell Biol . 14 , 7 \u2013 12 ( 2013 ) . 9 . Carlier , M . F . et al . Control of polarized assembly of actin \ufb01 laments in cell motility . Cell . Mol . Life Sci . 72 , 3051 \u2013 3067 ( 2015 ) . 10 . Shekhar , S . , Pernier , J . & Carlier , M . F . Regulators of actin \ufb01 lament barbed ends at a glance . J . Cell Sci . 129 , 1085 \u2013 1091 ( 2016 ) . 11 . Pruyne , D . et al . Role of formins in actin assembly : nucleation and barbed - end association . Science 297 , 612 \u2013 615 ( 2002 ) . 12 . Sagot , I . , Rodal , A . A . , Moseley , J . , Goode , B . L . & Pellman , D . An actin nucleationmechanism mediatedbyBni1andpro \ufb01 lin . Nat . Cell Biol . 4 , 626 \u2013 631 ( 2002 ) . 13 . Romero , S . et al . Formin is a processive motor that requires pro \ufb01 lin to accelerate actin assembly and associated ATP hydrolysis . Cell 119 , 419 \u2013 429 ( 2004 ) . 14 . Kovar , D . R . , Harris , E . S . , Mahaffy , R . , Higgs , H . N . & Pollard , T . D . Control of the assembly of ATP - and ADP - actin by formins and pro \ufb01 lin . Cell 124 , 423 \u2013 435 ( 2006 ) . 15 . Isenberg , G . , Aebi , U . & Pollard , T . D . An actin - binding protein from Acanthamoeba regulates actin \ufb01 lament polymerization and inter - actions . Nature 288 , 455 \u2013 459 ( 1980 ) . 16 . Edwards , M . et al . Capping protein regulators \ufb01 ne - tune actin assembly dynamics . Nat . Rev . Mol . Cell Biol . 15 , 677 \u2013 689 ( 2014 ) . 17 . Johnston , A . B . , Collins , A . & Goode , B . L . High - speed depolymer - izationatactin \ufb01 lamentendsjointlycatalysedbyTwin \ufb01 linandSrv2 / CAP . Nat . Cell Biol . https : / / doi . org / 10 . 1038 / ncb3252 ( 2015 ) . 18 . Shekhar , S . , Hoeprich , G . J . , Gelles , J . & Goode , B . L . Twin \ufb01 lin bypasses assembly conditions and actin \ufb01 lament aging to drive barbed end depolymerization . J . Cell Biol . https : / / doi . org / 10 . 1083 / jcb . 202006022 ( 2021 ) . 19 . Goode , B . L . , Drubin , D . G . & Lappalainen , P . Regulation of the cortical actin cytoskeleton in budding yeast by twin \ufb01 lin , a ubiqui - tous actin monomer - sequestering protein . J . Cell Biol . 142 , 723 \u2013 733 ( 1998 ) . 20 . Hakala , M . et al . Twin \ufb01 lin uncaps \ufb01 lament barbed ends to promote turnover of lamellipodial actin networks . Nat . Cell Biol . 23 , 147 \u2013 159 ( 2021 ) . 21 . Avenarius , M . R . et al . Heterodimeric capping protein is required for stereocilia length and width regulation . J . Cell Biol . 216 , 3861 \u2013 3881 ( 2017 ) . 22 . Peng , A . W . , Belyantseva , I . A . , Hsu , P . D . , Friedman , T . B . & Heller , S . Twin \ufb01 lin2regulatesactin \ufb01 lamentlengthsincochlearstereocilia . J . Neurosci . 29 , 15083 \u2013 15088 ( 2009 ) . 23 . Vartiainen , M . K . , Sarkkinen , E . M . , Matilainen , T . , Salminen , M . & Lappalainen , P . Mammals have two twin \ufb01 lin isoforms whose sub - cellular localizations and tissue distributions are differentially regulated . J . Biol . Chem . 278 , 34347 \u2013 34355 ( 2003 ) . 24 . Block , J . et al . FMNL2 drives actin - based protrusion and migration downstream of Cdc42 . Curr . Biol . 22 , 1005 \u2013 1012 ( 2012 ) . 25 . Myers , K . R . , Fan , Y . , McConnell , P . , Cooper , J . A . & Zheng , J . Q . Actin capping protein regulates postsynaptic spine development through CPI - motif interactions . Front . Mol . Neurosci . 15 , 1020949 ( 2022 ) . 26 . Zigmond , S . H . et al . Formin leaky cap allows elongation in the presenceoftightcappingproteins . Curr . Biol . 13 , 1820 \u2013 1823 ( 2003 ) . 27 . Bartolini , F . , Ramalingam , N . & Gundersen , G . G . Actin - capping protein promotes microtubule stability by antagonizing the actin activity of mDia1 . Mol . Biol . Cell 23 , 4032 \u2013 4040 ( 2012 ) . 28 . Kovar , D . R . , Wu , J . Q . & Pollard , T . D . Pro \ufb01 lin - mediated competition between capping protein and formin Cdc12p during cytokinesis in \ufb01 ssion yeast . Mol . Biol . Cell 16 , 2313 \u2013 2324 ( 2005 ) . 29 . Shekhar , S . et al . Formin andcappingprotein togetherembrace the actin \ufb01 lament in a menage a trois . Nat . Commun . 6 , 8730 ( 2015 ) . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 13 30 . Bombardier , J . P . et al . Single - molecule visualization of a formin - capping protein \u2018 decision complex \u2019 at the actin \ufb01 lament barbed end . Nat . Commun . 6 , 8707 ( 2015 ) . 31 . Watanabe , N . & Mitchison , T . J . Single - molecule speckle analysis of actin \ufb01 lament turnover in lamellipodia . Science 295 , 1083 \u2013 1086 ( 2002 ) . 32 . Shekhar , S . Micro \ufb02 uidics - assistedTIRFimagingtostudysingleactin \ufb01 lamentdynamics . Curr . Protoc . CellBiol . 77 , 121311 \u2013 121324 ( 2017 ) . 33 . Jegou , A . et al . Individual actin \ufb01 laments in a micro \ufb02 uidic \ufb02 ow reveal the mechanism of ATP hydrolysis and give insight into the properties of pro \ufb01 lin . PLoS Biol . 9 , e1001161 ( 2011 ) . 34 . Cao , L . et al . Modulation of formin processivity by pro \ufb01 lin and mechanical tension . Elife https : / / doi . org / 10 . 7554 / eLife . 34176 ( 2018 ) . 35 . Pollard , T . D . & Borisy , G . G . Cellularmotilitydrivenbyassemblyand disassembly of actin \ufb01 laments . Cell 112 , 453 \u2013 465 ( 2003 ) . 36 . Fujiwara , I . , Remmert , K . , Piszczek , G . & Hammer , J . A . Capping protein regulatory cycle driven by CARMIL and V - 1 may promote actin network assembly at protruding edges . Proc . Natl Acad . Sci . USA 111 , E1970 \u2013 E1979 ( 2014 ) . 37 . Mwangangi , D . M . , Manser , E . & Robinson , R . C . The structure of the actin \ufb01 lament uncapping complex mediated by twin \ufb01 lin . Sci . Adv . https : / / doi . org / 10 . 1126 / sciadv . abd5271 ( 2021 ) . 38 . Narita , A . , Takeda , S . , Yamashita , A . & Maeda , Y . Structural basis of actin \ufb01 lament capping at the barbed - end : a cryo - electron micro - scopy study . EMBO J . 25 , 5626 \u2013 5633 ( 2006 ) . 39 . Carman , P . J . , Barrie , K . R . , Rebowski , G . & Dominguez , R . Structures of the free and capped ends of the actin \ufb01 lament . Science https : / / doi . org / 10 . 1126 / science . adg6812 ( 2023 ) . 40 . Funk , J . et al . A barbed end interference mechanism reveals how capping protein promotes nucleation in branched actin networks . Nat . Commun . 12 , 5329 ( 2021 ) . 41 . Breitsprecher , D . etal . Rocketlauncher mechanism of collaborative actin assembly de \ufb01 ned by single - molecule imaging . Science 336 , 1164 \u2013 1168 ( 2012 ) . 42 . Hakala , M . , Kalimeri , M . , Enkavi , G . , Vattulainen , I . & Lappalainen , P . Molecular mechanism for inhibition of twin \ufb01 lin by phosphoinosi - tides . J . Biol . Chem . 293 , 4818 \u2013 4829 ( 2018 ) . 43 . Palmgren , S . , Ojala , P . J . , Wear , M . A . , Cooper , J . A . & Lappalainen , P . Interactions with PIP2 , ADP - actin monomers , and capping protein regulate the activity and localization of yeast twin \ufb01 lin . J . Cell Biol . 155 , 251 \u2013 260 ( 2001 ) . 44 . Paavilainen , V . O . et al . Structural basis and evolutionary origin of actin \ufb01 lament capping by twin \ufb01 lin . Proc . Natl Acad . Sci . USA 104 , 3113 \u2013 3118 ( 2007 ) . 45 . Wioland , H . et al . ADF / Co \ufb01 lin accelerates actin dynamics by severing \ufb01 laments and promoting their depolymerization at bothends . Curr . Biol . 27 , 1956 \u2013 1967 e1957 ( 2017 ) . 46 . Fujiwara , I . , Remmert , K . & Hammer , J . A . 3rd Direct observation of the uncapping of capping protein - capped actin \ufb01 laments by CAR - MIL homology domain 3 . J . Biol . Chem . 285 , 2707 \u2013 2720 ( 2010 ) . 47 . Johnston , A . B . et al . Anovelmodeof cappingprotein - regulation by twin \ufb01 lin . Elife https : / / doi . org / 10 . 7554 / eLife . 41313 ( 2018 ) . 48 . Kotila , T . et al . Mechanism of synergistic actin \ufb01 lament pointed end depolymerization by cyclase - associated protein and co \ufb01 lin . Nat . Commun . 10 , 5320 ( 2019 ) . 49 . Shekhar , S . , Chung , J . , Kondev , J . , Gelles , J . & Goode , B . L . Synergy between Cyclase - associated protein and Co \ufb01 lin accelerates actin \ufb01 lament depolymerization by two orders of magnitude . Nat . Com - mun . 10 , 5319 ( 2019 ) . 50 . Shekhar , S . & Carlier , M . F . Enhanced depolymerization of actin \ufb01 laments by ADF / Co \ufb01 lin and monomer funneling by capping pro - tein cooperate to accelerate barbed - end growth . Curr . Biol . 27 , 1990 \u2013 1998 e1995 ( 2017 ) . 51 . Lappalainen , P . , Kessels , M . M . , Cope , M . J . & Drubin , D . G . The ADF homology ( ADF - H ) domain : a highly exploited actin - binding mod - ule . Mol . Biol . Cell 9 , 1951 \u2013 1959 ( 1998 ) . 52 . Montaville , P . et al . Spire and Formin 2 synergize and antagonize in regulating actin assembly in meiosis by a ping - pong mechanism . PLoS Biol . 12 , e1001795 ( 2014 ) . 53 . Chesarone , M . , Gould , C . J . , Moseley , J . B . & Goode , B . L . Dis - placementofforminsfromgrowingbarbedendsbybud14iscritical for actin cable architecture and function . Dev . Cell 16 , 292 \u2013 302 ( 2009 ) . 54 . Pimm , M . L . , Marcin , A . G . , Haarer , B . K . , Eligio , M . A . & Henty - Ridilla , J . L . Coordination of actin plus - end dynamics by IQGAP1 , formin , and capping protein . https : / / doi . org / 10 . 1101 / 2023 . 05 . 04 . 539490 ( 2023 ) . 55 . Hoeprich , G . J . , Sinclair , A . N . , Shekhar , S . & Goode , B . L . Single - molecule imaging of IQGAP1 regulating actin \ufb01 lament dynamics . Mol . Biol . Cell 33 , ar2 ( 2022 ) . 56 . Garabedian , M . V . et al . A septin - Hof1 scaffold at the yeast bud neck binds and organizes actin cables . Mol . Biol . Cell 31 , 1988 \u2013 2001 ( 2020 ) . 57 . Schafer , D . A . , Jennings , P . B . & Cooper , J . A . Dynamics of capping protein and actin assembly in vitro : uncapping barbed ends by polyphosphoinositides . J . Cell Biol . 135 , 169 \u2013 179 ( 1996 ) . 58 . Meacham , C . E . , Ho , E . E . , Dubrovsky , E . , Gertler , F . B . & Hemann , M . T . In vivo RNAi screening identi \ufb01 es regulators of actin dynamics as key determinants of lymphoma progression . Nat . Genet . 41 , 1133 \u2013 1137 ( 2009 ) . 59 . Bockhorn , J . et al . MicroRNA - 30c targets cytoskeleton genes involved in breast cancer cell invasion . Breast Cancer Res . Treat . 137 , 373 \u2013 382 ( 2013 ) . 60 . Labat - de - Hoz , L . & Alonso , M . A . Formins in human disease . Cells https : / / doi . org / 10 . 3390 / cells10102554 ( 2021 ) . 61 . Li , Q . etal . AttenuationofmicroRNA - 1derepressesthecytoskeleton regulatory protein twin \ufb01 lin - 1 to provokecardiachypertrophy . J . Cell Sci . 123 , 2444 \u2013 2452 ( 2010 ) . 62 . Schindelin , J . et al . Fiji : an open - source platform for biological - image analysis . Nat . Methods 9 , 676 \u2013 682 ( 2012 ) . Acknowledgements S . S . is grateful to Bruce Goode for many years of actin - related discussions and for his mentorship and teaching , not only in doing research but also in managing a lab . S . S . thanks Marie - France Carlier for introducing him to actin and Jeff Gelles for teaching him single - molecule imaging . S . S . also thanks his graduate advi - sors Vinod Subramaniam and Hans Kanger for their mentorship . We thank Jim bear for critical reading of the manuscript and feedback . We thank Pekka Lappalainen for generously sharing twin \ufb01 lin plas - mids and Jan Faix for sharing VASP chimera plasmids . We thank Ankita and Sandeep Choubey for their help with statistical analysis . We thank Sandeep Choubey , Ekram Towsif , and Surbhi Garg for their comments on the manuscript . We thank Matt Winfree , Thomas Pitta and Lauryn Luderman for their tireless support towards our microscopy efforts . This work was supported by NIH NIGMS grant R35GM143050 to S . S . and startup funds from Emory University to S . S . Author contributions H . U . and I . G . conducted experiments and analyzed data . H . U . and S . S . prepared \ufb01 gures . S . S . designedexperimentsandsupervisedtheproject . S . S . and H . U . wrote the \ufb01 rst draft of the manuscript and all authors contributed to the editing . S . S . acquired funding . Competing interests The authors declare no competing interests . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 14 Additional information Supplementary information The online version contains supplementary material available at https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 . Correspondence and requests for materials should be addressed to Shashank Shekhar . Peer review information Nature Communications thanks Thomas Pol - lard , and the other , anonymous , reviewer ( s ) for their contribution to the peer review of this work . A peer review \ufb01 le is available . Reprints and permissions information is available at http : / / www . nature . com / reprints Publisher \u2019 s note Springer Nature remains neutral with regard to jur - isdictional claims in published maps and institutional af \ufb01 liations . Open Access This article is licensed under a Creative Commons Attribution 4 . 0 International License , which permits use , sharing , adaptation , distribution and reproduction in any medium or format , as long as you give appropriate credit to the original author ( s ) and the source , provide a link to the Creative Commons license , and indicate if changes were made . The images or other third party material in this article are included in the article \u2019 s Creative Commons license , unless indicated otherwise in a credit line to the material . If material is not included in the article \u2019 s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use , you will need to obtain permission directly from the copyright holder . To view a copy of this license , visit http : / / creativecommons . org / licenses / by / 4 . 0 / . \u00a9 The Author ( s ) 2023 Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 39655 - 3 Nature Communications | ( 2023 ) 14 : 3981 15",
    "yuEncouragingOutsideBox2016": "Encouraging \u201cOutside-\u00adthe-\u00adbox\u201d Thinking in Crowd Innovation Through Identifying Domains of Expertise Lixiu Yu , Aniket Kittur , Robert E . Kraut Carnegie Mellon University 5000 Forbes Avenue , Pittsburgh , PA 15213 { lixiuyu , nkittur , robert . kraut } @ cs . cmu . edu ABSTRACT People are more creative at solving difficult design problems when they use relevant examples from outside of the problem\u2019s domain as inspirations . However , finding such \u201coutside - the - box\u201d inspirations is difficult , particularly in large idea repositories such as the web , because without guidance people select domains to search based on surface similarity to the problem\u2019s domain . In this paper , we demonstrate an approach in which non - experts identify domains that have the potential to yield useful and non - obvious inspirations for solutions . We report an empirical study demonstrating how crowds can generate domains of expertise and that showing people an abstract representation rather than the original problem helps them identify more distant domains . Crowd workers drawing inspirations from the distant domains produced more creative solutions to the original problem than did those who sought inspiration on their own , or drew inspiration from domains closer to or not sharing structural correspondence with the original problem . Author Keywords Crowdsourcing ; problem solving ; design ; idea generation ACM Classification Keywords H . 5 . 3 Group and Organization Interfaces INTRODUCTION Important innovations and discoveries often come from drawing upon knowledge from domains outside of a target problem as an inspiration to solve the problem [ 8 , 10 ] . For example , using origami - folding techniques , NASA ' s space engineers designed a space array , which can be folded compactly and then deployed while in outer space . The new design solved the 50 - year - old space problem of transporting large objects in narrow rockets [ 2 ] . The emergence of large online idea repositories has the potential to radically accelerate innovation by increasing designers\u2019 access to analogical ideas . There now exists an enormous selection of ideas that could spark creative ideas through analogy . For example , InnoCentive ( innocentive . com ) contains more than 40 , 000 business , social , policy , scientific , and technical problems and solutions in various domains , hundreds of new product ideas are submitted to Quirky ( quirky . com ) every day by a pool of over a million inventors , and OpenIDEO has collected hundreds of solutions for a variety of social problems since 2010 ( openideo . com ) . More generally , information available online including scientific literature , patents , webpages , and images and videos represent a treasure trove of potential analogies in different domains . For example , a car mechanic adapted a method for extracting a lost cork from a wine bottle seen in a YouTube video to save a baby stuck in the birth canal , described as the most important innovation in birthing since forceps [ 17 , 22 ] . However , our ability to process this deluge of information to find and use analogies is severely bottlenecked by individual cognitive limits , as the speed and capacity with which individuals can learn and explore new domains have not kept up with the rapid growth in online information from which analogies can be discovered . Even when they have the appropriate knowledge , people often become fixated on surface - level details that prevent them from retrieving useful analogs from memory or external repositories or applying them for problem solving [ 9 ] . Previous work has shown that using crowds to mine online idea repositories for such analogous inspirations can enable search at a scale beyond the individual and has identified ways of reducing fixation by having different workers see different representations of the problem [ 24 , 25 ] . However , this work has looked at relatively small repositories of ideas , such as Quirky\u2019s , where searchers were only looking through hundreds of ideas . In contrast , the web holds orders of magnitude more potentially useful analogous ideas , not only in explicit idea repositories , such as the US patent Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page . Copyrights for components of this work owned by others than ACM must be honored . Abstracting with credit is permitted . To copy otherwise , or republish , to post on servers or to redistribute to lists , requires prior specific permission and / or a fee . Request permissions from Permissions @ acm . org . CSCW ' 16 , February 27 - March 02 , 2016 , San Francisco , CA , USA \u00a9 2016 ACM . ISBN 978 - 1 - 4503 - 3592 - 8 / 16 / 02 \u2026 $ 15 . 00 DOI : http : / / dx . doi . org / 10 . 1145 / 2818048 . 2820025 1214 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA database , but also in expert - generated content in nearly any domain . In this paper , we describe and evaluate a two - stage process that enables crowds to search the web for useful and novel do design ideas than alternative approaches . In the first stage , crowds identify domains of expertise remote from the initial problem but relevant enough to provide ideas to inspire useful and non - obvious solutions . They are best able to identify relevant but remote domains when the original concrete problem is represented as an abstract problem schema . In the second stage , crowds search in these domains to find inspirational examples they can adapt to solve the original problem . The key insight here is that a rich set of expert - generated ideas , solutions , and skills have already been documented on the web , and non - experts can find and appropriate these resources in the design process even if they did not possess the expertise to generate this knowledge in the first place . Identifying \u201cOutside-\u00adthe-\u00adbox\u201d Domains Although distant analogies can inspire innovation , systematically identifying them in distant domains is challenging . First , people must identify the domains where relevant inspirational sources might be found . The most useful inspirations often come from distant domains that have little surface similarity with the target problem [ 11 , 18 , 21 ] ; for example , in the NASA example , the problem comes from the aerospace field while paper folding is from origami art . Identifying such domains can be challenging for human , because they often become fixated on the surface details of the problem domain . Outside - the - box domains are also challenging for automated search algorithms to identify , since the algorithms also generally search on surface features , leading to same - domain recommendations [ 3 , 16 , 20 , 26 ] . Yu et al . [ 24 ] demonstrated a way for crowds to find distant domains more effectively by re - representing the problem . A critical step was abstracting surface details of the original problem to reduce people\u2019s fixation on surface features , such as domain - specific vocabulary and objects [ 7 , 15 ] . For example , in the previous NASA story , space and rockets are surface features while the abstract representation of the problem can be \u201chow to contain more content within a limited space\u201d . Using this abstract representation rather than the literal description of the problem frees people to identify remote areas of knowledge containing useful inspirations for the problem . Yu et al . then showed that crowd workers given the abstract representation of a problem ( i . e . , the schema ) rather than the concrete representation were able to find analogies in more distant domains that were more useful for solving the problem creatively . However , Yu et al . treated the search process as a black box , with no instructions on how to identify potentially fruitful domains to search within . This approach may have been successful on the tens to hundreds of ideas easily available on Quirky . com , but may not scale well to the billions of web pages available on the Internet , a subset of which could have a potentially relevant solution to the target problem . The difficulty of winnowing distant domains for useful analogs has posed a crucial challenge to many design - by - analogy approaches , causing \u201cthis influential technique to be limited to little more than interesting examples with accompanying direction to simply \u2018try to find analogies . \u2019\u201d [ 14 ] . Exploring \u201cOutside-\u00adthe-\u00adbox\u201d Domains A second problem is finding useful ideas within the identified domain . An expensive but powerful way to do so would be to have a panel of relevant experts on call from that domain . Indeed major design firms like IDEO are successful in part because they hire employees who are diverse in demographics , education and personal interests and can apply ideas from their areas of expertise to the problems clients bring them [ 10 ] . Although retaining experts may be necessary in research - investment intensive industries , like pharmaceuticals , for many more routine problem challenges in engineering and design , experts may have already documents relevant solutions in language accessible to laymen . Crowdsourcing approaches have been increasingly used to expand the pool that innovation companies can draw upon to solve problems , ranging from product ideation ( e . g . , Quirky . com ) to research and development ( e . g . , Innocentive . com ) to societal challenges ( OpenIDEO . com ) Companies describe their problems and then use what is called \u201cbroadcast search\u201d to invite anyone with a good idea to contribute solutions [ 6 , 13 ] . Here instead of contributing original ideas , we aim to use non - expert crowd workers to mine the web for ideas in various domains that might help solve the target problem . Knowledge , methods , principles , skills and tools that experts use in their domains have often been amply documented on the Internet . For example , the origami folding techniques used in the previously mentioned NASA example can be found through a search for \u201corigami patterns\u201d . Thus , problem solvers who might not possess the desired expertise themselves might profitably search for outside - the - box knowledge online . EXPERIMENT 1 : IDENTIFYING OUTSIDE-\u00adTHE-\u00adBOX DOMAINS Our overall goal in this research is to elicit distant domains and useful inspirations from a large , open - ended idea repository such as the web . In particular , we aim to go beyond previous design - by - analogy work in providing a structured process for identifying fruitful domains for exploration , rather than simply asking individuals to find useful examples directly . We hypothesize that doing so requires generating and exploring domains that are distant from the original problem domain , yet contain structurally similar and relevant inspirations . In Experiment 1 we 1215 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA explore a process for generating distant structurally similar domains , and in Experiment 2 we test whether the found domains yield relevant and useful inspirations . In a pilot study we asked workers directly to identify domains that might be useful for solving the schematic representation of the problem . However , the resulting domains were generally too vague to be useful ( e . g . , \u201cengineering\u201d ) . To address this problem we turned to a mechanism that people use in real life when they are looking for knowledge in a different domain : they look for referrals . The intuition is that even if people might not have the desired knowledge a problem needs and are not accustomed to thinking about \u201cdomains\u201d of knowledge , they may be able to identify a type of expert in a different domain who might deal with relevant problems . We thus reframed our elicitation from crowd workers , asking them to identify professions that could have useful perspectives . For example , if the goal were to design a new power strip , one might ask about professions that deal with problems of packing things into a limited space . Such professions could be as diverse as landscapers packing plants into a small yard ; warehouse loaders packing products into crates ; user interface designers packing information into interfaces ; or even contortionists packing their bodies into small containers . Some of these domains might yield interesting solutions , such as a terraced landscaping solution inspiring the design of a power strip with different height levels to avoid plugs obstructing each other . The above example also highlights the importance of abstraction in the domain elicitation process . Providing an abstract representation of a problem can increase the diversity of the resulting domains found [ 25 ] . For example , the abstract representation of the power strip\u2019s problem for \u201cpacking things into a limited space\u201d might elicit more diverse professions than a concrete representation such as \u201cfitting different sized plugs into a power strip\u201d . In Experiment 1 we manipulated the level of abstraction of the problem representation ( i . e . , original , concrete problem or abstracted problem ) and asked people to identify professions that might have interesting perspectives on the problem . Our goal was to find distant yet structurally relevant domains that might yield useful inspirations . Participants One hundred and twenty - two Amazon Mechanical Turk workers [ 12 ] participated in this experiment . Forty percent were women , and 93 % were native English speakers . Their average age was 33 and ranged from 19 to 68 . Design and Procedure We selected two design challenges from a crowd innovation website ( Quirky . com ) to use in the experiment ( see Table 1 ) . For each challenge we generated an abstract schematic representation by first abstracting the goal and the sub - goals , and replacing concrete objects with generic objects sharing essential attributes . For example , the original power strip problem listed on the website talks about plugs and outlets , while the schematic version talks about objects fitting into a container or blocking each other . While in this research the experimenters created the schematic problem representation for convenience and standardizing the input to the experiment , previous research shows schematization of a concrete problem can be done by lay crowd workers following instructions [ 25 ] . After accepting the task , participants were randomly assigned to either the original or schematic representation of the power strip or cup problem and asked to recommend types of experts for a design problem . Specifically , they were asked , \u201c Please read the design problem below and suggest three types of experts who might provide useful or interesting perspectives in solving it and explain why \u201d . Rating the Recommended Experts We hypothesized that the schematic representation would return a more diverse set of experts . Table 2 shows examples of the recommended experts for the two problems in the two conditions . Some professions appeared in both problems ( e . g . , carpenter , construction worker ) but for different reasons . For example , a rationale for a carpenter in the cup problem was that they could \u201c design a rack that separates the cups just enough so they can dry properly , but not take up too much space \u201d while for the power strip problem they \u201c Could design hidden compartments to run cords to other outlets that would also keep the cords hidden . \u201d Original description Schematic representation ( Power strip problem ) Have a look at the power strip under your desk . How many of its outlets are being used ? How many of them would you like to use , but you can ' t , because a giant power brick ( transformer ) in the adjacent outlet is blocking it ? How could you fit all the different plugs in all the outlets ? 1 . How can you fit objects of different sizes into a container ? ( goal ) 2 . Prevent objects blocking each other ( sub - goal ) 3 . Fully make use of a container\u2019s capacity ( sub - goal ) ( Cup problem ) When we finish washing cups and glasses , we have to either spread them out individually , but then they take up all the counter space . Alternatively , we can stack them , but then the cups never dry completely and it is hard to separate them from each other later . How can you dry many cups quickly so that they don\u2019t take up too much space and moisture doesn\u2019t get trapped between them ? 1 . How can you dry multiple stackable objects ? ( goal ) 2 . Prevent multiple similar objects from taking up much space ( sub - goal ) 3 . Separate multiple stackable objects easily ( sub - goal ) Table 1 . Original and schematic representations for the two design problems ( power strip , cup ) used in the experiments . 1216 SESSION : CROWD INNOVATION AND CROWDFUNDING In the original concrete formulation of the cup problem , most recommendations appeared to invoke people who would have direct knowledge about kitchens and cups \u2013 e . g . , housewives , chefs , bartenders , cup designers , and interior designers . Similarly , many participants who saw the original description of the power strip problem recommended electricians and electronic engineers . In contrast , the schematic representations of the cup problem appeared to return a wider variety of experts , including bakers , forklift operators , and sculptors . Participants who saw the power strip problem\u2019s schematic representation recommended a variety of experts who work with problems related to sizes , shapes , and blocking . For example , in the schematic representation of the power strip problem , a participant recommended a topology expert and explained , \u201c This person would know about the mathematical study of shapes and shapes in space which would help with deciding about objects of different sizes . \u201d To analyze whether the conditions differed in generating domains two judges rated each recommendation on two dimensions : whether the expert was unique among all recommendations for the same problem and how distant the expert\u2019s domain was from the original problem domain . To count unique experts , we combined similar recommendations into single one . For example , \u201cmover\u201d and \u201ca moving expert - moving peoples\u2019 possessions from one home to another\u201d were both classified as \u201cmover\u201d . To judge distance , two judges blind to experimental condition rated each recommended expert on a 7 - point Likert scale : \u201c How different is the above power strip problem from the problems that the recommended expert works with in his or her domain ? \u201d By this metric , for example , a \u201cmail sorter\u201d was considered further from the cup problem than was a dishwasher . The judges achieved a good inter - rater reliability of 0 . 78 , calculated using the Intraclass Correlation Coefficient ( ICC ) [ 5 ] . The final distance score was calculated by averaging the scores of the two judges . Analysis and Results Table 3 shows the means and standard errors of the percentages of unique experts and the distance scores . A proportion test showed participants identified a larger proportion of unique experts when shown schematic descriptions of the problem than when shown the more concrete , original problem descriptions ( z = 3 . 21 , p < 0 . 001 ) . Participants shown the schematic problem descriptions also recommended domains more distant from problem domain than did those shown the original problem description ( z = - 5 . 89 , p < 0 . 001 , by a Mann - Whitney U test used because of the skewed distribution ) . EXPERIMENT 2 : GENERATING IDEAS Although Experiment 1 suggests that that converting the original problem description into a schematic representation and eliciting domains in the form of professions led people to identify more distant structurally related domains . However , it fails to answer the key question of whether these distant domains yield more useful inspirations that lead to more creative solutions that the ones found in domains closer to the original problem domain . While more distant domains have the potential to yield inspirations that break fixation with the original problem and suggest interesting approaches from other domains , it is also possible that these domains might not be sufficiently relevant to the target problem to yield useful inspirations . For example , in Table 2 the schematic representation resulted in professions that at first glance might not seem relevant , such as \u201cmagician\u201d for the power strip problem or \u201cparking lot attendant\u201d for the cups problem . However , on closer inspection workers provided rationales that make these potentially plausible sources , such as \u201c Can make objects appear to do many things you might think illogical \u201d for magician , perhaps suggesting that some trick mechanics might help fit things in places you wouldn\u2019t expect ; or Cup problem Power strip problem Original Schematic Original Schematic Chef , Housewife , Bartender , Carpenter , Counter - designer , Cup designer , Dishwasher , Gardener , Glass expert , Home - builder , Maid , HVAC - technician , Interior - designer , Professional - organizer , Rack - specialist , Waiter Baker , Bodybuilder , Carpenter , Cashier , Construction \u2013 worker , Designer , Fireman , Forklift - operator , Landscaper , Mail sorter , Mathematician specializing in geometry , Meteorologist , Parking - lot - attendant , Pianist , Potter , Sculptor Electrician , Building - contractor , Computer - technician , Housewife , Network - engineer , Interior - designer , Artificial - intelligence - scientist , Building - maintainer , Cable installer , Carpenter , Construction - worker , Electrical - engineer Contortionist , Geometry - expert , Graphic - designer , Landscaper , Magician , Physicist , Sculptor , User - interface - designer , Architect , Warehouse - dock - loader , Artist , Expert - of - arithmetic , Expert - on - Japanese - aesthetics , Expert - on - topology Table 2 . Examples of the recommended expert domains from Experiment 1 in the Original and Schematic problem representation conditions . Condition Freq . Percent of unique experts Distance Mean S . E . Mean S . E Original 204 0 . 41 0 . 03 3 . 87 0 . 23 Schematic 162 0 . 60 0 . 04 5 . 74 0 . 16 Note on frequency : There were 68 participants in Original representation and 54 participants in Schematic representation . Each participant provided three recommendations . Table 3 . Experiment 1 : Means and standard errors . 1217 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA \u201c \u2026needs to be wise about how they position cars so that they can get in and out whenever needed\u2026would know the best way to position objects \u201d for the parking lot attendant . In this experiment we explore whether distant domains would yield useful inspirations by having participants generate new ideas using the inspirations found in the previous experiment . We are interested in whether participants can find more useful and interesting inspirational sources from an outside - the - box domain and solve the problem better than when searching on their own or when searching in a domain that shared surface similarities to the problem domain . Our hypothesis is that the benefits of far domains occur only when they are likely to contain ideas that have deep structural correspondences to the original problem . Thus , one purpose of this experiment is to insure that distance from the target domain by itself is not what leads to better problem solutions . The experiment involved four conditions . In the Problem - driven condition , participants searched for inspiration in a domain recommended from the original problem description . In the Schema - driven condition , they searched for inspiration in a domain recommended from the schematic problem representation . The Irrelevant schema - driven control condition tested whether far domains that are not structurally similar to the original problem have the same benefits as far domains that were structurally similar . Specifically , participants searched for inspirations for the cup problem in a domain recommended from the schematic problem representation of the power strip problem and vice versa . Finally , we also included a control condition where participants were asked to search for inspiration on their own , called the Self - selected condition . In all these conditions , participants first searched for inspirations and then tried to solve the original problem using the inspirations they found . Participants Overall 130 Amazon Mechanical Turk workers participated in Experiment 2 . Forty - six percent were women , and 96 % were native English speakers . Their average age was 34 and ranged from 18 to 66 . Participants from Experiment 1 were excluded from participating in Experiment 2 . Design and Procedure After participants accepted the task , they were randomly assigned to one of the four experimental conditions . They were asked to search for two inspirational examples for either the Cup or Power Strip problem and then generate a new product idea that solved the problem . They were either given no hint about where to search for inspiration ( Self - selected ) or were told to look for their inspirations from domains recommended in Experiment 1 from participants who saw the concrete version of an original problem ( Problem - driven ) , the schematized version of the original problem ( Schema - driven ) or a schematized version of an irrelevant problem ( Irrelevant schema - driven ) . Instructions for the conditions are described in more detail below : 1 . Self - selected . Participants saw either the Cup or Power strip problem as described on the left column in Table 1 and then asked to search for useful ideas on the web that would help them design a solution to the problem . Specifically , they were told , \u201c Please go to the Internet and find two useful ideas that could inspire good solutions for the above problem . These ideas could be knowledge , skills or methods other people use to solve a similar problem in their own domain . Please don ' t search for ideas related to this cup [ power strip ] problem . The useful ideas have to be about similar problems in a non - cup related domain . \u201d After they found two useful ideas , they were asked to solve the problem . Specifically they were told , \u201c Please generate a new product idea that solves the above design problem using the ideas you found . \u201d 2 . Problem - driven . After seeing one of the two design problems , participants were asked to search for useful ideas that an expert in the problem domain might have . For example , if they were assigned the housewife domain , their prompt would be , \u201c Please go to the Internet and find two useful ideas a housewife might have that could inspire good Inspiration Expert domains and solutions An architect : Cup with studs on sides , so that they can be stacked while allowing airflow to dry easily . A carpenter : Maybe instead of a linear power strip it could be built as an arc or circle that allows more plugs to fit in basically the same amount of space . By turning a line into a circle it opens up more room around each plug or AC adaptor . An expert on Japanese aesthetics : If you made an outlet that looks like a Miyabi style building . You could go up levels and it would leave more room for the plugs to go around the outlet . Table 5 . Examples of inspirations and resulting solutions from participants . Independent variable % creative Odds Ratio se Z Self - selected 0 . 26 5 . 44 * * 3 . 10 2 . 98 Problem - driven 0 . 12 13 . 53 * * * 8 . 91 3 . 95 Schema - driven 0 . 65 NA NA NA Irrelevant schema 0 . 25 5 . 45 * * 3 . 08 3 . 01 p < . 01 = * * p < . 001 = * * * Table 4 . The effects of inspiration condition on design quality . 1218 SESSION : CROWD INNOVATION AND CROWDFUNDING solutions for the above problem . These ideas could be knowledge , skills , or methods a housewife uses to solve a similar problem in his or her own domain . Please don ' t search for ideas related to this cup problem . The useful ideas have to be about problems in the domain that a housewife deals with . \u201d An expert was randomly selected an expert without replacement from the domains recommended in the Original representation condition in Experiment 1 , with a probability proportional to the number of times the expert was recommended in Experiment 1 . After finding two useful ideas , participants were asked to solve the problem using those ideas as inspiration . 3 . Schema - driven . This condition is identical to the Problem - driven condition except that participants were asked to find inspirations from domains identified in the Schematic representation condition in Experiment 1 . After finding two inspirations , they then solved the problem . 4 . Irrelevant schema - driven . This condition was identical to the Schema - driven condition except that participants were asked to find inspirations from the domains identified in the Schematic representation condition in Experiment 1 for an irrelevant problem . That is , if participants were to solve the cups problem , they were shown domains recommended for the power strip problem and vice versa . After finding two inspirations , they then solved the problem . There were roughly 90 recommendations identified in each condition in Experiment 1 . When selecting domains to give to participants , we coalesced domains to give related ones a common name and excluded generic recommendations such as \u201cscientist\u201d and \u201cengineer\u201d . Quality of the Design Solutions Judging the creative quality of an idea can be difficult problem involving significant subjectivity . To judge the ideas we draw on previous research establishing methods for robustly rating creative idea quality , which considers an idea as being creative if it is both novel and practical [ 4 ] . Novelty was defined as an idea that was not obvious and differed from existing products on the market . Practicality was defined as how realistically an idea achieved its goal and could be designed and manufactured . Two judges blind to experimental condition iterated on a rubric for rating product ideas , then used this rubric to independently rate each idea on two 7 - point Likert scales measuring novelty and practicality . The judges achieved good ICC inter - rater reliabilities of 0 . 79 and 0 . 61 for novelty and practicality respectively . The final creativity score of each idea was computed as the mean of its novelty and practicality scores . Some limitations of this approach are discussed in the discussion section . Below is an example of a solution to the power strip problem rated highly in terms of both novelty ( 6 . 0 ) and practicality ( 6 . 0 ) . Make a power strip that has some outlets raised about the others . That way a large plug can fit into the raised outlet and still leave room for another plug in the outlet that isn ' t raised . In contrast , the design idea below was rated poorly on both dimensions of novelty ( 1 . 0 ) and practicality ( 2 . 5 ) . I would propose a specially made power strip that is longer , and wider than any power strip on the market . Would that has enough space between each socket to fit any plug , or power brick comfortably . Of course it would be bigger , but that is a fair trade off . Following recommendations from [ 4 ] , we then classified each design as creative if it was above a media - split threshold of 3 . 0 for both novelty and practicality or non - creative otherwise . Analysis and Results To determine whether the source of the domains participants used as inspiration influenced the quality of their designs , we conducted a logistic regression model predicting with whether an idea was creative ( 1 ) or not ( 0 ) from the experimental conditions , along with level of education and whether the participants\u2019 language was English as controls . Table 4 summarizes the results , with Schema - driven domain as the reference group . Results indicate that searching for inspiration in domains far from the original problem significantly increased the probability of generating a creative solution compared to searching in Problem - driven domains , in Irrelevant schema - driven domains or when given no direction about which domains to search . The odds of producing a creative idea when given recommendations to search in a Schema - driven domain were 13 . 53 times the odds when given a Problem - driven domain , 5 . 45 times the odds when given an Irrelevant schema - driven domain and 5 . 44 times of the odds when participants searched for inspiration with guidance about domain ( Self - selected ) . To further examine whether the domain of search is responsible for the difference in the quality of ideas , we added the domain distance score ( judged in Experiment 1 ) as a mediator and re - run the logistic regression analysis . The results show that domain distance is significantly correlated with the odds of producing a creative idea : for every one - unit increase in the distance score , the odds of producing a creative idea increases by 2 . 07 times ( p < 0 . 01 ) . After adding domain distance as a mediator , the quality difference between the problem - driven condition and the schema - driven condition became non - signi \ufb01 cant ( b = - 0 . 96 , p = . 23 ) . However , the difference of idea quality still exists between the schema - driven condition and the irrelevant schema - driven condition : even holding domain distance constant , the odds of producing a creative idea when given recommendations to search in a schema - driven 1219 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA domain were 3 . 22 times of the odds when given an irrelevant schema - driven domain ( p < 0 . 05 ) . The mediation analysis suggests that the distance of inspiration from the target domain created by the abstraction of the problem schemas completely explains the higher idea quality in the schema - driven condition compared to the problem - driven condition . That is , compared to the problem - driven condition , people in the scheme - driven condition used examples from distant domains as inspiration , which in terms led them to generate creative ideas . However , the distance of inspiration from the target domain only partially explains why solutions inspired by schema - driven examples were more creative than ones inspired by irrelevant schema - driven ones . This failure of mediation shows that the benefits of far domains occur only when they are likely to contain ideas that have deep structural similarities to the original problem . Search Mechanisms In Experiment 2 , we found that people were able to find relevant inspirational examples for a problem in remote domains and were more likely to generate creative ideas by using the inspirations that was found from remote domains . These results , however , beg the question of how non - expert crowd workers were able to find useful ideas in unfamiliar domains . People found and used inspiration in the schema - driven domains using a variety of mechanisms . To illustrate , Table 5 shows the inspirations people found by searching in the domain of an architect , a carpenter , and an expert on Japanese aesthetics . Searching in an architect\u2019s domain , a participant found and adopted the energy - saving ventilation design of a building for the cup problem ; searching in a carpenter\u2019s domain , a participant was inspired by the circle - shape of a drawer ; and searching in the domain of Japanese aesthetics , a participant borrowed a design of the Japanese Miyabi . Other participants also utilized the mechanisms and principles of other domains . For example , a participant designed a power strip that has some outlets raised about the others by utilizing how warehouse dock loading systems work . Another participant designed a bendable power strip by searching in a contortionist\u2019s domain . These forms of inspiration are much different from those found through domains recommended based on the original problem description or when participants searched for inspirations on their own . In these conditions , people found ideas related to organizing clutter , drying dishware , or rack design . When searching based on irrelevant domains , people often found information that appeared random and their solutions were not well connected to the inspirations they found . The findings support the assumption that knowledge encoded on the web in various expert domains is useful in sparking creative solutions for a related problem in a different domain . However , the outside - the - box knowledge had to be structurally alignable and relevant to the problem as well . The distance and relevance were created through the schematic representation in Experiment 1 : the abstraction reduces the domain fixation leading to a distance between the original problem and the recommended domains , while goals and sub - goals keep the connection between them . The expert domains in Experiment 2 work as a mediation step to bridge the schematic representation and the useful ideas in the domains . DISCUSSION Previous research has shown that creative solutions to problems often emerge when experts from outside a problem domain apply their \u201cforeign\u201d skills , techniques and tools to the problem . This insight inspired our research in developing a systematic process to leverage crowd workers to harvest potentially applicable ideas from fields outside of a problem\u2019s domain . To do so , we had to get the workers to \u201cthink outside of the box\u201d and overcome the challenge of functional fixedness [ 1 ] , the cognitive bias that limits people to using objects only for their traditional purposes . We addressed this challenge by transforming the original , concrete problem statement to a more abstract representation . When combined with an instruction set framed in terms of \u201creferrals\u201d , this schematic representation of the problem allowed people to identify a more diverse set of domains to explore for potential solutions than did the original , concrete problem description . Moreover , when non - expert crowd workers searched for inspirations for solutions in domains primed by the abstract representations , they returned a rich set of relevant examples . Their examples were often very different from the original problem in terms of surface features , but shared structural features . For example , at the surface level a Miyabi style building has little in common with an electrical power strip , but its three dimensional structure and arrangement of rooms might suggest ways to fit different plugs in a power strip . Crowd workers who collected inspirations from these remote domains produced more creative problem solutions than did those who searched for inspiration without guidance about where to look or who looked for inspiration in domains primed by the original concrete problem statement or by an abstract representation of a problem that did not share structural correspondence with the target problem . We asked Amazon Mechanical Turk workers , who were not selected for engineering or design expertise , to solve relatively simple but real design challenges . The participants provided brief text descriptions of the approach they would take to a problem , such as building a power strip in two dimension ( a curve ) or three dimension ( a tower ) ; they were not required to provide details about implementation . Despite these limitations , the Cup and Power Strip problems were authentic design challenges taken from Quirky . com , the crowd innovation company 1220 SESSION : CROWD INNOVATION AND CROWDFUNDING specializing in consumer products . In the past , members of the Quirky community designed detailed solutions for these challenges , and Quirky itself manufactured and sold consumer products to solve these problems . Interestingly , some of the solutions provided by crowd workers , such as a bendable power strip inspired by the contortionist domain , paralleled those from the Quirky community itself , such as the bendable PivotPower ( one of Quirky\u2019s most popular products ) . Similarity , the architecture - inspired method for attaching studs to cups so that air could circulate between stacked cups to dry them is highly reminiscent of Totem Air Dry ( another Quirky product ) . Workers did not indicate being influenced by Quirky products and none returned a Quirky product as an inspiration . When they submitted design ideas that paralleled Quirky products , their designs had clear links to examples retrieved from their assigned source domains , although we cannot guarantee that participants had never seen the relevant Quirky products . However , assuming no direct influence , parallels between crowd workers\u2019 solutions and those manufactured by Quirky suggest that the design ideas participants came up with in this controlled experiment could be commercially valuable in the world . We consider our research , which challenged relatively unsophisticated workers to solve relatively simple design problems , only a first step . One area for profitable future research is investigating whether this process will be effective in crowd innovation sites like InnoCentive , TopCoder , or MathOverflow , which typically challenge sophisticated experts to solve complicated R & D problems . We believe the process of searching for inspirations outside of a problem\u2019s domain should be useful to overcome functional fixedness for sophisticated problems as well as simple ones and may actually be more useful for problems that require deep expertise . Abstracting away some of the non - essential features of the problem should help in identifying new domains in which to search for solutions . However , when the problem is complex or when the solution must be fully elaborated or reduced to practice , one may need to recruit true experts from the distant knowledge domains and not just novices reading what experts had previously written . Another limitation in the current research is that one of the two judges in Experiment 2 was the first author . While both were blind to experimental condition and used rubrics to reduce bias and to standardize their ratings , future experiments could benefit from separate judges to reduce any biases resulting from trying to infer which experimental condition each idea came . Future research could also examine the robustness of using non - designers as judges in this context , though prior work shows high agreement between designers and non - designers in judging consumer products [ e . g . , 23 ] . The approach of using schemas as described here could also benefit from further research into the boundary conditions for where schemas are useful . For example , it is not yet clear whether schemas would be as useful if the design process continued past ideation into the prototyping stage . Prototyping may force people to think more deeply about how to adapt specific mechanisms of inspirational examples to a solution , which may result in poor fits . For example , people may run into configurational or material use problem with designing a Japanese Miyabi - style power strip , which could limit the practicality of manufacturing it . Representing Problems in Crowd Innovation A core contribution of this work is that we use a problem\u2019s abstract representation as a cue to identify domains where a solution might be found , instead of directly searching for analogical ideas [ 25 ] . This opens up the search space from well - defined idea repositories ( as in [ 25 ] ) to the entire web . Starting a search for solutions in experts\u2019 domains can provide more concrete guidance about the type of knowledge needed to solve a problem than more abstract schematic representations of the problem , which can be ambiguous and difficult to interpret . Presenting different domains to different problem solvers opens up the search space and helps solves find more varied inspirations even from the same schematic representation . We argue that problem representation is an important research topic with the rise of crowd innovation . This new problem - solving model opens a rich research area on how to formulate problems so that a distributed group of people can solve them creatively . The schematic representation proposed in this paper is one approach constructing an alternative representation for a problem in order to encourage outside - the - box search for solutions . However , how to identify the most productive level of abstraction for a particular design challenge is still an open question . An interesting area for future study is whether more complex problems on sites such as InnoCentive could be re - represented and classified based on abstract goals and sub - problems , and thereby attract more actual outside - the - box experts to solve them . ACKOWLEDGEMENTS This work was supported by NSF grants IIS - 1526665 , IIS - 1149797 , IIS - 1217559 , OCI - 0943148 , IIS - 0968484 , IIS - 1111124 , Bosch , Google , and Microsoft , Heinz College Center for the Future of Work , and Carnegie Mellon\u2019s Center for the Future of Work . We thank reviewers for useful feedback . REFERENCES 1 . Robert E . Adamson . 1952 . Functional fixedness as related to problem solving : A repetition of three experiments . Journal of experimental psychology 44 , 4 : 288 . 2 . Meg Monk . 2013 . The Digital Universe . http : / / universe . byu . edu / 2013 / 12 / 12 / byu - engineers - use - origami - to - make - more - space - in - space / 1221 CSCW ' 16 , FEBRUARY 27 \u2013 MARCH2 , 2016 , SAN FRANCISCO , CA , USA 3 . Maryam Fazel - Zarandi and Eric Yu . 2008 . Ontology - based expertise finding . In Practical Aspects of Knowledge Management . Springer , 232 \u2013 243 . 4 . Ronald A . Finke , Thomas B . Ward , and Steven M . Smith . 1992 . Creative cognition : Theory , research , and applications . Bradford Book . 5 . Ronald Aylmer Fisher . 1954 . Statistical methods for research workers . Oliver and Boyd . 6 . Karsten Frey , Christian Luthje , and Simon Haag . 2011 . Whom Should Firms Attract to Open Innovation Platforms ? The Role of Knowledge Diversity and Motivation . Long Range Planning 44 , 5 \u2013 6 : 397 \u2013 420 . 7 . Morten Friis - Olivarius . 2013 . Overcoming design fixation through education and creativity methods . The 19th International Conference on Engineering Design . ( ICED\u201913 ) , 139 \u2013 148 . 8 . Dedre Gentner , Sarah Brem , Ron Ferguson , Philip Wolff , Arthur B . Markaman , and Ken Forbus . 1997 . Analogy and creativity in the works of Johannes Kepler . T . N . Ward , S . M . Smith , J . Vaid , eds . Creative Thought . Amer . Psych . Association , Washington , DC , 403 \u2013 460 . 9 . Dedre Gentner , Russell Landers , and others . 1985 . Analogical reminding : A good match is hard to find . In Proceedings of the international conference on systems , man and cybernetics , 607 \u2013 613 . 10 . Andrew Hargadon and Robert I . Sutton . 1997 . Technology brokering and innovation in a product development firm . Administrative science quarterly , 716 \u2013 749 . 11 . Keith James Holyoak . 1996 . Mental leaps : Analogy in creative thought . MIT press . 12 . Aniket Kittur , Ed H . Chi , and Bongwon Suh . Crowdsourcing user studies with Mechanical Turk . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI\u201908 ) , 453 \u2013 456 . 13 . Karim R . Lakhani , Lars Bo Jeppesen , Peter Andreas Lohse , and Jill A . Panetta . 2007 . The value of openess in scientific problem solving . Division of Research , Harvard Business School . 14 . J . S . Linsey , A . B . Markman , and K . L . Wood . 2012 . Design by analogy : A study of the WordTree method for problem re - representation . Journal of Mechanical Design 134 , 4 , 041009 . 15 . Tony McCaffrey . 2012 . Innovation relies on the obscure a key to overcoming the classic problem of functional fixedness . Psychological science , 0956797611429580 . 16 . David W . McDonald and Mark S . Ackerman . 2000 . Expertise recommender : a flexible recommendation system and architecture . In Proceedings of the 2000 ACM conference on Computer supported cooperative work ( CSCW \u201800 ) , 231 \u2013 240 . 17 . The New York Times . Car Mechanic Dreams Up a Tool to Ease Births . www . nytimes . com / 2013 / 11 / 14 / health / new - tool - to - ease - difficult - births - a - plastic - bag . html 18 . Tony Poze . 1983 . Analogical connections\u2014The essence of creativity . The Journal of creative behavior 17 , 4 : 240 \u2013 258 . 19 . Robert J . Sternberg . 1998 . Handbook of creativity . Cambridge University Press . 20 . Lynn Streeter and Karen E . Lochbaum . 1988 . An expert / expert - locating system based on automatic representation of semantic structure . In Proceedings of the Fourth Conference on Artificial Intelligence Applications ( IEEE \u201888 ) , 345 \u2013 350 . 21 . Thomas B . Ward . 1998 . Analogical distance and purpose in creative thought : Mental leaps versus mental hops . Advances in analogy research : Integration of theory and data from the cognitive , computational , and neural sciences , 221 \u2013 230 . 22 . Youtube , http : / / www . youtube . com / watch ? v = uL1ovAYtKuQ 23 . Lixiu Yu and Jeffrey V . Nickerson . 2011 . Cooks or cobblers ? : crowd creativity through combination . Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI\u201911 ) , 1393 \u2013 1402 . 24 . Lixiu Yu , Aniket Kittur , and Robert E . Kraut . 2014 . Distributed analogical idea generation : inventing with crowds . Proceedings of the SIGCHI conference on Human Factors in Computing Systems ( CHI\u201914 ) , 1245 \u2013 1254 . http : / / dx . doi . org / 10 . 1145 / 2556288 . 25573781 25 . Lixiu Yu , Aniket Kittur , and Robert E . Kraut . 2014 . Searching for analogical ideas with crowds . Proceedings of the 32nd annual ACM conference on Human factors in computing systems , ( CHI \u201814 ) , 1225 \u2013 1234 . http : / / dx . doi . org / 10 . 1145 / 2556288 . 2557378 26 . Jun Zhang and Mark S . Ackerman . 2005 . Searching for expertise in social networks : a simulation of potential strategies . Proceedings of the 2005 international ACM SIGGROUP conference on supporting group work , 71 - 80 1222 SESSION : CROWD INNOVATION AND CROWDFUNDING",
    "gickAnalogicalProblemSolving1980": "COGNITIVE PSYCHOLOGY 12 , 306 - 355 ( 1980 ) Analogical Problem Solving MARY L . GICK AND KEITH J . HOLYOAK University of Michigan The use of an analogy from a semantically distant domain to guide the problem - solving process was investigated . The representation of analogy in memory and processes involved in the use of analogies were discussed theoretically and explored in five experiments . In Experiment I oral protocols were used to exam - ine the processes involved in solving a problem by analogy . In all experiments subjects who first read a story about a military problem and its solution tended to generate analogous solutions to a medical problem ( Duncker\u2019s \u201cradiation prob - lem\u201d ) , provided they were given a hint to use the story to help solve the problem . Transfer frequency was reduced when the problem presented in the military story was substantially disanalogous to the radiation problem , even though the solution illustrated in the story corresponded to an effective radiation solution ( Experiment II ) . Subjects in Experiment III tended to generate analogous solutions to the radiation problem after providing their own solutions to the military problem . Subjects were able to retrieve the story from memory and use it to generate an analogous solution , even when the critical story had been memorized in the con - text of two distractor stories ( Experiment IV ) . However , when no hint to consider the story was given , frequency of analogous solutions decreased markedly . This decrease in transfer occurred when the story analogy was presented in a recall task along with distractor stories ( Experiment IV ) , when it was presented alone , and when it was presented in between two attempts to solve the problem ( Experi - ment V ) . Component processes and strategic variations in analogical problem solving were discussed . Issues related to noticing analogies and accessing them in memory were also examined , as was the relationship of analogical reasoning to other cognitive tasks . INTRODUCTION Where do new ideas come from ? What psychological mechanisms underlie creative insight ? This fundamental issue in the study of thought has received a great deal of informal discussion , but little empirical psychological investigation . The anecdotal reports of creative scientists and mathematicians suggest that the development of a new theory fre - quently depends on noticing and applying an analogy drawn from a differ - ent domain of knowledge ( Hadamard , 1954 ) . The hydraulic model of the blood circulation system , the planetary model of atomic structure , and the This research was supported by NSF Grant BNS77 - 01211 and NIMH Grant l - ROl - MH33278 - 01 to K . Holyoak . We thank Ellen Junn and Roger Kohlman for their assistance in testing subjects and scoring data , and David McArthur for helpful discussions . Robert Sternberg and an anonymous referee provided careful reviews of an earlier version of this paper . Experiments I and II were reported at the meeting of the Midwestern Psychological Association , May 1979 . Reprint requests should be sent to K . Holyoak at the University of Michigan , Human Performance Center , 330 Packard Road , Ann Arbor , MI 48104 . 306 OOlO - 0285 / 80 / 030306 - 50 $ 05 . 00 / O Copyright @ 1980 by Academic Press , Inc . AU rights of reproduction in any form reserved . ANALOGICAL PROBLEM SOLVING 307 \u201cbilliard ball\u201d model of gases all represent major scientific theories founded on analogies ( Boden , 1977 , Chap . 11 ) . As these examples suggest , fruitful analogies may be based on a map - ping of relations between two very disparate domains . In the brainstorm - ing technique of \u201cSynectics , \u201d problem - solving groups are trained to ac - tively search for analogies in areas other than that of the target problem ( Gordon , 1961 ) . It seems clear , however , that the \u201csemantic distance\u201d between analogous domains can vary a great deal . For example , Polya ( 1957 ) suggests that a useful strategy for solving geometry problems is to search for analogous problems within the domain of geometry . Analogies drawn within a domain may also play a role in categorization tasks . Clas - sification models based on comparisons of instances , which have been developed both in psychology ( Brooks , 1978 ; Medin & Schaffer , 1978 ) and in artificial intelligence ( Winston , 1975 ) , involve comparisons be - tween objects within a single domain ( e . g . , geometric patterns ) . Collins and his associates ( Collins , Warnock , Aiello , & Miller , 1975 ) have used protocol analyses to investigate the role of analogical reasoning by stu - dents who have incomplete knowledge of a problem domain such as geog - raphy . For example , a student might evaluate whether the region around Santiago , Chile is likely to produce wine by comparing it to a known wine - producing area , such as Northern California , on the relevant geo - graphic dimensions ( e . g . , latitude , proximity to ocean , type of terrain ) . While the process of solving analogy test items of the form A : B : : C : D has been studied quite extensively ( Sternberg , 1977a , 1977b ) , there has been little experimental investigation of analogical thinking in more com - plex problem - solving tasks . Some studies have examined transfer be - tween homomorphic or isomorphic versions of puzzle problems , such as the \u201cmissionaries and cannibals\u201d ( Reed , Ernst , & Banerji , 1974 ) and Tower of Hanoi ( Hayes & Simon , 1977 ) puzzles . These are relatively \u201cwell - defined\u201d problems ( Reitman , 1964 ; Simon , 1973 ) , in which the ini - tial conditions , legal operations , and goal state are explicitly specified . In contrast , anecdotal reports of the use of analogies typically involve prob - lems that are much less well defined . The present study was designed to investigate the use of analogies between disparate domains as a guide to finding solutions for an ill - defined problem . THE RADIATION PROBLEM AND ITS ANALOGIES Our basic experimental procedure was to provide subjects with a story analogy , describing a problem and its solution , and then to observe how subjects used the analogy in solving a subsequent target problem . The target problem was Duncker\u2019s ( 1945 ) \u201cradiation problem , \u201d which in our experiments was stated as follows . Suppose you are a doctor faced with a patient who has a malignant 308 GICK AND HOLYOAK tumor in his stomach . It is impossible to operate on the patient , but unless the tumor is destroyed the patient will die . There is a kind of ray that can be used to destroy the tumor . If the rays reach the tumor all at once at a sufficiently high intensity , the tumor will be destroyed . Unfortunately , at this intensity the healthy tissue that the rays pass through on the way to the tumor will also be destroyed . At lower intensities the rays are harm - less to healthy tissue , but they will not affect the tumor either . What type of procedure might be used to destroy the tumor with the rays , and at the same time avoid destroying the healthy tissue ? There are several reasons why the radiation problem seemed especially suitable for use in a study of analogical problem solving . First , it has all the hallmarks of the kind of \u201cill - defined\u201d problem for which an analogy from a remote domain might trigger a creative insight . The desired goal state is specified only at a relatively abstract level , and the permissible operations that might be used to achieve the goal are left very open ended . As a consequence , the possible solution proposals vary considerably . This made it possible to test for the use of analogies by attempting to influence the specific solutions that subjects would generate . In addition , we were able to benefit from Duncker\u2019s analyses of the performance of subjects who worked on the problem without receiving an analogy . Duncker identified three broad categories of proposed solutions to the radiation problem : ( 1 ) reducing the intensity of the rays as they pass through the healthy tissue ; ( 2 ) avoiding contact between the rays and healthy tissue ; and ( 3 ) altering the relative sensitivity to rays of the healthy tissue and the tumor ( e . g . , by immunizing the healthy tissue or sensitizing the tumor ) . Our analogies were designed to guide subjects toward specific versions of the first two classes of proposals . Our general aim in the present study , then , was to explore the process by which subjects use analogies between remote domains to generate problem solutions . Consequently , we wrote a series of stories far re - moved from the medical domain , each involving a military problem and its solution , which were analogous to the radiation problem . We will intro - duce the various stories as we proceed ; all are presented in the Appen - dixes . A FRAMEWORK FOR ANALOGICAL PROBLEM SOLVING It is important to develop a general conceptual framework within which specific issues concerning the role of analogies in problem solving can be formulated . What is meant by analogy , and how can an analogy be used to generate a problem solution ? In order to make our discussion more concrete we will consider the major story analogy used in the present experiments and the correspond - ANALOGICAL PROBLEM SOLVING 309 ing solution to the radiation problem . In the Attack - Dispersion story\u2019 a general wishes to capture a fortress located in the center of a country . There are many roads radiating outward from the fortress . All have been mined so that while small groups of men can pass over the roads safely , any large force will detonate the mines . A full - scale direct attack is there - fore impossible . The general\u2019s solution is to divide his army into small groups , send each group to the head of a different road , and have the groups converge simultaneously on the fortress . The analogous solution to the radiation problem is to simultaneously direct multiple low - intensity rays toward the tumor from different directions . In this way the healthy tissue will be left unharmed , but the effects of the multiple low - intensity rays will summate and destroy the tumor . At an intuitive level the parallels between the Attack - Dispersion story and the radiation problem are clear . Both situations involve an object that must be overcome , surrounded by objects that must be preserved . The target object in each case occupies a topographically central position in its environment . In each situation the protagonist has available a weapon with an effect proportional to the intensity or amount of the weapon that is used , and so on . How might people represent these analogical relationships and use them to generate a solution to the target problem ? This is not an easy question to answer . First of all , both the story and the problem must be read and understood . In attempting to describe this type of analogical problem solving we thus inherit all the problems associated with text comprehension . In particular , perception of analogy hinges on semantic knowledge and inference procedures . Since no general theory of language understanding is available , we must of necessity gloss over many impor - tant issues related to the understanding process . However , recent work on story comprehension ( Black & Bower , in press ; Kintsch , 1974 ; Kintsch & Van Dijk , 1978 ; Rumelhart , 1975 ; Schank & Abelson , 1977 ; Thorndyke , 1977 ) may offer some insights into how our story analogies might be represented in memory . Indeed , there appear to be close ties between the concept of analogy and the concept of \u201cschema , \u201d which has been widely applied in discussions of prose comprehension . In essence , both an analogy and a schema consist of an organized system of relations . Consequently , the framework for analogical problem solving presented here will draw its conceptual vocabulary from various schema - based models , as well as from Sternberg\u2019s ( 1977a , 1977b ) model of component processes involved in analogical reasoning . We will first consider how \u2019 In fact , three versions of this story were used in different experiments ( see Appendixes I - III ) . The versions differed only in minor points of wording of no consequence to the present discussion . 310 GICK AND HOLYOAK analogy might be represented , and then how this representation could be used to generate a solution to a problem . The Representation of Analogy A system of representation for analogy must be able to describe a fundamental property of such relational systems , namely , that analogy may be defined at multiple levels of abstraction . For example , at a rela - tively low level of abstraction the Attack - Dispersion story and the radia - tion problem have a variety of corresponding details ( e . g . , small groups of soldiers correspond to low - intensity rays ) . At a more abstract level , the story and the problem both involve the goal of overpowering an object located in a region that must be preserved . The multileveled nature of analogy can perhaps be understood in the context of Kintsch and Van Dijk\u2019s ( 1978 ) theory of prose representation . They argue that the understanding process may involve the iterative ap - plication of a set of inference rules that generate increasingly abstract \u201cmacrostructure\u201d representations of a prose passage . These macro - structures essentially correspond to summaries of the passage at various levels of generality . In the case of a problem - oriented story such as the Attack - Dispersion story , an abstract level of macrostructure might state a general solution principle ( e . g . , to destroy a target when direct application of a large force is harmful to the surrounding area , disperse the attacking forces , and have them converge at the target ) . The process of extracting a solution principle might thus be viewed as a special case of the process of deriving macrostructures for a body of information . While much remains to be learned about how this process operates , the three specific inference rules proposed by Kintsch and Van Dijk ( which they term \u201cdeletion , \u201d \u201cgeneralization , \u201d and \u201cconstruction\u201d ) would seem readily applicable to the type of story analogies we are considering here . Kintsch and Van Dijk emphasize that control processes are required to select a level of macrostructure analysis consistent with the person\u2019s processing goals . Similarly , we assume there is an optimal level of abstraction at which analogical relations may be represented in order to effectively guide the solution process . Indeed , an important empirical issue is to determine what factors influence this optimal level of abstrac - tion . We will now consider in more detail how an analogy between two relational systems might be represented , assuming an appropriate level of macrostructure has been derived . To pursue our example , Table 1 pre - sents our own summary of the Attack - Dispersion story , as well as a sum - mary of the radiation problem and its dispersion solution . These sum - maries are intended to reflect the major causal connections within both the story and the problem , and to illustrate the major analogical relations ANALOGICAL PROBLEM SOLVING 311 between them . The sentences in Table 1 are numbered to correspond to an approximate propositional analysis presented in Fig . 1 . Propositions from the story and from the radiation problem are matched to indicate analogical relations , and the propositions corresponding to the dispersion solution to the radiation problem ( which a person would be required to generate ) are italicized in both the table and the figure . Note that some of the propositions included in the story summary ( e . g . , proposition 11 ) are inferences , which are not directly stated in the original story ( see Appen - dix I ) . The notation in Fig . 1 consists of propositional functions , in which predicates are followed by one or more arguments ( enclosed in paren - theses ) . The arguments fill various semantic roles , such as agent , ob - TABLE 1 A Summary of Attack - Dispersion Story and of Corresponding Solution to Radiation Problem ( See Fig . 1 ) Proposition number Attack - Dispersion story l - 2 A fortress was located in the center of the country . 2a Many roads radiated out from the fortress . 3 - 4 A general wanted to capture the fortress with his army . 5 - 7 The general wanted to prevent mines on the roads from destroying his army and neighboring villages . 8 As a result the entire army could not attack the fortress along one road . 9 - 10 However , the entire army was needed to capture the fortress . 11 So an attack by one small group would not succeed . 12 The general therefore divided his army into several small groups . 13 He positioned the small groups at the heads of different roads . 14 - 15 The small groups simultaneously converged on the fortress . 16 In this way the army captured the fortress . l\u2019 - 2\u2019 3\u2019 - 4\u2019 5\u2019 - 7\u2019 8\u2019 Radiation problem and dispersion solution\u201d A tumor was located in the interior of a patient\u2019s body . A doctor wanted to destroy the tumor with rays . The doctor wanted to prevent the rays from destroying healthy tissue . As a result the high - intensity rays could not be applied to the tumor along one path . 9\u2019 - 10\u2019 11\u2019 12\u2019 13\u2019 14\u2019 - 15\u2019 16\u2019 However , high - intensity rays were needed to destroy the tumor . So applying one low - intensity ray would not succeed . The doctor therefore divided the rays into several low - intensity rays . He positioned the low - intensity rays at multiple locations around the patient\u2019s body . The low - intensity rays simultaneously converged on the tumor . In this way the rays destroyed the tumor . n Italicized propositions summarize the target dispersion solution . 312 GICK AND HOLYOAK ject , and location . Propositions may themselves serve as arguments , so in several cases one proposition is embedded in another . We make no claims about the logical adequacy or completeness of the representation in Fig . 1 ; indeed , many of the indicated arguments ( e . g . , \u201clow - intensity rays\u201d ) clearly could be further decomposed . However , separation of relations ( predicates ) and arguments serves to highlight the critical properties of analogy between remote domains : similarity of corresponding relations despite dissimilarity of corresponding arguments . The notation in Fig . 1 has been augmented by labeled arcs that repre - sent the major causal connections within the Attack - Dispersion story and the radiation problem . The labels are inspired by but not identical to the analysis of causal types offered by & hank and Abelson ( 1977 ) ( see also Black & Bower , in press ) . Roughly , a goal can be a \u201creason\u201d for an action ; a state or action can \u201cenable\u201d a subsequent action ; an action can \u201cresult\u201d in a subsequent state ; and a state or action can \u201cprevent\u201d a subsequent action . Again , the adequacy of this analysis is not really crit - ical for our present purpose , which is simply to make salient the corre - spondences in causal structure between the story and the problem , par - ticularly with respect to the target solution . Note that the labeled arcs are equivalent to higher - order predicates that embed numbered propositions as arguments . The representation in Fig . 1 can be used to highlight a variety of prop - erties of analogy , as well as some of the issues we have glossed over in constructing this representation . Fundamentally , an analogy consists of a yre @ nerd . ytue brmm fortress ) ) desm & ml . preat & shuy ( mines , army / v ~ llcqsd ) desm & ml . preat & shuy ( mines , army / v ~ llcqsd ) 6 5 6 5 allack ( enhre ormy , fortmrr . ona read ) atlack ( enhre ormy , fortmrr . ona read ) requre kapture hirers ) , sntwe army ) requre kapture hirers ) , sntwe army ) attack bnc small qwp , fwtm . ss ) attack bnc small qwp , fwtm . ss ) tgeenad , ormy , small grwpr ) locate iiatrerr , can * kou ~ trYl1 y2 I radiate ( roods , fortress ) $ s . e kkctw , B\u201d\u201d mys , tumw11 ( dacta , p $ vent $ strny ( roys , healthy t ! ssuaN ( one bw urtawty ray , tumw ) W groups , different roads ) konvenp bmatl groqs . fatressl ) FIG . 1 . Analogical correspondences between the Attack - Dispersion story and the radia - tion problem . ANALOGICAL PROBLEM SOLVING 313 mapping between two sets of propositions . Propositions are matched on the basis of similarity between the corresponding relations . Note that similarity need not imply identity . There need only be a consistent trans - formation that maps one set of relations onto the other . Boden ( 1977 ) gives the example of the Black Mass , which is based on systematic semantic reversals of the Catholic ritual . In the case of our story and problem , we wrote the summaries to maximize the similarity of the rela - tions . Yet dividing an army ( proposition 12 ) , for example , is clearly not quite the same as dividing high - intensity rays ( proposition 12\u2019 ) . To gener - ate this aspect of the parallel solution the person would need to take account of the relevant differences between an army and rays , perhaps explicitly introducing multiple machines that each emit low - intensity rays . In order to map proposition 1 onto 1\u2019 a person must have semantic knowl - edge of the relation between the meanings of \u201ccenter\u201d and \u201cinterior . \u201d Incidentally , note that we assume proposition 1\u2019 is included in the mac - restructure for the problem as an inference based on knowledge of where the stomach is located . In the case of propositions 16 and 16\u2019 , we assume the person will use semantic knowledge to transform the relation of \u201ccapturing\u201d into the relation of \u201cdestroying , \u201d operating on the common semantic core ( i . e . , \u201covercoming\u201d ) that links the two relations . As Fig . 1 indicates , there is clearly a high degree of correspondence between the propositions of the story and of the problem . At the same time , the systems are not perfectly isomorphic . It is probably the case that analogies used to guide problem solving are generally incomplete in some respects ( Gentner , Note 1 ) . For example , proposition 2a in the Attack - Dispersion story , which states that many roads radiate outward from the fortress , has no parallel in the statement of the radiation problem . Note that in the story this proposition serves as an important enabling condition for the solution . The absence of any explicit mention in the radiation problem of multiple paths to the target object would presumably hinder generation of the dispersion solution . For the above example it is plausible to argue that people must infer the fact that there are multiple potential \u201croutes\u201d to the tumor in the course of generating the dispersion solution , even though no such inference is represented in Fig . 1 . But in addition , Fig . 1 reveals at least one clear disanalogy between the story and the problem . In a complete analogy , there is a consistent mapping between pairs of arguments . That is , wher - ever argument A occurs in one relational system , argument A\u2019 occurs in the other . For example , in Fig . 1 the role the fortress plays in the story consistently maps onto the role the tumor plays in the problem . Note that the role of the army usually corresponds to that of the rays . However , this is not the case in propositions 5 and 5\u2019 . In the Attack - Dispersion story , sending the entire army down one road will result in destruction of the 314 GICK AND HOLYOAK army ( as well as neighboring villages ) by mines ; whereas in the radiation problem applying high - intensity rays to the tumor will result in destruc - tion of the healthy tissue by the rays . In other words , the army and the rays do not till corresponding semantic roles in propositions 5 and 5\u2019 ; rather , the army is the object of the process of destruction in 5 , while the rays are the instrument of the destruction in 5\u2019 . This example illustrates that degree of analogy in part depends on the level of abstraction at which the analogy is defined . In macrostructures slightly more abstract than those depicted in Fig . 1 , the fact that an attack by the entire army is impossible would map onto the fact that direct application of high - intensity rays is impossible . At this level the roles of the army and of the rays correspond appropriately . However , the story and the problem are disanalogous at the more specific level depicted in Fig . 1 ( i . e . , the level of the reas0n . s why the two respective courses of action are blocked ) . This observation suggests that for use in solving a problem the optimal level of abstraction for representing an analogy may be that which maximizes the degree of correspondence between the two relational systems . In many cases a very detailed representation will in - clude disanalogous relations , while a very abstract representation will omit information about important correspondences . The Process of Analogical Problem Solving So far we have been discussing how analogical relations may be repre - sented ; we must now consider how this information might be used to generate a solution to a problem . For our story analogy and target prob - lem the solution process appears to require three major steps . ( 1 ) A representation of the story analogy and of the target problem ( i . e . , its initial state and goal state ) must be constructed , as described above . ( 2 ) The representation of the story must be mapped onto that of the problem . If the story and the problem are drawn from remote domains , as in our example , the correspondences between arguments will not be im - mediately obvious . We would therefore expect the mapping process to be initiated by detection of similar relations in the two systems . For example , the person might notice that propositions 2 and 2\u2019 both involve location . Accordingly , a mapping between the two propositions will be established . This will automatically establish a mapping between the corresponding arguments ( i . e . , the fortress and the tumor , the center of the country and the interior of the body ) . Once a few such correspondences have been detected , the mapping process may proceed in a more \u201ctop - down\u201d man - ner , guided by expectations that previously mapped arguments will con - tinue to play parallel roles . For example , having mapped propositions 2 and 2\u2019 , the person might assume that 8 maps onto 8\u2019 because the role of the fortress should correspond to that of the tumor . ANALOGICAL PROBLEM SOLVING 315 ( 3 ) Finally , the person must use the mapping to generate the parallel solution to the target problem . This can be done by constructing a set of solution propositions for the target problem that correspond to the solu - tion propositions of the story . For example , consider how proposition 12\u2019 might be generated on the basis of proposition 12 . The mapping process will have identified the general with the doctor and the army with the rays . Accordingly , \u201cdoctor\u201d and \u201crays\u201d will be used to fill the argument slots corresponding to \u201cgeneral\u201d and \u201carmy . \u201d In addition , the relation be - tween the general and the army in 12 will be used to construct a parallel relation between the doctor and the rays in 12\u2019 . Thus the idea of the general dividing the army into several small groups will be transformed into the idea of the doctor dividing the rays into a number of low - intensity rays , initiating the dispersion solution to the radiation problem . ISSUES AND EXPERIMENTS A number of important questions arise within the framework of the process model we have outlined . A major issue , which we touched upon earlier , concerns the level of macrostructure at which the mapping pro - cess takes place . At one extreme the solution process might amount to abstracting a solution principle from the story and then applying it to the target problem . At the other extreme subjects might map the corre - spondences between the story and the problem at the most detailed possi - ble level . It is possible , of course , that the mapping process may actually proceed partially in parallel on several different levels . Even at a single level of macrostructure , there may be strategic varia - tions in the degree of mapping that takes place during the solution pro - cess . For example , subjects need not derive the entire set of corre - spondences outlined in Fig . 1 in order to generate the dispersion solution . One possibility , which we will term the \u201csolution - focusing\u201d strategy , is that subjects attempting to apply the story analogy will immediately iden - tify the solution propositions of the story . By doing the minimal amount of mapping required to match the arguments in these propositions with ar - guments in the radiation problem , the parallel solution could be gener - ated . Subjects using the solution - focusing strategy might thus solve the target problem without entirely grasping the correspondences between the problem statements in the story and in the radiation problem . Given the lack of empirical research on analogical problem solving , even more basic issues arise . We have sketched a model of how in princi - ple a problem might be solved on the basis of an analogy . However , we do not know whether subjects could actually execute this kind of process for our story analogies and target problem . There seem to be at least three distinct ways in which subjects who have a relevant analogy available might nonetheless fail to derive the parallel solution to a target problem . 316 GICK AND HOLYOAK The first and most basic is that subjects might be unable to successfully apply the story analogy even if they tried . Second , even if a story analogy is potentially useful , subjects might be unable to locate it in memory , especially if it had been encoded in the context of irrelevant distractor stories . Third , subjects might be able to retrieve a potentially useful anal - ogy and yet fail to spontaneously notice its relevance to the target problem . The experiments reported below were designed to explore these and related issues . In Experiments I - III subjects were presented with story analogies and given a hint to use them to solve the radiation problem . In addition to investigating whether subjects can in fact use analogies to generate problem solutions , Experiments I and II were intended to pro - vide some information about the processes involved in analogical problem solving . In Experiment I the oral protocols of subjects solving the target problem were analyzed , and in Experiment II the degree of analogical correspondence between the story and the target problem was varied . In Experiment III subjects were asked to first solve the problem presented in the story themselves ( rather than having its solution presented to them ) , and then to attempt to use their solutions as aids in solving the target problem . Experiments IV and V investigated the ability of subjects to solve the target problem on the basis of story analogies stored in memory . In addition , the latter two experiments varied whether or not subjects were provided with a hint to use a story analogy to help solve the target problem . These experiments thus examined the propensity of subjects to spontaneously notice and apply potential analogies between remote problem domains . EXPERIMENT I Experiment I was designed to demonstrate that subjects can use an analogy from a remote domain as a hint for solving a problem . Subjects first read a story analogy , and then attempted to propose as many solu - tions as possible to the radiation problem . By varying the nature of the solution suggested by the story , we hoped to influence the likelihood that subjects would generate specific solutions to the target problem . Subjects\u2019 \u201cthinking aloud\u201d protocols were tape recorded and later analyzed as a source of evidence regarding the process of analogical problem solving . Subjects in three experimental conditions read one of three stories about a military problem and its solution ( see Appendix I ) . Table 2 infor - mally illustrates the correspondences among the three stories and the radiation problem . The statement of the radiation problem ( see Introduc - tion ) was worded so as to minimize obvious lexical or syntactic corre - spondences with the story analogies . The Attack - Dispersion , Open Sup - ply Route , and Tunnel stories all have identical first paragraphs describing ANALOGICAL PROBLEM SOLVING 317 TABLE 2 Schematic Outline of Duncker\u2019s Radiation Problem Showing Correspondences with Analogous Stories Problem Statement Radiation problem Story analogies - Experiment I Problem setting Doctor has rays . General has army . Patient has tumor . Country has dictator . Tumor in stomach , surrounded Dictator in fortress in center of by healthy tissue . country , surrounded by villages . Roads radiate from fortress like spokes on a wheel . Desired goal Destroy tumor with rays . Capture fortress with army . Problem High - intensity rays destroy tumor Entire army can capture fortress , constraints but also destroy healthy tissue . but large group detonates mines on roads , destroying army and villages . Low - intensity rays destroy Small group of men can pass safely neither tumor nor healthy over roads but can not capture tissue . fortress . Impossible to operate . Solutions Type 1 Reduce intensity of rays on way Reduce size of group traveling to to tumor . fortress on one road . Dispersion ( 1 ) Many low intensity rays ( 1 ) Many small groups of men ( Attack - ( 2 ) From different directions ( 2 ) From different directions Dispersion ( 3 ) Simultaneously ( 3 ) Simultaneously story ) Type 11 Avoid contact between rays and Avoid contact between army and healthy tissue . mines . ( 1 ) Open pas - Send high - intensity rays through General discovers road that is not sage an open route , ( e . g . , esopha - mined , and sends entire army ( Open Supply pus ) . down this road . Route story ) ( 2 ) Operation Make an incision in stomach wall , Dig tunnel under mines , and send ( Tunnel story ) removing healthy tissue from entire army through . path of rays , and apply high intensity rays to tumor . \u201d Resulting goal state Radiation of high - intensity reaches tumor . Tumor destroyed . Healthy tissue intact . Entire army reaches fortress . Fortress captured . Army and villages preserved . n Incision violates constraint . the problem setting , desired goal , and the constraints on a solution . These aspects of the stories are analogous to the radiation problem , as discussed earlier ( see Fig . 1 ) . However , the stories differ in their second paragraphs , which state the 318 GICK AND HOLYOAK general\u2019s solution to his problem . In the Attack - Dispersion story ( de - scribed in the Introduction ) the general divides his army into small groups and sends them simultaneously down different roads to the fortress . The analogous solution to the radiation problem is the \u201cdispersion\u201d solution : have multiple low - intensity rays converge at the tumor . This is a very effective solution , 2 but one which subjects seldom generate spontane - ously . Duncker ( 1945 ) reported that only 2 of 42 subjects arrived at this dispersion solution , and both were prompted by the experimenter . A basic difftculty that appears to block generation of this solution is that people do not spontaneously think of rays as having the property of \u201cdi - visibility . \u201d In fact , Duncker found that the frequency of the dispersion solution increased when the term \u201cparticles\u201d was substituted for \u201crays\u201d ( presumably because particles are more obviously divisible ) . In the Open Supply Route story the general discovers an unblocked road leading to the fortress , and sends the entire army down this open road . An analogous radiation solution is to direct high - intensity rays down the esophagus ( or some other open passage , such as the intestines ) to the stomach . This solution was generated relatively frequently by the subjects tested by Duncker ( 29 % gave the open passage solution as opposed to only 5 % who gave the dispersion solution ) . In the Tunnel story the general digs an underground tunnel and sends his army through it to the fortress . Analogous radiation solutions might be to operate to expose the tumor to the rays , or to insert a tube through the stomach wall and send rays through it to the tumor . Many of Duncker\u2019s subjects ( 40 % ) spontaneously suggested such solutions . However , such procedures to create an open route to the tumor involve operating , and hence conflict with one of the constraints imposed on the radiation problem ( that it is impossible to operate ) . The Tunnel story is therefore a kind of \u201cfalse analogy\u201d to the radiation problem . That is , although the problem statements are analo - gous , the solution suggested by the story is inappropriate . If the analogy is nevertheless applied , subjects given the Tunnel story might be especially likely to momentarily disregard the problem constraints and propose an operation solution to the radiation problem . Although the above analysis of the analogous relationships between various solutions to the military problem and to the radiation problem was 2 This solution is functionally very similar to the standard medical procedure for radiation therapy , which is to rotate the radiation source around the patient ( or vice versa ) in such a way that the tumor is always the focal point of the radiation . The malignancy thus receives a cumulative dose of radiation while other areas receive a lesser amount . One difference is that our dispersion solution involves simultaneous application of the rays , whereas the medical procedure takes advantage of the fact that the effects of radiation summate over time . Even knowledge of the medical procedure would therefore be unlikely to lead to the exact solution corresponding to the Attack - Dispersion story . ANALOGICAL PROBLEM SOLVING 319 initially based on the experimenters\u2019 intuitions , we will see below that subjects\u2019 ratings essentially confirm the validity of this analysis . The primary prediction in Experiment I was that each story analogy would tend to increase the frequency of the analogous solution to the radiation problem , relative to the solution frequencies obtained for con - trol subjects given no prior story . However , there are additional ways in which the story analogies might influence the solutions given to the target problem . First , note that the problem statements for all three stories contain all the enabling conditions ( see Fig . 1 ) for generating the disper - sion solution ( e . g . , the central location of the fortress , the roads radiating outward , the fact that small groups can travel on the roads ) . Accordingly , subjects might spontaneously think of the dispersion solution to the gen - eral\u2019s problem , and then use it to generate the parallel solution to the radiation problem . If so , subjects given the Open Supply Route and Tun - nel stories might also produce the dispersion solution more often than would control subjects . It is also possible that giving subjects a story analogy may actually hinder the generation of nonanalogous solutions . That is , attempting to generate a parallel solution to the target problem may create a kind of \u201cset\u201d effect , so that other possible solutions ( e . g . , immunizing the healthy tissue to protect it from the rays ) will not be discovered . If such a set effect is obtained , control subjects should produce more total solu - tions than experimental subjects , and in addition there should be qualita - tive differences between the solutions produced by control subjects ver - sus subjects given story analogies . Method Subjects were divided into four conditions , each receiving either the Attack - Dispersion story , the Open Supply Route story , the Tunnel story , or no story ( the control group ) , prior to solving the radiation problem . The control condition used no story at all , rather than an irrelevant story , because it seemed possible that an irrelevant story would actually interfere with the solution process . ( A control condition using an irrelevant story was included in Experiment V below . ) Subjects were tested individually by a trained experimenter and sessions were tape recorded . All subjects first solved Duncker\u2019s ( 1945 ) \u201ccandle problem\u201d to familiarize them with the process of thinking aloud while problem solving . Subjects were told that the problems required some creativity for their solutions , and that they should not feel inhibited about making any suggestions that came to mind . They were also encouraged to give an ongoing account of what they were thinking about . Subjects were asked to begin by reading the problem out loud in order to get them used to speaking in front of the experimenter . Following the practice problem subjects in the experimental conditions were told that they would receive two further problems , the first of which would also have a solution given for it . They were told to read the first \u201cstory problem\u201d aloud and then to orally summarize the gist or point of the story . This part of the procedure was omitted for subjects in the control condition . All subjects then read the radiation problem and began to solve it . They were reminded to talk out loud , and encouraged to interrupt their reading of the problem at any 320 GICK AND HOLYOAK time if a solution occurred to them . Experimental subjects were told to try to use the first story problem as a hint in solving the second ( radiation ) problem . However , they were also told that it was not necessary to use the prior story in order to solve the problem . Subjects were allowed to reread the story analogy at any time . As subjects worked on the radiation problem the experimenter was prepared to intervene with an explicit hierarchy of prompts . If subjects were not explicit about the nature of a proposed solution , the experimenter asked them to clarify it , sometimes by drawing a diagram . Subjects in the experimental conditions who at first failed to generate the analo - gous solution were eventually prompted to reread the instructions . If they still did not produce the analogous solution , they were then reminded to use the prior story as a hint . At the end of the session , subjects were asked to rate each of their proposed solutions on two 7 - point scales , as to how \u201ccreative\u201d and how \u201cpractical\u201d the solutions are . A rating of 1 indicated maximum creativity or practicality . Forty undergraduates enrolled in introductory psychology at the University of Michigan served as subjects as part of a course requirement . Ten subjects were assigned to each of the four conditions . Results and Discussion Frequencies of analogous solutions . Subjects\u2019 protocols for the radia - tion problem were transcribed and scored for the presence of various types of proposed solutions , by two independent scorers . For this pur - pose any suggestion , even if it was eventually rejected by the subject , was counted as a proposed solution . The results of major interest concern the three types of proposals that are analogous to the solutions embodied in the story analogies - the dispersion solution , the open passage solution , and operation solutions . Table 3 presents the percentage of subjects in each condition who produced these various types of proposed solutions . The frequency of each solution was highest for subjects who received the relevant story analogy , i . e . , the dispersion solution was most frequent for the Attack - Dispersion condition , the open passage solution was most fre - quent for the Open Supply Route condition , and operation solutions were most frequent for the Tunnel condition . These differences in solution frequencies were most dramatic for the dispersion solution . All 10 subjects who were given the Attack - Dispersion story produced this solution , whereas not a single control subject did so . For this solution the frequency differences among the four conditions were highly significant , G2 ( 3 ) = 33 . 9 , p < . OOl , as was a comparison between the Attack - Dispersion condition versus all others , G2 ( 1 ) = 30 . 9 , p < . 001 . 3 The frequencies of the open passage solution and of operation solutions were also influenced by the story analogies . Seventy percent of subjects in the Open Supply Route condition produced the open passage solution , as opposed to 20 % of subjects in all other conditions , G2 ( 1 ) = 8 . 21 , p < 3 All contingency table analyses reported in the present paper use the GZ statistic ( maximum likelihood x2 ) ( Bishop , Fienberg , & Holland , 1975 ) . ANALOGICAL PROBLEM SOLVING 321 TABLE 3 Percentage of Subjects in Each Condition of Experiment I Who Proposed Various Solutions to the Radiation Problem Proposed solution Condition Dispersion Open Passage Operation\u201d Attack - Dispersion story 100 10 30 Open Supply Route story 10 70b 50 Tunnel story 20 30 80 Control 0 20 50 U This solution type includes proposals to operate to clear a path for the rays or to insert a tube through the stomach wall and send rays through to the tumor . b This value includes one subject who proposed the abstract idea of finding an open pas - sage through which the rays could be directed at the tumor , but who failed to suggest the esophagus or any other concrete possibility as a route . . Ol . Eighty percent of subjects in the Tunnel conditions produced opera - tion solutions , as opposed to 43 % of subjects in all other conditions , G2 ( 1 ) = 4 . 29 , p < . 05 . Ratings of stories as solution prompts . As we noted earlier , we initially used our own intuitions about the semantic parallels between the military and the medical solutions to predict the radiation solutions that would be triggered by the various story analogies . In order to assess whether sub - jects shared our intuitions we administered a rating task to 51 under - graduates , none of whom previously knew of any solutions to the radia - tion problem . For each of the three stories used in Experiment I , 17 subjects first read the story , then the radiation problem , and finally the six possible solution proposals listed in Table 4 . Subjects rated each of the proposals ( which were listed in a random order ) on a 9 - point scale as to how likely they thought it was that the story they had just read would make them think of the proposed solution . A rating of 9 indicated maximum likelihood that the story would suggest the solution . Subjects were instructed not to consider the practicality of the various solutions in making their judgments . The mean ratings are presented in Table 4 . Note that proposal 1 is the dispersion solution , 2 is the open passage solution , and 3 and 4 are opera - tion solutions . Proposals 5 and 6 are two other fairly common suggestions that we expected would be relatively unrelated to all of the three stories . ( Proposal 5 is a type of \u201caccumulation\u201d solution , which will be discussed in connection with Experiment III ) . The proposal ratings differed greatly depending on which story subjects had read , F ( 10 , 240 ) = 8 . 84 , p < . OOl . For each story analogy , Newman - Keuls tests were performed on the differences among the ratings for the six proposals . In the case of the 322 GICK AND HOLYOAK TABLE 4 Mean Ratings of Likelihood that Stories Would Help to Think of Various Radiation Solutions\u201d Proposed solutions 1 . Apply low - intensity rays from several different directions so they simultaneously converge at the tumor . 2 . Send high - intensity rays down the esophagus so they strike the tumor . 3 . Insert a tube through the healthy tissue to the tumor , and then send high - intensity rays through the tube to the tumor . 4 . Make an incision into the stomach to expose the tumor , and then apply high - intensity rays directly to the tumor . 5 . Treat the person with low - intensity rays , repeating the treatment a number of times . 6 . Treat the person with medium - intensity rays . Solution number story 1 2 3 4 5 6 Atack - Dispersion 8 . 41 2 . 53 3 . 71 2 . 94 4 . 88 3 . 47 Open Supply Route 3 . 59 6 . 65 5 . 82 4 . 58 2 . 94 2 . 29 Tunnel 5 . 12 4 . 47 6 . 47 5 . 35 2 . 71 1 . 82 ( 1 A rating of 9 indicated maximum likelihood that the story would suggest the solution . Attack - Dispersion story , subjects gave higher ratings to the dispersion solution than to any other ( p < . Ol ) . For the Open Supply Route story , the open passage solution received the highest rating , although it did not differ significantly from the ratings given the two operation solutions . However , these three solutions were all rated significantly higher than any of the others ( p < . 05 ) . For the Tunnel story , the \u201cinsert a tube\u201d solution ( proposal 3 ) received the highest rating , followed by the other operation solution ( proposal 4 ) , the dispersion solution , and the open passage solution . These four proposals did not differ significantly from each other . However , the two operation solutions received significantly higher ratings than proposals 5 and 6 ( p < . 05 ) . The fact that subjects gave the highest rating of all to the Attack - Dispersion story as a prompt for the dispersion solution is consistent with the fact that this story appeared to be the most effective analogy in Ex - periment I . The overlap in the ratings for the solutions analogous to the Open Supply Route and Tunnel stories ( proposals 2 , 3 , and 4 ) also reflects trends in the observed solution frequencies ( see Table 3 ) . Among the three story conditions , the Open Supply Route condition produced the second highest frequency of operation solutions , while the Tunnel condi - tion produced the second highest frequency of open passage solutions . Notice that both types of solutions involve avoiding contact between the rays and the healthy tissue , by directing the rays through an unobstructed ANALOGICAL PROBLEM SOLVING 323 route . The basic difference between the two solution types is simply that the open passage solution makes use of a preexisting route ( the esophagus ) , while the operation solutions require that the route be con - structed ( by some type of operation ) . At first glance the relatively high rating given to the Tunnel story as a prompt for the dispersion solution is more puzzling , since digging a tunnel seems very disanalogous to dispersion of rays . In fact , however , as Table 3 indicates , two subjects in the Tunnel condition actually did produce the relatively rare dispersion solution . Furthermore , as we will see when we discuss the protocols in more detail below , both of these subjects also spontaneously suggested that the general might have sent his army down multiple roads . Recall that the identical first paragraphs of all three stories contain the enabling conditions for the dispersion solution . However , in the Open Supply Route story the given solution involves use of only one road , which may create a set effect that blocks consideration of how multiple roads might be used . In contrast , the solution given in the Tunnel story avoids use of the roads altogether . Since some kind of attack by road is nonetheless an obvious possibility , the stated enabling conditions may lead subjects who read the Tunnel story to think of the alternative solution of using multiple roads , which in turn could prompt the disper - sion solution to the radiation problem . While some aspects of the above analysis are certainly speculative , it seems clear that subjects in the rating task were highly sensitive to factors that actually influenced the effective - ness of the story analogies in Experiment I . Frequencies of other solutions . The possibility of a set effect , men - tioned above , raises the general issue of whether story analogies actually block generation of qualitatively different solution proposals . To investi - gate this question an analysis of variance was performed on the total number of proposed solutions , other than the primary analogous solution , that were given by subjects in the various conditions . That is , dispersion solutions were excluded for the Attack - Dispersion condition , open pas - sage solutions were excluded for the Open Supply Route condition , oper - ation solutions were excluded for the Tunnel condition , while none of the above solutions were excluded for the control condition . In addition , a small number of proposals ( a mean of 0 . 4 per subject ) that did not involve use of the rays at all ( e . g . , Laetrile treatment ) were excluded for all conditions . The average number of nonanalogous solutions produced was 1 . 10 for the Attack - Dispersion condition , 1 . 60 for the Open Supply Route condi - tion , 2 . 70 for the Tunnel condition , and 2 . 00 for the control condition , F ( 3 , 36 ) = 4 . 17 , p < . 025 . Newman - Keuls tests revealed that only the difference between the Tunnel and Attack - Dispersion conditions was sig - nificant ( p < . Ol ) . As we noted earlier , subjects in the Tunnel condition 324 GICK AND HOLYOAK tended to produce a relatively high number of open passage and disper - sion solutions . The rating results ( Table 4 ) also indicated that the Tunnel story is a moderately effective prompt for several different solutions . The trends toward fewer alternative proposals in the Attack - Dispersion and Open Supply Route conditions , compared to the control condition , at least suggest the possibility of some kind of set effect . In order to obtain further evidence regarding a possible set effect , the frequencies of specific radiation solutions , other than those analogous to the various stories , were tabulated for each condition . The solutions ex - amined were proposals to treat the healthy tissue directly , rather than altering the route that the rays take . Specifically , these solutions suggested decreasing the sensitivity of the healthy tissue to the rays ( e . g . , by a chemical injection , or building up a tolerance to the rays ) , or covering the healthy tissue with a barrier to protect it from the rays ( e . g . , by inserting a lead shield to protect the healthy tissue ) . Such solutions were produced by 30 % of the subjects in the control condition , 10 % of the subjects in the Tunnel condition , and none of the subjects in the Attack - Dispersion and Open Supply conditions . While the numbers involved were too small to be statistically reliable , these results suggest that an analogy may tend to block generation of alternative types of solutions . Practicality and creativity ratings . The practicality and creativity rat - ings that subjects gave for their own solutions were examined . Within each story condition , the ratings given to the analogous solution were compared to the means of the ratings given to other solutions by the same subjects . In the Attack - Dispersion condition seven subjects produced and rated the dispersion solution and at least one other proposal . The disper - sion solution was rated as much more practical than other solutions ( 1 . 43 versus 4 . 64 ) , F ( 1 , 6 ) = 27 . 2 , p < . Ol . However , the creativity ratings did not differ significantly between the dispersion solution and other propos - als ( 3 . 43 versus 3 . 64 ) , F < 1 . It seems likely that subjects did not perceive their dispersion solutions as especially creative because they were aware of using the prior story to generate an analogous solution . Parallel analyses for the Open Supply Route and Tunnel conditions revealed no significant differences . Collapsing over subjects in all condi - tions , the dispersion solution tended to be rated as most practical ( a mean of 2 . 0 ) , followed by the open passage ( 3 . 9 ) and operation solutions ( 5 . 4 ) . We also asked an independent group of 20 undergraduates ( none of whom previously knew of the radiation problem ) to choose which solution was more practical : the dispersion solution ( \u201capply low - intensity rays from several different directions so they simultaneously converge at the tumor\u201d ) , or the open passage solution ( \u201csend high - intensity rays down the esophagus so they strike the tumor\u201d ) . Fifteen of the twenty subjects selected the dispersion solution as more practical @ < . 05 by a sign test ) . ANALOGICAL PROBLEM SOLVING 325 When asked to justify their decision , 11 of these 15 mentioned the prob - lem of preventing the rays from destroying the tissue lining the esophagus . It may be that some subjects in Experiment I implicitly generated the open passage solution ( or an operation solution ) on the basis of the rele - vant story analogy , but then failed to mention it ( despite the instruction to mention any possibility ) because they recognized its inadequacy . We will consider other sources of difficulty in using the Open Supply Route and Tunnel stories below , when subjects\u2019 protocols are discussed in more detail . Problem - solving protocols . The results discussed so far demonstrate that story analogies can play a major role in directing the problem - solving process . However , they reveal little about the process by which subjects arrive at an analogous solution . We therefore supplemented the quantita - tive analysis of solution types , reported above , with a more qualitative analysis of subjects\u2019 problem - solving protocols . Several aspects of the protocols were examined . Occasions when the experimenter prompted the subjects to use the story were noted , as were correspondences be - tween the story and the target problem that subjects mentioned in the course of generating solutions . This analysis was , of course , constrained by the overall quality and quantity of the protocols . For example , some subjects insisted that talking aloud hindered their thinking , and con - sequently did not say very much . Rather than presenting an exhaustive analysis of all the protocols , we will therefore concentrate on particularly suggestive excerpts . While this type of protocol analysis has obvious limitations , it may at least provide some hints about the process of analogical problem solving , and in fact served in part to motivate sub - sequent experiments . A major issue , raised earlier , concerns the degree of mapping subjects perform in the process of generating an analogous solution . Do subjects make use of detailed correspondences between the story and the target problem , or do they focus directly on the solution embedded in the story and attempt to apply it to the target problem ? Of the 10 subjects in the Attack - Dispersion condition , 7 produced the dispersion solution without any prompt , and 3 produced it after being prompted to refer back to the story . In some respects the protocols for prompted subjects are poten - tially more informative , since what such subjects say is more likely to reflect an ongoing solution process , rather than the result of a process already completed . The protocols of 2 of the 3 prompted subjects suggested use of a solution - focusing strategy . Table 5 presents an excerpt from the protocol of one of these subjects , S15 . After the prompt to use the story , this subject clearly focuses on the solution of dividing up the army into groups and immediately generates the parallel solution to the radiation problem . There is no apparent map - 326 GICK AND HOLYOAK TABLE 5 Portion of Protocol for Sl5 ( Attack - Dispersion Condition ) Subject reads radiation problem . S : Alright I , what I most , what I\u2019d probably do is send in the ray at sufficiently high intensity and then taking the risk that the tissues , the healthy tissues that would be destroyed , could be repaired later on . Trying to relate this to the other problem , I could say that you could give multiple treatments of low - intensity ray . But from this problem it seems that they won\u2019t have effect on the tumor so . . . so I don\u2019t think that would work . Later . . . E : Okay . And as a last question can you give me a , tell me ways in which your solution would satisfy the constraints of the experiment ? S : What are the constraints of the experiment ? E : Okay , i . e . , that the healthy tissue will not be destroyed , and the tumor will be ? S : Alright , in that way my first suggestion would probably not be the way to go at it . Because that way you\u2019re getting low intensity so it won\u2019t destroy the tissue and hopefully over a period of time the additive effect of low - intensity rays would kill the tumor . But from reading the article , I don\u2019t know if that would work or not , because it says that a low - intensity ray doesn\u2019t have any effect on the tumor at all . So I don\u2019t know . I don\u2019t know any other possible ways of doing it . E : Would it help to possibly go back to the story and see whether you can apply that ? S : Well , that\u2019s what I was trying to do here . It says here he divides his army into different small groups . Okay , may . . . possibly . What they could do , but this is a whole new solution now , possibly what they could do is attack the tumor from a multiple of directions with lower intensity rays and then , since you\u2019re coming in from all different directions , the healthy , with small - intensity rays you\u2019re not going to be destroying the healthy tissue but you\u2019re , and they\u2019ll all converge at the point of the tumor which will hopefully destroy the tumor . ping between the initial problem stated in the story and the target problem . Notice also that the solution S 15 proposes prior to the prompt involves the idea of applying many low - intensity rays . After the prompt , the sub - ject produces the dispersion solution by augmenting this aspect of the earlier solution with the idea of sending rays from many angles . This pattern of gradual solution development was also evident in the protocol of another prompted subject in the Attack - Dispersion condition . In such cases it appears that the subjects were working in the appropriate direc - tion ( in the sense of Maier , 1930 ) prior to producing the complete disper - sion solution . That is , at first they seemed to have the abstract idea that the solution should involve reducing the intensity of the rays on the way ANALOGICAL PROBLEM SOLVING 327 to the tumor , but only later were able to develop a concrete solution satisfying the problem constraints . S15 appeared to use the story analogy to develop the early partial solution ; however , other subjects apparently generated an abstract partial solution independently of the analogy . The details of the problem - solving process are less evident in the pro - tocols of the seven unprompted subjects , since they expressed the solu - tion all at once . Three of these subjects simply stated the solution and alluded to the usefulness of the prior story ( saying , for example , \u201cconsid - ering the problem before\u201d ) . These subjects did not mention any specific correspondences between the story and the target problem , and hence their protocols were quite unrevealing with respect to the solution pro - cess . However , two other unprompted subjects did spontaneously mention correspondences between the problems . Immediately after reading the radiation problem , S23 stated : Like in the first problem , the impenetrable fortress , the guy had put bombs all around , and the bombs could be compared to the destruction of healthy tissue . And so they had to , they couldn\u2019t go in in mass through one road , they had to split up so as not to destroy the healthy tissue . Because if there\u2019s only a little bit of ray it doesn\u2019t damage the tissue , but it\u2019s all focused on the same spot . Table 6 presents a portion of the protocol for S28 , another unprompted subject . This subject\u2019s protocol is particularly interesting because he claimed that he generated the dispersion solution on the basis of an anal - ogy between the radiation problem and an actual problem he had read about the previous night ( using lasers to fuse the filament in a lightbulb without breaking the glass ) . This may be an unanticipated instance of the use of a \u201creal world\u201d analogy to solve a problem . However , when later questioned by the experimenter , S28 clearly was also aware of corre - spondences between the radiation problem and the prior story\u2019 . It is therefore uncertain which analogy initially triggered generation of the dispersion solution . Also note that at the beginning of this excerpt S28 remarks that the dispersion solution made it hard to think of alternative solutions , suggesting that he was aware of a set effect . No other subject explicitly mentioned such an inhibitory effect of the prior story . It is clear in the above two cases that the subjects noticed some corre - spondences involving the initial conditions and constraints of the story and target problem . However , it is difficult to tell whether these aspects of the mapping process were instrumental in generating the analogous solu - tion , or whether subjects simply mentioned the correspondences to justify the adequacy of the solution , after it had already been generated . In general it was not clear what particular correspondences were central to the solution process . However , several subjects alluded to the importance of the phrase \u201clike spokes on a wheel . \u201d Recall that the existence of 328 GICK AND HOLYOAK TABLE 6 Portion of Protocol for S28 ( Attack - Dispersion Condition ) Subject reads radiation problem and states dispersion solution . Experimenter asks for other solutions and subject suggests operating to expose tumor . S : I like my first solution so much that it\u2019s hard to come up with any others . E : Can you tell me how you arrived at your first solution ? S : To tell you the truth I was thinking about that problem last night . ( Experiment asks for clarification ) . S : I remembered reading an ad at one time on one . Some company has really expensive lightbulbs and when the filament breaks inside it\u2019s really expensive to replace the lightbulb so what they do is take lasers from all different angles and , like shooting through they don\u2019t disturb the glass but when they concentrate they fuse the filament . E : Oh , I see , very interesting . S : A . . . other than that I would , I really couldn\u2019t , don\u2019t think I could come up with other solutions . E : Okay , can you tell me whether you applied any of the hints from the previous story at all ? S : Yeah , I would say , yeah , the fortress is similar to the tumor and the army is the same as X rays or rays that destroy the tumor , and they cannot all pass through the organism , i . e . , countryside , at the same time or they blow up . multiple routes is a critical enabling condition for the solution embodied in the Attack - Dispersion story , and it has no explicit parallel in the state - ment of the radiation problem . This aspect of the story analogy may therefore serve to generate the critical insight that it is possible to send rays from multiple directions . One illustrative example is the following excerpt from the protocol of S7 in the Attack - Dispersion condition , which begins immediately after the subject had read the radiation problem : Well , I already know the answer . I knew it when I read it . I might as well stop and say that . What you do is you have a bunch of rays that are weaker , and you point them so that they concentrate on one point . So you just have many different angles . It could not only be two dimensional , the analogy of the spokes on the fortress . But you could even have it three dimensional , so you could have a whole ball of spokes going in . And you would have a high intensity right at the tumor . In addition , the protocols of all three subjects in the Open Supply Route and Tunnel conditions who produced the dispersion solution suggested that it was triggered by the idea of multiple converging routes . For exam - ple , immediately after S2 in the Open Supply Route condition read the problem , she expressed the idea of using a \u201ccircular approach\u201d ( which in her earlier story summary she explicitly related to the phrase \u201cspokes on a wheel\u201d ) . This idea then led to the multidirectional aspect of the disper - sion solution to the radiation problem . ANALOGICAL PROBLEM SOLVING 329 Two subjects in the Tunnel condition who produced the dispersion solution did so after first spontaneously remarking that the general might have sent his men down multiple roads . One other subject , in the Open Supply Route condition , also suggested the dispersion solution to the military problem , but failed to apply it to the radiation problem . The use of self - generated solutions to help solve an analogous problem will be investigated more systematically in Experiment III . Subjects\u2019 protocols also provided information about some of the dif - ficulties they encountered in attempting to apply analogies based on the Open Supply Route and Tunnel stories . As we pointed out earlier , the open passage solution is not especially practical and some subjects may have thought of this solution without mentioning it . For example , S32 in the Tunnel condition gave the open passage solution as an afterthought at the very end of the interview , and also outlined the problems with it : that the esophagus is not straight , and that there would be \u201crefraction off the esophagal walls , or absorption of the rays , \u201d which would destroy tissue . Three subjects attempted to overcome this difficulty by suggesting that a ray - proof tube through which the rays could be directed should be in - serted down the throat . In addition , the nature of the analogy suggested by the Open Supply Route story is somewhat different from that suggested by the other two stories . The solutions embodied in both of the latter stories suggest pro - cedures that can be used to generate parallel solutions to the radiation problem ( dividing the rays in the case of the Attack - Dispersion story , operating in the case of the Tunnel story ) . In contrast , the Open Supply Route story only suggests that an existing open passage might be used . The subject must then search memory to find a concrete example of such an open passage to the stomach ( e . g . , the esophagus ) . Applying the anal - ogy thus involves two steps : mapping the abstract idea of an open passage from the story to the target problem , and then thinking of a concrete example of such a passage . The difficulty of applying the analogy may account for the fact that four of the seven subjects in the Open Supply Route condition who gave the open passage solution had to be prompted to use the story . Table 7 presents a portion of the protocol for S19 in the Open Supply Route condition . This subject works through a rather detailed mapping of the correspondences between the story and the radiation problem . But while she clearly develops the abstract idea of finding an open passage , she fails in the attempt to think of a concrete example . The partial solution produced by S19 can be contrasted with the complete lack of success apparent in the protocol of S37 in the Open Supply Route condition : The only thing that is apparent to me is that the general had other information that he knew that one of the thoroughfares would be left open , and so he was able to 330 GICK AND HOLYOAK TABLE 7 Portion of Protocol for S19 ( Open Supply Route Condition ) E : It might help if you reread the instructions here . This part . ( S rereads radiation problem . ) S : Okay , so what was the first problem ? The spokes of the wheel - right ? E : Right . S : So the center fortress deal would be the idea of the tumor . That\u2019s . . . E : Okay . S : And then the spokes that blow up a little would be like the healthy tissue that blows up a little bit . And so with that one the guy had one route that was gonna do it . I guess in this one that\u2019s what you have to do is find the one route that would work . E : Okay . S : And , I think , and not use the other ways . E : Okay . What would that be ? S : That would mean we have to find one approach that was going to get to the tumor without getting the healthy tissue . And I don\u2019t see how you could do that . Cause it\u2019s not - it doesn\u2019t seem like it\u2019s the same thing . E : What doesn\u2019t seem like the same thing ? S : Well the idea that road , with a road its possible to have one road unguarded but without , in your body there\u2019s always going to be , unless the tumor was right on the outside , there would always be some tissue you would have to go through to get to it . use that . But , unless the doctor had some new information or some other treat - ment , I don\u2019t see any other applications from the first problem to the second problem . Notice that S37 appears to have mapped the story and the target prob - lem at a very abstract level of macrostructure , so that the perceived analogy ( the general had new information , so perhaps the doctor might also ) was too vague to yield a specific solution proposal for the radiation problem . In the case of the Tunnel condition four of the eight subjects who generated operation solutions received a prompt to use the story before they did so . Some subjects may have been reluctant to suggest an opera - tion solution because they were aware that it violated a constraint given in the problem statement . 4 An excerpt from the protocol of S24 , presented in Table 8 , illustrates the kind of difficulty encountered in this condition . The protocol suggests that the subject was quite carefully mapping com - ponents of the story onto components of the radiation problem . However , 4 Actually , some subjects pointed out that the problem statement is somewhat vague on this point . The statement that \u201cit is impossible to operate on the patient\u201d might be inter - preted as meaning only that it is impossible to operate and remove the tumor , rather than that an operation of any kind is impossible . However , protocols and practicality ratings indicated that most subjects who suggested operating to expose the tumor considered the proposal dubious at best . ANALOGICAL PROBLEM SOLVING 331 TABLE 8 Portion of Protocol for S24 ( Tunnel Condition ) E : If you read your instructions , it says that this story might give you some hints . . . What are you thinking ? S : Well , I remember that the main way they solved this problem was they dug under the fortress and went around it . So , possibly in this situation , you could go around the healthy tissue . I don\u2019t know how you\u2019d do that I see an analogy , it\u2019s not real clear . E : Why isn\u2019t it clear ? S : Because when I picture healthy tissue in my mind , healthy tissue all over and , you know , just like a tumor in between all this healthy tissue . But here , the mines they\u2019re on top near the surface of the ground , so they can , you can dig under those and you won\u2019t really have any problem . But here no matter where you go , like a circle around the healthy tissue . . . maybe an operation . E : Except that one of the constraints of the experiment says that you can\u2019t operate . S : Okay , that\u2019s not possible . . . maybe . . . I was thinking that maybe you could give intervals of high intensity , but I don\u2019t know , that still would probably destroy the healthy tissue . E : Can you think of anything else ? . . . Is this problem , the previous story , is that distracting ? S : ( mumbles ) Again , I\u2019m looking for an analogy between the two . And kind of set up the same problem and solve it the same way , but I don\u2019t know if I can or not . E : So , can you think of any other possibilities ? S : ( long pause ) No . the subject was unable to generate a satisfying parallel solution to the target problem . The overall impression created by the problem - solving protocols is that the generation of analogous solutions involves a conscious process of mapping correspondences between the story and the target problem . The degree of mapping required seems to vary a great deal . Sometimes mapping was done in considerable detail , particularly if the subject was having difficulty producing a parallel solution . In other cases noticing one or two major points of correspondence seemed sufficient to generate the solution . In some instances , particularly for dispersion and open passage solutions , aspects of the solution were clearly generated in sequential steps . The protocols of control subjects differed in several ways from those of subjects in the story conditions . Control subjects more often were prompted to talk . Sometimes they seemed confused and asked if there really was a solution to the problem . In addition , control subjects received fewer prompts to clarify their solutions . The latter finding raises the ques - tion of the extent to which the use of story analogies may depend on verbal or nonverbal feedback from the experimenter . Since a primary concern in Experiment I was to obtain interpretabie oral protocols , a 332 GICK AND HOLYOAK considerable amount of probing by the experimenter was essential . How - ever , it would clearly be desirable to demonstrate that people can use analogies to generate problem solutions without interacting with the ex - perimenter . Experiment II was designed to serve this purpose , and also to provide additional information about the degree of mapping required to produce a solution on the basis of a story analogy . EXPERIMENT II In order to avoid any possibility that subjects could be led to a particu - lar solution by the experimenter , Experiment II used a noninteractive procedure in which story analogies and the radiation problem were ad - ministered in booklet form . Experiment II was also designed to assess the degree of mapping required to generate a solution on the basis of an analogy . For this purpose different subjects were given one of two matched stories . While both stories embodied the same solution ( the dis - persion solution ) , they differed in the degree of correspondence that existed between their problem statements and the radiation problem . If correspondences between the problem statements play an important role in generating a parallel solution to the target problem , lessening the degree of analogy between the story problem and the radiation problem should reduce the probability that subjects will be able to use the story to pro - duce the analogous solution . Method Materials . The stories used in Experiment II were the Attack - Dispersion story ( ver - sion 2 ) and the Parade - Dispersion story ( see Appendix II ) . Table 9 presents a schematic outline of the Parade - Dispersion story , which can be compared to the outlines of the Attack - Dispersion story and radiation problem ( Table 2 ) . The Attack - Dispersion story used in Experiment II is essentially identical to that used in Experiment I . The story was rewritten slightly to match it as closely as possible with the Parade story in length and wording . Note that the setting information is nearly the same in the Attack and Parade stories . In particular , the critical enabling conditions for the dispersion solution ( centrally located fortress , multi - ple roads radiating outward ) are present in both stories . In addition , the second paragraphs of the two stories , which describe the general\u2019s dispersion solution , are identical ( except for the tinal sentences , which state the goal that has been achieved ) . However , the problem statement for the Parade story is substantially different from that for the Attack story , and it is disanalogous to the radiation problem . In the Parade story the genera1 is not trying to attack the dictator in the fortress , but rather to stage a parade that meets the dictator\u2019s specifications . The constraint of the mined roads has been removed ; and use of the entire army is required not to produce a sufftciently strong assault , but rather to produce a sufficiently impressive display . The problem statements of the Parade story and of the radiation problem are thus disanalogous in several respects . The Parade problem lacks the basic element of \u201cdesire to overcome a target object . \u201d The basis of the constraint against sending the entire army down one road ( the genera1 would lose his rank ) does not parallel the basis of the constraint against using high - intensity rays ( healthy tissue would be destroyed ) . Only the setting information and the constraint against sending the full army down one route remain analogous to features of the radiation problem . Accordingly , the mapping process ANALOGICAL PROBLEM SOLVING 333 TABLE 9 Schematic Outline of Parade - Dispersion Story Problem statement Problem setting General has army . Country has dictator . Dictator in fortress in center of country , surrounded by villages . Desired goal Problem constraints Roads radiate from fortress like spokes on a wheel . Produce impressive parade that can be seen and heard throughout entire country . Sending entire army down one road fails to produce impres - sive parade . If parade fails to impress dictator , general will lose his rank . Solution Dispersion Divide up parade so that each part of country sees part of parade . Use Resulting goal state ( 1 ) Many small groups of men ( 2 ) From different directions ( 3 ) Simultaneously Parade seen and heard simultaneously throughout country . General preserves his rank . subjects can perform in order to use the Parade story to generate the radiation dispersion solution will be limited by the lack of correspondence between the problem statements for the story and the target problem . As in the Attack story , the general in the Parade story solves his problem by dividing his troops and sending them down multiple roads to the fortress . But in the Parade story the procession of soldiers to the fortress directly constitutes achievement of the goal state , whereas in the Attack story the movement of troops is simply the means by which the final goal ( capture of the fortress ) is achieved . Furthermore , in the Parade story the fact that the troops converge on the fortress is a more or less incidental aspect of the solution procedure , whereas in the Attack story this aspect of the solution is critical . Thus even though the surface description of the solution is the same in both stories , the solution contexts differ . Such contextual differences may influence the precise interpretation subjects give to the solution . In order to assess whether the Parade and Attack stories in fact suggest similar solutions to the radiation problem , two independent groups of 17 undergraduates rated the stories as prompts for various radiation solutions . ( The Parade story was simply included as an addi - tional condition in the story rating task discussed under Results of Experiment I . ) The six solutions that were rated are listed in Table 4 . While there was a slight trend for the Attack story to be rated higher than the Parade story as a prompt for the dispersion solution , the two stories did not produce significantly different ratings , either as a main effect , F < 1 , or as an interaction , F ( . 5 , 160 ) = 1 . 32 , p > . 25 . These rating results thus confirm that the two stories , despite the differences in their problem statements , suggest essentially the same solution to the radiation problem . Experiment II was designed to determine whether subjects actually solving the radiation problem would be hindered by a disanalogous problem statement . Procedure and subjects . Subjects were divided into three conditions , receiving either the Attack story , the Parade story , or no story ( control condition ) prior to working on 334 GICK AND HOLYOAK the radiation problem . The test booklet consisted of instructions , story analogy ( for experi - mental conditions ) , radiation problem , a solution sheet , and a final questionnaire , each on a separate page . Experimental subjects first read the story and then wrote a brief summary of it , referring back to the story if they wished . All subjects then attempted to solve the radiation problem . Subjects in the experimental conditions were told that the first story might give them some hints for solving the test problem , although it was not necessary to use the prior story to solve the problem . They were allowed to refer back to the story at any time . All subjects were instructed to write as many solutions as possible in the order they came to mind . They were asked to write down every idea they considered , even those later rejected . Subjects were told to make their proposals as explicit as possible , but not to be concerned with technical medical considerations . The final questionnaire asked subjects to indicate how helpful the story problem was in solving the radiation problem ( \u201cnot helpful , \u201d \u201csomewhat helpful , \u201d or \u201cvery helpful\u201d ) , and in what way it was helpful . Subjects were also asked if they had known the solution to the radiation problem prior to the experiment . Subjects were 143 undergraduates tested in five introductory psychology classes . Forty - seven subjects served in the Attack - Dispersion condition , 46 in the Parade - Dispersion con - dition , and 50 in the control condition . Results and Discussion The data for one subject in the Parade - Dispersion condition who was familiar with the radiation problem and dispersion solution were dis - carded . The remaining subjects\u2019 solution proposals were analyzed for the presence of various solution types , particularly the dispersion solution . This analysis was done by two independent scorers , each blind to the conditions in which subjects served . Disagreements were resolved by discussion . Unlike subjects in Experiment I , subjects in Experiment II were not prompted to fully explicate their solutions . As a result , a number of subjects produced incomplete versions of the dispersion solution . In order to be scored as a complete dispersion solution , three features had to be present in the proposal : ( 1 ) the rays are applied to the tumor from different directions , ( 2 ) at low intensity , and ( 3 ) simultaneously . A partial solution had to contain at least the first feature , the critical element of dispersion . However , partial solutions might omit features 2 and / or 3 . In addition , a partial solution might include reference to some degree of damage to healthy tissue . Presence of the latter feature was taken as an indication that the subject did not entirely understand the implications of the dispersion solution , since the problem statement specified that low - intensity rays are harmless . The percentage of subjects producing complete or partial dispersion solutions differed substantially across the three conditions , as shown in Table 10 . Dispersion solutions were produced by 76 % of the subjects in the Attack - Dispersion condition , 49 % of the subjects in the Parade - Dispersion condition , and only 8 % of the subjects in the control condition , G2 ( 2 ) = 53 . 1 , p < . OOl . The Attack story produced significantly more dispersion solutions than did the Parade story , G2 ( 2 ) = 7 . 70 , p < . Ol , while the two story conditions together produced significantly more dis - ANALOGICAL PROBLEM SOLVING 335 TABLE 10 Percentage of Subjects in Each Condition of Experiment II Who Proposed the Dispersion Solution to the Radiation Problem Dispersion solution Condition Complete Partial Total N Attack - Dispersion story Parade - Dispersion story Control 57 19 76 47 31 18 49 45 8 0 8 50 persion solutions than did the control condition , GZ ( l ) = 45 . 5 , p < . OOl . The Attack story also prompted more complete dispersion solutions than did the Parade story . The four subjects in the control condition who spontaneously generated dispersion solutions gave complete solutions . Collapsing over all conditions , dispersion solutions tended to be produced relatively early in the sequence of proposals given by a subject ( i . e . , prior to the proposal of median rank , p < . 025 by a sign test ) . Table 11 presents the percentage of subjects in the two story conditions who rated the story as \u201cnot helpful , \u201d \u201csomewhat helpful , \u201d or \u201cvery helpful , \u201d as a function of whether the subject produced a complete dis - persion solution , partial solution , or no dispersion solution . The results obtained with such post - hoc questionnaires must be interpreted with cau - tion , since they may reflect hindsight rather than accurate memories of the solution process . Nevertheless , these rating results are at least a source of converging evidence . Most subjects who produced a dispersion solution rated the story as \u201cvery helpful . \u201d In contrast , the majority of those subjects in both conditions who failed to produce a dispersion solu - tion rated the story as \u201cnot helpful at all . \u201d There were no reliable differ - ences between the rating patterns produced by subjects in the Attack versus Parade conditions . The total number of proposed solutions ( other than the dispersion solu - tion ) was tabulated for subjects in each of the three conditions . The mean number of proposals was 1 . 17 for the Attack condition , 1 . 40 for the Parade condition , and 2 . 12 for the control condition , F ( 2 , 139 ) = 12 . 7 , p < . OOl . A Newman - Keuls test indicated that only the difference between the control condition and the two story conditions was significant , p < . Ol . This decline in the number of alternative proposals given by subjects prompted to generate the dispersion solution replicated the comparable trend obtained in Experiment I . As in Experiment I , the frequency of a specific class of alternative solutions ( decreasing the sensitivity of the healthy tissue to the rays , or T AB L E 11 P e r c en t age o f S ub j e c t s G i v i ng E a c h H e l p f u l ne ss B a t i ng a s a F un c t i on o f S t o r y C ond i t i on and T y pe o f D i p s e r s i on S o l u t i on G i v en 0 T y pe o f d i s pe r s i on s o l u t i on g i v en 2 C o m p l e t e d i s pe r s i on P a r t i a l d i s pe r s i on N o d i s p e r s i on > s o l u t i on s o l u t i on s o l u t i on 2 N o t S o m e w ha t V e r y N o t S o m e w ha t V e r y N o t S o m e w ha t V e r y ! 3 C ond i t i on he l p f u l he l p f u l he l p f u l he l p f u l he l p f u l he l p f u l he l p f u l he l p f u l he l p f u l 2 A tt a ck - D i s pe r s i on 0 22 78 11 22 67 64 27 0 9 s t o r y k P a r ade - D i s pe r s i on 0 29 71 0 38 63 57 39 4 s t o r y ANALOGICAL PROBLEM SOLVING 337 somehow protecting the tissue ) was tabulated for each condition . Whereas 28 % of the subjects in the control condition produced such solu - tions , only 2 % of those in the Attack condition and 16 % of those in the Parade condition did so , G2 ( 2 ) = 14 . 6 , p < . Ol . Subjects in the Parade condition produced more solutions of this type than did subjects in the Attack condition , G2 ( 1 ) = 5 . 78 , p < . 02 . This result is a significant repli - cation of a trend obtained in Experiment I : the more effective the story analogy is in prompting the analogous solution , the more it inhibits pro - duction of . alternative , disanalogous proposals . The basic results of Experiment II are thus extremely clear . First , subjects can readily use story analogies to guide their attempts to solve the radiation problem , even without feedback from the experimenter . Second , the effectiveness of analogies in prompting a specific solution is a matter of degree . In particular , a story with a problem statement analo - gous to that of the radiation problem ( the Attack story ) was more likely to trigger the dispersion solution than was a story with a problem statement less related to that of the radiation problem ( the Parade story ) . This was true even though both stories embodied similar setting information and solution statements . EXPERIMENT III Experiments I and II demonstrated that subjects can use a story anal - ogy , describing a problem and its solution , to guide generation of an analogous solution to the radiation problem . A natural question is whether subjects could also use their own solutions to the initial story problem to help them solve the target problem . You may recall that Experiment I provided some evidence for this possibility . Three subjects who received stories other than the Attack - Dispersion story spontaneously suggested that the general might have divided his troops and sent them down multi - ple roads ; two of these subjects then went on to produce the dispersion solution to the radiation problem . Accordingly , in Experiment III subjects were first given just the problem statement from the Attack - Dispersion story , and asked to suggest what the general might do to capture the fortress . Subjects were then asked to solve the radiation problem , using their solutions to the initial problem as hints . Method The problems were administered in booklet form as in Experiment II . The instructions stated that the subject would have to solve \u201ctwo verbal problems requiring some creativity for their solutions . \u201d The first problem consisted of the first paragraph of the Attack - Dispersion story ( version 2 , Appendix II ) , followed by the question \u201cWhat could the general do in order to capture the fortress . 7\u201d Subjects were to provide as many possible solutions as they could think of . Following this , subjects proceeded to give solutions to the radiation problem . The instructions stated , \u201c , . . you may find that the first problem that you solved 338 GICK AND HOLYOAK gives you some hints for solving the second problem , so you should try and use it if you can . . . However , it is not necessary to use the first problem in order to solve the second . \u201d Subjects were told that they could look back to the first problem and their solutions to it at any time . On the final page of the booklet subjects were asked to indicate to what extent the first story problem helped in solving the ray problem ( \u201cnot at all , \u201d \u201csomewhat , \u201d \u201cvery much\u201d ) , and in what way it was helpful . They were also asked to indicate if they had known the solution to the radiation problem prior to the experiment . The experiment was administered to 46 students in two introductory psychology classes . Results and Discussion Data from one subject who indicated prior knowledge of the dispersion solution to the radiation problem were discarded . The frequencies of vari - ous solutions to the two problems were tabulated for the remaining 45 subjects by two independent scorers ; disagreements were resolved by discussion . As might be expected given the extremely loose constraints imposed by the problem statement , the proposed solutions to the military problem were quite varied . Many of these ( e . g . , airlifting troops into the fortress by balloon ) had no apparent correspondence to any potential solution to the radiation problem . However , 22 of the 45 subjects ( 49 % ) produced the gist of the dispersion solution : the general should divide the troops into small groups and send them down different roads . The major question of interest was whether those subjects who pro - duced the dispersion solution to the military problem would be especially likely to then produce the dispersion solution to the radiation problem . Of the 22 subjects who produced the dispersion solution to the military problem , 9 ( 41 % ) subsequently produced either complete or partial dis - persion solutions to the radiation problem . Five of these subjects pro - duced a complete solution . In contrast , only 3 of the remaining 23 subjects ( 13 % ) produced this solution G2 ( 1 ) = 4 . 61 , p < . 05 . Two of these subjects produced a complete solution . The value of 41 % is significantly higher than the 8 % of subjects who produced the dispersion solution to the radiation problem without any prior story problem ( Control condition of Experiment II ) , G2 ( 1 ) = 10 . 4 , p < . Ol . The above results are correlational in nature , since subjects who ar - rived at the dispersion solution to the military problem did so of their own accord - no experimental manipulation determined the subjects that would produce the critical solution . It could therefore be argued that subjects who produced dispersion solutions to both problems did so be - cause of some general factor related to problem - solving skills or strategies . However , additional evidence suggests that self - generated so - lutions to the military problem had a causal influence on generation of the radiation dispersion solution . First , dispersion solutions were produced significantly more frequently in Experiment III ( regardless of whether a parallel solution was produced for the military problem ) than in the con - ANALOGICAL PROBLEM SOLVING 339 trol condition of Experiment II . Second , the questionnaire results also reflected the influence of the solutions produced for the military problem . Of the nine subjects who generated the dispersion solution to both prob - lems , all but one indicated that the military story was \u201csomewhat\u201d or \u201cvery\u201d helpful in solving the radiation problem ( one other subject did not respond to this question ) . In contrast , a majority ( 58 % ) of the remaining subjects indicated that the story was not helpful at all . Interestingly , two of the three subjects who produced the radiation dispersion solution with - out first producing the military dispersion solution indicated that the story was somewhat or very helpful ( the third subject failed to answer the question ) . All three of these subjects had suggested that the general could send small groups in succession to the fortress . This solution may have triggered the idea of dividing forces that is critical to the radiation disper - sion solution . While self - generated solutions to the simpler military problem thus facilitated discovery of the radiation dispersion solution , the degree of transfer was less than perfect . The value of 41 % radiation dispersion solutions , based on a self - generated military dispersion solution , is sig - nificantly less than the 76 % radiation dispersion solutions produced by subjects given a military dispersion solution written by the experimenter ( Attack - Dispersion condition of Experiment II ) , G2 ( 1 ) = 8 . 25 , p < . Ol . There are a number of possible explanations for this difference . Subjects in Experiment III almost always proposed more than one solution for the military problem ( a mean of 3 . 04 ) . Consequently , their own dispersion solution , if they produced it , was usually embedded in other \u201cdistractor\u201d solutions . Subjects who did not systematically consider each of their pro - posed military solutions with respect to the radiation problem may there - fore have sometimes missed the critical analogy . Indeed , some subjects may have simply failed to apply their prior solutions at all when working on the radiation problem . In addition , it is possible that the self - generated dispersion solutions did not always entirely capture the analogy to the corresponding radiation solution . All of the self - generated dispersion solutions expressed the basic idea of sending small groups down multiple roads . However , in some versions the element of simultaneity of attack was absent or not clearly expressed . In other cases subjects suggested that the small groups would meet and regroup near the fortress prior to initiating an attack . This added feature of delay is disanalogous to the radiation dispersion solution ( since low - intensity rays cannot \u201cwait and meet up\u201d near the tumor prior to striking it ) . Of the 22 self - generated dispersion solutions , 10 either lacked a clear expression of simultaneity or added the element of delay . How - ever , since 4 of these 10 subjects succeeded in generating the radiation dispersion solution , we were unable to find any effect of quality of the 340 GICK AND HOLYOAK military dispersion solution on subsequent transfer to the radiation prob - lem . Nonetheless , it remains possible that in various ways the subjects\u2019 dispersion solutions ( which were often written in a cryptic fashion ) were incomplete in comparison with our own version of the solution . At any rate , it is clear that further research will be required to assess how problem - solving activity per se affects subsequent transfer of a solution to an analogous problem . In addition to the dispersion solution , another frequently proposed so - lution to the military problem was to have the general send successive small groups of men down a single road . This \u201csuccessive groups\u201d solu - tion appeared to trigger an analogous solution to the radiation problem . We defined proposed solutions to the latter problem that involved spreading out ray applications over time ( regardless of the intensity of the rays or whether treatment was intermittent or continuous ) as \u201caccumula - tion\u201d solutions . Of 24 subjects who gave the successive groups solution to the military problem , 10 ( 42 % ) suggested an accumulation solution to the radiation problem . In contrast , only 3 of the remaining 21 subjects ( 14 % ) suggested an accumulation solution to the radiation problem , G2 ( l ) = 4 . 28 , p < . 05 . Furthermore , 2 of the latter 3 subjects had produced the dispersion solution to the military problem . Their accumulation solutions may therefore have been triggered by a relatively abstract analogy be - tween \u201cdispersion over space\u201d and \u201cdispersion over time . \u201d This possi - bility is supported by the fact that an accumulation solution was rated the second most likely ( after the dispersion solution ) to be prompted by the Attack - Dispersion story ( see Table 4 ) . This tendency for subjects to generate analogous accumulation solu - tions is particularly intersting because such proposals are clearly ineffec - tive solutions . Simply distributing ray treatments over time will not over - come the basic problem that the rays will have equal effects on both the tumor and the healthy tissue . Since the instructions to subjects stated that they should write down any proposal they could think of , even if they later rejected it , we do not know whether subjects who gave an accumu - lation solution actually believed it would work . However , this result at least raises the possibility that a \u201cfalse analogy\u201d between problems ( sim - ilar to the Tunnel condition in Experiment I ) may foster errors in evaluat - ing a parallel solution to a transfer problem . EXPERIMENT IV Our central concern in the experiments reported so far was to deter - mine if people can use an analogy to generate a solution to a target prob - lem , and to investigate how analogical problem solving proceeds . Con - sequently , we simplified the subjects\u2019 task in several important ways . First , subjects were always allowed to reread the story analogy at any ANALOGICAL PROBLEM SOLVING 341 time , so that their performance would not be limited by memory factors . Second , the story was always presented alone , so that subjects would have no problem identifying the relevant analogy . Third , subjects were always explicitly told to try to use the story as an aid in solving the target problem . This hint was quite nonspecific ; at no time were subjects told the nature of the analogous relationship between the story and problem . Nevertheless , the hint eliminated the need for subjects to spontaneously notice the possible analogy . In many cases of everyday problem solving in which an analogy could help , the person would have to spontaneously notice the correspondence between the target problem and some analogous problem , either of which might be stored in memory . The two experiments reported below begin to investigate the effect of such additional processing requirements on analogical problem solving . Method In Experiment IV subjects first memorized the Attack - Dispersion story ( version 3 , Ap - pendix HIS ) in the guise of a story recall experiment , and then went on to work on the radiation problem . In addition , subjects also first memorized two additional \u201cdistractor\u201d stories written by the experimenters ( \u201cThe Wine Merchants\u201d and \u201cThe Identical Twins\u201d ; see Appendix IV ) . These additional stories were matched closely in length with the Attack - Dispersion story , and also describe problems and their solutions . However , the two dis - tractor stories were intended to be as disanalogous to the radiation problem as possible . The experiment was administered in booklet form to small groups of subjects . The initial instructions to all subjects stated that the experiment would have two parts , first a story recall task and then a problem - solving task . For the story recall task , subjects first received one story , and were given 3 min to study it . The stories were then collected , and subjects had up to 15 min to recall the story . They were asked to recall the story in as close to its original form as possible , but to give the gist of it even if they couldn\u2019t remember the exact wording . All subjects completed their recall attempts within 15 min . When all subjects in a group were finished , the next story was distributed and the study - recall procedure was repeated . The stories were always administered in the order \u201cThe Wine Merchants , \u201d the Attack Dispersion story ( entitled \u201cThe General\u201d for subjects ) , and \u201cThe Identical Twins . \u201d The critical story analogy was placed in the middle serial position in order to maximize the difficulty of later using the memorized story to solve the radiation problem . Following the story recall tasks subjects were given a short ( 3 to 5 min ) break , and then asked to write solutions to the radiation problem . However , subjects were divided into two conditions that received slightly different instructions . For subjects in the \u201cHint\u201d condition , the instructions on solving the radiation problem included the following sentence : \u201cIn solv - ing this problem you may find that one of the stories you read before will give you a hint for a solution of this problem . \u201d For subjects in the \u201cNo Hint\u201d condition , this sentence was deleted from the instructions . The instructions used in the Hint condition were thus compa - rable to those used in the earlier experiments , whereas the instructions used in the No Hint condition for the first time did not call attention to the potential relevance of the stories to the target problem . 5 The wording of the story was modified slightly in this version to match it with other story analogies we intended to use in a larger experimental design . For various reasons these additional conditions were never actually tested . 342 GICK AND HOLYOAK After completing their solution attempts subjects were given a final questionnaire . Sub - jects in the No Hint condition were asked whether it occurred to them \u201cto try and use any of the stories from the first experiment to help solve the ray problem\u201d ; and if so , how helpful each of the three stories was ( \u201cnot at all , \u201d \u201csomewhat , \u201d \u201cvery\u201d ) , and in what way . Sub - jects in the Hint condition answered only the latter part of the above question . All subjects were also asked if they had known the solution to the ray problem prior to the experiment . Twenty - seven undergraduates from the Human Performance Center paid subject pool served as paid subjects . Twelve subjects were tested in the Hint condition and fifteen in the No Hint condition . Results and Discussion Two independent scorers were entirely in agreement in identifying dis - persion solutions to the radiation problem . For the Hint condition , 11 out of 12 subjects ( 92 % ) produced the complete dispersion solution . This percentage is , of course , far higher than would be expected for subjects who did not receive the Attack - Dispersion story ( under 10 % in both Ex - periments I and II ) . In addition , 10 of the 12 subjects indicated that the Attack - Dispersion story was \u201cvery helpful , \u201d one indicated that it was \u201csomewhat helpful , \u201d and one did not answer the question . In contrast , 50 % of the subjects indicated that \u201cThe Wine Merchants\u201d was \u201cnot helpful at all , \u201d and 91 % ( 10 of 11 who answered the question ) indicated that \u201cThe Identical Twins\u201d was \u201cnot helpful at all . \u201d Subjects clearly had no serious difficulty in identifying the critical story analogy in memory and applying it to generate the dispersion solution to the radiation problem . However , this picture changes dramatically when the results for the No Hint condition are examined . Whereas 92 % of the subjects in the Hint condition produced the dispersion solution , only 20 % ( 3 out of 15 ) of those in the No Hint condition did so , G2 ( 1 ) = 15 . 5 , ~ < . OOl . Furthermore , 2 of these 3 subjects gave only partial solutions ( as defined in Experiment II ) , and indicated that they did not consider using the stories . It is therefore possible , and in fact rather likely , that only 1 of the 15 subjects spontane - ously noticed the critical analogy and successfully applied it to produce the dispersion solution . In response to the questionnaire , 12 of the 15 No Hint subjects indi - cated that it had not occurred to them to use the stories . The 3 who said they did use the stories all indicated that the Attack - Dispersion story was \u201cvery helpful\u201d ; however , only one of these subjects actually produced the dispersion solution . The written comments of the other two subjects who said they used the Attack - Dispersion story suggested they were at most sensitive to some very vague , abstract analogy with the radiation problem . For example , one wrote that the story \u201cshowed a unique way to attack a problem using methods that were not very obvious or usual . \u201d ( Similar comments were given by subjects who rated the Wine Merchants ANALOGICAL PROBLEM SOLVING 343 story as somewhat or very helpful . ) It therefore seems that no more than 3 subjects , and perhaps only 1 , actually noticed the critical analogy without a hint from the experimenter . Since subjects were randomly assigned to the two conditions , degree of memory for the critical story should have been equalized across the two conditions . To confirm this , the protocols for the Attack - Dispersion story were scored for gist recall . For this purpose the story was divided into 43 propositions ( see Appendix III ) . This propositional division was made using the procedure outlined by Thorndyke ( 1977 ) , with the addition that adjectives and prepositional phrases that seemed intuitively important to the story were counted as separate propositions . As anticipated , the mean number of propositions recalled did not differ significantly between the Hint and No Hint groups ( 32 . 08 versus 33 . 53 ) , f < 1 . While we did not analyze the recall in detail , a few aspects are worth noting . First , performance appeared quite good in an absolute sense , which is consistent with the fact that subjects were almost invariably successful in using the story to generate a solution to the radiation prob - lem ( as long as they were told to try ) . Second , as many investigators of story recall have reported , some propositions were much more memora - ble than others . Three propositions were recalled by all 27 subjects . These were proposition 1 ( \u201cA small country was ruled by a king\u201d ) and proposi - tions 5 and 6 ( \u201cMany roads radiated outward from the fortress like spokes on a wheel\u201d ) . As we argued earlier , the latter pair of propositions estab - lish a critical enabling condition for the general\u2019s solution to his problem . This high level of recall obtained for setting information causally related to the solution is consistent with the problem - based model of story recall proposed by Black and Bower ( in press ) . EXPERIMENT V The results of Experiment IV demonstrated that subjects can identify a relevant story analogy encoded into memory in the context of distractor stories , and can use the analogy to generate a solution to a subsequent target problem . However , when the experimental instructions did not provide a hint that the stories might help to solve the target problem , subjects seldom noticed or used the analogy . This suggests that the pro - cess of analogical problem solving is neither automatic nor invariably applied by college students as a conscious strategy . The knowledge ac - quired in the context of the story recall phase of the experiment seemed to be encapsulated in such a way that its pertinence to the problem - solving task was not recognized . An important question is whether this type of encapsulation of experi - ence is more or less absolute , or whether there are factors that would make a relevant analogy more likely to be noticed even though it was 344 GICK AND HOLYOAK initially encoded in a recall context . Experiment V modified the design of Experiment IV in order to examine two such possible factors . First , the total memory load was reduced by eliminating the two distractor stories from the recall phase ; and second , in one condition the story analogy was presented after subjects had read and begun to work on the radiation problem . The latter condition can be viewed as an experimental analog of a situation in which a person \u201cstumbles upon\u201d relevant information in the course of working on a problem , as is often reported in anecdotes de - scribing the \u201cEureka\u201d experiences of creative thinkers . Method The experiment was administered in booklet form as in Experiment IV . Subjects were assigned to one of three conditions . The initial procedure for the Story First condition was identical to that for the No Hint condition of Experiment IV , except that the recall task involved only the critical Attack - Dispersion story . After working on the radiation problem subjects were asked to indicate whether it had occurred to them to use the story from the first experiment to help solve the problem . If they responded \u201cno , \u201d they were then asked to try and use the story to generate additional solutions to the radiation problem . The proce - dure thus involved three steps : ( 1 ) reading and recall of the story ; ( 2 ) an attempt to solve the radiation problem without a hint to use the story ; and ( 3 ) a final attempt to solve it with a hint . The Story Second condition was presented in the guise of an \u201cincubation\u201d experiment . Subjects were told they would be given a problem to work on , then would be interrupted to perform a different task ( story recall ) , and then would again work on the problem . The actual procedure involved four steps : ( 1 ) an initial IO - min attempt to solve the radiation problem ; ( 2 ) reading and recall of the Attack - Dispersion story ; ( 3 ) a second attempt at the radiation problem , without any hint to use the story ; and ( 4 ) a final attempt to solve the radiation problem after the hint was given . This procedure thus differed from that for the Story First condition only by the addition of step 1 , the initial attempt to solve the problem . If subjects in the Story Second condition were to produce more dispersion solutions than subjects in the Story First condition immediately after recalling the story , this would suggest that initial exposure to the problem makes it more likely that the person will notice a relevant analogy . However , one might argue that such a result could be interpreted as a real \u201cincu - bation effect . \u201d That is , regardless of the nature of the story presented during the intervening recall task , perhaps simply taking a break from the problem would be sufficient to increase the probability of generating the dispersion solution . To control for this possibility , subjects in the Incubation Control condition received exactly the same procedure as did those in the Story Second condition , except that their recall task used one of the distractor stories from Experiment IV ( \u201cThe Identical Twins\u201d ) . Accordingly , any tendency for the Incubation Control condition to produce more dispersion solutions after presentation of the story would presumably be due to a beneficial effect of incubation per se , rather than to use an analogy . Forty - seven undergraduates from the Human Performance Center subject pool served as paid subjects . Seventeen subjects were assigned to the Story First condition , 20 to the Story Second condition , and 10 to the Incubation Control condition . Results and Discussion Table 12 presents the percentage of subjects in each condition who produced the dispersion solution during the various steps of the proce - ANALOGICAL PROBLEM SOLVING 345 TABLE 12 Percentage of Subjects in Experiment V Who Produced Dispersion Solution at Each Step of the Procedure Condition Before story After story ( no hint ) After story ( with hint ) Never N Story First - 41 35 24 17 Story Second 10 35 30 25 20 Incubation Control 10 0 0 90 10 dure . There was no evidence that the manipulation of presenting the problem prior to the story analogy ( Story Second condition ) increased the probability that subjects would notice or use the analogy . In the Story First condition , 41 % of the subjects gave the dispersion solution on their first attempt following recall of the story ; while in the Story Second con - dition , 35 % of the subjects produced this solution immediately after reading the story . One subject in the Story First condition gave a partial solution ; all other solutions were complete . For the Story First condition we cannot clearly separate subjects who used the story to produce the solution from those who may have produced it spontaneously ( as did those subjects in the Story Second condition who gave the dispersion solution prior to seeing the story ) . However , of the seven subjects in the Story First group who gave the dispersion solution immediately after story recall , six reported that they used the story to help solve the prob - lem . If we accept these reports at face value , it appears that the per - centages of subjects in the two conditions who spontaneously noticed and used the analogy were identical ( 35 % in both conditions ) . The percentages of all subjects who reported that it occurred to them to use the story were also similar across the two conditions ( 47 % in the Story First condition , 40 % in the Story Second condition ) . It is clear that in both conditions not all of the subjects who eventually proved able to use the story to generate the dispersion solution did so spontaneously . Collapsing over the Story First and Story Second condi - tions , 43 % of the subjects generated the dispersion solution prior to re - ceiving the hint , while a total of 76 % eventually succeeded in producing the critical solution . Note that there was a trend toward a higher percent - age of subjects generating the solution without a hint in Experiment V ( 43 % ) , where no distractor stories were used , than in Experiment IV ( 20 % ) , where two distractor stories were included in the recall task . How - ever , this trend was not significant , G2 ( 1 ) = 2 . 64 , p = . lO . It should be noted that the above comparison confounds number of distractor stories with serial position of the critical story ( which was always presented in the middle position in Experiment IV ) . It is possible that distractor stories 346 GICK AND HOLYOAK would have a less detrimental effect on the likelihood of subjects noticing the analogy if the critical story were presented last , just prior to the problem - solving task . As in all previous experiments , the story analogy clearly played a crit - ical role in generating the dispersion solution . Whereas 76 % of the sub - jects in the two experimental conditions eventually produced the disper - sion solution , only 10 % ( one subject ) produced it in the Incubation Con - trol condition , G2 ( 1 ) = 15 . 0 , p < . OOl . Since the one successful subject in the latter group produced the target solution prior to receiving the story , there was not the slightest suggestion that simply taking a break from the problem was sufficient to stimulate discovery of the dispersion solution . GENERAL DISCUSSION The present study provides an experimental demonstration that a solu - tion to a problem can be developed by using an analogous problem from a very different domain . Our results substantiate anecdotal descriptions of the role that analogical thinking may play in creative problem solving , and at the same time provide some information about the mental processes involved in analogical problem solving . The results of Experiments I and II indicated that there is considerable variation in the degree of mapping required to generate an analogous solution . In particular , the intermediate frequency of dispersion solutions produced in Experiment II by the Parade story , which was only partially analogous to the radiation prob - lem , supports two important conclusions about the mapping process in - volved in analogical problem solving . First , subjects in the Parade condi - tion were much more likely to generate dispersion solutions than were control subjects . Thus it seems that subjects can often generate an analo - gous solution even though a complete mapping between aspects of the prior story and the target problem is impossible . In such cases it seems that a solution - focusing strategy may be sufficient to produce the parallel solution . Second , the Parade story was not as effective as the more com - pletely analogous Attack story in prompting the dispersion solution . This suggests that subjects can also perform a more detailed mapping between the problem statements of the story and of the target problem , and that these additional correspondences are sometimes critical in determining whether the subject arrives at the analogous solution . However , the types of correspondences between the two problem statements that are most critical in developing a solution are not entirely clear . Numerous subjects in our experiments commented on the impor - tance of the reference in the story to roads radiating outward \u201clike spokes on a wheel . \u201d Intuitively , this phrase seems to elicit a spatial image that represents those essential aspects of the dispersion solution that can be applied to both the military and the medical problems . Even though the ANALOGICAL PROBLEM SOLVING 347 stories and the target problem were always presented verbally in our experiments , the problems essentially describe spatial relationships . Our use of a propositional representation to describe the correspondences between the stories and the radiation problem does not preclude the pos - sibility that some form of analog representation plays an important role in the mapping process . For example , the mapping process may in part depend on interpretive procedures that are applied to a mediating spatial image . Further research is needed to explore the role of spatial repre - sentation in analogical problem solving . It is clear that our understanding of the use of analogies in problem solving remains severely limited in many important respects . We certainly need to be cautious in generalizing on the basis of the present study , which used only one target problem and a very limited set of story analogies . While it seems reasonable to expect that comparable results would be obtained with other ill - defined \u201cinsight\u201d problems , for which a solution hinges on a small number of critical inferences , this remains to be demonstrated . It is still less clear whether analogies can be used in a similar fashion to help solve more \u201ccomputational\u201d problems , for which the solution con - sists of a series of discrete steps . Reed et al . ( 1974 ) were unable to demon - strate positive transfer between two homomorphic \u201criver crossing\u201d problems , except when the correspondences between the arguments of the two problems were described to subjects . In addition , most subjects in the Reed et al . study reported making little or no use of the first problem when solving the second . It is possible that the mapping process required in such multimove problems places excessive demands on memory ca - pacity . However , various procedural differences make it difficult to di - rectly compare the Reed et al . results to those obtained in the present study . For example , subjects in the Reed et al . study solved two succes - sive problems , while in our experiments the solution to the first problem was described in the context of a story ( except in Experiment III , in which the probability of transfering the analogous solution was somewhat re - duced when subjects solved the first problem themselves ) . In addition , it is possible that people are able to use analogies more easily in solving some computational problems than in solving others . For example , Hayes and Simon ( 1977 ) have demonstrated positive transfer between isomor - phic versions of the Tower of Hanoi puzzle , another computational prob - lem . Clearly much remains to be learned about the influence of problem characteristics on problem solving by analogy . In addition to investigating the effects of problem type , we need to learn more about the ways in which the use of analogies may interact with other strategies ( e . g . , means - ends analysis ) used in problem solving . 348 GICK AND HOLYOAK Noticing and Accessing Potential Analogies A number of important questions for future research involve the closely related issues of the spontaneous noticing of analogies , and the accessing of potential analogies stored in memory . The results of Experiments IV and V suggest that one of the major blocks to successful use of an analogy may be failure to spontaneously notice its pertinence to the target prob - lem . When subjects were not told to try to use the prior stories to help solve the radiation problem , only a minority succeeded in generating the analogous solution . This decline in transfer performance cannot be attrib - uted to faulty encoding of the story analogy , since most subjects were able to produce the analogous solution once they were given a hint to apply the story . Also , the problem of spontaneous noticing was not limited to stories previously encoded into memory . In the Story Second condition of Experiment V , many subjects failed to notice the relevance of the story even though they had to read , memorize , and recall it after beginning to work on the target problem . However , even when subjects are not given an explicit hint to use a story analogy to solve a problem , the analogy itself can be viewed as a hint about a possible solution to the target problem . Previous work on the general topic of hints in problem solving , most notably that of Maier ( 1930 , 193 l ) , has used hints that involve objects that are incorporated into the solution directly , rather than analogically . It is therefore difftcult to compare the present results with earlier research on hints . For example , in one investigation of the use of hints in the \u201ctwo - string\u201d problem , the experimenter \u201caccidentally\u201d brushed against one string and set it in mo - tion ( Maier , 1931 ) . This hint was very effective in eliciting the \u201cpen - dulum\u201d solution ( attaching objects as weights and setting the strings in motion ) ; yet subjects rarely reported being aware of using the hint , unless they gave the solution in separate stages , rather than all at once . This experiment is most comparable to the conditions in Experiments IV and V in which an analogy was made available to subjects in the guise of a recall task , but subjects were not told that the story analogy was a hint to help solve the radiation problem . In apparent contrast to Maier\u2019s results , most subjects in these conditions who generated the dispersion solution re - ported using the story to do so . Thus it seems that our subjects were usu - ally aware of the usefulness of the prior story , although the precise time at which they noticed its relevance is unclear from the data ( since these were post - hoc reports ) . Why should subjects so often fail to notice the relevance of a story analogy to a target problem when a hint to use the story is not provided ? One might argue that this result is not particularly surprising , since the ANALOGICAL PROBLEM SOLVING 349 story was presented in a different experimental context ( a story recall experiment ) . The difficulty of the recall context may be related to the problem of identifying the optimal level of abstraction for representing an analogy , as we discussed in the Introduction . A recall task , with its em - phasis on memory for specific wording , may lead the person to represent the story at a level of macrostructure too detailed to maximize its analogi - cal correspondence with the target problem . A hint to use the story may lead the person to derive a more abstract level of macrostructure , better suited for the problem - solving task . But in any case , the issue of how analogies are noticed is a very general one . A potential analogy may often be encoded in a very different context from that in which the target problem appears . Indeed , the basic problem in using an analogy between remote domains is to connect two bodies of information from disparate semantic contexts . More generally , successful transfer of learning generally involves overcoming contextual barriers . This may not be easy ; for example , it is all too common for a student to fail to notice the relevance of knowledge acquired in one class to a problem encountered in another . The problem of how analogies are noticed is closely related to the issue of how analogies are accessed in memory . Noticing that information in memory is relevant to a target problem is part of the process of retrieving an analogy . These problems were side - stepped in Experiments I - III , since subjects received a hint to use the story analogies and were allowed to reread them at any time . The problem of memory access was greatest in Experiment IV , in which the relevant story analogy was memorized in the context of two irrelevant distractor stories . Subjects in this experiment seemed to have little difficulty in identifying the appropriate story in memory , and applying it to the target problem , as long as they were instructed to do so . However , subjects may have performed this task by simply testing each of the three stories to see if it suggested a solution to the target problem . Such a strategy would presumably be impractical in most everyday problem - solving situations , where virtually any piece of information in memory might potentially afford a useful analogy . How might potential analogies be accessed in memory ? Is the memory search process directed , and if so , how ? At one extreme the problem solver may not actually search memory at all ; rather , he or she may simply \u201cstumble upon\u201d an analogy . That is , after a piece of knowledge has for some reason become the focus of attention , the person may spon - taneously notice its analogous relationship to a problem yet to be solved . It also seems plausible , however , that people may sometimes locate use - ful analogies in memory on the basis of a conscious search process . It may be possible to use a representation of the current problem as a retrieval cue for accessing analogous problems . Perhaps in some cases the person 350 GICK AND HOLYOAK first begins working on a problem and arrives at an abstract characteriza - tion of a potential solution , as we discussed in Experiment I . This solution representation might then be used to retrieve an analogous problem with that type of solution , which could then be used to help generate a more concrete solution to the target problem . The latter possibility is related to the solution - focusing strategy discussed in connection with Experiment I . A better understanding of how analogies are retrieved and noticed is clearly essential in order to effectively teach the use of analogies as a heuristic strategy for problem solving ( Polya , 1957 ) . The Generality of the Mapping Process The mapping process involved in the use of analogies may play a role in a variety of cognitive skills . Using an analogy involves mapping the repre - sentations of two ( or perhaps more ) instances onto one another . Similar processes may also be involved in abstracting the relational structure common to a set of particular instances . In the domain of problem solving , for example , a person who encounters several analogies to the radiation problem might eventually derive a schema for \u201cdispersion - type\u201d prob - lems . This schema would presumably be structured much like a concrete instance of a dispersion problem ( cf . Figure l ) , except that the predicates and arguments would be more abstract . A person equipped with such a general schema could then solve new dispersion - type problems by map - ping them directly onto it . These observations suggest that similar map - ping processes may be involved in three distinct but interrelated ac - tivities : ( 1 ) comparing one instance to another ; ( 2 ) deriving a schema for a class of instances ; and ( 3 ) comparing an instance to a general schema . Note that the above description of the role of mapping potentially applies not just to problem solving , but to a wide range of cognitive skills requiring concept learning and classification of instances . Such skills are involved in tasks that vary a great deal in terms of both complexity and cognitive domain . For example , the mapping of correspondences between relational structures is involved in the use of schemata for story under - standing ( Rumelhart , 1975 ) , frames for scene perception ( Minsky , 1975 ) , and scripts for understanding of social behavior ( Abelson , 1975 ) . Such structures all serve to describe our ability to deal with novel instances of familiar situations . Theories in each domain must explain how abstract structures can be derived from a set of instances , and how instances can be related to each other and to abstract structures . If similar mapping processes are involved in analogical problem solving and other cognitive skills , then the study of the use of analogies to solve problems has implications that extend to other domains . We mentioned at the beginning of this paper that an analogy may often serve as a model to guide the development of a new theory . In a similar fashion a theory of ANALOGICAL PROBLEM SOLVING 351 analogical problem solving might serve as a useful model in developing theories in other areas of cognition . APPENDIX I Story Analogies Used in Experiment I First Paragraph ( All Stories ) A small country fell under the iron rule of a dictator . The dictator ruled the country from a strong fortress . The fortress was situated in the middle of the country , surrounded by farms and villages . Many roads radiated outward from the fortress like spokes on a wheel . A great general arose who raised a large army at the border and vowed to capture the fortress and free the country of the dictator . The general knew that if his entire army could attack the fortress at once it could be captured . His troops were poised at the head of one of the roads leading to the fortress , ready to attack . However , a spy brought the general a disturbing report . The ruthless dictator had planted mines on each of the roads . The mines were set so that small bodies of men could pass over them safely , since the dictator needed to be able to move troops and workers to and from the fortress . However , any large force would detonate the mines . Not only would this blow up the road and render it impassable , but the dictator would then destroy many villages in retaliation . A full - scale direct attack on the fortress therefore appeared impossible . Second Paragraph , Attack - Dispersion Story ( Version 1 ) The general , however , was undaunted . He divided his army up into small groups and dispatched each group to the head of a different road . When all was ready he gave the signal , and each group charged down a different road . All of the small groups passed safely over the mines , and the army then attacked the fortress in full strength . In this way , the general was able to capture the fortress and overthrow the dictator . Second Paragraph , Open Supply Route Story The general , however , was undaunted . He knew that one major thoroughfare leading to the fortress was always kept open as a supply route . He led his army to the head of the supply route . When all was ready he gave the signal , and the entire army charged down the open route . The army avoided the mines and attacked the fortress in full strength . In this way , the general was able to capture the fortress and overthrow the dic - tator . Second Paragraph , Tunnel Story The general , however , was undaunted . He and his men dug an under - ground tunnel beneath the mines following the route of the road to the 352 GICK AND HOLYOAK fortress . When the tunnel was dug , the men crawled through it until they arrived safely at the foot of the fortress . Here they all gathered together and attacked the fortress in full strength . In this way , the general was able to capture the fortress and overthrow the dictator . APPENDIX II Story Analogies Used in Experiment II Attack - Dispersion Story ( Version 2 ) A small country was controlled by a dictator . The dictator ruled the country from a strong fortress . The fortress was situated in the middle of the country , surrounded by farms and villages . Many roads radiated out - ward from the fortress like spokes on a wheel . A general arose who raised a large army and vowed to capture the fortress and free the country of the dictator . The general knew that if his entire army could attack the fortress at once it could be captured . The general\u2019s troops were gathered at the head of one of the roads leading to the fortress , ready to attack . However , a spy brought the general a disturbing report . The ruthless dictator had planted mines on each of the roads . The mines were set so that small bodies of men could pass over them safely , since the dictator needed to be able to move troops and workers to and from the fortress . However , any large force would detonate the mines . Not only would this blow up the road and render it impassable , but the dictator would then destroy many villages in retaliation . It therefore seemed impossible to mount a full - scale direct attack on the fortress . The general , however , knew just what to do . He divided his army up into small groups and dispatched each group to the head of a different road . When all was ready he gave the signal , and each group marched down a different road . Each group continued down its road to the fortress , so that the entire army finally arrived together at the fortress at the same time . In this way , the general was able to capture the fortress , and thus overthrow the dictator . Parade - Dispersion Story A small country was controlled by a dictator . The dictator ruled the country from a strong fortress . The fortress was situated in the middle of the country , surrounded by farms and villages . Many roads radiated out - ward from the fortress like spokes on a wheel . To celebrate the anniver - sary of his rise to power , the dictator ordered his general to conduct a full - scale military parade . On the morning of the anniversary , the gen - eral\u2019s troops were gathered at the head of one of the roads leading to the fortress , ready to march . However , a lieutenant brought the general a disturbing report . The dictator was demanding that this parade had to be more impressive than any previous parade . He wanted his army to be ANALOGICAL PROBLEM SOLVING 353 seen and heard at the same time in every region of the country . Further - more , the dictator was threatening that if the parade was not sufficiently impressive he was going to strip the general of his medals and reduce him to the rank of private . But it seemed impossible to have a parade that could be seen throughout the whole country . The general , however , knew just what to do . He divided his army up into small groups and dispatched each group to the head of a different road . When all was ready he gave the signal , and each group marched down a different road . Each group continued down its road to the fortress , so that the entire army finally arrived together at the fortress at the same time . In this way , the general was able to have the parade seen and heard through the entire country at once , and thus please the dictator . APPENDIX III Story Analogy Used in Experiments IV and V Attack - Dispersion Story ( Version 3 ) ( 1 ) A small country was ruled ( 2 ) [ from a strong fortress ] by a king . ( 3 ) The fortress was situated in the middle of the country ( 4 ) surrounded by farms and villages . ( 5 ) Many roads radiated outward from the fortress ( 6 ) like spokes on a wheel . ( 7 ) A rebel general vowed ( 8 ) to capture the fortress . ( 9 ) The general knew that ( 10 ) an attack ( 11 ) [ by his entire army ] would capture the fortress . ( 12 ) He gathered his army ( 13 ) at the head of one of the roads . ( 14 ) However , the general learned that ( 15 ) the king had planted mines ( 16 ) on each of the roads . ( 17 ) The mines were set ( 18 ) so that ( 19 ) [ small ] bodies of men could pass over them safely , ( 20 ) since the king needed ( 21 ) to move his troops and workers to and from the fortress . ( 22 ) However , any ( 23 ) [ large ] force would detonate the mines . ( 24 ) Not only would this blow up the road ( 25 ) and render it impassable , ( 26 ) but it would also destroy many neighboring villages . ( 27 ) It therefore seemed impossible ( 28 ) to mount a ( 29 ) [ full - scale direct ] attack on the fortress . ( 30 ) The general , however , knew just what to do . ( 31 ) He divided his army up into small groups ( 32 ) and dispatched each group ( 33 ) to the head of a different road . ( 34 ) When all was ready ( 35 ) he gave the signal ( 36 ) and each group marched down a different road . ( 37 ) Each group continued down its road to the fortress ( 38 ) so that the ( 39 ) [ entire ] army finally arrived ( 40 ) [ together ] at the fortress at the same time . ( 41 ) The fortress fell ( 42 ) and the king was forced ( 43 ) to flee into exile . APPENDIX IV Distractor Stories Used in Experiment IV The Wine Merchants One day a rich man found that his wine cellar was empty . So he sent out messengers to announce a generous offer . The first person to bring the 354 GICK AND HOLYOAK rich man a barrel of wine would be given a brick of solid gold . However , the offer would expire at sundown . Two wine merchants heard the news . Each had a horse - drawn cart loaded with large barrels of wine . They both set out for the duke\u2019s palace at once . An hour before sundown they came to a place where the bridge had been washed out by a raging river . The first merchant drove his horses and cart into the flood in a desperate attempt to reach the other side . But the horses were already exhausted and could not fight the cur - rent . The cart overturned , and the horses , wine , and driver were washed away . The second merchant tried a different tactic . He poured the wine out of all but one of his barrels , and lashed them together to form a raft ; then he loaded the one full barrel , a horse , and himself on top . He set the raft adrift and floated downstream . In a few minutes the raft came to rest on the shore in front of the town where the rich man lived . The merchant disembarked , loaded the wine barrel on the horse , and led it to the rich man\u2019s house . He arrived just as the sun was setting , and collected the gold brick as a reward for his efforts . The Identical Twins Once there were identical twins who were continually playing pranks on their family , friends , and teachers . The annual school picnic was al - ways a big event for the twins . There were races and other athletic events in which the twins won lots of prizes . One year a new student arrived who was a star runner . The twins wanted to win the main event : the 2 - mile race through the woods behind the school . So they secretly devised a plan which would enable them to outdo the newcomer . The day of the race arrived . Each runner was to pick his own path through the woods to a clearing , where a teacher stood posted to deter - mine the winner . One twin entered the race , while the other excused himself on the grounds that he had hurt his leg in an earlier broadjumping event . The race began and the students rushed into the woods . The twin rushed into the woods and waited until the others had passed out of sight . Then he went back to the school using a path hidden from the picnic area . Shortly after , the other twin , who had been hiding behind a rock near the finish line of the race , burst out and ran into the clearing ahead of the other runners . The teacher named him the winner and marveled at the speed of his running . Next year the twins switched places and thereafter maintained their status on this event . REFERENCES Abelson , R . P . Concepts for representing mundane reality in plans . In D . G . Bobrow & A . Collins ( Eds . ) , Representation and understanding : Studies in cognitive science . New York : Academic Press , 1975 . Bishop , Y . M . M . , Fienberg , S . E . , & Holland , P . W . Discrete multivariate analysis : Theory and practice . Cambridge : MIT Press , 1975 . ANALOGICAL PROBLEM SOLVING 355 Black , J . , & Bower , G . H . Story understanding as problem - solving . Poetics , in press . Boden , M . Arri\u2019cial intelligence and natural man . New York : Basic Books , 1977 . Brooks , L . Nonanalytic concept formation and memory for instances . In E . Roscb & B . B . Lloyd ( Eds . ) , Cognition and categorization . Hillsdale , NJ : Erlbaum , 1978 . Collins , A . , Warnock , E . H . , Aiello , N . , & Miller , M . L . Reasoning from incomplete knowl - edge . In D . G . Bobrow & A . Collins ( Eds . ) , Representation and understanding : Studies in cognitive science . New York : Academic Press , 1975 . Duncker , K . On problem solving . Psychological Monographs , 1945 , 58 ( Whole No . 270 ) . Gordon , W . J . J . Synectics . New York : Harper & Row , 1961 . Hadamard , J . The psychology of invention in the mathematical field . Princeton , NJ : Princeton Univ . Press , 1945 . Hayes , J . R . , & Simon , H . A . Psychological differences among problem isomorphs . In N . J . Castellan , Jr . , D . B . Pisoni , & G . R . Potts ( Eds . ) , Cognitive theory . Hillsdale , NJ : Erlbaum , 1977 . Vol . 2 . Kintsch , W . The representafion of meaning in memory . Hillsdale , NJ : Erlbaum , 1974 . Kintsch , W . , & Van Dijk , T . A . Toward a model of text comprehension and production . Psychological Review , 1978 , 85 , 363 - 394 . Maier , N . Reasoning in humans . I . On direction . Journal of Comparative Psychology , 1930 , 10 , 115 - 143 . Maier , N . Reasoning in humans . II . The solution of a problem and its appearance in con - sciousness . Journal of Comparative Psychology , 1931 , 12 , 181 - 194 . Medin , D . L . , & Schaffer , M . M . Context theory of classification learning . Psychological Review , 1978 , 85 , 207 - 238 . Minsky , M . A framework for representing knowledge . In P . H . Winston ( Ed . ) , The psychol - ogy of computer vision . New York : McGraw - Hill , 1975 . Polya , G . How to solve if . Princeton , NJ : Princeton Univ . Press , 1957 . Reed , S . K . , Ernst , G . W . , & Banerji , R . The role of analogy in transfer between similar problem states . Cognitive Psychology , 1974 , 6 , 436 - 450 . Reitman , W . Heuristic decision procedures , open constraints , and the structure of ill - defined problems . In M . W . Shelley & G . L . Bryan ( Eds . ) , Human judgment and opti - ma & y . New York : Wiley , 1964 . Rumelhart , D . E . Notes on a schema for stories . In D . G . Bobrow & A . Collins ( Eds . ) , Representation and understanding : Studies in cognitive science . New York : Academic Press , 1975 . Schank , R . , & Abelson , R . P . Scripts , plans , goals , and understanding . Hillsdale , NJ : Erlbaum , 1977 . Simon , H . A . The structure of ill - structured problems . Artificial Intelligence , 1973 , 4 , 181 - 201 . Sternberg , R . J . Intelligence , information processing and analogical reasoning : The com - ponential analysis of human abiliries . Hillsdale , NJ : Erlbaum , 1977 . ( a ) Sternberg , R . J . Component processes in analogical reasoning . Psychological Review , 1977 , 84 , 353 - 378 . ( b ) Thorndyke , P . W . Cognitive structures in comprehension and memory of narrative dis - course . Cognitive Psychology , 1977 , 9 , 77 - 110 . Winston , P . H . Learning structural descriptions from examples . In P . H . Winston ( Ed . ) , The psychology of computer vision . New York : McGraw - Hill , 1975 . REFERENCE NOTE 1 . Gentner , D . The role of analogical models in learning scientific topics . Technical re - port , Bolt , Beranek and Newman , Inc . , 1979 . ( Accepted October 2 , 1979 )",
    "boudreauFieldExperimentSearch2017": "The Review of Economics and Statistics V OL . XCIX O CTOBER 2017 N UMBER 4 A FIELD EXPERIMENT ON SEARCH COSTS AND THE FORMATION OF SCIENTIFIC COLLABORATIONS Kevin J . Boudreau , Tom Brady , Ina Ganguli , Patrick Gaule , Eva Guinan , Anthony Hollenberg , and Karim R . Lakhani * Abstract\u2014 We present the results of a \ufb01eld experiment conducted at Har - vard Medical School to understand the extent to which search costs affect matching among scienti\ufb01c collaborators . We generated exogenous varia - tion in search costs for pairs of potential collaborators by randomly assigning individuals to 90 - minute structured information - sharing ses - sions as part of a grant funding opportunity . We estimate that the treat - ment increases the probability of grant co - application of a given pair of researchers by 75 % . The \ufb01ndings suggest that matching between scien - tists is subject to considerable friction , even in the case of geographically proximate scientists working in the same institutional context . I . Introduction T HE primary unit of scienti\ufb01c knowledge production has become the team or collaboration rather than the lone scientist ( Jones , 2009 ) . Indeed , teams are not only growing in frequency , but also in size and impact relative to single authors ( Wuchty , Jones , & Uzzi , 2007 ) . Unlike set - tings inside \ufb01rms , where executives and managers play a central role in organizing and forming teams ( Lazear & Shaw , 2007 ) , academic scientists have greater freedom and autonomy in selecting their collaborators and their topics of inquiry ( Stephan 2012 ) . Although there is a growing body of research on the productivity and outcomes of scienti\ufb01c teams once formed ( e . g . , Adams et al . , 2005 ; Wuchty , Jones , & Uzzi , 2007 ; Agrawal , Goldfarb , & Teodoridis , 2016 ) , we know relatively little about the largely decentra - lized process by which scienti\ufb01c teams come into existence ( Stephan , 2012 ) . In this paper , we investigate the role of one particular mechanism , search costs and frictions , on these matching outcomes . The role of search costs and resulting frictions in the for - mation of scienti\ufb01c collaborations is not well understood . On the one hand , the growing prominence of teams and fall - ing communications and collaboration costs in science ( Agrawal & Goldfarb , 2008 ; Ding et al . , 2010 ) might sug - gest forces favorable to novel team formation . On the other hand , geography and distance are regularly documented to play a role in shaping collaborations , even today ( e . g . , Rosenthal & Strange , 2001 ; Glaeser , 2010 , Catalini , 2016 ) , and rather than continually forming novel collaborations , scientists most often work with partners in the same institu - tion , in a similar knowledge domain , and within preexisting social networks ( Baccara & Yariv , 2013 ; Freeman , Ganguli , & Murciano - Goroff , 2015 ; Freeman & Huang , 2014 ; Fafchamps , Goyal , & Van der Leij , 2010 ; Azoulay , Liu , & Stuart , 2009 ) . Moreover , past collaborations remain an important predictor of future ones . Although these patterns might be explained by any number of factors , they raise the question of whether search costs play a \ufb01rst - order role in shaping the organization of scientists into teams . The high information requirements for forming matches suggest that search frictions may be an important considera - tion . A large number of factors can play a role in decisions to collaborate , and these factors may be nuanced or dif\ufb01cult to observe . This includes factors such as the complementar - ity of skills of prospective partners , current research interests and priorities , access to broader sets of relevant resources ( funding , equipment , research personnel ) , timing and sche - duling constraints , and personal chemistry and disposition Received for publication March 11 , 2016 . Revision accepted for publi - cation December 2 , 2016 . Editor : Brigitte C . Madrian . * Boudreau : Northeastern University ; Brady : Massachusetts General Hospital and Harvard Medical School ; Ganguli : University of Massachu - setts Amherst ; Gaule : CERGE - EI , a joint workplace of Charles University and the Economics Institute of the Academy of Sciences of the Czech Republic ; Guinan : Dana - Farber Cancer Institute and Harvard Medical School ; Hollenberg : Beth Israel Deaconess Medical Center and Harvard Medical School ; Lakhani : Harvard Business School and NBER . K . J . B . , I . G . , P . G . , E . G . , and K . R . L . designed , developed , and executed the experiment and contributed to the manuscript . I . G . and P . G . con - ducted data analysis . T . B . and A . H . contributed to the experimental devel - opment and execution . We appreciate helpful comments from Pierre Azoulay , Marcel Fafchamps , Lee Fleming , Ben Golub , Shane Greenstein , Ben Jones , Nicola Lacetera , Josh Lerner , Paula Stephan , and Scott Stern , and seminar participants at the NBER Summer Institute Innovation Meet - ings , the NBER Productivity Lunch , the Georgia Tech REER Conference , MIT Sloan School of Management , Northwestern University , University of California , Berkeley , University of Rochester , SITE Stockholm School of Economics , CIRCLE Lund University , Graduate Institute Geneva , and Universidad Carlos III de Madrid . This work was conducted with support from Harvard Catalyst / The Harvard Clinical and Translational Science Center ( National Center for Advancing Translational Sciences , National Institutes of Health Awards UL1TR001102 , UL1TR000170 , and UL1RR025758 - 02S4 ) , NASA Tournament Lab at Harvard University , Harvard Business School Division of Research and Faculty Development , and \ufb01nancial contributions from Harvard University and its af\ufb01liated aca - demic health care centers . We also thank Harvard Catalyst for support and cooperation in implementing the experiment , particularly Lee Nadler , William Chin , Laura Weisel , and David Frank . Amy Webber and Wei Zhong provided valuable assistance throughout the implementation pro - cess . Michael Menietti and Christoph Riedl assisted with execution of the experiment . The Harvard Innovation Laboratory generously provided us with meeting space for the experiment . A supplemental appendix is available online at http : / / www . mitpress journals . org / doi / suppl / 10 . 1162 / REST _ a _ 00676 . The Review of Economics and Statistics , October 2017 , 99 ( 4 ) : 565 \u2013 576 (cid:2) 2017 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology doi : 10 . 1162 / REST _ a _ 00676 ( Stephan , 2012 ) . If acquiring and evaluating this information is costly , signi\ufb01cant search frictions will appear , as has been found in other matching markets ( Mortensen & Pissarides , 1999 ) . Observed patterns of collaboration might then be interpreted as re\ufb02ecting limited information in decision making\u2014and therefore may constitute a suboptimal alloca - tion of human resources . To understand whether and to what extent search costs can affect the formation of collaborations among research collaborators , we carried out a \ufb01eld experiment with the goal of introducing exogenous variation in the informa - tion available to research scientists concerning potential collaborators . Our research team worked closely with Harvard Medical School\u2019s ( HMS ) clinical and administra - tive executives to modify and redesign existing internal grant processes so that causal inferences could be drawn in the context of an $ 800 , 000 grant opportunity for re - searchers at Harvard University and the HMS system of hospitals and research centers to encourage the develop - ment of clinical applications of advanced medical imaging technologies . The \ufb01eld experiment involved designing a research sym - posium ( repeated on three consecutive nights ) that was part of the grant process , where investigators were to get details about the grant rules and administration , learn about advanced technologies underlying the grant , and meet other researchers through structured information - sharing ses - sions . Participation in one of the symposia ( and only one ) was mandatory for submitting a grant application , which was due four weeks after the symposia . Each symposium consisted of a 30 - minute general introduction followed by 90 minutes of information sharing in independent and phy - sically separated breakout rooms . Breakout rooms facili - tated face - to - face interactions by having half of the researchers circulate about the room while the other half \u2018\u2018broadcast\u2019\u2019 their research ideas in a standardized poster format . We reduced the cost of initial face - to - face interac - tions for random subsets of scientists by randomly assigning the roughly 400 researchers who took part to independent breakout rooms . Therefore , we can evaluate the effect of the treatment by simply comparing the likelihood of colla - boration for pairs of researchers assigned to the same room ( treatment ) with the likelihood of pairs assigned to different rooms ( control ) . It is important to note that estimates of search costs in this context might be interpreted as occurring under best - case conditions . We study prospective collaborators operat - ing within a shared institutional context , with potential funding availability , within the same geographic area , and in a context in which information systems and tools facili - tate the search for prospective collaborators . Yet our results suggest that matching between scientists is subject to considerable frictions even in this best - case context . We estimate that assignment to the same breakout room increased the probability of forming a collaboration by 75 % , increasing the probability from 0 . 16 % in the control group to 0 . 28 % in the treatment . We estimate the effect to be signi\ufb01cant at the 5 % or 10 % level , depending upon model speci\ufb01cation . ( The 95 % con\ufb01dence interval around the point estimate ranges from (cid:2) 4 % to (cid:2) 112 % . ) To put this into perspective , the point estimate of the treatment effect is about one - third of the effect of working in the same hospital or of performing research in the same clinical area . This is a substantial effect for what is arguably a relatively small ( 90 - minute information sharing ) treatment . This main \ufb01nding is consistent with the view that large search costs and frictions play a \ufb01rst - order role in shaping the process of searching for collaborators and suggests the important function of information - rich face - to - face encoun - ters in catalyzing collaborations . Consistent with the inter - pretation of a signi\ufb01cant effect of search costs , the treatment effect was especially strong for pairs of researchers working in the same clinical area , where presumably search costs might be construed as lower given similar backgrounds and training . The \ufb01ndings therefore suggest the possibility that current observed patterns of collaborations in academic science are perhaps highly constrained by the availability of information and search costs . This is plausibly an important source of inef\ufb01ciency . However , we cannot observe impli - cations of this inef\ufb01ciency within this analysis . The \ufb01nding of the \ufb01rst - order role played by search costs also offers one plausible explanation for the prevalence of homophily ( McPherson , Smith - Lovin , & Cook 2001 ; Currarini , Jackson , & Pin 2009 ) in forming collaborations , where like scientists tend to coauthor , and repeatedly , as both tendencies may economize on search costs . The \ufb01nd - ings also imply potentially important differences between the formation of collaborations versus the execution of dis - tributed collaborations . The formation and execution of col - laborations may be considered as representing altogether different kinds of coordination problems\u2014one of matching and the other of joint production . Whereas evidence sug - gests research collaborations may be able to be carried out at a distance through decreased communication and travel costs and increasingly sophisticated collaboration platforms ( see Agrawal & Goldfarb 2008 ; Jones , Wuchty , & Uzzi 2008 ; Adams et al . , 2005 ; Catalini , Fons - Rosen , & Gaule 2016 ) , the process of forming collaborations may still be especially highly in\ufb02uenced and informed by information - rich , interpersonal interactions . The paper proceeds as follows . We \ufb01rst describe our experimental design , including details of the grant program and research symposia in section II . In section III , we describe the data . The empirical strategy and results follow in sections IV and V , respectively . Section VI concludes . II . The Field Experiment A . Harvard Medical School and Its Af\ufb01liated Hospitals Our \ufb01eld experiment involved faculty and researchers from Harvard University and its af\ufb01liated hospitals and 566 THE REVIEW OF ECONOMICS AND STATISTICS institutions . Harvard Medical School and its seventeen af\ufb01liated hospitals and research institutes ( including Massa - chusetts General Hospital , Children\u2019s Hospital Boston , Brigham and Women\u2019s Hospital , Beth Israel Deaconess Medical Center , and the Dana - Farber Cancer Institute ) are a major force in biomedical research . Collectively , they employ more than 11 , 000 faculty and receive in excess of $ 1 . 5 billion in annual funding from the U . S . National Insti - tutes of Health ( NIH ) . Harvard researchers account for around 5 % of scienti\ufb01c articles published in the top four medical journals , a larger share than Germany or Canada as a whole . 1 Fifteen researchers have shared in nine Nobel prizes awarded for work done while at Harvard Medical School . While our experiment is set entirely within the Harvard University system , in fact its researchers work in distinct organizations and research centers . The Harvard - af\ufb01liated hospitals are separately owned and managed and appear as separate entities in hospital rankings and lists of NIH grant recipients . Four of the \ufb01ve largest hospitals are located in the Longwood Medical Area campus in Boston , while Mas - sachusetts General Hospital has its own campus about 3 miles away ( and approximately 20 minutes by institutional shuttle bus ) . B . Harvard Catalyst and Advanced Imaging Closing the gap between research \ufb01ndings and clinical applications ( \u2018\u2018bench to bedside\u2019\u2019 ) is a major priority for the NIH . This has resulted in the establishment of a new insti - tute , National Center for Advancing Translational Sciences , that provides signi\ufb01cant research funding to universities and hospitals that undertake collaborative translational activities to accelerate treatment development . As part of Harvard\u2019s efforts to promote clinical and translational research , the Harvard Clinical and Translational Center , Harvard Catalyst , provides seed funding in the form of pilot grants to support nascent research efforts . These pilot grants are awarded competitively to faculty within Harvard Uni - versity . They emphasize early - stage research with the potential to improve human health . Pilot grant funding enables researchers to generate the preliminary data that are essential for larger grant applications to the NIH . Our \ufb01eld experiment was layered onto a Harvard Catalyst pilot grant program . This particular grant opportunity , which offered $ 50 , 000 per award , was centered on proposals to devise or improve methods for using advanced medical ima - ging technologies\u2014speci\ufb01cally , physiological magnetic resonance ( MR ) , positron emission tomography ( PET ) , and optical imaging\u2014to address unmet clinical needs . A major challenge in the \ufb01eld of advanced imaging is that progress requires both expertise in the latest imaging tools and tech - nologies and a deep understanding of the health problems to which they could be applied , with these different types of knowledge typically being held by people with different dis - ciplinary backgrounds . Thus , advanced imaging is an arche - typical example of a problem often found in modern science where advancing the knowledge frontier requires combining knowledge embodied in different individuals ( Jones , 2009 ) . We worked in close collaboration with HMS administra - tors and executives to redesign their pilot grant process so that we could obtain causal inferences about the role of search costs in \ufb01nding collaborators . While the grant process was primarily focused on identifying and funding promising early - stage translational research in the \ufb01eld of advanced imaging , Harvard Catalyst leaders also perceived a need for familiarizing clinicians with recent developments in advanced imaging and for Harvard - wide community build - ing among researchers . This provided us with the opportu - nity to create a new interactive research symposium where we could exogenously shift search costs for certain pairs of individuals by building in randomized face - to - face interac - tions . Hence we modi\ufb01ed the Harvard Catalyst grant process by requiring potential applicants to attend an interactive research symposium that would be a forum to learn about new technologies , understand the grant process , and exchange ideas among fellow researchers across Harvard . This was the \ufb01rst time such an interactive Harvard - wide symposium on a new research grant opportunity was offered . In November 2011 , all Harvard University life sciences faculty and researchers were invited to participate in a unique funding opportunity centered on advanced imaging technologies using a directed e - mail campaign , outreach to departmental clinical and research directors , and marketing messages on various internal websites and through posters across facilities . Up to $ 800 , 000 was available to support \ufb01fteen pilot grants . There was the additional potential for researchers to apply for several concept development prizes of $ 2 , 000 each . The concept prizes were meant to stimulate innovative thinking and future investigation in areas in which imaging had not been previously considered as an intervention and did not require any implementation plan . In the \ufb01rst stage , investigators who were interested in applying for the grants were asked to submit a statement of interest in which they brie\ufb02y described a speci\ufb01c medical problem that advanced imaging techniques could poten - tially address . Basic biographical information ( e . g . , degree , institution , department appointment ) was collected at this stage . Information distributed about the funding opportunity speci\ufb01ed that eligibility to submit a \ufb01nal application was conditional on attending an advanced imaging symposium on one of three preannounced dates . Applicants could indi - cate at this stage if there were any dates during which they could not attend a symposium . It was also communicated to applicants that the symposia would be studied by Harvard Catalyst to develop better insights about scienti\ufb01c team for - mation and that data on interaction patterns among indivi - duals would be collected . 1 Journals included are the New England Journal of Medicine , Journal of the American Medical Association ( JAMA ) , Nature Medicine , and Lan - cet . Authors\u2019 calculations based on research articles published during from 2000 to 2009 . Fractional counting was used when coauthors belonged to different institutions . 567 SEARCH COSTS AND THE FORMATION OF SCIENTIFIC COLLABORATIONS C . Randomization and the Advanced Imaging Symposium The initial call generated 471 statement of interest appli - cations , of which 435 applicants were invited to attend an advanced imaging symposium and thus proceed in the grant application process . 2 Forty - one applicants ( 9 . 4 % ) failed to RSVP or otherwise show up at the event . 3 Invitations were also extended to several individuals with world - class exper - tise in advanced imaging , bringing the total number of par - ticipants to 402 . The symposium was structured so that participants would come to the event prepared to discuss their idea with other participants in small breakout rooms of thirty to forty peo - ple . The treatment was intended to introduce exogenous var - iation in search costs to some pairs of participants at the symposium by having them be present in the same breakout rooms at the event . Each participant was randomly allocated to a breakout room in advance so that a random subset of all possible pairs among all participants would receive the treat - ment . Three symposia were held on sequential nights and were identically structured , with four breakout rooms per night . 4 We also randomized the participants across nights ; however , we respected the blackout dates for which appli - cants had previously indicated they would not be available . 5 The events were held January 31 , February 1 , and Febru - ary 2 , 2012 , at the Harvard Innovation Lab , located on Har - vard\u2019s Allston campus . The program began with a 30 - minute address by the program leadership describing the pilot grant opportunity and the agenda for the evening , including an introduction to advanced imaging tools and technologies . The breakout sessions then began in separate rooms . The number of participants in each room varied from 28 to 43 . The breakout room sessions were split into two periods of 45 minutes each , with a 15 - minute break in the middle dur - ing which all participants could mingle in a common space where refreshments were provided . The rooms provided a venue for presentation of the participants\u2019 ideas in the form of posters . Each poster followed a standard format describ - ing each participant\u2019s submitted idea from the statement of interest ( based on information they had provided prior to the event ) and was placed in the breakout room in advance . 6 The posters were intended to foster information sharing among participants and included the following details related to the statement of interest idea : ( a ) What is your question ? ( b ) Why does it matter ? and ( c ) What is needed for your research to succeed ? A 300 - character limit was imposed for each question . Posters were prepared in a stan - dard size and format by Harvard Catalyst , and each was placed on a separate whiteboard that allowed for the possibi - lity of visual explanations and note taking . Participants within each breakout room were randomly split into two groups . Participants from group 1 were asked to stand by their poster during the \ufb01rst period , while group 2 participants circulated . The two groups then switched roles during period two : group 1 participants circulated around the room , while group 2 participants stood near their own posters . The placement of each individual\u2019s poster in the room was also randomly determined in advance . D . Grant Applications Shortly after the symposia , all participants received via e - mail an invitation to submit applications for the pilot grants or concept awards by the deadline of March 8 , 2012 . At this time , they also received PDF booklets with the names , contact information , and posters of all researchers who participated over the three nights . 7 Note that as of 2008 , Harvard Catalyst had already deployed the Harvard Catalyst Pro\ufb01les website , an online , publicly accessible interactive database that includes contact information , cur - rent appointments , individual publication records , and other information for faculty and researchers across Harvard Medical School , and can be searched by name or keyword . Thus , much of the information contained in the PDF book - let could already easily be acquired online . However , the booklet also included information on current research inter - ests , related to the grant , that was less easily available ; the booklet may also have increased the salience of the publicly available information for the included individuals . Our intention was to provide identical information to all partici - pants apart from information acquired speci\ufb01cally in the breakout rooms at the symposia . Consistent with previous Harvard Catalyst pilot grant processes , applications had to include a principal investiga - tor and at least one co - investigator . Concept award applica - tions similarly had to include at least two individuals . Researchers with faculty appointments could apply as prin - cipal investigator on only one pilot grant but could apply as 2 Thirty - six statements of interests were outside the parameters of the request for applications in terms of area of inquiry ( e . g . , proposing ultra - sound or X - ray computed tomography [ CT ] techniques ) and the submit - ters were not invited to attend the symposium . 3 We do not include these individuals in the analysis . 4 The randomization was carried out by generating a unique random number for each participant , ranking the numbers , and then assigning par - ticipants to breakout rooms on nights based on their rank . We assigned 32 participants to the \ufb01rst three rooms each night , and the remainder ( 41 \u2013 48 ) to the last room , which was slightly larger . 5 Participants with blackout dates were a minority , but to guard against the potential endogeneity of selection into nights for this group , the analy - sis focuses on comparisons within nights . 6 Participants were provided with the following information in the e - mailed invitations to attend a symposium : \u2018\u2018You do not need to bring any particular items with you . We have a poster prepared with your submitted answers to the three questions based on your statement of interest . Posters will be displayed at the symposium to facilitate talking about your idea with other attendees . There will be no formal presentations of any kind . \u2019\u2019 7 The following information was included in e - mail communication with participants immediately following the event : \u2018\u2018Attached to this email is a PDF booklet with the names and contact information of all researchers who participated over the three nights and their posters . We hope this is of use in contacting individuals that you met during the eve - ning and in identifying additional potential collaborations and col - laborators . . . . As described at the symposium , your proposal or your col - laborators can be the same as suggested in your Statement of Interest or can be somewhat or entirely different . You can participate in multiple applications . \u2019\u2019 568 THE REVIEW OF ECONOMICS AND STATISTICS co - investigator on an unlimited number of additional appli - cations . Researchers without a faculty appointment could not be principal investigators on a pilot grant application , but they could be co - investigators on an unlimited number of applications . All attendees were eligible to apply for a concept award grant and could appear on an unlimited num - ber of applications . Finally , at least one applicant on any grant application had to have attended the symposium . The grant application did not need to be based on the initial statement of interest . Extra care was taken to ensure that the symposium pro - cess did not somehow prime participants to seek collabora - tions only in their breakout rooms . Participants were informed that the composition of their teams would not be communicated to reviewers and would not be considered as a criterion for awarding the grant . They were also told to remove any personal identifying information about the submission teams from their proposals ( including self - references and indications of special access to technologies ) . 8 This differed from the typically single - blinded process used in NIH and Catalyst grants , in case the identi\ufb01cation of sub - mission applicants might have an impact on collaboration choices . In the end , the majority of participants chose not to apply with other symposium participants : 66 % of the applications included only one symposium participant as a coapplicant . III . Data A . Sources and Construction Registration data . Faculty and researchers interested in taking part in the funding opportunity were asked to submit a short statement of interest describing in 250 words or less a speci\ufb01c medical problem that advanced imaging techni - ques could potentially address . Registration data also included basic biographical information ( rank , education history , hospital af\ufb01liation , department ) . Participants were also asked to identify themselves as primarily an imager or primarily a clinician . Clinical area and imaging modality were coded from the statement of interest documents . Publications . We matched participants to Harvard Cat - alyst Pro\ufb01les , an online , publicly accessible database that includes individual publication records and other informa - tion for faculty and researchers across Harvard Medical School . Grant applications . Our main outcome variable comes from the pilot grant and concept award applications . We received 224 applications for pilot grants or concept awards . 9 Of those , 148 included one symposium participant in the applicant list , 49 included two symposium participants , and 27 included more than two symposium participants . Constructing the pair - level data . Our analysis is con - ducted at the dyadic level . We constructed dyads by creat - ing every possible pairwise combination of researchers attending a symposium on the same night ( 26 , 604 dyads ) . We constructed our main outcome at the dyadic level , colla - boration , from grant applications . A collaboration was de\ufb01ned as any pairs of symposium participants appearing on the same application . In some parts of the analysis , we restricted our sample to the 52 pairs of scientists who attended on the same night and coapplied . We also con - structed a number of dyadic - level control variables from the registration and publication data , which are described in more detail in section IV . B . Summary Statistics and Randomization Check Table 1 provides individual - level summary statistics for symposium participants . 10 Of the 402 attendees , 29 % were women , 42 % identi\ufb01ed themselves as imagers , and 73 % held Harvard faculty appointments ( the others were post - doctoral fellows or clinical fellows ) . Over 80 % of attendees came from the four largest Harvard - af\ufb01liated hospitals : Massachusetts General Hospital ( MGH ) , Brigham and T ABLE 1 . \u2014S UMMARY S TATISTICS , A TTENDEES Sample Mean Female 0 . 29 Faculty member 0 . 73 Imager 0 . 42 Longwood 0 . 51 HospitalMassachusetts General Hospital 0 . 37 Brigham and Women\u2019s Hospital 0 . 19 Beth Israel Deaconess Medical Center 0 . 14 Children\u2019s Hospital Boston 0 . 13 Other 0 . 17 Clinical area ( SOI ) Neurology 0 . 25 Oncology 0 . 25 Neuropsychiatric 0 . 10 Cardiovascular 0 . 06 Gastroenterology 0 . 04 Transplantation 0 . 04 Ophthalmology 0 . 03 Other 0 . 23 Attended January 31 0 . 35 Attended February 1 0 . 32 Attended February 2 0 . 33 Observations 402 SeesectionIII inthe textfor adetaileddescriptionofthevariables . 8 The following directions to applicants were highlighted in the grant request for applications : \u2018\u2018As the initial review will be blinded in regard to the applicant ( s ) , do not refer to yourself , other participants or institu - tions by name ( e . g . , substitute \u2018our optical imaging experts , \u2019 \u2018our cardiol - ogy collaborators , \u2019 \u2018our laboratory\u2019 or \u2018the genomics core\u2019 for speci\ufb01c individuals or facilities ) . \u2019\u2019 9 Seventy - eight percent of applications were for pilot grants , and 22 % were for concept awards . 10 Across the three nights , 394 individuals were in attendance . How - ever , 5 people with special expertise in advanced imaging attended the event on more than one night ; we count them as different participants on each night , bringing the total number of participants to 402 . 569 SEARCH COSTS AND THE FORMATION OF SCIENTIFIC COLLABORATIONS Women\u2019s Hospital , Children\u2019s Hospital Boston , and the Beth Israel Deaconess Medical Center . The most prevalent clinical expertise areas were neurology ( 25 % ) , oncology ( 25 % ) , and neuropsychiatry ( 10 % ) . We can also compare the participants to the general population of researchers at Harvard Medical School . Appendix table A1 provides summary statistics on partici - pants and nonparticipants based on information in the Har - vard Catalyst Pro\ufb01les database . In terms of degree types , there was no signi\ufb01cant difference in the share of MDs among attendees and the overall HMS population , but there was a larger share of PhDs among attendees ( 49 % PhDs among attendees versus 38 % at HMS ) . We would expect a greater representation of PhDs at the event since it was part of a research grant opportunity , and academic PhDs are very often focused on research , while academic MDs have a larger array of potential roles . Attendees also had more prior publications on average ( approximately four publica - tions more than the typical HMS researcher ) . We also see some signi\ufb01cant differences in the distribution across ranks , with attendees more likely to be instructors and assistant or associate professors relative to the overall dis - tribution at HMS and less likely to be full professors and postdocs . Attendees were also more likely to come from MGH . One reason for this is that MGH houses a large advanced imaging center , the Martinos Center for Biome - dical Imaging , and the focus of the grant opportunity was advanced imaging . For the same reason , individuals from radiology departments were overrepresented among atten - dees . To verify that the randomization generated balance across covariates , we present summary statistics in table 2 for the pairs in our sample assigned to the same breakout room and those assigned to different breakout rooms . The unit of observation here is a dyad of researchers , and the sample is every possible pairwise combination of research - ers attending on the same night ( 26 , 604 dyads ) . Treated pairs and control pairs look very similar , with the exception of pairs of previous coauthors , pairs with both members from the same hospital , and pairs including one woman , which are statistically different across treated and control pairs . 11 In our regression analysis , we control for these cov - ariates . The last row of table 2 includes collaboration , our out - come variable . The incidence of collaboration is signi\ufb01 - cantly larger in the treated group , which we investigate in a regression framework in the next section . It is notable that the incidence of collaboration is less than 0 . 2 % in our sam - ple . While this may seem low , the likelihood that any two HMS faculty members will copublish in a given year is 0 . 06 % and , thus , of the same order of magnitude . 12 Viewed through the lens of all pair - wise combinations of scientists who could collaborate , collaboration is indeed a relatively rare event . Table 3 shows the characteristics of the subset of colla - borating dyads . Here we restrict the sample of every possi - ble pairwise combination of researchers attending on the same night ( 26 , 604 dyads ) to those who coapplied ( 52 dyads ) . Among attendees who attended on the same night but were not in the same breakout room , 33 pairs co - applied . Among pairs in the same room at the event , 19 pairs coapplied . 13 T - tests show that among the same - room collaborations , there was a higher incidence of pairs with one postdoc and of pairs researching the same clinical area . It is important to note that some of the within - room colla - borations would have occurred in the absence of any treat - ment effect . Extrapolating the across - room incidence rate ( 0 . 16 % ) to the number of within - room pairs ( 7 , 149 ) , we would expect eleven collaborations to have occurred within rooms in the absence of any treatment effect . T ABLE 2 . \u2014D YADS BY T REATMENT S TATUS Treatment : Control : Sample Means Same Room Different Room Difference One postdoc 0 . 404 0 . 396 (cid:2) 0 . 007 Both postdocs 0 . 072 0 . 075 0 . 003 One female 0 . 403 0 . 418 0 . 015 * Both female 0 . 085 0 . 082 (cid:2) 0 . 004 Same hospital 0 . 198 0 . 208 0 . 010 * Both Longwood 0 . 266 0 . 258 (cid:2) 0 . 010 * One imager \u00fe one clinician 0 . 492 0 . 489 (cid:2) 0 . 003 Both imagers 0 . 175 0 . 176 0 . 001 Same clinical area ( SOI ) 0 . 123 0 . 119 (cid:2) 0 . 004 Previous coauthor 0 . 001 0 . 002 0 . 001 * Observations 6 , 702 19 , 962 The unit of observation is a dyad of researchers . We construct dyads by creating every possible pairwise combination of researchers attending on the same night ( 26 , 604 dyads ) . The category Treatment : Same Room refers to participants in the same room at the event ; it was randomized across pairs of participants attending on the same night . Collaboration indicates whether the pair appeared on any common pilot grant or conceptawardapplications . See sectionIII foradetaileddescriptionofthevariables . Asterisksindicatethe results oftestsforequalityofmeans . * p < 0 . 10 , * * p < 0 . 05 , * * * p < 0 . 01 . 11 The relatively large difference between the percentage of pairs of previous coauthors across treatment and control groups can be explained by the very small number of pairs of previous coauthors in our sample ( 40 out of more than 20 , 000 ) . Thus , randomization could easily result in a dif - ferent incidence of pairs with coauthors across treatment and control groups , as it did in our case . 12 Authors\u2019 calculation based on publication data from Harvard Catalyst Pro\ufb01les . 13 The nineteen pairs who coapplied from the same room correspond to eighteen separate grant applications . 570 THE REVIEW OF ECONOMICS AND STATISTICS IV . Estimation Strategy A . Speci\ufb01cation We use the simplest possible estimation strategy to describe differences between treatment and control groups\u2014 and the effect of exogenous variation in search costs in our context . The approach of our statistical analysis is to study the incidence of collaborations among all possible pairs of participants attending on the same night within our experi - mental group of 402 individuals . This reduced - form approach suits our interest in studying the extent to which observed behaviors deviate from fully informed equilibrium out - comes . 14 This approach also allows us to deal with relatively small numbers of actual within - room collaborations in a most straightforward and conservative manner . Thus , the unit of analysis is the scientist pair , and the data set includes every possible pair of scientists across all nights . We use a linear probability model to describe how the incidence or probability of collaborations differs across treatment and control groups ( i . e . , those in the same versus different breakout rooms ) . Random assignment of pairs within the research design allows us to interpret differ - ences as causally related to exogenous variation in search costs . We are also able to regress the incidence of colla - borations on other covariates of researcher pairs to further describe associations with the incidence of collaborations . To measure whether treatment effects varied across sub - groups , we interacted Same Room indicator with pair - level variables . Thus , to estimate the impact of being in the same room at the event on the likelihood of collaboration between pairs , we ran linear regressions with the following speci\ufb01cation : Collaboration ij \u00bc a \u00fe b Same room ij \u00fe h Same room ij (cid:3) Distance ij \u00fe p Distance ij \u00fe d X ij \u00fe e ij ; ( 1 ) where the key explanatory variable associated with the treat - ment effect , Same Room ij , is an indicator variable that equals 1 if both researcher i and j were randomly assigned to the same breakout room at the symposium . 15 Collaboration ij is an indi - cator variable for whether i and j appeared on any common pilot grant or concept award applications . X ij is a vector of observable pair - level characteristics that can affect the likeli - hood of collaboration and includes measures of gender and professional rank . The vector Distance ij includes measures of differences in professional rank , as well as geographic , scien - ti\ufb01c , and past coauthoring , described below . The model also includes \ufb01xed effects for each night of the symposium . The estimation of dyadic regressions raises an inference problem : dyadic observations may not be independent of each other as the same individual appears in many dyads . To address this inference problem , Fafchamps and Gubert ( 2007 ) have developed a network inference method adapted from spatial econometrics that corrects dyadic correlation of errors and also possible heteroskedasticity . We estimate and report equation ( 1 ) using an OLS regression with grouped dyadic standard errors according to Fafchamps - Gubert . 16 T ABLE 3 . \u2014C OLLABORATING D YADS BY T REATMENT S TATUS Sample Means Collaborations within the Same Room CollaborationsacrossRooms Difference One postdoc 0 . 421 0 . 212 (cid:2) 0 . 209 Both postdocs 0 . 000 0 . 030 0 . 030 One female 0 . 474 0 . 303 (cid:2) 0 . 140 Both female 0 . 158 0 . 061 (cid:2) 0 . 097 Same hospital 0 . 579 0 . 636 0 . 057 Both Longwood 0 . 158 0 . 303 0 . 145 One imager \u00fe one clinician 0 . 474 0 . 485 0 . 011 Both imagers 0 . 316 0 . 394 0 . 078 Same clinical area ( SOI ) 0 . 579 0 . 273 (cid:2) 0 . 306 * * Previous coauthor 0 . 105 0 . 121 0 . 016 Observations 19 33 The unit of observation is a dyad of researchers . We construct dyads by creating every possible pairwise combination of researchers attending on the same night ( 26 , 604 dyads ) , but here we focus on researchers who attend on the same night and appeared on a common pilot grant or concept award application ( 52 dyads ) . See section III in the text for a detailed description of the variables . Asterisks indicate the r esults of t - testsforequality ofmeans . * p < 0 . 10 , * * p < 0 . 05 , * * * p < 0 . 01 . 14 Structural matching models that contemplate competitive equilibria in matching are an alternative approach to modeling the equilibrium for - mation of collaborations . However , pursuing such an approach requires we make structural assumptions regarding equilibrium search process and outcomes\u2014which goes against our interests in this study , given our inter - est in investigating frictions . Therefore , it is more appropriate in this instance to proceed with a reduced - form description of patterns to better describe any implications of search costs . Although this creates the possi - bility of downwardly biased estimates on the treatment effect , any such effect is likely to be vanishingly small : competition in matching is likely to have played only a small role , if much at all , as the absolute incidence of collaborations in these data is rather low and individuals were not lim - ited in the number of collaborations they could form . 15 There are several other ways to study and model search costs in this setting . We could , for example , study the effect of attending the sympo - sium on the same night . Furthermore , since participants\u2019 posters were also randomized within the breakout rooms , we could study if immediate neighbors in the breakout room at the event had an impact on collabora - tion . However , neither of these approaches had a signi\ufb01cant impact on our outcome of interest , grant coapplications . 16 We implement this using the ngreg ado \ufb01le available on Marcel Faf - champs\u2019s website : http : / / web . stanford . edu / (cid:4) fafchamp / resources . html . Note that we obtain very similar standard errors and con\ufb01dence intervals when using Eicker - White heteroskedasticity - robust standard errors instead . We do not cluster standard errors by night of attendance , since assignment to nights is itself random ( conditional on blackout dates for a minority of participants ) ( Cameron & Miller , 2015 ) . 571 SEARCH COSTS AND THE FORMATION OF SCIENTIFIC COLLABORATIONS B . Covariates Several additional covariates describing pairs are also included in the model . Inclusion of these covariates should not affect the point estimate of the treatment effect but should increase its precision and offer further opportunity for interpretation . Our vector of pair - level covariates , X ij , includes variables for gender and professional rank . Gender is captured by indicator variables Both female , One female , and Both male . Research indicates that women have a greater propensity to work with other women ( Boschini & Sjogren , 2007 ) and more generally have more limited aca - demic networks ( see Ding et al . , 2010 ) . For professional rank , we include indicators for One postoc in the pair and Both postdocs . Postdocs were eligible to apply for either the concept or pilot grants ; however , two postdocs could colla - borate on a pilot grant application only if a third team mem - ber with a faculty appointment assumed the role of princi - pal investigator . The vector Distance ij includes measures of differences in professional rank and geographic , scienti\ufb01c , and past coau - thoring . Given the potential relevance of these various forms of distance to search costs , coef\ufb01cients estimated on these variables provide at least some broad and rough means of judging the importance of any estimated treatment effect by direct comparison with coef\ufb01cients on these variables . Distances in professional range are measured with indi - cator variables corresponding to possible combinations of differences . In relation to geographic distance , we create an indicator for Same hospital , which indicates whether pair members\u2019 primary appointments are in the same Harvard - af\ufb01liated hospital or institute . We also create an indicator for Both Longwood , indicating that both members of the pair work on the same campus , 17 as the largest concentra - tion of researchers are located in hospitals and institutes either on the Longwood Medical Area ( LMA ) campus or at the MGH campus . The campuses are located approximately 3 miles apart ( with a travel time of about 20 minutes during normal traf\ufb01c ) . We also create a direct measure of geo - graphic distance by geocoding exact locations of of\ufb01ces and calculating pairwise distances in miles . In relation to scienti\ufb01c or intellectual distance , we create indicator variables for Both imagers , One imager \u00fe one clinician , and Both clinicians . We construct this variable using the information attendees themselves reported during the initial stage of the application process . We also con - structed indicator variables for Same clinical area and Same imaging modality ( physiological MR , PET , or optical ima - ging ) . These were coded from the statement of Interest documents submitted in the \ufb01rst stage of the application process . We also create measures of scienti\ufb01c distance using overlap in the Medical Subject Heading ( MeSH ) terms from each individual\u2019s publications and overlaps in the keywords of each individual\u2019s statement of interest . 18 A \ufb01nal measure of distances is whether the pair had pre - viously collaborated , indicator variable Previous coauthors . We also distinguish cases of one single past copublication with more than one past copublication with indicator variables . V . Analysis and Results A . Does Reducing Search Costs Increase the Propensity to Collaborate ? We \ufb01rst analyze whether our 90 - minute breakout treat - ment had an effect on the incidence of collaborations and the magnitude of any such effects . OLS estimates with robust standard errors are presented in table 4 . ( The same results are presented using probit estimation in table A2 . ) Column 1 shows the basic result , regressing the incidence of collabora - tions on our treatment effect indicator and a constant . The baseline probability of collaborations is captured by the con - stant coef\ufb01cient of 0 . 0016 or 0 . 16 % . The point estimate shows that the treatment increases the likelihood of collabor - ating on an application by approximately 75 % ( increasing T ABLE 4 . \u2014M AIN E FFECT OF T REATMENT ON C OLLABORATION DV \u00bc Collaboration ( 1 ) ( 2 ) ( 3 ) Same Room 0 . 0012 * 0 . 0012 * 0 . 0014 * ( 0 . 0007 ) ( 0 . 0007 ) ( 0 . 0007 ) One postdoc (cid:2) 0 . 0008 ( 0 . 0005 ) Both postdocs (cid:2) 0 . 0014 * * ( 0 . 0007 ) One is female 0 . 0002 ( 0 . 0005 ) Both are female 0 . 0010 ( 0 . 0011 ) Same hospital 0 . 0042 * * * ( 0 . 0010 ) Both Longwood (cid:2) 0 . 0001 ( 0 . 0007 ) One imager \u00fe one clinician 0 . 0008 * ( 0 . 0005 ) Both imagers 0 . 0025 * * ( 0 . 0010 ) Same clinical area ( SOI ) 0 . 0042 * * * ( 0 . 0014 ) Previous coauthor 0 . 1176 * * ( 0 . 0468 ) Constant 0 . 0016 * * * 0 . 0012 * * * (cid:2) 0 . 0010 ( 0 . 0003 ) ( 0 . 0004 ) ( 0 . 0007 ) Night \ufb01xed effects No Yes Yes R 2 0 . 000 0 . 000 0 . 017 Number of observations 26 , 664 26 , 664 26 , 664 The unit of analysis is a dyad of researchers . We construct dyads by creating every possible pairwise combination of researchers attending on the same night ( 26 , 604dyads ) . Thedependent variable is Colla - boration , an indicator variable for whether the pair appeared on any common pilot grant or concept award applications . The main variable of interest is Same room , which was randomized across pairs attending on the same night . All estimation is by OLS . Grouped dyadic standard errors in parentheses . * p < 0 . 10 , * * p < 0 . 05 , * * * p < 0 . 01 . 17 The LMA includes eight hospitals or institutes in our sample , and the MGH campus includes two hospitals or institutes . The other hospitals or institutes in the sample are considered to be individual campuses . 18 We include these other measures of scienti\ufb01c distance in our regres - sion analysis , but since these measures rely on prior publications ( and some individuals in the sample have no or few publications ) , our pre - ferred measure is self - reported clinical area . 572 THE REVIEW OF ECONOMICS AND STATISTICS the likelihood of a pair collaborating from 0 . 16 % to 0 . 28 % ) . 19 The estimate is signi\ufb01cant at the 10 % level . The advanced imaging symposia were held on three differ - ent nights . We thus include \ufb01xed effects for the night of the event ( January 31 , February 1 , or February 2 ) in column 2 to account for any differences across nights . The night \ufb01xed effects are not signi\ufb01cant , and their inclusion has very little impact on the same room coef\ufb01cient ( or its standard error ) . In column 3 we introduce pair - level variables to account for gender composition , differences in rank , as well as geo - graphic , scienti\ufb01c , and past coauthoring distance . The ran - dom assignment ensures that being in the same room is orthogonal asymptotically to any observable or unobserva - ble pair characteristic . 20 Correspondingly , introducing co - variates does not statistically change the estimated treatment effect . The standard error is not palpably changed , but the signi\ufb01cance marginally increases on account of a small increase in the point estimate . The point estimate for the effect of being in the same room increases slightly from 0 . 0012 to 0 . 0014 . ( Note that we also include additional con - trols in other speci\ufb01cations , including dummies for whether a pair was in the same group within a breakout room\u2014 group 1 or 2\u2014and their proximity to one another in the room\u2014whether the pair had posters next to each other\u2014 but the results do not change . ) We conclude this section by brie\ufb02y discussing the sign of the point estimates of the control variables in table 4 , column 3 . Working in the same clinical area , being af\ufb01liated with the same hospital , and being a coauthor in the past are positively and signi\ufb01cantly correlated with collaboration . Consistent with the related literature , these results suggest that geo - graphic , scienti\ufb01c , and past coauthoring are all positively related to collaboration . Pairs of one imager and one clinician were signi\ufb01cantly more likely to collaborate than pairs of clinicians only , but collaborations were even more likely to form when both members of the pair were imagers . Colla - boration was signi\ufb01cantly less likely to occur between pairs consisting of two postdocs , which is possibly explained by the fact that two postdocs could collaborate on a pilot grant appli - cation only if a third team member with a faculty appointment assumed the role of principal investigator . Overall , the single largest correlate is whether scientists had previously coau - thored a publication . This association is at least an order of magnitude larger than for each of the other correlates . Therefore , our estimated treatment effect of being in the same breakout room on collaboration is over 30 % of the effect of being from the same hospital ( 0 . 0044 ) and of researching the same clinical area ( 0 . 0040 ) . Relative to the single most important correlate , past coauthorship , it is only about 1 % of the magnitude ( 0 . 1126 ) . ( The probit estimates in table A2 of the appendix show similar results . ) B . For Which Pairs Does Reducing Search Costs Have the Greatest Effect ? Next , we investigate whether the treatment had an effect for different types of pairs . Unlike earlier estimates of cor - relations with covariates , interaction terms can be inter - preted causally . Probit estimates are reported in table A3 of the appendix . We introduce the interactions between co - variates with the treatment effect individually in columns 1 to 7 and then simultaneously in column 8 of table 5 . In introducing each of the interaction terms , we also of course reintroduce the direct effect of the covariate in the regres - sions ; however , our focus here is on interactions terms . In the results of columns 1 through 7 , the only interaction term found to be signi\ufb01cant is that in column 6 , which reports a positive and signi\ufb01cant coef\ufb01cient on the interac - tion between the treatment effect and the indicator for researchers being in the same clinical area . The coef\ufb01cient on the direct treatment effect term Same Room becomes sta - tistically indistinguishable from 0 when introducing this interaction . Results in column 8 corroborate this result , as introducing all covariates and all interaction terms at once in the model produces an almost identical estimate on this interaction term . In column 8 , which includes all interac - tions , the treatment increases the likelihood of collaborating for pairs researching the same clinical area from 0 . 35 % to 0 . 94 % relative to pairs researching different clinical areas . 21 There are several possible explanations for the effect , but it suggests that researchers had limited information about these potential collaborations\u2014about either the potential interest of other researchers in certain types of projects or the potential bene\ufb01ts of collaborating with these indivi - duals . If they did , the information provided at the event should not have any independent effect for these pairs . It may also be the case that discussions were more bene\ufb01cial for clinically proximate pairs because they shared common ground , allowing them to convert their discussions into col - laborations . Another possible explanation is that it is quite costly to switch clinical areas ( specialization and training in medicine occurs on the basis of clinical areas\u2014e . g . , dermatology , neurology , oncology ) , and therefore , even if researchers talked to people with interesting ideas in other clinical areas at the event , the bene\ufb01ts to collaboration were highest for those in the same clinical area . We fail to detect evidence of the signi\ufb01cance of other inter - actions . Our results on the interaction between being in the same room at the event and other pair characteristics are not 19 However , our point estimates regarding the magnitude of the effect are imprecise . The con\ufb01dence interval ranges from 4 % to 112 % . 20 Being in the same room is orthogonal to pair characteristics ex ante . However , ex post , being in the same room at the event could be correlated with pair characteristics by chance . While this is much less of a concern than in observational data ( Leamer , 2010 ) , it is nonetheless useful to con - trol for relevant , observable pair characteristics to address the possibility that the effect of being in the same room is affected by differences in observable pair characteristics . Introducing controls has the added bene\ufb01t of improving the precision of the Same room estimate by reducing the unexplained variance . 21 The sample average incidence of collaboration for pairs working in the same clinical area but not in the same room is 0 . 35 . 573 SEARCH COSTS AND THE FORMATION OF SCIENTIFIC COLLABORATIONS conclusive . While the point estimates for some interactions are positive , they are not signi\ufb01cant up to the 10 % level . 22 We also investigated various alternative speci\ufb01cations such as including more \ufb01nely grained measures of geo - graphic distance , scienti\ufb01c distance , or past coauthoring , 23 as well as controlling more \ufb02exibly for ranks and rank dif - ferences between pair members . We included dummies for whether a pair was in the same group ( 1 or 2 ) , proximity in the room ( whether the pair had posters next to each other ) , and the number of total individuals in the room ( to test whether density mattered ) , but the results were not signi\ufb01 - cant , and the same clinical area result is consistent and stable across these speci\ufb01cations . VI . Summary and Conclusions Teams are a primary unit of knowledge production , and scientists in large part self - organize into research teams . Yet we know little regarding the matching of scientists into teams . In this paper , we present the results of a \ufb01eld experi - ment to investigate the role of search costs in the formation of scienti\ufb01c teams by comparing the incidence of collabora - T ABLE 5 . \u2014T REATMENT AND I NTERACTIONS WITH M EASURES OF D ISTANCE DV \u00bc Collaboration ( 1 ) ( 2 ) ( 3 ) ( 4 ) ( 5 ) ( 6 ) ( 7 ) ( 8 ) Same Room 0 . 0008 0 . 0000 0 . 0008 0 . 0018 * 0 . 0012 (cid:2) 0 . 0000 0 . 0011 (cid:2) 0 . 0019 ( 0 . 0009 ) ( 0 . 0009 ) ( 0 . 0006 ) ( 0 . 0009 ) ( 0 . 0009 ) ( 0 . 0006 ) ( 0 . 0007 ) ( 0 . 0014 ) One postdoc (cid:2) 0 . 0016 * * * (cid:2) 0 . 0012 * * ( 0 . 0006 ) * ( 0 . 0005 ) Same room (cid:3) One postdoc 0 . 0015 0 . 0019 ( 0 . 0013 ) ( 0 . 0013 ) Both postdocs (cid:2) 0 . 0016 * (cid:2) 0 . 0012 ( 0 . 0008 ) ( 0 . 0008 ) Same room (cid:3) Both postdocs (cid:2) 0 . 0014 (cid:2) 0 . 0011 ( 0 . 0011 ) ( 0 . 0011 ) One is female (cid:2) 0 . 0008 (cid:2) 0 . 0004 ( 0 . 0006 ) ( 0 . 0006 ) Same room (cid:3) One female 0 . 0021 0 . 0024 ( 0 . 0016 ) ( 0 . 0015 ) Both are female (cid:2) 0 . 0007 (cid:2) 0 . 0000 ( 0 . 0013 ) ( 0 . 0013 ) Same room (cid:3) Both female 0 . 0040 0 . 0035 ( 0 . 0025 ) ( 0 . 0025 ) Same hospital 0 . 0044 * * * 0 . 0037 * * * ( 0 . 0014 ) ( 0 . 0013 ) Same room (cid:3) Same hospital 0 . 0024 0 . 0023 ( 0 . 0026 ) ( 0 . 0025 ) Both Longwood 0 . 0005 0 . 0004 ( 0 . 0007 ) ( 0 . 0007 ) Same room (cid:3) Both Longwood (cid:2) 0 . 0020 (cid:2) 0 . 0018 ( 0 . 0014 ) ( 0 . 0015 ) One imager \u00fe one clinician 0 . 0009 \u00fe 0 . 0008 * ( 0 . 0005 ) ( 0 . 0004 ) Same room (cid:3) One imager (cid:2) 0 . 0000 (cid:2) 0 . 0000 ( 0 . 0014 ) ( 0 . 0012 ) Both imagers 0 . 0031 * * * 0 . 0025 * ( 0 . 0011 ) ( 0 . 0011 ) Same room (cid:3) Both imager 0 . 0002 (cid:2) 0 . 0002 ( 0 . 0026 ) ( 0 . 0025 ) Same clinical area ( SOI ) 0 . 0025 * 0 . 0018 ( 0 . 0011 ) ( 0 . 0012 ) Same room (cid:3) Same clin area 0 . 0095 * * 0 . 0094 * * ( 0 . 0045 ) ( 0 . 0042 ) Previous coauthor 0 . 0917 * 0 . 0889 * * ( 0 . 0450 ) ( 0 . 0442 ) Same room (cid:3) Prev coauthor 0 . 2389 0 . 2363 ( 0 . 1981 ) ( 0 . 1967 ) Night \ufb01xed effects Yes Yes Yes Yes Yes Yes Yes Yes R 2 0 . 001 0 . 000 0 . 003 0 . 000 0 . 001 0 . 002 0 . 019 0 . 024 Number of observations 26 , 664 26 , 664 26 , 664 26 , 664 26 , 664 26 , 664 26 , 664 26 , 664 The unit of analysis is a dyad of researchers . We construct dyads by creating every possible pairwise combination of researchers attending on the same night ( 26 , 604 dyads ) . The dependent variable is Collabora - tion , an indicator variable for whether the pair appeared on any common pilot grant or concept award applications . The main variable of interest is Same room , which was randomized across pairs attending on the samenight . Allestimationis byOLS . Groupeddyadicstandarderrorsin parentheses . * p < 0 . 10 , * * p < 0 . 05 , * * * p < 0 . 01 . 22 The interaction between being in the same room and pairs with one woman are marginally signi\ufb01cant with p - values of 0 . 093 in the probit speci\ufb01cation and 0 . 133 in the OLS speci\ufb01cation . A differential effect for pairs with a woman would be consistent with the \ufb01ndings of Ding et al . ( 2010 ) , who show that the introduction of information technology bene - \ufb01ted collaborations more for female scientists than for male scientists , since women tend to have less diverse networks , lower job mobility , and more constraints to attending conferences and seminars . These factors would similarly lead women to bene\ufb01t more from mixing with other researchers at the event in terms of \ufb01nding coauthors . 23 We considered , for instance , whether pair members investigated the same imaging modality , the extent of the overlap of scienti\ufb01c keywords in previous publications , and whether pair members shared a common coauthor . 574 THE REVIEW OF ECONOMICS AND STATISTICS tions among researchers who participated in the same breakout rooms within an interactive research symposium as part of a grant proposal process versus those who were assigned to different breakout rooms . We thus randomly varied search costs for a set of prospective collaborators , observing both the collaborations that did form along with those that did not . We \ufb01nd that the small , focused treatment signi\ufb01cantly increased the incidence of collaboration on subsequent grant proposals by 75 % in relation to the baseline probabil - ity of collaboration between pairs of researchers ( increasing from 0 . 16 % in the control group to 0 . 28 % in the treated group ) . The magnitude of this effect is equivalent to roughly a third of the boost in the probability of collabora - tion associated with working in the same hospital or , alter - natively , the probability of working in the same clinical area . In this regard , the point estimate can be viewed as rather large , despite the relatively small and focused nature of the treatment : a 90 - minute breakout session . It is in fact notable that we \ufb01nd any effect at all , let alone such a large effect in the context of scientists who are already geograph - ically proximate and working within a common institutional context , where online resources and information systems already exist to facilitate collaboration . We interpret these large effects as showing that even when working in relatively favorable conditions , search costs and frictions continue to powerfully shape ( and limit ) the formation of collaborations between scientists . Whereas a great deal of collaborative work might potentially be per - formed at a distance , the formation of collaborations appears to be highly sensitive to information - rich face - to - face interactions . In this sense , the question of the \u2018\u2018death of distance\u2019\u2019 and the role of collocation and information tech - nology , for example , might be reconsidered at least in rela - tion to questions of forming collaborations . The \ufb01nding is consistent with the complex and manifold set of variables on which collaboration decisions might be based and the effectiveness of face - to - face interactions in rapidly conveying information through high - frequency rapid feedback and visual and nonverbal cues ( Gaspar & Glaeser , 1998 ; Storper & Venables , 2004 ) . For example , given our existing communications technologies , it may remain dif\ufb01cult to wholly codify current research interests , complementarity in knowledge and skills , access to resources , and timing and scheduling constraints , let alone questions of personal chemistry and disposition or subtler questions of one\u2019s intellectual outlook . The result is also consistent with face - to - face interactions potentially trigger - ing or credibly signaling commitments , establishing trust and personal chemistry ( Azoulay et al . , 2009 ) . Further consistent with the role of search costs in our results , the treatment effect was most pronounced on sub - sets of scientist pairs who are less distant , working within the same clinical area , and therefore perhaps needing to overcome lower information and search cost hurdles . We also found positive associations between the likelihood of forming collaborations as prospective collaborators coming from the same hospital , and the single most important pre - dictor of collaborations , in terms of coef\ufb01cient magnitude , was whether individuals had previously collaborated . In documenting an important role played by search costs in in\ufb02uencing the formation of collaborations , we leave open a range of related questions . For example , in this paper , we did not study or observe longer - run outcomes of scienti\ufb01c productivity such as subsequent publications . ( Initial analysis of reviewers\u2019 assessments of the grant applications indicates no statistical difference between scores of applications submitted by pairs in the same room versus pairs not in the same room at the event ; see appendix table A4 . ) Also , we demonstrate what are arguably large effects of the particular treatment we implemented here . However , the treatment exploited here is not necessarily optimal and could be subject to further improvement . Such insights could be relevant in devising improved means of designing supporting information systems and matching facilities . An additional and potentially rather important series of questions falling outside the scope of this study concerns how individuals develop their own stock of matching - relevant information and heuristics in the \ufb01rst place ( apart from effects of situational or episodic shocks in information , as were explored here ) . The patterns documented here also raise questions regarding the extent to which homophily ( Lazarsfeld & Merton , 1954 ; Baccara & Yariv , 2013 ) exempli\ufb01ed by increased likelihood for scientists to form ties with other scientists possessing similar personal characteristics , might , at least in large part , be the result of search costs\u2014rather than re\ufb02ecting collaboration preferences ( Boudreau & Lakhani , 2012 ) or lower coordination costs when collabor - ating with similar partners ( Reagans & Zuckerman , 2001 ; Reagans , Zuckerman , & McEvily , 2004 ) . Despite these limitations , we see this study as a step toward opening the black box of how scienti\ufb01c collabora - tions form . In recent years , there has been considerable interest in the policy arena in fostering collaborations , and especially interdisciplinary collaborations , in particular by the U . S . government agencies funding fundamental research and development ( a combined budget of $ 36 bil - lion in 2011 ) , the NIH and the National Science Founda - tion . Yet there is scant evidence indicating how to do this in practice . On a methodological level , we are\u2014to the best of our knowledge\u2014the \ufb01rst to bring \ufb01eld experimental meth - ods to a workplace setting where the participants are engaged in scienti\ufb01c knowledge production . Evidence from randomized experiments on the scienti\ufb01c community such as ours will presumably be increasingly valuable to policy - makers as they consider reforms to scienti\ufb01c institutions ( Azoulay , 2012 ) , and more generally our study provides a template for the design of randomized control trials in inno - vation research ( Boudreau & Lakhani , 2016 ) . We show that creating settings where scientists meet face - to - face to dis - cuss early - stage research ideas can be useful for fostering 575 SEARCH COSTS AND THE FORMATION OF SCIENTIFIC COLLABORATIONS collaboration . However , time spent in such \u2018\u2018mixer\u2019\u2019 events has opportunity costs , and we thus remain agnostic on the effect of such activities on scienti\ufb01c productivity and on welfare more generally . REFERENCES Adams , James D . , Grant C . Black , J . Roger Clemmons , and Paula E . Stephan , \u2018\u2018Scienti\ufb01c Teams and Institutional Collaborations : Evi - dence from U . S . Universities , \u2019\u2019 Research Policy 34 ( 2005 ) , 259 \u2013 285 . Agrawal , Ajay , and Avi Goldfarb , \u2018\u2018Restructuring Research : Communica - tion Costs and the Democratization of University Innovation , \u2019\u2019 American Economic Review 98 ( 2008 ) , 1578 \u2013 1590 . Agrawal , Ajay , Avi Goldfarb , and Florenta Teodoridis , \u2018\u2018Does the Accu - mulation of Knowledge Cause an Increase in Returns to Collabora - tion ? Evidence from the Collapse of the Soviet Union , \u2019\u2019 American Economic Journal : Applied Economics 8 ( 2016 ) 100 \u2013 128 . Azoulay , Pierre , \u2018\u2018Research Ef\ufb01ciency : Turn the Scienti\ufb01c Method on Ourselves , \u2019\u2019 Nature 484 ( 2012 ) , 31 \u2013 32 . Azoulay Pierre , Christopher C . Liu , and Toby E . Stuart , \u2018\u2018Social In\ufb02uence Given ( Partially ) Deliberate Matching : Career Imprints in the Creation of Academic Entrepreneurs , \u2019\u2019 HBS working paper 9 - 136 ( 2009 ) . Baccara , Mariagiovanna , and Leeat Yariv , \u2018\u2018Homophily in Peer Groups , \u2019\u2019 American Economic Journal : Microeconomics 5 ( 2013 ) , 69 \u2013 96 . Boschini , Anne , and Anna Sjogren ; \u2018\u2018Is Team Formation Gender Neutral ? Evidence from Coauthorship Patterns , \u2019\u2019 Journal of Labor Econom - ics 25 ( 2007 ) , 325 \u2013 365 . Boudreau , Kevin J . , and Karim R . Lakhani , \u2018\u2018The Confederacy of Hetero - geneous Software Organizations and Heterogeneous Developers : Field Experimental Evidence on Sorting and Worker Effort\u2019\u2019 ( pp . 483 \u2013 502 ) , in Josh Lerner and Scott Stern , eds . , The Rate and Direction of Inventive Activity Revisited ( Chicago : University of Chicago Press , 2012 ) . \u2014\u2014\u2014 \u2018\u2018Innovation Experiments : Researching Technical Advance , Knowledge Production and the Design of Supporting Institutions , \u2019\u2019 in Josh Lerner and Scott Stern , eds . , Innovation Policy and the Economy ( Chicago : University of Chicago Press , 2016 ) . Cameron , A . Colin , and Douglas L . Miller \u2018\u2018A Practitioner\u2019s Guide to Cluster - Robust Inference , \u2019\u2019 Journal of Human Resources 50 ( 2015 ) , 317 \u2013 372 . Catalini , Christian , \u2018\u2018Microgeography and the Direction of Inventive Activity , \u2019\u2019 Rotman School of Management working paper 2126890 ( 2016 ) . Catalini , Christian , Christian Fons - Rosen , and Patrick Gaule , \u2018\u2018Did Cheaper Flights Change the Geography of Scienti\ufb01c Collabora - tion ? \u2019\u2019 MIT Sloan research paper 5172 - 16 ( 2016 ) . Currarini , Sergio , Matthew O . Jackson , and Paolo Pin , \u2018\u2018An Economic Model of Friendship : Homophily , Minorities , and Segregation , \u2019\u2019 Econometrica 77 ( 2009 ) , 1003 \u2013 1045 . Ding , Waverly W . , Sharon G . Levin , Paula E . Stephan , and Anne E . Winkler , \u2018\u2018The Impact of Information Technology on Academic Scientists\u2019 Productivity and Collaboration Patterns , \u2019\u2019 Management Science 56 ( 2010 ) , 1439 \u2013 1461 . Fafchamps , Marcel , Sanjeev Goyal , and Marco J . Van der Leij , \u2018\u2018Matching and Network Effects , \u2019\u2019 Journal of the European Eco - nomic Association 8 ( 2010 ) , 203 \u2013 231 . Fafchamps , Marcel , and Flore Gubert , \u2018\u2018The Formation of Risk Sharing Networks , \u2019\u2019 Journal of Development Economics 83 ( 2007 ) , 326 \u2013 350 . Freeman , Richard B . , Ina Ganguli , and Raviv Murciano - Goroff , \u2018\u2018Why and Wherefore of Increased Scienti\ufb01c Collaboration , \u2019\u2019 in Adam Jaffe and Ben Jones , eds . , The Changing Frontier : Rethinking Science and Innovation Policy ( Cambridge , MA : National Bureau of Economic Research , 2015 ) . Freeman , Richard B . , and W . Huang , \u2018\u2018Collaborating with People Like Me : Ethnic Co - Authorship within the US , \u2019\u2019 NBER working paper 19905 ( 2014 ) . Gaspar , Jess , and Edward L . Glaeser , \u2018\u2018Information Technology and the Future of Cities , \u2019\u2019 Journal of Urban Economics 43 ( 1998 ) , 136 \u2013 156 . Glaeser , Edward L . , Agglomeration Economics ( Chicago : University of Chicago Press , 2010 ) . Jones , Benjamin F . , \u2018\u2018The Burden of Knowledge and the \u2018Death of the Renaissance Man\u2019 : Is Innovation Getting Harder ? \u2019\u2019 Review of Eco - nomic Studies 76 ( 2009 ) , 283 \u2013 317 . Jones , Benjamin F . , Stefan Wuchty , and Brian Uzzi , \u2018\u2018Multi - University Research Teams : Shifting Impact , Geography , and Strati\ufb01cation in Science , \u2019\u2019 Science 322 ( 2008 ) , 1259 . Lazarsfeld , Paul F . , and Robert K . Merton , \u2018\u2018Friendship as a Social Pro - cess : A Substantive and Methodological Analysis , \u2019\u2019 Freedom and Control in Modern Society 18 : 1 ( 1954 ) , 18 \u2013 66 . Lazear , Edward P . , and Kathrin L . Shaw , \u2018\u2018Personnel Economics : The Economist\u2019s View of Human Resources , \u2019\u2019 Journal of Economic Perspectives 21 ( 2007 ) , 91 \u2013 114 . Leamer , Edward E . , \u2018\u2018Tantalus on the Road to Asymptotia , \u2019\u2019 Journal of Economic Perspectives 24 ( 2010 ) , 31 \u2013 46 . McPherson , Miller , Lynn Smith - Lovin , and James M . Cook , \u2018\u2018Birds of a Feather : Homophily in Social Networks , \u2019\u2019 Annual Review of Sociology 27 ( 2001 ) , 415 \u2013 444 . Mortensen , Dale T . , and Christopher A . Pissarides , \u2018\u2018New Developments in Models of Search in the Labor Market\u2019\u2019 ( pp . 2567 \u2013 2627 ) , in Orley C . Ashenfelter and David Card , eds . , Handbook of Labor Economics , vol . 3b ( Amsterdam : Elsevier , 1999 ) . Reagans , Ray , and Ezra W . Zuckerman , \u2018\u2018Networks , Diversity , and Pro - ductivity : The Social Capital of Corporate R & D Teams , \u2019\u2019 Organi - zation Science 12 ( 2001 ) , 502 \u2013 517 . Reagans , Ray , Ezra W . Zuckerman , and Bill McEvily , \u2018\u2018How to Make the Team : Social Networks vs . Demography as Criteria for Designing Effective Teams , \u2019\u2019 Administrative Science Quarterly 49 ( 2004 ) , 101 \u2013 133 . Rosenthal , Stuart S . , and William C . Strange , \u2018\u2018The Determinants of Agglomeration , \u2019\u2019 Journal of Urban Economics 50 ( 2001 ) , 191 \u2013 229 . Stephan , Paula , How Economics Shapes Science ( Cambridge , MA : Har - vard University Press , 2012 ) . Storper , Michael , and Anthony J . Venables , \u2018\u2018Buzz : Face - to - Face Contact and the Urban Economy , \u2019\u2019 Journal of Economic Geography 4 ( 2004 ) , 351 \u2013 370 . Wuchty Stefan , Benjamin F . Jones , and Brian Uzzi , \u2018\u2018The Increasing Dominance of Teams in Production of Knowledge , \u2019\u2019 Science 316 ( 2007 ) , 1036 . 576 THE REVIEW OF ECONOMICS AND STATISTICS",
    "paletzUncoveringUncertaintyDisagreement2016": "Uncovering Uncertainty through Disagreement SUSANNAH B . F . PALETZ 1 * , JOEL CHAN 2 and CHRISTIAN D . SCHUNN 3 1 Center for Advanced Study of Language , University of Maryland , College Park , College Park , USA 2 Human \u2013 Computer Interaction Institute , Carnegie Mellon University , Pittsburgh , USA 3 Learning Research and Development Center , University of Pittsburgh , Pittsburgh , USA Summary : This study explored the association between different types of brief disagreements and subsequent levels of expressed psychological uncertainty , a fundamental cognitive aspect of complex problem solving . We examined 11hours ( 11861 utterances ) of conversations in expert science teams , sampled across the \ufb01 rst 90days of the Mars Exploration Rover mission . Utterances were independently coded for micro - con \ufb02 icts and expressed psychological uncertainty . Using time - lagged hierarchical linear modeling applied to blocks of 25 utterances , we found that micro - con \ufb02 icts regarding rover planning were followed by greater uncertainty . Brief disagreements about science issues were followed by an increase in expressed uncertainty early in the mission . Examining the potential reverse temporal association , uncertainty actually predicted fewer subsequent disagreements , ruling out indirect , third variable associations of con \ufb02 ict and uncertainty . Overall , these \ufb01 ndings suggest that some forms of disagreement may serve to uncover important areas of uncertainty in complex teamwork , perhaps via revealing differences in mental models . Copyright \u00a9 2016 John Wiley & Sons , Ltd . INTRODUCTION The most important scienti \ufb01 c and technological advances are increasingly being produced by teams ( Jones , 2009 ; Singh & Fleming , 2009 ; Wuchty , Jones , & Uzzi , 2007 ) . Organiza - tions are also looking to interdisciplinary teams to address especially dif \ufb01 cult global problems ( Beck , 2013 ; Derry , Schunn , & Gernsbacher , 2005 ) . However , interdisciplinary teams have challenges : they must communicate across un - shared norms and mental models , which may result in dis - agreement and gaps in understanding ( Paletz & Schunn , 2010 ) . Furthermore , real - world problems are notable for their high levels of uncertainty . Problem solving with com - plex , ill - de \ufb01 ned problems often depends on signi \ufb01 cant effort devoted to the detection and resolution of uncertainty ( Schunn & Trafton , 2012 ) : Detection strategies are necessary to identify what is unknown ( either because information is missing or because presented information is misleading ) , and resolution of the uncertainty is important for ultimately solving the complex problem . This project studies how real - world science teams deal with uncertainty in problem solving . By taking a micro - process approach , this study examines how and whether brief intrateam disagreements can help teams detect uncer - tainty and then compares the association between disagree - ments and subsequent uncertainty in conversations at two different time phases within a team \u2019 s life . The past few de - cades have seen an increased interest in examining how teams \u2019 problem - solving processes unfold over time ( Cronin , Weingart , & Todorova , 2011 ; Kurtzberg & Mueller , 2005 ; McGrath , 1991 ; McGrath & Tschan , 2004 ) . At the most micro - temporal level ( at the timescale of seconds ) , brief speech acts , such as disagreement , may cumulate to have large effects on subsequent communication and cognition ( e . g . , Chiu , 2008a , 2008b ; Paletz , Schunn , & Kim , 2013 ) . Uncertainty Research in cognitive psychology has revealed that uncer - tainty is a central aspect of complex problem solving ( for a review , see Schunn & Trafton , 2012 ) . Psychological uncer - tainty is the recognition or feeling of missing , vague , or incomplete information ( Schunn , 2010 ) . Uncertainty is ubiq - uitous within and central to real - world problem solving , including science and engineering , management , medicine , and education ( Kahneman & Tversky , 1982 ; Schunn , 2010 ) . The research \ufb01 eld of naturalistic decision making unpacks in - dividual and team judgment in uncertain situations such as avi - ation ( e . g . , Bearman , Paletz , Orasanu , & Thomas , 2010 ; Klein , 1989 ) . Indeed , one of the main differences between real - world and experimental tasks is the level of uncertainty , as well as uncertainty about uncertainty , in real - world tasks ( Chinn & Malhotra , 2002 ; Kirschenbaum , Trafton , Schunn , & Trickett , 2014 ; Schunn & Trafton , 2012 ) . Across real - world domains , problem - solving success de - pends on the effective detection , understanding , and resolu - tion of uncertainty ( Chan , Paletz , & Schunn , 2012 ; Downey & Slocum , 1975 ; Kahneman & Tversky , 1982 ; Schunn , 2010 ; Schunn & Trafton , 2012 ) . For example , product devel - opment leaders have to deal with design uncertainties , while doctors have to deal with uncertainties regarding the true causes or diseases associated with a group of symptoms . People productively deal with task - related uncertainty in many ways ( Jordan & Babrow , 2013 ; Lipshitz & Strauss , 1997 ; Schunn & Trafton , 2012 ) , especially by taking it into account in decisions ( e . g . , avoiding irreversible action and explicitly noting con \ufb01 dence levels ) , communicating directly about uncertainty , or reducing uncertainty through additional problem solving ( e . g . , collecting additional information and making analogies ; Chan et al . , 2012 ) . Although the importance of the resolution of uncertainty is obvious to problem solving , uncertainty detection is also important . For instance , initial diagnostic certainty can be correct or incorrect . If incorrect , raising uncertainty about the initial idea will enable the re - examination of data and en - courage information search . Indeed , past literature suggests * Correspondence to : Susannah B . F . Paletz , Center for Advanced Study of Language , University of Maryland , College Park , College Park , MD , USA . E - mail : sbfpaletz @ gmail . com Copyright \u00a9 2016 John Wiley & Sons , Ltd . Applied Cognitive Psychology , Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) Published online 17 February 2016 in Wiley Online Library ( wileyonlinelibrary . com ) DOI : 10 . 1002 / acp . 3213 that individuals are often bad at articulating their assumptions and identifying genuine uncertainties , especially in complex situations ( Dunbar , 1997 ; Evans , 1989 ; Gorman , 1986 ) . Uncertainty is also important in the earliest stages of problem solving . Problem \ufb01 nding , or how a problem is initially discovered and structured , is a vital precursor to problem solving ( Mumford , Reiter - Palmon , & Redmond , 1994 ; Paletz & Peng , 2009 ) , especially in ill - de \ufb01 ned domains where un - certainty abounds ( Runco , 1994 ; Runco & Nemiro , 1994 ) . Put simply , teams and individuals need to know what they do not know in order to move forward and innovate . Con \ufb02 ict Researchers from different \ufb01 elds have examined con \ufb02 ict in many different forms , from violent acts perpetrated on behalf of nations to con \ufb02 icts within oneself . We examine con \ufb02 ict as disagreements between individuals within teams , drawing from a tradition within social and organizational psychology ( Barki & Hartwick , 2004 ) . Teams are collections of individ - uals that perform interdependent tasks that affect others , are embedded in a larger social system , and perceive themselves , and are perceived by others , as a social entity ( Guzzo & Dickson , 1996 ) . Intrateam con \ufb02 ict can vary in its focus , length , and intensity , and these variations can change the effects of con \ufb02 ict on team processes ( Paletz , Schunn , & Kim , 2011 ; Todorova , Bear , & Weingart , 2014 ; Weingart , Behfar , Bendersky , Todorova , & Jehn , 2014 ) . Under certain conditions , con \ufb02 ict has the potential to increase creativity ( e . g . , Chiu , 2008a ; Nemeth , 1986 ; Nemeth , Personnaz , Personnaz , & Goncalo , 2004 ; Paletz , Miron - Spektor , & Lin , 2014 ) . However , it can also hinder creativity , including when the con \ufb02 ict is only observed by another ( Miron - Spektor , Efrat - Treister , Rafaeli , & Schwartz - Cohen , 2011 ) . This study focuses on micro - con \ufb02 icts , or brief , expressed disagreements , in order to unpack and illuminate team pro - cesses in detail . Prior research on micro - con \ufb02 icts has exam - ined complex interaction patterns and con \ufb02 ict management ( e . g . , Kuhn & Poole , 2000 ; Poole & Dobosh , 2010 ) , affec - tive con \ufb02 ict interactions within families or married couples ( e . g . , Gottman & Notarius , 2000 ; Vuchinich , 1987 ) , and the immediate effects of brief disagreements on creativity , problem solving , and analogy ( e . g . , Chiu , 2008a , 2008b ; Paletz , Schunn , et al . , 2013 ) . Intragroup con \ufb02 ict researchers distinguish between task con \ufb02 ict , or disagreements about the work being performed ; process con \ufb02 ict , or how to go about doing the task ( e . g . , pri - oritization and scheduling ) ; and relationship con \ufb02 ict , which is about personal issues ( e . g . , Barki & Hartwick , 2004 ; De Dreu & Weingart , 2003 ; Jehn , 1995 , 1997 ) . Relationship con \ufb02 ict is generally thought to be harmful ( e . g . , Amason , 1996 ; De Dreu & Weingart , 2003 ; He , Ding , & Yang , 2014 ; Jehn , 1997 ; De Wit , Greer , & Jehn , 2012 ; De Wit , Jehn , & Scheepers , 2013 ) . Process con \ufb02 ict is also generally thought to be problematic ( e . g . , De Dreu & Weingart , 2003 ) , but how destructive it is may depend on when it comes during the team \u2019 s process ( e . g . , Goncalo , Polman , & Maslach , 2010 ) . Task con \ufb02 ict is now thought to have a cur - vilinear effect on performance on intellectual tasks ( creativ - ity , Fahr , Lee , & Farh , 2010 ; decision quality , Parayitam & Dooley , 2011 ) . These studies suggest that moderate levels of task con \ufb02 ict are best for team outcomes . In addition , the dynamics and implications of con \ufb02 ict may vary across a team \u2019 s life cycle . Con \ufb02 ict occurs throughout the life of a team , but it may be more common during early phases of a team \u2019 s life cycle ( Paletz et al . , 2011 ; Poole & Garner , 2006 ) . Con \ufb02 ict earlier in the team \u2019 s life may be func - tional , helping set up processes and tasks : early process con - \ufb02 ict may help forestall premature collective ef \ufb01 cacy , leading to better team performance later ( Goncalo et al . , 2010 ) . Most research on work teams examines con \ufb02 ict using retro - spective self - report surveys ( e . g . , De Dreu & Weingart , 2003 ; Fahr et al . , 2010 ; Greer , Jehn , & Mannix , 2008 ; Parayitam & Dooley , 2011 ; Shaw et al . , 2011 ) . However , self - report sur - veys or interviews fail to capture phenomena that require exact recall and are too fast or complex to be noticed by participants ( Goh , Goodman , & Weingart , 2013 ; Gottman & Notarius , 2000 ; Weingart , 1997 ) . Retrospection is often inaccurate , and con \ufb01 dence about memories may be unrelated to accuracy ( e . g . , Chua , Hannula , & Ranganath , 2012 ; Wong , Cramer , & Gallo , 2012 ) . Further , surveys lose the \ufb01 ne - grained temporal structure of communication processes that occur during con - versations . For example , micro - process research has found that polite disagreements are positively associated with micro - creativity ( Chiu , 2008a ) and subsequent correct contributions ( Chiu , 2008b ) , whereas rude disagreements after a wrong idea can increase teammate creativity ( Chiu , 2008a ) . This current study follows this growing interest in examin - ing the immediate consequences of disagreement , or micro - con \ufb02 icts , on team processes ( e . g . , Chiu , 2008a , 2008b ; Kauffeld & Lehmann - Willenbrock , 2012 ; Paletz , Schunn , et al . , 2013 ) . We examine individual utterances for the pres - ence of disagreement and then aggregate them into micro - con \ufb02 ict events . Con \ufb02 ict and subsequent uncertainty We embed our research questions within shared mental models theory to propose that expressed micro - con \ufb02 icts , or brief disagreements ( Paletz et al . , 2011 ) , may in \ufb02 uence sub - sequent uncertainty , depending on the type of con \ufb02 ict and phase of the team \u2019 s life cycle . One aspect of teams as they develop is the creation of shared mental models ( e . g . , Burke , Stagl , Salas , Peirce , & Kendall , 2006 ; Johnson - Laird , 1980 ; Mathieu , Heffner , Goodwin , Salas , & Cannon - Bowers , 2000 ) . Shared mental models are the similarity in cognitive schemas regarding teamwork , norms , and tasks . They are important for effective plan execution and timely problem solving , among other team functions ( e . g . , Burke et al . , 2006 ; Jones , Stevens , & Fischer , 2000 ) . Interdisciplinary work is often fraught with con \ufb02 ict , as in - dividuals with different backgrounds , skills , and perspec - tives are brought together ( Mannix & Neale , 2005 ; Paletz & Schunn , 2010 ) . Con \ufb02 ict may be caused by underlying differences in shared mental models ( Bearman et al . , 2010 ; Cronin & Weingart , 2007 ) . Disagreement is generally expressed because an individual thinks or feels something con - tradictory to what was just said or implied . Indeed , disagree - ment can enable team members to hear previously unshared information from team members \u2019 different mental models . 388 S . B . F . Paletz et al . Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e On the one hand , additional information gained via con - \ufb02 ict can decrease team uncertainty . Gathering additional information is , after all , an uncertainty - reduction technique ( Lipshitz & Strauss , 1997 ; Schunn & Trafton , 2012 ) . Members of multidisciplinary teams may have information that resolves others \u2019 uncertainty . New information unco - vered via con \ufb02 ict may enable team members to update their mental models and create new certainties . On the other hand , the non - overlapping mental models that arise from interdisciplinary teams also may increase un - certainty . Individuals may not see ambiguity or errors in their own thinking that are then noticed by other scientists ( Dama & Dunbar , 1996 ) . Particularly in multidisciplinary contexts , a disagreement may raise and question underlying assumptions held by other members of the team , leaving pre - viously ( supposedly ) known information to be suddenly up for debate . A period of uncertainty may then occur when mental models incorrectly thought to be shared are reassessed and re - formed . Team context may affect whether disagreement decreases subsequent uncertainty or not . In this study , we focus on temporal context ( early versus later time periods in a team \u2019 s planned life cycle ) , a contextual factor that seems especially likely to interact with team con \ufb02 ict and uncertainty . Teams often go through iterative cycles of sorting out their pro - cesses , self - evaluating , and focusing on the work ( Gersick , 1988 ; Morgan , Salas , & Glickman , 1994 ; Tuckman , 1965 ; Tuckman & Jensen , 1977 ) . Early in a team \u2019 s life cycle , there is more uncertainty and con \ufb02 ict , as processes are established and iterated ( e . g . , Paletz et al . , 2011 ; Poole & Garner , 2006 ; Tuckman & Jensen , 1977 ) . In other words , early on , the members are formulating their shared mental models of their tasks and processes . During this phase , disagreement may lead to more uncertainty . Later in a team \u2019 s life cycle , dis - agreements are less likely to reveal unknown , major underly - ing differences in assumptions and mental models . Thus , there may be different relations between con \ufb02 ict and uncer - tainty early on , when everything is in \ufb02 ux and being established , versus later in the team \u2019 s lifecycle . Finally , different types of con \ufb02 ict may have different effects on uncertainty , just as they have different impacts on team success ( De Wit et al . , 2012 ) . Task con \ufb02 ict is most relevant to exposing gaps in mental models about the task , and thus most likely to be related to subsequent uncertainty . But process con \ufb02 ict may also reveal gaps , if perhaps more indirectly . For example , because mental models about the task support speci \ufb01 c processes , arguments about processes may en - courage others to share task information about these particular processes . It is unclear whether relationship con \ufb02 icts should have an effect on uncertainty , and these will not be a focus of this study for practical reasons ( as discussed later ) . Research questions This research is one of a small number of studies to examine \ufb01 ne - grained temporal processes of con \ufb02 ict ( e . g . , Chiu , 2008a , 2008b ; Paletz , Schunn , et al . , 2013 ; Poole & Dobosh , 2010 ) and the \ufb01 rst to examine the association between con - \ufb02 ict and subsequent uncertainty . The prior literature reviewed earlier suggests an association is plausible , but the exact form is dif \ufb01 cult to predict , and it is likely depen - dent on both the time phase in the group \u2019 s life cycle and the type of con \ufb02 ict . Our research questions are as follows : Research Question 1 : What is the association between the different types of expressed micro - con \ufb02 icts and immedi - ate , subsequent levels of uncertainty in speech ? Research Question 2 : How might these relations differ during an early phase of a team \u2019 s life cycle versus a later phase ? In addition , to further probe the nature of observed associ - ations , we also examine whether the relations between micro - con \ufb02 icts and subsequent uncertainty are bidirectional or potentially caused by a third factor . We accomplish this by testing for an association between uncertainty and subse - quent micro - con \ufb02 icts . METHODS Research context and data collection Research context This study examined informal , task - relevant conversations occurring between scientists working on the multidisciplin - ary Mars Exploration Rover ( MER ) mission ( Squyres , 2005 ) . The MER mission aimed to , and succeeded in , dis - covering evidence for a history of liquid water on Mars via sending two rovers to separate sides of the planet . The mis - sion was initially scheduled to last 90 Martian days , and so the scientists spent those days working at the Jet Propulsion Laboratory with workstations and screens enabling collabo - ration . Each rover had its own , roughly independent science team ( MER A and B ) , which was further broken up into dif - ferent science sub - teams . The discovery of the history of liq - uid water involved geological and atmospheric science , complex task planning , and general human work processes . Data collection Data were collected on 20 data collection trips over the 90 initial mission days . Researchers placed video cameras on top of large shared screens near different subgroups \u2019 work - station clusters . Each trip involved 3 days of about 8 hours of data collection each day , resulting in roughly 400 h of video . The scientists became habituated to the cameras , at times discussing personal matters . Data sampling and transcription We sampled 11 hours and 25 minutes of informal , task - relevant conversation clips from days early and later in the nominal mission ( \ufb01 rst 90 days ) . Re \ufb02 ecting the common real - ities of complex work unfolding over long timescales and distributed over multiple physical workspaces , much of the 400 hours of video ended up recording individuals working in silence , empty chairs , and / or such poor audio \u2013 video that it could not be transcribed . We excluded structured meetings from the sample because they were conducted in a formal round - robin presentation style , with most participants simply listening to people talking offscreen . During the normal workday , scientists sitting at their workstations would strike Uncovering uncertainty through disagreement 389 Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e up task - relevant conversations , and these would end ( as clips ) when the conversation stopped or the speakers moved away together ( e . g . , to get food ) . Task - relevant talk was de \ufb01 ned as anything relating to the mission at all , including process , re - lational , and organizational issues ( e . g . , organizational strug - gles between subgroups ) , whereas off - task talk was de \ufb01 ned as exclusively personal matters such as family vacations . While the larger MER project consisted of over 100 scientists , \u2018 sub - teams \u2019 of 2 \u2013 10 MER scientists assembled organically through - out the project . These sub - teams were most often task focused , engaged in interdependent tasks , and had a shared social iden - tity ( as MER scientists ) . Because the sub - teams tended to be task focused and \ufb02 uid in number , rather than centrally planned , con \ufb01 gurations of individuals often overlapped imperfectly be - tween sub - team conversations ( Paletz & Schunn , 2011 ) . This study examined conversations from these sub - teams , which occurred in video clips , representing natural within - team , interpersonal con \ufb02 icts and uncertainty as they arose . The over - whelming majority of the conversations sampled ( 93 % ) were inherently multidisciplinary , as they came from video of the Long Term Planning area , the location of a cross - disciplinary group that was the hub for different groups \u2019 conversations re - garding future science activities . The video clips were transcribed into 12 336 utterances , which were then 100 % double coded as to whether they were related to the MER mission at all ( on - task ) or not ( Cohen \u2019 s kappa = 0 . 96 ) . On - task talk was de \ufb01 ned in the same way as in our clip sampling strategy . Trained transcribers broke ( unit - ized ) the conversations into utterances during the course of transcription . Utterances were operationalized as thought state - ments or clauses and are thus often briefer than turns of talk ( Chi , 1997 ) . Our analyses are conducted on the resulting 11861 on - task utterances ( roughly 11hours ) , coming from 114 clips that ranged from 8 to 760 utterances long ( M = 104 . 0 , median = 66 . 5 , SD = 121 . 8 ) . Twenty - six per cent of the video clips had a mix of male and female scientists , whereas 74 % of video clips were all male . The number of speakers in the clip ranged from 2 to 10 . These data have been previously analyzed to examine the nature of micro - con \ufb02 icts ( Paletz et al . , 2011 ) , the temporal relations between micro - con \ufb02 icts and analogy ( Paletz , Schunn , et al . , 2013 ) , and the association between analogy and uncertainty ( Chan et al . , 2012 ) . Measures The primary variables for this study were micro - con \ufb02 icts and their types , psychological uncertainty , and temporal context . To reduce noise from the dif \ufb01 cult coding task , both uncer - tainty and micro - con \ufb02 icts were double coded ( 100 % overlap in assessment by two independent coders ) with mismatches resolved through discussion ( Smith , 2000 ) . To remove the in \ufb02 uence of cross - category coding cues , each variable was coded by a different pair of independent coders . All coders were blind to the results of the other coding and also to the goals and questions of this study . Temporal context : early versus later in the \ufb01 rst 90 days of the mission The scientists were relative novices to real - time Mars opera - tions initially but gained expertise dramatically over the \ufb01 rst 90 days ( Paletz , Kim , Schunn , Tollinger , & Vera , 2013 ) . The mission was initially planned to be 90 Martian days long . We chose 50 days as our cutoff point to divide clips into ear - lier versus later in the mission because the initial few days of each mission involved tests regarding rover health . Also , the scientists experienced a dramatic speedup of their scheduled science planning over the course of these 90 days : 51 days was when one of the rover teams decreased their scheduled shift ( and planning time ) from 6 to 4 . 5 hours ( Paletz , Kim , et al . , 2013 ; Tollinger , Schunn , & Vera , 2006 ) . Early in the mission , the scientists were more likely to grapple with the mission processes . Indeed , process micro - con \ufb02 icts were less likely during the second half of the mission ( Paletz et al . , 2011 ) . Further , by the second half of the mission , not only were the scientists more practiced at running a mission on Mars , but they had also discovered strong evidence of liquid water and had granted a news conference on the topic . Sixty - \ufb01 ve per cent ( 74 ) of the conversation clips were from before Day 50 ( early phase ) , and 35 % ( 40 ) were drawn from after or on Day 50 ( Days 50 \u2013 90 , later phase ) . Finer - grained temporal analyses were not possible : the sampling was uneven owing to mission events intermittently disrupting data collection . Micro - con \ufb02 icts We coded for micro - con \ufb02 icts , or brief disagreements , using a coding scheme detailed elsewhere for this sample ( Paletz et al . , 2011 ) . This scheme involved assessing each utterance for the presence ( 1 ) or absence ( 0 ) of con \ufb02 ict , as contrasted with other schemes that code 30 - second to 1 - minute seg - ments in their entirety ( e . g . , Poole , 1989 , used in Poole & Dobosh , 2010 ) and those that involve coding by sentence into one of several exclusive categories , collectively captur - ing all types of team interactions ( e . g . , act4teams , used in Lehmann - Willenbrock , Meyers , Kauffeld , Neininger , & Henschel , 2011 ) . Each utterance was \ufb02 agged as containing a micro - con \ufb02 ict whenever a speaker explicitly disagreed with something said earlier in the conversational video clip . Simply stating something controversial was not suf \ufb01 cient , but supporting arguments were counted as con \ufb02 ict . The coders read the transcript and referred to the audio \u2013 video re - cording to resolve differences and to clarify what occurred , paying close attention to participants \u2019 body language , tone , and , when possible , facial expressions ( con \ufb02 ict event kappa = 0 . 62 ) . For example , a group of scientists were discussing the probable geological nature of a small area on Mars . One noted that they would be able to \ufb01 nd out more by looking at an outcrop closely . A second scientist disagreed , saying , \u2018 but I think you gonna see , I think if it \u2019 s there , you \u2019 re gonna see it remotely anyway \u2019 ( via viewing by satellites ) . The original scientist tried to explain his posi - tion , \u2018 I think the , uh , I mean the scale of detail \u2019 , and the sec - ond scientist continued his argument with \u2018 no , I think , because up close \u2014 up close you \u2019 re gonna see the broken , you know , jumble \u2019 . The \ufb01 rst disagreed , \u2018 nah , I think it \u2019 s worse than that \u2019 , and the second continued his thought , \u2018 if you can get it \u2019 . At that point , a third scientist laughed and said , \u2018 you \u2019 re , you \u2019 re , I think you \u2019 re both wrong . The prob - lem is from looking at it from just sort of across , I don \u2019 t think you can see close enough to see the lamination \u2019 . This exam - ple is re \ufb02 ective of the natural micro - con \ufb02 icts from this 390 S . B . F . Paletz et al . Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e sample : overlapping ideas , low affect , and spontaneous . All disagreements between coders were resolved by discussion , with a consensus code assigned to each utterance . The coding scheme was developed iteratively on this dataset , with nuances noted in the coding scheme and past clips recoded as necessary ( Paletz et al . , 2011 ) . The micro - con \ufb02 icts were also coded by utterance for type of con \ufb02 ict , using categories described widely in the social and organizational con \ufb02 ict literature ( task , process , and rela - tionship con \ufb02 ict , e . g . , Jehn , 1997 ) . Because the MER con - text clearly distinguished between the two major work task types of science ( e . g . , data interpretation ) and rover planning ( e . g . , which instrument readings to do ) , the coders assessed each con \ufb02 ict utterance as to whether the con \ufb02 ict was about science ( task ) , rover planning ( task ) , work process , or rela - tionship matters ( see Table 1 for examples ) . To reduce noise from reliability issues , the data were exhaustively double coded , and discrepancies were resolved through discussion ( Smith , 2000 ) . These utterances were clustered into con \ufb02 ict events ( and blocks made up of continuous con \ufb02 ict events , as discussed later ) based on speci \ufb01 c topic : for instance , a speci \ufb01 c process micro - con \ufb02 ict could then lead to another process con \ufb02 ict , another type of con \ufb02 ict , or no con \ufb02 ict at all . Expressed relationship con \ufb02 ict was too rare in this sam - ple to be included in the analyses ( only three of the 688 total blocks had relationship con \ufb02 ict ) ; leaving out relationship con \ufb02 ict , the con \ufb02 ict - type coding had a reliability kappa of 0 . 67 at the utterance level ( up from 0 . 48 with relationship con \ufb02 ict included , likely because the rarity of relationship con \ufb02 ict decreased the overall kappa ; Bakeman , Quera , McArthur , & Robinson , 1997 ) . Uncertainty Two coders used a preexisting coding scheme to assess each utterance for the presence of uncertainty , using \u2018 hedge words \u2019 ( e . g . , \u2018 I guess \u2019 , \u2018 I think \u2019 , \u2018 possibly \u2019 , \u2018 maybe \u2019 , \u2018 I believe \u2019 , de - tailed in Trickett , Trafton , Saner , & Schunn , 2007 ) as cues to potential uncertainty . All coder disagreements were resolved Table 1 . Examples of micro - con \ufb02 icts between Mars Exploration Rover scientists ( key words in bold ) , excerpts from Paletz et al . ( 2011 ) Coded as Utterance Example 1 : science con \ufb02 ict No con \ufb02 ict S2 : \u2026 I \u2019 m afraid that this very low angle interior wall re \ufb02 ects that No con \ufb02 ict there is no bedrock there . Science con \ufb02 ict S3 : Well , Bereonies got a low slope also . Science con \ufb02 ict It could be that the bedrock is just not very conical . Example 2 : rover planning con \ufb02 ict No con \ufb02 ict S3 : Then the issue becomes we all , we may not have enough time to do any mineralogy or this spectral stuff No con \ufb02 ict if we \u2019 re going to have a long look at the salt . No con \ufb02 ict S2 : You think of this No con \ufb02 ict as we \u2019 ve only been there for two weeks . Rover planning con \ufb02 ict There \u2019 s a time for everything now . Rover planning con \ufb02 ict How do we spend it such that Rover planning con \ufb02 ict maybe your stuff , half the peoples \u2019 stuff doesn \u2019 t get done on one sol , Rover planning con \ufb02 ict but the other half can really get something that \u2019 s good . Rover planning con \ufb02 ict S3 : Right , but I need to do the stuff I want to do , which is \u2026 Example 3 : process con \ufb02 ict ( with \u2018 no \u2019 ) No con \ufb02 ict S2 : You showed me that already . Process con \ufb02 ict S3 : No , I showed you a single frame . Example 4 : relationship con \ufb02 ict Rover planning con \ufb02 ict S1 : Well , what I \u2019 m saying is common things occur commonly Rover planning con \ufb02 ict and rare things occur rarely Rover planning con \ufb02 ict and to say that we landed on the rare thing rather than the common thing Relationship con \ufb02 ict is a totally subjective decision that we make that goes against logic . Relationship con \ufb02 ict That \u2019 s all I \u2019 m saying . Rover planning con \ufb02 ict S2 : No , I agree with that , but \u2014 [ is cut off ] Example 5 : con \ufb02 ict with no con \ufb02 ict embedded Science con \ufb02 ict S6 : But to me they look like each one of these things No con \ufb02 ict Even though they \u2019 re really wide , Science con \ufb02 ict They do sort of look like single event types of features Science con \ufb02 ict Because the sides of each of them are sort of parallel to each other Science con \ufb02 ict which you wouldn \u2019 t get if there was a swarm coming in from the same place . Example 6 : use of \u2018 no \u2019 without con \ufb02 ict No con \ufb02 ict S3 : Is there any change in the plan here ? No con \ufb02 ict S1 : No . Example 7 : use of \u2018 but \u2019 without con \ufb02 ict ( additions rather than disagreement ) No con \ufb02 ict S3 : Let me ask . No con \ufb02 ict S2 : Well , we need some , No con \ufb02 ict but I think we need to pick a number , you know , No con \ufb02 ict and I think we need to pick a number of rocks No con \ufb02 ict and there \u2019 s certainly at least two types that I can think of . No con \ufb02 ict But we need to have \u2026 S2 , S1 , and so on are speaker numbers . The \ufb01 rst speaker in the clip is S1 and so on . S2 in one clip may not be S2 in another clip . Uncovering uncertainty through disagreement 391 Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e by discussion , with a consensus code assigned . The uncer - tainty coding in this dataset is described in more detail else - where , in a paper that found that problem - related analogies may serve an uncertainty - reduction function ( Cohen \u2019 s kappa = 0 . 75 ; Chan et al . , 2012 ; see Table 2 for examples ) . Schunn ( 2010 ) established the validity of this measure - ment approach using convergent and discriminant validity methods . For example , use of these hedge words corresponded also to the use of uncertainty gestures in team conversations at the utterance level , well above base rates in the local temporal context ( Schunn , 2010 ) . Further , uncertainty and imprecision ( purposeful approximations like \u2018 thirty something \u2019 ) could be reliably separated in speech and gestures , and they had differ - ent relations to other cognitive processes ( Schunn , 2010 ; Schunn , Saner , Kirschenbaum , Trafton , & Littleton , 2007 ) . 1 Both con \ufb02 ict and uncertainty were analyzed at the block level ( i . e . , aggregated across utterances ) . We describe our block creation approach in the subsequent section . Analyses Blocks and video clips To unpack temporal patterns , we broke the video clips ( con - versations ) into segmented blocks made up of continuous , multiple utterances . Intuitively , segmenting video clips into blocks of utterances allows us to track the ebb and \ufb02 ow of uncertainty through conversations by examining how prop - erties of one block relate to uncertainty in subsequent blocks . We created blocks by \ufb01 rst identifying contiguous con \ufb02 ict ut - terances and events . These events would become the initial blocks . For example , the con \ufb02 ict described previously would become a single block made up of multiple utterances . We then identi \ufb01 ed the 25 utterances before and after each con \ufb02 ict event as two additional blocks ( Figure 1 ) . The rest of the clips were broken up into successive blocks of 25 utterances each , each ending at the 25th utterance , the beginning / end of a video clip , or with the next con \ufb02 ict event . Analyses of temporal patterns in uncertainty in this and other complex problem - solving data ( e . g . , Chan et al . , 2012 ; Christensen & Schunn , 2009 ) suggest that uncertainty tends to show signi \ufb01 cant rises and declines in windows of 15 \u2013 30 utterances . This approach to creating utterance blocks also enables the transcript and analyses to be relatively homoge - nous on the independent variable ( con \ufb02 ict ) and standardized in length on the dependent variable blocks ( as with Paletz , Schunn , et al . , 2013 ) . This segmentation method resulted in 688 blocks nested within 114 video clips . Block - level measures As noted , we used aggregated measures of con \ufb02 ict subtype and uncertainty at the block level in our analyses . We \ufb02 agged each con \ufb02 ict event block as to whether it contained at least one of each speci \ufb01 c type of con \ufb02 ict ( at the utterance level ) to create three separate , dichotomous con \ufb02 ict - type vari - ables ( science , planning , and process , 1 = present , 0 = absent ) . Uncertainty was de \ufb01 ned as the number of utterances in a given block coded as uncertain . While the con \ufb02 ict subtype codes at the block level were theoretically not mutually exclusive ( the same block could be marked as a 1 , present , for multiple of the three con \ufb02 ict - type variables ) , blocks were often relatively homogeneous with respect to con \ufb02 ict subtype . Using chi - squared tests , lagged process con \ufb02 icts were signi \ufb01 cantly negatively associ - ated with both lagged planning con \ufb02 icts , \u03c7 2 ( 1 , 115 ) = 32 . 84 , p < . 001 , phi = (cid:1) 0 . 53 , and lagged science con \ufb02 icts , \u03c7 2 ( 1 , 115 ) = 21 . 30 , p < . 001 , phi = (cid:1) 0 . 43 , and planning and science lagged micro - con \ufb02 icts were also negatively related to each other , \u03c7 2 ( 1 , 115 ) = 17 . 97 , p < . 001 , phi = (cid:1) 0 . 40 . 1 To rule out the possibility that our uncertainty coding scheme merely re \ufb02 ected politeness norms , we tested differences by gender and status . Men in our dataset expressed uncertainty ( proportion of uncertainty out of all utterances spoken by men , M = 0 . 14 , SE = 0 . 003 ) at a slightly higher rate than women ( proportion of uncertainty out of utterances spoken by women , M = 0 . 11 , SE = 0 . 01 ) , z = 1 . 96 , p = . 05 . Facultyexpresseduncertainty ( M = 0 . 14 , SE = 0 . 003 ) at asimilar ratetograduate students ( M = 0 . 10 , SE = 0 . 02 ) , z = 1 . 44 , p = . 15 , and older and younger faculty ( M = 0 . 14 , SE = 0 . 00 ) used uncertainty at the same rate , z = 0 . 07 , p = . 94 . Table 2 . Examples of uncertainty from Mars Exploration Rover scientists ( key words in bold ) in conversational contexts , excerpts from Chan et al . ( 2012 ) Speaker Turn Excerpt 1 1 I would have thought that the ventifacting would have been very late and it would have destroyed the surface . Actually maybe that \u2019 s why we don \u2019 t see the black surface of the other two rocks , which also were ventifacts , right ? Because maybe they just happened , we happened to hit faces where it had been scarred 2 And of course that , all that , would mean , if that is an injected block , all of that took place after ; it did not take place at < missing words > 1 Yea , well I mean , I don \u2019 t think that we can rule out that this isn \u2019 t some kind of desert varnish , although I don \u2019 t understand how desert varnish forms Excerpt 2 4 Can you do something with an acid fog ? I know we got sulfate and sulfurs and chlorines ; is there some way we can < missing words > 1 Certainly , yea , maybe that \u2019 s a good explanation of it 4 Maybe it \u2019 s a coating of some sort of sulfate chlorine , you know , crude Excerpt 3 1 Ok , so the idea is go , and then we \u2019 re going to at the end of sol [ Martian day ] 90 . At the end of the afternoon we deploy the M\u00f6ssbauer to capture it . 2 Is this going to be the < missing words > 1 I don \u2019 t know , I think so . And then we let it integrate all night . And then , or maybe , we let it integrate , we let it integrate until some other pass at 3 o \u2019 clock in the morning or something and then 392 S . B . F . Paletz et al . Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e Indeed , 96 % of the 118 con \ufb02 ict blocks had only one subtype . Thus , our data enable us to tease apart the associations between con \ufb02 ict , uncertainty , and temporal context by con \ufb02 ict subtype . Statistical analyses We addressed our main research questions using time - lagged , variable exposure overdispersed Poisson models . These models were tested with two - level hierarchical linear modeling ( Raudenbush & Bryk , 2002 ) via the HLM 7 ( 7 . 01 ) software , which uses penalized quasi - likelihood estimation ( Raudenbush , Bryk , & Congdon , 2013 ) . We report the unit - speci \ufb01 c models with robust standard errors , as is appro - priate for these analyses . Dependent variables that are count data ( i . e . , number of uncertainty utterances in the block ) require Poisson models , and we used the overdispersed setting because the variance of uncertainty ( 5 . 0 ) was greater than the mean ( 2 . 3 ; \u03c3 2 was 1 . 3 instead of 1 ) . Because the blocks varied in their number of utterances because of the endings of clips , beginnings of new con \ufb02 ict events , and so on , we used variable exposure Poisson . As an offset variable , we used ( 1 + ln [ number of utterances ] ) because the offset variable ( i . e . , ln [ 1 ] ) cannot equal zero . We utilized two - level hierarchical linear modeling of blocks ( Level 1 ) within video clips ( Level 2 ) to capture the dependence inherent in the data ( greater similarity in , and non - independence of , conversational dynamics within clips ) and manage unequal cell sizes . Supporting our assertion that uncertainty was in \ufb02 uenced by conversation - level factors , a null model of the uncertainty dependent variable 2 showed that there was signi \ufb01 cant variance at Level 2 ( video clip level ) , tau = 0 . 096 , \u03c7 2 ( 113 ) = 223 . 68 , p < . 001 ( intraclass correlation = 6 . 8 % ) . Because our research questions centered on temporal as - sociations , we used time - lagged analyses to test whether un - certainty in one block was predicted by the presence / absence of con \ufb02 ict , or a particular type of con \ufb02 ict , in the block ( 25 utterances ) before . In creating the models to test our research questions , the time - lagged con \ufb02 ict - to - uncertainty tests were at Level 1 ( blocks ) . Temporal context ( early versus later in the mission ) was at the conversational clip level ( Level 2 ) , and the moderation effect of early / later phase was a cross - level interaction between the clip and block level ( interaction between Level 2 and Level 1 variables ; Figure 1 ) . Before examining our independent variables , we tested for signi \ufb01 cant associations between uncertainty and several co - variates . These covariates were plausibly connected to levels of uncertainty or con \ufb02 ict and therefore represent possible confounds . First , the covariates at Level 1 ( block level ) were tested , then those that were signi \ufb01 cant at Level 2 ( video clip ) , controlling for the signi \ufb01 cant ones at Level 1 . These tested variables had to be signi \ufb01 cant to be retained in the \ufb01 nal model . Three Level 1 variables passed these tests : the average number of words per utterance ( across all the blocks , M = 7 . 0 , median = 6 . 86 , SD = 2 . 16 ; t ( 563 ) = 6 . 75 , B = 0 . 11 , SE = 0 . 02 , event ratio = 1 . 12 [ 1 . 08 , 1 . 15 ] , p < . 001 ) and the two topic vector variables to account for the three topics of discussion ( vector # 1 : t ( 563 ) = (cid:1) 1 . 66 , B = (cid:1) 0 . 14 , SE = 0 . 09 , event ratio = 0 . 87 ( 0 . 73 , 1 . 03 ) , p = . 098 ; vector # 2 , t ( 563 ) = (cid:1) 2 . 21 , B = (cid:1) 0 . 25 , SE = 0 . 11 , event ratio = 0 . 78 [ 0 . 63 , 0 . 97 ] , p = . 027 ) . Although one of the topic vectors was not signi \ufb01 - cant , we kept both vectors because together they represented all three topics of discussion ( science , planning , and work process ; see Paletz et al . , 2013 , for more detail on this covariate ) . The number of words per utterance was a control because the simple length of an utterance could in \ufb02 ate the likelihood of any feature , including uncertainty , in the block . The topic vectors were controlled for because topic alone could have been a third variable that could affect both uncer - tainty and con \ufb02 ict : different topics may elicit or have more con \ufb02 ict or uncertainty regarding them . We tested the following covariates , any of which could have reasonably been related to uncertainty or con \ufb02 ict , and found them not to be signi \ufb01 cant ( p s > . 05 ) : at Level 1 ( block ) , the average number of people present ; and at Level 2 ( video clip ) , the general age / status composition of the scientists present in the clip using two vector variables representing graduate students , young faculty , and older faculty ; the average number of speakers in the clip ; gender composition ; type of science team ( Long Term Planning versus other types of science teams ) ; number of utterances in the clip ; the rover team ( A versus B ) ; and whether the clip was from earlier or later in the initial 90 days . As appropriate to testing interaction effects , nonsigni \ufb01 cant relevant main effects ( i . e . , temporal context ) are included in models to control for main effects when testing interaction effects . 2 A null hierarchical linear model tests the dependent variable without any predictor variables using chi - squared estimation to determine whether there are signi \ufb01 cant higher - level components . Figure 1 . Data structure : blocks nested within clips , and time - lagged analyses of micro - con \ufb02 icts to uncertainty moderated by temporal context Uncovering uncertainty through disagreement 393 Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e RESULTS Frequency of uncertainty and micro - con \ufb02 icts At the block level , uncertainty occurred relatively frequently , with 75 % ( 518 ) of the 688 blocks having at least one uncertain utterance ( M = 2 . 3 , median = 2 . 0 , SD = 2 . 2 , range is 0 \u2013 10 ) . The frequency of the lagged versions of these different task and process con \ufb02 ict types is shown in Table 3 ( see also Paletz et al . , 2011 ; Paletz , Schunn , et al . , 2013 ) . Associations between types of con \ufb02 ict , team factors , and subsequent uncertainty Overall con \ufb02 ict We \ufb01 rst tested whether lagged con \ufb02 ict had a signi \ufb01 cant asso - ciation with subsequent uncertainty and had a signi \ufb01 cant in - teraction with temporal context , as well as whether temporal context had a main effect . The interaction of overall con \ufb02 ict with temporal context was not signi \ufb01 cant , and the model removing it showed that overall lagged con \ufb02 ict was only a marginally signi \ufb01 cant predictor of uncertainty ( Table 4 ) . We expected different forms of con \ufb02 ict to have different effects on uncertainty , so we then unpacked the effects by subtype . Micro - con \ufb02 ict by subtype Next , we conducted moderated tests of temporal context ( early / later phase ) for the three topic subtypes of lagged con - \ufb02 ict : science con \ufb02 ict , planning con \ufb02 ict , and process con \ufb02 ict . Lagged process con \ufb02 icts were unrelated to uncertainty either as a main effect , t ( 462 ) = 1 . 18 , B = 0 . 15 , SE = 0 . 13 , event ra - tio = 1 . 17 [ 0 . 90 , 1 . 51 ] , p = . 239 , or in interaction with tempo - ral context , t ( 462 ) = (cid:1) 1 . 07 , B = (cid:1) 0 . 35 , SE = 0 . 33 , event ratio = 0 . 70 [ 0 . 37 , 1 . 35 ] , p = . 287 ( Figure 2A ) . Planning con \ufb02 ict had a signi \ufb01 cant , positive main effect on subsequent uncertainty , but no signi \ufb01 cant interaction with temporal con - text , t ( 462 ) = (cid:1) 0 . 25 , B = (cid:1) 0 . 05 , SE = 0 . 21 , event ratio = 0 . 95 [ 0 . 63 , 1 . 43 ] , p = . 801 ( see Table 5 for the \ufb01 nal model without the interaction vector ; Figure 2B ) . In contrast , lagged science con \ufb02 ict had a signi \ufb01 cant interaction with temporal context ( early / later phase in the \ufb01 rst 90 days of the mission ) . Early in the mission , lagged science con \ufb02 ict had a positive associ - ation with uncertainty ( similar to planning con \ufb02 ict ) , whereas the association was not signi \ufb01 cant later in the mission ( Table 6 , Figure 2C ) . Two examples of con \ufb02 ict ( planning micro - con \ufb02 icts and early science micro - con \ufb02 icts ) leading to increased uncer - tainty are in Table 7 . In the \ufb01 rst example , two scientists are debating what a sensible plan of instrument readings for Table 3 . Frequency of presence of lagged science , planning , pro - cess , quickly resolved , positive , and negative micro - con \ufb02 icts by block Micro - con \ufb02 ict type Frequency % ( n of 569 lagged blocks ) Any micro - con \ufb02 ict 20 . 2 ( 115 ) Process micro - con \ufb02 ict 8 . 0 ( 46 ) Rover planning ( task ) micro - con \ufb02 ict 8 . 2 ( 47 ) Science ( task ) micro - con \ufb02 ict 4 . 4 ( 25 ) Table 4 . Any lagged micro - con \ufb02 ict and temporal context on uncertainty Variable B SE t Event rate ratio [ 95 % con \ufb01 dence ratio ] df p Intercept , \u03b3 000 (cid:1) 1 . 20 0 . 15 (cid:1) 8 . 21 0 . 30 [ 0 . 23 , 0 . 40 ] 97 < . 001 Average words per utterance 0 . 11 0 . 02 6 . 23 1 . 11 [ 1 . 08 , 1 . 15 ] 463 < . 001 Topic Vector 1 (cid:1) 0 . 09 0 . 09 (cid:1) 1 . 01 0 . 91 [ 0 . 76 , 1 . 09 ] 463 . 315 Topic Vector 2 (cid:1) 0 . 22 0 . 12 (cid:1) 1 . 80 0 . 81 [ 0 . 64 , 1 . 02 ] 463 . 073 Early / later phase 0 . 14 0 . 08 1 . 76 1 . 15 [ 0 . 98 , 1 . 35 ] 97 . 082 Lagged con \ufb02 ict 0 . 13 0 . 07 1 . 67 1 . 13 [ 0 . 98 , 1 . 31 ] 463 . 095 Figure 2 . Predicted number of uncertain utterances controlling for covariates by time - lagged ( A ) process micro - con \ufb02 icts , ( B ) rover planning micro - con \ufb02 icts , and ( C ) science micro - con \ufb02 icts for early and later in the \ufb01 rst 90days of the mission 394 S . B . F . Paletz et al . Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e the rover would be . After the \ufb01 rst speaker strongly explains his or her opinion and the second responds with disagree - ment , the \ufb01 rst starts to express uncertainty about his or her proposed plan . In the second example , a scientist disagrees using uncertainty hedge words , but then after the disagree - ment , that scientist and two others express uncertainty about the nature of the geology they are observing . In both cases , the disagreements precede a signi \ufb01 cant increase in uncer - tainty rates . In summary , micro - con \ufb02 icts about rover planning and science con \ufb02 icts ( the aspects most central to work in the MER context ) were predictive of increases in subsequent psychological uncertainty , but there was no relation between process con \ufb02 icts and subsequent uncertainty . The associa - tion between science con \ufb02 icts and subsequent uncertainty depended on whether the science micro - con \ufb02 ict came earlier or later in the mission , with only early science micro - con \ufb02 icts increasing uncertainty . Association between uncertainty and subsequent con \ufb02 ict Testing the predictiveness of con \ufb02 ict for subsequent uncer - tainty is only half the story in understanding the ebb and \ufb02 ow of con \ufb02 ict and uncertainty . It is theoretically possible that high levels of uncertainty before con \ufb02 icts were related to high levels after the con \ufb02 ict and that the uncertainty was causing the con \ufb02 ict . Therefore , we analyzed our data using the same multilevel analysis approach ( except with logistic regression to account for the binary nature of the dependent variable , the presence of con \ufb02 ict in blocks , and using R \u2019 s multilevel modeling module lme4 ) , and controlling for the same set of covariates ( adding lagged within - domain analo - gies to predict process and science micro - con \ufb02 icts because they were found in prior research to be signi \ufb01 cantly related to these micro - con \ufb02 icts ; Paletz , Schunn , et al . , 2013 ) . We found that uncertainty was , in general , signi \ufb01 cantly nega - tively related to subsequent and concurrent micro - con \ufb02 icts . Indeed , we found that concurrent and previous uncertainty were signi \ufb01 cantly negatively related to rover planning : con - current uncertainty , t ( 462 ) = (cid:1) 2 . 54 , B = (cid:1) 0 . 23 , SE = 0 . 09 , odds ratio = 0 . 79 ( 0 . 66 , 0 . 95 ) , p = . 011 , and previous uncer - tainty , t ( 462 ) = (cid:1) 3 . 02 , B = (cid:1) 0 . 29 , SE = 0 . 10 , odds ratio = 0 . 75 ( 0 . 62 , 0 . 90 ) , p = . 003 . Similarly , process con \ufb02 icts were nega - tively predicted by concurrent uncertainty , t ( 461 ) = (cid:1) 3 . 36 , B = (cid:1) 0 . 40 , SE = 0 . 12 , odds ratio = 0 . 67 ( 0 . 53 , 0 . 84 ) , p < . 001 , and previous uncertainty , t ( 461 ) = (cid:1) 2 . 65 , B = (cid:1) 0 . 27 , SE = 0 . 10 , odds ratio = 0 . 76 ( 0 . 62 , 0 . 93 ) , p = . 008 . Previous uncertainty , but not concurrent uncertainty ( p > . 20 ) , was neg - atively related to subsequent science con \ufb02 ict , t ( 461 ) = (cid:1) 2 . 60 , B = (cid:1) 0 . 34 , SE = 0 . 13 , odds ratio = 0 . 71 ( 0 . 55 , 0 . 92 ) , p = . 010 . These \ufb01 ndings suggest that con \ufb02 ict of any type is less likely to occur at the same time as or immediately following relatively high levels of uncertainty . DISCUSSION Unpacking brief temporal relations between micro - processes is important for gaining a deeper understanding of team problem - solving processes . This study analyzed conversa - tions as a key medium for unpacking team problem solving , focusing on exploring the temporal association between expressed brief disagreements and subsequent uncertainty . Summary and interpretation of \ufb01 ndings We discovered that two different types of preceding task con \ufb02 ict predicted subsequent increases in psychological uncertainty : Planning task con \ufb02 ict was associated with an increase in uncertainty without any interaction with temporal phase , and science con \ufb02 ict was associated with an increase in uncertainty , but only early in the mission . Process con \ufb02 ict was unrelated to subsequent uncertainty . Testing for a poten - tial reverse association revealed that process , rover planning , and science con \ufb02 ict all decreased following uncertainty , suggesting that our \ufb01 ndings for con \ufb02 ict predicting increases in expressed uncertainty could not be due to a general Table 5 . Final model of lagged planning con \ufb02 ict and covariates on uncertainty Variable B SE t Event rate ratio [ 95 % con \ufb01 dence ratio ] df p Intercept , \u03b3 000 (cid:1) 1 . 18 0 . 15 (cid:1) 8 . 00 0 . 31 [ 0 . 23 , 0 . 41 ] 97 < . 001 Average words per utterance 0 . 11 0 . 02 6 . 27 1 . 11 [ 1 . 08 , 1 . 15 ] 463 < . 001 Topic Vector 1 (cid:1) 0 . 12 0 . 09 (cid:1) 1 . 29 0 . 89 [ 0 . 74 , 1 . 07 ] 463 . 198 Topic Vector 2 (cid:1) 0 . 22 0 . 12 (cid:1) 1 . 85 0 . 80 [ 0 . 63 , 1 . 01 ] 463 . 065 Early / later phase 0 . 13 0 . 08 1 . 69 1 . 14 [ 0 . 98 , 1 . 33 ] 97 . 095 Lagged planning con \ufb02 ict 0 . 22 0 . 11 2 . 09 1 . 25 [ 1 . 01 , 1 . 54 ] 463 . 037 Table 6 . Final model of lagged science con \ufb02 ict , temporal context , their interaction , and covariates on uncertainty Variable B SE t Event rate ratio [ 95 % con \ufb01 dence ratio ] df p Intercept , \u03b3 000 (cid:1) 1 . 19 0 . 15 (cid:1) 7 . 97 0 . 31 [ 0 . 23 , 0 . 41 ] 97 < . 001 Average words per utterance 0 . 11 0 . 02 6 . 10 1 . 11 [ 1 . 08 , 1 . 15 ] 462 < . 001 Topic Vector 1 (cid:1) 0 . 08 0 . 09 (cid:1) 0 . 91 0 . 92 [ 0 . 77 , 1 . 10 ] 462 . 366 Topic Vector 2 (cid:1) 0 . 22 0 . 12 (cid:1) 1 . 82 0 . 81 [ 0 . 64 , 1 . 02 ] 462 . 069 Early / later phase 0 . 16 0 . 08 2 . 05 1 . 17 [ 1 . 01 , 1 . 37 ] 97 . 043 Lagged science con \ufb02 ict 0 . 30 0 . 18 1 . 64 1 . 35 [ 0 . 94 , 1 . 93 ] 462 . 102 Lagged Science Con \ufb02 ict\u00d7 Early / Later Phase (cid:1) 0 . 51 0 . 23 (cid:1) 2 . 23 0 . 60 [ 0 . 39 , 0 . 94 ] 462 . 027 Uncovering uncertainty through disagreement 395 Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e positive association between uncertainty and con \ufb02 ict through some alternate factor . Taken together , these \ufb01 ndings suggest that individuals were more likely to disagree when they expressed more con \ufb01 dence and fewer hedge words and that con \ufb02 ict exposed underlying team uncertainties . Uncertainty is central to problem solving ( Kahneman & Tversky , 1982 ; Schunn , 2010 ; Schunn & Trafton , 2012 ) . Resolving uncertainty often entails sharing information previously unknown , whereas detecting uncertainty involves discovering what is unknown ( Schunn & Trafton , 2012 ) . Task con \ufb02 ict is commonly argued to be the least harmful to team success , at best ( e . g . , de Wit et al . , 2012 ; Nemeth , 1986 ) . Uncertainty is generally thought to be problematic , a relatively weak information state that needs to be improved . Does this \ufb01 nding imply that task con \ufb02 ict has yet another potentially unproductive function ? Such a value judgment is unwarranted given the grain size of the current analysis . Micro - con \ufb02 icts and expressed uncertainty are part of complex team dynamics that iterate many times and gradually unfold . In fact , the discovered dynamic may simply be a core characteristic of normal team functioning . To interpret these \ufb01 ndings , we examine the MER context more closely and draw on the shared mental models theory ( e . g . , Burke et al . , 2006 ; Mathieu et al . , 2000 ) , which de - scribes the development of team cognition . We suggest that task con \ufb02 ict in this case may have uncovered disconnects between the multidisciplinary scientists \u2019 mental models , and the rise in uncertainty following task con \ufb02 ict may have re \ufb02 ected the identi \ufb01 cation of newly exposed problems . There was a great deal of uncertainty in conducting science on Mars in general . Although levels of uncertainty were not generally different between early and later in the \ufb01 rst 90 days of the mission , disagreements about science were more likely to involve an increase in subsequent uncertainty Table 7 . Examples of micro - con \ufb02 icts preceding increases in uncertainty Coded con \ufb02 ict Coded uncertainty a Speaker number Utterance Example 1 : planning micro - con \ufb02 ict preceding more uncertainty Planning 0 1 No , no , no , no Planning 0 1 If you want to do the integration on the RAT [ Rock Abrasion Tool ] , Planning 0 1 you \u2019 re going to have to do some additional brushing Planning 0 1 because you don \u2019 t want crap in the brushing in the gravel . Planning 0 1 So you pretty much go through the standard sequence , Planning 0 1 and the advantage of that is that you now have looked at all the different depths . Planning 0 2 But what do you , what do you do with that ? ? Planning 0 2 So after you \u2019 ve done the integration , Planning 0 2 then you do the \ufb01 nal brush with backup . None 0 1 Right , you can \u2019 t do any more , more brushing None 0 1 You don \u2019 t want to contaminate the hole None 1 1 So that could be one way you could get that Minitest [ another instrument ] observation , None 0 1 and not a lot of < missing word > in one sol . None 0 1 The only thing that \u2019 s scary about that is that you drive away not knowing what you got in the Minitest observation , None 0 1 but it \u2019 s just a backup you know of 185 . None 0 1 I \u2019 ll show you what we \u2019 re talking about . None 0 1 So that \u2019 s one option . None 0 1 Then after you start to drive , None 1 1 but I guess you \u2019 re on for what , two more sols , so pretty much . Example 2 : science micro - con \ufb02 ict early in mission ( sol 16 ) preceding more uncertainty None 0 3 So these aren \u2019 t exactly perfect . None 0 3 These are actually old < missing words > weathered out . Science 0 4 Well , but if you think about it , Science 1 4 the history of this thing post - impact , you know , maybe , maybe these have , um , formed over time . None 0 1 Well , do you think it could be impact generated ? None 1 4 No , not impact generated , \u2019 cause they maybe a certain , well I don \u2019 t know . Science 0 1 Well , the problem is they look very speci \ufb01 c , it \u2019 s not even penetrative < missing word > . None 0 4 No , you \u2019 re right , None 1 4 there would be if , there would have to be some fundamental difference in the hardness of those layers . None 0 2 You know , one thing you can do on that is to say None 1 2 that maybe these were , that there is a sulfate cement in there < missing words > , None 1 2 but they don \u2019 t really look like < missing words > . None 0 1 See , I love all this < missing words > None 0 1 If you were on Earth you would just simply call that a nodular fabric None 0 1 and go about your business None 0 1 and then you \u2019 d speculate about whether it was nodular because of early cementation , pearling , or pressure solution None 1 1 and probably here we can rule out the last two . a Zero is not coded as having uncertainty , 1 is coded as having uncertainty . 396 S . B . F . Paletz et al . Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e early in the mission . The scientists during that phase were likely still developing their mental models ( individually and jointly ) of how and whether liquid water had existed on Mars \u2014 the high - level topic of the science con \ufb02 icts . Roughly a third of the way through the original mission , the scientists felt con \ufb01 dent that they had discovered evidence of historical liquid water on Mars . At this stage , mental models regarding the science may have become more shared across the scientists , affecting the interpretation of new \ufb01 nd - ings . This turning point in scienti \ufb01 c discovery may have constrained the types of new information that could arise from disagreement , such that disagreements after that point were no longer positively , signi \ufb01 cantly related to uncertainty . However , issues of what instrument readings and activi - ties to do with the rovers ( i . e . , planning activities ) remained both contentious and uncertain throughout the mission . The different subgroups of scientists ( e . g . , soil science and atmospheres ) each wanted to conduct their own instrument readings , and each rover team had to prioritize and plan what the rovers would do . For instance , the scientists would debate whether they should have a rover ( i . e . , Rover A , Spirit ) stay in one location and conduct additional instrument readings or drive to another location . Thus , these kinds of arguments could be due to differences in explicit goals in addition to differences in perceptions of the utility of doing instrument readings in a particular location . Furthermore , there was still a great deal unknown about what the best course of action would be , either scienti \ufb01 cally or logistically . The scientists did not know at the time that the rovers would last several years past their initial 90 - day life . Thus , the underlying differences in goals , assumptions , and mental models of what the best courses of action would be may have continued throughout the \ufb01 rst 90 days of the mission , unchanged by the discovery of historical liquid water . Disagreements would still lead to greater uncertainty , because they would point out the missing information inherent in trying to plan rover activities on Mars . Uncertainty could , and did , lead to effective problem solv - ing , even within the moment . Indeed , previous research we conducted on this dataset found a signi \ufb01 cant positive associa - tion between uncertainty and subsequent problem - related anal - ogy , with a return to baseline uncertainty after the analogy was mentioned ( Chan et al . , 2012 ) . This study also found that the most common topic of uncertainty was science problem solv - ing , suggesting that much of the uncertainty was central to their primary problem solving . For example , one scientist expressed uncertainty about the cause of observed ventifacting ( a geomorphic feature on rocks ) and uncertainty about a poten - tial mechanism for the ventifacting , desert varnish ( Chan et al . , 2012 , p . 1361 ) . That scientist and two others proceeded to discuss the problem . One of the scientists made an analogy to how desert varnish works on Earth and a later analogy to \ufb01 ndings from a Viking mission . Between the three scientists , they identi \ufb01 ed several possible reasons for ventifacting and the known causes of desert varnish that could be occurring on Mars . This anecdote and the data supporting it illustrate how uncertainty in the MER context could , and did , precede problem solving ( Chan et al . , 2012 ) . Thus , if task con \ufb02 ict is leading to uncertainty in this domain , that uncertainty then has the potential to lead to problem solving . Previous \ufb01 ndings suggest that self - reported assessments of task con \ufb02 ict have a curvilinear ( inverse - U ) association with team performance ( e . g . , Fahr et al . , 2010 ; Shaw et al . , 2011 ) . This study unpacks one possible function of task con \ufb02 ict at the process level : uncovering areas of uncertainty . Relating to the curvilinear pattern with success , it may be that low levels of task con \ufb02 ict leave critical misunderstan - dings or mental model gaps hidden and that very high levels of task con \ufb02 ict unearth more uncertainty than can be resolved . Alternately , these very high levels of con \ufb02 ict inhibit the resolution of the uncertainty via limiting the team \u2019 s ability to build shared mental models . These hypo - theses should be investigated in future research . Strengths and limitations A bene \ufb01 t of this study is that it studied real - world problem solving at the micro - temporal level . Too often , detailed pro - cess studies of actual behaviors are limited to the research lab with undergraduate participants completing arti \ufb01 cial tasks over limited time periods , and self - report methods ( survey or interviews ) are used to study actual work contexts . Even the rare communication study of con \ufb02 ict as it arises from natural conversations does not ask the research ques - tions posed here ( e . g . , Poole & Dobosh , 2010 ) . However , one limitation of the methods used in this study involves the lower reliabilities associated with coding micro - con \ufb02 ict types . There are multiple factors underlying lower measure reliability than that typically found with some other approaches . For one , the coding scheme for this study follows traditional examinations of con \ufb02 ict types as distinct ( i . e . , task , process , and relationship ; Jehn , 1995 ) , but these dimensions were developed on self - report measures , inter - views , and larger grain sizes in observations of teams . It can be dif \ufb01 cult for even trained coders to distinguish between these types at the utterance . To reduce the negative effects of lowered reliability on the statistical power of the analyses , the target behavior was 100 % double coded , with all disagreements resolved by discussion , giving us more con \ufb01 dence in our eventual resolutions ( Smith , 2000 ) . Further , the operationalization of micro - con \ufb02 icts was at an aggregated level , not the utterance level , helping to reduce noise due to lower reliabilities . Another bene \ufb01 t of the methods used in this study is the di - versity of individuals and tasks . For example , it examined in - dividuals across a range of professions ( university scientist , government lab scientist , and industry engineer ) and along a range of skill levels ( e . g . , from famous full professors to early graduate students ) . The scientists also accomplished a large variety of tasks , including dealing with anomalies , planning , scheduling , data analysis , e - mailing , taking photo - graphs , and so on . Thus , even though these conversations were taken from one larger context ( i . e . , the MER mission ) , the sample was quite diverse in many other ways . At the same time , there were some unique features of the sample that may limit its generalizability . The scientists were overall an extremely successful team . They had complex processes and worked during the \ufb01 rst 90 days on Mars time schedules , getting progressively out of sync with Earth day / night cycles . While some of these features may be Uncovering uncertainty through disagreement 397 Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e generalizable ( complex processes and expert team ) , some are unique ( e . g . , working on Mars time ) , which means the \ufb01 nd - ings may not transfer to other settings . Furthermore , even given transcribing and coding over 11 hours of conversa - tions , we acknowledge that these conversations were only a sample of what the scientists discussed , on and off camera . Future studies could examine data evenly across the life of a team and so uncover with greater precision when the temporal moderation effects occur . Implications and future directions Scienti \ufb01 c problem solving often attempts to resolve uncer - tainty , and uncertainty may drive scienti \ufb01 c discovery ( Schunn , 2010 ; Schunn & Trafton , 2012 ) . Even before prob - lem solving is used to resolve uncertainty , however , what is unknown needs to be identi \ufb01 ed . Problem \ufb01 nding is a neces - sary precursor to problem solving . This paper illuminates a possible , contingent association between brief disagreements and immediate , subsequent psychological uncertainty , join - ing with other studies that examine the immediate conse - quences of brief disagreements ( e . g . , Chiu , 2008a , 2008b ; Kauffeld & Lehmann - Willenbrock , 2012 ; Paletz et al . , 2013 ) . Future research could continue to unpack potential contextual and other moderators of this micro - con \ufb02 ict / uncertainty connection . This study supports theorizing that different types of con \ufb02 ict may play different roles in team processes . For instance , process con \ufb02 ict early in a team \u2019 s life cycle may be positively related to eventual team perfor - mance , perhaps because it prevents premature closure ( Goncalo et al . , 2010 ) . A \ufb01 ne - grained analysis , as conducted in this study , of multiple teams \u2019 process micro - con \ufb02 icts in their early stages can shed light onto how they may in \ufb02 uence uncertainty and closure on speci \ufb01 c decisions . This paper also has implications for the role of uncertainty in problem - solving conversations . Future directions might also include qualitative studies of different types of uncer - tainty ( e . g . , Lipshitz & Strauss , 1997 ) and how the uncer - tainty that arises after a con \ufb02 ict may serve a different role in problem - solving conversations compared with other types of uncertainty . For instance , how is decision making impacted by whether or not a con \ufb02 ict \u2013 uncertainty pair is present ? Additional studies could also examine whether the uncertainty that arises after a micro - con \ufb02 ict is related to problem \ufb01 nding and restructuring , exposing gaps in informa - tion and / or freeing individuals to share their concerns . This paper advances future theorizing about team pro - cesses by examining a new association at a micro - temporal level . Of all the potential functions of disagreement , uncertainty reduction or increase has been barely studied . Examining micro - con \ufb02 icts serves to shift the lens of scienti \ufb01 c examination from mostly static snapshots of broad team perceptions to examining the interplay of different types of quick conversational acts . Practically , such research and theory could help guide the creation of team facilitation techniques and suggestions for how teams can better leverage their disagreements for speci \ufb01 c problem - solving processes . Much of the literature on con \ufb02 ict focuses on reducing or mitigating it . This paper suggests that con \ufb02 ict resolution training could also be expanded to emphasize the useful cognitive functions of small disagreements , such as questioning underlying assumptions and restructuring problems . Once uncertainty is brought into the open , new problems can be explicitly addressed . Problem solving , particularly in multidisciplinary teams , is vital for scien - ti \ufb01 c innovation , economic progress , and \ufb01 nding of solutions to pressing societal problems . By unpacking problem - solving conversations as they occur in real - world successful expert teams , we discover more about their functioning . ACKNOWLEDGEMENTS This research was supported in part by the United States National Science Foundation ( NSF ) Science of Science and Innovation Policy Program grants # SBE - 1064083 to the \ufb01 rst author and NSF grant # SBE - 0830210 to the \ufb01 rst and third authors when the \ufb01 rst author was at the Learning Research and Development Center at the University of Pittsburgh . We wish to thank Kevin Kim and Mathilda du Toit for their statistical advice . The authors are grateful to Alonso Vera , Irene Tollinger , Preston Tollinger , Mike McCurdy , and Tyler Klein for their assistance in data collection ; Lauren Gunn , Julia Cocchia , Carl Fenderson , Justin Krill , Kerry Glassman , Tiemoko Ballo , Rebecca Sax , Michael Ye , and Candace Smalley for coding ; and Carmela Rizzo for research management support . These data were coded while the \ufb01 rst and second authors were at the University of Pittsburgh . REFERENCES Amason , A . C . ( 1996 ) . Distinguishing the effects of functional and dysfunc - tional con \ufb02 ict on strategic decision making : Resolving a paradox for top management teams . Academy of Management Journal , 39 , 123 \u2013 148 . DOI : 10 . 2307 / 256633 . Bakeman , R . , Quera , V . , McArthur , D . , & Robinson , B . F . ( 1997 ) . Detecting sequentialpatternsanddetermining theirreliability withfallibleobservers . Psychological Methods , 2 , 357 \u2013 370 . DOI : 10 . 1037 / 1082 - 989X . 2 . 4 . 357 . Barki , H . , & Hartwick , J . ( 2004 ) . Conceptualizing the construct of interpersonal con \ufb02 ict . International Journal of Con \ufb02 ict Management , 15 , 216 \u2013 244 . DOI : 10 . 1108 / eb022913 . Bearman , C . R . , Paletz , S . B . F . , Orasanu , J . , & Thomas , M . J . W . ( 2010 ) . The breakdown of coordinated decision making in distributed systems . Human Factors , 52 , 173 \u2013 188 . DOI : 10 . 1177 / 0018720810371338 . Beck , S . J . ( 2013 ) . Moving beyond disciplinary differences in group research . Small Group Research , 44 , 195 \u2013 199 . DOI : 10 . 1177 / 1046496412471862 . Burke , C . S . , Stagl , K . C . , Salas , E . , Peirce , L . , & Kendall , D . ( 2006 ) . Under - standing team adaptation : A conceptual analysis and model . Journal of Applied Psychology , 91 , 1189 \u2013 1207 . doi : 10 . 1037 / 0021 - 9010 . 91 . 6 . 1189 Chan , J . , Paletz , S . B . F . , & Schunn , C . D . ( 2012 ) . Analogy as a strategy for supporting complex problem solving under uncertainty . Memory and Cognition , 40 , 1352 \u2013 1365 . DOI : 10 . 3758 / s13421 - 012 - 0227 - z . Chi , M . ( 1997 ) . Quantifying qualitative analyses of verbal data : A practical guide . Journal of the Learning Sciences , 6 , 271 \u2013 315 . DOI : 10 . 1207 / s15327809jls0603 _ 1 . Chinn , C . A . , & Malhotra , B . A . ( 2002 ) . Epistemologically authentic inquiry in schools : A theoretical framework for evaluating inquiry tasks . Science Education , 86 ( 2 ) , 175 \u2013 218 . DOI : 10 . 1002 / sce . 10001 . Chiu , M . M . ( 2008a ) . Effects of argumentation on group micro - creativity : Statistical discourse analyses of algebra students \u2019 collaborative problem - solving . Contemporary Educational Psychology , 33 , 382 \u2013 402 . DOI : 10 . 1016 / j . cedpsych . 2008 . 05 . 001 . Chiu , M . M . ( 2008b ) . Flowing toward correct contributions during group problem solving : A statistical discourse analysis . Journal of the Learning Sciences , 17 , 415 \u2013 463 . DOI : 10 . 1080 / 10508400802224830 . 398 S . B . F . Paletz et al . Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e Christensen , B . T . , & Schunn , C . D . ( 2009 ) . The role and impact of mental simulation in design . Applied Cognitive Psychology , 23 , 327 \u2013 344 . DOI : 10 . 1002 / acp . 1464 . Chua , E . F . , Hannula , D . E . , & Ranganath , C . ( 2012 ) . Distinguishing highly con \ufb01 dent accurate and inaccurate memory : Insights about relevant and ir - relevant in \ufb02 uences on memory con \ufb01 dence . Memory , 20 , 48 \u2013 62 . DOI : 10 . 1080 / 09658211 . 2011 . 633919 . Cronin , M . A . , & Weingart , L . R . ( 2007 ) . Representational gaps , informa - tion processing , and con \ufb02 ict in functionally diverse teams . Academy of Management Review , 32 , 761 \u2013 773 . DOI : 10 . 5465 / AMR . 2007 . 25275511 . Cronin , M . A . , Weingart , L . R . , & Todorova , G . ( 2011 ) . Dynamics in groups : Are we there yet ? Academy of Management Annals , 5 , 571 \u2013 612 . DOI : 10 . 1080 / 19416520 . 2011 . 590297 . Dama , M . , & Dunbar , K . ( 1996 ) . Distributed reasoning : An analysis of where social and cognitive worlds fuse . In : Proceedings of the Eighteenth Annual Conference of the Cognitive Science Society ( pp . 166 \u2013 170 ) . De Dreu , C . K . W . , & Weingart , L . R . ( 2003 ) . Task versus relationship con - \ufb02 ict , team performance , and team member satisfaction : A meta - analysis . Journal of Applied Psychology , 88 , 741 \u2013 749 . doi : 10 . 1037 / 0021 - 9010 . 88 . 4 . 741 Derry , S . J . , Schunn , C . D . , & Gernsbacher , M . A . ( Eds ) ( 2005 ) . Interdisciplin - ary collaboration : An emerging cognitive science . Mahwah , NJ : Erlbaum . De Wit , F . R . C . , Greer , L . L . , & Jehn , K . A . ( 2012 ) . The paradox of intragroup con \ufb02 ict : A meta - analysis . Journal of Applied Psychology , 97 , 360 \u2013 390 . DOI : 10 . 1037 / a0024844 . De Wit , F . R . C . , Jehn , K . A . , & Scheepers , D . ( 2013 ) . Task con \ufb02 ict , infor - mation processing , and decision - making : The damaging effect of rela - tionship con \ufb02 ict . Organizational Behavior and Human Decision Processes , 122 , 177 \u2013 189 . DOI : 10 . 1016 / j . obhdp . 2013 . 07 . 002 . Downey , H . K . , & Slocum , J . W . ( 1975 ) . Uncertainty : Measures , research , and sources of variation . Academy of Management Journal , 18 , 562 \u2013 578 . DOI : 10 . 2307 / 255685 . Dunbar , K . ( 1997 ) . How scientists think : On - line creativity and conceptual change in science . In T . B . Ward & S . M . Smith ( Eds . ) , Creative thought : An investigation of conceptual structures and processes ( pp . 461 \u2013 493 ) . Washington , DC : American Psychological Association . Evans , J . S . B . T . ( 1989 ) . Bias in human reasoning : Causes and conse - quences . London : Erlbaum . Fahr , J . - L . , Lee , C . , & Farh , C . I . C . ( 2010 ) . Task con \ufb02 ict and team creativ - ity : A question of how much and when . Journal of Applied Psychology , 95 , 1173 \u2013 1180 . DOI : 10 . 1037 / a0020015 . Gersick , C . J . G . ( 1988 ) . Time and transition in work teams . Toward a new model of group development . Academy of Management Journal , 31 , 9 \u2013 41 . DOI : 10 . 2307 / 256496 . Goh , K . T . , Goodman , P . S . , & Weingart , L . R . ( 2013 ) . Team innovation processes : An examination of activity cycles in creative project teams . Small Group Research , 44 , 159 \u2013 194 . DOI : 10 . 2307 / 256496 . Goncalo , J . A . , Polman , E . , & Maslach , C . ( 2010 ) . Can con \ufb01 dence come too soon ? Collective ef \ufb01 cacy , con \ufb02 ict and group performance over time . Organizational Behavior and Human Decision Processes , 113 , 13 \u2013 24 . DOI : 10 . 1016 / j . obhdp . 2010 . 05 . 001 . Gorman , M . E . ( 1986 ) . How the possibility of error affects falsi \ufb01 cation on a task that models scienti \ufb01 c problem solving . British Journal of Psychol - ogy , 77 , 85 \u2013 96 . DOI : 10 . 1111 / j . 2044 - 8295 . 1986 . tb01984 . x . Gottman , J . M . , & Notarius , C . I . ( 2000 ) . Decade review : Observing marital interaction . Journal of Marriage and the Family , 62 , 927 \u2013 947 . DOI : 10 . 1111 / j . 1741 - 3737 . 2000 . 00927 . x . Greer , L . L . , Jehn , K . A . , & Mannix , E . A . ( 2008 ) . Con \ufb02 ict transformation : A longitudinal investigation of the relationships between different types of intragroup con \ufb02 ict and the moderating role of con \ufb02 ict resolution . Small Group Research , 39 , 278 \u2013 302 . DOI : 10 . 1177 / 1046496408317793 . Guzzo , R . A . , & Dickson , M . W . ( 1996 ) . Teams in organizations : Recent re - search on performance and effectiveness . Annual Review of Psychology , 47 , 307 \u2013 338 . He , Y . , Ding , X . - H . , & Yang , K . ( 2014 ) . Unpacking the relationships between con \ufb02 icts and team innovation : Empirical evidence from China . Manage - ment Decision , 52 , 1533 \u2013 1548 . DOI : 10 . 1108 / MD - 03 - 2014 - 0127 . Jehn , K . A . ( 1995 ) . A multimethod examination of the bene \ufb01 ts and detri - ments of intragroup con \ufb02 ict . Administrative Science Quarterly , 40 , 256 \u2013 282 . Jehn , K . A . ( 1997 ) . A qualitative analysis of con \ufb02 ict types and dimensions in organizational groups . Administrative Science Quarterly , 42 , 530 \u2013 557 . Johnson - Laird , P . N . ( 1980 ) . Mental models in cognitive science . Cognitive Science , 4 , 71 \u2013 115 . DOI : 10 . 1207 / s15516709cog0401 _ 4 . Jones , B . F . ( 2009 ) . The burden of knowledge and the \u2018 death of the renais - sance man \u2019 : Is innovation getting harder ? Review of Economic Studies , 76 ( 1 ) , 283 \u2013 317 . DOI : 10 . 1111 / j . 1467 - 937X . 2008 . 00531 . x . Jones , R . G . , Stevens , M . J . , & Fischer , D . L . ( 2000 ) . Selection in team con - texts . In J . F . Kehoe ( Ed . ) , Managing selection in changing organizations : Human resource strategies ( pp . 210 \u2013 241 ) . San Francisco : Jossey - Bass . Jordan , M . E . , & Babrow , A . S . ( 2013 ) . Communication in creative collaborations : The challenges of uncertainty and desire related to task , identity , and relational goals . Communication Education . Online \ufb01 rst , doi : 10 . 1080 / 03634523 . 2013 . 769612 Kahneman , D . , & Tversky , A . ( 1982 ) . Variants of uncertainty . Cognition , 11 , 143 \u2013 157 . DOI : 10 . 1016 / 0010 - 0277 ( 82 ) 90023 - 3 . Kauffeld , S . , & Lehmann - Willenbrock , N . ( 2012 ) . Meetings matter : Effects of team meetings on team and organizational success . Small Group Research , 43 , 130 \u2013 158 . DOI : 10 . 1177 / 1046496411429599 . Kirschenbaum , S . S . , Trafton , J . G . , Schunn , C . D . , & Trickett , S . B . ( 2014 ) . Visualizing uncertainty : The impact on performance . Human Factors , 56 ( 3 ) , 509 \u2013 520 . DOI : 10 . 1177 / 0018720813498093 . Klein , G . A . ( 1989 ) . Strategies of decision making ( edn , pp . 56 \u2013 64 ) . May : Military Review . Kuhn , T . , & Poole , M . S . ( 2000 ) . Do con \ufb02 ict management styles affect group decision making ? Evidence from a longitudinal \ufb01 eld study . Human Communication Research , 26 , 558 \u2013 590 . DOI : 10 . 1111 / j . 1468 - 2958 . 2000 . tb00769 . x . Kurtzberg , T . R . , & Mueller , J . S . ( 2005 ) . The in \ufb02 uence of daily con \ufb02 ict on perceptions of creativity : A longitudinal study . International Journal of Con \ufb02 ict Management , 16 , 335 \u2013 353 . Lehmann - Willenbrock , N . , Meyers , R . A . , Kauffeld , S . , Neininger , A . , & Henschel , A . ( 2011 ) . Verbal interaction sequences and group mood : Ex - ploring the role of team planning communication . Small Group Research , 42 , 639 \u2013 668 . DOI : 10 . 1177 / 1046496411398397 . Lipshitz , R . , & Strauss , O . ( 1997 ) . Coping with uncertainty : A naturalistic decision - making analysis . Organizational Behavior and Human Decision Processes , 69 , 149 \u2013 163 . DOI : 10 . 1006 / obhd . 1997 . 2679 . Mannix , E . , & Neale , M . A . ( 2005 ) . Whatdifferences makea difference ? The promise and reality of diverse teams in organizations . Psychological Sci - ence in the Public Interest , 6 , 2 . DOI : 10 . 1111 / j . 1529 - 1006 . 2005 . 00022 . x . Mathieu , J . E . , Heffner , T . S . , Goodwin , G . F . , Salas , E . , & Cannon - Bowers , J . A . ( 2000 ) . The in \ufb02 uence of shared mental models on team process and performance . Journal of Applied Psychology , 85 , 273 \u2013 283 . DOI : 10 . 1037 / 0021 - 9010 . 85 . 2 . 273 . McGrath , J . ( 1991 ) . Time , interaction , andperformance ( TIP ) : Atheoryofgroups . Small Group Research , 22 , 147 \u2013 174 . DOI : 10 . 1177 / 1046496491222001 . McGrath , J . , & Tschan , F . ( 2004 ) . Temporal matters in social psychology : Examining the role of time in the lives of groups and individuals . Washington , DC : American Psychological Association . Miron - Spektor , E . , Efrat - Treister , D . , Rafaeli , A . , & Schwartz - Cohen , O . ( 2011 ) . Others \u2019 anger makes people work harder not smarter : The effect of observing anger and sarcasm on complex thinking . Journal of Applied Psychology , 96 , 1065 \u2013 1075 . DOI : 10 . 1037 / a0023593 . Morgan , B . B . , Salas , E . , & Glickman , A . S . ( 1994 ) . An analysis of team evolution and maturation . Journal of General Psychology , 120 , 277 \u2013 291 . DOI : 10 . 1080 / 00221309 . 1993 . 9711148 . Mumford , M . D . , Reiter - Palmon , R . , & Redmond , M . R . ( 1994 ) . Problem construction and cognition : Applying problem representations in ill - de \ufb01 ned domains . In M . A . Runco ( Ed . ) , Problem \ufb01 nding , problem solving , and creativity ( pp . 3 \u2013 39 ) . Norwood , NJ : Ablex . Nemeth , C . J . ( 1986 ) . Differential contributions of majority and minority in \ufb02 u - ence . Psychological Review , 93 , 23 \u2013 32 . DOI : 10 . 1037 / 0033 - 295X . 93 . 1 . 23 . Nemeth , C . J . Personnaz , B . , Personnaz , M . , & Goncalo , J . A . 2004 The liber - ating role of con \ufb02 ict in group creativity : A study intwo countries . European Journal of Social Psychology , 34 , 365 \u2013 374 . doi : 10 . 1002 / ejsp . 210 Paletz , S . B . F . , Kim , K . H . , Schunn , C . D . , Tollinger , I . , & Vera , A . ( 2013a ) . Reuse and recycle : The development of adaptive expertise , rou - tine expertise , and novelty in a large research team . Applied Cognitive Psychology , 27 , 415 \u2013 428 . DOI : 10 . 1002 / acp . 2928 . Paletz , S . B . F . , Miron - Spektor , E . , & Lin , C . - C . ( 2014 ) . A cultural lens on interpersonal con \ufb02 ict and creativity in multicultural environments . Psychology of Aesthetics , Creativity , and the Arts , 8 ( 2 ) , 237 \u2013 252 . DOI : 10 . 1037 / a0035927 . Uncovering uncertainty through disagreement 399 Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e Paletz , S . B . F . , & Peng , K . ( 2009 ) . Problem \ufb01 nding and contradiction : Examining the relationship between na\u00efve dialectical thinking , ethnicity , and creativity . Creativity Research Journal , 21 , 139 \u2013 151 . DOI : 10 . 1080 / 10400410902858683 . Paletz , S . B . F . , & Schunn , C . D . ( 2010 ) . A social - cognitive framework of multidisciplinary team innovation . Topics in Cognitive Science , 2 , 73 \u2013 95 . DOI : 10 . 1111 / j . 1756 - 8765 . 2009 . 01029 . x . Paletz , S . B . F . , & Schunn , C . D . ( 2011 ) . Assessing group level participation in \ufb02 uid teams : Testing a new metric . Behavior Research Methods , 32 , 522 \u2013 536 . Paletz , S . B . F . , Schunn , C . D . , & Kim , K . H . ( 2011 ) . Con \ufb02 ict under the mi - croscope : Micro - con \ufb02 icts in naturalistic team discussions . Negotiation and Con \ufb02 ict Management Research , 4 , 314 \u2013 351 . DOI : 10 . 1111 / j . 1750 - 4716 . 2011 . 00085 . x . Paletz , S . B . F . , Schunn , C . D . , & Kim , K . H . ( 2013b ) . The interplay of con \ufb02 ict and analogy in multidisciplinary teams . Cognition , 126 , 1 \u2013 19 . DOI : 10 . 1016 / j . cognition . 2012 . 07 . 020 . Parayitam , S . , & Dooley , R . S . ( 2011 ) . Is too much cognitive con \ufb02 ict in stra - tegic decision - making teams too bad ? International Journal of Con \ufb02 ict Management , 22 , 342 \u2013 357 . DOI : 10 . 1108 / 10444061111171350 . Poole , M . S . ( 1989 ) . Group working relationships coding system . Retrieved 29 October 2014 from : https : / / www . ideals . illinois . edu / handle / 2142 / 14539 . Poole , M . S . , & Dobosh , M . ( 2010 ) . Exploring con \ufb02 ict management pro - cesses in jury deliberations through interaction analysis . Small Group Re - search , 41 , 408 \u2013 426 . DOI : 10 . 1177 / 1046496410366310 . Poole , M . S . , & Garner , J . T . ( 2006 ) . Perspectives on workgroup con \ufb02 ict and communication . In J . G . Oetzel & S . Ting - Toomey ( Eds . ) , The Sage handbook of con \ufb02 ict communication : Integrating theory , research , and practice ( pp 267 \u2013 292 ) . Thousand Oaks , CA : Sage . Raudenbush , S . W . , & Bryk , A . S . ( 2002 ) . Hierarchical linear models : Appli - cations and data analysis methods ( 2nd edn ) . Newbury Park , CA : Sage . Raudenbush , S . W . , Bryk , A . S . , & Congdon , R . ( 2013 ) . HLM 7 for Windows [ computer software ] . Skokie , IL : Scienti \ufb01 c Software International , Inc . . Runco , M . A . ( 1994 ) . Problem \ufb01 nding , problem solving , and creativity . Stamford , CT : Ablex . Runco , M . A . , & Nemiro , J . ( 1994 ) . Problem \ufb01 nding , creativity , and gifted - ness . Roeper Review , 16 , 235 \u2013 241 . DOI : 10 . 1080 / 02783199409553588 . Schunn , C . D . ( 2010 ) . From uncertainty exact to certainly vague : Epistemic uncertainty and approximation in science and engineering problem solv - ing . In B . H . Ross ( Ed . ) , The psychology of learning and motivation , vol . 53 ( pp . 227 \u2013 252 ) . Burlington : Academic Press . Schunn , C . D . , Saner , L . D . , Kirschenbaum , S . K . , Trafton , J . G . , & Littleton , E . B . ( 2007 ) . Complex visual data analysis , uncertainty , and representation . In M . C . Lovett , & P . Shah ( Eds . ) , Thinking with data . Mahwah , NJ : Erlbaum . Schunn , C . D . , & Trafton , J . G . ( 2012 ) . The psychology of uncertainty in scienti \ufb01 c data analysis . In G . Feist & M . Gorman ( Eds . ) , Handbook of the psychology of science ( pp . 461 \u2013 483 ) . New York , NY : Springer Publishing . Shaw , J . D . , Zhu , J . , Duffy , M . K . , Scott , K . L . , Shih , H . - A . , & Susanto , E . ( 2011 ) . A contingency model of con \ufb02 ict and team effectiveness . Journal of Applied Psychology , 96 , 391 \u2013 400 . DOI : 10 . 1037 / a0021340 . Singh , J . , & Fleming , L . ( 2009 ) . Lone inventors as sources of breakthroughs : Myth or reality ? Management Science , 56 ( 1 ) , 41 \u2013 56 . DOI : 10 . 1287 / mnsc . 1090 . 1072 . Squyres , S . ( 2005 ) . Roving Mars : Spirit , Opportunity , and the exploration of the red planet . New York : Hyperion . Smith , C . P . ( 2000 ) . Content analysis and narrative analysis . In H . T . Reis & C . M . Judd ( Eds . ) , Handbook of research methods in social and person - ality psychology ( pp . 313 \u2013 335 ) . Cambridge , UK : Cambridge University Press . Todorova , G . , Bear , J . B . , & Weingart , L . R . ( 2014 ) . Can con \ufb02 ict be ener - gizing ? A study of task con \ufb02 ict , positive emotions , and job satisfaction . Journal of Applied Psychology , 99 , 451 \u2013 467 . DOI : 10 . 1037 / a0035134 Tollinger , I . , Schunn , C . D . , & Vera , A . H . ( 2006 ) . What changes when a large team becomes more expert ? In Proceedings of the 28th Annual Conference of the Cognitive Science Society . Mahwah , NJ : Erlbaum . Trickett , S . B . , Trafton , J . G . , Saner , L . D . , & Schunn , C . D . ( 2007 ) . \u2018 I don \u2019 t know what \u2019 s going on there \u2019 : The use of spatial transformations to deal with and resolve uncertainty in complex visualizations . In M . C . Lovett & P . Shah ( Eds . ) , Thinking with data ( pp . 65 \u2013 86 ) . Mahwah , NJ : Erlbaum . Tuckman , B . W . ( 1965 ) . Developmental sequence in small groups . Psycho - logical Bulletin , 63 , 384 \u2013 399 . doi : 10 . 1037 / h0022100 Tuckman , B . W . , & Jensen , M . A . C . ( 1977 ) . Stages of small - group devel - opment revisited . Group Organization Management , 2 , 419 \u2013 427 . DOI : 10 . 1177 / 105960117700200404 . Vuchinich , S . ( 1987 ) . Starting and stopping spontaneous family con \ufb02 icts . Journal of Marriage and the Family , 49 , 591 \u2013 601 . Weingart , L . R . ( 1997 ) . How did they do that ? The ways and means of studying group processes . Research in Organizational Behavior , 19 , 189 \u2013 239 . Weingart , L . R . , Behfar , K . , Bendersky , C . , Todorova , G . , & Jehn , K . A . ( 2014 ) . The directness and oppositional intensity of con \ufb02 ict expression . Academy of Management Review . Advance online publication : doi : 10 . 5465 / amr . 2013 . 0124 Wong , J . T . , Cramer , S . J . , & Gallo , D . A . ( 2012 ) . Age - related reduction of the con \ufb01 dence \u2013 accuracy relationship in episodic memory : Effects of rec - ollection quality and retrieval monitoring . Psychology and Aging , 27 , 1053 \u2013 1065 . DOI : 10 . 1037 / a0027686 . Wuchty , S . , Jones , B . F . , & Uzzi , B . ( 2007 ) . The increasing dominance of teams in production of knowledge . Science , 316 , 1036 \u2013 1039 . DOI : 10 . 1126 / science . 1136099 . 400 S . B . F . Paletz et al . Copyright \u00a9 2016 John Wiley & Sons , Ltd . Appl . Cognit . Psychol . 30 : 387 \u2013 400 ( 2016 ) 10990720 , 2016 , 3 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / ac p . 3213 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 03 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e",
    "chanBestDesignIdeas2015": "Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? Joel Chan , Learning Research and Development Center , University of Pittsburgh , LRDC Room 823 , 3939 O\u2019Hara St , Pittsburgh , PA 15260 , USA Steven P . Dow , Human e Computer - Interaction Institute , Carnegie Mellon University , Pittsburgh , PA , USA Christian D . Schunn , Learning Research and Development Center , University of Pittsburgh , Pittsburgh , PA , USA Design ideas often come from sources of inspiration ( e . g . , analogous designs , prior experiences ) . In this paper , we test the popular but unevenly supported hypothesis that conceptually distant sources of inspiration provide the best insights for creative production . Through text analysis of hundreds of design concepts across a dozen di\ufb00erent design challenges on a Web - based innovation platform that tracks connections to sources of inspiration , we \ufb01nd that citing sources is associated with greater creativity of ideas , but conceptually closer rather than farther sources appear more bene\ufb01cial . This inverse relationship between conceptual distance and design creativity is robust across di\ufb00erent design problems on the platform . In light of these \ufb01ndings , we revisit theories of design inspiration and creative cognition . (cid:1) 2014 Elsevier Ltd . All rights reserved . Keywords : innovation , design cognition , creative design , conceptual design , sources of inspiration W here do creative design ideas come from ? Cognitive scientists have discovered that people inevitably build new ideas from their prior knowledge and experiences ( Marsh , Ward , & Landau , 1999 ; Ward , 1994 ) . While these prior experiences can serve as sources of inspiration ( Eckert & Stacey , 1998 ) and drive sustained creation of ideas that are both new and have high potential for impact ( Hargadon & Sutton , 1997 ; Helms , Vattam , & Goel , 2009 ) , they can also lead designers astray : for instance , de - signers sometimes incorporate undesirable features from existing solutions ( Jansson & Smith , 1991 ; Linsey et al . , 2010 ) , and prior knowledge can make it di\ufb03cult to think of alternative approaches ( German & Barrett , 2005 ; Wiley , 1998 ) . This raises the question : what features of potential inspi - rational sources can predict their value ( and / or potential harmful e\ufb00ects ) ? In this paper , we examine how the conceptual distance of sources relates to their inspirational value . Corresponding author : Joel Chan joc59 @ pitt . edu joelchuc @ cs . cmu . edu www . elsevier . com / locate / destud 0142 - 694X Design Studies - - ( 2014 ) - - e - - http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 1 (cid:1) 2014 Elsevier Ltd . All rights reserved . Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 1 Background 1 . 1 Research base What do we mean by conceptual distance ? Consider the problem of e - waste accumulation : the world generates 20 e 50 million metric tons of e - waste every year , yielding environmentally hazardous additions to land\ufb01lls . A designer might approach this problem by building on near sources like smaller - scale electronics reuse / recycle e\ufb00orts , or by drawing inspiration from a far source like edible food packaging technology ( e . g . , to design re - usable electronics parts ) . What are the relative bene\ufb01ts of di\ufb00erent levels of source conceptual distance along a continuum from near to far ? Many authors , principally those studying the role of analogy in creative prob - lem solving , have proposed that conceptually far sources d structurally similar ideas with many surface ( or object ) dissimilarities d are the best sour - ces of inspiration for creative breakthroughs ( Gentner & Markman , 1997 ; Holyoak & Thagard , 1996 ; Poze , 1983 ; Ward , 1998 ) . This proposal d here called the Conceptual Leap Hypothesis d is consistent with many anecdotal accounts of creative breakthroughs , from Kekule\u2019s discovery of the structure of benzene by visual analogy to a snake biting its tail ( Findlay , 1965 ) , to George Mestral\u2019s invention of Velcro by analogy to burdock root seeds ( Freeman & Golden , 1997 ) , to more recent case studies ( Enkel & Gassmann , 2010 ; Kalogerakis , Lu , & Herstatt , 2010 ) . However , empirical support for this proposal is mixed . Some studies have shown an advantage of far over near sources for novelty , quality , and \ufb02ex - ibility of ideation ( Chan et al . , 2011 ; Chiu & Shu , 2012 ; Dahl & Moreau , 2002 ; Gonc\u00b8alves , Cardoso , & Badke - Schaub , 2013 ; Hender , Dean , Rodgers , & Jay , 2002 ) ; but , some in vivo studies of creative cognition have not found strong connections between far sources and creative mental leaps ( Chan & Schunn , 2014 ; Dunbar , 1997 ) , and other experiments have demonstrated equivalent bene\ufb01ts of far and near sources ( Enkel & Gassmann , 2010 ; Malaga , 2000 ) . Relatedly , Tseng , Moss , Cagan , and Kotovsky ( 2008 ) showed that far sources were more impactful after ideation had already begun ( vs . before ideation ) , providing more functionally distinct ideas than near or control , but both far and near sources led to similar levels of novelty . Similarly , Wilson , Rosen , Nelson , and Yen ( 2010 ) showed no advantage of far over near sources for novelty of ideas ( although near but not far sources decreased variety of ideas ) . Fu et al . ( 2013 ) even found that far sources led to lower novelty and quality of ideas than near sources . Thus , more empirical work is needed to determine whether the Conceptual Leap Hypothesis is well supported . Further , Fu et al . ( 2013 ) argue there is an inverted U - shape function in which moderate distance is best , suggesting 2 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 the importance of conceptualizing and measuring distance along a continuum . 1 . 2 Impetus for the current work Key methodological shortcomings in prior work further motivate more and better empirical work . Prior studies may be too short ( typically 30 min to 1 h ) to convert far sources into viable concepts . To successfully use far sources , designers must spend considerable cognitive e\ufb00ort to ignore irrelevant surface details , attend to potentially insightful structural similarities , and adapt the source to the target context . Additionally , many far sources may yield shallow or unusable inferences ( e . g . , due to non - alignable di\ufb00erences in structural or surface features ; Perkins , 1997 ) ; thus , designers might have to sift through many samples of far sources to \ufb01nd \u2018hidden gems . \u2019 These higher processing costs for far sources might partially explain why some studies show a negative impact of far sources on the number of ideas generated ( Chan et al . , 2011 ; Hender et al . , 2002 ) . In the context of a short task , these processing costs might take up valuable time and resources that could be used for other important as - pects of ideation ( e . g . , iteration , idea selection ) ; in contrast , in real - world design contexts , designers typically have days , weeks or even months ( not an hour ) to consider and process far sources . A second issue is a lack of statistical power . Most existing experimental studies have N (cid:1) 12 per treatment cell ( Chiu & Shu , 2012 ; Hender et al . , 2002 ; Malaga , 2000 ) ; only four studies had N (cid:3) 18 ( Chan et al . , 2011 ; Fu et al . , 2013 ; Gonc\u00b8alves et al . , 2013 ; Tseng et al . , 2008 ) , and they are evenly split in support / opposition for the bene\ufb01ts of far sources . Among the few correlational studies , only Dahl and Moreau ( 2002 ) had a well powered study design in this regard , with 119 participants and a reasonable range of concep - tual distance . Enkel and Gassmann ( 2010 ) only examined 25 cases , all of which were cases of cross - industry transfer ( thus restricting the range of con - ceptual distance being considered ) . This lack of statistical power may have led to a proliferation of false negatives ( potentially exacerbated by small or potentially zero e\ufb00ects at short time scales ) , but possibly also severely over - estimated e\ufb00ect sizes or false positives ( Button et al . , 2013 ) ; more adequately powered studies are needed for more precise estimates of the e\ufb00ects of con - ceptual distance . A \ufb01nal methodological issue is problem variation . Many experimental studies focused on a single design problem . The inconsistent outcomes in these studies may be partially due to some design problems having unique characteristics , e . g . , coincidentally having good solutions that overlap with concepts in far sources . Indeed , Chiu and Shu ( 2012 ) , who examined multiple design prob - lems , observed inconsistent e\ufb00ects across problems . Other investigations of design stimuli have also observed problem variation for e\ufb00ects ( Goldschmidt & Smolkov , 2006 ; Liikkanen & Perttula , 2008 ) . Inspiration source distance and design ideation 3 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 This paper contributes to theories of design inspiration by 1 ) reporting the re - sults of a study that addresses these methodological issues to yield clearer ev - idence , and 2 ) ( to foreshadow our results ) re - examining theories of design inspiration and conceptual distance in light of accumulating preponderance of evidence against the Conceptual Leap Hypothesis . 2 Methods 2 . 1 Overview of research context The current work is conducted in the context of OpenIDEO ( www . openideo . - com ) , a Web - based crowd - sourced innovation platform that addresses a range of social and environmental problems ( e . g . , managing e - waste , increasing accessibility in elections ) . The OpenIDEO designers , with expertise in design processes , guide contributors to the platform through a structured design pro - cess to produce concepts that are ultimately implemented for real - world impact ( \u2018Impact Stories , \u2019 n . d . ) . For this study , we focus on three crucial early stages in the process : \ufb01rst , in the inspiration phase ( lasting between 1 . 5 and 4 weeks , M \u00bc 3 . 1 ) , contributors post inspirations ( e . g . , descriptions of solutions to analogous problems and case studies of stakeholders ) , which help to de\ufb01ne the problem space and identify promising solution approaches ; then , in the concepting phase ( lasting the next 2 e 6 weeks , m \u00bc 3 . 4 ) , contributors post con - cepts , i . e . , speci\ufb01c solutions to the problem . Figure 1 shows an example concept ; it is representative of the typical length and level of detail in concepts , i . e . , w 150 words on average , more detail than one or two words / sentences / sketches , but less detail than a full - \ufb02edged design report / presentation or patent application . Finally , a subset of these concepts is shortlisted by an expert panel ( composed of the OpenIDEO designers and a set of domain experts / stake - holders ) for further re\ufb01nement , based on their creative potential . In later stages , these concepts are re\ufb01ned and evaluated in more detail , and then a sub - set of them is selected for implementation . We focus on the \ufb01rst three stages given our focus on creative ideation ( the later stages involve many other design processes , such as prototyping ) . The OpenIDEO platform has many desirable properties as a research context for our work , including the existence of multiple design problems , thousands of concepts and inspirations , substantive written descriptions of ideas to enable e\ufb03cient text - based analyses , and records of feedback received for each idea , another critical factor in design success . A central property for our research question is the explicit nature of sources of inspiration in the OpenIDEO work\ufb02ow . The site encourages contributors to build on others\u2019 ideas . Importantly , when posting concepts or inspirations , contributors are prompted to cite any concepts or inspirations that serve as sources of inspira - tion for their idea . Also , when browsing other concepts / inspirations , they are able to also see concepts / inspirations the given concept / inspiration \u2018built upon\u2019 ( i . e . , cited as explicit sources of inspiration ; see Figure 2 ) . This culture 4 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 Figure 1 Example concept illustrating the typical amount of detail per concept Figure 2 Depiction of OpenIDEO citation work\ufb02ow . When posting concepts / inspirations , users are prompted to cite concepts / inspirations they \u2018build upon\u2019 by dragging bookmarked concepts / inspirations ( middle panel ) to the citation area ( left panel ) . Users can also search for related concepts / inspirations at this step ( middle panel ) . These cited sources then show up as metadata for the concept / inspiration ( right panel ) Inspiration source distance and design ideation 5 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 of citing sources is particularly advantageous , given that people generally forget to monitor or cite their sources of inspiration ( Brown & Murphy , 1989 ; Marsh , Landau , & Hicks , 1997 ) , and our goal is to study the e\ufb00ects of source use . While users might still forget to cite sources , these platform fea - tures help ensure higher rates of source monitoring than other naturalistic ideation contexts . We note that this operationalization of sources as self - identi\ufb01ed citations precludes consideration of implicit stimulation ; however , the Conceptual Leap Hypothesis may be more applicable to conscious inspi - ration processes ( e . g . , analogy , for which conscious processing is arguably an important de\ufb01ning feature ; Schunn & Dunbar , 1996 ) . 2 . 2 Sample and initial data collection The full dataset for this study consists of 2341 concepts posted for 12 completed challenges by 1190 unique contributors , citing 4557 unique inspira - tions ; 241 ( 10 % ) of these concepts are shortlisted for further re\ufb01nement . See Table 2 for a description of the 12 challenges ( with some basic metadata on each challenge ) . Figure 3 shows the full - text design brief for two challenges . Figure 3 Full - text of challenge briefs from two OpenIDEO challenges 6 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 With administrator permission , we downloaded all inspirations and concepts ( which exist as individual webpages ) and used an HTML parser to extract the following data and metadata : 1 ) Concept / inspiration author ( who posted the concept / inspiration ) 2 ) Number of comments ( before the re\ufb01nement phase ) 3 ) Shortlist status ( yes / no ) 4 ) List of cited sources of inspiration 5 ) Full - text of concept / inspiration Not all concepts cited inspirations as sources . Of the 2341 concepts , 707 ( posted by 357 authors ) cited at least one inspiration , collectively citing 2245 unique inspirations . 110 of these concepts ( w 16 % ) were shortlisted ( see Table 1 for a breakdown by challenge ) . This set of 707 concepts is the primary sample for this study ; the others serve as a contrast to examine the value of explicit building at all on prior sources , and to aid in interpretation of any negative or positive e\ufb00ects of variations in distance . Because we only collected publicly available data , we do not have complete information on the expertise of all contributors : however , based on their public pro\ufb01les on OpenIDEO , at least 1 / 3 of the authors in this sample are professionals in design - related disci - plines ( e . g . , user experience / interaction design , communication design , archi - tecture , product / industrial design , entrepreneurs and social innovators , etc . ) and / or domain experts or stakeholders ( e . g . , urban development researcher Table 1 Descriptions and number of posts for OpenIDEO challenges in final analysis sample Name / description # of Inspirations # of Concepts ( shortlisted ) How might we increase the number of registered bone marrow donors to help save more lives ? 186 71 ( 7 ) How might we inspire and enable communities to take more initiative in making their local environments better ? 160 44 ( 11 ) How can we manage e - waste & discarded electronics to safeguard human health & protect our environment ? 60 26 ( 8 ) How might we better connect food production and consumption ? 266 147 ( 10 ) How can technology help people working to uphold human rights in the face of unlawful detention ? 248 62 ( 7 ) How might we identify and celebrate businesses that innovate for world bene\ufb01t and inspire other companies to do the same ? 122 24 ( 13 ) How might we use social business to improve health in low - income communities ? 131 46 ( 11 ) How might we increase social impact with OpenIDEO over the next year ? 67 40 ( 12 ) How might we restore vibrancy in cities and regions facing economic decline ? 558 119 ( 13 ) How might we design an accessible election experience for everyone ? 241 47 ( 8 ) How might we support web entrepreneurs in launching and growing sustainable global businesses ? 88 49 ( 7 ) How can we equip young people with the skills , information and opportunities to succeed in the world of work ? 118 32 ( 3 ) Inspiration source distance and design ideation 7 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 contributing to the vibrant - cities challenge , education policy researcher contributing to the youth - employment challenge , medical professional contributing to the bone - marrow challenge ) . Collectively , these authors ac - counted for approximately half of the 707 concepts in this study . We analyze the impact of the distance of inspirations ( and not cited concepts ) given our focus on ideation processes during \u2018original\u2019 or non - routine design , where designers often start with a problem and only \u2018inspirations\u2019 ( e . g . , infor - mation about the problem or potentially related designs ) rather than routine design ( e . g . , con\ufb01guration or parametric design ) , where designers might be modifying or iterating on existing solutions rather than generating novel ones ( Chakrabarti , 2006 ; Dym , 1994 ; Gero , 2000 ; Ullman , 2002 ) . The Concep - tual Leap Hypothesis maps most clearly to non - routine design . 2 . 3 Measures 2 . 3 . 1 Creativity of concepts We operationalize concept creativity as whether a concept gets shortlisted . Shortlisting is done by a panel of expert judges , including the original chal - lenge sponsors , who have spent signi\ufb01cant time searching for and learning about existing approaches , and the OpenIDEO designers , who are experts in the general domain of creative design , and who have spent considerable time upfront with challenge sponsors learning about and de\ufb01ning the problem space for each challenge . An expert panel is widely considered a \u2018gold standard\u2019 for measuring the crea - tivity of ideas ( Amabile , 1982 ; Baer & McKool , 2009 ; Brown , 1989 ; Sawyer , 2012 ) . Further , we know from conversations with the OpenIDEO team that the panel\u2019s judgments combines consideration of both novelty and useful - ness / appropriateness ( here operationalized as potential for impact ; A . Jablow , personal communication , May 1 , 2014 ) , the standard de\ufb01nition of creativity ( Sawyer , 2012 ) . Since OpenIDEO challenges are novel and unsolved , success - ful concepts are di\ufb00erent from ( and , perhaps more importantly , signi\ufb01cantly better than ) existing unsatisfactory solutions . We use shortlist ( rather than win status ) given our focus on the ideation phase in design ( vs . convergence / re\ufb01nement , which happens after concepts are shortlisted , and can strongly in\ufb02uence which shortlisted concepts get selected as \u2018winners\u2019 for implementation ) . 2 . 3 . 2 Conceptual distance 2 . 3 . 2 . 1 Measurement approach . Measuring conceptual distance is a major methodological challenge , especially when studying large samples of ideation processes ( e . g . , many designs across many design problems ) . The complex and multifaceted nature of typical design problems can make it di\ufb03cult to distin - guish \u2018within\u2019 and \u2018between\u2019 domain sources in a consistent and principled 8 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 manner . Further , using only a binary scale risks losing variance information that could be critical for converging on a more precise understanding of the e\ufb00ects of conceptual distance ( e . g . , curvilinear e\ufb00ects across the continuum of distance ) . Continuous distance measures are an attractive alternative , but can be extremely costly to obtain at this scale , especially for naturalistic sour - ces ( e . g . , relatively developed text descriptions vs . simple sketches or one - to - two sentence descriptions ) . Human raters may su\ufb00er from high levels of fa - tigue , resulting in poor reliability or drift of standards . We address this methodological challenge with probabilistic topic modeling ( Blei , 2012 ; Steyvers & Gri\ufb03ths , 2007 ) , a major computational approach for understanding large collections of unstructured text . They are similar to other unsupervised machine learning methods d e . g . , K - means clustering , and Latent Semantic Analysis ( Deerwester , Dumais , Furnas , & Landauer , 1990 ) d but distinct in that they emphasize human understanding of not just the relationship between documents in a collection , but the \u2018reasons\u2019 for the hy - pothesized relationships ( e . g . , the \u2018meaning\u2019 of particular dimensions of vari - ation ) , largely because the algorithms underlying these models tend to produce dimensions in terms of clusters of tightly co - occurring words . Thus , they have been used most prominently in applications where understanding of a corpus , not just information retrieval performance , is a high priority goal , e . g . , knowl - edge discovery and information retrieval in repositories of scienti\ufb01c papers ( Gri\ufb03ths & Steyvers , 2004 ) , describing the structure and evolution of scienti\ufb01c \ufb01elds ( Blei & La\ufb00erty , 2006 , 2007 ) , and discovering topical dynamics in social media use ( Schwartz et al . , 2013 ) . We use Latent Dirichlet Allocation ( LDA ; Blei , Ng , Jordan , & La\ufb00erty , 2003 ) , the simplest topic model . LDA assumes that documents are composed of a mixture of latent \u2018topics\u2019 ( occurring with di\ufb00erent \u2018weights\u2019 in the mixture ) , which in turn generate the words in the documents . LDA de\ufb01nes topics as probability distributions over words : for example , a \u2018genetics\u2019 topic can be thought of as a probability distribution over the words { phenotype , popula - tion , transcription , cameras , quarterbacks } , such that words closely related to the topic { phenotype , population , transcription } have a high probability in that topic , and words not closely related to the topic { cameras , quarter - backs } have a very low probability . Using Bayesian statistical learning algo - rithms , LDA infers the latent topical structure of the corpus from the co - occurrence patterns of words across documents . This topical structure in - cludes 1 ) the topics in the corpus , i . e . , the sets of probability distributions over words , and 2 ) the topic mixtures for each document , i . e . , a vector of weights for each of the corpus topics for that document . We can derive concep - tual similarity between any pair of documents by computing the cosine be - tween their topic - weight vectors . In essence , documents that share dominant topics in similar relative proportions are the most similar . Inspiration source distance and design ideation 9 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 Here , we used the open - source MAchine Learning for LanguagE Toolkit ( MALLET ; McCallum , 2002 ) to train an LDA model with 400 topics for all documents in the full dataset , i . e . , 2341 concepts , 4557 inspirations , and 12 challenge briefs ( 6910 total documents ) . Additional technical details on the model - building procedure are available in Appendix A . Resulting cosines be - tween inspirations and the challenge brief ranged from 0 . 01 to 0 . 91 ( M \u00bc 0 . 21 , SD \u00bc 0 . 18 ) , a fairly typical range for large - scale information retrieval applica - tions ( Jessup & Martin , 2001 ) . 2 . 3 . 2 . 2 Validation . Since we use LDA\u2019s measures of conceptual distance as a substitute for human judgments , we validate the adequacy of our topic model using measures of \ufb01t with human similarity judgments on a subset of the data by trained human raters . Five trained raters used a Likert - type scale to rate 199 inspirations from one OpenIDEO challenge for similarity to their challenge brief , from 1 ( very dissim - ilar ) to 6 ( extremely similar ) . Raters were given the intuition that the rating would approximately track the proportion of \u2018topical overlap\u2019 between each inspiration and the challenge brief , or the extent to which they are \u2018about the same thing . \u2019 The design challenge context was explicitly deemphasized , so as to reduce the in\ufb02uence of individual di\ufb00erences in perceptions of the \u2018relevance\u2019 of sources of inspiration . Thus , the raters were instructed to treat all the docu - ments as \u2018documents\u2019 ( e . g . , an article about some topics , vs . \u2018problem solution\u2019 ) and consciously avoid judging the \u2018value\u2019 of the inspirations , simply focusing on semantic similarity . Raters listed major topics in the challenge brief and eval - uated each inspiration against those major topics . To ensure internal consis - tency , the raters also sorted the inspirations by similarity after every 15 e 20 judgments . They then inspected the rank ordering and composition of inspira - tions at each point in the scale , and made adjustments if necessary ( e . g . , if an inspiration previously rated as \u20181\u2019 now , in light of newly encountered inspira - tions , seemed more like a \u20182\u2019 or \u20183\u2019 ) . Although the task was di\ufb03cult , the mean ratings across raters had an acceptable aggregate consistency intra - class corre - lation coe\ufb03cient ( ICC ( 2 , 5 ) ) of 0 . 74 ( mean inter - coder correlation \u00bc 0 . 36 ) . LDA cosines correlated highly , at r \u00bc 0 . 51 , 95 % CI \u00bc [ 0 . 40 , 0 . 60 ] , with the contin - uous human similarity judgments ( see Figure 4A ) . We note that this correlation is better than the highest correlation between human raters ( r \u00bc 0 . 48 ) , reinforc - ing the value of automatic coding methods for this di\ufb03cult task . For comparability with prior work , we also measure \ufb01t with binary ( within - vs . between - domain ) distance ratings . Two raters also classi\ufb01ed 345 inspirations from a di\ufb00erent challenge as either within - or between - domain . Raters \ufb01rst collaboratively de\ufb01ned the problem domain , focusing on the question , \u2018What is the problem to be solved ? \u2019 before rating inspirations . Within - domain inspi - rations were information about the problem ( e . g . , stakeholders , constraints ) 10 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 and existing prior solutions for very similar problems , while between - domain inspirations were information / solutions for analogous or di\ufb00erent problems . Reliability for this measure was acceptable , with an overall average kappa of 0 . 78 ( 89 % agreement ) . All disagreements were resolved by discussion . Similar to the continuous similarity judgments , the point biserial correlation between the LDA - derived cosine and the binary judgments was also high , at 0 . 50 , 95 % CI \u00bc [ 0 . 42 , 0 . 58 ] . The mean cosine to the challenge brief was also higher for within - domain ( M \u00bc 0 . 49 , SD \u00bc 0 . 25 , N \u00bc 181 ) vs . between - domain inspi - rations ( M \u00bc 0 . 23 , SD \u00bc 0 . 20 , N \u00bc 164 ) , d \u00bc 1 . 16 , 95 % CI \u00bc [ 1 . 13 , 1 . 19 ] ( see Figure 4B ) , further validating the LDA approach to measuring distance . Figure 5 shows examples of a near and far inspiration ( from the e - waste chal - lenge ) , along with the top 3 LDA topics ( represented by the top 5 words for that latent topic ) , computed cosine vs . its challenge brief , and human similarity rat - ing . The top 3 topics for the challenge brief are { waste , e , recycling , electronics , electronic } , { waste , materials , recycling , recycled , material } , and { devices , elec - tronics , electronic , device , products } , distinguishing e - waste , general recycling , and electronics products topics . These examples illustrate how LDA is able to e\ufb00ectively extract the latent topical mixture of the inspirations from their text ( inspirations with media also include textual descriptions of the media , miti - gating concerns about loss of semantic information due to using only text as input to LDA ) and also capture intuitions about variations in conceptual dis - tance among inspirations : a document about di\ufb00erent ways of assigning value to possessions is intuitively conceptually more distant from the domain of e - waste than a document about a prior e\ufb00ort to address e - waste . The near and far examples depicted in Figure 5 also represent the range of con - ceptual distance measured in this dataset , with the near inspiration\u2019s cosine of 0 . 64 representing approximately the 90th percentile of similarity to the chal - lenge domain , and the far inspiration\u2019s cosine of 0 . 01 representing approxi - mately the 10th percentile of similarity to the challenge domain . Thus , the range of conceptual distance of inspirations in this data spans approximately Figure 4 ( A ) Scatterplot of LDA cosines vs . averaged human continuous similarity judgments for inspirations in the e - waste challenge . ( B ) . Mean cosine against the challenge brief for within - vs . between - domain inspirations Inspiration source distance and design ideation 11 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 from sources that are very clearly within the domain ( e . g . , an actual solution for the problem of electronic waste involving recycling of materials ) to sources that are quite distant , but not obviously random ( e . g . , an observation of how people assign emotional value to relationships and artifacts ) . This range most likely excludes the \u2018too far\u2019 example designs studied in Fu et al . ( 2013 ) or the \u2018opposite stimuli\u2019 used in Chiu and Shu ( 2012 ) . 2 . 3 . 2 . 3 Final distance measures . The challenge briefs varied in length and speci\ufb01city across challenges , as did mean raw cosines for inspirations . But , these di\ufb00erences in mean similarity were much larger , d \u00bc 1 . 90 , 95 % CI \u00bc [ 1 . 85 e 1 . 92 ] ( for 80 inspirations from 4 challenges with maximally di\ufb00erent mean cosines ) , than for human similarity judgments ( coded sepa - rately but with the same methodology as before ) , d \u00bc 0 . 18 , 95 % CI \u00bc [ e 0 . 05 to 0 . 43 ] . This suggested that between - challenge di\ufb00erences were more an artifact of variance in challenge brief length / speci\ufb01city . Thus , to ensure meaningful comparability across challenges , we normalized the cosines by computing the z - score for each inspiration\u2019s cosine relative to other inspi - rations from the same challenge before analyzing the results in the full dataset . However , similar results are found using raw cosines , but with more uncer - tainty in the statistical coe\ufb03cient estimates . We then subtracted the cosine z - score from zero such that larger values meant more distant . From these \u2018reversed\u2019 cosine z - scores , two di\ufb00erent distance Figure 5 Topics found by LDA within examples of near and far inspirations for the e - waste challenge 12 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 measures were computed to tease apart possibly distinct e\ufb00ects of source dis - tance : 1 ) max distance ( DIST MAX ) , i . e . , the distance of a concept\u2019s furthest source from the problem domain and 2 ) mean distance ( DIST MEAN ) of the concept\u2019s sources . DIST MAX estimates \u2018upper bounds\u2019 for the bene\ufb01ts of dis - tance : do the best ideas really come from the furthest sources ? DIST MEAN cap - italizes on the fact that many concepts relied on multiple inspirations and estimates the impact of the relative balance of relying on near vs . far sources ( e . g . , more near than far sources , or vice versa ) . 2 . 3 . 3 Control measures Given our correlational approach , it is important to identify and rule out or adjust for other important factors that may in\ufb02uence the creativity of concepts ( particularly in the later stages , where prototyping and feedback are especially important ) and may be correlated with the predictor variables . Feedback . Given the collaborative nature of OpenIDEO , we reasoned that feedback in the form of comments ( labeled here as FEEDBACK ) in\ufb02uences success . Comments can o\ufb00er encouragement , raise issues / questions , or pro - vide speci\ufb01c suggestions for improvement , all potentially signi\ufb01cantly enhancing the quality of the concept . Further , feedback may be an alternate pathway to success via source distance , in that concepts that build on far sour - ces may attract more attention and therefore higher levels of feedback , which then improve the quality of the concept . Quality of cited sources . Concepts that build on existing high - quality concepts ( e . g . , those who end up being shortlisted or chosen as winners ) have a partic - ular advantage of being able to learn from the mistakes and shortcomings , good ideas , and feedback in these high - quality concepts . Thus , as a proxy mea - sure of quality , the number of shortlisted concepts a given concept builds upon ( labeled SOURCESHORT ) could be a large determinant of a concept\u2019s success . 2 . 4 Analytic approach We are interested in predicting the creative outcomes of 707 concepts , posted by 357 authors for 12 di\ufb00erent design challenges . Authors are not cleanly nested within challenges , nor vice versa ; our data are cross - classi\ufb01ed , with concepts cross - classi\ufb01ed within both authors and challenges ( see Figure 6 ) . This cross - classi\ufb01ed structure violates assumptions of uni - form independence between concepts : concepts posted by the same author or within the same challenge may be more similar to each other . Failing to account for this non - independence could lead to overestimates of the sta - tistical signi\ufb01cance of model estimates ( i . e . , make unwarranted claims of sta - tistically signi\ufb01cant e\ufb00ects ) . This issue is exacerbated when testing for small e\ufb00ects . Additionally , modeling between - author e\ufb00ects allows us to separate author - e\ufb00ects ( e . g . , higher / lower creativity ) from the impact of sources on Inspiration source distance and design ideation 13 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 individual concepts Thus , we employ generalized linear mixed models ( also called hierarchical generalized linear models ) to model both \ufb01xed e\ufb00ects ( of our independent and control variables ) and random e\ufb00ects ( potential varia - tion of the outcome variable attributable to author - or challenge - nesting and also potential between - challenge variation in the e\ufb00ect of distance ) on short - list status ( a binary variable , which requires logistic , rather than linear , regression ) . An initial model predicting the outcome with only the intercept and between - challenge and - author variation con\ufb01rms the presence of signi\ufb01cant non - independence , with between - author and between - challenge variation in short - list outcomes estimated at 0 . 44 , and 0 . 50 , respectively . The intra - class correla - tions for author - level and challenge - level variance in the intercept are w 0 . 11 and 0 . 13 , respectively , well above the cuto\ufb00 recommended by Raudenbush and Bryk ( 2002 ) . 1 3 Results 3 . 1 Descriptive statistics On average , 16 % of concepts in the sample get shortlisted ( see Table 2 ) . DIS - T MEAN is centered approximately at 0 , re\ufb02ecting our normalization procedure . Both DIST MAX and DIST MEAN have a fair degree of negative skew . SOUR - CESHORT and FEEDBACK have strong positive skew ( most concepts either have few comments or cite 0 or 1 shortlisted concepts ) . There is a strong positive relationship between DIST MAX and DIST MEAN ( see Table 3 ) . All variables have signi\ufb01cant bivariate correlations with SHORT - LIST except for DIST MAX ; however , since it is a substantive variable of inter - est , we will model it nonetheless . Controlling for other variables might enable us to detect subtle e\ufb00ects . Figure 6 Illustrated cross - classi\ufb01ed structure of the data 14 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 3 . 2 Statistical models We estimated separate models for the e\ufb00ects of DIST MAX and DIST MEAN , each controlling for challenge - and author - nesting , FEEDBACK , and SHORTSOURCE . 3 . 2 . 1 Max distance Our model estimated an inverse relationship between DIST MAX and Pr ( shortlist ) , such that a 1 - unit increase in DIST MAX predicted a 0 . 33 decrease in the log - odds of being shortlisted , after accounting for the e\ufb00ects of FEEDBACK , SHORTSOURCE , and challenge - and author - level nesting , p < . 05 ( see Appendix B for technical details on the statistical models ) . How - ever , this coe\ufb03cient was estimated with considerable uncertainty , as indi - cated by the large con\ufb01dence intervals ( coe\ufb03cient could be as small as (cid:4) 0 . 06 or as large as (cid:4) 0 . 60 ) ; considering also the small bivariate correlation with SHORTLIST , we are fairly certain that the \u2018true\u2019 coe\ufb03cient is not pos - itive ( contra the Conceptual Leap Hypothesis ) , but we are quite uncertain about its magnitude . Figure 7 visually displays the estimated relationship between DIST MAX and Pr ( shortlist ) , evaluated at mean values of feedback and shortlisted sources . To aid interpretation , we also plot the predicted Pr ( shortlist ) for concepts that cite no sources using a horizontal gray bar ( bar width indicates uncertainty in estimate of Pr ( shortlist ) ) : concepts with approximately equivalent amounts of feedback ( i . e . , mean of 8 . 43 ) , have a predicted Pr ( shortlist \u00bc 0 . 09 , 95 % CI \u00bc [ 0 . 07 e 0 . 11 ] ; using a logistic model , the coe\ufb03cient for \u2018any citation\u2019 ( con - trolling for feedback ) is 0 . 31 , 95 % CI \u00bc [ 0 . 01 e 0 . 62 ] ) . This bar serves as an approximate \u2018control\u2019 group , allowing us to interpret the e\ufb00ect not just in terms of the e\ufb00ects of far sources relative to near sources , but also in comparison with using no sources . Comparing the \ufb01tted curve with this bar highlights how the Table 2 Descriptive statistics Variable Valid N Min Max Mean Median SD SHORTLIST 707 0 . 00 1 . 00 0 . 16 0 . 00 0 . 36 DIST MAX 707 (cid:4) 3 . 85 1 . 90 0 . 45 0 . 76 0 . 85 DIST MEAN 707 (cid:4) 3 . 85 1 . 67 (cid:4) 0 . 10 0 . 01 0 . 85 SOURCESHORT 707 0 11 0 . 51 0 0 . 96 FEEDBACK 707 0 67 8 . 43 6 9 . 45 Table 3 Bivariate correlations Variable DIST MAX DIST MEAN SOURCESHORT FEEDBACK SHORTLIST (cid:4) 0 . 05 (cid:4) 0 . 10 * 0 . 11 * * 0 . 33 * * * DIST MAX 0 . 77 * * * 0 . 05 0 . 07 m DIST MEAN (cid:4) 0 . 05 0 . 01 SOURCESHORT 0 . 12 * * m p < . 10 ; * p < . 05 ; * * p < . 01 ; * * * p < . 001 . Inspiration source distance and design ideation 15 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 advantage of citing vs . not citing inspirations seems to be driven mostly by cit - ing relatively near inspirations : Pr ( shortlist ) for concepts that cite far inspira - tions converges on that of no - citation concepts . We emphasize again that , despite the uncertainty in the degree of the negative relationship between DIS - T MAX and Pr ( shortlist ) , the data do not support an inference that the best ideas are coming from the farthest inspirations : rather , relying on nearer rather than farther sources seems to lead to more creative design ideas . Importantly , this pattern of results was robust across challenges on the platform : the model esti - mated essentially zero between - challenge variation in the slope of DIST MAX . c 2 ( 2 ) \u00bc 0 . 05 , p \u00bc . 49 ( see Figure 8 ) . 3 . 2 . 2 Mean distance Similar results were obtained for DIST MEAN . There was a robust inverse rela - tionship between DIST MEAN and Pr ( shortlist ) , such that a 1 - unit increase in DIST MEAN was associated with a decrease of approximately 0 . 40 in the log - odds of being shortlisted , p < . 05 . The estimates of this e\ufb00ect were obtained with similarly low precision regarding the magnitude of the e\ufb00ect , with 95 % CI upper limit of at most B \u00bc (cid:4) 0 . 09 ( but as high as (cid:4) 0 . 71 ) . As shown in Figure 9 , as DIST MEAN increases , Pr ( shortlist ) approaches that of non - citing concepts , again suggesting ( as with DIST MAX ) that the most bene\ufb01cial sources appear to be ones that are relatively close to the challenge domain . Again , as with DIST MAX , this pattern of results did not vary across challenges : our model estimated essentially zero between - challenge variation in the slope of DIST MEAN , c 2 ( 2 ) \u00bc 0 . 07 , p \u00bc . 48 ( see Figure 10 ) . 4 Discussion 4 . 1 Summary and interpretation of \ufb01ndings This study explored how the inspirational value of sources varies with their conceptual distance from the problem domain along the continuum from near to far . The study\u2019s \ufb01ndings provide no support for the notion that the best ideas come from building explicitly on the farthest sources . On the Figure 7 Model - \ufb01tted rela - tionship between DIST MAX and Pr ( shortlist ) , evaluated at mean values of feedback and source shortlist . Grayed lines are \ufb01ts with upper and lower limits for 95 % CI for e\ufb00ect of DIST MAX 16 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 Figure 8 Overall and by - challenge model - \ufb01tted rela - tionship between DIST MAX and Pr ( shortlist ) . Fitted values evaluated at mean values of feedback and source shortlist . Grayed lines are \ufb01ts for each individual challenge Figure 9 Model - \ufb01tted rela - tionship between DIST MEAN and Pr ( shortlist ) , evaluated at mean values of feedback and source shortlist . Grayed lines are \ufb01ts with upper and lower limits for the 95 % CI for the e\ufb00ect of DIST MEAN Figure 10 Overall and by - challenge model - \ufb01tted rela - tionship between DIST MEAN and Pr ( shortlist ) . Fitted values evaluated at mean values of feedback and source shortlist . Grayed lines are \ufb01ts for each individual challenge Inspiration source distance and design ideation 17 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 contrary , the bene\ufb01ts of building explicitly on inspirations seem to accrue mainly for concepts that build more on near than far inspirations . Impor - tantly , these e\ufb00ects were consistently found in all of the challenges , addressing concerns raised about potential problem variation , at least among non - routine social innovation design problems . 4 . 2 Caveats and limitations Some caveats should be discussed before addressing the implications of this study . First , the statistical patterns observed here are conditional : i . e . , we \ufb01nd an inverse relationship between conceptual distance of explicitly cited inspiration sources and Pr ( shortlist ) . Our data are silent on the e\ufb00ects of dis - tance for concepts that did not cite sources ( where lack of citation could indi - cate forgetting of sources or lack of conscious building on sources ) . There is a potential concern over range restriction or attrition due to our reli - ance on self - identi\ufb01ed sources . However , several features of the data help to ameliorate this concern . First , concepts that did not cite sources were overall of lower quality ; thus , it is unlikely that the inverse e\ufb00ects of distance are solely due to attrition ( e . g . , bene\ufb01cial far inspirations not being observed ) . Second , the integration of citations and building on sources into the overall OpenI - DEO work\ufb02ow and philosophy of ideation also helps ameliorate concerns about attrition of far sources . Finally , the dataset included many sources that were quite far away , providing su\ufb03cient data to statistically test the e\ufb00ects of relative reliance on far sources ( even if they are overall under - reported ) . Nevertheless , we should still be cautious about making inferences about the impact of unconscious sources ( since sources in this data are explicitly cited and therefore consciously built upon ) . However , as we note in the methods , Table 4 Model estimates and fit statistics for cross - classified multilevel logistic regressions of Pr ( shortlist ) on DIST MAX , with comparison to baseline model ( controls only ) Baseline model ( controls only ) DIST MAX , \ufb01xed slope DIST MAX , random slope Fixed e\ufb00ects g 00 , intercept (cid:4) 2 . 66 [ (cid:4) 3 . 28 , (cid:4) 2 . 03 ] (cid:4) 2 . 57 [ (cid:4) 3 . 29 , (cid:4) 2 . 05 ] (cid:4) 2 . 57 [ (cid:4) 3 . 29 , (cid:4) 2 . 05 ] g 10 , FEEDBACK 0 . 09 * * * [ 0 . 07 , 0 . 12 ] 0 . 10 * * * [ 0 . 07 , 0 . 12 ] 0 . 10 * * * [ 0 . 07 , 0 . 12 ] g 20 , SOURCESHORT 0 . 14 [ (cid:4) 0 . 08 , 0 . 36 ] 0 . 15 [ (cid:4) 0 . 07 , 0 . 38 ] 0 . 15 [ (cid:4) 0 . 07 , 0 . 38 ] g 30 , DIST MAX (cid:4) 0 . 33 * [ (cid:4) 0 . 60 , (cid:4) 0 . 06 ] (cid:4) 0 . 32 * [ (cid:4) 0 . 59 , (cid:4) 0 . 06 ] Random e\ufb00ects u 0 authorj for intercept 0 . 29 0 . 31 0 . 32 u 0 challengek for intercept 0 . 75 0 . 76 0 . 74 u 3 challengek for DIST MAX 0 . 00 Model \ufb01t statistics Deviance 511 . 39 506 . 04 505 . 99 AIC 521 . 39 518 . 04 521 . 99 m p < . 10 ; * p < . 05 ; * * p < . 01 ; * * * p < . 001 ; 95 % CI ( Wald ) \u00bc [ lower , upper ] . 18 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 the Conceptual Leap Hypothesis maps most cleanly to conscious inspiration processes ( e . g . , analogy ) . Finally , some may be concerned that we have not measured novelty here . Conceivably , the bene\ufb01ts of distance may only be best observed for the novelty of ideas , and not necessarily quality , consistent with some recent work ( Franke , Poetz , & Schreier , 2014 ) . However , novelty per se does not produce creativity ; we contend that to fully understand the e\ufb00ects of distance on design creativity , we must consider its impacts on both novelty and quality together ( as our shortlist measure does ) . 4 . 3 Implications and future directions Overall , our results consistently stand in opposition to the Conceptual Leap Hypothesis . In tandem with prior opposing \ufb01ndings ( reviewed in the introduc - tion ) , our work lends strength to alternative theories of inspiration by theorists like Perkins ( 1983 ) , who argues that conceptual distance does not matter , and Weisberg ( 2009 , 2011 ) , who argues that within - domain expertise is a primary driver of creative cognition . We should be clear that our \ufb01ndings do not imply that no creative ideas come from far sources ( indeed , in our data , some creative ideas did come from far sources ) ; rather , our data suggest that the most crea - tive design ideas are more likely to come from relying on a preponderance of nearer rather than farther sources . However , our data do suggest that highly creative ideas can often come from relying almost not at all on far sources ( as evidenced by the analyses with maximum distance of sources ) . These good ideas may arise from iterative , deep search , a mechanism for creative breakthroughs that may be often overlooked but potentially at least as impor - tant as singular creative leaps ( Chan & Schunn , 2014 ; Dow , Heddleston , & Table 5 Model estimates and fit statistics for cross - classified multilevel logistic regressions of Pr ( shortlist ) on DIST MEAN , with comparison to baseline model ( controls only ) Baseline model ( controls only ) DIST MEAN , \ufb01xed slope DIST MEAN , random slope Fixed e\ufb00ects g 00 , intercept (cid:4) 2 . 66 [ (cid:4) 3 . 28 , (cid:4) 2 . 03 ] (cid:4) 2 . 74 [ (cid:4) 3 . 36 , (cid:4) 2 . 11 ] (cid:4) 2 . 74 [ (cid:4) 3 . 36 , (cid:4) 2 . 11 ] g 10 , FEEDBACK 0 . 09 * * * [ 0 . 07 , 0 . 12 ] 0 . 10 * * * [ 0 . 07 , 0 . 12 ] 0 . 10 * * * [ 0 . 07 , 0 . 12 ] g 20 , SOURCESHORT 0 . 14 [ (cid:4) 0 . 08 , 0 . 36 ] 0 . 13 [ (cid:4) 0 . 09 , 0 . 35 ] 0 . 13 [ (cid:4) 0 . 09 , 0 . 35 ] g 30 , DIST MEAN (cid:4) 0 . 40 * [ (cid:4) 0 . 71 , (cid:4) 0 . 09 ] (cid:4) 0 . 40 * [ (cid:4) 0 . 73 , (cid:4) 0 . 07 ] Random e\ufb00ects u 0 authorj for intercept 0 . 29 0 . 31 0 . 30 u 0 challengek for intercept 0 . 75 0 . 73 0 . 73 u 3 challengek for DIST MEAN 0 . 03 Model \ufb01t statistics Deviance 511 . 39 505 . 13 505 . 06 AIC 521 . 39 517 . 13 521 . 06 m p < . 10 ; * p < . 05 ; * * p < . 01 ; * * * p < . 001 ; 95 % CI ( Wald ) \u00bc [ lower , upper ] . Inspiration source distance and design ideation 19 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 Klemmer , 2009 ; Mecca & Mumford , 2013 ; Rietzschel , Nijstad , & Stroebe , 2007 ; Sawyer , 2012 ; Weisberg , 2011 ) . In light of this and our \ufb01ndings , it may be fruitful to deemphasize the privileged role of far sources and mental leaps in theories of design inspiration and creative cognition . How might this proposed theoretical revision be reconciled with the relatively robust \ufb01nding that problem solvers from outside the problem domain can often produce the most creative ideas ( Franke et al . , 2014 ; Hargadon & Sutton , 1997 ; Jeppesen & Lakhani , 2010 ) ? Returning to our re\ufb02ections on the potential costs of processing far sources , one way to reconcile the two sets of \ufb01ndings might be to hypothesize that expertise in the distant source domain enables the impact of distant ideas by bypassing the cognitive costs of deeply understanding the far domain , and \ufb01lters out shallow inferences that are not likely to lead to deep insights . Hargadon and Sutton\u2019s ( 1997 ) \ufb01nd - ings from their in - depth ethnographic study of the consistently innovative IDEO design \ufb01rm are consistent with an expertise - mediation claim : the \ufb01rm\u2019s cross - domain - inspired innovations appeared to \ufb02ow at the day - to - day process level mainly from deep immersion of its designers in multiple disciplines , and \u2018division of expertise\u2019 within the \ufb01rm , with brainstorms acting as crucial cata - lysts for involving experts from di\ufb00erent domains on projects . However , studies directly testing expertise - mediation are scarce or non - existent . Further , the weight of the present data , combined with prior studies showing no advantage of far sources , suggests that considering alternative mechanisms of outside - domain advantage may be more theoretically fruitful : for instance , perhaps the advantage of outside - domain problem - solvers arises from the di\ufb00erent perspectives they bring to the problem d allowing for more \ufb02exible and alternative problem representations , which may lead to breakthrough in - sights ( Kaplan & Simon , 1990 ; Knoblich , Ohlsson , Haider , & Rhenius , 1999 ; \u20ac Ollinger , Jones , Faber , & Knoblich , 2012 ) . Domain - outsiders may also have a looser attachment to the status quo or prior successful solutions by virtue of being a \u2018newcomer\u2019 to the domain ( Choi & Levine , 2004 ) d leading to higher readiness to consider good ideas that challenge existing assumptions within the domain d rather than knowledge and transfer of di\ufb00erent solutions per se . Finally , it would be interesting to examine potential moderating in\ufb02uences of source processing strategies . In our data , closer sources were more bene\ufb01cial , but good ideas also did come from far sources ; however , as we have argued , it can be more di\ufb03cult to convert far sources into viable concepts . Are there common strategies for e\ufb00ective conversion of far sources , and are they di\ufb00erent from strategies for e\ufb00ectively building on near sources ? For example , one e\ufb00ective strategy for building on sources while avoiding \ufb01xation is to use a schema - based strategy ( i . e . , extract and transfer abstract functional principles rather than concrete solution features ; Ahmed & Christensen , 2009 ; Yu , Kraut , & Kittur , 2014 ) . Are there processing strategies that expert creative 20 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 designers apply uniquely to far sources ( e . g . , to deal with potentially un - alignable di\ufb00erences ) ? Answering this question can shed further light on the variety of ways designers can be inspired by sources to produce creative design ideas . We close by noting the methodological contribution of this work . While we are not the \ufb01rst to use topic modeling to explore semantic meaning in a large collection of documents , we are the \ufb01rst to our knowledge to validate this method in the context of large - scale study of design ideas . We have shown that the topic model approach adequately captures human intuitions about the semantics of the design space , while providing dramatic savings in cost : indeed , such an approach can make more complex research questions ( e . g . , exploring pairwise distances between design idea or , tracing conceptual paths / moves in a design ideation session ) much more feasible without sacri - \ufb01cing too much quality . We believe this approach can be a potentially valuable way for creativity researchers to study the dynamics of idea generation at scale , while avoiding the ( previously inevitable ) tradeo\ufb00 between internal validity ( e . g . , having adequate statistical power ) and external validity ( e . g . , using real , complex design problems and ideas instead of toy problems ) . Appendix A . Topic model technical details A . 1 . Document preprocessing All documents were \ufb01rst tokenized using the TreeBank Tokenizer from the open - source Natural Language Tool Kit Python library ( Bird , Klein , & Loper , 2009 ) . To improve the information content of the document text , we removed a standard list of stopwords , i . e . , highly frequent words that do not carry semantic meaning on their own ( e . g . , \u2018the\u2019 , \u2018this\u2019 ) . We used the open - source MAchine Learning for LanguagE Toolkit\u2019s ( MALLET ; McCallum , 2002 ) stopword list . A . 2 . Model parameter selection We used MALLET to train our LDA model , with asymmetric priors for the topic - document and topic - word distributions , which allows for some words to be more prominent than others and some topics to be more prominent than others , typically improving model \ufb01t and performance ( Wallach , Mimno , & McCallum , 2009 ) . Priors were optimized using MALLET\u2019s in - package optimization option . LDA requires that K ( the number of topics ) be prespeci\ufb01ed by the modeler . Model \ufb01t typically improves with K , with diminishing returns past a certain point . Intuitively , higher K leads to \ufb01ner - grained topical distinctions , but too high K may lead to uninterpretable topics ; on the other hand , too low K would yield too general topics . Further , traditional methods of optimizing K ( computing \u2018perplexity\u2019 , or the likelihood of observing the distribution of Inspiration source distance and design ideation 21 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 words in the corpus given a topic model of the corpus ) do not always correlate with human judgments of model quality ( e . g . , domain expert evaluations of topic quality ; Chang , Gerrish , Wang , Boyd - graber , & Blei , 2009 ) . We explored the following settings of K : [ 12 , 25 , 50 , 100 , 200 , 300 , 400 , 500 , 600 , 700 ] . Because the optimization algorithm for the prior parameters is nondeterministic , models with identical K might produce noticeably di\ufb00erent topic model solutions , e . g . , if the optimization search space is rugged , the al - gorithm might get trapped in di\ufb00erent local maxima . Therefore , we ran 50 models at each K , using identical settings ( i . e . , 1000 iterations of the Gibbs sampler , internally optimizing parameters for the asymmetric priors ) . Figure 11 shows the mean \ufb01t ( with both continuous and binary similarity judg - ments ) at each level of K . Model \ufb01t is generally fairly high at all levels of K , with the continuous judg - ments tending to increase very slightly with K , tapering out past 400 . Fit with binary judgments tended to decrease ( also very slightly ) with K , probably re\ufb02ecting the decreasing utility of increasingly \ufb01ner - grained distinctions for a binary same / di\ufb00erent classi\ufb01cation . Because we wanted to optimize for \ufb01t with human judgments of conceptual distance overall , we selected the level of K at which the divergent lines for \ufb01t with continuous and binary judgments \ufb01rst begin to cross ( i . e . , at K \u00bc 400 ) . Subsequently , we created a combined \u2018\ufb01t\u2019 mea - sure ( sum of the correlation coe\ufb03cients for \ufb01t vs . continuous and binary judg - ments ) , and selected the model with K \u00bc 400 that had the best overall \ufb01t measure . However , as we report in the next section , the results of our analyses are robust to di\ufb00erent settings of K . Figure 11 Mean \ufb01t ( with (cid:5) 1 SE ) vs . human judgments for LDA cosines by level of K Appendix B . Statistical modeling technical details B . 1 . Statistical modeling approach All models were \ufb01tted using the lme4 package ( Bates , Maechler , Bolker , & Walker , 2013 ) in R ( R Core Team , 2013 ) , using full maximum likelihood esti - mation by the Laplace approximation . The following is the general structure of these models ( in mixed model notation ) : 22 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 h i \u00f0 authorjchallengek \u00de \u00bc g 00 \u00fe X q g q 0 X qi \u00fe u 0 authorj \u00fe u 0 challengek where (cid:6) h i \u00f0 authorjchallengek \u00de is the predicted log odds of being shortlisted for the ith concept posted by the j th author in the k th challenge (cid:6) g 00 is the grand mean log odds for all concepts (cid:6) g q 0 is a vector of q predictors ( q \u00bc 0 for our null model ) (cid:6) u 0 authorj and u 0 challengek are the random e\ufb00ects contribution of variation between - authors and between - challenges for mean g 00 ( i . e . , how much a given author or challenge varies from the mean ) A baseline model with only control variables and variance components was \ufb01rst \ufb01tted . Then , for the models for both DIST MAX and DIST MEAN , we \ufb01rst estimated a model with a \ufb01xed e\ufb00ect of distance , and then a random e\ufb00ect ( to test for problem variation ) . These random slopes models include the addi - tional parameter u 1 challengek that models the between - challenge variance component for the slope of distance . B . 2 . Model selection Estimates and test statistics for each step in our model - building procedure are shown in Tables 4 and 5 . We \ufb01rst \ufb01tted a model predicting Pr ( shortlist ) with our control variables to serve as a baseline for evaluating the predictive power of our distance measures . The baseline model estimates a strong positive e\ufb00ect of FEEDBACK , estimated with high precision : each additional comment added 0 . 10 [ 0 . 07 , 0 . 12 ] to the log - odds of being shortlisted , p < . 001 . The model also estimated a positive e\ufb00ect of SHORTSOURCE , B \u00bc 0 . 14 [ e 0 . 08 , 0 . 36 ] but with poor precision , and falling short of conventional statistical signi\ufb01 - cance , p \u00bc . 21 ; nevertheless , we leave it in the model for theoretical reasons . The baseline model is a good \ufb01t to the data , reducing deviance from the null model ( with no control variables ) by a large and statistically signi\ufb01cant amount , c 2 ( 1 ) \u00bc 74 . 35 , p \u00bc . 00 . For the \ufb01xed slope model for DIST MAZ , adding the coe\ufb03cient for results in a signi\ufb01cant reduction in deviance from the baseline model , c 2 ( 2 ) \u00bc 0 . 13 , p \u00bc . 47 . The random slope model did not signi\ufb01cantly reduce deviance in com - parison with the simpler \ufb01xed slope model , c 2 ( 2 ) \u00bc 0 . 05 , p \u00bc . 49 ( p - value is halved , heeding common warnings that a likelihood ratio test discriminating two models that di\ufb00er on only one variance component may be overly conser - vative , e . g . , Pinheiro & Bates , 2000 ) . Also , the Akaike Information Criterion ( AIC ) increases from the \ufb01xed to random slope model . Thus , we select the \ufb01xed slope model ( i . e . , no problem - variation ) as our best estimate of the e\ufb00ects of DIST MAX . This \ufb01nal model has an overall deviance reduction vs . null at c 2 ( 3 ) \u00bc 79 . 71 , p \u00bc . 00 . Inspiration source distance and design ideation 23 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 We used the same procedure for model selection for the DIST MEAN models . The \ufb01xed slope model results in a small but signi\ufb01cant reduction in deviance from the baseline model , c 2 ( 1 ) \u00bc 6 . 27 , p \u00bc . 01 . Adding the variance compo - nent for the slope of DIST MEAN increases the AIC , and does not signi\ufb01cantly reduce deviance , c 2 ( 2 ) \u00bc 0 . 07 , p \u00bc . 48 ( again , p - value here is halved to correct for overconservativeness ) . Thus , again we select the \ufb01xed slope model as our \ufb01nal model for the e\ufb00ects of DIST MEAN . This \ufb01nal model has an overall reduc - tion in deviance from the null model of about c 2 ( 3 ) \u00bc 80 . 61 , p \u00bc . 00 . B . 3 . Robustness and sensitivity We tested the robustness of our coe\ufb03cient estimates by calculating outlier in - \ufb02uence statistics using the in\ufb02uence . measures method in the stats package in R , applied to logistic regression model variants of both the DIST MEAN and DIST MAX models ( i . e . , without author - and challenge - level variance compo - nents ; coe\ufb03cient estimates are almost identical to the \ufb01xed slope multilevel models ) : DFBETAS and Cook\u2019s Distance measures were below recommended thresholds for all data points ( Fox , 2002 ) . Addressing potential concerns about sensitivity to topic model parameter set - tings , we also \ufb01tted the same \ufb01xed slope multilevel models using recomputed conceptual distance measures for the top 20 ( best - \ufb01tting ) topic models at K \u00bc 200 , 300 , 400 , 500 , and 600 ( total of 100 models ) . All models produced negative estimates for the e\ufb00ect of both DIST MEAN and DIST MAX , with poorer precision for lower K . Thus , our results are robust to di\ufb00erent settings of K for the topic models . We also address potential concerns about interactions with expertise by \ufb01tting a model that allowed the slope of distance to vary by authors . In this model , the overall mean e\ufb00ect of distance remained almost identical ( B \u00bc (cid:4) 0 . 46 ) , and the model\u2019s \ufb01t was not signi\ufb01cantly better than the \ufb01xed slope model , c 2 ( 3 ) \u00bc 3 . 44 , p \u00bc . 16 , indicating a lack of statistically signi\ufb01cant between - author variability for the slope of distance . Finally , we also \ufb01tted models that considered not just immediately cited inspi - rations , but also indirectly cited inspirations ( i . e . , inspirations cited by cited in - spirations ) , and they too yielded almost identical coe\ufb03cient estimates and con\ufb01dence intervals . References Ahmed , S . , & Christensen , B . T . ( 2009 ) . An in situ study of analogical reasoning in novice and experienced designer engineers . Journal of Mechanical Design , 131 ( 11 ) , 111004 . Amabile , T . M . ( 1982 ) . Social psychology of creativity : a consensual assessment technique . Journal of Personality and Social Psychology , 43 ( 5 ) , 997 e 1013 . Baer , J . , & McKool , S . S . ( 2009 ) . Assessing creativity using the consensual assess - ment technique . In C . S . Schreiner ( Ed . ) , Handbook of research on assessment 24 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 technologies , methods , and applications in higher education ( pp . 65 e 77 ) , Her - shey , PA . Bates , D . , Maechler , M . , Bolker , B . , & Walker , S . ( 2013 ) . Lme4 : Linear mixed - e\ufb00ects models using eigen and S4 . R package version 1 . 0 - 5 . [ Computer soft - ware ] . Retrieved from . http : / / CRAN . R - project . org / package \u00bc lme4 . Bird , S . , Klein , E . , & Loper , E . ( 2009 ) . Natural language processing with python . O\u2019Reilly Media Inc . Blei , D . M . ( 2012 ) . Probabilistic topic models . Communications of the ACM , 55 ( 4 ) , 77 e 84 . Blei , D . M . , & La\ufb00erty , J . D . ( 2006 ) . Dynamic topic models . In Proceedings of the 23rd international conference on machine learning ( pp . 113 e 120 ) . Blei , D . M . , & La\ufb00erty , J . D . ( 2007 ) . A correlated topic model of science . The Annals of Applied Statistics 17 e 35 . Blei , D . M . , Ng , A . Y . , Jordan , M . I . , & La\ufb00erty , J . ( 2003 ) . Latent Dirichlet allocation . Journal of Machine Learning Research 993 e 1022 . Brown , R . T . ( 1989 ) . Creativity : what are we to measure ? In J . A . Glover , R . R . Ronning , & C . R . Reynolds ( Eds . ) , Handbook of creativity ( pp . 3 e 32 ) , New York , NY . Brown , A . S . , & Murphy , D . R . ( 1989 ) . Cryptomnesia : delineating inadvertent plagiarism . Journal of Experimental Psychology : Learning , Memory , and Cognition , 15 ( 3 ) , 432 e 442 . Button , K . S . , Ioannidis , J . P . A . , Mokrysz , C . , Nosek , B . A . , Flint , J . , Robinson , E . S . J . , et al . ( 2013 ) . Power failure : why small sample size under - mines the reliability of neuroscience . Nature Reviews Neuroscience , 14 ( 5 ) , 365 e 376 . http : / / dx . doi . org / 10 . 1038 / nrn3475 . Chakrabarti , A . ( 2006 ) . De\ufb01ning and supporting design creativity . In Proceedings of the 9th international design conference DESIGN 2006 ( pp . 479 e 486 ) . Chan , J . , Fu , K . , Schunn , C . D . , Cagan , J . , Wood , K . L . , & Kotovsky , K . ( 2011 ) . On the bene\ufb01ts and pitfalls of analogies for innovative design : ideation perfor - mance based on analogical distance , commonness , and modality of examples . Journal of Mechanical Design , 133 , 081004 . Chang , J . , Gerrish , S . , Wang , C . , Boyd - graber , J . L . , & Blei , D . M . ( 2009 ) . Reading tea leaves : how humans interpret topic models . Advances in neural in - formation processing systems 288 e 296 . Chan , J . , & Schunn , C . ( 2014 ) . The impact of analogies on creative concept gen - eration : lessons from an in vivo study in engineering design . Cognitive Science . http : / / dx . doi . org / 10 . 1111 / cogs . 12127 . Chiu , I . , & Shu , H . ( 2012 ) . Investigating e\ufb00ects of oppositely related semantic stimuli on design concept creativity . Journal of Engineering Design , 23 ( 4 ) , 271 e 296 . http : / / dx . doi . org / 10 . 1080 / 09544828 . 2011 . 603298 . Choi , H . S . , & Levine , J . M . ( 2004 ) . Minority in\ufb02uence in work teams : the impact of newcomers . Journal of Experimental Social Psychology , 40 ( 2 ) , 273 e 280 . Dahl , D . W . , & Moreau , P . ( 2002 ) . The in\ufb02uence and value of analogical thinking during new product ideation . Journal of Marketing Research , 39 ( 1 ) , 47 e 60 . Deerwester , S . , Dumais , S . T . , Furnas , G . W . , & Landauer , T . K . ( 1990 ) . Indexing by latent semantic analysis . Journal of the American Society for Information Science , 41 ( 6 ) , 1990 . Dow , S . P . , Heddleston , K . , & Klemmer , S . R . ( 2009 ) . The e\ufb03cacy of prototyping under time constraints . In Proceedings of the 7th ACM conference on creativity and cognition . Dunbar , K . N . ( 1997 ) . How scientists think : on - line creativity and conceptual change in science . In T . B . Ward , S . M . Smith , & J . Vaid ( Eds . ) , Creative Inspiration source distance and design ideation 25 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 thought : An investigation of conceptual structures and processes ( pp . 461 e 493 ) , Washington , D . C . Dym , C . L . ( 1994 ) . Engineering design : A synthesis of views . New York , NY : Cam - bridge University Press . Eckert , C . , & Stacey , M . ( 1998 ) . Fortune favours only the prepared mind : why sources of inspiration are essential for continuing creativity . Creativity and Innovation Management , 7 ( 1 ) , 1 e 12 . Enkel , E . , & Gassmann , O . ( 2010 ) . Creative imitation : exploring the case of cross - industry innovation . R & D Management , 40 ( 3 ) , 256 e 270 . Findlay , A . ( 1965 ) . A hundred years of chemistry ( 3rd ed . ) . London : Duckworth . Fox , J . ( 2002 ) . An R and s - plus companion to applied regression . Sage . Franke , N . , Poetz , M . K . , & Schreier , M . ( 2014 ) . Integrating problem solvers from analogous markets in new product ideation . Management Science , 60 ( 4 ) , 1063 e 1081 . Freeman , A . , & Golden , B . ( 1997 ) . Why didn\u2019t I think of that ? Bizarre origins of ingenious inventions we couldn\u2019t live without . New York : John Wiley . Fu , K . , Chan , J . , Cagan , J . , Kotovsky , K . , Schunn , C . , & Wood , K . ( 2013 ) . The meaning of \u201cnear\u201d and \u201cfar\u201d : the impact of structuring design databases and the e\ufb00ect of distance of analogy on design output . Journal of Mechanical Design , 135 ( 2 ) , 021007 . http : / / dx . doi . org / 10 . 1115 / 1 . 4023158 . Gentner , D . , & Markman , A . B . ( 1997 ) . Structure mapping in analogy and sim - ilarity . American Psychologist , 52 ( 1 ) , 45 e 56 . German , T . P . , & Barrett , H . C . ( 2005 ) . Functional \ufb01xedness in a technologically sparse culture . Psychological Science , 16 ( 1 ) , 1 e 5 . Gero , J . S . ( 2000 ) . Computational models of innovative and creative design pro - cesses . Technological Forecasting and Social Change , 64 ( 2 ) , 183 e 196 . Goldschmidt , G . , & Smolkov , M . ( 2006 ) . Variances in the impact of visual stimuli on design problem solving performance . Design Studies , 27 ( 5 ) , 549 e 569 . Gonc\u00b8alves , M . , Cardoso , C . , & Badke - Schaub , P . ( 2013 ) . Inspiration peak : exploring the semantic distance between design problem and textual inspira - tional stimuli . International Journal of Design Creativity and Innovation 1 e 18 , ( ahead - of - print ) . Gri\ufb03ths , T . L . , & Steyvers , M . ( 2004 ) . Finding scienti\ufb01c topics . Proceedings of the National Academy of Sciences of the United States of America , 101 ( Suppl . 1 ) , 5228 e 5235 . http : / / dx . doi . org / 10 . 1073 / pnas . 0307752101 . Hargadon , A . , & Sutton , R . I . ( 1997 ) . Technology brokering and innovation in a product development \ufb01rm . Administrative Science Quarterly , 42 ( 4 ) , 716 . http : / / dx . doi . org / 10 . 2307 / 2393655 . Helms , M . , Vattam , S . S . , & Goel , A . K . ( 2009 ) . Biologically inspired design : pro - cess and products . Design Studies , 30 ( 5 ) , 606 e 622 . Hender , J . M . , Dean , D . L . , Rodgers , T . L . , & Jay , F . F . ( 2002 ) . An examination of the impact of stimuli type and GSS structure on creativity : brainstorming versus non - brainstorming techniques in a GSS environment . Journal of Man - agement Information Systems , 18 ( 4 ) , 59 e 85 . Holyoak , K . J . , & Thagard , P . ( 1996 ) . Mental leaps : Analogy in creative thought . Cambridge , MA . Impact Stories . ( n . d . ) . Impact stories . [ Web page ] . Retrieved from http : / / www . o - penideo . com / content / impact - stories . Jansson , D . G . , & Smith , S . M . ( 1991 ) . Design \ufb01xation . Design Studies , 12 ( 1 ) , 3 e 11 . Jeppesen , L . B . , & Lakhani , K . R . ( 2010 ) . Marginality and problem - solving e\ufb00ec - tiveness in broadcast search . Organization Science , 21 ( 5 ) , 1016 e 1033 . 26 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 Jessup , E . R . , & Martin , J . H . ( 2001 ) . Taking a new look at the latent semantic analysis approach to information retrieval . Computational information retrieval . Philadelphia : SIAM 121 e 144 . Kalogerakis , K . , Lu , C . , & Herstatt , C . ( 2010 ) . Developing innovations based on analogies : experience from design and engineering consultants . Journal of Product Innovation Management , 27 , 418 e 436 . Kaplan , C . , & Simon , H . A . ( 1990 ) . In search of insight . Cognitive Psychology , 22 ( 3 ) , 374 e 419 . Knoblich , G . , Ohlsson , S . , Haider , H . , & Rhenius , D . ( 1999 ) . Constraint relaxa - tion and chunk decomposition in insight problem solving . Journal of Experi - mental Psychology : Learning , Memory , and Cognition , 25 ( 6 ) , 1534 e 1555 . Liikkanen , L . A . , & Perttula , M . ( 2008 ) . Inspiring design idea generation : insights from a memory - search perspective . Journal of Engineering Design , 21 ( 5 ) , 545 e 560 . Linsey , J . , Tseng , I . , Fu , K . , Cagan , J . , Wood , K . , & Schunn , C . ( 2010 ) . A study of design \ufb01xation , its mitigation and perception in engineering design faculty . Journal of Mechanical Design , 132 ( 4 ) . 041003 - 1 - 12 . Malaga , R . A . ( 2000 ) . The e\ufb00ect of stimulus modes and associative distance in in - dividual creativity support systems . Decision Support Systems , 29 ( 2 ) , 125 e 141 . Marsh , R . L . , Landau , J . D . , & Hicks , J . L . ( 1997 ) . Contributions of inadequate source monitoring to unconscious plagiarism during idea generation . Journal of Experimental Psychology : Learning , Memory , and Cognition , 23 ( 4 ) , 886 e 897 . Marsh , R . L . , Ward , T . B . , & Landau , J . D . ( 1999 ) . The inadvertent use of prior knowledge in a generative cognitive task . Memory & Cognition , 27 ( 1 ) , 94 e 105 . McCallum , A . K . ( 2002 ) . MALLET : a machine learning for language toolkit . [ Computer software ] . Retrieved from . http : / / mallet . cs . umass . edu . Mecca , J . T . , & Mumford , M . D . ( 2013 ) . Imitation and creativity : bene\ufb01cial ef - fects of propulsion strategies and speci\ufb01city . The Journal of Creative Behavior . http : / / dx . doi . org / 10 . 1002 / jocb . 49 . \u20ac Ollinger , M . , Jones , G . , Faber , A . H . , & Knoblich , G . ( 2012 ) . Cognitive mecha - nisms of insight : the role of heuristics and representational change in solving the eight - coin problem . Journal of Experimental Psychology : Learning , Mem - ory , and Cognition . http : / / dx . doi . org / 10 . 1037 / a0029194 . Perkins , D . N . ( 1983 ) . Novel remote analogies seldom contribute to discovery . The Journal of Creative Behavior , 17 ( 4 ) , 223 e 239 . Perkins , D . N . ( 1997 ) . Creativity\u2019s camel : the role of analogy in invention . In T . B . Ward , S . M . Smith , & J . Vaid ( Eds . ) , Creative thought : An investigation of conceptual structures and processes ( pp . 523 e 538 ) . Washington , D . C . : Amer - ican Psychological Association . Pinheiro , J . C . , & Bates , D . M . ( 2000 ) . Linear mixed - e\ufb00ects models : Basic concepts and examples . Springer . Poze , T . ( 1983 ) . Analogical connections d the essence of creativity . The Journal of Creative Behavior , 17 ( 4 ) , 240 e 258 . Raudenbush , S . W . , & Bryk , A . S . ( 2002 ) . Hierarchical linear models : Applications and data analysis methods ( 2nd ed . ) . Thousand Oaks , CA . R Core Team . ( 2013 ) . R : A language and environment for statistical computing . [ Computer software ] . Vienna , Austria : R Foundation for Statistical Computing . Retrieved from . http : / / www . R - project . org / . Rietzschel , E . F . , Nijstad , B . A . , & Stroebe , W . ( 2007 ) . Relative accessibility of domain knowledge and creativity : the e\ufb00ects of knowledge activation on the quantity and originality of generated ideas . Journal of Experimental Social Psychology , 43 ( 6 ) , 933 e 946 . Inspiration source distance and design ideation 27 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 Sawyer , R . K . ( 2012 ) . Explaining creativity : The science of human innovation ( 2nd ed . ) . New York : Oxford University Press . Schunn , C . D . , & Dunbar , K . N . ( 1996 ) . Priming , analogy , and awareness in com - plex reasoning . Memory & Cognition , 24 ( 3 ) , 271 e 284 . Schwartz , A . H . , Eichstaedt , J . C . , Kern , M . L . , Dziurzynski , L . , Ramones , S . M . , Agrawal , M . , et al . ( 2013 ) . Personality , gender , and age in the language of social media : the open - vocabulary approach . PLoS ONE , 8 ( 9 ) , e73791 . Steyvers , M . , & Gri\ufb03ths , T . ( 2007 ) . Probabilistic topic models . In T . Landauer , D . McNamara , S . Dennis , & W . Kintsch ( Eds . ) , Handbook of latent semantic analysis ( pp . 424 e 440 ) . New York , NY : Lawrence Erlbaum . Tseng , I . , Moss , J . , Cagan , J . , & Kotovsky , K . ( 2008 ) . The role of timing and analogical similarity in the stimulation of idea generation in design . Design Studies , 29 ( 3 ) , 203 e 221 . Ullman , D . ( 2002 ) . The mechanical design process . New York , NY ( 3rd ed . ) . Wallach , H . M . , Mimno , D . M . , & McCallum , A . ( 2009 ) . Rethinking LDA : why priors matter . In NIPS , Vol 22 ( pp . 1973 e 1981 ) . Ward , T . B . ( 1994 ) . Structured imagination : the role of category structure in exemplar generation . Cognitive Psychology , 27 ( 1 ) , 1 e 40 . Ward , T . B . ( 1998 ) . Analogical distance and purpose in creative thought : mental leaps versus mental hops . In K . J . Holyoak , D . Gentner , & B . Kokinov ( Eds . ) , Advances in analogy research : Integration of theory and data from the cognitive , computational , and neural sciences ( pp . 221 e 230 ) , So\ufb01a , Bulgaria . Weisberg , R . W . ( 2009 ) . On \u201cout - of - the - box\u201d thinking in creativity . In A . B . Markman , & K . L . Wood ( Eds . ) , Tools for innovation ( pp . 23 e 47 ) , New York , NY . Weisberg , R . W . ( 2011 ) . Frank lloyd wright\u2019s fallingwater : a case study in inside - the - box creativity . Creativity Research Journal , 23 ( 4 ) , 296 e 312 . http : / / dx . doi . org / 10 . 1080 / 10400419 . 2011 . 621814 . Wiley , J . ( 1998 ) . Expertise as mental set : the e\ufb00ects of domain knowledge in cre - ative problem solving . Memory & Cognition , 26 ( 4 ) , 716 e 730 . Wilson , J . O . , Rosen , D . , Nelson , B . A . , & Yen , J . ( 2010 ) . The e\ufb00ects of biological examples in idea generation . Design Studies , 31 ( 2 ) , 169 e 186 . Yu , L . , Kraut , B . , & Kittur , A . ( 2014 ) . Distributed analogical idea generation : innovating with crowds . In Proceedings of the ACM conference on human fac - tors in computing systems ( CHI\u201914 ) . Zeger , S . L . , Liang , K . - Y . , & Albert , P . S . ( 1988 ) . Models for longitudinal data : a generalized estimating equation approach . Biometrics 1049 e 1060 . Endnote Although concept - level variance is not estimated in mixed logistic regressions , we follow Zeger , Liang , and Albert\u2019s ( 1988 ) suggestion of ( 15 / 16 ) p 3 / 3 as a reason - able approximation for residual level - 1 variance ( the concept level in our case ) . 28 Design Studies Vol - - No . - - Month 2014 Please cite this article in press as : Chan , J . , et al . , Do the best design ideas ( really ) come from conceptually distant sources of inspiration ? , Design Studies ( 2014 ) , http : / / dx . doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001",
    "deguchi2023direct": "BIOPHYSICS Direct observation of motor protein stepping in living cells using MINFLUX Takahiro Deguchi 1 , Malina K . Iwanski 2 , Eva - Maria Schentarra 1 , 3 , Christopher Heidebrecht 1 , 3 , Lisa Schmidt 1 , 3 , Jennifer Heck 4 , Tobias Weihs 5 , Sebastian Schnorrenberg 6 , Philipp Hoess 1 , Sheng Liu 1 , 7 , Veronika Chevyreva 1 , 8 , Kyung - Min Noh 4 , Lukas C . Kapitein 2 , Jonas Ries 1 * Dynamic measurements of molecular machines can provide invaluable insights into their mechanism , but these measurements have been challenging in living cells . Here , we developed live - cell tracking of single fluorophores with nanometer spatial and millisecond temporal resolution in two and three dimensions using the recently introduced super - resolution technique MINFLUX . Using this approach , we resolved the precise stepping motion of the motor protein kinesin - 1 as it walked on microtubules in living cells . Nanoscopic tracking of motors walking on the microtubules of fixed cells also enabled us to resolve the architecture of the microtubule cytoskeleton with protofilament resolution . M olecular machines drive many pro - cesses essential for life . For example , members of the myosin , kinesin , and dynein families are adenosine 5 \u2032 - triphosphate ( ATP ) \u2013 driven processive molecular motors that recognize the intrinsic structural polarity of cytoskeletal filaments to drive directed movement important for cellular processes such as intracellular transport and cell division ( 1 \u2013 3 ) . Over the past decades , the mi - crotubule plus - end directed motor kinesin - 1 , hereafter called kinesin , has served as a key model both for the understanding of motor dynamics and for the development of improved single - molecule methods such as optical tweez - ers and single - particle tracking ( 4 \u2013 7 ) . These studies have revealed that kinesinmoves in a hand - over - hand manner , in which each step along the microtubule encompasses a 16 - nm displacement of the N - terminal motor do - main , which leads to an 8 - nm displacement of the C - terminal cargo - binding domain ( Fig . 1A ) . Despite the great success of existing single - molecule techniques in studying purified mo - tors in well - controlled in vitro reconstitution experiments , performing such experiments in thelivingcellhasprovenchallenging . Thelarge labels required for optical tweezers typically bind multiple motors in undefined ways , and single fluorophores do not provide sufficient spatial and temporal resolution to resolve the fast stepping behavior of motors under phys - iologicalATPconcentrations . Asaworkaround , bright nanoparticles taken up through endo - cytosis into transport vesicles have been used as proxies to study motor dynamics in living cells ( 8 , 9 ) . However , in such experiments , the identity and copy numbers of the motors that drive transport are unknown . Therefore , the stepping dynamics of specific motors inside living cells has remained experimen - tally inaccessible . MINFLUX , a super - resolution microscopy technology ( 10 ) , holds great potential to over - come these limitations . By efficiently using the limited photon budget of single fluoro - phores , MINFLUX enables high spatial res - olution for imaging ( 10 \u2013 12 ) and temporal resolutionfor fluorophore tracking ( 13 , 14 ) . In MINFLUX tracking , a donut - shaped excitation beam is scanned around a single fluorophore ( Fig . 1B ) . From the intensities measured at specific positions , the coordinate of the fluo - rophore is calculated and the scanning pattern is recentered on this position before the next iteration . Keeping the fluorophore close to the dark center of the beam results in a high localization precision and minimizes photo - bleaching . In a companion paperinthisissue , MINFLUX was used to dissect the conforma - tional stages of kinesin stepping on in vitro polymerized microtubules ( 15 ) . We now show that this technique can also enable the high - resolution tracking of fast molecular motors in living cells . Optimization of MINFLUX for motor protein tracking To establish MINFLUX tracking of molecular motors in living cells , we first optimized the workflow on single molecules and fluorescent beads in vitro to maximize precision and speed ( fig . S1 , A to C ) and achieved a localization precision of \u2248 2 nm with a submillisecond tem - poral resolution . We then optimized kinesin tracking using motor - PAINT ( 16 ) . In this method , cells are permeabilized and fixed before fluorescently labeled kinesin motors are added that walk along microtubules toward the plus end ( Fig . 1C and movie S1 ) . Unlike high - resolution in vitro assays that use large beads as labels ( 5 , 17 ) , we used small fluorescent tags that reduced the linkage error to \u2248 3 nm [ as predicted by Alphafold2 ( 18 , 19 ) ] , which was comparable to the system resolution . Motor - PAINT is less challenging than live - cell tracking because it allows us to precisely control the concentration of kinesin motors and , importantly , their speed , by ad - justingtheATPconcentration . UsingMINFLUX motor - PAINT , we were able to reconstruct cel - lular microtubules with a precision of \u2248 2 nm ( Fig . 1 , D and E , and table S1 ) . Additionally , the directionality of kinesin reveals the ori - entation of the microtubules . Compared with standard motor - PAINT with a wide - field mi - croscope , the use of MINFLUX improved the localization precision 5 - fold , the tempo - ral resolution 50 - fold , and the number of localizations per track by more than one order of magnitude . In neurons , this allowed us to better resolve individual microtubules inside dendrites compared with our earlier motor - PAINT study ( Fig . 1D ) ( 16 ) . In human osteosarcoma ( U2OS ) cells , we could resolve individual trajectories of the purified motors in the crowded area around the centrosome with near - protofilament resolution ( Fig . 1 , E and F ) . Tracks , which were just 12 nm apart , were easily resolvable , and we regularly ob - served side stepping between different pro - tofilaments ( Fig . 1 , F and G , and movie S2 ) . These side steps often occurred after stalling events , suggesting that motors were circum - venting obstructions such as microtubule - associated proteins ( MAPs ) that became fixed to the microtubules or microtubule defects from the fixation process . A closer inspection of individual tracks showed clusters of localizations that cor - respond to the 8 - nm steps of the labeled C terminus . Indeed , these steps become obvious when plotting the position of the motor along the microtubule over time ( Fig . 1H ) . This al - lowed us to quantify the precise step - size and dwell - timedistributionsundersaturating ( phys - iological ) ATP concentrations ( Fig . 1 , I and J ) . From956stepsin49tracks , wemeasuredastep size of 7 . 8 \u00b1 2 . 7 ( SD ) \u00b1 0 . 09 ( SEM ) nm and an averagedwelltimeof30 . 8ms . Toinvestigatethe stepping behavior in greater detail , we reduced the ATP concentration to slow down the motors ( 20 ) , resulting in a similar step size but a re - duced rate constant for ATP binding ( fig . S2 , A and E , and tables S1 and S2 ) . Under these conditions , we could measure hundreds to thousands of localizations per step ( Fig . 1 , K and L , and movie S3 ) . Averaging over the co - ordinates allowed us to calculate the position of each step with subangstrom precision ( SEM ) . RESEARCH Deguchi et al . , Science 379 , 1010 \u2013 1015 ( 2023 ) 10 March 2023 1 of 6 1 Cell Biology and Biophysics Unit , European Molecular Biology Laboratory , Heidelberg , Germany . 2 Cell Biology , Neurobiology and Biophysics , Department of Biology , Faculty of Science , Utrecht University , Utrecht , Netherlands . 3 Faculty of Biosciences , University of Heidelberg , Heidelberg , Germany . 4 Genome Biology Unit , European Molecular Biology Laboratory , Heidelberg , Germany . 5 Abberior Instruments GmbH , G\u00f6ttingen , Germany . 6 EMBL Imaging Centre , European Molecular Biology Laboratory , Heidelberg , Germany . 7 Department of Physics and Astronomy , University of New Mexico , Albuquerque , NM , USA . 8 The FIRC Institute of Molecular Oncology , Milano , Italy . * Corresponding author . Email : jonas . ries @ embl . de D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on A p r il 15 , 2023 Deguchi et al . , Science 379 , 1010 \u2013 1015 ( 2023 ) 10 March 2023 2 of 6 1 2 3 4 5 6 7 L permeabilization & fixation kinesin motor addition A B 8 nm 16 nm 8 nm 0 1 2 3 4 5 6 7 8 9 0 50 100 150 po s i t i on ( n m ) time ( s ) 100 ms 68 7 9 7 8 117 888 7 88 10 8 6 7 18 nm 6 nm 29 nm 8 nm 25 nm 12 nm 12 nm 26 nm 12 16 step size ( nm ) 00 4 8 100 c ou n t s dwell time ( ms ) c oun t s 20 40 60 80 20 40 60 80 100 0 40 20 0 60 80 100120 0 200 400 600 time ( ms ) 0 50 100 po s i t i on ( n m ) 9 9 7 8 7 8 117 8 7 8 7 low [ ATP ] saturating [ ATP ] x ( nm ) 00 20 40 60 10 20 30 40 50 c oun t s d = 12 nm \u03c3 = 5 nm E H D K F G C datamoving average ( 20 ms ) steps I J L M 1 2 towards cell body towards periphery 2 3 4 4 3 1 1 2 3 Fig . 1 . MINFLUX tracking of kinesin in fixed cells . ( A ) Kinesin walks on microtubules in a hand - over - hand manner . The apparent step size is 8 nm when the label is attached to the C - terminal tail domain and 16 nm when it is attached to the N - terminal motor domain . ( B ) 2D MINFLUX tracking of a single molecule . A donut beam probes seven positions around a fluorophore to determine its location with nanometer precision . The scan pattern is iteratively centered on the fluorophore during tracking . ( C ) Motor - PAINT approach to track kinesin in fixed cells . Cells are first permeabilized to extract cell contents and then gently fixed to preserve microtubules . Purified fluorescently labeled kinesins [ Dm KHC ( 1 - 421 ) - SNAP - tag - 6xHis ] are added and tracked as they walk toward the plus ends of the microtubules ( movie S1 ) . Panels ( D ) to ( M ) show MINFLUX motor - PAINT in fixed cells . ( D ) Confocal images of a neuron and overlaid kinesin trajectories in four neurites . Most of the neurites show kinesin trajectories in both directions , i . e . , toward ( magenta ) and away ( cyan ) from the soma , as expected for dendrites . ( E ) Confocal microscopy images of green fluorescent protein ( GFP ) \u2013 a - tubulin in U2OS cells showing what appears to be a centrosome and overlaid kinesin trajectories with color - coded walking directions . ( F ) Tracks as indicated in ( E ) show side stepping . ( G ) Tracks as close as 12 nm are clearly resolved and display kinesin switching laterally between neighboring microtubules or protofilaments ( movie S2 ) . ( H ) Representative track and the corresponding time versus position plot at saturating ATP concentrations ( > 1 mM , here 6 mM ) showing 8 - nm walking steps . ( I ) Histogram of step size at saturating ATP concentrations from seven experiments , 49 tracks , and 956 steps with a Gaussian fit ( 7 . 8 nm \u00b1 2 . 7 SD \u00b1 0 . 09 nm SEM ; red line ) . ( J ) Dwell time histogram and fit with a convolution of two exponential functions ( average dwell time of 30 . 8 ms ; red line ) . ( K and L ) Representative track at low ATP concentrations ( 10 m M ) and a corresponding time versus position , raw data ( gray ) , and 20 - ms running mean ( black ) clearly showing 8 - nm walking steps ( see movie S3 ) . ( M ) Representative track showing a zigzag trajectory , indicating an asymmetric arrangement of the label within a kinesin molecule ( see fig . S2 , A and E , and movie S4 ; see fig . S3 for additional examples ) . Scale bars : ( L ) and ( M ) , 10 nm ; ( F ) to ( H ) , 100 nm ; ( D ) and ( E ) , 1 m m . RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on A p r il 15 , 2023 Deguchi et al . , Science 379 , 1010 \u2013 1015 ( 2023 ) 10 March 2023 3 of 6 GFP - \u03b1 - tubulin 21 16 1714 13 15 16 15 1713 16 161817 915 161620 14 1019 18 16715 17 16 16 14 15 19 17 20 16 18 1518 17 13 18 16 16 15 1916 1912 17 17181616 13 16 19131619 1812 19 19 16 22 1417 18 19 20 1416 15 1816 16 161913 16 13 21 8 15 14 21 17 1517 19 18 12 15 po s i t i on time 16 n m 100 ms a c d b a a c c d d e f g b b a c d e f g b HaloTag ( JF646 ) - K560 kinesin - 1 16 nm GFP - \u03b1 - tubulin HaloTag ( JF646 ) - FL kinesin - 1 B 00 1 po s i t i on ( n m ) 50 100 150 200 13 18 14 15 16 14 15 16 16 17 18 17 0 . 5 1 . 5 time ( s ) A H G F J K L M I D E po s i t i on time 16 n m 100 ms 17191816 1421 16 1915 13 17 18 151713191714 12 19 1616 1713 181416 161217132113 191321 17 11 15 15 161516 17 15 1419161915 161615 a a b b c c B a c b step size ( nm ) c oun t s c oun t s 50 100 150 0 dwell time ( ms ) 40 20 0 60 80 100 120 C 8 4 0 12 16 20 24 28 32 0 50 100 150 200 Fig . 2 . MINFLUX tracking of kinesin in live cells . Panels ( A ) to ( D ) show tracking of full - length kinesin labeled N - terminally with a HaloTag bound to JF646 in live U2OS cells . ( A ) Confocal images of GFP \u2013 a - tubulin in untreated live U2OS cells and overlaid full - length human kinesin trajectories . ( B ) Kinesin RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on A p r il 15 , 2023 Currently , the accuracy of the measurements is not limited by the detected photons , but rather by the stability of the sample and microscope , as well as the offset of the label from the microtubule - binding site . We observed that 58 % ofthetracksdisplayedzigzag motion , with every other step displaced perpendicular to the track center by , on average , 3 . 6 nm ( Fig . 1M ; fig . S3 , A and C ; and movie S4 ) . This motion is likely caused by an asymmetric positioning of the fluorophore with respect tothe two motor domains imaged in a top view ( fig . S3D ) ( 21 ) , demonstrating that MINFLUX can reveal in - tricate details of the conformational dynamics of individual motor proteins . Deguchi et al . , Science 379 , 1010 \u2013 1015 ( 2023 ) 10 March 2023 4 of 6 track in which the localizations are rendered as a super - resolution image in the region indicated in ( A ) . ( C ) Line plot connecting each localization . ( D ) Time versus position plot of the highlighted portion of the track in ( C ) showing steps of 16 nm . Panels ( E ) to ( J ) show tracking of truncated kinesin ( HaloTag - K560 ) in Taxol - treated live U2OS cells . ( E ) Confocal images of GFP \u2013 a - tubulin and overlaid kinesin tracks . ( F and G ) The tracks indicated in ( E ) rendered as a super - resolution image ( F ) and line plots connecting each localization ( G ) ( see movie S9 ) showing clear walking steps ( localization precision : 2 nm ; temporal resolution : 1 ms ) . ( H ) Time versus position plots of representative kinesin tracks as indicated in ( E ) showing clear 16 nm stepwise movements . ( I ) Step - size histogram ( 161 experiments , 330 tracks , and 2887 steps ) and a Gaussian fit ( 16 . 2 \u00b1 3 . 8 SD \u00b1 0 . 07 SEM nm ) . ( J ) Dwell - time histogram , fit with a convolution of four exponential functions ( average dwell time of 27 . 5 ms ; red line ) . Panels ( K ) to ( M ) show tracking of kinesin ( HaloTag - K560 ) in untreated live primary mouse cortical neurons . ( K ) Confocal images of GFP \u2013 a - tubulin and overlaid kinesin tracks . ( L and M ) Representative tracks corresponding to those indicated in ( K ) as line plots ( L ) and time versus position plots ( M ) showing 16 - nm stepwise movements ( see fig . S2 , C and G , for step size and dwell time histograms ) . Scale bars : ( B ) , ( C ) , ( F ) , ( G ) , and ( L ) , 100 nm ; ( A ) , ( E ) , and ( K ) , 1 m m . xy z x z A C D 90\u00b0 90\u00b0 90\u00b0 90\u00b0 x y z motor - PAINT a c b B E F x - y x - y x - z x - y x - z x - y x - z x - z po s i t i on time 16 n m 100 ms 11 1614 15 12 1617 15 16 15 15 16 13 1710 19 14 13 16 19 17 12 12 16 11 15 13 12 14 20 17 21 18 live - cell a b c Fig . 3 . 3D MINFLUX tracking of kinesin . Panels ( A ) and ( B ) show 3D MINFLUX tracking . ( A and B ) A 3D donut beam ( A ) probes the intensity at seven 3D - distributed positions around a fluorophore ( B ) . Panels ( C ) and ( D ) show 3D tracking with motor - PAINT in fixed U2OS cells . ( C ) 3D rendering of kinesin tracks at crossing microtubules with a volumetric size of 1 . 2 \u00d7 1 . 2 \u00d7 1 m m . ( D ) Selected tracks from ( C ) in top and side views , including ascending and descending trajectories and two trajectories in which motors switch microtubules ( arrows ) . Panels ( E ) and ( F ) show 3D tracking in live cells . ( E ) Representative kinesin tracks in live U2OS cells in top and side views , showing stepwise movements both in the x \u2013 y plane and along the z axis ( see fig . S2 , D and H , for histograms and fig . S9 for confocal overview images ; also see movie S10 ) . ( F ) Position versus time plots of the tracks from ( E ) showing 16 - nm steps . Scale bars : ( E ) , 100 nm ; ( C ) and ( D ) , 200 nm . RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on A p r il 15 , 2023 Application of MINFLUX in living cells We next attempted MINFLUX tracking of kinesin in living cells . To this end , we ex - pressed HaloTag - kinesin ( full - length ) in U2OS cells ( movie S5 ) and labeled at most a single motor domainper dimer by addition of the dye JF646 at very low concentrations . Individual tracks clearly revealed the 16 - nm steps of the motor domains ( Fig . 2 , A to D , and movie S6 ) . On average , we found a step size of 15 . 7 \u00b1 3 . 8 SD \u00b1 0 . 25 SEM nm and an average dwell time of 46 . 8 ms ( fig . S2 , B and F ) . We also observed tracks with frequent switching between microtubules , back - slipping potentially caused by multiple competing mo - tors ( movie S7 ) , and , unlike in motor - PAINT , tracks without clear steps ( fig . S4 ) . The latter potentially stem from kinesins that are attached to dynamic microtubules or cargoes driven by other motors and are thus passively dragged along . However , the number of tracks that we couldacquirewaslow , likelybecauseofkinesins assuminganautoinhibitedformwithonlyalowfractionintheprocessivestate ( 22 , 23 ) . To increasethethroughput , weusedthetruncated kinesin variant K560 , in which cargo binding and autoinhibition are removed , and treated cells with Taxol to increase the number of stabilized microtubules preferred by kinesin ( movie S8 ) ( 24 ) . With these changes , we could readily observe multiple tracks in a single field of view ( Fig . 2 , E to H , and movie S9 ) . This allowed us to measure precise in vivo step - size anddwell - timedistributionsfrom2887stepsin 330 tracks ( Fig . 2 , I and J ) . These measure - ments revealed that although the average step sizeof16 . 2\u00b13 . 8SD\u00b10 . 07SEMnmwassimilar to that of full - length kinesin , the average dwell time of 27 . 5 ms was much shorter , consistent with the higher speeds that we observed with K560 ( table S2 ) . To test whether our approach could be extended to more complex and sen - sitive cell types , we examined kinesin dynam - ics in the axons of live primary mouse cortical neurons ( Fig . 2 , K to M ) , which critically depend onmotor - driventransport . Here , wecouldclearly quantify the 16 - nm stepping dynamics of kinesin withoutTaxoltreatment ( stepsize = 15 . 7\u00b13 . 7SD\u00b1 0 . 21 SEM nm ; average dwell time 29 . 2 ms ; fig . S2 , C and G ) , demonstrating that MINFLUX re - veals conformational dynamics of individual motor proteins in complex cellular systems . Three - dimensional tracking in live cells Because most biological structures extend into three dimensions , only three - dimensional ( 3D ) tracking can capture the true dynamics and avoid projection artifacts that limit the accu - racy in 2D tracking . Unfortunately , standard single - particle tracking provides , at best , poor z resolution ( 25 ) . MINFLUX has been used to image cellular structures in 3D ( 11 ) , but for tracking , it has so far remained limited to 2D . We therefore adapted MINFLUX for 3D track - ing by scanning a3D donut beam in 3D ( Fig . 3 , A and B ) . We achieved a localization precision of 2 . 5 , 3 . 1 , and 3 . 9 nm in the x , y , and z di - rections , respectively ( fig . S1D ) . When used with motor - PAINT , we could resolve many tracks on crossingmicrotubules ( Fig . 3C ) , including jumps between microtubules ( arrow in Fig . 3D ) . We could also establish 3D tracking in live cells with a similar spatial and temporal resolution ( 3 . 9 nm and 3 . 0 ms , respectively ; table S1 ) , al - lowing us to resolve the 16 - nm steps of kinesin in 3D ( Fig . 3 , E and F ; fig . S2 , D and H ; and movie S10 ) . When we investigated these tra - jectories in the cross - sectional views , we found dynamic movements along the z axis , includ - ing side steps and vertical trajectories . 3D tracking allowed us to extract accurate step sizes from a strongly inclined trajectory ( aver - age 15 . 1 nm ) , which , when analyzed in 2D , showed a bias toward smaller step sizes ( average 9 . 0 nm ) ( fig . S5 ) . Thus , MINFLUX tracking opens the possibility to quantify the precise 3D dynamics of molecular machines in living cells . Discussion Here , we established MINFLUX tracking of kinesin with nanometer spatial and submilli - second temporal resolution and demonstrated that we could directly resolve steps of individ - ual motors in live cells . In contrast to recent in vitro MINFLUX measurements ( 15 ) , we did not observe clear 4 - nm substeps of kinesin in motor - PAINT , likely because of insufficient spatial resolution . However , in live - cell exper - iments with our N - terminally labeled motor , we could occasionally observe 8 - nm substeps ( fig . S6 ) . This encouraged us to also image C - terminally labeled kinesin in living cells . Here , the faster dynamics and higher back - ground fluorescence made the measurements challenging , but we could observe tracks with the expected 8 - nm step size ( fig . S7 ) . We found that these C - terminally labeled motors moved with a slightly higher average velocity ( table S2 ) , an effect that could be the result of motor domain labeling , an important consideration for future experiments . Our study paves the way for investigating how the stepping kinetics of motors in cells are modulated by the presence or absence of different MAPs or cargo adaptors ( 26 ) . Such measurements could help to explain the ob - served discrepancy of kinesin stepping be - haviors such as dwell time , stalling , and side stepping in motor - PAINT ( in vitro \u2013 like ) and live cells ( e . g . , the presence of regulatory MAPs such as MAP7 , higher salt concentrations , etc . ) ( tables S1 and S2 ) . One parameter that was consistent across studies was the average step size despite the fact that in live cells the mo - tion of the microtubules themselves could contribute to the measured step size . As ex - pected , the average step size was \u2248 16 nm , both in U2OS cells [ which have some micro - tubule sliding ( 27 ) ] and in neuronal axons [ where microtubules are largely immobile ( 28 ) ] ( table S1 ) . The impact of microtubule movement is thus likely to be minimal at the time scale of kinesin steps . MINFLUX tracking is not limited to kinesin , but can also be used to study the precise motion of any protein in living cells with high spatiotemporal resolution and minimal perturbance because of its compatibility with single - fluorophore labels ( seefig . S8 for track - ing of Myosin - V ) . In the future , developing MINFLUX to simultaneously track two colors will enable monitoring of the relative 3D po - sitions of labeled protein domains with nano - meter spatial and submillisecond temporal resolution . Such measurements of confor - mational changes of molecular machines in their native environment will provide im - portant insights into their function and regulation . REFERENCES AND NOTES 1 . J . Howard , A . J . Hudspeth , R . D . Vale , Nature 342 , 154 \u2013 158 ( 1989 ) . 2 . A . D . Mehta et al . , Nature 400 , 590 \u2013 593 ( 1999 ) . 3 . C . Shingyoji , H . Higuchi , M . Yoshimura , E . Katayama , T . Yanagida , Nature 393 , 711 \u2013 714 ( 1998 ) . 4 . K . Svoboda , C . F . Schmidt , B . J . Schnapp , S . M . Block , Nature 365 , 721 \u2013 727 ( 1993 ) . 5 . S . Sudhakar et al . , Science 371 , eabd9944 ( 2021 ) . 6 . A . Yildiz , M . Tomishige , R . D . Vale , P . R . Selvin , Science 303 , 676 \u2013 678 ( 2004 ) . 7 . W . L . Stepp , Z . \u00d6kten , Life Sci . Alliance 2 , e201900456 ( 2019 ) . 8 . C . S . Peng et al . , https : / / www . biorxiv . org / content / 10 . 1101 / 2022 . 01 . 05 . 475120v1 ( 2022 ) . 9 . X . Nan , P . A . Sims , X . S . Xie , ChemPhysChem 9 , 707 \u2013 712 ( 2008 ) . 10 . F . Balzarotti et al . , Science 355 , 606 \u2013 612 ( 2017 ) . 11 . K . C . Gwosch et al . , Nat . Methods 17 , 217 \u2013 224 ( 2020 ) . 12 . J . K . Pape et al . , Proc . Natl . Acad . Sci . U . S . A . 117 , 20607 \u2013 20614 ( 2020 ) . 13 . Y . Eilers , H . Ta , K . C . Gwosch , F . Balzarotti , S . W . Hell , Proc . Natl . Acad . Sci . U . S . A . 115 , 6117 \u2013 6122 ( 2018 ) . 14 . R . Schmidt et al . , Nat . Commun . 12 , 1478 ( 2021 ) . 15 . J . O . Wolff et al . , Science 379 , 1004 \u2013 1010 ( 2023 ) . 16 . R . P . Tas et al . , Neuron 96 , 1264 \u2013 1271 . e5 ( 2017 ) . 17 . H . Isojima , R . Iino , Y . Niitani , H . Noji , M . Tomishige , Nat . Chem . Biol . 12 , 290 \u2013 297 ( 2016 ) . 18 . J . Jumper et al . , Nature 596 , 583 \u2013 589 ( 2021 ) . 19 . M . Mirdita et al . , Nat . Methods 19 , 679 \u2013 682 ( 2022 ) . 20 . M . J . Schnitzer , S . M . Block , Nature 388 , 386 \u2013 390 ( 1997 ) . 21 . D . Liu , X . Liu , Z . Shang , C . V . Sindelar , eLife 6 , e24490 ( 2017 ) . 22 . D . L . Coy , W . O . Hancock , M . Wagenbach , J . Howard , Nat . Cell Biol . 1 , 288 \u2013 292 ( 1999 ) . 23 . D . S . Friedman , R . D . Vale , Nat . Cell Biol . 1 , 293 \u2013 297 ( 1999 ) . 24 . D . Cai , D . P . McEwen , J . R . Martens , E . Meyhofer , K . J . Verhey , PLOS Biol . 7 , e1000216 ( 2009 ) . 25 . J . Andrecka et al . , eLife 4 , e05413 ( 2015 ) . 26 . P . J . Hooikaas et al . , J . Cell Biol . 218 , 1298 \u2013 1318 ( 2019 ) . 27 . K . I . Jansen , M . Burute , L . C . Kapitein , https : / / www . biorxiv . org / content / 10 . 1101 / 2021 . 06 . 23 . 449589v1 ( 2021 ) [ if now published , omit the URL and provide only a standard reference ] . 28 . M . Burute , K . I . Jansen , M . Mihajlovic , T . Vermonden , L . C . Kapitein , Sci . Adv . 8 , eabo2343 ( 2022 ) . ACKNOWLEDGMENTS We thank K . Watanabe and K . I . Jansen for generating the full - length wild - type Myosin - V and K560 - HaloTag constructs , respectively ; Abberior Instruments , specifically R . Schmidt for MINFLUX technical support ; F . Fichtner , A . Pacheco , U . Matti , and L . Perez for help with sample preparations ; and the EMBL Deguchi et al . , Science 379 , 1010 \u2013 1015 ( 2023 ) 10 March 2023 5 of 6 RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on A p r il 15 , 2023 Imaging Centre for access to the MINFLUX instrument . JF646 and JFX646 HaloTag ligands were a kind gift of L . Lavis ( HHMI Janelia Research Campus ) . Funding : This work was supported by H2020 Marie Sk \u0142 odowska - Curie Actions ( RobMin grant no . 101031734 to T . D . , EMBL ARISE fellowship no . 945405 to S . L . , and EIPOD4 program grant no . 847543 to J . H . ) ; the European Research Council ( grant no . ERC CoG - 724489 to J . R . and grant no . CoG - 819219 to L . C . K . ) ; and the European Molecular Biology Laboratory ( T . D . , E . - M . S . , C . H . , L . S . , J . H . , S . S . , P . H . , S . L , V . C . , K . - M . N . , and J . R . ) . Authorcontributions : J . R . conceivedtheproject . L . S . , P . H . , and V . C . designed and generated constructs . M . K . I . purified proteins for motor - PAINT . J . H . cultured neurons . T . D . , M . K . I . , E . - M . S . , C . H . , L . S . , and J . H . prepared samples . T . D . , T . W . , and S . S . established MINFLUX trackingprotocols . T . D . andS . L . acquired MINFLUX tracking data . E . - M . S . and C . H . acquired fluorescence images . J . R . wrote the analysis software . T . D . , M . K . I . , E . - M . S . , C . H . , L . C . K . , and J . R . analyzedthe data . K . - M . N . , L . C . K . , and J . R . supervised the research . T . D . , M . K . I , E . - M . S . , C . H . , J . H . , P . H . , L . C . K . , and J . R . wrote the manuscript with input from all authors . Competing interests : The authors declare no competing interests . Data and materials availability : All MINFLUX tracking data are available at BioStudies ( https : / / www . ebi . ac . uk / biostudies / ) under accession no . S - BIAD608 . License information : Copyright \u00a9 2023 the authors , some rights reserved ; exclusive licensee American Association for the Advancement of Science . No claim to original US government works . https : / / www . science . org / about / science - licenses - journal - article - reuse SUPPLEMENTARY MATERIALS science . org / doi / 10 . 1126 / science . ade2676 Materials and Methods Figs . S1 to S9 Tables S1 and S2 References ( 29 \u2013 37 ) MDAR Reproducibility Checklist Movies S1 to S10 View / request a protocol for this paper from Bio - protocol . Submitted 8 August 2022 ; accepted 23 January 2023 10 . 1126 / science . ade2676 Deguchi et al . , Science 379 , 1010 \u2013 1015 ( 2023 ) 10 March 2023 6 of 6 RESEARCH | RESEARCH ARTICLE D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on A p r il 15 , 2023 Use of this article is subject to the Terms of service Science ( ISSN ) is published by the American Association for the Advancement of Science . 1200 New York Avenue NW , Washington , DC 20005 . The title Science is a registered trademark of AAAS . Copyright \u00a9 2023 The Authors , some rights reserved ; exclusive licensee American Association for the Advancement of Science . No claim to original U . S . Government Works Direct observation of motor protein stepping in living cells using MINFLUX Takahiro Deguchi , Malina K . Iwanski , Eva - Maria Schentarra , Christopher Heidebrecht , Lisa Schmidt , Jennifer Heck , Tobias Weihs , Sebastian Schnorrenberg , Philipp Hoess , Sheng Liu , Veronika Chevyreva , Kyung - Min Noh , Lukas C . Kapitein , and Jonas Ries Science , 379 ( 6636 ) , . DOI : 10 . 1126 / science . ade2676 Zeroing in on motor proteins The super - resolution microscopy technique MINFLUX enables localization of fluorophores using a minimal number of photons . Two studies now expand on the development and implementation of MINFLUX to track motor protein dynamics in vitro and in cells ( see the Perspective by Fei and Zhou ) . Wolff et al . refined the precision of MINFLUX such that single - fluorophore tracking with nanometer precision was possible with only tens of photons . They tracked the movement of kinesin - 1 on microtubules and were able to see individual 4 - nanometer substeps and rotation of the protein during stepping in their analysis . Deguchi et al . applied MINFLUX with a labeling and tracking approach called motor - PAINT to monitor stepping of motor proteins on microtubules in living and fixed cells in both two and three dimensions . \u2014MAF View the article online https : / / www . science . org / doi / 10 . 1126 / science . ade2676 Permissions https : / / www . science . org / help / reprints - and - permissions D o w n l o a d e d fr o m h tt p s : / / www . s c i e n ce . o r g a t U n i v e r s it y o f W a s h i ng t on on A p r il 15 , 2023",
    "chanSemanticallyFarInspirations2017": "Semantically Far Inspirations Considered Harmful ? Accounting for Cognitive States in Collaborative Ideation Joel Chan 1 , Pao Siangliulue 2 , Denisa Qori McDonald 3 , Ruixue Liu 3 , Reza Moradinezhad 3 , Safa Aman 3 , Erin T . Solovey 3 , Krzysztof Z . Gajos 2 , Steven P . Dow 4 , 1 Carnegie Mellon University , 2 Harvard University , 3 Drexel University , 4 University of California San Diego , joelchuc @ cs . cmu . edu , { paopow , kgajos } @ seas . harvard . edu , { denisa . qori , rl498 , reza . mora , sra68 , erin . solovey } @ drexel . edu , spdow @ ucsd . edu ABSTRACT Collaborative ideation systems can help people generate more creative ideas by exposing them to ideas different from their own . However , there are competing theoretical views on whether and when such exposure is helpful . Associationist theory suggests that exposing ideators to ideas that are semantically far from their own maximizes novel combinations of ideas . In contrast , SIAM theory cautions that systems should offer far ideas only when ideators reach an impasse ( a cognitive state in which they have exhausted ideas within a particular category ) , and offer near ideas during productive ideation ( a cognitive state in which they are actively exploring ideas within a category ) , which maximizes exploration within categories . Our research compares these theoretical recommendations . In an online experiment , 245 participants generated ideas for a themed wedding ; we detected and validated participants\u2019 cognitive states using a combination of behavioral and neuroimaging data . Receiving far ideas during productive ideation resulted in slower ideation and less within - category exploration , without significant benefits for novelty , compared to receiving no inspirations . Participants were also more likely to hit an impasse when receiving far ideas during productive ideation . These findings suggest that far inspirational ideas can harm creativity if received during productive ideation . Author Keywords Creativity ; creativity support tools ; brainstorming ; examples ; collaborative ideation ACM Classification Keywords Human - centered computing ~ Collaborative and social computing INTRODUCTION Large - scale collaborative ideation platforms , like Climate CoLab and OpenIDEO , draw hundreds to thousands of contributors to collaboratively generate and develop solutions for creative problems . The promise of these platforms is that more breakthrough ideas can be developed by facilitating collaboration and remixing ideas at much higher levels of scale and diversity than before . However , the scale and diversity of crowd ideation also presents unique challenges for collaborative inspiration . In small groups , ideation can be improved by simply exposing all contributors to all ideas [ 14 ] ; however , at crowd scale , it is not uncommon to have hundreds or thousands of contributions . Ideators do not have sufficient time or cognitive resources to sift through that many ideas to select and build on ideas that are most helpful for their thinking [ 26 , 27 ] . Instead , ideators in these settings often resort to superficial processing of a few ideas [ 26 , 27 ] . Consequently , systems that can find and deliver potentially inspiring content to ideators are an important area of technical research and development for large - scale collaborative ideation platforms . From a technical standpoint , significant progress has been made on the problem of how to structure large collections of ideas to enable exploration and navigation of the solution space [ 1 , 18 , 23 , 33 , 43 , 44 ] . Other studies have explored how dedicated community managers [ 6 ] or facilitators [ 8 ] might deliver appropriate inspiration to ideators . In this paper , we consider a central human factors question facing designers of these inspiration delivery systems : how should inspiration delivery systems take into account the semantic distance of other people\u2019s ideas from the target user\u2019s ideas ? Are the benefits of collaborative inspiration maximized by promoting cross - pollination of ideas ( e . g . , exposing ideators to ideas that are very different from their own ) , or by promoting deeper exploration of shared solution approaches ( e . g . , iterating on each other\u2019s ideas Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page . Copyrights for components of this work owned by others than the author ( s ) must be honored . Abstracting with credit is permitted . To copy otherwise , or republish , to post on servers or to redistribute to lists , requires prior specific permission and / or a fee . Request permissions from Permissions @ acm . org . C & C ' 17 , June 27 - 30 , 2017 , Singapore , Singapore Copyright is held by the owner / author ( s ) . Publication rights licensed to ACM . ACM 978 - 1 - 4503 - 4403 - 6 / 17 / 06\u2026 $ 15 . 00 DOI : http : / / dx . doi . org / 10 . 1145 / 3059454 . 3059455 Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 93 within a particular semantic region of the solution space ) ? Understanding the relative impact these approaches on the quantity , novelty , and diversity of generated ideas can inform the design of these platforms . To preview our results , we find that semantically far inspirations can negatively impact creative processes ( speed , amount of within - category exploration ) and products ( novelty of ideas ) if they are received during particular cognitive states . Theoretical Foundations Before we describe our experiment and results in detail , it is useful to consider the theoretical foundations of our investigation . Two prominent theories of creativity offer different answers to the question of semantic distance of inspirational stimuli from target user\u2019s own ideas . We selected these theories for relevance to the specific question of inspiration delivery . Our goal is not to arbitrate between competing overall theories of creativity , but rather to advance theoretical foundations for designing effective inspiration delivery systems . Associationist Theory of Creativity On the one hand , the associationist theory of creativity [ 21 , 31 , 34 , 42 ] argues that creativity arises from combining ideas that are very different from one another . From a cognitive standpoint , the associationist view notes a few key mechanisms by which this can be accomplished . First , creators can use serendipity to activate disparate portions of semantic memory at the same time . Mednick [ 34 ] noted an illustrative example of a physicist who \u201creduced serendipity to a method by placing in a fishbowl large numbers of slips of paper , each inscribed with a physical fact . He regularly devotes some time to randomly drawing pairs of these facts from the fishbowl , looking for new and useful combinations\u201d ( p . 221 - 222 ; emphasis ours ) . A related mechanism is mediation , which connects disparate concepts by finding deep structural similarities between them ( e . g . , by analogy [ 19 ] ) . It is worth mentioning that the original associationist theory also cites a mechanism of similarity , which connects disparate concepts by \u201cspreading activation\u201d through similar / related concepts in between . This mechanism is not shared by other associationist theories ( e . g . , conceptual combination , analogy ) , so we focus on the first two mechanisms\u2014serendipity and mediation\u2014that emphasize directly connecting semantically disparate concepts . These theoretical assertions imply that , to improve creativity , inspiration should maximize the probability of making interesting remote associations , either by directly providing candidate combinations , or by activating distant portions of semantic memory through semantic priming [ 12 , 35 ] . This can be accomplished by delivering ideas that are semantically distant from the user\u2019s own ideas . According to the associationist theory , presenting ideators with such semantically far stimuli should increase the diversity and novelty of ideas they subsequently generate . Search for Ideas in Associative Memory ( SIAM ) Model On the other hand , the SIAM ( Search for Ideas in Associative Memory ) model of creative idea generation [ 37 ] argues that the answer depends on the user\u2019s cognitive state . The SIAM model posits that ideation proceeds by alternating between two kinds of cognitive states : productive ideation and impasses . During productive ideation , an ideator fluently accesses idea components from memory , and actively develops new ideas from those idea components . In this state , temporally adjacent ideas are often relatively close in semantic space , with some being close variations or elaborations of prior ideas , and some assembled from neighboring components in associative memory . In this state , ideators can develop more novel and useful ideas through deeper iteration and elaboration [ 11 , 15 , 36 , 41 ] . After exhausting resources within a semantic region , people enter an impasse state , and commence search for new semantic regions of memory from which to sample idea components . Idea generation during this state is slower and more effortful ( as measured by time intervals between ideas , and subjective reports and / or neurophysiological measures of cognitive effort ) , and temporally adjacent ideas tend to be semantically distant from each other . Overall , SIAM posits that , in addition to exploring new semantic regions ( when appropriate ) , creativity can be maximized with fluent exploration within categories , which enables people to move past common , shallow ideas to more interesting , less obvious ideas . The theoretical assertions of the SIAM model imply that semantic distance should be accounted for in different ways for these states . During productive ideation , inspiration should enrich the local semantic region directly , or activate other ( potentially valuable ) idea components in the near neighborhood through semantic priming [ 12 , 35 ] . This can be accomplished by delivering conceptually near stimuli , which can help people more deeply explore the local region ( beneficial for reaching more creative ideas [ 11 , 15 , 36 , 41 ] ) . In this state , exposure to far stimuli might even be harmful : attending to those stimuli could shift attention away from the current memory region ( again , through processes like semantic priming ) , perhaps prematurely terminating productive chains of thought . Further , understanding and adapting far stimuli may require significant cognitive effort , and ideators may not be motivated to expend this cognitive effort if they are productively ideating within a category . During impasses , inspiration should suggest new semantic regions to explore , rather than refocusing ideators on the depleted semantic region . This can be accomplished by delivering semantically far stimuli , which could help accelerate the process of finding new productive regions to explore by activating more diffuse portions of memory . From Theories to Inspiration Delivery System Designs The associationist and SIAM theories therefore predict very different best and worst inspiration delivery approaches ( see Table 1 ) . Associationist theory suggests an ALWAYS - FAR inspiration approach would be best for ideation , where Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 94 the system strives to deliver semantically distant ideas to the user , without accounting for cognitive states . According to the associationist theory , the least helpful approach would be the ALWAYS - NEAR inspiration strategy , which would be predicted to harm ideation by always constraining the user to a semantically adjacent region . In contrast , SIAM suggests a MATCH - STATE inspiration approach would be best for ideation , where the system delivers near stimuli during productive ideation , and far stimuli during impasses . SIAM further predicts that a MISMATCH - STATE inspiration approach will be least helpful , since it presents the user with the opposite of their theoretically predicted inspiration needs during each state ( i . e . , potentially distracting far stimuli during productive ideation , and constraining near stimuli during impasses ) . Interestingly , SIAM also offers a competing prediction for the associationist\u2019s best and worst approaches : the ALWAYS - FAR and ALWAYS - NEAR approaches should yield similar levels of novelty . This is because SIAM predicts that , in the ALWAYS - FAR approach , the increase in novelty from providing pointers to new areas of exploration should be offset by losses in novelty due to hindered within - category exploration . In the ALWAYS - NEAR approach , SIAM predicts that decreased novelty from lack of pointers to new areas of exploration should be offset by increased novelty from fluent within - category exploration [ 5 , 36 ] . To advance research and development of inspiration delivery systems for collaborative ideation platforms , empirical work is needed to tease apart whether the associationist or SIAM theories ( or neither ) are more useful theoretical guides for how to appropriately account for semantic distance of potential inspirational stimuli . Overview and Contributions of The Present Study In this paper , we report the results of an empirical test of these theories by comparing each of their predicted best and worst inspiration delivery approaches against a NO - STIMULI baseline . In an online ideation experiment , 245 participants generated ideas for themed weddings in one of the four inspiration conditions ( ALWAYS - FAR , ALWAYS - NEAR , MATCH - STATE , and MISMATCH - STATE ) , or in the NO - STIMULI baseline condition . We detected changes between productive ideation and impasse states through a simple self - report mechanism . This approach was validated by behavioral ( ideation was significantly slower right before an impasse ) and neuroimaging data ( which showed neuroimaging markers of significantly elevated cognitive effort right before an impasse ) . Consistent with SIAM predictions , MISMATCH - STATE and ALWAYS - FAR participants generated ideas at a slower rate than NO - STIMULI participants , and ALWAYS - FAR participants iterated less within categories ( as measured by mean similarity between subsequent ideas ) compared to NO - STIMULI participants . Further , participants who received near stimuli during productive ideation ( MATCH - STATE , ALWAYS - NEAR ) were less likely to face impasses than participants who received far stimuli during productive ideation ( ALWAYS - FAR , MISMATCH - STATE ) . Contrary to associationist predictions , ALWAYS - FAR ideas were not significantly more diverse or novel than NO - STIMULI ideas ; instead , ALWAYS - FAR ideas were marginally statistically significantly less novel than NO - STIMULI ideas . This paper contributes new insights for how to best promote creative inspiration on collaborative ideation platforms . Specifically , our findings show that far inspirational ideas\u2014though considered to be generally useful for creative inspiration\u2014can harm creativity if received during productive ideation . Our findings also imply that the SIAM model\u2019s state - contingent view of inspiration needs is more useful as a theoretical starting point than the associationist theory of creativity for guiding the design of collaborative inspiration systems . Predicted best approach Predicted worst approach Associationist theory ALWAYS - FAR : deliver far stimuli , regardless of cognitive state ; increases novelty and diversity of ideas by promoting remote associations ( SIAM predicts neutral effect on novelty : gains in novelty from pointers to new ideas offset by hindering exploration within categories ) ALWAYS - NEAR : deliver near stimuli , regardless of cognitive state ; decreases novelty and diversity of ideas by suppressing remote associations ( SIAM predicts neutral effect on novelty : loss of novelty from lack of pointers to new ideas offset by gains from fluent within - category exploration ) SIAM model MATCH - STATE : deliver near stimuli during productive ideation , and far stimuli during impasses ; increases novelty of ideas by promoting fluent exploration within categories , and providing pointers to new areas of exploration at the appropriate time MISMATCH - STATE : deliver far stimuli during productive ideation , and near stimuli during impasses ; decreases novelty of ideas by hindering fluent exploration within categories , and suppressing pointers to new areas of exploration at the appropriate time Table 1 . Best and worst approaches for choosing semantic distance of inspirational stimuli , as predicted by the associationist and SIAM theories of creativity . Associationist theory predicts that an ALWAYS - FAR approach is best , and an ALWAYS - NEAR approach is worst ; SIAM predicts that a MATCH - STATE approach is best , and a MISMATCH - STATE approach is worst . SIAM also makes competing predictions for the associationist theory\u2019s predicted best and worst approaches . Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 95 EXPERIMENT Participants We recruited 245 participants ( mean age = 33 . 5 years , SD = 10 . 4 , 51 % female ) from Amazon Mechanical Turk ( MTurk ) . All participants were located in the U . S . and had 95 % approval on at least 100 MTurk tasks . Participants were paid $ 1 . 25 for their time ( approximately $ 6 / hr wage , given average completion times of 12 \u2013 13 minutes ) . Study Design We used a between - subjects design . Participants were randomly assigned to one of the 5 conditions : 1 ) the NO - STIMULI baseline ( N = 54 ) , 2 ) associationist theory\u2019s predicted best ALWAYS - FAR condition ( N = 47 ) , 3 ) associationist theory\u2019s predicted worst ALWAYS - NEAR condition ( N = 48 ) . 4 ) SIAM theory\u2019s predicted best MATCH - STATE condition ( N = 51 ) , and 5 ) SIAM theory\u2019s predicted worst MISMATCH - STATE condition ( N = 45 ) . Brainstorming Task Participants generated ideas for a themed wedding , where each idea consisted of 1 ) a theme , 2 ) a main prop to be used for guest activities , and 3 ) a freeform description of how the prop would be incorporated into the wedding . We chose this task structure to maximize our ability to accurately tailor conceptual similarity based on participants\u2019 current thinking ( and therefore experimentally isolate our intervention ) in real - time . Achieving real - time semantic tailoring of potential stimuli to unstructured participant ideas of varying length and specificity is challenging to accomplish with a high degree of accuracy . To address this concern , our brainstorming task is semi - structured : participants separately specify theme ( e . g . , \u201cmedieval\u201d ) and prop ( e . g . , \u201csilver spoons\u201d ) components of their themed wedding idea . This allows us to perform fast and accurate tailoring based on those single or compound words where computational similarity measurements tend to do better . For example , models like Pennington et al\u2019s [ 38 ] Global Vectors for Word Representation ( GloVe ) model \u2014 which uses an unsupervised learning algorithm to learn vector representations for words from global word - to - word co - occurrence statistics within a corpus \u2014 are able to achieve between 60 % and 84 % accuracy on word analogy tasks . This brainstorming task also achieves a degree of ecological validity since developing ideas for themed weddings is a common real - world creative task . Sampling Inspirations based on Conceptual Distance We used pre - trained GloVe vectors ( trained on approximately 6 billion tokens ( Wikipedia 2014 and Gigaword 5 corpora , with 300 dimensions ) provided by Pennington et al [ 38 ] to perform similarity matching . While other vector - space models like Latent Semantic Analysis ( LSA ) have a longer history in cognitive science for measuring semantic distance [ 32 ] , we opted to use GloVe , a recent state - of - the - art model that typically agrees well with classic models like LSA [ 17 ] , while being capable of modeling more nuanced semantics , such as simple four - term analogies ( e . g . , man : woman = king : queen ) [ 38 ] . Our database of potential inspirational stimuli consisted of 455 themes and 655 props collected from pilot runs of this study ( with 207 MTurk workers ; none of these workers also participated in the main study ) . We showed inspirations as sets of 3 themes and 3 props , assembled in real - time and tailored to participants\u2019 last generated idea . Near stimuli are sampled to be near in semantic space , but no nearer than cosine similarity of 0 . 5 in the GloVe vector space . We selected this threshold to avoid duplicates and very close matches that are likely already activated , and also to activate the periphery of the current location in the semantic network in order to enrich the semantic region with additional potentially active idea components . Far stimuli were sampled to be as far from current thinking as possible . To ensure our sampling approach focused on conceptual distance , we also controlled for diversity of inspiration sets ( the relative distance between inspirations in a set ) because the diversity of inspirational examples has been shown to impact creative performance [ 4 , 43 ] . Conceptual distance and diversity tend to be positively correlated , but it is hard to generate diverse sets for near stimuli . All things being equal , ideas that are all close to a seed idea will be relatively close to each other , compared to ideas sampled from distant semantic regions . Therefore , we restricted the diversity of sets to be relatively low . We use a simple sampling algorithm to ensure low diversity of sets : For each query , we first sampled a seed inspiration ( whether near or far ) . Then , we found two nearest neighbors of that seed inspiration ( where the cosine similarity of those neighbors to the seed and each other were less than 0 . 5 ) . This completed a set of 3 inspirations for the query . The following are examples of near and far inspirations sampled by our approach for two different themes : For \u201cfootball\u201d , Near : [ season , fun and games , fourth of July ] , and Far : [ toga , hula , prom ] . For \u201csteam punk\u201d , Near : [ album , light of love , rock ] , and Far : [ minions , knight and damsel , ghostbusters ] . Validating Stimuli Sampling Approach To validate our approach , we randomly sampled 100 themes generated by participants in the study , along with the near and far sets of inspirations actually retrieved for those themes during the experiment . We then generated a new set of inspirations that was the opposite distance ( either near or far ) from each theme . A trained research assistant ( blind to which sets were deemed near or far by the algorithm ) then went through each theme and marked which of the two sets ( left or right ) was \u201cnearest\u201d to the theme . A second judge ( one of the authors , also blind to the algorithm\u2019s predictions ) completed judgments for a random subset of 40 of the items , and agreement between the human judges was very high , Cohen\u2019s \ud835\udf05 = 0 . 95 . The Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 96 research assistant judged the remaining items . The model\u2019s selection of near stimuli corresponded well to the judge\u2019s selection , Cohen\u2019s \ud835\udf05 = 0 . 84 , validating our semantic tailoring manipulation . In terms of \u201cabsolute\u201d distance , the near stimuli in our dataset were , on average , 4 nodes away from the participant\u2019s last idea in Wordnet\u2019s association network ( e . g . , WOLF - - > canine - - > carnivore - - > feline - - > CAT ) , compared to 9 nodes away for far stimuli . To validate our diversity control mechanism , we sampled 855 inspiration sets actually provided to participants during our experiment , and measured their pairwise distances using GloVe . In this sample , far sets were not more diverse than near sets ; in fact , there was a tendency for near sets to be more diverse ( M = . 39 , SE = . 00 ) than far sets ( M = . 27 , SE = . 00 ; t = 18 . 67 , p < . 01 ) . This suggests that our diversity control mechanism successfully removed the usual coupling between distance and diversity of inspiration sets . Inferring Ideators\u2019 Cognitive States Our experiment requires that we accurately infer participants\u2019 cognitive state at each moment . Automatic detection is likely to be noisy ; people can be idle for different reasons , and may be productively thinking while not typing . In contrast , prior work shows that people can notice when they are stuck [ 45 ] . However , ideally we do not want to burden participants with constantly monitoring their own cognitive state when they are productive . Therefore , we designed a partially user - driven approach to infer cognitive states . The default state is productive ideation . The participant triggers a state change to impasse by requesting a set of inspirations , with the intuition that participants in this state would naturally seek out new stimulation . The system then infers a state change back to productive ideation once the participant submits a new idea . Validating User - Driven Inferring of Cognitive States Since this user - driven approach is novel , we sought to validate that it succeeds at differentiating between cognitive states . We conducted a small pilot study in which participants brainstormed while wearing functional near - infrared spectroscopy ( fNIRS ) brain sensors . fNIRS is a neuroimaging method that detects changes in the concentration of oxygenated and deoxygenated blood in the brain , relative to a reference point ( e . g . , during resting state , or a fixed time interval prior to an \u201cevent\u201d ) . These changes in blood oxygen concentration can be used to infer changes in brain activity in particular regions of the brain [ 49 ] , similar to the blood oxygen level dependent ( BOLD ) measure used in fMRI . Neuroimaging Validation . The fNIRS device , manufactured by ISS , Inc . contained six measurement channels with 3 - cm source - detector distances . Participants experienced the same set of procedures as our online participants , except they generated ideas for 20 minutes ( to maximize the amount of data points per participant ) , and participated in a post - task semi - structured interview . We preprocessed the fNIRS data for analysis using Homer2 [ 24 ] , a MATLAB - based application for processing fNIRS data . Preprocessing involved converting the raw light intensity data to oxygenated ( HbO ) and deoxygenated hemoglobin ( HbR ) changes . A high - pass filter was also applied at 0 . 5 Hz . Our sample consists of 6 participants , with a total of 23 instances of inferred impasses . We focused our analysis on comparing brain activity 10 seconds immediately before an inspiration request ( which we assume would be an impasse state ) and 10 seconds immediately after the first idea submission after an inspiration request ( which we assume would be a productive ideation state ) . Our hypothesis is that we should see brain activity that indicates higher levels of cognitive effort during the inferred impasse state , compared to the productive ideation state . The expected hemodynamic response would be a negative change in HbR during increase cognitive load [ 25 , 46 ] . Thus , we operationalize cognitive effort as the maximum change in HbR relative to the 2s just prior to the event . We found a significant difference in the maximum HbR signal between the two states , with the pre - stuck period being lower than the post - stuck period across all of the channels we measured ( see Figure 1 ; all p < . 01 ) . This finding suggests participants were exerting high levels of cognitive effort during inferred impasses ( i . e . , right before requesting inspirations ) . Behavioral Validation . These neuroimaging results were further corroborated by comments participants made during the post - task semi - structured interview . For example , one participant said she clicked to request more inspirations \u201cmostly just like after I had like exhausted the ideas in my mind and I was like OK I don\u2019t know what could possibly Figure 1 . Participants in our pilot study showed higher levels of cognitive effort during system - inferred impasses ( green bars ) compared to system - inferred productive ideation ( red bars ) , as indicated by lower levels of \u0394 HbR ( in micromolars , \u03bc M ) in regions of the prefrontal cortex ( measured using functional near - infrared spectroscopy ) . Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 97 be next\u201d . Another participant said he requested inspirations when he \u201cwas just running out of ideas\u201d . Further , participants in our online experiment took significantly more time between subsequent idea submissions just before an inferred impasse ( M = 79 . 6 seconds , SE = 7 . 3 ) , compared to idea transitions not temporally adjacent to an inferred impasse event ( M = 55 . 8 , SE = 5 . 0 ) , t ( 19 . 4 ) = 3 . 3 , p < . 01 ( see Figure 2 ) . Note that these do not include the interval between the last idea submission before an inspiration request and the request , or the interval between the request and the first idea submission after the request . The intervals are therefore indicative of the speed with which ideas are generated , and not the time it takes to perform extra tasks , such as ( deciding to ) request more inspirations . This difference in inter - idea interval is consistent with SIAM\u2019s predictions of faster ideation during productive ideation and slower ideation during impasses . Altogether , these results suggest that our user - driven approach successfully detects cognitive state changes ( i . e . , inspiration requests signal a transition to an impasse state ) . System and Interface Figure 3 shows the ideation interface used in our experiment . Participants enter ideas in a semi - structured format ( separate fields for themes , props , and descriptions ) . In the left pane , the system automatically retrieves a new set of themes and props after each idea entry . The participant is assumed to be in a productive ideation state unless they click the \u201cGive me other inspirations\u201d button . When this button is clicked , the system infers the participant is in an impasse state , and retrieves a new set of inspirations accordingly . Participants can refresh the inspiration feed during the impasse state as many times as they wish . Each button click during this state retrieves a new set of inspirations . Once a participant submits a new idea after requesting inspirations , the system infers the participant has returned to a productive ideation state . The system was built in Meteor . js ( a Node . js - based web application framework ) . The system communicates with a similarity engine via a RESTful API to select inspirations based on semantic relatedness to participants\u2019 current idea focus . Inspiration retrievals took about 1 \u2013 2 seconds . Procedure After providing informed consent , participants were randomly assigned to one of the 5 conditions . Next , they entered a warm - up task screen where they generated alternative uses for a brick for one minute . After this , they completed a short tutorial that highlighted the key features of the interface . Finally , they generated ideas for the main problem for 8 minutes . The system automatically took them to a final survey page after 8 minutes . MEASURES Within - Category Fluency We operationalize two measures of within - category fluency that capture related but distinct aspects of the theoretical construct of within - category fluency : inter - idea interval ( probability of an impasse ) , and transition similarity . Inter - Idea Interval SIAM posits that ideation within a category is more rapid than generating ideas in - between categories ; thus , participants who have higher within - category fluency should , on average , have shorter intervals between idea submissions . Thus , we operationalize inter - idea interval as the median number of seconds between subsequent idea submissions , as logged by our system . We report median inter - idea interval because this measure is insensitive to long inter - idea intervals during impasses and instead , reflects how rapidly a participant was generating ideas while they were in a productive ideation state . We also statistically control in our analysis for the number of ideas generated . This allows us to more cleanly capture the degree of within - category fluency . Transition Similarity Ideas within a category tend to be more semantically similar to each other than ideas between categories : thus , participants who have higher within - category fluency should , on average , have higher similarity between successive ideas . We operationalized transition similarity as Figure 3 . Participants enter ideas in a semi - structured interface . Inspirations in the left pane automatically update during productive ideation ; when participants request new inspirations , this signals an impasse state to the system , and new inspirations are retrieved . The control condition interface is identical except for the absence of an inspiration feed . Figure 2 . Participants generated ideas more slowly just before an impasse compared to other points in their ideation session . Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 98 the median GloVe cosine between themes and props of subsequent ideas . As with inter - idea interval , reporting the median allows us more cleanly capture the degree of within - category fluency . Overall Fluency : Number of Ideas Generated Fluency was operationalized as the number of ideas generated for the problem . Diversity : Mean Pairwise Distance Between Ideas Diversity was operationalized as mean pairwise distance ( the reverse of similarity ) between a participant\u2019s ideas as measured by GloVe . Novelty We recruited 185 workers from MTurk to rate the novelty of the generated ideas on a scale of 1 ( Extremely Obvious ) to 7 ( Extremely Novel ) . Each worker rated a random subset of approximately 30 ideas . Computing correlations between each judges\u2019 ratings and the overall aggregate score yielded an average aggregate - judge correlation of r = . 64 . To deal with potential differences in usage of the rating scale across raters ( e . g . , some might only use the upper end of the scale ) , we compute standardized scores ( a . k . a . \u201cz - scores\u201d ) within raters ( i . e . , a mean score of a rater was subtracted from that rater\u2019s each individual score and the difference was divided by the standard deviation of that rater\u2019s scores ) . An example of a high novelty idea is \u201c [ Chemistry ] [ Lab experiment ] The couple could conduct a common laboratory experiment combining two substances to create a third as part of their ceremony symbolizing and celebrating their union . \u201d ( z - score = 1 . 61 ) . An example of a low novelty idea is \u201c [ formal ] [ gift ] It would be a typical wedding\u201d ( z - score = \u2013 1 . 94 ) Each participant\u2019s novelty score was the highest novelty score across his / her ideas . This conceptualization of novelty is a good fit for the predictions of both associationist and SIAM theories for novelty , which emphasize novelty as an outcome of the ideation process ( i . e . , the most novel idea that was generated , instead of average novelty of all ideas ) . Control Measure : Baseline Fluency In ideation studies , it is important to control for pre - existing differences in participants\u2019 creative capacities , such as baseline fluency of ideation [ 7 , 9 ] . Baseline fluency was operationalized as the number of alternative uses generated for a brick during the warm - up task that participants completed prior to the main ideation task . This measure is intended to capture both aspects of baseline creative fluency [ 20 ] , and aspects of participant motivation and comfort with the interface ( all important for creative productivity ) . RESULTS System - Usage Statistics Participants generated a total of 1 , 574 ideas across conditions . 85 % of participants in the inspiration conditions reported using the inspiration feature in some way ( e . g . , attending to inspirations , requesting inspirations ) . Of those who did , 68 % self - reported interacting with the inspirations ( e . g . , attending to , using as inspiration ) at least \u201csomewhat frequently\u201d ( 3 on a scale of 1 to 5 ) when they weren ' t actively clicking to get more inspirations , M = 3 . 0 ( SE = 0 . 1 ) . Across the inspiration conditions , 49 % of participants requested inspirations at least once . Interestingly , however , likelihood of an inspiration request was not equal across conditions . Participants who received near stimuli during productive ideation ( ALWAYS - NEAR and MATCH - STATE ) were less likely to request an inspiration at least once ( M = . 42 , SE = . 05 ) than participants who received far stimuli during productive ideation ( ALWAYS - FAR and MISMATCH - STATE , M = . 57 , SE = . 05 ) . A logistic regression model , predicting the probability of inspiration request as a function of stimuli distance during productive ideation , showed that the difference between the near and far distance groups was statistically significant , z = 1 . 96 , p = . 05 . Since this analysis was conducted in response to seeing the data ( vs . hypothesized in advance , as with the primary analyses in the subsequent section ) , we wish to clearly mark this finding as exploratory ( rather than confirmatory ) . We revisit this finding in the Discussion . Overall , these numbers suggest that the features of the system relating to inspiration were used reasonably frequently , providing an adequate test of our manipulations . Primary Analyses For each dependent measure , we estimate an ANCOVA model with baseline fluency as a control covariate ( if it is a statistically significant predictor of the dependent measure ) . For median inter - idea interval , we also include number of ideas generated as a theoretically motivated control covariate ( if it is statistically significant ) . All significant main effects of condition are followed up with planned contrasts against the NO - STIMULI condition , using Dunnett\u2019s procedure [ 16 ] to control Type I error inflation from multiple comparisons . Table 2 summarizes model - adjusted means and standard errors for each dependent measure by condition . Slower Ideation with Always - Far and Mismatch - State To investigate ideation pace in each condition , we estimated an ANCOVA predicting median inter - idea interval as a function of condition , controlling for number of ideas ( which was significantly negatively correlated with inter - idea interval , r = \u2013 . 63 , p < . 01 ) . Recall that controlling for number of ideas provides a finer - grained measure of within - category fluency , allowing us to discern qualitative differences ( low vs . high within - category fluency ) between quantitatively similar ( overall number of ideas ) ideation traces . The model showed a significant main effect of condition on median seconds between subsequent ideas , F ( 4 , 233 ) = 3 . 2 , p = . 01 . Planned contrasts showed that ALWAYS - FAR and MISMATCH - STATE participants had significantly longer median inter - idea intervals ( t = 2 . 8 , p = . 02 and t = 3 . 1 , p = . 01 , respectively ) compared to NO - STIMULI participants . Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 99 Lower Transition Similarity in Always - Far Condition Transition similarity was not significantly correlated with baseline fluency ( r = . 02 , p = . 79 ) . Therefore , we estimated an ANOVA with condition as the only factor . The model showed a significant effect of condition , F ( 4 , 218 ) = 4 . 9 , p < . 01 . Planned contrasts showed that ALWAYS - FAR participants had significantly lower median transition similarity ( t = - 3 . 1 , p < . 01 ) than participants in the NO - STIMULI condition . A post - hoc Tukey test also showed that ALWAYS - FAR participants had lower median transition similarity than both MATCH - STATE ( p = . 02 ) and ALWAYS - NEAR participants ( p < . 01 ) . Fluency : Equal Number of Ideas Across Conditions An ANCOVA controlling for baseline fluency showed no significant main effect of condition on the number of ideas , F ( 4 , 239 ) = 0 . 5 , p = 0 . 77 . Always - Far Leads to More Diversity than Always - Near No theoretical covariates were statistically significantly related to diversity . Therefore , we estimated an ANOVA with condition as the only factor . This ANOVA showed a significant main effect of condition on diversity , F ( 4 , 233 ) = 2 . 97 , p < . 01 , but planned contrasts did not show any differences between the inspiration conditions and the control condition . However , a post - hoc Tukey test showed that participants in the ALWAYS - NEAR conditions had significantly lower diversity of ideas compared to the ALWAYS - FAR condition ( p = . 01 ) . No Benefits for Novelty in Always - Far Condition Baseline fluency was not significantly correlated with novelty ( r = . 03 , p = . 68 ) . Thus , we estimated an ANOVA with condition as the only factor . This model showed a significant main effect of condition , F ( 4 , 239 ) = 2 . 5 , p = . 04 . Planned contrasts showed no significant differences between ALWAYS - FAR and NO - STIMULI participants , t = \u2013 2 . 3 , p = . 07 ; however , the mean trends were in the opposite direction predicted by the associationist theory , with the most novel ideas of ALWAYS - FAR participants rated as less novel ( M = 0 . 64 , SE = 0 . 07 ) than NO - STIMULI participants\u2019 most novel ideas ( M = 0 . 88 , SE = 0 . 07 ) . DISCUSSION In this study , we explored how the semantic distance of inspirations from the target user\u2019s own ideas impacts their creative performance . Specifically , we compared two competing theoretical recommendations from creativity theories : 1 ) the associationist view , which predicted that always providing far stimuli would be most beneficial , and 2 ) the SIAM model of creative ideation , which predicted that a state - contingent inspiration delivery ( where near stimuli are delivered during productive ideation , and far stimuli during impasses ) would be most beneficial . Figure 3 summarizes our findings and their implications for the two competing theories . Consistent with the associationist view , ideas generated in the ALWAYS - FAR condition were significantly more diverse than those generated in the ALWAYS - NEAR condition ; however , ALWAYS - FAR ideas were not significantly more diverse than NO - STIMULI ideas . Further , contrary to associationist predictions , ALWAYS - FAR ideas were not significantly more novel than NO - STIMULI ideas ; instead , the mean trends showed that ALWAYS - FAR ideas were possibly less novel than NO - STIMULI ideas . In contrast , consistent with SIAM predictions , MISMATCH - STATE and ALWAYS - FAR participants generated ideas at a slower rate than NO - STIMULI participants , and ALWAYS - FAR participants iterated less within categories compared to NO - STIMULI participants ( indicating that far stimuli hinder within - category exploration ) . Finally , an exploratory analysis showed that participants who received near stimuli during productive ideation ( MATCH - STATE , ALWAYS - NEAR ) were less likely to request inspirations ( which our system used to detect impasses ) than participants who received far stimuli during productive ideation ( ALWAYS - FAR , MISMATCH - STATE ) , suggesting that near stimuli could extend productive ideation chains ( relative to far stimuli ) . However , MATCH - STATE participants did not have greater within - category fluency , overall fluency , or novelty of ideas than NO - STIMULI participants . In summary , we conclude that the SIAM model\u2019s state - contingent view may be more useful as a theoretical starting Inter - idea interval Transition similarity Overall Fluency Diversity Novelty NO - STIMULI 64 . 2 ( 5 . 3 ) 0 . 19 ( 0 . 01 ) 6 . 5 ( 0 . 5 ) 0 . 84 ( 0 . 01 ) 0 . 88 ( 0 . 07 ) ALWAYS - FAR 86 . 2 ( 5 . 7 ) * 0 . 12 ( 0 . 02 ) * * 6 . 1 ( 0 . 5 ) 0 . 86 ( 0 . 01 ) 0 . 64 ( 0 . 07 ) m ALWAYS - NEAR 74 . 3 ( 5 . 6 ) 0 . 20 ( 0 . 02 ) 6 . 4 ( 0 . 5 ) 0 . 81 ( 0 . 01 ) 0 . 67 ( 0 . 07 ) MATCH - STATE 76 . 6 ( 5 . 5 ) 0 . 19 ( 0 . 01 ) 7 . 0 ( 0 . 5 ) 0 . 83 ( 0 . 01 ) 0 . 88 ( 0 . 07 ) MISMATCH - STATE 88 . 7 ( 5 . 8 ) * * 0 . 14 ( 0 . 02 ) 6 . 5 ( 0 . 5 ) 0 . 84 ( 0 . 01 ) 0 . 79 ( 0 . 07 ) Table 2 . Model - adjusted means and standard errors for each dependent measure by condition . m p < . 10 , * p < . 05 , and * * p < . 01 for contrasts with the NO - STIMULI baseline , with Dunnett\u2019s correction for multiple comparisons . Contrasts show that ALWAYS - FAR resulted in significantly longer inter - idea intervals and significantly lower transition similarity and novelty than NO - STIMULI ; MISMATCH - STATE resulted in significantly longer inter - idea intervals than NO - STIMULI . Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 100 point than the associationist theory of creativity for guiding the design of inspiration delivery systems . In practical terms , our findings suggest that following the associationist recommendation to deliver semantically far inspirations throughout ideation ( including during productive ideation , as in the ALWAYS - FAR condition ) is inadvisable , as it provides uncertain benefits for diversity , and relatively certain costs for novelty and within - category exploration . The theoretical assertions of SIAM suggest that the reduction in novelty arises from disruption of the deep - exploration pathway to novel ideas [ 11 , 36 , 41 ] . However , it is still unclear if delivering near stimuli during productive ideation and far stimuli during impasses maximizes benefits for ideation . In the limitations section , we suggest methodological changes that might help future studies explore this issue further . We remind the reader that we make no claims about the relative merits of the associationist or SIAM theories for explaining creativity in general : we restrict our claims to their relative merits for guiding the design of effective inspiration delivery systems . Limitations One limitation of our study is that not all participants entered an impasse state , partially limiting our ability to observe the effects of stimuli during that state . This might be one reason we did not detect an advantage for SIAM\u2019s hypothesized best condition ( MATCH - STATE ) , since\u2014 possibly due to the benefits of receiving near stimuli during productive ideation\u2014many of those participants did not have the opportunity to receive interesting pointers to new areas of exploration ( only available in the impasse state in that condition ) . Perhaps a longer time scale than 8 minutes ( to increase the probability that most participants would run out of ideas ) or induced breaks in the session would provide a better opportunity to study the effects of stimuli during impasses . Relatedly , a longer time scale might provide a clearer test of the potential cumulative benefits of inspiration during productive ideation ( which might come at a slight cognitive cost ) . Finally , our sample consisted of MTurk workers paid $ 6 / hr : while many participants expressed enjoyment in the task , low baseline levels of skill / knowledge are still a possibility , which may have suppressed positive effects of inspirations ( to the extent that skill / motivation is required to adequately benefit from the inspirations ) . These are possible reasons that the four treatment conditions did not significantly outperform the NO - STIMULI participants on any measure . We therefore urge caution generalizing these results to other settings with a longer time scale and with more skilled / knowledgeable / motivated participants . Our system also had a relatively slow response time ( ~ 1 - 2s ) when retrieving inspirations . While this response time is likely to be acceptable from a usability perspective , the inspirations may not always have arrived before participants started writing down their next idea , potentially adding noise to the intervention . A quicker response time might enable a cleaner test of the potential benefits of near stimuli during productive ideation . Broader Implications and Future Work Inspirations should be Delivered at the Right Moments Our study is consistent with previous work that suggests that potentially helpful inspirations ( e . g . , analogous ideas from other domains [ 48 ] , simple hints [ 28 ] , or diverse ideas of others [ 45 ] ) are only helpful when delivered under particular circumstances . These findings underscore the importance of considering not just which inspirations should be delivered to improve creative ideation , but also when they should be delivered . Considering this larger body of findings yields important issues for further research . Figure 3 . Summary of model - adjusted mean contrasts across dependent measures for each inspiration condition against the NO - STIMULI baseline condition ( vertical dashed lines ) . Mean contrasts are reported on the original scale of the dependent measure . Error bars are 95 % confidence intervals . Significant contrasts from the NO - STIMULI baseline ( by Dunnett\u2019s t - test ) are shown in orange . Contrasts that support a theory\u2019s prediction are marked green ; contrasts that contradict ( i . e . , go in the opposite direction of ) a theory\u2019s prediction are marked red . Here , SIAM\u2019s predictions for the ALWAYS - FAR ( for inter - idea interval , transition similarity ) , ALWAYS - NEAR ( for inter - idea interval , transition similarity , fluency ) , and MISMATCH - STATE conditions ( for inter - idea interval ) are supported , while the associationist theory\u2019s predictions for ALWAYS - NEAR is contradicted ( for novelty ) . Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 101 One important issue to consider is whether ideators should receive any inspiration during productive ideation . Does receiving any stimuli during productive ideation amount to mere disruptive interruptions [ 2 , 3 , 45 ] ) ? One state - contingent strategy might be to avoid offering any stimuli during productive ideation , and only offer far stimuli during impasses . Our findings are only partially consistent with this view of inspirations : rather than global deficits associated with any stimulation , we observed meaningful theoretically predicted differences between the conditions , finding that far stimuli was harmful , but near stimuli did not lead to statistically significant deficits . The SIAM model of ideation still provides some theoretical reason to doubt this ( ideation depends on having \u201cidea elements\u201d to recombine ; having more \u201cidea elements\u201d should improve ideation ) , as does prior studies on the benefits of seeing the ideas of others during collaborative brainstorming in small groups [ 14 ] . Future studies that , like our study , vary not just timing but theoretically meaningful variations in the kind of stimuli presented could provide more clarity on this issue . Another important issue is the implications of a state - contingent view for computational creativity support tools that operate at longer time scales ( e . g . , weeks , the lifespan of a project ) than what we examined in this body of work on timing ( i . e . , seconds and minutes ) . At longer time scales , the lines between cognitive states and between real - time inspiration and user - driven search for inspiration might be blurred . For example , how do users get \u201cstuck\u201d , or enter a state of creative flow , within the larger context of a project ? How might we we appropriately tailor the behavior of the inspiration tools to these states at those time scales ( e . g . , patent database search engines [ 30 ] ) ? Finally , if accounting for cognitive states of users is important , how might systems effectively detect when users are in particular cognitive states ? In this study , we used a partially user - driven approach to detecting users\u2019 cognitive states . While this is a reasonable approach for detecting switches to impasse states , future work might explore the use of behavioral markers that predict the onset of an impasse ( e . g . , slowed inter - idea interval , excessively high inter - idea similarity ) and prevent , rather than respond to it . These behavioral markers could be augmented with physiological markers ( e . g . , the fNIRS signals we obtained in our validation study ) , to obtain more nuanced and accurate representations of user states . Advances in the portability and wearability of these physiological sensors open up exciting new avenues for designing creativity support systems that respond to users\u2019 \u201cimplicit input\u201d [ 47 ] . This approach could be especially productive to the extent that changes in cognitive states happen more on a continuum than a binary state shift : systems that can detect early / mild stages of impasse and pre - emptively introduce interventions to prevent impasses might represent a new class of creativity support tools that promote extended states of creative \u201cflow\u201d [ 13 ] . Far Stimuli Should be Used with Caution Our findings also have broader implications for the role of semantic distance in creative inspiration . Far ( rather than near ) stimuli have generally been thought to be more useful for provoking creative mental leaps [ 19 , 22 , 29 , 40 , 50 ] . However , recent investigations are beginning to challenge and refine that claim [ 10 , 39 , 51 ] , pointing out , for example , that overreliance on semantically far stimuli can harm creative performance [ 10 ] . Much of this recent work has focused on the impact of inspirations that are far from one\u2019s problem domain ( e . g . , drawing inspiration from pendulum motions in grandfather clocks when generating ideas for a new approaches to generating electricity in developing countries ) . In this work , we extend the notion that far stimuli should be used with caution to the related but distinct notion of semantic distance of inspirations from one\u2019s current thinking . We do not mean to argue that far stimuli are unimportant or that they should be avoided entirely ; rather , our findings , together with other work on semantic distance , suggest that future research should explore when and how creators can best take advantage of semantically far inspirations . Dual Pathways to Creative Outcomes Finally , our finding that far stimuli during productive ideation not only reduced within - category fluency , but also reduced novelty of ideas ( as predicted by SIAM ) , lends support to the \u201cdual pathway\u201d view of creative ideation [ 11 , 36 , 41 ] , which posits that iteration and deep exploration within categories is an important pathway to creative ( not just better quality ) ideas , perhaps just as important as creative \u201cmental leaps\u201d to remote regions of a solution space [ 50 ] , or combining semantically very different ideas [ 34 ] . Future research on large - scale collaborative ideation could build on this view to explore how to coordinate the crowd to deeply explore within solution approaches , as a complement to promoting cross - pollination of ideas . CONCLUSION In this paper , we empirically examined competing theoretical recommendations for how inspirational delivery systems on collaborative ideation platforms should account for semantic distance of inspirational stimuli . In an online ideation experiment , we find that following the associationist theory\u2019s recommendation to always provide far stimuli yields uncertain benefits for idea diversity and relatively certain costs for within - category fluency and idea novelty . Our research suggests that far inspirations can be harmful for creativity if delivered during productive ideation , and that collaborative inspiration systems could be improved by accounting for ideators\u2019 cognitive states . ACKNOWLEDGEMENTS We thank all of the people ( both in - person and on MTurk ) who participated in the experiment , particularly those who gave valuable feedback on early versions of the experiment . We are also grateful for support from National Science Foundation awards IIS - 1122206 and IIS - 1122320 . Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 102 REFERENCES 1 . Salvatore Andolina , Khalil Klouche , Diogo Cabral , Tuukka Ruotsalo , and Giulio Jacucci . 2015 . InspirationWall : Supporting Idea Generation Through Automatic Information Exploration . In Proceedings of the 2015 ACM SIGCHI Conference on Creativity and Cognition ( C & C \u201915 ) , 103 \u2013 106 . https : / / doi . org / 10 . 1145 / 2757226 . 2757252 2 . Brian P . Bailey and Shamsi T . Iqbal . 2008 . Understanding Changes in Mental Workload During Execution of Goal - directed Tasks and Its Application for Interruption Management . ACM Trans . Comput . - Hum . Interact . 14 , 4 : 21 : 1 \u2013 21 : 28 . https : / / doi . org / 10 . 1145 / 1314683 . 1314689 3 . Brian P Bailey , Joseph Konstan , John V Carlis , and others . 2000 . Measuring the effects of interruptions on task performance in the user interface . In Systems , Man , and Cybernetics , 2000 IEEE International Conference on , 757 \u2013 762 . 4 . Jonali Baruah and Paul B . Paulus . 2011 . Category assignment and relatedness in the group ideation process . Journal of Experimental Social Psychology 47 , 6 : 1070 \u2013 1077 . 5 . Roger E . Beaty and Paul J . Silvia . 2012 . Why do ideas get more creative across time ? An executive interpretation of the serial order effect in divergent thinking tasks . Psychology of Aesthetics , Creativity , and the Arts 6 , 4 : 309 \u2013 319 . 6 . Osvald M . Bjelland and Robert Chapman Wood . 2008 . An Inside View of IBM\u2019s\u2019 Innovation Jam\u2019 . MIT Sloan management review 50 , 1 : 32 \u2013 40 . 7 . Joel Chan , Steven C . Dang , and Steven P . Dow . 2016 . Improving crowd innovation with expert facilitation . In Proceedings of the ACM Conference on Computer - Supported Cooperative Work & Social Computing . https : / / doi . org / 10 . 1145 / 2818048 . 2820023 8 . Joel Chan , Steven Dang , and Steven P . Dow . 2016 . IdeaGens : Enabling Expert Facilitation of Crowd Brainstorming . In Proceedings of the 19th ACM Conference on Computer Supported Cooperative Work and Social Computing Companion ( CSCW \u201916 Companion ) , 13 \u2013 16 . https : / / doi . org / 10 . 1145 / 2818052 . 2874313 9 . Joel Chan , Steven Dang , and Steven P . Dow . 2016 . Comparing Different Sensemaking Approaches for Large - Scale Ideation . In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems , 2717 \u2013 2728 . 10 . Joel Chan , Steven P . Dow , and Christian D . Schunn . 2015 . Do The Best Design Ideas ( Really ) Come From Conceptually Distant Sources Of Inspiration ? Design Studies 36 : 31 \u2013 58 . https : / / doi . org / 10 . 1016 / j . destud . 2014 . 08 . 001 11 . Joel Chan and Christian D . Schunn . 2015 . The importance of iteration in creative conceptual combination . Cognition 145 : 104 \u2013 115 . https : / / doi . org / 10 . 1016 / j . cognition . 2015 . 08 . 008 12 . Allan M . Collins and Elizabeth F . Loftus . 1975 . A spreading - activation theory of semantic processing . Psychological Review 82 , 6 : 407 \u2013 428 . 13 . Mihaly Csikszentmihalyi . 1997 . Flow and the Psychology of Discovery and Invention . HarperPerennial , New York , NY . 14 . Darleen M . DeRosa , Carter L . Smith , and Donald A . Hantula . 2007 . The medium matters : Mining the long - promised merit of group interaction in creative idea generation tasks in a meta - analysis of the electronic group brainstorming literature . Computers in Human Behavior 23 , 3 : 1549 \u2013 1581 . https : / / doi . org / 10 . 1016 / j . chb . 2005 . 07 . 003 15 . Steven P . Dow , Alana Glassco , Jonathan Kass , Melissa Schwarz , Daniel L . Schwartz , and Scott R . Klemmer . 2010 . Parallel Prototyping Leads to Better Design Results , More Divergence , and Increased Self - efficacy . ACM Trans . Comput . - Hum . Interact . 17 , 4 : 18 : 1 \u2013 18 : 24 . https : / / doi . org / 10 . 1145 / 1879831 . 1879836 16 . Charles W . Dunnett . 1955 . A Multiple Comparison Procedure for Comparing Several Treatments with a Control . Journal of the American Statistical Association 50 , 272 : 1096 \u2013 1121 . https : / / doi . org / 10 . 1080 / 01621459 . 1955 . 10501294 17 . Manaal Faruqui and Chris Dyer . 2015 . Non - distributional Word Vector Representations . In Proceedings of ACL . 18 . Katherine Fu , Joel Chan , Christian Schunn , Jonathan Cagan , and Kenneth Kotovsky . 2013 . Expert representation of design repository space : A comparison to and validation of algorithmic output . Design Studies 34 , 6 : 729 \u2013 762 . https : / / doi . org / 10 . 1016 / j . destud . 2013 . 06 . 002 19 . Dedre Gentner and Arthur B . Markman . 1997 . Structure mapping in analogy and similarity . American Psychologist 52 , 1 : 45 \u2013 56 . 20 . Joy P . Guilford . 1950 . Creativity . American Psychologist 5 : 444 \u2013 454 . 21 . Nitin Gupta , Yoonhee Jang , Sara C . Mednick , and David E . Huber . 2012 . The Road Not Taken : Creative Solutions Require Avoidance of High - Frequency Responses . Psychological Science . 22 . Keith J . Holyoak and Paul Thagard . 1996 . Mental leaps : Analogy in creative thought . Cambridge , MA . 23 . Seth Hunter and Pattie Maes . 2008 . WordPlay : A table - top interface for collaborative brainstorming and decision making . Proceedings of IEEE Tabletops and Interactive Surfaces : 2 \u2013 5 . 24 . Theodore J . Huppert , Solomon G . Diamond , Maria A . Franceschini , and David A . Boas . 2009 . HomER : a review of time - series analysis methods for near - infrared spectroscopy of the brain . Applied Optics 48 , 10 : D280 . https : / / doi . org / 10 . 1364 / AO . 48 . 00D280 25 . Theodore J . Huppert , Rick D . Hoge , Solomon G . Diamond , Maria A . Franceschini , and David A . Boas . 2006 . A temporal comparison of BOLD , ASL , and NIRS hemodynamic responses to motor stimuli in adult Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 103 humans . NeuroImage 29 , 2 : 368 \u2013 382 . https : / / doi . org / 10 . 1016 / j . neuroimage . 2005 . 08 . 065 26 . Elahe Javadi and Wai - Tat - . T . Fu . 2011 . Idea Visibility , Information Diversity , and Idea Integration in Electronic Brainstorming . In Proceedings of the 6th International Conference on Foundations of Augmented Cognition : Directing the Future of Adaptive Systems ( FAC\u201911 ) , 517 \u2013 524 . 27 . Elahe Javadi , Joseph Mahoney , and Judith Gebauer . 2013 . The impact of user interface design on idea integration in electronic brainstorming : an attention - based view . Journal of the Association for Information Systems 14 , 1 : 1 \u2013 21 . 28 . Craig A . Kaplan and Herbert A . Simon . 1990 . In search of insight . Cognitive Psychology 22 , 3 : 374 \u2013 419 . 29 . Stefan W . Knoll and Graham Horton . 2011 . The Impact of Stimuli Characteristics on the Ideation Process : An Evaluation of the Change of Perspective \u201cAnalogy . \u201d In Proceedings of the 44th Hawaii International Conference on System Sciences . 30 . S . Koch , H . Bosch , M . Giereth , and T . Ertl . 2009 . Iterative Integration of Visual Insights during Scalable Patent Search and Analysis . In Proceedings of IEEE Transactions on Visualization and Computer Graphics . 31 . Arthur Koestler . 1964 . The act of creation . Macmillan , Oxford , England . 32 . T . K . Landauer and S . T . Dumais . 1997 . A solution to Plato\u2019s problem : The latent semantic analysis theory of acquisition , induction , and representation of knowledge . Psychological Review 104 , 2 : 211 \u2013 240 . 33 . Brian Lee , Savil Srivastava , Ranjitha Kumar , Ronen Brafman , and Scott R . Klemmer . 2010 . Designing with Interactive Example Galleries . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201910 ) , 2257 \u2013 2266 . https : / / doi . org / 10 . 1145 / 1753326 . 1753667 34 . Sara A . Mednick . 1962 . The associative basis of the creative process . Psychological Review 69 , 3 : 220 \u2013 232 . 35 . Ana Sofia Morais , Henrik Olsson , and Lael J . Schooler . 2012 . Mapping the Structure of Semantic Memory . Cognitive Science . https : / / doi . org / 10 . 1111 / cogs . 12013 36 . Bernard A . Nijstad , Carsten K . W . De Dreu , Eric F . Rietzschel , and Matthijs Baas . 2010 . The dual pathway to creativity model : Creative ideation as a function of flexibility and persistence . European Review of Social Psychology 21 : 34 \u2013 77 . https : / / doi . org / 10 . 1080 / 10463281003765323 37 . Bernard A . Nijstad and Wolfgang Stroebe . 2006 . How the group affects the mind : a cognitive model of idea generation in groups . Personality and Social Psychology Review 10 , 3 : 186 \u2013 213 . https : / / doi . org / 10 . 1207 / s15327957pspr1003 _ 1 38 . Jeffrey Pennington , Richard Socher , and Christopher D Manning . 2014 . Glove : Global vectors for word representation . Proceedings of the Empiricial Methods in Natural Language Processing ( EMNLP 2014 ) 12 : 1532 \u2013 1543 . 39 . David N . Perkins . 1983 . Novel Remote Analogies Seldom Contribute to Discovery . The Journal of Creative Behavior 17 , 4 : 223 \u2013 239 . 40 . Tony Poze . 1983 . Analogical connections : The essence of creativity . The Journal of creative behavior 17 , 4 : 240 \u2013 258 . 41 . Eric F . Rietzschel , Bernard A . Nijstad , and Wolfgang Stroebe . 2007 . Relative accessibility of domain knowledge and creativity : The effects of knowledge activation on the quantity and originality of generated ideas . Journal of Experimental Social Psychology 43 , 6 : 933 \u2013 946 . https : / / doi . org / 10 . 1016 / j . jesp . 2006 . 10 . 014 42 . Albert Rothenberg . 1979 . The emerging goddess : The creative process in art , science , and other fields . University of Chicago Press Chicago . 43 . Pao Siangliulue , Kenneth C . Arnold , Krzysztof Z . Gajos , and Steven P . Dow . 2015 . Toward Collaborative Ideation at Scale : Leveraging Ideas from Others to Generate More Creative and Diverse Ideas . In Proceedings of the ACM Conference on Computer Supported Cooperative Work & Social Computing . https : / / doi . org / 10 . 1145 / 2675133 . 2675239 44 . Pao Siangliulue , Joel Chan , Steven P . Dow , and Krzysztof Z . Gajos . 2016 . IdeaHound : Improving Large - scale Collaborative Ideation with Crowd - powered Real - time Semantic Modeling . In Proceedings of 2016 ACM Conference on User Interface Software and Technology ( UIST 2016 ) . 45 . Pao Siangliulue , Joel Chan , Krzysztof Gajos , and Steven P . Dow . 2015 . Providing timely examples improves the quantity and quality of generated ideas . In Proceedings of the ACM Conference on Creativity and Cognition . https : / / doi . org / 10 . 1145 / 2757226 . 2757230 46 . Erin Treacy Solovey , Francine Lalooses , Krysta Chauncey , Douglas Weaver , Margarita Parasi , Matthias Scheutz , Angelo Sassaroli , Sergio Fantini , Paul Schermerhorn , Audrey Girouard , and Robert J . K . Jacob . 2011 . Sensing Cognitive Multitasking for a Brain - based Adaptive User Interface . In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems ( CHI \u201911 ) , 383 \u2013 392 . https : / / doi . org / 10 . 1145 / 1978942 . 1978997 47 . Erin Treacy Solovey , Daniel Afergan , Evan M . Peck , Samuel W . Hincks , and Robert J . K . Jacob . 2015 . Designing Implicit Interfaces for Physiological Computing : Guidelines and Lessons Learned Using fNIRS . ACM Trans . Comput . - Hum . Interact . 21 , 6 : 35 : 1 \u2013 35 : 27 . https : / / doi . org / 10 . 1145 / 2687926 48 . Ian Tseng , Jarrod Moss , Jonathan Cagan , and Kenneth Kotovsky . 2008 . The role of timing and analogical similarity in the stimulation of idea generation in design . Design Studies 29 , 3 : 203 \u2013 221 . 49 . Arno Villringer and Britton Chance . 1997 . Non - invasive optical spectroscopy and imaging of human brain function . Trends in Neurosciences 20 , 10 : 435 \u2013 442 . https : / / doi . org / 10 . 1016 / S0166 - 2236 ( 97 ) 01132 - 6 Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 104 50 . Thomas B . Ward . 1998 . Analogical distance and purpose in creative thought : Mental leaps versus mental hops . In Advances in Analogy Research : Integration of Theory and Data from the Cognitive , Computational , and Neural Sciences , Keith J . Holyoak , Dedre Gentner and B . Kokinov ( eds . ) . Sofia , Bulgaria , 221 \u2013 230 . 51 . R . W . Weisberg . 2009 . On \u201cout - of - the - box\u201d thinking in creativity . In Tools for innovation , A . B . Markman and K . L . Wood ( eds . ) . New York , NY , 23 \u2013 47 . Session : Crowd Ideation C & C 2017 , June 27 \u2013 30 , 2017 , Singapore 105",
    "chanBenefitsPitfallsAnalogies2011": "Joel Chan University of Pittsburgh , LRDC Room 823 , 3939 O\u2019Hara Street , Pittsburgh , PA 15260 e - mail : joc59 @ pitt . edu Katherine Fu Carnegie Mellon University , 5000 Forbes Avenue , Pittsburgh , PA 15213 e - mail : kfu @ andrew . cmu . edu Christian Schunn University of Pittsburgh , LRDC Room 821 , 3939 O\u2019Hara Street , Pittsburgh , PA 15260 e - mail : schunn @ pitt . edu Jonathan Cagan Carnegie Mellon University , Scaife Hall 419 , 5000 Forbes Avenue , Pittsburgh , PA 15213 e - mail : jcag @ andrew . cmu . edu Kristin Wood University of Texas - Austin , 1 University Station , ETC 4 . 146B , M / C C2200 , Austin , TX 78712 - 1063 e - mail : ood @ mail . utexas . edu Kenneth Kotovsky Carnegie Mellon University , Baker Hall 342F , Pittsburgh , PA 15213 e - mail : kotovsky @ andrew . cmu . edu On the Benefits and Pitfalls of Analogies for Innovative Design : Ideation Performance Based on Analogical Distance , Commonness , and Modality of Examples Drawing inspiration from examples by analogy can be a powerful tool for innovative design during conceptual ideation but also carries the risk of negative design outcomes ( e . g . , design \ufb01xation ) , depending on key properties of examples . Understanding these properties is critical for effectively harnessing the power of analogy . The current research explores how variations in analogical distance , commonness , and representa - tion modality in\ufb02uence the effects of examples on conceptual ideation . Senior - level engi - neering students generated solution concepts for an engineering design problem with or without provided examples drawn from the U . S . Patent database . Examples were crossed by analogical distance ( near - \ufb01eld vs . far - \ufb01eld ) , commonness ( more vs . less - common ) , and modality ( picture vs . text ) . A control group that received no examples was included for comparison . Effects were examined on a mixture of ideation process and product var - iables . Our results show positive effects of far - \ufb01eld and less - common examples on novelty and variability in quality of solution concepts . These effects are not modulated by modal - ity . However , detailed analyses of process variables suggest divergent inspiration path - ways for far - \ufb01eld vs . less - common examples . Additionally , the combination of far - \ufb01eld , less - common examples resulted in more novel concepts than in the control group . These \ufb01ndings suggest guidelines for the effective design and implementation of design - by - anal - ogy methods , particularly a focus on far - \ufb01eld , less - common examples during the ideation process . [ DOI : 10 . 1115 / 1 . 4004396 ] Keywords : design cognition , design methods , conceptual design , innovation , analogy 1 Introduction Innovation , de\ufb01ned as the capacity to generate ideas or products that are both novel and useful , is a critical component of success - ful design in today\u2019s economy [ 1 , 2 ] . A number of investigators have argued that innovation can be best managed in the \u201cfuzzy front end\u201d of the design process [ 3 , 4 ] , notably in the ideation phase , where concepts are created either intuitively or through systematic processes . While many approaches exist to create ideas and concepts as part of ideation , the search for and use of analo - gies have been shown to be quite powerful [ 5 \u2013 8 ] . Analogy is a mapping of knowledge from one domain to another enabled by a supporting system of relations or representations between situa - tions [ 9 ] . This process of comparison between situations fosters new inferences and promotes construing problems in new insight - ful ways . This process likewise is dependent on how the problem is represented , encouraging multiple representations to more fully enable analogical reasoning [ 10 , 11 ] . As an illustrative example , the design concept for the bipolar plate of a fuel cell could be use - fully informed by analogy to a plant leaf due to its similarity in functionality . The most signi\ufb01cant functions affecting the current generation capability of a bipolar plate are \u201cdistribute \ufb02uid , \u201d \u201cguide \ufb02uid , \u201d and \u201cdisperse \ufb02uid . \u201d The plant leaf possesses a sim - ilar function chain , where the veins and lamina perform the func - tions . As a result of this analogy , the bipolar plate \ufb02ow \ufb01eld can be designed to mimic the structure of a leaf [ 10 , 11 ] . Design - by - analogy is clearly a powerful tool in the conceptual design process , and a number of methods have been developed to harness its power , such as Synectics [ 12 ] \u2014group design through analogy types ; French\u2019s work on inspiration from nature [ 13 ] ; Biomimetic concept generation [ 14 ] \u2014a systematic tool to index biological phenomena that links to textbook information ; and analogous design using the Function and Flow Basis [ 15 , 16 ] \u2014 analogous and nonobvious product exploration using the func - tional and \ufb02ow basis . However , fundamental questions surround the proper use of design - by - analogy methods . Most critical , and the problems that are the focus in our work , are what should one analogize over , and what reasoning modalities and associated rep - resentations make innovative design - by - analogy more likely ? While these questions have remained largely unanswered in speci\ufb01c knowledge domains such as engineering design , there is related research literature in the domain of psychological studies of creativity , reasoning , and problem solving . In what follows , we review the relevant literature that motivate our present hypothe - ses , describe the methods and \ufb01ndings of our cognitive study , and then discuss the insights and implications of our work . 2 Background 2 . 1 Analogical Distance of Example Designs . One key vari - able of interest with respect to the question of what one should analogize over is analogical distance . This variable can be Contributed by the Design Education Committee of ASME for publication in the J OURNAL OF M ECHANICAL D ESIGN . Manuscript received December 20 , 2010 ; \ufb01nal manuscript received June 7 , 2011 ; published online August 1 , 2011 . Assoc . Editor : Janis Terpenny . Journal of Mechanical Design AUGUST 2011 , Vol . 133 / 081004 - 1 Copyright V C 2011 by ASME Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm conceptualized as ranging over a continuum from far - \ufb01eld ( from a different problem domain ) to near - \ufb01eld ( from the same or very similar problem domain ) , where analogies closer to the far - \ufb01eld end point share little or no surface features with the target domain , while analogies closer to the near - \ufb01eld end point share a signi\ufb01 - cant number of surface features . The potential for creative insights seems clearest when the two domains being compared are very different on the surface [ 17 ] . Classic accounts of creative discov - eries and inventions often highlight the potential of far - \ufb01eld anal - ogies for creative insights , including George Mestral\u2019s invention of Velcro via analogy to burdock root seeds , and Niels Bohr\u2019s dis - covery of the structure of atoms via analogy to the solar system . Empirical work has also supported a link between far - \ufb01eld analo - gies and innovative outcomes . For instance , it has been shown that the number of far - \ufb01eld analogies used by designers during ideation is positively related to the originality of proposed solu - tions , as rated by a sample of potential customers [ 18 ] . Further , exposure to surface dissimilar design examples increases idea novelty relative to using no examples , and exposure to surface similar examples decreases the variety of ideas generated relative to surface dissimilar examples [ 19 ] . On the other hand , far - \ufb01eld analogies can be dif\ufb01cult to retrieve from memory [ 20 ] or notice as relevant to one\u2019s target problem [ 5 ] . In addition , some investigators have disputed the privileged role of far - \ufb01eld analogies in prominent inventions and discoveries [ 21 , 22 ] . As such , it is an open question whether far - \ufb01eld analogies are always bene\ufb01cial to the design process . One way to tease apart possible ways in which far - \ufb01eld and near - \ufb01eld analogies might help or hinder designers is to use multiple measures of ideation processes , including novelty and variety of ideas , as well as aver - age quality and variance in idea quality . An initial testable hy - pothesis is that providing far - \ufb01eld examples would allow one to generate more novel ideas relative to near - \ufb01eld or no examples . 2 . 2 Commonness of Example Designs . Another potential variable of interest is the commonness of example designs ( i . e . , how common the designs are found in designers\u2019 worlds ) . The commonness of the example design in its respective design space increases the probability that a designer would have had prior ex - posure and / or experience with the design . Psychologically , the commonness of an example design is related to the degree to which it activates relevant prior knowledge of a designer . This knowledge can come from exposure to instances ( since designed objects exist in the world ) , or from deliberately structured experi - ences , such as in engineering coursework or in the course of pro - fessional design [ 23 ] . The psychological literature on creativity and problem solving suggests that prior experience with an artifact might in\ufb02uence one\u2019s ability to \ufb02exibly re - represent and use it and combine it with other concepts in a novel fashion . Take for instance , Duncker\u2019s [ 24 ] classic candle problem , where the task is to \ufb01x a lighted candle on a wall in such a way that the candle wax will not drip onto a table below , and the given materials are a can - dle , a book of matches , and a box of thumb - tacks . A correct solu - tion involves emptying the box of tacks and using it as a platform for the candle ; however , this solution eludes most solvers because it requires recognizing an unconventional use of the box as a plat - form . In fact , when the box is presented to solvers empty , with the tacks beside it , solvers are much more likely to \ufb01nd the unconven - tional solution [ 25 ] . Similarly , in Maier\u2019s [ 26 ] two string problem , where the task is to tie two strings together that are hanging from the ceiling just out of arm\u2019s reach from each other using various objects available ( e . g . , a chair , a pair of pliers , etc . ) , people often fail to recognize the solution of tying the pair of pliers to one string and swinging it like a pendulum and catching it while stand - ing on a chair between the strings . These \ufb01ndings demonstrate the phenomenon of \u201cfunctional \ufb01xedness , \u201d where individuals have dif\ufb01culty seeing unusual alternative uses for an artifact . Another potentially relevant \ufb01nding in the psychological litera - ture is that individuals who acquire experience with classes of in - formation and procedures tend to represent them in relatively large , holistic \u201cchunks\u201d in memory , organized by deep functional and relational principles [ 27 \u2013 29 ] . Many researchers have argued that this ability to \u201cchunk\u201d underlies expertise and skill acquisi - tion [ 27 , 30 , 31 ] . However , if the task at hand requires the individ - ual to perceive or represent information in novel ways , e . g . , to stimulate creative ideation in design , representation of that infor - mation in chunks might become a barrier to success , particularly if processing of component parts of the information chunks helps with re - representation [ 32 \u2013 34 ] . These \ufb01ndings lead to a hypothesis that less - common example designs , which designers are less likely to have been exposed to , might present a unique advantage over more - common example designs in terms of the potential for stimulating creative ideation . Speci\ufb01cally , it could be that less - common examples are more likely to support multiple interpretations , and thus facilitate broader search through the space of possible solutions . Addition - ally , given that the commonness of example designs in the world ( e . g . , in practice , curriculum , etc . ) is related to its representation in designers\u2019 long - term memory , e . g . , ease / probability of recall , one could hypothesize that less - common examples might confer an advantage in terms of the novelty of solution paths they inspire . However , the literature gives no a priori reason to expect effects of commonness on mean quality of solution concepts . 2 . 3 Modality of Example Designs . With respect to the question of optimal reasoning modalities , a potential variable of interest is the contrast between pictorial and text - based represen - tations of examples . One possible reason to investigate this con - trast is that pictorial representations , e . g . , sketches , photographs , and engineering drawings , often contain a higher degree of su - per\ufb01cial features than text - based representations of the same in - formation . This might be detrimental to conceptual design , as the presence of representations with a high degree of super\ufb01cial detail , such as in detailed prototypes , in the physical design envi - ronment tend to restrict the retrieval of far - \ufb01eld analogies from memory [ 7 ] . On the other hand , some investigators argue that pictorial - based representations are better for conceptual design ; for example , it has been shown that novice designers who are presented with sketches of example designs produce more novel and higher quality solution concepts on average relative to being presented with text - based example designs [ 35 ] . At a pragmatic level , too , in creating design - by - analogy tools , one ultimately has to decide on a representation format for potential analogies ; thus , it is important to investigate if it matters whether they are represented in pictorial or text - based formats [ 10 , 11 ] . Addition - ally , it is important to know if the effects of example analogical distance or commonness are modulated by their representation modality . 2 . 4 Summary . In summary , a review of the relevant psycho - logical literature suggests that investigating variations in example analogical distance , commonness , and modality might shed some important light on the questions regarding what to analogize over and whether there are optimal reasoning modalities . Prior work tentatively supports a hypothesis favoring far - \ufb01eld over near - \ufb01eld examples . With respect to commonness , to our knowledge , no studies have directly tested the effects of example commonness on conceptual ideation ; however , the literature does suggest a hy - pothesis favoring less - common over more - common examples . Importantly , the theoretical and empirical literature suggest that there might be different effects of example analogical distance and commonness along different dimensions of the ideation pro - cess , thus motivating a \ufb01ne - grained analytic approach to ensure that the effects of these variables can be clearly understood . Finally , the literature appears to be relatively equivocal about the contrast between pictorial and text - based representations ; thus , our investigation of this variable in the present study is more ex - ploratory than hypothesis - driven . 081004 - 2 / Vol . 133 , AUGUST 2011 Transactions of the ASME Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm 3 Experimental Methods 3 . 1 Design . To investigate the effects of example analogical distance and commonness on conceptual design processes and possible interactions with modality , we conducted a 2 ( distance : far - \ufb01eld vs . near - \ufb01eld ) ! 2 ( commonness : more - common vs . less - common ) ! 2 ( modality : pictures vs . text ) factorial experiment , where participants , i . e . , senior - level engineering students , were given a real - world design problem and were asked to generate so - lution concepts \ufb01rst brie\ufb02y without examples , such that they understood the problem , and then with examples , to evaluate the effects of examples on problem solving . To establish whether examples of different types enabled or hindered problem solving , a control group of students executed a similar procedure but received no examples . 3 . 2 Participants . Participants were 153 students ( predomi - nantly mechanical engineering undergraduates ) enrolled at two research universities in the United States . Participants were recruited from classes and were given either extra credit or com - pensation of $ 15 for their participation . Participants ranged from 20 to 38 years in age ( M \u00bc 22 , SD \u00bc 1 . 89 ) . 70 % were male . 87 % were undergraduate engineering students ( 95 % mechanical engi - neering , 5 % electrical engineering and others ) and 13 % masters students in disciplines related to product design ( e . g . , mechanical engineering , product development , business administration ) . 66 % of the participants had at least 1 \u2013 6 months of engineering intern - ship experience , and all but 2 out of the 153 students had experi - ence with at least one prior design project in their engineering curriculum . Approximately 82 % of the students had taken at least one course where a structured approach to design was taught . Thus , most of the participants had relevant mechanical engineer - ing domain knowledge and design experience . Participants were randomly assigned to one of the nine possible conditions in each class by distributing folders of paper materials prior to students arriving in class . The obtained distribution of par - ticipants across the nine conditions is shown in Table 1\u2014the sam - ple populations , N s , are unequal not because of dropout but rather from stochasticity in where students chose to sit down . With these sample populations , statistical power for detecting three - way interactions ( not our theoretical goal ) is modest , but power for detecting two - way interactions and main effects is good . 3 . 3 Design Problem . The design problem was to design a low cost , easy to manufacture , and portable device to collect energy from human motion for use in developing and impover - ished rural communities , e . g . , India , many African countries . This design problem was selected to be meaningful and challenging to our participants . The problem was meaningful in the sense that real - world engineering \ufb01rms are seeking solutions to this problem and the problem involves social value ; thus , students would be appropriately engaged during the task [ 36 \u2013 38 ] . The problem was challenging in the sense that a dominant or accepted set of solu - tions to the problem has yet to be developed ( so students would not simply retrieve past solutions ) , but it was not so complex as to be a hopeless task requiring a large design team and very detailed task analysis . 3 . 4 Selection of Examples . Examples were patents selected from the U . S . Patent Database . Candidate patents were retrieved using keyword search on the U . S . Patent and Trade Of\ufb01ce web - site . The keywords used were basic physical principles , such as induction , heat transfer , potential energy , as well as larger cate - gorical terms like mechanical energy . The \ufb01nal set of eight patents was selected by two PhD - level mechanical engineering faculty based on two sets of criteria : ( 1 ) balanced crossing of the analogi - cal distance and commonness factors , such that there would be two patents in each of the four possible combinations , and ( 2 ) overall applicability to the design problem , over and above ana - logical distance and commonness . Each participant in the analogy conditions received two examples of a particular type , roughly balanced across conditions for applicability . The patents for each of the conditions are shown in Table 2 . With respect to the \ufb01rst set of criteria , the speci\ufb01c guidelines for selection were as follows : 1 . Distance : Far - \ufb01eld patents were devices judged to be not directly for the purpose of generating electricity , while near - \ufb01eld patents were those judged to be directly for the purpose of generating electricity . 2 . Commonness : More - common patents were devices judged likely to be encountered by our target population in their standard engineering curriculum and / or everyday life , while less - common patents were those judged unlikely to be seen previously by the participants under typical circumstances . With respect to the modality factor , in the picture conditions , participants received a representative \ufb01rst \ufb01gure from the patent , which typically provides a good overview of the device , while in the text conditions , participants received the patent abstract . In some cases , abstracts differed substantially in length ; to equate for quantity of text across conditions , overly brief abstracts were aug - mented with additional text from the body of the patent , which elaborated on the details of the design and technology . To provide some foundational context , all text - and - picture - condition partici - pants also received the patent title . 3 . 5 Experimental Procedure . The experiments were con - ducted during class . Participants generated solution concepts in three phases and subsequently completed a background survey . Participants proceeded through the phases using a sequence of envelopes to carefully control timing of the task and exposure to examples across conditions . In particular , we wanted to ensure that design examples were received only after participants had made some substantial progress in ideation , since prior work has shown that examples and potential analogies are most helpful when received after ideation has already begun [ 39 , 40 ] . The over - all time allowed for this task was suf\ufb01cient to allow for broad ex - ploration of the concept space , but not enough to develop Table 1 Distribution of participants across conditions Near - field Far - field More - common Less - common More - common Less - common Picture 13 17 15 16 Text 17 16 16 17 Control 24 Table 2 Patents for each condition Near - field Far - field More - common - Waterwheel - driven generating assembly ( 6208037 ) - Escapement mechanism for pendulum clocks ( 4139981 ) - Recovery of geothermal energy ( 4030549 ) - Induction loop vehicle detector ( 4568937 ) Less - common - Apparatus for producing electrical energy from ocean waves ( 4266143 ) - Accelerometer ( 4335611 ) - Freeway power generator ( 4247785 ) - Earthquake isolation \ufb02oor ( 4402483 ) Journal of Mechanical Design AUGUST 2011 , Vol . 133 / 081004 - 3 Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm particular ideas in depth , matching our focus on the ideation process . Analogy and control groups executed the same overall sequence , but differed in the particular activities in the second phase of ideation ( see Fig . 1 for a comparison of the procedures ) . In general , the sequence of phases was to : ( 1 ) read design problem and generate solution concepts , ( 2 ) either ( a ) review two patents and write / draw solutions / ideas that come to mind when looking at the patents or ( b ) continue generating concepts , and ( 3 ) generate more solution concepts . Each phase lasted 10 min . With respect to idea generation , participants were instructed to generate and record as many solution concepts to the design prob - lem as they could , including novel and experimental ones , using words and / or sketches to describe their solution concepts . 4 Ideation Metrics The experiment generated 1321 total ideas . To thoroughly explore the range of effects of varying the analogical distance , commonness , and modality of design examples on conceptual design processes , we applied a range of ideation metrics to these ideas : ( 1 ) the extent to which solution features were transferred from examples , ( 2 ) quantity of ideation , ( 3 ) breadth of search through the space of possible solutions , ( 4 ) quality of solution concepts , and ( 5 ) novelty of solution concepts . The \ufb01rst three met - rics provided measures of the ideation process of participants and how they processed the examples : examining solution transfer provides insight into the mechanisms by which participants might be stimulated by the examples , e . g . , did they actually use solution elements ; measuring quantity of ideation gave a sense of how par - ticipants were exploring the design space , i . e . , whether they were generating and re\ufb01ning a small number of ideas , or exploring mul - tiple concepts and variations of concepts , which is associated with higher likelihood of generating high - quality concepts [ 4 ] ; \ufb01nally , breadth of search was taken to be a measure of the ability to gen - erate a wide variety of ideas , which is associated with the ability to restructure problems , an important component of creative abil - ity [ 41 \u2013 43 ] . The \ufb01nal two metrics focused on the ideation prod - ucts of participants . We investigated quality because in design , a baseline requirement is that concepts must meet customer speci\ufb01 - cations ; design concepts that are novel but do not meet customer speci\ufb01cations cannot be considered acceptable designs , let alone creative ones [ 41 ] . We investigated novelty because there is a high degree of consensus in the literature that creative products are at least novel [ 41 , 42 ] . 4 . 1 Data Preprocessing . The raw output of each participant was in the form of sketches and / or verbal descriptions of concepts . Examples of participant - generated solution concepts are shown in Fig . 2 . A number of preprocessing steps were necessary to prepare the data for coding and analysis . First , each participant\u2019s raw output was segmented by a trained coder into solution concepts . A sketch and / or verbal description was segmented as one solution concept if it was judged to describe one distinct solution to the design problem . Variations of solutions ( e . g . , with minor modi\ufb01cations ) were counted as distinct solution concepts . Segmentation was independently checked by a second coder . Inter - rater agreement was high ( 96 % ) , and all dis - agreements were resolved by discussion . Next , sets of two senior mechanical engineering students rated each solution concept as meeting or not meeting the minimum constraints of the design problem , as described above , to remove off - topic inspirations gen - erated by the patent examples , especially in the second phase . Inter - rater agreement was acceptable , with an average Cohen\u2019s kappa of 0 . 72 . All disagreements were resolved through discus - sion . The 1066 solution concepts remaining after preprocessing constituted the \ufb01nal data set for analysis . 4 . 2 Solution Transfer . Solution transfer was de\ufb01ned as the degree to which a given participant\u2019s idea set contained solution features from the examples she / he received . The process of pro - ducing a solution transfer score for each participant was as fol - lows . First , key features were generated by one of the co - authors for each of the eight patent examples , and the list was cross - checked for relevance by the other co - authors . Recall that each participant received two examples ; however , since picture and text examples were essentially the same examples ( only in differ - ent representations ) , the 2 ! 2 ! 2 design reduced to a 2 ! 2 design , leaving a total of eight examples . A total of 39 key fea - tures were identi\ufb01ed . Because some features overlapped across examples ( e . g . , \u201cbuilt into ground , stationary , or permanent\u201d was Fig . 1 Comparison of experimental procedures for analogy vs . control groups Fig . 2 Example participant solution concepts 081004 - 4 / Vol . 133 , AUGUST 2011 Transactions of the ASME Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm associated with four patent examples ) , there was not a simple one - to - one mapping of features to examples . The number of features associated with each of the eight examples ranged from 4 to 7 ( M \u00bc 4 . 9 , SD \u00bc 1 . 0 ) . Second , each participant solution concept was coded for the presence or absence of a set of the features found in the full set of patent examples presented to participants . The \ufb01rst 50 % of solution concepts was double - coded by two sen - ior mechanical engineering students to establish reliability . Later , all coding was completed by one student only . Test - retest meas - ures of reliability were obtained in lieu of inter - rater reliabilities . Cohen\u2019s kappa averaged across features was 0 . 57 . Because some features had low coding reliability or high overlap of features across many of the patents or simply were common elements of most proposed solutions across all conditions , the initial set of 39 features was \ufb01ltered down to 23 features according to three criteria : 1 . Acceptable inter - rater agreement , i . e . , Cohen\u2019s kappa greater than 0 . 4 . 2 . Not shared by more than three examples . 3 . Not too common , i . e . , base rate ( collapsed across conditions ) less than 0 . 5 . After \ufb01ltering , the number of features ranged from 1 to 5 ( M \u00bc 2 . 9 , SD \u00bc 1 . 4 ) per example and from 4 to 8 ( M \u00bc 5 . 8 , SD \u00bc 1 . 7 ) per each of the four conditions in the distance by com - monness 2 ! 2 design . Cohen\u2019s kappa averaged across the \ufb01ltered set of features was 0 . 66 . To produce solution transfer scores for each participant , the fol - lowing procedure was used . First , for each cell in the 2 ! 2 ( distance ! commonness ) , we computed for each participant the proportion of his / her ideas that had at least one solution feature from the examples she / he received . Next , this proportion was con - verted into a standardized z - score by subtracting the mean and dividing by the standard deviation of proportion scores for all par - ticipants who were not in that 2 ! 2 cell . The reason for using this transformation was that solution features from examples could occur in participants\u2019 ideas even if they never saw the relevant examples ; this transformation allows us to separate the probability of participants using solution features from examples they have seen from the probability of using those solution features even if they had never seen the examples . For each participant , the trans - fer score was the z - score of each feature relevant to the examples they actually received . The solution transfer score thus gave a measure of the degree to which a given participant\u2019s idea set differed from \u201cnormal\u201d in terms of the proportion of ideas with at least one feature from the examples she / he received . To illustrate , suppose participant 1001 had a z - score of 1 . 34 for far - \ufb01eld , more - common examples . This number would say that the proportion of 1001\u2019s ideas with at least one solution feature from the examples s / he received was 1 . 34 standard deviations higher than the mean proportion of ideas with at least one solution from those examples under \u201cnormal\u201d circum - stances ( i . e . , without having seen either of the two far - \ufb01eld , more - common examples ) . 4 . 3 Quantity of Ideation . Quantity of ideation was de\ufb01ned as the number of solution concepts generated post analogy , i . e . , from the second phase of ideation onwards , that met the minimum constraints of the design problem , viz . ( 1 ) the device generates electricity , and ( 2 ) it uses human motion as the primary input . As noted in the introduction , quantity is often taken to be a key com - ponent of creativity . Quantity was de\ufb01ned at the level of the par - ticipant , i . e . , each participant received a single quantity score . Because we were primarily interested in the effects of examples on quantity , analyses concentrated on the number of solution con - cepts generated after receiving examples ( i . e . , after the \ufb01rst phase ) adjusting for the number of solution concepts generated in the \ufb01rst phase ( which acted as a covariate to adjust for baseline variation in quantity across participants ) . 4 . 4 Breadth of Search . Breadth of search was conceptual - ized in our study as the proportion of the space of possible solu - tions searched by a given participant . To determine the space of possible solutions , the design problem was \ufb01rst functionally decomposed into potential subfunctions by one of the authors , drawing from the reconciled function and \ufb02ow basis of Hirtz and colleagues [ 16 ] . Due to the open - ended nature of the design problem , a rela - tively large number of subfunctions were initially generated , as follows : 1 . Import / accept human interaction 2 . Transform human energy to mechanical energy 3 . Transform human energy to alternative energy 4 . Import other material 5 . Contain / store other material 6 . Transfer other material 7 . Import alternative energy source 8 . Transform alternative energy source into mechanical energy 9 . Transform alternative energy source to alternative energy 10 . Transform collected energy to mechanical energy 11 . Transmit mechanical energy 12 . Transform mechanical energy 13 . Store mechanical energy 14 . Transform mechanical to alternative energy 15 . Transform alternative energy to electrical energy 16 . Actuate / deactuate energy 17 . Transform mechanical energy to electrical energy 18 . Condition electrical energy 19 . Store electrical energy 20 . Supply electrical energy 21 . Transmit electrical energy 22 . Convert electrical to light or EM Each subfunction solution consisted of a how and what compo - nent , where the former speci\ufb01es the component of the solution concept that implements the subfunction , and the latter speci\ufb01es either the input or the output of the subfunction ( whichever is the less speci\ufb01ed ) . For example , a solution for the subfunction \u201cimport human\u201d might be \u201c foot with pedals . \u201d Two senior mechanical engineering students independently coded the solutions to the subfunctions for each solution concept . The solution types for the how and what components of each sub - function were generated bottom - up by the students as they coded , with each new solution type being added to a running list of solu - tion types ; the running list of solution types for each subfunction constituted the coding scheme . Inter - rater reliability was high , with an average Cohen\u2019s kappa across subfunctions of 0 . 84 . All disagreements were resolved by discussion . While the nature of the design problem was open - ended , a core set of subfunctions emerged from the dataset : only a small subset of the initial set of subfunctions occurred often enough for stable estimates of breadth and novelty ( i . e . , base rate greater than 0 . 1 , collapsed across conditions ) : 1 . Import human 2 . Transform human energy to mechanical energy 3 . Import alternative energy 4 . Transform alternative energy to mechanical energy 5 . Transform mechanical energy to electrical energy 6 . Store electrical energy Upon more detailed analysis , it turned out that there were only two solution types for the subfunction \u201cstore electrical energy , \u201d namely \u201cbattery\u201d or \u201ccapacitor , \u201d and the frequency of occurrence for each solution type was relatively equivalent ; thus , novelty scores for this subfunction would be unlikely to differentiate between participants . Furthermore , since the design problem was focused on the problem of harvesting ( vs . storing ) energy , data for this subfunction were not included in computations of breadth . Journal of Mechanical Design AUGUST 2011 , Vol . 133 / 081004 - 5 Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm We de\ufb01ned the space of possible solutions for each of the what and how components of each subfunction by enumerating the number of distinct solution types generated by participants across all phases of ideation . A breadth score b j for each participant on subfunction j was then computed with b j \u00bc X n k \u00bc 1 w jk ! C jk T jk ( 1 ) where C jk is the total number of solution types generated by the participant for level k of subfunction j , T jk is the total number of solution types produced by all participants for level k of subfunc - tion j , and w k is the weight assigned level k . To give priority to breadth of search in the what space ( types of energy / material manipulated ) , we gave a weight of 0 . 66 to the what level ( which was assigned to k \u00bc 1 ) , and a weight of . 33 to the how level ( which was assigned to k \u00bc 2 ) . An overall breadth score for each partici - pant was given by the average of breadth scores for each of the three subfunctions j . 4 . 5 Quality . Quality of solution concepts was measured using holistic ratings on a set of subdimensions of quality . Two other senior mechanical engineering students independently coded solution concepts on 5 - point scales ranging from 0 to 4 ( 0 is unac - ceptable and 4 is excellent ) for six subdimensions of quality , cor - responding to a set of possible customer speci\ufb01cations : 1 . Cost 2 . Feasibility of materials / cost / manufacturing 3 . Feasibility of energy input / output ratio 4 . Number of people required to operate device at a given moment 5 . Estimated energy output 6 . Portability 7 . Time to set up and build , assuming all parts already avail - able at hand These subdimensions were generated by the second author , who is a Ph . D . candidate in mechanical engineering focusing on design methods and cognition , and checked for validity by two other authors , who are mechanical engineering faculty specializing in en - gineering design . For each subdimension , each point on the 5 - point scale was anchored with a unique descriptor . For example , for the \u201cfeasibility of energy input / output ratio\u201d subdimension , 0 was \u201cunfeasible design or input energy completely dwarfs output , \u201d 1 was \u201cinput less than output\u201d , 2 was \u201cI / O about even , \u201d 3 was \u201csustainable / little surplus output ; human input easy , \u201d and 4 was \u201coutput signi\ufb01 - cantly higher than input . \u201d Inter - rater agreement was computed using a Pearson correlation between the ratings of the two coders for each subdimension . The average of correlations across subdimensions was 0 . 65 , and the range was from 0 . 49 to 0 . 77 . An overall quality score was computed for each solution concept , as given by Q \u00bc P n j \u00bc 1 q j ! r j Q max ( 2 ) where q j is the quality score for quality subdimension j , r j is the reliability of the coding for that subdimension , and Q max is the max - imum possible overall quality score , which would be given by set - ting q j to 4 for each subdimension . The contributions of subdimension scores to the overall quality score were weighted by reliability to minimize the in\ufb02uence of measurement error . Since the overall quality score was a proportion of the maximum possible quality score , the score ranged from 0 to 1 . Agreement between coders at the level of this composite score was acceptable ( r \u00bc 0 . 68 ) . 4 . 6 Novelty . Novelty was de\ufb01ned as the degree to which a particular solution type was unusual within a space of possible solutions . This approach allowed us to avoid the dif\ufb01culties of judging the novelty of thousands of solution concepts via holistic rating methods . Recall that for the breadth metric , the space of possible solutions was de\ufb01ned in terms of a set of \ufb01ve core sub - functions for the design problem ; recall further that each subfunc - tion was decomposed further into what and how components , where the former speci\ufb01es the component of the solution concept that implements the subfunction , and the latter speci\ufb01es either the input or the output of the subfunction ( whichever is the less speci - \ufb01ed ) . Rather than computing novelty scores for solutions to each level of each subfunction ( the what and how levels ) , we chose to compute novelty scores for the conjunction of what and how solu - tion components for each subfunction . For example , rather than computing the relative unusualness of the solution components \u201cfoot\u201d and \u201cpedals\u201d separately for the solution \u201cfoot with pedals\u201d for the subfunction \u201cimport human interaction , \u201d the relative unusualness of the solution \u201cfoot with pedals\u201d relative to other solutions would be computed . The rationale for this choice was that these words in conjunction as a solution have a speci\ufb01c mean - ing that needed to be considered . Novelty scores were computed for each subfunction solution using Eq . ( 3 ) , which is a formula adapted from Ref . [ 39 ] N i \u00bc T i # C i T i ( 3 ) where T i is the total number of solution tokens generated for sub - function i in the \ufb01rst phase of ideation ( collapsed across all partic - ipants ) , and C i is the total number of solution tokens of the current solution type in the \ufb01rst phase of ideation . Because this measure was essentially a measure of proportion , the novelty score for each idea ranged from 0 to 1 , with 0 representing solution types found in every solution ( this extreme was never observed ) and 1 representing solution types that never occurred in the \ufb01rst phase . The initial set of solution concepts ( generated in the \ufb01rst phase of ideation ) was taken to be the original design space of the partici - pants since it corresponded to concepts generated prior to receiv - ing examples . The \ufb01nal novelty score for each solution concept was the average of its subfunction novelty scores . 5 Results 5 . 1 Relationships Between Metrics . Analysis of the inter - relationships between the ideation metrics suggested a preliminary process model that could account for these correlations and help to conceptually organize the results ( see Fig . 3 ) . Of course , corre - lations per se do not guarantee causation and other causal models are possible . The preliminary process model is as follows : \u2022 Increased solution transfer results in decreased quantity , pos - sibly because many participants had trouble thinking of solu - tions beyond the ones presented . \u2022 A high quantity of ideation allows for greater breadth of search , even if only on a statistical sampling basis . Fig . 3 Summary of intermetric correlations . Numbers shown are Pearson\u2019s r . All correlations are signi\ufb01cant at p < 0 . 01 . 081004 - 6 / Vol . 133 , AUGUST 2011 Transactions of the ASME Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm \u2022 Greater breadth of search , perhaps also only on a statistical sampling basis , in turn allows for the generation of higher novelty and higher quality solution concepts . \u2022 Repeatedly searching on the fringes of the design space ( as measured by high average novelty ) further increases the probability of \ufb01nding a highly novel concept . \u2022 Finally , increasing the variability of the quality of solution concepts increases the probability of generating a high - qual - ity solution concept . This last relationship is in accord with the work of Ulrich and colleagues in the \ufb01eld of innovation management , who have argued and showed empirically that one way to increase the likelihood of \ufb01nding high market potential product concepts is to increase the variance of the quality of the concepts that are generated [ 4 , 44 ] . 5 . 2 Effects of Analogy Manipulations on Ideation Metrics . We now present our \ufb01ndings by manipulation ( distance , commonness , and modality ) , using the preliminary process model as an organizational framework . Effects of manipulations on the ideation metrics will be described following the \ufb02ow of the process model , \ufb01rst considering solution transfer , quantity , and breadth , followed by consideration of effects on quality and novelty of ideation . Separate 3 - way ( distance ! commonness ! modality ) analysis of variance ( ANOVA ) models were computed for each process variable in the model . In some cases ( indicated in each case ) , the level of that variable during the pre - analogy phase was used as a covariate in the analysis because the baseline measure was a signi\ufb01cant predictor of postanalogy performance . 5 . 2 . 1 Analogical Distance of Examples . There was a main effect of example distance ( p < 0 . 01 , g 2 \u00bc 0 . 08 ) on solution trans - fer , where participants who received far - \ufb01eld examples were much more likely than participants who received near - \ufb01eld exam - ples to use solution elements from the examples they received ( d \u00bc 0 . 60 ) ; 1 in fact , solution features from near - \ufb01eld examples were no more likely to be present in participant solutions after processing examples relative to the pre - example phase ( see Fig . 4 , bottom left ) . There was also a main effect on quantity ( p < 0 . 01 , g \u00bc 0 . 05 ) , where participants who received far - \ufb01eld examples generated sig - ni\ufb01cantly fewer solution concepts relative to participants who received near - \ufb01eld examples ( p < 0 . 05 , d \u00bc # 0 . 30 ; see Fig . 4 , upper left ) . There were no signi\ufb01cant differences in terms of quantity between receiving no examples ( control ) and receiving either far - or near - \ufb01eld examples . However , the small effect of distance on quan - tity did not translate into an effect on breadth : there were no reliable effects of distance on breadth of search ( p \u00bc 0 . 78 , g 2 \u00bc 0 . 00 ) . With respect to quality of solution concepts , there were no effects of distance on either mean or maximum quality . However , there was a main effect of distance of the variability in quality of participants\u2019 solution concepts ( p < 0 . 05 , g 2 \u00bc 0 . 06 ; see Fig . 4 , lower right ) , where participants who received far - \ufb01eld examples had a larger standard deviation in quality of solution concepts Fig . 4 Summary of effects of example distance . * p < 0 . 05 and * * p < 0 . 01 . Control group data are shown in white bars . Error bars are 6 1 standard error . 1 d statistics estimate the size of the difference in group means in terms of the av - erage standard deviation of the two groups in the contrast ; in this case , d \u00bc 0 . 60 esti - mates that the mean probability of transfer is greater with far - \ufb01eld vs near\ufb01eld examples by 0 . 60 of a standard deviation ( a moderate to large difference ) . Journal of Mechanical Design AUGUST 2011 , Vol . 133 / 081004 - 7 Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm than participants who received either near - \ufb01eld examples ( p < 0 . 05 , d \u00bc 0 . 64 ) or no examples ( p < 0 . 05 , d \u00bc 0 . 78 ) . There were no signi\ufb01cant differences between receiving near - \ufb01eld examples vs . no examples . Finally , there was a main effect of distance on mean novelty ( p < 0 . 05 , g 2 \u00bc 0 . 04 ) , where participants who received far - \ufb01eld examples generated solution concepts that were more novel on av - erage relative to participants who received near - \ufb01eld examples ( p < 0 . 05 , d \u00bc 0 . 56 ; see Fig . 4 , upper right ) . Similar patterns of effects were found with maximum novelty of solution concepts ( p < 0 . 05 , g 2 \u00bc 0 . 04 ) , where the most novel solution concept of participants who received far - \ufb01eld examples was more novel on average relative to the most novel solution concept of participants who received near - \ufb01eld examples ( p < 0 . 05 , d \u00bc 0 . 56 ) . There were no signi\ufb01cant differences between participants who received no examples ( control ) vs . near - or far - \ufb01eld examples on either mean or maximum novelty . In summary ( see Fig . 4 ) , example distance appeared to have signi\ufb01cant effects on multiple aspects of ideation . Speci\ufb01cally , novelty and variability in quality of concepts increased as a func - tion of receiving far - \ufb01eld examples , although only in the latter case was the contrast with control statistically signi\ufb01cant . The so - lution transfer metric suggests that these increases might be asso - ciated with incorporating solution elements from the far - \ufb01eld examples . However , the bene\ufb01ts of far - \ufb01eld examples came with a slight cost , viz . a reduction in quantity : in meaningful terms , the cost of processing far - \ufb01eld examples given a standard time for ideation appeared to be , on average , about one solution concept . 5 . 2 . 2 Commonness of Examples . Turning now to the main effects of commonness in the same ANOVAs , there were no reliable effects on solution transfer ( p = 0 . 30 , g 2 \u00bc 0 . 01 ) . How - ever , there was a main effect on quantity ( p < 0 . 01 , g 2 \u00bc 0 . 12 ) , where participants who received more - common examples gen - erated signi\ufb01cantly fewer solution concepts relative to partici - pants who received either more - common examples ( p < 0 . 01 , d \u00bc # 0 . 67 ) or no examples ( p < 0 . 01 , d \u00bc # 0 . 76 ; Fig . 5 , upper left ) . There were no signi\ufb01cant differences in quantity between participants who received less - common vs . no examples ( con - trol ) . There was also a main effect of on breadth of search ( p < 0 . 01 , g 2 \u00bc 0 . 07 ) , where participants who received more - common examples searched less of the design space than par - ticipants who received either less - common examples ( p < 0 . 05 , d \u00bc # 0 . 61 ; Fig . 5 , lower middle ) or no examples ( p < 0 . 01 , d \u00bc # 1 . 03 ) . There were no signi\ufb01cant differences in breadth of search between participants who received less - common vs . no examples ( control ) . With respect to quality of solution concepts , there were no reli - able effects of commonness on either mean or max quality . How - ever , there was a main effect on variability in quality of participants\u2019 solution concepts ( p < 0 . 05 , g 2 \u00bc 0 . 06 ; see Fig . 5 , lower right ) , where participants who received less - common exam - ples had a larger standard deviation in quality of solution concepts than participants who received either more - common examples ( p < 0 . 05 , d \u00bc 0 . 62 ) or no examples ( p < 0 . 05 , d \u00bc 0 . 68 ) . There were no signi\ufb01cant differences between receiving more - common examples vs . no examples . Fig . 5 Summary of effects of example commonness . * p < 0 . 05and * * p < 0 . 01 . Control group data are shown in white bars . Error bars are 6 1 standard error . 081004 - 8 / Vol . 133 , AUGUST 2011 Transactions of the ASME Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm Finally , there were main effects on mean novelty ( p < 0 . 01 , g 2 \u00bc 0 . 10 ) , where participants who received less - common exam - ples generated solution concepts that were more novel on average relative to participants who received more - common examples ( p < 0 . 01 , d \u00bc 0 . 61 ; see Fig . 5 upper right ) and maximum novelty ( p < 0 . 01 , g 2 \u00bc 0 . 96 ) , where the most novel solution concept of participants who received less - common examples was more novel on average relative to the most novel solution concept of partici - pants who received more - common examples ( p < 0 . 01 , d \u00bc 0 . 61 ) . There were no signi\ufb01cant differences between participants who received no examples ( control ) vs . more - or less - common exam - ples on either mean or maximum novelty . In summary ( see Fig . 5 ) , example commonness also appeared to have signi\ufb01cant effects on ideation . Less - common examples were associated with more positive ideation processes and prod - ucts relative to more - common examples , with bene\ufb01ts for quan - tity and breadth of ideation , variability in solution quality , and novelty of solution concepts , although only in the case of vari - ability in solution quality was the contrast with control statisti - cally signi\ufb01cant . 5 . 2 . 3 Joint Effects of Example Distance and Commonness on Novelty . While far - \ufb01eld and less - common examples separately increased novelty of ideas , neither far - \ufb01eld examples as a whole nor less - common examples as a whole were signi\ufb01cantly different from control , which sat in the middle . To examine whether the combination of far - \ufb01eld and less - common properties increased novelty over control , we used a Dunnett\u2019s multiple comparison post hoc test . Since there were no effects of modality on novelty ( described below ) , we collapsed across the picture and text factors and conducted the post hoc test comparing each of the combina - tions in the 2 ! 2 matrix ( distance x commonness ) with the control condition as a reference group . The post hoc test showed that the combination of far - \ufb01eld , less - common examples did in fact increase novelty vs . control , for both mean ( d \u00bc 1 . 14 ; see Fig . 6 ) and max ( d \u00bc 1 . 29 ) . 5 . 3 Effects of Example Modality . Turning to the effects of modality in the overall ANOVAs , there was a main effect of example modality ( p < 0 . 01 , g 2 \u00bc 0 . 09 ) on solution transfer , where participants who received their examples in text form were more likely to use solution elements from the examples they received , regardless of distance or commonness of the example ( d \u00bc 0 . 60 ; Fig . 7 , lower left ) . There was also a main effect of on quantity ( p < 0 . 01 , g \u00bc 0 . 12 ; Fig . 7 , upper left ) , where participants who received text examples generated signi\ufb01cantly fewer solution concepts relative to partici - pants who received either picture examples ( p < 0 . 01 , d \u00bc # 0 . 67 ) or no examples ( control ; p < 0 . 05 , d \u00bc # 0 . 56 ) . There were no sig - ni\ufb01cant differences between participants who received picture examples vs . no examples ( control ) . Thus , receiving examples in text form increased the likelihood of being able to use solution Fig . 6 Mean novelty of solution concepts by example distance and commonness . * p < 0 . 05 . Error bars are 6 1 standard error . Fig . 7 Summary of effects of example modality . * p < 0 . 05 and * * p < 0 . 01 . Control group data are shown in white bars . Error bars are 6 1 standard error . Journal of Mechanical Design AUGUST 2011 , Vol . 133 / 081004 - 9 Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm elements from those examples relative to picture form , but also decreased quantity by an average of about two concepts relative to receiving either picture or text examples . There were no additional effects of modality on the other de - pendent measures ( breadth , p \u00bc 0 . 11 , g 2 \u00bc 0 . 03 ; mean novelty , p \u00bc 0 . 20 , g 2 \u00bc 0 . 02 ; max novelty , p \u00bc 0 . 49 , g 2 \u00bc 0 . 00 ; quality vari - ability , p \u00bc 0 . 44 , g 2 \u00bc 0 . 01 ) . Thus , modality had little impact on the key end - state outputs of the ideation process , unlike the effects of example commonness or example analogical distance . 6 Discussion 6 . 1 Optimal Example Types . Our \ufb01ndings demonstrate that the analogical distance and commonness of examples signi\ufb01cantly in\ufb02uences their impact on designers\u2019 ideation . With respect to an - alogical distance , augmenting ideation with far - \ufb01eld examples brings signi\ufb01cant bene\ufb01ts vis - a ` - vis the kinds of concepts that can be generated ; speci\ufb01cally , ideation with far - \ufb01eld examples enhan - ces the ability to generate highly novel solution concepts and also allows for more variability in the quality of concepts , which may increase the likelihood of generating high quality concepts . It is interesting to note that , even though the far - \ufb01eld examples we gave participants were not energy - generating devices , they were still able to bene\ufb01t from the concepts and solution elements in the devices . This sort of transfer is greater in distance than typically seen in the analogy literature , where far - \ufb01eld analogies in problem solving are usually from cases in other domains that are surface dissimilar but still solve the same basic problem [ 20 , 45 ] . However , the use of far - \ufb01eld examples was not without some cost . Far - \ufb01eld examples reduced overall quantity of ideation rela - tive to near - \ufb01eld or no examples . This \ufb01nding can be interpreted in terms of processing dif\ufb01culty . When we computed an addi - tional 3 - way ANOVA model on quantity for only the \ufb01nal phase of ideation , removing from consideration quantity of ideation while processing examples , the effects of distance were no longer present ( p \u00bc 0 . 47 ) . This suggests that the reduction in quantity can be attributed to the time taken to map the far - \ufb01eld example to the design problem . Thus , it appears that far - \ufb01eld examples not only carry with them the potential to increase novelty and quality of design concepts generated but also carry an initial processing cost in terms of time taken to map them to the target problem . With respect to commonness of examples , we found that the use of less - common examples positively impacts ideation . Less - common examples resulted in increased quantity of ideation , breadth of search , and higher novelty of ideas relative to more - common examples . In a follow - up analysis analyzing quantity for only the \ufb01nal phase of ideation , the positive effects of less - com - mon examples relative to more - common examples were still pres - ent ( p < 0 . 05 , d \u00bc 0 . 56 ) , suggesting that the effects cannot be explained simply in terms of initial processing costs , as in the case of distance effects on quantity . Thus , it seems that less - common examples might be more bene\ufb01cial for stimulating ideation , par - ticularly in terms of novelty of concepts generated . This \ufb01nding is in accord with some work in the domain of artistic creativity , where it has been shown that copying novel artworks has a posi - tive effect on the ability of art students to \ufb02exibly re - interpret art - work and increases the novelty of the artworks produced [ 46 ] . While distance and commonness had some similar effects on ideation processes and products , our \ufb01ne - grained analytic approach suggests some potentially important distinctions . The critical contrast seems to be with respect to effects on quantity and breadth of ideation . Far - \ufb01eld examples increased novelty of solutions and variability in solution quality , but appeared to do so via solution transfer , and resulted in decreased quantity ; in con - trast , less - common examples also increased novelty and quality variability , but appeared to do so via broadening the search space and increasing quantity . One way to interpret this contrast is that example distance and commonness have different mechanisms of inspiration . Based on the results , one could hypothesize that far - \ufb01eld examples inspire designers by moving them into one or two novel regions of the design space ( high solution transfer , high novelty ) , which they then explore in more depth ( low quantity , no bene\ufb01ts on breadth ) ; in contrast , one could hypothesize that less - common examples inspire designers by moving them into multi - ple different regions of the design space via re - interpretation of design functions and features ( low solution transfer , high breadth , and quantity ) . 6 . 2 Optimal Representation Modality of Examples . With regard to the outcome measures of novelty and quality of solution concepts , we found that the representation modality of examples did not change the effects of the distance and commonness factors on ideation . However , we did \ufb01nd evidence for a negative effect of text representations on overall quantity of ideation relative to picture or no examples . Similar to the effects of distance on quan - tity , this suppression effect of text representations can be inter - preted in terms of initial processing costs : when we analyzed only the last phase of ideation , the effect of modality was weaker ( pic - tures vs . text , d \u00bc 0 . 32 ; pictures vs . control , d \u00bc 0 . 45 ) and no lon - ger statistically signi\ufb01cant ( p \u00bc 0 . 07 ) . As an ancient proverb puts it , one picture may be worth 10 , 000 words with respect to convey - ing design concepts . 6 . 3 Caveats . The current work comes with a number of cav - eats . First , we have examined only one design problem . Although a real design problem of some complexity , examples may have different effects on more complex design problems . Second , we examined the effects of particular examples rather than a range of examples sampled multiple times from a class of examples . This experimental design choice made it more feasible to analyze solu - tion transfer but raises possibilities of effects being caused by odd examples or example descriptions . To reduce this threat , we had two examples per condition , and the factorial design of the study permits for multiple replications of main effects . Third , our partic - ipants were senior - level engineering students , for the most part , rather than expert designers , and there is some research to suggest that novices have more dif\ufb01culty with analogical mappings [ 5 , 47 ] . However , design teams sometimes include less experienced designers . Finally , our study focused only on the earliest ideation phase , and future work will have to examine the effects of exam - ples on downstream , and in particular \ufb01nished , solutions . This restriction was most salient in the analyses of quality in that many of the ideas were not feasible or not \ufb02eshed out suf\ufb01ciently to determine feasibility . However , a number of studies point to early ideation as a key moment for intervention to generate innovative designs [ 3 , 4 ] . 6 . 4 Practical Implications and Future Work . The overall focus of this study was on whether particular kinds of examples are more helpful than others for stimulating ideation . However , with the inclusion of a control group , which received no exam - ples , we were able to answer a separate but related question : all things considered , does analogizing over examples confer bene\ufb01ts over and above ideating without examples ? In other words , is design - by - analogy worth the extra time and effort ? Our \ufb01ndings suggest that if the goal of conceptual ideation is to ultimately gen - erate and develop a concept that is high quality and novel , then the answer is yes . There are also implications for the design of tools and methods to support design - by - analogy . As noted in the introduction , a range of previous design - by - analogy methods have been devel - oped ; of particular interest is the development of computational tools that automate the search for analogies [ 48 ] . It is well known in the psychological literature that retrieving far - \ufb01eld analogies is cognitively dif\ufb01cult ; remindings tend to be signi\ufb01cantly con - strained by surface similarity [ 49 ] , reducing the probability of retrieving potentially relevant surface dissimilar analogies . Thus , computational tools that are able to de\ufb01ne and compute functional 081004 - 10 / Vol . 133 , AUGUST 2011 Transactions of the ASME Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm and surface similarity between items in a design space in a prin - cipled manner relative to the current design problem would hold excellent potential as aids for inspiration . These tools might be able to maximize the potential bene\ufb01ts of analogies by retrieving and delivering to the designer in a timely manner surface dissimi - lar analogies and potentially ( as our \ufb01ndings suggest ) even analo - gies that do not necessarily provide direct solutions to the target problem . Additionally , if these systems are able to give priority to analogies that are relatively unusual or infrequently encountered , the potential for inspiration might be even higher . Currently , the state of the art for computational design - by - anal - ogy tools has not reached the point of being able to provide \ufb02exi - ble and real - time support in this manner . The present work provides an impetus for investment into this important research area , as the potential bene\ufb01ts to engineering practice and to soci - ety via increased innovation is high Acknowledgment This work is supported by grants from the National Science Foundation , Grant Nos . CMMI - 0855326 , CMMI - 0855510 , and CMMI - 0855293 , and , in part , by the University of Texas at Austin Cockrell School of Engineering and the Cullen Trust Endowed Professorship in Engineering No . 1 . Any opinions , \ufb01ndings , or recommendations are those of the authors and do not necessarily re\ufb02ect the views of the sponsors . References [ 1 ] Pisano , G . , and Shih , W . , 2009 , \u201cRestoring American Competitiveness , \u201d Harv . Bus . Rev . , 87 , pp . 114 \u2013 125 . [ 2 ] National Academy of Engineering , 2005 , Engineering Research and America\u2019s , Future : Meeting the Challenges of a Global Economy , National Academies Press , Washington , DC . [ 3 ] Vogel , C . M . , Cagan , J . , and Boatwright , P . , 2005 , The Design of Things to Come , Wharton School Pub , Upper Saddle River , NJ . [ 4 ] Terwiesch , C . , and Ulrich , K . T . , 2009 , Innovation Tournaments , Harvard Busi - ness School Pub . [ 5 ] Casakin , H . , and Goldschmidt , G . , 1999 , \u201cExpertise and the use of visual anal - ogy : Implications for design education , \u201d Des . Stud . , 20 ( 2 ) , pp . 153 \u2013 175 . [ 6 ] Goel , A . , 1997 , \u201cDesign , Analogy and Creativity , \u201d IEEE Expert , 12 ( 3 ) , pp . 62 \u2013 70 . [ 7 ] Christensen , B . T . , and Schunn , C . D . , 2007 , \u201cThe Relationship of Analogical Distance to Analogical Function and Pre - Inventive Structure : The Case of Engi - neering Design . Mem . Cognit . , 35 ( 1 ) , pp . 29 \u2013 38 . [ 8 ] Linsey , J . , Murphy , J . , Laux , J . , Markman , A . , and Wood , K . L . , 2009 , \u201cSupporting Innovation by Promoting Analogical Reasoning , \u201d Tools of Innova - tion , Oxford University Press , New York , NY . [ 9 ] Gentner , D . , 1983 , \u201cStructure - Mapping : A Theoretical Framework for Ana - logy , \u201d Cogn . Sci . , 7 , pp . 155 \u2013 170 . [ 10 ] Linsey , J . , Murphy , J . , Markman , A . , Wood , K . L . , and Kortoglu , T . , 2006 , \u201cRepresenting Analogies : Increasing the Probability of Innovation , \u201d ASME International Design Theory and Method Conference , Philadelphia , PA . [ 11 ] Linsey , J . , Wood , K . L . , and Markman , A . , 2008 , \u201cModality and Representation in Analogy , \u201d Artif . Intell . Eng . Des . Anal . Manuf . ( Special Issue on Multi - modal Design ) , 22 ( 2 ) , pp . 85 \u2013 100 . [ 12 ] Gordon , W . J . J . , 1961 , Synectics : The Development of Creative Capacity , Harper and Brothers , New York . [ 13 ] French , M . , 1988 , Invention and Evolution : Design in Nature and Engineering , Cambridge University Press , Cambridge , UK . [ 14 ] Hacco , E . , and Shu , L . H . , 2002 , \u201cBiomimetic Concept Generation Applied to Design for Remanufacture , \u201d 2002 ASME Design Engineering Technology Con - ference and Company and Information in Engineering Conference , Montreal , Quebec , Canada . [ 15 ] McAdams , D . A . , and Wood , K . L . , 2000 , \u201cQuantitative Measures for Design by Analogy , \u201d , DETC\u2019 00 , 2000 ASME Design Engineering Technology Confer - ence Baltimore , Maryland . [ 16 ] Hirtz , J . , Stone , R . B . , and McAdams , D . A . , 2002 , \u201cA Functional Basis for En - gineering Design : Reconciling and Evolving Previous Efforts , \u201d Res . Eng . Des . , 13 , pp . 65 \u2013 82 . [ 17 ] Gentner , D . , and Markman , A . B . , 1997 , \u201cStructure Mapping in Analogy and Similarity , \u201d Am . Psychol . , 52 , pp . 45 \u2013 56 . [ 18 ] Dahl , D . W . , and Moreau , P . , 2002 , \u201cThe In\ufb02uence and Value of Analogical Tthinking During New Product Ideation , \u201d J . Mark . Res . , 39 ( 1 ) , pp . 47 \u2013 60 . [ 19 ] Wilson , J . O . , Rosen , D . , Nelson , B . A . , and Yen , J . , 2010 , \u201cThe Effects of Bio - logical Examples in Idea Generation , \u201d Des . Stud . , 31 , pp . 169 \u2013 186 . [ 20 ] Gick , M . L . , and Holyoak , K . J . , 1980 , \u201cAnalogical Problem Solving , \u201d Cogn . Psychol . , 12 ( 3 ) , pp . 306 \u2013 355 . [ 21 ] Dunbar , K . , 1997 , \u201cHow Scientists Think : On - Line Creativity and Conceptual Change in Science , \u201d Creative thought : An investigation of conceptual struc - tures and processes , T . B . Ward , S . M . Smith , and J . Vaid , ed . , Amer . Psych . Assoc . , Washington , DC , pp . 461 \u2013 493 . [ 22 ] Weisberg , R . W . , 2009 , \u201cOn \u201cOut - of - the - Box , \u201d Thinking in Creativity , \u201d Tools for Innovation A . B . Markman and K . L . Wood , eds . , Oxford University Press , New York . [ 23 ] Purcell , A . T . , and Gero , J . S . 1992 , \u201cEffects of Examples on the Results of a Design Activity , \u201d Knowledge - Based Syst . , 5 ( 1 ) , pp . 82 \u2013 91 . [ 24 ] Duncker , K . , 1945 , On Problem Solving , Amer . Psych . Assoc . Washington , DC . [ 25 ] Adamson , R . E . , 1952 , \u201cFunctional Fixedness as Related to Problem Solving : A Repetition of Three Experiments , \u201d J . Exp . Psychol . , 44 ( 4 ) , pp . 288 \u2013 291 . [ 26 ] Maier , N . R . F . , 1931 , \u201cReasoning in Humans . II . The Solution of a Prob - lem and Its Appearance in Consciousness , \u201d J . Comp . Psychol . , 12 , pp . 181 \u2013 194 . [ 27 ] Chase , W . G . , and Simon , H . A . , 1973 , \u201cThe Mind\u2019s Eye in Chess , \u201d Visual In - formation Processing , W . G . Chase , ed . , Academic Press , New York . [ 28 ] Chi , M . T . H . , Feltovich , P . J . , and Glaser , R . , 1981 , \u201cCategorization and Repre - sentation of Physics Problems by Experts and Novices , \u201d Cogn . Sci . , 5 , pp . 121 \u2013 152 . [ 29 ] Chi , M . T . H . , and Koeske , R . D . , 1983 , \u201cNetwork Representation of a Child\u2019s Dinosaur Knowledge , \u201d Dev . Psychol . , 19 ( 1 ) , pp . 29 \u2013 39 . [ 30 ] Newell , A . , 1990 , Uni\ufb01ed Theories of Cognition , Harvard University Press , Cambridge , MA . [ 31 ] Anderson , J . R . , and Schunn , C . D . , 2000 , \u201cImplications of the ACT - R Learning Theory : No Magic Bullets , \u201d Advances in Instructional Psychology , R . Glaser , ed . , Lawrence Erlbaum , Mahwah , NJ . [ 32 ] Kaplan , C . A . , and Simon , H . A . , 1990 , \u201cIn Search of Insight , \u201d Cogn . Psychol . , 22 , pp . 374 \u2013 419 . [ 33 ] Ohlsson , S . , 1992 , \u201cInformation - Processing Explanations of Insight and Related Phenomena , \u201d Advances in the Psychology of Thinking , Vol . 1 , M . T . Keane and K . J . Gilhooly , eds . , Harvester Wheatsheaf , Hertfordshire , UK . [ 34 ] Knoblich , G . , Ohlsson , S . , Haider , H . , and Rhenius , D . , 1999 , \u201cConstraint Relaxation and Chunk Decomposition in Insight Problem Solving , \u201d J . Exp . Psych . Learn . Mem . Cogn . , 25 ( 6 ) , pp . 1534 \u2013 1555 . [ 35 ] McKoy , F . L . , Vargas - Hernandez , N . , Summers , J . D . , and Shah , J . J . , 2001 , \u201cIn\ufb02uence of Design Representation on Effectiveness of Idea Generation , \u201d DETC \u201801 : ASME 2001 Des . Eng . Tech . Conf . and Comp . and Inf . In Eng . Conf . , Pittsburgh , PA . [ 36 ] Green , M . , Dutson , A . , Wood , K . L . , Stone , R . , and McAdams , D . , 2002 , \u201cIntegrating Service - Oriented Design Projects in the Engineering Curriculum , \u201d Proceedings of the 2002 American Society for Engineering Education Annual Conference and Exposition . [ 37 ] Green , M . , and Wood , K . L . , 2004 , \u201cService - Learning Approaches to Interna - tional Humanitarian Design Projects : Assessment of Spiritual Impact , \u201d Pro - ceedings of the 2004 Christian Engineering Education Conference . [ 38 ] White , C . , and Wood , K . L . , 2010 , \u201cIn\ufb02uences and Interests in Humanitarian Engineering , \u201d Proceedings of the ASEE Annual Conference , Lexington , KY , June 2010 , AC 2010 - 652 ; Proceedings of the Global Colloquium on Engineer - ing Education , Singapore , October 2010 . [ 39 ] Moss , J . , Kotovsky , K . , and Cagan , J . , 2007 , \u201cThe In\ufb02uence of Open Goals on the Acquisition of Problem Relevant Information , \u201d J . Exp . Psych . Learn . Mem . Cogn . 33 ( 5 ) , pp . 876 \u2013 891 . [ 40 ] Tseng , I . , Moss , J . , Cagan , J . , and Kotovsky , K . , 2008 , \u201cThe Role of Timing and Analogical Similarity in the Stimulation of Idea Generation in Design , \u201d Des . Stud . , 29 , pp . 203 \u2013 221 . [ 41 ] Markman , A . B . , and Wood , K . L . , eds . , 2009 , Tools for Innovation : The Sci - ence Behind Practical Methods That Drive New Ideas , Oxford University Press , New York . [ 42 ] Boden , M . A . , 2004 , The Creative Mind : Myths and Mechanisms , 2nd ed . , Routledge , London . [ 43 ] Shah , J . J . , Vargas - Hernandez , N . , and Smith , S . M . , 2003 , \u201cMetrics for Meas - uring Ideation Effectiveness , \u201d Des . Stud . , 24 , pp . 111 \u2013 134 . [ 44 ] Girotra , K . , Terwiesch , C . , and Ulrich , K . T . , 2010 , \u201cIdea Generation and the Quality of the Best Idea , \u201d Manage . Sci . , 56 ( 4 ) , pp . 591 \u2013 605 . [ 45 ] Blanchette , I . , and Dunbar , K . , 2000 , \u201cHow Analogies are Generated : The Roles of Structural and Super\ufb01cial Similarity , \u201d Mem . Cognit . , 28 ( 1 ) , pp . 108 \u2013 124 . [ 46 ] Ishibashi , K . , and Okada , T . , 2006 , \u201cExploring the Effect of Copying Incompre - hensible Exemplars on Creative Drawings , \u201d R . Sun , ed . , Proceedings 28th Ann . Conf . Cog . Sci . Society . Vancouver , Canada . [ 47 ] Novick , L . R . , 1988 , \u201cAnalogical Transfer , Problem Similarity , and Expertise , \u201d J . Exp . Psych . Learn . Mem . Cogn . , 14 ( 3 ) , pp . 510 \u2013 520 . [ 48 ] Chakrabarti , A . , Sarkar , P . , Leelavathamma , B . , and Nataraju , B . S . , 2005 , \u201cA Functional Representation for Biomimetic and Arti\ufb01cial Inspiration of New Ideas , \u201d AIEDAM , 19 , pp . 113 \u2013 132 . [ 49 ] Forbus , K . D . , Gentner , D . , and Law , K . , 1994 , \u201cMAC / FAC : A Model of Similarity - Based Retrieval , \u201d Cogn . Sci . , 19 , pp . 141 \u2013 205 . Journal of Mechanical Design AUGUST 2011 , Vol . 133 / 081004 - 11 Downloaded 03 Aug 2011 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm",
    "chanSituativeCreativityLarger2016": "29 docs . lib . purdue . edu / jps 2016 | Volume 9 Journal of Problem Solving Special iSSue Situative Creativity : Larger Physical Spaces Facilitate Thinking of Novel Uses for Everyday Objects Joel Chan 1 and Timothy J . Nokes - Malach 2 1 Human - Computer Interaction Institute , Carnegie Mellon University , 2 Learning Research and Development Center , University of Pittsburgh People often use spatial metaphors ( e . g . , think \u201claterally , \u201d \u201coutside the box\u201d ) to describe explo - ration of the problem space during creative problem solving . In this paper , we probe the potential cognitive underpinnings of these spatial metaphors . Drawing on theories of situ - ative cognition , semantic foraging theory , and environmental psychology , we formulate and test the hypothesis that larger physical spaces can facilitate divergent ( but not convergent ) processes in problem space exploration . Across two experiments , participants worked on a battery of problem solving tasks intended to represent divergent ( alternative uses , shape invention ) and convergent ( remote associates , letter extrapolation ) problem solving processes in either a large or a small room . In Experiment 1 , participants in the larger room produced more novel alternative uses for everyday objects , and created more novel shape inventions , but generated less practical alternative uses , than participants in the smaller room . In Experi - ment 2 , participants in the larger room ( including a variant larger room ) also produced more novel alternative uses for everyday objects , and less practical alternative uses , than partici - pants in a small room , but did not create more novel shape inventions . These results sug - gest that spatial metaphors for problem space exploration may reflect meaningful cognitive phenomena : People may be able to search more broadly in a problem space if they are in an environment where broad physical search is a salient affordance ; however , this effect appears to be relatively small and may depend on having sufficiently motivated participants . Correspondence : Correspondence concerning this article should be addressed to Joel Chan at the Human - Computer Interaction Institute , Carnegie Mellon University , 2504B Newell - Simon Hall , 5000 Forbes Ave , Pittsburgh , PA 15213 , via email to joelchuc @ cs . cmu . edu , or via telephone to ( 479 ) 647 - 0575 . Keywords : situative cognition , creative problem solving , metaphor Acknowledgments : We thank Chris Wiltrout , Emily Schmidt , Courtney Stein , and Jeremy Addison for help with data collection and coding , and Efosa Osazuwa and Tina Liu for help with data coding . We also thank the edi - tor and our two anonymous referees for very helpful comments on earlier drafts of the manuscript . INTRODUCTION A key component of creative problem solving is exploration of the problem space . The problem space is typically described as the mental representation of the problem , including the initial problem description , goal , and operators ( i . e . , strate - gies ) to move from the initial state to the goal state ( Newell & Simon , 1972 ) . Theories of creative problem solving posit that the effective initial exploration of the problem space\u2014some - times called \u201cdivergence\u201d or \u201cdivergent thinking\u201d ( Guilford , 1956 ) \u2014is critical to produce a successful solution ( Amabile , 1983 ; Finke , Ward , & Smith , 1996 ; Sawyer , 2012 ; Simonton , 2011 ; Wallas , 1926 ; Warr & O\u2019Neill , 2005 ) . Effective problem space exploration can be supported by considering many dif - ferent solution approaches ( Ad\u00e1nez , 2005 ; Chan et al . , 2011 ; Parnes & Meadow , 1959 ; Shah , Millsap , Woodward , & Smith , 2012 ; Torrance , 1988 ) , increasing the variance in the quality of solutions considered ( e . g . , being willing to consider \u201cwild\u201d ideas ; Chan et al . , 2011 ; Girotra , Terwiesch , & Ulrich , 2010 ; Terwiesch & Ulrich , 2009 ) , considering solutions and perspec - tives from outside one\u2019s discipline or problem domain ( Chan et al . , 2011 ; Gentner & Markman , 1997 ; Ward , 1998 ) , relaxing inferred constraints about the problem description ( Knoblich , Ohlsson , Haider , & Rhenius , 1999 ; Ohlsson , 1992 ) , and explor - ing alternative conceptualizations of the problem ( Kaplan & Simon , 1990 ; MacGregor & Cunningham , 2009 ) . Divergence can also be facilitated by modulation of attention : For example , a reduction of attentional control or focus has been identified as a key mechanism for achieving divergent thinking and mak - ing remote associations in creative problem solving ( Aiello , Jarosz , Cushen , & Wiley , 2012 ; Haarmann , George , Smaliy , & Dien , 2012 ; Martindale , 1997 ; Wiley & Jarosz , 2012 ) . Successful exploration is often described with spatial lan - guage and imagery . For example , people commonly encour - age one another to think \u201claterally , \u201d not \u201cvertically\u201d ( Bono , 1970 ) , and \u201coutside the box , \u201d or to explore \u201cbroadly\u201d ( Wiley , 1998 ) and make \u201cremote\u201d associations in semantic memory ( Mednick , 1962 ) . In this paper , our goal is to probe the poten - tial cognitive underpinnings of these spatial metaphors . Are these metaphors arbitrary , or merely artifacts of human convention ? Or do they identify real cognitive phenomena ? Could embodying variations in these spatial metaphors ( e . g . , large vs . small physical environments ) influence the nature of people\u2019s search patterns in semantic space ? The present investigation is inspired by a growing body of literature across a diverse range of tasks that suggests that http : / / dx . doi . org / 10 . 7771 / 1932 - 6246 . 1184 docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 30 people\u2019s embodied physical context can have significant impli - cations for information processing . For example , people per - ceive slopes as steeper if they wear a heavy vs . a light backpack ( Bhalla & Proffitt , 1999 ) , can transfer knowledge and skills across contexts when there is high interconnectedness across activities and practices in those contexts ( Brown , Collins , & Duguid , 1989 ) , and make more \u201cholistic\u201d decisions ( i . e . , inte - grating multiple sources of data and abstraction ) in rooms with higher vs . lower ceilings ( Meyers - Levy & Zhu , 2007 ) . The literature provides two potential theoretical motivations for suspecting that the spatial metaphors of creative search have grounding in cognitive phenomena . The first account , which we call \u201cdirect priming , \u201d is exemplified by Hills and col - leagues\u2019 argument that goal - directed search for resources in external spaces and search for resources in internal spaces ( e . g . , semantic memory ) share a common neural substrate ( Hills , 2006 ; Hills , Jones , & Todd , 2012 ; Hills , Todd , & Goldstone , 2008 ) . Specifically , that they share dopaminergic modulation of area - restricted search such that search is narrow in situa - tions where the target resources have been frequently found in the past and search becomes broad in situations where the target resources are encountered less frequently . One intrigu - ing implication of this argument is that expectations about the structure of search environments in external spaces can shape search patterns in internal spaces , or vice versa . To test this implication , Hills et al . ( 2008 ) studied how search patterns on an anagram task ( i . e . , search for as many words as possible for a given letter set ) might be shaped by prior experience with a spatial foraging task ( i . e . , search for high - value pixels in a simple 2 - D computer maze ) . In their experiment , participants completed spatial foraging tasks with either \u201cclumpy\u201d\u2014many pixels concentrated in a few patches\u2014or \u201cdiffuse\u201d\u2014pixels evenly distributed in the envi - ronment\u2014resource distributions , and subsequently tried to find as many anagrams as they could . They found that par - ticipants who had just experienced a \u201cclumpy\u201d distribution of pixels took longer to switch between letter sets when search - ing for anagrams , consistent with expectations for a \u201cclumpy\u201d distribution of anagrams ( i . e . , expecting letter sets to contain more anagrams . They inferred from this that the distribution of resources in the spatial environment primed expectations for the distribution of \u201cresources\u201d in the semantic space . This analysis suggests that the affordances in the external , physical environment ( e . g . , the distribution of resources ) may shape the mind\u2019s internal search in semantic space . Following this line of thought , we reason that , to the extent that large physical environments afford free movement and exploration , they may also better facilitate divergent problem solving ( i . e . , exploration of semantic space ) relative to smaller , constrained spaces . Rather than simply fostering increased performance via increased effort , people might be sensitive to how larger physical spaces afford freer exploration , and consequently adopt a semantic search strategy that better matches this resource distribution , such as by relaxing their focus of attention from more clearly relevant or high - quality responses to more semantically distant and varied ( and likely more novel ) associations ( Aiello et al . , 2012 ; Haarmann et al . , 2012 ; Martindale , 1997 ; Wiley & Jarosz , 2012 ) . This direct priming of attentional focus might occur without conscious awareness , similar to some varieties of top - down modulation of visual attention ( Koch & Tsuchiya , 2007 ) . This direct prim - ing mechanism can also be related to the notion of \u201cframes\u201d in research on situative cognition , for example , expectations ( whether explicit or tacit ) about a given situation that are influ - enced by the affordances and constraints of particular environ - ments , and go to shape cognition and interaction ( Goffman , 1974 ; Greeno , 1994 ; Greeno & Middle - School Mathemat - ics through Applications Project Group , 1998 ; Maclachlan & Reid , 1994 ; Nokes - Malach & Mestre , 2013 ; Scherr & Hammer , 2009 ) . Because both varieties of direct priming mechanisms can occur without conscious awareness , we do not expect facilitation of divergent performance to be associated with more effortful performance . Indeed , to the extent that people relax their focus of attention to search more broadly , we might even expect to see a decreased perception of task difficulty as measured by cognitive load ( e . g . , Antonenko , Paas , Grabner , & Gog , 2010 ; Chandler & Sweller , 1991 ) . The second line of reasoning , which we call the \u201cconcept activation\u201d account , comes from research in environmental psychology that explores how certain configurations of physical environments can prime certain psychological states or ideas , which can then influence later information processing . For example , Hall ( 1966 ) argues that small and contained spaces ( e . g . , chapels ) can evoke notions of confinement or restricted - ness , while larger spaces ( e . g . , cathedrals ) can prime notions of freedom and openness . Similarly , Moore , Lane , Hill , Cohen , and McGinty ( 1994 ) suggest that lower ceilings may invoke more restricted play , while higher ceilings may encourage \u201cfreer\u201d play . In the Meyers - Levy and Zhu ( 2007 ) study men - tioned previously , the effect of the ceiling height manipulation on decision making was mediated by activation of the concept of \u201cfreedom\u201d vs . \u201cconfinement . \u201d This line of reasoning presents an indirect mechanism by which larger spaces prime concepts of \u201cfreedom\u201d and \u201cbroadness , \u201d which in turn induces infor - mation processing that is also \u201cless constrained , \u201d for example , more holistic , as in Meyers - Levy & Zhu ( 2007 ) , thereby facili - tating divergent processing during problem solving . In contrast to the direct priming account , concept activa - tion may also be marked by affective changes ( e . g . , increases in positive affect , decreases in negative affect ) , since concepts related to \u201cfreedom\u201d may have positive valence , while con - cepts related to \u201cconfinement\u201d may have negative valence . For example , a recent affective norming project found that the word \u201cfreedom\u201d had a highly positive valence score of docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 31 7 . 72 on a 1 to 9 valence scale ( 1 is highly negative , 9 is highly positive ) , while the word \u201crestrict\u201d had a much more nega - tive valence of 3 . 48 out of 9 ( Warriner , Kuperman , & Brys - baert , 2013 ) . Therefore , measuring changes in affect may be a way to distinguish between direct priming or concept activation accounts of potential associations between physi - cal surroundings and divergent / convergent problem solving processes : Increased divergent performance in large physi - cal spaces accompanied by increases in positive affect ( and decreases in negative affect ) would be more consistent with a concept activation account of the cognitive basis of spatial imagery for divergent exploration . Synthesizing these ideas , we test the hypothesis that larger spaces will have a facilitation effect on divergent problem solving processes ( i . e . , processes that have similar cognitive characteristics to the exploration stage of the creative pro - cess ) , but not \u201cconvergent\u201d problem solving processes ( i . e . , processes that focus on \u201cconverging\u201d on a single \u201ccorrect\u201d or canonical answer ) . To the extent that increased divergence may be at odds with convergent processes ( Goldenberg , Lar - son , & Wiley , 2013 ) , larger spaces might also hinder con - vergent problem solving . We further hypothesize that this facilitation would be accompanied by decreases in perceived task difficulty . Measures of affect might help distinguish between the direct priming and concept activation expla - nations of observed effects . In summary , in this paper , we examine the following three main hypotheses . H1 : Divergent performance will be higher in larger vs . smaller physical spaces . H2 : Convergent performance will be lower in larger vs . smaller physical spaces . H3 : Perceived task difficulty for divergent tasks will be lower in larger vs . smaller physical spaces . We conducted two experiments to test these hypotheses , first with a sample of paid volunteers , and then with a larger sample of psychology subject pool participants and an expan - sion of the range of physical spaces and problem solving stimuli that are tested . To preview our results , we find partial support for the first two hypotheses across both experiments , and find that these effects are not associated with changes in affect . EXPERIMENT 1 In this study , we provide a first test of the three hypotheses . The basic experimental approach is to have participants work on a battery of problem solving tasks intended to represent both divergent and convergent processing in either large or small rooms . As noted , our hypotheses are that divergent problem solving performance will be facilitated by being in a large ( vs . small ) room , while convergent performance will be hindered by being in the large room . METhODs Participants Forty - seven people ( 20 males , 27 females ; ages 19 \u2013 66 , average age 27 ) from the community at a large research university in the northeastern United States participated in this study . Thirty - five of the participants were undergraduate or graduate students at the university . Most of the other participants were recent gradu - ates or employees of the university or businesses on campus . All participants were recruited through fliers posted around cam - pus and were compensated 10 dollars for their time . Four participants didn\u2019t produce valid data on one of the problem solving tasks ( 3 did not produce any valid inven - tions , and 1 did not produce any valid uses ) , and were there - fore dropped from our analyses : There were two each from the large and small rooms respectively . 1 Therefore , our final dataset consisted of 43 participants . Materials Room size manipulation . To manipulate room size , we had partici - pants complete their problem solving tasks in one of two rooms on campus . The \u201clarge\u201d room was a conference auditorium ( see Fig 1 , left panel ) . The dimensions of the room were approximately 15\u2019 W \u00d7 30\u2019 L \u00d7 15\u2019 H . Participants completed their tasks on a desk in the front of the auditorium facing toward the audience seats so that the size of the room would be salient . Other than the desk and chair the participants used , and the other chairs facing the front of the auditorium , the auditorium was empty . The \u201csmall\u201d room was a former office space that was emp - tied out for the experiment ( Fig 1 : next page , right panel ) . The dimensions of the room were approximately 8\u2019 W \u00d7 10\u2019 L \u00d7 8\u2019 H . Participants completed their tasks on a desk facing one of the walls . It was empty except for the desk and chair the participants used . Other than the size of the room , we made sure that the two rooms were similar in a number of important ways , including amount of stimuli encountered on the walk to the room ( both rooms were in the same building ) , ambient noise ( we chose rooms that were far from other offices in the build - ing ) , and temperature ( both rooms shared the same central air conditioning system ) . The one potentially salient differ - ence was the tone of lighting : the large room used incandes - cent lighting , while the small room used fluorescent lighting . Problem - solving tasks Participants completed a battery of four problem - solving tasks intended to represent both divergent and convergent processing : 1 ) an alternative uses task , 2 ) a shape invention task , 3 ) a version of the Remote Associates Test ( RAT ) , and 4 ) a letter series extrapolation task . In this section we describe each task along with the hypothesized processes involved . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 32 Alternative uses . The alternative uses task is patterned after Guilford\u2019s ( 1967 ) classic alternate uses task in which the problem solver is asked to list as many uses as possible for a common object ( e . g . , think of as many uses as you can for a brick ) . It has been hypothesized to measure divergent thinking processes because the output of the task is a range of responses rather than one correct or incorrect response . Task output is typically measured in terms of the fluency and flexibility ( e . g . , novelty ) of the responses . However , this task may also involve convergent processes . People may initially think of a variety of responses , and subsequently evaluate and select only the uses that are both novel and practical . This corresponds to the selection / evaluation processes / phases in various theories of cognitive and creative production , such as the convergent production component of Guilford\u2019s ( 1956 ) \u201cstructure of intellect\u201d theory , the response validation stage of Amabile\u2019s ( 1983 ) process model of creativity , and the Explore phase of Finke and colleagues\u2019 ( 1996 ) Geneplore model . Therefore , we hypothesize that fluency and novelty are mea - sures of divergent thinking on this task , whereas practical - ity is a measure of convergent thinking . It is useful to note that convergence may not necessarily always follow diver - gence : Convergence can also reflect attentional focus on and rapid selection of the most readily accessible responses , which are often the most successful or appropriate ( Bilali\u0107 , McLeod , & Gobet , 2008 ; Guilford , 1956 ; Luchins , 1942 ) . We used \u201cSHOE\u201d and \u201cNEWSPAPER\u201d as our common object items and gave the following instructions to partici - pants : \u201cIn this part of the experiment , your task is to list as many uses as you can for an object ( named below ) . For exam - ple , if the object is \u2018BRICK , \u2019 you could say \u2018building material , doorstop , anchor , etc . \u2019 The goal is to come up with as many uses of an object as possible . There are 2 of these problems , and you will have 4 minutes for each . \u201d Shape invention . In the shape invention task ( Finke et al . , 1996 ) , the problem solver is given three three - dimensional shapes to combine together to create as many useful objects that belong to one of three given categories ( e . g . , toys and games , transportation ) . Similar to the alternative uses task , we hypothesize that this task includes elements of both diver - gent and convergent processing . Again , we hypothesize that Figure 1 Picture of large room ( left ) and small room ( right ) in Experiment 1 . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 33 the novelty of the items generated would reflect divergent thinking processes , whereas the practicality of the objects would reflect convergent processes . Some versions of this task allow the problem solver to choose the shapes and / or the object categories ; we elected to randomly select both shapes and categories and present them as givens for all participants . Part of the motivation for this was that Finke and colleagues ( 1996 ) found that this con - dition stimulated the most creative responses , and we wanted to give our participants the best chance of displaying creative performance . Finke and colleagues\u2019 original set of shapes includes 15 different shapes , ranging from spheres to cylin - ders , to wires , wheels , and flat squares ; each shape belongs to a subcategory set of 5 shapes , ordered by how difficult it is ( normed from their studies ) to incorporate into an invented object ( easy , medium , and hard ) . We randomly sampled 3 shapes from this list , with \u201ceasy\u201d shapes having a probability = 0 . 10 of being selected , \u201cmedium\u201d shapes having a probability = . 07 of being selected , and \u201chard\u201d shapes having a probabil - ity = 0 . 03 of being selected . Finke and colleagues\u2019 original set of object categories consisted of 8 categories : weapons , toys and games , appliances , transportation , scientific instru - ments , tools and utensils , furniture , and personal items . We randomly selected three categories from this list to give to participants . We ended up with the sphere , tube , and cone objects ( see Figure 2 ) , and the following object categories : tools and utensils , toys and games , and personal items . We gave the following instructions to participants : \u201cIn this part of the experiment , your task is to try to use the follow - ing \u2018parts\u2019 to \u2018construct\u2019 as many useful objects as you can . These objects can be existing things or things you invent . You will have 8 minutes to do this . The rules for using the parts to construct objects are as follows : 1 ) you are allowed to vary the size , position , or orientation of any part , but you may not bend or deform the parts ( except the tube ) , 2 ) the parts can be put inside one another , 3 ) you decide if the parts are hollow or solid , and 4 ) you decide what material the parts are made of\u2014 they can be made of any material , including wood , metal , plas - tic , rubber , or glass , or any combination of these materials . \u201d RAT . To complete the RAT ( Bowers , Regehr , Balthazard , & Parker , 1990 ; Mednick , 1962 ; Mednick & Mednick , 1967 ) , solvers must generate a target word that is related to a list of three cue words . For example , a correct response for the cue words Elephant - Lapse - Vivid would be the target word \u201cmemory . \u201d While the RAT is most commonly used as a mea - sure of creativity ( e . g . , Wiley , 1998 ) , and several prior studies have examined the role of divergent processes in RAT per - formance ( Aiello et al . , 2012 ; Haarmann et al . , 2012 ) , we rea - soned that it also heavily taps convergent processes since the final output is a single answer that is compared to a predeter - mined correct answer . We are not the first to treat the RAT in this way : Other recent studies have also studied the RAT as primarily a convergent task in contrast to the alternate uses task ( treated as primarily measuring divergent processes ) , and found both dissociable effects on these tasks from their manipulations and a lack of correlation between performance on these tasks ( Colzato , Ozturk , & Hommel , 2012 ; Oppezzo & Schwartz , 2014 ; Radel , Davranche , Fournier , & Dietrich , 2015 ) . We gave participants 32 items , of varying difficulty , drawn from Mednick and Mednick ( 1967 ) and Bowers et al . ( 1990 ) , chosen to reflect a range of difficulty levels . The full list of items can be seen in the appendix . Letter series . In the letter series task , the problem solver is given a series of letters and is asked to generate additional letters to \u201cN\u201d places to complete the pattern exemplified in the given series . For example , given the series \u201caaabbbccc , \u201d the correct extrapolation to N = 3 places is \u201cddd . \u201d Similar to the RAT , this task requires the identification and genera - tion of a single answer . Although participants may consider several possible patterns when solving the problem , they must eventually converge or decide on one to extrapolate . We gave participants 18 items drawn from prior studies with the letter series task ( Nokes , 2009 ; Simon & Kotovsky , 1963 ) , which were chosen to reflect a range of difficulty levels . The items had initial series ranging from 9 to 16 letters in length ; all items required participants to extrapolate the series to N = 10 places . The full list of items can be seen in the appendix . Dependent measures In the previous section we described each task we used for the experiment . Some of the tasks ( specifically the uses and invention tasks ) were hypothesized to include both divergent and convergent processing . We now describe how we mea - sured divergent and convergent processing across the tasks . Divergent measures . Both the uses and invention tasks were scored for fluency and novelty to yield our primary diver - gent measures . Fluency was defined as the number of uses or inventions generated . Novelty was rated on a scale from 1 ( not at all novel ) to 4 ( extremely novel ) . Examples of low and high novelty uses are \u201cuse SHOE to protect feet\u201d and \u201cuse SHOE as boat for termites\u201d ; examples of low and high Situative Creativity 13 following object categories : tools and utensils , toys and games , and personal items . We gave the following instructions to participants : \u201cIn this part of the experiment , your task is to try to use the following \u2018parts\u2019 to \u2018construct\u2019 as many useful objects as you can . These objects can be existing things or things you invent . You will have 8 minutes to do this . The rules for using the parts to construct objects are as follows : 1 ) you are allowed to vary the size , position , or orientation of any part , but you may not bend or deform the parts ( except the tube ) , 2 ) the parts can be put inside one another , 3 ) you decide if the parts are hollow or solid , and 4 ) you decide what material the parts are made of\u2014they can be made of any material , including wood , metal , plastic , rubber , or glass , or any combination of these materials . \u201d RAT . To complete the RAT ( Bowers , Regehr , Balthazard , & Parker , 1990 ; Mednick , 1962 ; Mednick & Mednick , 1967 ) , solvers must generate a target word that is related to a list of three cue words . For example , a correct response for the cue words Elephant - Lapse - Vivid would be the target word \u201cmemory . \u201d While the RAT is most commonly used as a measure of creativity ( e . g . , Wiley , 1998 ) , and several prior studies have examined the role of divergent processes in RAT performance ( Aiello et al . , 2012 ; Haarmann et al . , 2012 ) , we reasoned that it also heavily taps convergent processes since the final output is a single answer that is compared to a predetermined correct answer . We are not the first to treat the RAT in this way : Other recent studies have also studied the RAT as primarily a convergent task in contrast to the alternate uses task ( treated as primarily measuring divergent processes ) , and found both dissociable effects on Figure 2 . Shapes for invention task in Experiment 1 . Figure 2 Shapes for invention task in Experiment 1 . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 34 novelty inventions are shown in Figure 3 . Two trained coders evaluated the uses , with high inter - rater reliability , ICC ( 2 , 2 ) = . 89 . Three trained coders evaluated the inventions , with high inter - rater reliability , ICC ( 2 , 3 ) = . 85 . Each use / invention\u2019s novelty score was the arithmetic mean of all judges\u2019 scores for that use / invention . Novelty scores were then aggregated into participant - level measures in the following ways : mean nov - elty ( how novel their uses were , on average ) , and max novelty ( what was the highest novelty score they achieved ) . Convergent measures . Both the uses and invention tasks were also scored for practicality , to reflect convergent processing on those tasks . Practicality scoring for the uses task was initially done with a 4 - point scale ( 1 \u2013 unlikely to work at all , 2 \u2013 will work less well than conventional means , 3 \u2013 will work as well as conventional means , 4 \u2013 will work better than conven - tional means ) , but was collapsed to a 3 - point scale because there were almost no ( agreed - upon ) generated uses that war - ranted a 4 . Two trained coders evaluated the uses , with good inter - rater reliability , ICC ( 2 , 2 ) = . 79 . In contrast , scoring for the invention category had slightly higher variance , allowing us to code inventions on a 5 - point scale ( 1 \u2013 extremely bad example of its invention category , to 5 \u2013 exceeds expectations for a good example of its invention category ) . Seven trained coders evaluated the inventions for practicality , with accept - able inter - rater reliability , ICC ( 2 , 7 ) = . 61 . Examples of low and high practicality uses are \u201cuse SHOE as tent stake\u201d and \u201cuse SHOE as slapping device to bring someone back to their senses\u201d ; examples of low and high practicality inventions are shown in Figure 3 . Participant - level practicality measures were created by taking the arithmetic mean of practicality scores achieved to create a mean practicality measure for both uses and invention tasks , separately . Performance on the RAT and letter series tasks were intended to primarily reflect convergent processing , since both tasks sought the production of a single \u201cbest\u201d response . One trained coder scored the RAT responses as either correct or incorrect , using the answer key from Mednick & Mednick ( 1967 ) and Bowers et al . ( 1990 ) . Percent correct was used for analysis . Letters series task performance was measured by marking responses as either correct or incorrect , using canonical answers from the prior references ( Nokes , 2009 ; Simon & Kotovsky , 1963 ) , and the percent correct was used for analysis . Other measures Perceived task difficulty . To measure perceived task difficulty , we adapted two items from prior research with cognitive load ( Jang & Schunn , 2012 ) . The measure was about the task just completed . The first item asked , \u201cHow easy or difficult was this task ? \u201d and participants were asked to answer using a 1 to 9 scale , where 1 was anchored as \u201cVery , very easy , \u201d and 9 was anchored as \u201cVery , very difficult . \u201d The second item asked \u201cHow much mental effort ( e . g . , searching , remembering , thinking , deciding ) did the task take ? \u201d and participants were asked to answer using a 1 to 9 scale , with 1 anchored as \u201cVery , very low mental effort , \u201d and 9 anchored as \u201cVery , very high mental effort . \u201d Positive and negative affect . To measure affect , we used the Positive and Negative Affect Schedule ( PANAS ; Watson , Clark , & Tel - legen , 1988 ) , in which a subject is given 20 words that describe different feelings and emotions , and\u2014using a scale of 1 ( very slightly or not at all ) to 5 ( extremely ) \u2014rates the extent to which she feels that way \u201cright now\u201d ( i . e . , at the present moment ) . Situative Creativity 15 inventions generated . Novelty was rated on a scale from 1 ( not at all novel ) to 4 ( extremely novel ) . Examples of low and high novelty uses are \u201cuse SHOE to protect feet\u201d and \u201cuse SHOE as boat for termites\u201d ; examples of low and high novelty inventions are shown in Figure 3 . Two trained coders evaluated the uses , with high inter - rater reliability , ICC ( 2 , 2 ) = . 89 . Three trained coders evaluated the inventions , with high inter - rater reliability , ICC ( 2 , 3 ) = . 85 . Each use / invention\u2019s novelty score was the arithmetic mean of all judges\u2019 scores for that use / invention . Novelty scores were then aggregated into participant - level measures in the following ways : mean novelty ( how novel their uses were , on average ) , and max novelty ( what was the highest novelty score they achieved ) . Convergent measures . Both the uses and invention tasks were also scored for practicality , to reflect convergent processing on those tasks . Practicality scoring for the uses task was initially done with a 4 - point scale ( 1 \u2013 unlikely to work at all , 2 \u2013 will work less well than conventional means , 3 \u2013 will work as well as conventional means , 4 \u2013 will work better than conventional Figure 3 . Example of low and high novelty / practicality inventions . The low and high novelty inventions are a \u201cfunnel\u201d and a device that \u201cslows elevators with centripetal force . \u201d The low and high practicality inventions are an \u201cunstable martini glass\u201d and a \u201ctool to catch water to measure the rain . \u201d Note that the shapes used for invention are a cone , sphere , and tube . Figure 3 Example of low and high novelty / practicality inventions . The low and high novelty inventions are a \u201cfunnel\u201d and a device that \u201cslows elevators with centripetal force . \u201d The low and high practicality inventions are an \u201cunstable martini glass\u201d and a \u201ctool to catch water to measure the rain . \u201d Note that the shapes used for invention are a cone , sphere , and tube . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 35 Procedure Participants were greeted and brought to either the large or small room , depending on their assignment . They were then informed that they would complete four problem - solving tasks , grouped into two blocks with two problem sets each . The uses and invention tasks formed one block , while the RAT and letter series task formed the other block . The tasks were counterbalanced by block , and specific problem solving tasks within the block ( RAT and letters ; uses and invention ) . They had eight minutes to complete each problem set , and a one - minute warning was given before the time was up . After each problem set , participants completed the untimed cognitive load survey . Before they began the problem sets , participants were asked to complete the first PANAS to get a baseline measure . They were asked to complete a second and third PANAS after the first and second blocks of problem sets , respectively . Overall , the experiment ran no longer than 45 minutes . The experimenter remained in the room during the length of the experiment , seated behind the participant . Design This study had a between - subjects design . The independent variable was room size , with two levels ( large or small ) . Par - ticipants were randomly assigned to conditions . In the final dataset , there were 21 participants in the large room and 22 participants in the small room . REsUlTs Table 1 shows the descriptive statistics for all problem solv - ing measures . Table 2 shows the intercorrelations between participants\u2019 problem solving task performance measures , collapsed across Experiments 1 and 2 . In general , the correla - tions were in the expected directions ( e . g . , significant positive correlations between divergent measures ; significant positive correlations between letter series and RAT measures ; sig - nificant negative correlations between uses practicality and the divergent measures ) . However , the correlations are gen - erally low , explaining small amounts of common variance , and some correlations are missing ( e . g . , no significant cor - relations between practicality and the letters series and RAT measures , and no significant correlations between invention practicality and any of the other measures ) . This suggests that the measures do not necessarily primarily reflect the two constructs of divergence and convergence ( as we had hypoth - esized ) . Therefore , we analyze each measure separately . Alternative uses There was an effect of room size on mean fluency , with par - ticipants in the large room generating more uses ( M = 25 . 0 , SE = 1 . 8 ) than participants in the small room ( M = 19 . 3 , SE = 1 . 8 ) , d = 0 . 68 , 95 % CI = [ 0 . 03 , 1 . 33 ] , F ( 1 , 41 ) = 5 . 0 , p = . 03 . There was also an effect of room size on mean novelty , with participants in the large room producing higher mean novelty with their uses ( M = 1 . 8 , SE = . 09 ) than those in the small room ( M = 1 . 5 , SE = . 08 ) , d = 0 . 78 [ 0 . 13 , 1 . 44 ] , F ( 1 , 41 ) = 6 . 6 , p = . 01 ( Fig . 4 , left panel ) . The results were similar for max novelty : Participants in the large room achieved marginally higher max novelty scores ( M = 3 . 4 , SE = . 19 ) than those in the small room ( M = 2 . 9 , SE = . 18 ) , d = 0 . 61 [ - 0 . 04 , 1 . 26 ] , F ( 1 , 41 ) = 4 . 0 , p = . 05 ( Fig . 4 , middle panel ) . In contrast , participants in the large room generated alter - native uses that were significantly less practical ( M = 2 . 5 , SE = . 04 ) than those from the small room ( M = 2 . 7 , SE = . 04 ) , d = \u2013 1 . 00 [ - 1 . 68 , - 0 . 33 ] , F ( 1 , 41 ) = 10 . 9 , p = . 00 ( Fig . 4 , right panel ) . Shape invention Participants in the large room generated slightly more inven - tions ( M = 5 . 5 , SE = . 42 ) than participants in the small room ( M = 4 . 6 , SE = . 41 ) , d = 0 . 47 [ - 0 . 17 , 1 . 11 ] , but this difference was not statistically significant , F ( 1 , 41 ) = 2 . 3 , p = . 13 . However , there Table 1 Descriptive statistics for Experiment 1 . Situative Creativity 19 the intercorrelations between participants\u2019 problem solving task performance measures , collapsed across Experiments 1 and 2 . In general , the correlations were in the expected directions ( e . g . , significant positive correlations between divergent measures ; significant positive correlations Table 1 Descriptive statistics for experiment 1 . Mean Median Min Max SE Divergent measures Uses fluency 22 . 07 21 7 42 1 . 32 Uses novelty mean 1 . 67 1 . 59 1 2 . 58 0 . 06 Uses novelty max 3 . 12 3 1 4 0 . 13 Invention fluency 5 . 07 5 2 9 0 . 29 Invention novelty mean 2 . 74 2 . 83 1 . 56 3 . 83 0 . 08 Invention novelty max 3 . 58 3 . 67 2 4 0 . 08 Convergent measures Uses practicality 2 . 62 2 . 64 2 . 03 3 0 . 04 Invention practicality 3 . 15 3 . 21 1 . 79 3 . 86 0 . 06 RAT 0 . 59 0 . 62 0 1 0 . 04 Letters 0 . 61 0 . 71 0 1 0 . 05 Table 2 Intercorrelations between measures , collapsed across Experiments 1 and 2 . Divergent Convergent Uses novel mean Uses novel max Invent fluen . Invent novel mean Invent novel max Uses pract . Invent pract . RAT Letters U fluency 0 . 50 * 0 . 50 * 0 . 37 * 0 . 24 * 0 . 26 * - 0 . 43 * - 0 . 07 - 0 . 04 - 0 . 01 U novel mean 0 . 79 * 0 . 31 * 0 . 27 * 0 . 29 * - 0 . 89 * - 0 . 02 0 . 01 - 0 . 04 U novel max 0 . 17 * 0 . 25 * 0 . 22 * - 0 . 66 * 0 . 01 0 . 03 - 0 . 12 I fluency 0 . 09 0 . 46 * - 0 . 32 * 0 . 00 - 0 . 07 - 0 . 06 I novel mean 0 . 75 * - 0 . 23 * - 0 . 11 - 0 . 02 - 0 . 03 I novel max - 0 . 28 * - 0 . 05 - 0 . 01 - 0 . 02 U practicality 0 . 06 - 0 . 06 0 . 00 I practicality 0 . 14 m 0 . 00 RAT 0 . 21 * m p < . 10 ; * p < . 05 docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 36 was an effect of room size on mean novelty , with participants in the large room , on average , creating more novel inventions ( M = 2 . 9 , SE = . 10 ) than participants in the small room ( M = 2 . 6 , SE = . 10 ) , d = 0 . 76 [ 0 . 10 , 1 . 41 ] , F ( 1 , 41 ) = 6 . 2 , p = . 02 . Similarly , there was an effect of max novelty , with the most novel inventions of participants in the large room being , on average , more novel ( M = 3 . 8 , SE = . 10 ) than the most novel inventions of participants in the small room ( M = 3 . 4 , SE = . 10 ) , d = 0 . 76 [ 0 . 10 , 1 . 41 ] , F ( 1 , 41 ) = 6 . 2 , p = . 02 . For practi - cality , there was no effect of room size , with inventions from the large room condition about as practical ( M = 3 . 2 , SE = . 08 ) as those from the small room ( M = 3 . 1 , SE = . 08 ) , d = 0 . 32 [ - 0 . 31 , 0 . 96 ] , F ( 1 , 41 ) = 1 . 1 , p = . 29 . RAT There was no effect of room size , with participants in the large room having about the same mean proportion cor - rect ( M = . 57 , SE = . 05 ) as participants in the small room ( M = . 60 , SE = . 05 ) , d = - 0 . 12 [ - 0 . 75 , 0 . 51 ] , F ( 1 , 41 ) = 0 . 2 , p = . 69 . Letter series Participants in the large room generated a slightly higher proportion of correct responses ( M = . 65 , SE = . 07 ) than participants in the small room ( M = . 57 , SE = . 07 ) , d = 0 . 22 [ - 0 . 41 , 0 . 85 ] , but this difference was not statistically signifi - cant , F ( 1 , 41 ) = . 05 , p = . 48 . Perceived task difficulty There was a marginal effect of room size on perceived dif - ficulty for the uses task , with participants in the large room self - reporting slightly lower levels of difficulty ( M = 4 . 1 , SE = . 36 ) compared to participants in the small room ( M = 5 . 1 , SE = . 36 ) , d = - 0 . 53 [ - 1 . 15 , 0 . 08 ] , F ( 1 , 45 ) = 3 . 4 , p = . 07 . For the invention task , there was a main effect of room size , with lower self - reported difficulty in the large ( M = 5 . 2 , SE = 0 . 3 ) vs . small room ( M = 6 . 8 , SE = 0 . 3 ) , d = - 1 . 16 [ - 1 . 81 , - 0 . 51 ] , F ( 1 , 45 ) = 15 . 7 , p = 0 . 00 . In contrast , for the RAT , participants in the large room self - reported about the same levels of perceived difficulty ( M = 7 . 1 , SE = . 30 ) as participants in the small room ( M = 7 . 4 , SE = . 29 ) , d = - 0 . 16 [ - 0 . 77 , 0 . 43 ] , F ( 1 , 45 ) = 0 . 34 , p = . 56 . Similarly , for the letters task , participants in the large room self - reported about the same levels of perceived difficulty ( M = 6 . 3 , SE = . 36 ) as participants in the small room ( M = 5 . 9 , SE = . 36 ) , d = 0 . 24 [ - 0 . 37 , 0 . 84 ] , F ( 1 , 45 ) = 0 . 65 , p = . 42 . Positive and negative affect There was no effect of room size on positive affect , with par - ticipants in the large room self - reporting about the same levels of positive affect ( M = 27 . 6 , SE = 1 . 5 ) as participants in the small room ( M = 27 . 4 , SE = 1 . 5 ) , d = 0 . 04 [ - 0 . 59 , 0 . 68 ] , F ( 1 , 41 ) = 0 . 02 , p = . 89 . Similarly , there was no effect of room size on negative affect , with participants in the large room self - reporting about the same levels of negative affect ( M = 12 . 3 , SE = . 47 ) as participants in the small room ( M = 11 . 8 , SE = . 46 ) , d = 0 . 22 [ - 0 . 41 , 0 . 86 ] , F ( 1 , 41 ) = 0 . 53 , p = . 47 . DIsCUssION Experiment 1 yielded evidence consistent with Hypothesis 1 ( see Figure 5 for a summary of the observed effects ) . As pre - dicted , participants\u2019 performance was higher on the diver - gent problem solving measures in the larger room than in the smaller room ( e . g . , uses fluency , uses novelty , invention nov - elty ) . In contrast , we found only partial support for Hypothesis 2 , that is , that participants in the large rooms would perform worse on convergence measures than participants in the small rooms . Consistent with the hypothesis , we found that partic - ipants in the larger room showed lower performance on the Situative Creativity 19 the intercorrelations between participants\u2019 problem solving task performance measures , collapsed across Experiments 1 and 2 . In general , the correlations were in the expected directions ( e . g . , significant positive correlations between divergent measures ; significant positive correlations Table 1 Descriptive statistics for experiment 1 . Mean Median Min Max SE Divergent measures Uses fluency 22 . 07 21 7 42 1 . 32 Uses novelty mean 1 . 67 1 . 59 1 2 . 58 0 . 06 Uses novelty max 3 . 12 3 1 4 0 . 13 Invention fluency 5 . 07 5 2 9 0 . 29 Invention novelty mean 2 . 74 2 . 83 1 . 56 3 . 83 0 . 08 Invention novelty max 3 . 58 3 . 67 2 4 0 . 08 Convergent measures Uses practicality 2 . 62 2 . 64 2 . 03 3 0 . 04 Invention practicality 3 . 15 3 . 21 1 . 79 3 . 86 0 . 06 RAT 0 . 59 0 . 62 0 1 0 . 04 Letters 0 . 61 0 . 71 0 1 0 . 05 Table 2 Intercorrelations between measures , collapsed across Experiments 1 and 2 . Divergent Convergent Uses novel mean Uses novel max Invent fluen . Invent novel mean Invent novel max Uses pract . Invent pract . RAT Letters U fluency 0 . 50 * 0 . 50 * 0 . 37 * 0 . 24 * 0 . 26 * - 0 . 43 * - 0 . 07 - 0 . 04 - 0 . 01 U novel mean 0 . 79 * 0 . 31 * 0 . 27 * 0 . 29 * - 0 . 89 * - 0 . 02 0 . 01 - 0 . 04 U novel max 0 . 17 * 0 . 25 * 0 . 22 * - 0 . 66 * 0 . 01 0 . 03 - 0 . 12 I fluency 0 . 09 0 . 46 * - 0 . 32 * 0 . 00 - 0 . 07 - 0 . 06 I novel mean 0 . 75 * - 0 . 23 * - 0 . 11 - 0 . 02 - 0 . 03 I novel max - 0 . 28 * - 0 . 05 - 0 . 01 - 0 . 02 U practicality 0 . 06 - 0 . 06 0 . 00 I practicality 0 . 14 m 0 . 00 RAT 0 . 21 * m p < . 10 ; * p < . 05 Table 2 Intercorrelations between measures , collapsed across Experiments 1 and 2 . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 37 uses practicality measure . However , no differences were found across the two groups on invention practicality and perfor - mance on the RAT and letter series task . In the general discus - sion we discuss possible reasons for why we did not observe stronger negative effects of a large space on convergence mea - sures . Taken together , these results show that the benefits of the larger room for divergent performance were not simply due to a general facilitation effect of being in the larger room ; rather , there seems to be a specific effect of being in a larger room on the cognitive processes that enable divergent performance . Analysis of the additional measures yielded additional insights . In support of Hypotheses 3 , the perceived task dif - ficulty results suggest that participants in the larger room not only performed better on the uses and invention tasks ( in terms of divergent performance measures ) , but also found the task overall to be less cognitively taxing ( compared to participants in the smaller rooms ) , suggesting that some of the performance benefits might be due to unconscious mechanisms ( e . g . , automatic attunement of semantic search patterns to search affordances in the physical environment ) . Further , analysis of the survey responses for PANAS suggests that the differences are not explained by positive boosts to affect in the larger room ( or increased negative affect in the smaller room ) . This result is consistent with the direct prim - ing hypothesis and not the concept activation account . EXPERIMENT 2 : REPlICATION AND EXTENsION Given the novelty of our hypotheses , we conducted a second study to replicate and extend the results of Experiment 1 . The focus of the extension is to ensure that the effects were not due to idiosyncrasies of the particular configurations of the large room or problem solving stimuli . To this end , we slightly altered the large room manipulation from Experiment 1 ( par - ticipants sat at the top of the auditorium rather than at the bot - tom ) , and added a second new large room that had a lower ceiling height but was still spacious horizontally . To maximize statistical power , we treated them as a single condition in our analyses . 2 We also changed the objects used for the uses task , as well as the categories and shapes used for the invention task . METhODs Participants One hundred and nine undergraduates ( 61 females ; ages 18 \u2013 31 , average age 19 ) enrolled in Introduction to Psy - chology at a large research university in the northeastern Situative Creativity 20 between letter series and RAT measures ; significant negative correlations between uses practicality and the divergent measures ) . However , the correlations are generally low , explaining small amounts of common variance , and some correlations are missing ( e . g . , no significant correlations between practicality and the letters series and RAT measures , and no significant correlations between invention practicality and any of the other measures ) . This suggests that the measures do not necessarily primarily reflect the two constructs of divergence and convergence ( as we had hypothesized ) . Therefore , we analyze each measure separately . Alternative uses There was an effect of room size on mean fluency , with participants in the large room generating more uses ( M = 25 . 0 , SE = 1 . 8 ) than participants in the small room ( M = 19 . 3 , SE = 1 . 8 ) , d = 0 . 68 , 95 % CI = [ 0 . 03 , 1 . 33 ] , F ( 1 , 41 ) = 5 . 0 , p = . 03 . There was also an effect of room size on mean novelty , with participants in the large room producing higher mean novelty with Figure 4 . Novelty and practicality of alternative uses by room size in Experiment 1 . Error bars are \u00b11 SE . Figure 4 Novelty and practicality of alternative uses by room size in Experiment 1 . Error bars are \u00b11 SE . Situative Creativity 24 showed lower performance on the uses practicality measure . However , no differences were found across the two groups on invention practicality and performance on the RAT and letter series task . In the general discussion we discuss possible reasons for why we did not observe stronger negative effects of a large space on convergence measures . Taken together , these results show that the benefits of the larger room for divergent performance were not simply due to a general facilitation effect of being in the larger room ; rather , there seems to be a specific effect of being in a larger room on the cognitive processes that enable divergent performance . Analysis of the additional measures yielded additional insights . In support of Hypotheses 3 , the perceived task difficulty results suggest that participants in the larger room not only performed better on the uses and invention tasks ( in terms of divergent performance measures ) , but also found the task overall to be less cognitively taxing ( compared to participants in the smaller rooms ) , suggesting that some of the performance benefits might be due to unconscious mechanisms ( e . g . , automatic attunement of semantic search patterns to search affordances in the Figure 5 . Summary of effects in Experiment 1 . Error bars are 95 % CIs . Figure 5 Summary of effects in Experiment 1 . Error bars are 95 % CIs . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 38 United States participated in this study . All participants were recruited through the university\u2019s psychology subject pool , and were compensated with course credit . Three participants ( assigned to the large rooms ) did not produce any valid responses to the invention task , and were dropped from all analyses , leaving us with 106 total partici - pants in our final dataset . 3 Materials Room size manipulation . In this study , the \u201clarge\u201d setting of our room size manipulation included two different rooms : 1 ) the same conference auditorium as in Experiment 1 , with the only difference being that participants sat at the top of the room , rather than the bottom of the room , and 2 ) another conference room in the same building ( see Figure 6 ) . The dimensions of the new large room were approximately 15\u2019 W \u00d7 30\u2019 L \u00d7 8\u2019 H . Other than the desk and chair the participants used , along with the other desks and chairs in the room , the room was empty . Note also that the lighting here is fluores - cent , similar to the small rooms in both experiments . The \u201csmall\u201d room was another former office space in the same building that was emptied out for the experiment . The dimensions of the room were the same as in Experiment 1 , approximately 8\u2019 W \u00d7 10\u2019 L \u00d7 8\u2019 H . As in Experiment 1 , participants completed their tasks on a desk facing one of the walls . It was empty except for the desk and chair the partici - pants used . Problem - solving tasks . As in Experiment 1 , participants com - pleted the alternative uses , shape invention , RAT , and let - ter series tasks . The only differences from Experiment 1 are with the stimuli for the alternative uses and shape invention tasks . The objects used for the alternative uses were \u201cCUP\u201d and \u201cTABLE . \u201d A different set of categories and shapes were randomly sampled for the invention task , using the same procedure as in Experiment 1 . The new categories were \u201cTransportation , \u201d \u201cFurniture , \u201d and \u201cWeapons , \u201d and the new shapes were \u201crectangular block , \u201d \u201cring , \u201d and \u201chalf - sphere\u201d ( see Figure 7 ) . Dependent measures As in Experiment 1 , we obtained measures of fluency and novelty for both the uses and invention tasks . Inter - rater reli - ability was high for novelty scoring across both tasks , ICC ( 2 , 2 ) = . 83 for uses novelty , and ICC ( 2 , 3 ) = . 84 for invention nov - elty . We then aggregated scores at the participant level into mean and max novelty . Also as in Experiment 1 , we evaluated uses and inventions for practicality . Inter - rater reliability was high for uses practicality , ICC ( 2 , 2 ) = . 82 , and acceptable for invention practicality , ICC ( 2 , 4 ) = . 67 . We then aggregated scores into participant - level measures of mean uses and invention practicality . Both the RAT and letter series tasks were scored identically to Experiment 1 ( i . e . , percent correct ) . Procedure The procedure was identical to Experiment 1 , except that partic - ipants did not complete the PANAS measurement at any point . Design This experiment had a between - subjects design . The inde - pendent variable was room size , with two levels ( large or small ) . Participants were randomly assigned to conditions . In the final dataset , there were 68 participants in the large rooms and 38 participants in the small room . REsUlTs Descriptive statistics for all measures are shown in Table 3 . Note that performance on all divergent measures ( except for uses fluency ) was significantly lower than that observed in Experiment 1 . In contrast , performance on the practi - cality measures for both the uses and invention tasks were Situative Creativity 26 Three participants ( assigned to the large rooms ) did not produce any valid responses to the invention task , and were dropped from all analyses , leaving us with 106 total participants in our final dataset . 3 Materials Room size manipulation . In this study , the \u201clarge\u201d setting of our room size manipulation included two different rooms : 1 ) the same conference auditorium as in Experiment 1 , with the only difference being that participants sat at the top of the room , rather than the bottom of the room , and 2 ) another conference room in the same building ( see Fig . 6 ) . The dimensions of the new large room were approximately 15\u2032 \u00d7 30\u2032 \u00d7 8\u2032 . Other than the desk and chair the participants used , along with the other desks and chairs in the room , the room was empty . Note also that the lighting here is fluorescent , similar to the small rooms in both experiments . The \u201csmall\u201d room was another former office space in the same building that was emptied 3 Results are the same with partial data from these participants included . Figure 6 . Picture of new large room in Experiment 2 . Figure 6 Picture of new large room in Experiment 2 . Situative Creativity 27 out for the experiment . The dimensions of the room were the same as in Experiment 1 , i . e . , approximately 8 \u2032 W \u00d7 10 \u2032 L \u00d7 8 \u2032 H . As in Experiment 1 , participants completed their tasks on a desk facing one of the walls . It was empty except for the desk and chair the participants used . Problem - solving tasks . As in Experiment 1 , participants completed the alternative uses , shape invention , RAT , and letter series tasks . The only differences from Experiment 1 are with the stimuli for the alternative uses and shape invention tasks . The objects used for the alternative uses were \u201cCUP\u201d and \u201cTABLE . \u201d A different set of categories and shapes were randomly sampled for the invention task , using the same procedure as in Experiment 1 . The new categories were \u201cTransportation , \u201d \u201cFurniture , \u201d and \u201cWeapons , \u201d and the new shapes were \u201crectangular block , \u201d \u201cring , \u201d and \u201chalf - sphere\u201d ( see Fig . 7 ) . Dependent measures As in Experiment 1 , we obtained measures of fluency and novelty for both the uses and invention tasks . Inter - rater reliability was high for novelty scoring across both tasks , ICC ( 2 , 2 ) = . 83 for uses novelty , and ICC ( 2 , 3 ) = . 84 for invention novelty . We then aggregated scores at the participant level into mean and max novelty . Also as in Experiment 1 , we evaluated uses and inventions for practicality . Inter - rater reliability was high for uses practicality , ICC ( 2 , 2 ) = . 82 , and acceptable for invention practicality , ICC ( 2 , 4 ) = . 67 . We then aggregated scores into Figure 7 . Shapes for invention task in Experiment 2 . Figure 7 Shapes for invention task in Experiment 2 . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 39 significantly higher than in Experiment 1 . However , letters performance was also lower than in Experiment 1 . We return to this issue in the discussion when interpreting the relation - ship between the results across the two experiments . Alternative uses In contrast to Experiment 1 , there was no effect of room size on mean fluency , F ( 1 , 1 04 ) = 0 . 4 , p = . 53 , with participants in the large rooms generating about the same number of uses ( M = 20 . 3 , SE = 0 . 9 ) as participants in the small room ( M = 19 . 3 , SE = 1 . 2 ) , d = 0 . 13 [ - 0 . 28 , 0 . 54 ] . In contrast , similar to Experiment 1 , mean trends for novelty of uses were in the hypothesized direction . However , the mean differences did not reach statistical significance . Mean novelty of uses was nonsignificantly higher in the large room ( M = 1 . 4 , SE = 0 . 0 ) compared to the small room ( M = 1 . 3 , SE = 0 . 0 ) , d = 0 . 34 [ - 0 . 07 , 0 . 75 ] , F ( 1 , 104 ) = 2 . 8 , p = . 10 ( Fig . 8 , left panel ) . Max novelty was marginally higher in the large rooms ( M = 2 . 9 , SE = 0 . 1 ) compared to the small room ( M = 2 . 6 , SE = 0 . 1 ) , d = 0 . 37 [ - 0 . 03 , 0 . 78 ] , F ( 1 , 104 ) = 3 . 4 , p = . 07 ( Fig . 8 , middle panel ) . Similar to Experiment 1 , there was an effect of room size on uses practicality , F ( 1 , 104 ) = 4 . 1 , p = . 04 , with participants in the larger rooms generating less practical uses ( M = 2 . 8 , SE = 0 . 01 ) than participants in the small room ( M = 2 . 9 , SE = 0 . 02 ) , d = - 0 . 41 [ - 0 . 82 , 0 . 00 ] ( Fig 8 , right panel ) . Shape invention There was no effect of room size on mean fluency , F ( 1 , 104 ) = 0 . 9 , p = . 36 , with participants in the large room generating about the same number of inventions ( M = 3 . 4 , SE = 0 . 3 ) as participants in the small room ( M = 3 . 8 , SE = 0 . 3 ) , d = - 0 . 19 [ - 0 . 59 , 0 . 22 ] . In contrast to Experiment 1 , there were no reli - able effects of room size on novelty of inventions : Participants in the larger rooms had similar mean novelty scores ( M = 2 . 5 , SE = 0 . 1 ) as participants in the small room ( M = 2 . 5 , SE = 0 . 1 ) , d = 0 . 14 [ - 0 . 26 , 0 . 55 ] , F ( 1 , 104 ) = 0 . 5 , p = . 49 . Similarly , partici - pants in the larger rooms achieved about the same max novelty scores ( M = 3 . 2 , SE = 0 . 1 ) as participants in the small room ( M = 3 . 1 , SE = 0 . 1 ) , d = 0 . 06 [ - 0 . 35 , 0 . 46 ] , F ( 1 , 104 ) = 0 . 1 , p = 0 . 79 . There was no effect of room size on invention practical - ity , with participants in the large room generating inventions that were about as practical ( M = 3 . 4 , SE = 0 . 1 ) as partici - pants in the small room ( M = 3 . 5 , SE = 0 . 1 ) , d = - 0 . 15 [ - 0 . 56 , 0 . 25 ] , F ( 1 , 104 ) = 0 . 6 , p = . 46 . RAT There was no effect of room size , F ( 1 , 104 ) = 0 . 02 , p = . 90 . Participants in the larger rooms had about the same mean proportion correct ( M = . 56 , SE = . 03 ) as participants in the small room ( M = . 55 , SE = . 04 ) , d = 0 . 03 [ - 0 . 37 , 0 . 43 ] . Situative Creativity 29 Alternative uses In contrast to Experiment 1 , there was no effect of room size on mean fluency , F ( 1 , 1 04 ) = 0 . 4 , p = . 53 , with participants in the large rooms generating about the same number of uses ( M = 20 . 3 , SE = 0 . 9 ) as participants in the small room ( M = 19 . 3 , SE = 1 . 2 ) , d = 0 . 13 [ - 0 . 28 , 0 . 54 ] . In contrast , similar to Experiment 1 , mean trends for novelty of uses were in the hypothesized direction . However , the mean differences did not reach statistical significance . Mean novelty of uses was nonsignificantly higher in the large room ( M = 1 . 4 , SE = 0 . 0 ) compared to the small room ( M = 1 . 3 , SE = 0 . 0 ) , d = 0 . 34 [ - 0 . 07 , 0 . 75 ] , F ( 1 , 104 ) = 2 . 8 , p = . 10 ( Fig . 8 , left panel ) . Max novelty was marginally higher in the large rooms ( M = 2 . 9 , SE = 0 . 1 ) compared to the small room ( M = 2 . 6 , SE = 0 . 1 ) , d = 0 . 37 [ - 0 . 03 , 0 . 78 ] , F ( 1 , 104 ) = 3 . 4 , p = . 07 ( Fig . 8 , middle panel ) . Similar to Experiment 1 , there was an effect of room size on uses practicality , F ( 1 , 104 ) = 4 . 1 , p = . 04 , with participants in the larger rooms generating less practical uses ( M = 2 . 8 , SE = Table 3 Descriptive statistics for experiment 2 . Mean Median Min Max SE Uses fluency 19 . 93 19 6 37 0 . 69 Uses novelty mean 1 . 38 V 1 . 35 1 2 . 12 0 . 02 Uses novelty max 2 . 77 V 3 1 4 0 . 08 Invention fluency 3 . 51 V 3 1 13 0 . 21 Invention novelty mean 2 . 51 V 2 . 52 1 . 25 3 . 78 0 . 05 Invention novelty max 3 . 17 V 3 . 33 1 . 67 4 0 . 07 Uses practicality 2 . 83 ^ 2 . 86 2 . 47 3 0 . 01 Invention practicality 3 . 43 ^ 3 . 50 1 . 5 4 . 33 0 . 04 RAT 0 . 56 0 . 55 0 1 0 . 02 Letters 0 . 50 V 0 . 52 0 0 . 78 0 . 01 V p < . 05 lower than Experiment 1 ; ^ p < . 05 higher than Experiment 1 . Table 3 Descriptive statistics for Experiment 2 . Situative Creativity 30 0 . 01 ) than participants in the small room ( M = 2 . 9 , SE = 0 . 02 ) , d = - 0 . 41 [ - 0 . 82 , 0 . 00 ] ( Fig 8 , right panel ) . Shape invention There was no effect of room size on mean fluency , F ( 1 , 104 ) = 0 . 9 , p = . 36 , with participants in the large room generating about the same number of inventions ( M = 3 . 4 , SE = 0 . 3 ) as participants in the small room ( M = 3 . 8 , SE = 0 . 3 ) , d = - 0 . 19 [ - 0 . 59 , 0 . 22 ] . In contrast to Experiment 1 , there were no reliable effects of room size on novelty of inventions : Participants in the larger rooms had similar mean novelty scores ( M = 2 . 5 , SE = 0 . 1 ) as participants in the small room ( M = 2 . 5 , SE = 0 . 1 ) , d = 0 . 14 [ - 0 . 26 , 0 . 55 ] , F ( 1 , 104 ) = 0 . 5 , p = . 49 . Similarly , participants in the larger rooms achieved about the same max novelty scores ( M = 3 . 2 , SE = 0 . 1 ) as participants in the small room ( M = 3 . 1 , SE = 0 . 1 ) , d = 0 . 06 [ - 0 . 35 , 0 . 46 ] , F ( 1 , 104 ) = 0 . 1 , p = 0 . 79 . Figure 8 . Novelty and practicality of alternative uses by room size in Experiment 2 . Error bars are \u00b11 SE . Figure 8 Novelty and practicality of alternative uses by room size in Experiment 2 . Error bars are \u00b11 SE . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 40 Letter series There was no effect of room size , F ( 1 , 104 ) = 0 . 07 , p = . 79 . Participants in the larger rooms had about the same mean proportion correct ( M = . 50 , SE = . 02 ) as participants in the small room ( M = . 50 , SE = . 02 ) , d = - 0 . 05 [ - 0 . 46 , 0 . 35 ] . Perceived task difficulty In contrast to Experiment 1 , there was no effect of room size on perceived difficulty for the uses task , F ( 1 , 104 ) = 0 . 0 , p = . 98 . Participants in the larger rooms self - reported the same levels of cognitive load ( M = 5 . 1 , SE = . 18 ) as participants in the small room ( M = 5 . 1 , SE = . 24 ) , d = - 0 . 00 [ - 0 . 40 , 0 . 41 ] . Similarly , for the invention task , participants self - reported the same level of difficulty in the large ( M = 5 . 9 , SE = 0 . 21 ) and small rooms ( M = 5 . 9 , SE = 0 . 28 ) , d = - 0 . 00 [ \u2013 0 . 41 , 0 . 40 ] , F ( 1 , 104 ) = 0 . 0 , p = 0 . 98 . Results were the same as Experiment 1 for the RAT and letters tasks . For the RAT , perceived difficulty was about the same in the large ( M = 7 . 5 , SE = 0 . 15 ) and small rooms ( M = 7 . 3 , SE = 0 . 20 ) , d = 0 . 13 [ - 0 . 28 , 0 . 53 ] , F ( 1 , 104 ) = 0 . 4 , p = 0 . 53 . Similarly , perceived difficulty of the letters task was the same in the large ( M = 0 . 50 , SE = 0 . 02 ) and small rooms ( M = 0 . 50 , SE = 0 . 02 ) , d = - 0 . 05 [ - 0 . 46 , 0 . 35 ] , F ( 1 , 104 ) = 0 . 1 , p = 0 . 79 . DIsCUssION In Experiment 2 , we sought to replicate and extend the findings from Experiment 1 . See Figure 8 for a summary of the effects . We observed very similar patterns of effects for the alternative uses , RAT , and letter series tasks . Similar to Experiment 1 , novelty ( both mean and max ) of uses was higher in the larger vs . small room , although the effect size was substantially smaller than Experiment 1 ( approximately half the size ) . These trends are in the predicted direction of Hypothesis 1 , and consistent with both the direct priming and concept activation accounts . Overall performance on the divergent measures were worse in both conditions in this experiment compared to Experiment 1 , which may suggest floor effects . In the general discussion we further consider possible reasons for the partial replication . Partial support and replication was found for Hypothesis 2 , with participants in the large room showing worse perfor - mance on the practicality measure of the alternative uses task compared to those in the small room . Also similar to Experi - ment 1 , there was no effect of room size on the RAT or letter series task . However , perceived difficulty patterns did not rep - licate from Experiment 1 . The failure to replicate the difference in perceived difficulty means that findings do not support the direct priming hypothesis more so than the concept activation account . In sum , we observed a partial replication of the results in Experiment 1 ( mainly with novelty and practicality of uses ) . GENERAl DIsCUssION In this paper , we sought to explore the potential cognitive bases of the spatial metaphor that initial exploration of a creative space should be \u201cbroad . \u201d Across two experiments , we tested the hypothesis that larger physical spaces facilitate divergent , but not convergent , processes in problem - solving . Experiment 1 provided support for Hypothesis 1 and 2 , and Experiment 2 partially replicated the findings . Smaller room sizes facilitated the generation of more practical uses of everyday objects across both experiments . Larger room sizes facilitated the generation of more novel uses of everyday objects ( specifically mean and max novelty ) in Experiment 1 ; although the same trends were seen in Experiment 2 , the effects did not reach significance . As noted in our discussion of the descriptive statistics in Experiment 2 , there was a significant drop in performance across many of our measures from Experiment 1 , suggest - ing that there may have been important differences in the two samples . One potential explanation is that Experiment 2 participants were exclusively undergraduate students par - ticipating for course credit , whereas Experiment 1 partici - pants were paid volunteers and included a wider range of demographics ( not just undergraduate students ) . Reduced motivation may have led to floor effects , potentially reducing the sensitivity of our measures . Motivation differences might have been especially important since we did not provide direct instructions to \u201cbe creative\u201d in either experiment . For example , while the mean rated novelty of alternative uses was close to \u201cnot at all novel\u201d across both experiments , the mean and variability was higher in Experiment 1 ( M = 1 . 67 , SD = 0 . 42 ) compared to Experiment 2 ( M = 1 . 38 , SD = 0 . 24 ) . The generally low novelty scores ( and high practicality scores ) with relatively low variance suggest that participants were in general defaulting to more convergent processing . This Situative Creativity 32 7 . 3 , SE = 0 . 20 ) , d = 0 . 13 [ - 0 . 28 , 0 . 53 ] , F ( 1 , 104 ) = 0 . 4 , p = 0 . 53 . Similarly , perceived difficulty of the letters task was the same in the large ( M = 0 . 50 , SE = 0 . 02 ) and small rooms ( M = 0 . 50 , SE = 0 . 02 ) , d = - 0 . 05 [ - 0 . 46 , 0 . 35 ] , F ( 1 , 104 ) = 0 . 1 , p = 0 . 79 . Discussion In Experiment 2 , we sought to replicate and extend the findings from Experiment 1 . See Figure 8 for a summary of the effects . We observed very similar patterns of effects for the alternative uses , RAT , and letter series tasks . Similar to Experiment 1 , novelty ( both mean and max ) of uses was higher in the larger vs . small room , although the effect size was substantially smaller than Experiment 1 ( approximately half the size ) . These trends are in the predicted direction of Hypothesis 1 , and consistent with both the direct priming and concept activation accounts . Overall performance on the divergent measures were worse in both conditions in this experiment compared to Experiment 1 , which may suggest floor effects . In the general Figure 9 . Summary of effects in Experiment 2 . Error bars are 95 % CIs . Figure 9 Summary of effects in Experiment 2 . Error bars are 95 % CIs . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 41 observation is consistent with prior research on the \u201cpath of least resistance\u201d in creative production ( Ward , 1994 ) , and other work that has shown that instructions to \u201cbe creative\u201d can yield substantial improvements to creative output ( Nus - baum , Silvia , & Beaty , 2014 ) : People generally need to expend cognitive effort to overcome initial biases toward less cre - ative responses ( e . g . , using cognitive control to inhibit more accessible but less creative responses ; Beaty , Silvia , Nusbaum , Jauk , & Benedek , 2014 ) . In light of this , it may be useful to think of these results as describing the effects of room size on \u201cdefault\u201d problem solving ( i . e . , when participants are not necessarily actively trying to be creative in their responses ) . We believe that pooling the data from the two experi - ments provides the clearest picture ( e . g . , robust across a wide range of participants and problem solving stimuli ) of whether there is a relationship between room size and diver - gent and convergent problem solving processes . The pooled data indicate reliable evidence that larger physical spaces facilitate novelty and hinder practicality of solutions on the alternative uses task ( see Figure 9 ) . In the pooled data , mean novelty of uses is higher in the larger rooms ( M = 1 . 5 , SE = 0 . 03 ) compared to the small rooms ( M = 1 . 4 , SE = 0 . 04 ) , d = 0 . 34 [ 0 . 01 , 0 . 67 ] , F ( 1 , 147 ) = 4 . 13 , p = . 04 . Similarly , max novelty of uses is higher in the larger rooms ( M = 3 . 0 , SE = 0 . 09 ) compared to the small rooms ( M = 2 . 7 , SE = 0 . 11 ) , d = 0 . 37 [ 0 . 04 , 0 . 71 ] , F ( 1 , 147 ) = 5 . 13 , p = . 03 . In contrast , mean practicality of uses was lower in the larger rooms ( M = 2 . 7 , SE = 0 . 02 ) compared to the small rooms ( M = 2 . 8 , SE = 0 . 02 ) , d = - 0 . 38 [ - 0 . 71 , - 0 . 04 ] , F ( 1 , 147 ) = 5 . 11 , p = . 03 . Our primary goal in this study is to document a psycholog - ical phenomenon : We provide an initial test of whether there is an association between room size and divergent problem Situative Creativity 35 We believe that pooling the data from the two experiments provides the clearest picture ( e . g . , robust across a wide range of participants and problem solving stimuli ) of whether there is a relationship between room size and divergent and convergent problem solving processes . The pooled data indicate reliable evidence that larger physical spaces facilitate novelty and hinder practicality of solutions on the alternative uses task ( see Figure 9 ) . In the pooled data , mean novelty of uses is higher in the larger rooms ( M = 1 . 5 , SE = 0 . 03 ) compared to the small rooms ( M = 1 . 4 , SE = 0 . 04 ) , d = 0 . 34 [ 0 . 01 , 0 . 67 ] , F ( 1 , 147 ) = 4 . 13 , p = . 04 . Similarly , max novelty of uses is higher in the larger rooms ( M = 3 . 0 , SE = 0 . 09 ) compared to the small rooms ( M = 2 . 7 , SE = 0 . 11 ) , d = 0 . 37 [ 0 . 04 , 0 . 71 ] , F ( 1 , 147 ) = 5 . 13 , p = . 03 . In contrast , mean practicality of uses was lower in the larger rooms ( M = 2 . 7 , SE = 0 . 02 ) compared to the small rooms ( M = 2 . 8 , SE = 0 . 02 ) , d = - 0 . 38 [ - 0 . 71 , - 0 . 04 ] , F ( 1 , 147 ) = 5 . 11 , p = . 03 . Our primary goal in this study is to document a psychological phenomenon : We provide an initial test of whether there is an association between room size and divergent problem solving Figure 10 . Novelty and practicality of alternative uses by room size , pooled across experiments . Error bars are \u00b11 SE . solving performance . While this effect appears to be relatively small and may depend on having sufficiently motivated par - ticipants , some aspects of our results provide hints for future theoretical refinement . While our results from the problem solving measures ( and non - replication of the hypothesized perceived task difficulty results from Experiment 1 to Experi - ment 2 ) are consistent with both a direct priming explanation ( Hills , 2006 ; Hills et al . , 2008 , 2012 ) , as well as the concept activation explanation ( Meyers - Levy & Zhu , 2007 ) , the affect results in Experiment 1 help to partially arbitrate between the explanations . The lack of effect on positive / negative affect is more consistent with a direct priming explanation , since concepts of \u201cfreedom\u201d or \u201copenness\u201d are expected to engen - der more positive affect , whereas attunement to resource distribution patterns are not . We therefore suggest that , to the extent that this effect proves reliable , it may be a conse - quence of automatic attunement of semantic search patterns to search affordances in the physical environment . That is , people may be responding to the physical search affordances of the physical environment by defocusing their attention to enable broader search in semantic memory , shifting from a tight focus on a few highly relevant responses to consider - ing more semantically distant and varied responses . However , alternative explanations are possible : For example , it is possi - ble that the small room reminded people of traditional office environments , which may have invoked a \u201cwork schema\u201d that primed more focused attention , impairing divergent think - ing . Further investigations are necessary to tease apart the psychological underpinnings of this effect . Although we intended the letter series task and RAT to be measures of convergent problem solving processes based on the nature of the response required ( many = divergent , single Figure 10 Novelty and practicality of alternative uses by room size , pooled across experiments . Error bars are \u00b11 SE . docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 42 = convergent ) , in hindsight , it is probably best to consider both the RAT and the letter series task as a more even mix of diver - gent and convergent processes than the fluency and novelty measures for the uses and invention tasks ( which quite cleanly measure divergent processes ) . For example , in the RAT , one might first search broadly for possible meanings ( strong and weak associates ) of the target words and then only later con - verge on the common target that links across all three . Similarly , the letter series task may first require divergent search for sev - eral possible patterns before converging on the single , correct pattern to extrapolate . The mixture of both divergent and con - vergent processes in the letter series task and RAT might explain why we did not find a harmful effect of large rooms on letter series and RAT performance , and why we only found a harmful effect of large rooms on practicality of uses ( which turned out empirically to be our cleanest measures of convergent processes , based on the intercorrelations between measures ) . Our data have broader implications for the psychology of creative problem solving . For example , our observed strong negative correlations between novelty and practicality of uses corroborate prior arguments that originality and prac - ticality in creative thought are cognitively at odds with each other ( Goldenberg et al . , 2013 ) . Our results also have impli - cations for how we should think about the RAT as a measure of creativity . In this study , we departed from a number of prior studies that have examined the divergent aspects of the RAT , for example , studying the relationship between defo - cused attention and RAT performance ( Aiello et al . , 2012 ; Haarmann et al . , 2012 ) . However , the differing patterns of results and lack of correlation between the RAT and diver - gent problem solving measures for the invention and uses task suggest that the RAT may involve more convergence than is typically described . Noting these findings might lead to more fruitful theoretical examinations of the relation - ship between the RAT and problem solving and creativity . For example , Goel , Eimontaite , Goel , and Schindler ( 2015 ) recently argued that insight problems ( such as the RAT ) are a subset of well - structured problems , while divergent prob - lem solving tasks ( such as the alternative uses and invention tasks ) are a subset of ill - structured problems . Our results also echo a number of recent studies that have demonstrated the psychological separability of divergent and convergent prob - lem solving processes ( Chermahini & Hommel , 2010 ; Col - zato et al . , 2012 ; Hommel , Colzato , Fischer , & Christoffels , 2011 ; Oppezzo & Schwartz , 2014 ; Radel et al . , 2015 ) . We join these more nuanced arguments to call for more careful anal - yses of the components of creative performance ( e . g . , sepa - rating divergent vs . convergent processes ) in future research on creativity . NOTEs 1 Results are the same with partial data from these partici - pants included . 2 There were no statistical differences between participants in the two large rooms on any of the measures . 3 Results are the same with partial data from these partici - pants included . REFERENCEs Ad\u00e1nez , A . M . ( 2005 ) . Does quantity generate quality ? Testing the fundamental principle of brainstorming . Spanish Journal of Psychology , 8 ( 2 ) , 215 \u2013 220 . http : / / doi . org / 10 . 1017 / S1138741600005096 Aiello , D . , Jarosz , A . , Cushen , P . , & Wiley , J . ( 2012 ) . Firing the executive : When an analytic approach to problem solving helps and hurts . Journal of Problem Solving , 4 ( 2 ) . http : / / doi . org / 10 . 7771 / 1932 - 6246 . 1128 Amabile , T . M . ( 1983 ) . The social psychology of creativ - ity : A componential conceptualization . Journal of Per - sonality and Social Psychology , 45 ( 2 ) , 357 . http : / / dx . doi . org / 10 . 1037 / 0022 - 3514 . 45 . 2 . 357 Antonenko , P . , Paas , F . , Grabner , R . , & Gog , T . van . ( 2010 ) . using electroencephalography to measure cognitive load . Educational Psychology Review , 22 ( 4 ) , 425 \u2013 438 . http : / / doi . org / 10 . 1007 / s10648 - 010 - 9130 - y Beaty , R . E . , Silvia , P . J . , Nusbaum , E . C . , Jauk , E . , & Benedek , M . ( 2014 ) . The roles of associative and executive processes in creative cognition . Memory & Cognition , 42 ( 7 ) , 1186 \u2013 1197 . http : / / doi . org / 10 . 3758 / s13421 - 014 - 0428 - 8 Bhalla , M . , & Proffitt , D . R . ( 1999 ) . Visual - motor reca - libration in geographical slant perception . Journal of Experimental Psychology : Human Perception and Perfor - mance , 25 ( 4 ) , 1076 \u2013 1096 . http : / / dx . doi . org / 10 . 1037 / 0096 - 1523 . 25 . 4 . 1076 Bilali\u0107 , M . , McLeod , P . , & Gobet , F . ( 2008 ) . Why good thoughts block better ones : The mechanism of the perni - cious Einstellung ( set ) effect . Cognition , 108 ( 3 ) , 652 \u2013 661 . http : / / dx . doi . org / 10 . 1016 / j . cognition . 2008 . 05 . 005 Bono , E . D . ( 1970 ) . Lateral thinking : Creativity step by step . New York : Harper Colophon . Bowers , K . S . , Regehr , G . , Balthazard , C . , & Parker , K . ( 1990 ) . Intuition in the context of discovery . Cogni - tive Psychology , 22 ( 1 ) , 72 \u2013 110 . http : / / dx . doi . org / 10 . 1016 / 0010 - 0285 ( 90 ) 90004 - N Brown , J . S . , Collins , A . , & Duguid , P . ( 1989 ) . Situated cognition and the culture of learning . Educational Researcher , 18 ( 1 ) , 32 \u2013 42 . http : / / dx . doi . org / 10 . 3102 / 0013189X018001032 Chan , J . , Fu , K . , Schunn , C . D . , Cagan , J . , Wood , K . L . , & Kotovsky , K . ( 2011 ) . On the benefits and pitfalls of analo - gies for innovative design : Ideation performance based on docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 43 analogical distance , commonness , and modality of exam - ples . Journal of Mechanical Design , 133 , 081004 . http : / / doi . org / 10 . 1115 / 1 . 4004396 Chandler , P . , & Sweller , J . ( 1991 ) . Cognitive load theory and the format of instruction . Cognition and Instruction , 8 ( 4 ) , 293 \u2013 332 . http : / / dx . doi . org / 10 . 1207 / s1532690xci0804 _ 2 Chermahini , S . A . , & Hommel , B . ( 2010 ) . The ( b ) link between creativity and dopamine : Spontaneous eye blink rates predict and dissociate divergent and convergent think - ing . Cognition , 115 ( 3 ) , 458 \u2013 465 . http : / / doi . org / 10 . 1016 / j . cognition . 2010 . 03 . 007 Colzato , L . S . , Ozturk , A . , & Hommel , B . ( 2012 ) . Meditate to create : The impact of focused - attention and open - monitoring training on convergent and divergent think - ing . Frontiers in Psychology , 3 ( April ) , 116 . http : / / dx . doi . org / 10 . 3389 / fpsyg . 2012 . 00116 Finke , R . A . , Ward , T . B . , & Smith , S . M . ( 1996 ) . Creative cog - nition : Theory , research , and applications . Cambridge , MA : MIT Press . Gentner , D . , & Markman , A . B . ( 1997 ) . Structure mapping in analogy and similarity . American Psychologist , 52 ( 1 ) , 45 \u2013 56 . http : / / dx . doi . org / 10 . 1037 / 0003 - 066X . 52 . 1 . 45 Girotra , K . , Terwiesch , C . , & Ulrich , K . T . ( 2010 ) . Idea gener - ation and the quality of the best idea . Management Science , 56 , 591 \u2013 605 . http : / / dx . doi . org / 10 . 1287 / mnsc . 1090 . 1144 Goel , V . , Eimontaite , I . , Goel , A . , & Schindler , I . ( 2015 ) . Dif - ferential modulation of performance in insight and diver - gent thinking tasks with tDCS . Journal of Problem Solving , 8 ( 1 ) . http : / / doi . org / 10 . 7771 / 1932 - 6246 . 1172 Goffman , E . ( 1974 ) . Frame analysis : An essay on the organiza - tion of experience ( Vol . ix ) . Cambridge , MA , US : Harvard University Press . Goldenberg , O . , Larson , J . R . , & Wiley , J . ( 2013 ) . Goal instructions , response format , and idea generation in groups . Small Group Research , 44 ( 3 ) , 227 \u2013 256 . http : / / doi . org / 10 . 1177 / 1046496413486701 Greeno , J . G . ( 1994 ) . Gibson\u2019s affordances . Psychologi - cal Review , 101 ( 2 ) , 336 \u2013 342 . http : / / dx . doi . org / 10 . 1037 / 0033 - 295X . 101 . 2 . 336 Greeno , J . G . , & Middle - School Mathematics through Appli - cations Project Group . ( 1998 ) . The situativity of knowing , learning , and research . American Psychologist , 53 ( 1 ) , 5 . http : / / dx . doi . org / 10 . 1037 / 0003 - 066X . 53 . 1 . 5 Guilford , J . P . ( 1956 ) . The structure of intellect . Psychological Bulletin , 53 ( 4 ) , 267 \u2013 293 . http : / / doi . org / 10 . 1037 / h0040755 Guilford , J . P . ( 1967 ) . The nature of human intelligence . New York : McGraw - Hill . Haarmann , H . , George , T . , Smaliy , A . , & Dien , J . ( 2012 ) . Remote associates test and alpha brain waves . Journal of Problem Solving , 4 ( 2 ) . http : / / doi . org / 10 . 7771 / 1932 - 6246 . 1126 Hall , E . T . ( 1966 ) . The hidden dimension ( 1st ed . ) ( Vol . xii ) . New York : Doubleday & Co . Hills , T . T . ( 2006 ) . Animal foraging and the evolution of goal - directed cognition . Cognitive Science , 30 ( 1 ) , 3 \u2013 41 . http : / / dx . doi . org / 10 . 1207 / s15516709cog0000 _ 50 Hills , T . T . , Jones , M . N . , & Todd , P . M . ( 2012 ) . Optimal for - aging in semantic memory . Psychological Review , 119 ( 2 ) , 431 \u2013 440 . http : / / dx . doi . org / 10 . 1037 / a0027373 Hills , T . T . , Todd , P . M . , & Goldstone , R . L . ( 2008 ) . Search in external and internal spaces : Evidence for generalized cog - nitive search processes . Psychological Science , 19 ( 8 ) , 802 \u2013 808 . http : / / dx . doi . org / 10 . 1111 / j . 1467 - 9280 . 2008 . 02160 . x Hommel , B . , Colzato , L . S . , Fischer , R . , & Christoffels , I . K . ( 2011 ) . Bilingualism and creativity : Benefits in convergent thinking come with losses in divergent thinking . Frontiers in Psychology , 2 ( November ) , 273 . http : / / doi . org / 10 . 3389 / fpsyg . 2011 . 00273 Jang , J . , & Schunn , C . D . ( 2012 ) . Physical design tools sup - port and hinder innovative engineering design . Jour - nal of Mechanical Design , 134 ( 4 ) , 041001 . http : / / dx . doi . org / 10 . 1115 / 1 . 4005651 Kaplan , C . , & Simon , H . A . ( 1990 ) . In search of insight . Cognitive Psychology , 22 ( 3 ) , 374 \u2013 419 . http : / / dx . doi . org / 10 . 1016 / 0010 - 0285 ( 90 ) 90008 - R Knoblich , G . , Ohlsson , S . , Haider , H . , & Rhenius , D . ( 1999 ) . Constraint relaxation and chunk decomposition in insight problem solving . Journal of Experimental Psychol - ogy : Learning , Memory , and Cognition , 25 ( 6 ) , 1534 \u2013 1555 . http : / / dx . doi . org / 10 . 1037 / 0278 - 7393 . 25 . 6 . 1534 Koch , C . , & Tsuchiya , N . ( 2007 ) . Attention and consciousness : Two distinct brain processes . Trends in Cognitive Sciences , 11 ( 1 ) , 16 \u2013 22 . http : / / doi . org / 10 . 1016 / j . tics . 2006 . 10 . 012 Luchins , A . S . ( 1942 ) . Mechanization in problem solving : The effect of Einstellung . Psychological Monographs , 54 ( 6 ) , i \u2013 95 . http : / / doi . org / 10 . 1037 / h0093502 MacGregor , J . , & Cunningham , J . ( 2009 ) . The effects of number and level of restructuring in insight prob - lem solving . Journal of Problem Solving , 2 ( 2 ) . http : / / doi . org / 10 . 7771 / 1932 - 6246 . 1062 Maclachlan , G . L . , & Reid , I . ( 1994 ) . Framing and interpreta - tion . Melbourne , Australia : Melbourne University Press . Martindale , C . ( 1997 ) . Creativity and connectionism . In S . M . Smith , T . B . Ward , & R . A . Finke ( Eds . ) , The creative cognition approach ( pp . 249 \u2013 268 ) . Cambridge , MA : MIT Press . Mednick , S . A . ( 1962 ) . The associative basis of the creative process . Psychological Review , 69 ( 3 ) , 220 \u2013 232 . http : / / dx . doi . org / 10 . 1037 / h0048850 Mednick , S . A . , & Mednick , M . T . ( 1967 ) . Examiner\u2019s man - ual : Remote Associates Test . Boston , MA . Meyers - Levy , J . , & Zhu , R . J . ( 2007 ) . The influence of ceiling height : The effect of priming on the type of pro - cessing that people use . Journal of Consumer Research , 34 ( 2 ) , 174 \u2013 186 . http : / / dx . doi . org / 10 . 1086 / 519146 docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 44 Moore , G . T . , Lane , C . G . , Hill , A . B . , Cohen , U . , & McGinty , T . ( 1994 ) . Recommendations for child care centers . Revised edition . Retrieved from http : / / eric . ed . gov / ? id = ED417797 Newell , A . , & Simon , H . A . ( 1972 ) . Human problem solving . Englewood Cliffs , NJ : Prentice - Hall . Nokes - Malach , T . J . , & Mestre , J . P . ( 2013 ) . Toward a model of transfer as sense - making . Educational Psychologist , 48 ( 3 ) , 184 \u2013 207 . http : / / doi . org / 10 . 1080 / 00461520 . 2013 . 807556 Nokes , T . J . ( 2009 ) . Mechanisms of knowledge trans - fer . Thinking & Reasoning , 15 ( 1 ) , 1 \u2013 36 . http : / / dx . doi . org / 10 . 1080 / 13546780802490186 Nusbaum , E . C . , Silvia , P . J . , & Beaty , R . E . ( 2014 ) . Ready , set , create : What instructing people to \u201cbe creative\u201d reveals about the meaning and mechanisms of divergent think - ing . Psychology of Aesthetics , Creativity , and the Arts , 8 ( 4 ) , 423 \u2013 432 . http : / / doi . org / 10 . 1037 / a0036549 Ohlsson , S . ( 1992 ) . Information - processing explanations of insight and related phenomena . In M . T . Keane & K . J . Gil - hooly ( Eds . ) , Advances in the psychology of thinking ( Vol . 1 , pp . 1 \u2013 44 ) . Hertfordshire , UK : Harvester Wheatsheaf . Oppezzo , M . , & Schwartz , D . L . ( 2014 ) . Give your ideas some legs : The positive effect of walking on creative thinking . Jour - nal of Experimental Psychology : Learning , Memory , and Cog - nition , 40 ( 4 ) , 1142 \u2013 1152 . http : / / doi . org / 10 . 1037 / a0036577 Parnes , S . J . , & Meadow , A . ( 1959 ) . Effects of \u201cbrainstorm - ing\u201d instructions on creative problem solving by trained and untrained subjects . Journal of Educational Psychology , 50 ( 4 ) , 171 \u2013 176 . http : / / dx . doi . org / 10 . 1037 / h0047223 Radel , R . , Davranche , K . , Fournier , M . , & Dietrich , A . ( 2015 ) . The role of ( dis ) inhibition in creativity : Decreased inhibi - tion improves idea generation . Cognition , 134 , 110 \u2013 120 . http : / / dx . doi . org / 10 . 1016 / j . cognition . 2014 . 09 . 001 Sawyer , R . K . ( 2012 ) . Explaining creativity : The science of human innovation ( 2nd ed . ) . New York : Oxford University Press . Scherr , R . E . , & Hammer , D . ( 2009 ) . Student behavior and epis - temological framing : Examples from collaborative active - learning activities in physics . Cognition and Instruction , 27 ( 2 ) , 147 \u2013 174 . http : / / doi . org / 10 . 1080 / 07370000902797379 Shah , J . J . , Millsap , R . E . , Woodward , J . , & Smith , S . M . ( 2012 ) . Applied tests of design skills part 1 : Divergent thinking . Journal of Mechanical Design , 134 ( 2 ) , 021005 \u2013 021005 \u2013 10 . http : / / doi . org / 10 . 1115 / 1 . 4005594 Simon , H . A . , & Kotovsky , K . ( 1963 ) . Human acquisition of concepts for sequential patterns . Psychological Review , 70 ( 6 ) , 534 \u2013 546 . http : / / dx . doi . org / 10 . 1037 / h0043901 Simonton , D . K . ( 2011 ) . Creativity and discovery as blind variation : Campbell\u2019s ( 1960 ) BVSR model after the half - century mark . Review of General Psychology , 15 ( 2 ) . http : / / dx . doi . org / 10 . 1037 / a0022912 Terwiesch , C . , & Ulrich , K . T . ( 2009 ) . Innovation tourna - ments : Creating and selecting exceptional opportunities . Boston , MA : Harvard Business Press . Torrance , E . P . ( 1988 ) . The nature of creativity as manifest in its testing . In R . J . Sternberg ( Ed . ) , The nature of creativity : Contemporary psychological perspectives ( pp . 43 \u2013 75 ) . New York : Cambridge University Press . Wallas , G . ( 1926 ) . The art of thought . New York : Harcourt , Brace and Company Ward , T . B . ( 1994 ) . Structured imagination : The role of cat - egory structure in exemplar generation . Cognitive Psy - chology , 27 ( 1 ) , 1 \u2013 40 . http : / / dx . doi . org / 10 . 1006 / cogp . 1994 . 1010 Ward , T . B . ( 1998 ) . Analogical distance and purpose in cre - ative thought : Mental leaps versus mental hops . In K . J . Holyoak , D . Gentner , & B . Kokinov ( Eds . ) , Advances in analogy research : Integration of theory and data from the cognitive , computational , and neural sciences ( pp . 221 \u2013 230 ) . Sofia , Bulgaria . Warr , A . , & O\u2019Neill , E . ( 2005 ) . Understanding design as a social creative process . In Proceedings of the 5th Confer - ence on Creativity & Cognition ( pp . 118 \u2013 127 ) . New York : ACM . http : / / doi . org / 10 . 1145 / 1056224 . 1056242 Warriner , A . B . , Kuperman , V . , & Brysbaert , M . ( 2013 ) . Norms of valence , arousal , and dominance for 13 , 915 English lemmas . Behavior Research Methods , 45 ( 4 ) , 1191 \u2013 1207 . http : / / doi . org / 10 . 3758 / s13428 - 012 - 0314 - x Watson , D . , Clark , L . A . , & Tellegen , A . ( 1988 ) . Develop - ment and validation of brief measures of positive and negative affect : The PANAS scales . Journal of Personality and Social Psychology , 54 ( 6 ) , 1063 \u2013 1070 . http : / / dx . doi . org / 10 . 1037 / 0022 - 3514 . 54 . 6 . 1063 Wiley , J . ( 1998 ) . Expertise as mental set : The effects of domain knowledge in creative problem solving . Memory & Cognition , 26 ( 4 ) , 716 \u2013 730 . http : / / dx . doi . org / 10 . 3758 / BF03211392 Wiley , J . , & Jarosz , A . F . ( 2012 ) . Working memory capacity , attentional focus , and problem solving . Current Direc - tions in Psychological Science , 21 ( 4 ) , 258 \u2013 262 . http : / / doi . org / 10 . 1177 / 0963721412447622 docs . lib . purdue . edu / jps 2016 | Volume 9 J . Chan & T . J . Nokes - Malach Situative Creativity 45 APPENDIX A : FUll lIsT OF ITEMs FOR RAT TAsK APPENDIX B : FUll lIsT OF ITEMs FOR lETTER sERIEs TAsK Situative Creativity 46 APPENDIX A : FULL LIST OF ITEMS FOR RAT TASK Situative Creativity 47 APPENDIX B : FULL LIST OF ITEMS FOR LETTER SERIES TASK 1 . aaabbbcccdd _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 2 . atbataatbat _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 3 . abmcdmefmghm _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 4 . defgefghfghi _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 5 . qxapxbqxa _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 6 . aduacuaeuabuafua _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 7 . mabmbcmcdm _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 8 . urtustuttu _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 9 . abyabxabwab _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 10 . rscdstdetuef _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 11 . npaoqapraqsa _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 12 . wxaxybyzczadab _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 13 . jkqrklrslmst _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 14 . pononmnmlmlk _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 15 . lmzmlymnx _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 16 . efsferfgq _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 17 . cdqdcpdeo _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 18 . ijwjivjku _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _",
    "chen2023novo": "Article De novo protein identi\ufb01cation in mammalian sperm using in situ cryoelectron tomography and AlphaFold2 docking Graphical abstract Highlights d In situ cryo - ET revealed native structures of doublet microtubules in mammalian sperm d Subtomogram averaging led to 6 - A\u02da reconstruction of microtubule doublets d Protein discovery by matching structures with the AlphaFold2 - predicted mouse proteome d Sperm doublets feature a plastic and partially redundant Tektin 5 network Authors Zhen Chen , Momoko Shiozaki , Kelsey M . Haas , . . . , Robyn M . Kaake , Ronald D . Vale , David A . Agard Correspondence zhen . chen @ ucsf . edu ( Z . C . ) , valer @ janelia . hhmi . org ( R . D . V . ) , david @ agard . ucsf . edu ( D . A . A . ) In brief Chen et al . reported the reconstruction of microtubule doublets in mammalian sperm using in situ cryoelectron tomography . Unbiased matching of sperm - speci\ufb01c structures with the AlphaFold2 library allowed de novo protein identi\ufb01cation of microtubule inner proteins . A transgenic mouse model revealed the functional importance and partial redundance of Tektin 5 in sperm . Chen et al . , 2023 , Cell 186 , 5041 \u2013 5053 November 9 , 2023 \u00aa 2023 The Authors . Published by Elsevier Inc . https : / / doi . org / 10 . 1016 / j . cell . 2023 . 09 . 017 ll Article De novo protein identi\ufb01cation in mammalian sperm using in situ cryoelectron tomography and AlphaFold2 docking Zhen Chen , 1 , 2 , * Momoko Shiozaki , 3 Kelsey M . Haas , 2 , 4 , 5 Will M . Skinner , 6 Shumei Zhao , 3 Caiying Guo , 3 Benjamin J . Polacco , 2 , 5 Zhiheng Yu , 3 Nevan J . Krogan , 2 , 4 , 5 Polina V . Lishko , 6 , 7 Robyn M . Kaake , 2 , 4 , 5 Ronald D . Vale , 2 , 3 , * and David A . Agard 1 , 5 , 8 , * 1 Department of Biochemistry and Biophysics , University of California , San Francisco , San Francisco , CA , USA 2 Department of Cellular and Molecular Pharmacology , University of California , San Francisco , San Francisco , CA , USA 3 Janelia Research Campus , Howard Hughes Medical Institute , Ashburn , VA , USA 4 J . David Gladstone Institutes , San Francisco , CA , USA 5 Quantitative Biosciences Institute ( QBI ) , University of California , San Francisco , San Francisco , CA , USA 6 Department of Molecular and Cell Biology , University of California , Berkeley , Berkeley , CA , USA 7 Present address : Department of Cell Biology and Physiology , Washington University School of Medicine in St . Louis , St . Louis , MO , USA 8 Lead contact * Correspondence : zhen . chen @ ucsf . edu ( Z . C . ) , valer @ janelia . hhmi . org ( R . D . V . ) , david @ agard . ucsf . edu ( D . A . A . ) https : / / doi . org / 10 . 1016 / j . cell . 2023 . 09 . 017 SUMMARY To understand the molecular mechanisms of cellular pathways , contemporary work\ufb02ows typically require multiple techniques to identify proteins , track their localization , and determine their structures in vitro . Here , we combined cellular cryoelectron tomography ( cryo - ET ) and AlphaFold2 modeling to address these questions and understand how mammalian sperm are built in situ . Our cellular cryo - ET and subtomogram averaging provided 6 . 0 - A\u02da reconstructions of axonemal microtubule structures . The well - resolved tertiary structures allowed us to unbiasedly match sperm - speci\ufb01c densities with 21 , 615 AlphaFold2 - predicted pro - tein models of the mouse proteome . We identi\ufb01ed Tektin 5 , CCDC105 , and SPACA9 as novel microtubule - associated proteins . These proteins form an extensive interaction network crosslinking the lumen of axonemal doublet microtubules , suggesting their roles in modulating the mechanical properties of the \ufb01la - ments . Indeed , Tekt5 (cid:1) / (cid:1) sperm possess more deformed \ufb02agella with 180 (cid:3) bends . Together , our studies pre - sented a cellular visual proteomics work\ufb02ow and shed light on the in vivo functions of Tektin 5 . INTRODUCTION Natural fertilization requires the rhythmic beating motion of sperm \ufb02agella to propel the cell toward the egg . 1 , 2 This coordi - nated and repetitive bending of sperm \ufb02agella relies on macro - molecular machinery to generate periodic force and endure me - chanical stresses . Genetic analyses of infertility have so far offered only an incomplete list of protein candidates in sperm . 1 Additionally , we currently lack high - resolution information of sperm macromolecular complexes to understand their assem - blies and functions at the molecular level . Eukaryotic motile cilia and \ufb02agella share a conserved \ufb01lamen - tous structure known as the axoneme , characterized by an over - all architecture of nine doublet microtubules ( doublets ) sur - rounding two singlet microtubules . 3 \u2013 6 These cytoskeletal \ufb01laments are extensively decorated externally and internally by proteins required for the various beating motions and the struc - tural integrity of \ufb02agella . 3 , 4 Notably , sperm from different species can differ substantially in their morphologies , functions ( e . g . , swimming behaviors ) , and genetics . 7 \u2013 9 In particular , mammalian sperm \ufb02agella are much longer and wider and must withstand larger bending torques compared with other motile cilia . 10 , 11 Despite the crucial roles of sperm axonemes in fertility and speciation , our understanding of their unique adaptations re - mains limited . Isolation of axonemal complexes from non - sperm motile cilia combined with single - particle cryoelectron microscopy ( cryo - EM ) analyses has led to high - resolution reconstructions ( better than4A\u02da ) . 12 \u2013 16 Therichstructural information onthetertiarystruc - tures and side chains has allowed con\ufb01dent assignments of pro - teinidentitiesintheEMreconstructions . However , carefuloptimi - zation of puri\ufb01cation strategies is required to avoid partial loss of components . 14 , 16 , 17 The truly intact structures are not guaran - teed by the end of puri\ufb01cation . On the other hand , direct visuali - zation of macromolecular complexes in mammalian sperm using cryogenic focused ion beam - scanning electron microscopy ( cryo - FIB - SEM ) and in situ cryoelectron tomography ( cryo - ET ) indicated that there are indeed mammalian sperm - speci\ufb01c ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 , November 9 , 2023 \u00aa 2023 The Authors . Published by Elsevier Inc . 5041 This is an open access article under the CC BY - NC - ND license ( http : / / creativecommons . org / licenses / by - nc - nd / 4 . 0 / ) . features . 18 \u2013 20 However , thecurrentreconstructionsfromcryo - ET subtomogram averaging are limited to (cid:4) 10 \u2013 20 A\u02da resolutions , mainly due to alignment inaccuracies . At such resolutions , ter - tiary structures of the proteins are rarely resolved , and multi - pro - tein complexes appear as blobs , making it challenging to deter - mine the identities of individual sperm proteins . Here , our in situ cryo - ET and subtomogram averaging has achieved up to 6 . 0 - A\u02da reconstructions of native microtubule structures in mouse and human sperm . The well - resolved ter - tiary structures in our cryo - EM maps allowed us to survey 21 , 615 AlphaFold2 - predicted protein models of the mouse pro - teome and unbiasedly identify matching ones . Such a visual pro - teomics approach helped us to discover novel microtubule - associated proteins in mammalian sperm , localize them in cells , and determine their native structures and interaction network without cell disruption and biochemical puri\ufb01cation . We also generated CRISPR - knockout mouse lines and showed that the newly identi\ufb01ed Tektin 5 is important for the structural integrity of the \ufb02agella . RESULTS In situ structures of sperm doublets at subnanometer resolutions Freshly extracted mouse sperm were treated with the dynein in - hibitor erythro - 9 - ( 2 - hydroxy - 3 - nonyl ) adenine ( EHNA ; 10 mM ) , which immediately stopped the beating motion of sperm \ufb02agella . 21 Subsequently , the inhibited sperm were vitri\ufb01ed on EM grids . To facilitate cryo - ET imaging , which is limited by sam - ple thickness , lamellae of (cid:4) 300 nm thickness were generated by cryo - FIB - SEM milling . Tilt series were recorded using a 300 kV Krios cryo transmission electron microscope ( cryo - TEM ) and a dose - symmetric scheme with a tilting increment of 4 (cid:3) . The 4 (cid:3) tilt increment , instead of the commonly used 1 (cid:3) \u2013 3 (cid:3) , 18 , 19 , 22 , 23 was used to improve the signal - to - noise ratio of each tilt image while keeping the similar angular ranges and total dose ( Fig - ure S1 ) . Three - dimensional classi\ufb01cation and re\ufb01nement of sub - tomograms corresponding to the 96 nm - repeating units were carried out as reported previously . 19 We then performed local re\ufb01nement on the 48 nm - repeating structures of doublets , focusing on the microtubules , aiming to reach the highest possible resolution ( see the work\ufb02ow shown in Figure S1 ) . The newly developed RELION4 was used to re\ufb01ne the 3D recon - structions and achieved 7 . 7 A\u02da overall resolution ( Fourier shell correlation [ FSC ] = 0 . 143 ) ( Figures 1A , S2A , and S2B ; see STAR Methods ) . 24 The improvement of resolution compared with RELION3 comes from more accurate contrast transfer func - tion ( CTF ) estimation and alignment of tilt series as these param - eters of each tilt image were iteratively re\ufb01ned relative to the 3D reconstructions ( Figure S2A ) . 24 Densities of microtubule inner proteins ( MIPs ) from mouse axonemes that repeat every 16 nm were observed despite of the overall periodicity of 48 nm ( Figures 1C \u2013 1E ) . In addition , we reprocessed a previously collected human sperm dataset and achieved 10 . 3 A \u02da for the 48 nm - repeating structures of the doublets ( FSC = 0 . 143 ) ( Figures 1B , S3A , and S3B ) . 19 Individual a - helices for the tubulins and MIPs are resolved in both maps . These maps were then compared with the published cryo - EM map of isolated doublets from bovine trachea reconstructed by single - particle cryo - EM , 13 which were low - pass \ufb01ltered to comparable resolutions of 7 . 5 and 10 A\u02da , respectively ( Figures S2C and S3C ) . Similar levels of detail in secondary and tertiary structures were resolved , vali - dating the resolution estimates of our 3D reconstructions . To our knowledge , these resolutions are currently the highest achieved by cryo - ET for any in situ axonemal structures ( 12 - A\u02da maps were reported previously for equivalent structures from Tetrahymena cilia 23 , 25 ) . Our 3D reconstructions of mouse and human sperm doublets reveal densities similar to the ones from bovine trachea dou - blets , 13 as well as sperm - speci\ufb01c densities ( colored densities in Figure 1 ) . Inside the A - tubule of mouse sperm axonemes , twelve helical bundles form a \ufb01lamentous core parallel to the lon - gitudinal axis , whereas only eight helical bundles , identi\ufb01ed as Tektins 1 \u2013 4 , are present in bovine trachea cilia ( Figures 1A and S2D ) . 13 Among the four mouse sperm - speci\ufb01c helical bundles , only one is a continuous 3 - helix bundle that runs along the entire length of the doublets ( Figures 1A and 1C ) . This continuous bundle is also observed in human sperm doublets ( Figures 1A \u2013 1C ) . The other three bundles , the two broken bundles and the curved bundles , all have breaks within the 48 - nm periodic struc - ture and appear different in mouse and human sperm ( Figures S2D and S3D \u2013 S3G ) . In particular , the two broken straight 3 - helix bundles have very low occupancy in the human sperm doublet ( Figures 1B , S3D , and S3E ) , while one of the curved helical bundles is connected to the microtubule lumen in human but not mouse sperm ( Figures S2D , S3F , and S3G ) . In the mouse sperm doublet , we also observed unique \u2018\u2018oblique\u2019\u2019 helical densities oriented (cid:4) 45 (cid:3) relative to the \ufb01lament axis and a globular domain next to it every 16 nm ( Figure 1C ) . These com - parisons indicate there are sperm - speci\ufb01c MIPs compared with mammalian trachea cilia and also diversi\ufb01cations among mammalian sperm in the A - tubule of the doublets . Outside the A - tubule , novel densities that are conserved in both mouse and human sperm doublets were resolved . A contin - uous 3 - helix bundle with multiple protrusions is situated at the external interface between A11 and A12 proto\ufb01laments , previ - ously named the \u2018\u2018ribbon\u2019\u2019 region of doublets ( Figures 1A , 1B , and 1D ) . 26 Inside the B - tubule , there are groups of 4 - helix bun - dles lining the inner surface of tubulins from the B4 to B9 proto - \ufb01laments . These 4 - helix bundles are stacked along the helical pitch of the microtubule , consistent with the previously reported striation density at lower resolutions ( Figure 1E ) . 19 Together , these data reveal that the mammalian sperm doublets have the most extensive MIP network of any microtubule structure observed to date . Although the B - tubule appears similar , mouse sperm have more MIPs than human sperm in the A - tubule . De novo protein identi\ufb01cation using AlphaFold2 The clearly resolved secondary and tertiary structures allowed us to interpret maps and build pseudo - atomic models . First , we were able to identify densities corresponding to the 29 MIPs observed in bovine trachea cilia ( Figure S4 ) , 13 suggesting their orthologs or homologs are likely present in sperm axo - nemes . We then sought to identify proteins contributing to the conserved sperm densities in mouse and human doublets ( high - lighted in Figures 1C \u2013 1E ) . Because most of these features repeat ll OPEN ACCESS 5042 Cell 186 , 5041 \u2013 5053 , November 9 , 2023 Article Figure 1 . The 3D reconstructions of mouse and human sperm doublets revealed novel MIPs ( A and B ) Transverse cross - section views of the doublets of mouse ( A ) and human ( B ) sperm . Conserved sperm MIP densities are highlighted ( pink , blue , and green ) , andthecorrespondingviewinganglesof ( C ) \u2013 ( E ) areindicated ( coloredarrowheads ) . The3 - helixdensitiesinA - tubulesharedwithbovinetracheadoublets ( EMD - 24664 ) are colored ( yellow ) . 13 Divergent sperm densities are also indicated ( red dashed shapes ) . Individual proto\ufb01laments of the doublets are labeled as A1 \u2013 A13 and B1 \u2013 B10 . ( C \u2013 E ) Zoom - in views of the conserved sperm MIP densities along the longitudinal axis . In ( C ) , mouse sperm - speci\ufb01c densities are indicated and labeled ( red dashed shapes ; see more in Figures S2 and S3 ) . In ( E ) , although the striations are 8 nm apart from one another , the overall periodicity is 48 nm . See also Figures S1 , S2 , and S3 . ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 , November 9 , 2023 5043 Article every 16 nm along the axoneme axis , we performed focused re\ufb01nement of the 16 nm - repeating structures in the A - and B - tubules of mouse doublets separately , and the larger number of subtomograms further improved the resolutions of recon - structions to 6 . 0 and 6 . 7 A \u02da , respectively ( Figure S5 ) . We then aimed to develop a general strategy to assign or narrow down the protein identities of the densities in the 6 \u2013 7 - A \u02da reconstructions . Unassigned densities from our maps were manually isolated and unbiasedly matched to the predicted models from the AlphaFold2 mouse proteome library ( 21 , 615 proteins ) using the COLORES program from the SITUS package ( Figure 2A ) . 27 , 28 The best poses for 21 , 615 mouse proteins were scored and ranked by the cross - correlation scores calculated by COLORES . 28 We tested this work\ufb02ow using densities corre - sponding to conserved single - domain MIPs ( CFAP20 and PACRG ) and a multi - domain MIP ( NME7 ) as controls . The cor - rect PDBs corresponding to the selected densities all came up as the top hits using this unbiased proteome - wide search ( Data S1 ) , indicating this visual proteomic approach can reliably identify proteins with matching tertiary structures . Figure 2 . De novo protein identi\ufb01cation of sperm MIPs assisted by AlphaFold2 ( A ) Conserved densities in mouse and human sperm were segmented from the averages of 16 - nm repeats of mouse sperm doublets and searched in the AlphaFold2 library of the mouse proteome ( 21 , 615 proteins ) . ( B ) The predicted structure of Tektin 5 based on AlphaFold2 was \ufb01tted into the continuous 3 - helix bundle . ( C ) Modeling of a complex formed by a full - length Tektin 5 and a truncated one ( N - Tekt 5 : aa 1 \u2013 149 ) using ColabFold . 32 ( D ) Fitting andmodelingof Tektin5s into the3 - helix bundle densities intheA - tubule . Thenearby densities accounted forby other proteins arealso shown ( yellow ribbon ) . ( E ) An unbiased search in the AlphaFold2 library identi\ufb01ed CCDC105 as the candidate for the continuous 3 - helix density at the ribbon . The three conserved proline - rich loops among CCDC105 orthologs could account for the protrusion densities but were not modeled ( see Figures S6C and S6D ) . ( F ) Modeling of a complex formed by a full - length CCDC105 and a truncated one ( N - CCDC105 : aa 1 \u2013 135 ) using ColabFold . 32 ( G ) Fitting and modeling of CCDC105 into the 3 - helix bundle density at the ribbon . The nearby densities accounted for by other proteins are also shown ( yellow ribbon ) . ( H ) The AlphaFold2 model for SPACA9 was directly \ufb01tted into the density and viewed from different angles . ( I ) TwoorthogonalviewsofthestriationsofSPAC9intheB - tubule . DifferentSPACA9moleculesarecoloredwithdifferentshadesofgreen . Theleftpanelshowed a particular striation indicated in the right panel ( the dashed rectangle ) . See also Figures S1 \u2013 S6 . ll OPEN ACCESS 5044 Cell 186 , 5041 \u2013 5053 , November 9 , 2023 Article For the continuous 3 - helix densities in the A - tubule , the best hit was Tektin 5 ( Figure S6A ) , previously found only in the mammalian testis and sperm in previous proteomic studies ( Figures 2B , S6A , and S6B ; see STAR Methods for details ) . 29 \u2013 31 Although Tektin 5 has no reported structure , AlphaFold2 pre - dicted that it possesses single - helix , 3 - helix , and 2 - helix seg - ments ( Figure 2B ) , 27 a tertiary structure that is almost identical to the ones reported for Tektin 1 \u2013 4 in bovine trachea cilia . 13 The ColabFold , an AlphaFold2 - based Google notebook , was then used to model how two copies of Tektin 5 molecules interact . 32 The resulting complexes suggest the single - helix N - terminal region of one Tektin 5 could interact with the 2 - helix C termini of the other molecule ( Figure 2C ) , indicating its poten - tial to self - polymerize and form a quasi - continuous 3 - helix bundle . Indeed , multiple Tektin 5 could be \ufb01tted into the contin - uous 3 - helix densities with 16 - nm periodicity , with minor adjustments of the orientations of individual a - helices of the orig - inal AlphaFold2 model ( Figure 2D ) . Upon manual inspection of the hit list , Tektin 1 \u2013 4 were also among the top 10 hits ( Figures S6A and S6B ) , and this \ufb01nding corroborated the robust - ness of our search method in \ufb01nding proteins with matching tertiary structures . We assigned these densities as Tektin 5 because it is uniquely present in mammalian sperm based on previous proteomic studies , 30 , 31 and such densities are absent in bovine trachea cilia that only contain Tektin 1 \u2013 4 . 13 For the continuous 3 - helix densities with protrusions at the ribbon , CCDC105 ( coiled - coil - domain - containing protein 105 ) was identi\ufb01ed in the top 20 hits from the unbiased search in AlphaFold2 mouse proteome library , along with Tektin 1 \u2013 5 ( see Data S2 for the top 30 hits ) . Previous proteomic studies revealed that CCDC105 was found in mammalian sperm and the testis but not other tissues . 29 \u2013 31 CCDC105 adopts a similar overall tertiary structure as Tektins 1 \u2013 5 based on AlphaFold2 prediction ( Fig - ure 2E ) , 27 suggesting it is a yet uncharacterized Tektin homolog . However , there are three proline - rich loops in CCDC105 that are uniquely conserved across CCDC105 orthologs ( Figures 2E , S6C , and S6D ) , 33 and they likely form structured loops like the ones observed in other axonemal complexes . 34 , 35 We also modeled how two copies of CCDC105 would interact using AlphaFold2 / ColabFold . 32 The predicted interface again involves coiled - coil interactions between the single - helix segment of one CCDC105 and the 2 - helix segment of the other ( Figure 2F ) . Furthermore , CCDC105 \ufb01ts well into the continuous 3 - helix densities at the ribbon , with their characteristic proline - rich loops matching the protrusions in our density maps ( Figure 2G ) . Notably , we could not swap the \ufb01tting of Tektin 5 and CCDC105 into these two 3 - helix bundles after extensive trials , mostly due to the different orientations and lengths of the a - helices ( Figure S6F ) . We also extracted the 4 - helix bundles at the B - tubule striations and performed an unbiased search against the AlphaFold2 library of the mouse proteome . Sperm acrosome - associated protein 9 ( SPACA9 ) was found to be the best hit ( Figures 2H and 2I ; see Data S3 for the top 30 hits ) . SPACA9 was previously found in various ciliated organs in humans ( testis , fallopian tubes , and lung ) using proteomics . 29 The tertiary fold of SPACA9 is so unique that no other homologous protein was found in the top 50 ranked structures from the unbiased search . Interestingly , no match was found when the search was done against the CATH library that curates non - redundant domains of published PDBs . 36 Thus , the capability of AlphaFold2 to pre - dict protein structures accurately , especially for the ones without published homologous structures , is critical to carrying out the unbiased proteome - wide survey . We next focused on the mouse sperm - speci\ufb01c densities . There are two 3 - helix bundles that appear to be similar to the ones formed by Tektins 1 \u2013 5 , apart from the discontinuous sec - tions ( Figures 3A and S2D ) . We also applied the unbiased search method to the slanted and curved helical bundles ( Figures 1C and S2D ) . Intriguingly , Tektin 1 \u2013 5 and CCDC105 were found to be among the top 30 hits in both cases , whereas no other PDB among the top 200 \ufb01ts better , albeit only parts of the structures are observed for the densities ( see Data S4 ) . Other hits among the top 200 do not match the secondary structures of the target densities upon visual inspections , suggesting Tektin 1 \u2013 5 and CCDC105 are the only proteins in mouse proteome that adopt such conformations . For the slanted helical densities , there is an additional a - helix connecting to the position where the missing single helix was expected to originate and is folded back by (cid:4) 180 (cid:3) ( Figures 3B and 3C ) . Interestingly , Tektin 5 , but not Tektin 1 \u2013 4 , has multiple conserved Gly residues among its orthologs at this turning region , making it plausible that Tektin 5 could adopt the bent - helix conformation . At a lower threshold , this bent helix is connected to a nearby globular domain that also repeats every 16 nm ( Figure S7A ) . The unbiased proteome - wide search suggests this globular domain matches the tertiary struc - ture of multiple dual speci\ufb01city phosphatase ( DUSP ) proteins ( DUSP 3 , 13 , 14 , 18 , 21 , and 29 ) ( Figures 3C and S7B ) . For the curved helical bundles , there are three 16 - nm groups of densities within every 48 - nm repeat ( Figures 3D and S2D ) . These densities can be explained by three modi\ufb01ed Tektin 5 mol - ecules , in which the two intermolecular interfaces near the two NME7s ( a previously known MIP shared with bovine trachea cilia ) are disrupted ( Figure 3D ) . The \ufb01rst and second Tektin 5s lack densities for the single - helix segment before the conserved glycine residues near Gly137 ( mouse ) , whereas the second and third Tektin 5s possess curved 2 - helix segments ( Figure 3D ) . Both modi\ufb01cations of Tektin 5s are necessary to avoid direct ste - ric clashes with the two NME7s , which adopt similar conforma - tions in bovine trachea and mouse sperm doublets . As curved bundles were not observed in the bovine trachea cilia that only contain Tektins 1 \u2013 4 , 13 we hypothesize that Tektin 5 has evolved to adopt multiple conformations and positions within sperm ax - onemes ( Figure 4E ) . In summary , the various helical conformers of Tektin 5 , together with the more uniform 3 - helix bundles of Tektin 1 \u2013 4 , are arranged with different polarities ( Figures 4A and 4B ) and ori - entations ( Figure 4C ) , forming the most extensive MIP network inside microtubules discovered to date . Mass spectrometry analyses of mouse sperm To further validate our de novo protein assignments , we used mass - spectrometry - based proteomics . Mouse sperm were iso - lated and extracted using salt buffers with increasing denaturing capabilities ( Extraction1 or E1 : 0 . 1 % Triton , E2 : NaCl , E3 : KSCN , E4 : urea , and E5 : 10 % SDS ) , and the extractions were analyzed ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 , November 9 , 2023 5045 Article using SDS - PAGE and western blotting ( Figure S8A ) . a - Tubulins could be detected in KSCN and urea extractions but not in others ( Figure S8B ) , suggesting the microtubule doublets are disas - sembled and the MIP candidates are likely present in these two extractions . After analyzing the E1 \u2013 E5 fractions by mass spectrometry ( MS ) , proteins with signi\ufb01cant changes in abun - dance between fractions were clustered into six distinct groups based on the correlation of intensity pro\ufb01le ( Figures S8C and S8D ; Tables S1 and S2 ) . 37 , 38 Gene ontology ( GO ) analyses sug - gest that cluster 4 is enriched for proteins involved in cilium , cilium assembly , cytoskeleton , and the axoneme and shows increased intensities in fraction E3 , whereas cluster 6 is enriched for cilia assembly and shows increased intensities in fraction E4 ( Figure S8E ; Table S3 ) . 39 The overall change in protein Figure 3 . Conformational plasticities of Tektin 5 ( A ) The two broken 3 - helix bundles could be explained by two complete and a third partial copies of Tektin 5 ( dashed rectangles ) per 48 - nm repeat , instead of three Tektins in the continuous 3 - helix bundle . ( B ) TheAlphaFold2modelof mouseTektin 5was\ufb01tted into theslantedhelicaldensities . Sequence alignmentofTektin5from M . musculus , H . sapiens , B . taurus , and F . catus is shown from Q133 to F151 ( the numbering of amino acids is based on M . musculus Tektin 5 ) . The conserved Gly137 , Gly143 , and Gly150 are near the turning point of the bent a - helix . ( C ) The \ufb01tting of Tektin 5 and DUSP3 protein ( its homologs are also possible candidates ) into the 16 nm - repeating features ; see the same view of the map in Figure 1C . ( D ) Three modi\ufb01ed Tektin 5 were \ufb01tted into the densities of the curved bundle in the mouse sperm doublet ( as indicated in Figure 1A ) . The intact intermolecular interaction interface , N termini of the Tektin 5s , and curved 2 - helix segments are indicated ( arrows ) . Nearby MIPs shared between mouse sperm \ufb02agella and bovine trachea cilia are also colored and labeled ( NME7 , CFAP161 , and SPAG8 ) . ( E ) The cross - section schematic is shown . The highlighted models of ( A ) \u2013 ( D ) are indicated using arrows . See also Figures S1 , S2 , S4 , S6 , S7 , and S8 . ll OPEN ACCESS 5046 Cell 186 , 5041 \u2013 5053 , November 9 , 2023 Article abundance is consistent with the idea that E3 and E4 buffers ex - tracted microtubule - associated proteins in the axonemes . Indeed , 28 of the 29 previously identi\ufb01ed MIPs in bovine trachea doublets were reproducibly identi\ufb01ed in all three biological repli - cates of mouse sperm extractions ( Table S4 ) . The almost com - plete list of MIPs highlighted the coverage of our biochemical and MS analyses . Importantly , SPACA9 was reproducibly iden - ti\ufb01ed in fraction E3 , whereas Tektin 1 \u2013 5 and CCDC105 were reproducibly identi\ufb01ed in fractions E3 and E4 , with high protein intensity and a range of 3 to 34 unique peptide identi\ufb01cations per replicate ( Figure S8F ; Table S4 ) . Only one DUSP protein , DUSP3 , was identi\ufb01ed in two of three replicates in fraction E4 fraction ( Figure S8F ; Table S4 ) . However , additional analyses are required to identify these proteins . Moreover , AlphaFold2 models for the other candidates from fractions E1 \u2013 E5 were in - spected , but no additional candidates could explain the various densities of helical bundles described above . Tektin 5 strengthens the \ufb02agella and is partially redundant In order to analyze the functions of Tektin 5 , we generated knockout mice carrying null alleles of Tekt5 using CRISPR tech - nologies . To our surprise , the F2 homozygous knockout males are still fertile when they are mated with the wild - type ( WT ) fe - males ( litter size : 7 . 3 \u00b1 1 . 4 , n = 6 mating trials ) , suggesting some levels of functional redundancy . However , the sperm ex - tracted from the mutant males have a lower fraction of motile cells compared with WT controls ( 64 % \u00b1 3 % vs . 77 % \u00b1 4 % ) and a higher percentage of defective \ufb02agella with 180 (cid:3) bends ( 30 % \u00b1 3 % vs . 13 % \u00b1 3 % ) ( Figures 5A and 5B ) , suggesting the mechanical integrity of the cellular structures inside \ufb02agella is compromised in the mutant sperm . We then analyzed the doublet structure of the mutant sperm and compared it with the WT counterpart . We \ufb01rst focused on the various densities that were assigned to be Tektin 5 ( as shown in Figure 3 ) . The continuous 3 - helix bundle remains in the mutant Figure 4 . Sperm doublets are composed of microtubules and extensive coiled - coil bun - dles ( A ) The plus and minus ends of Tektin 5 were named based on the N and C termini of the protein . ( B ) The cross - section view of the mouse sperm doublets shows the polarities of 3 - helix bundles pointing toward the readers . ( C ) The orientations for each 3 - helix bundle were represented by a vector starting from the middle point of the2 - helix segment and pointingtoward the single - helix segment of the other Tektin molecule . ( Figure 5C ) , suggesting a Tektin homolog could substitute on the same position in the absence of Tektin 5 . By contrast , the occupancy of densities corresponding to broken and curved helical bundles is much lower compared with the surround - ing proteins , such as the tubulins , in the mutant sperm ( Figures 5D and 5E ) , whereas in WT sperm the occupancies are comparable ( Figures 3C and 3D ) . These comparisons suggest the compensation for the lack of Tektin 5 by other Tektin homologs is low at these sites . For the slanted helical densities that repeat every 16 nm in WT sperm , we observed that two of the slanted helical densities are partially occupied while the last one is almost absent ( Fig - ure 5C ) . Interestingly , the densities corresponding to the DUSP domains next to the slanted helical densities were barely resolved , suggesting the lack of Tektin 5 would decrease the recruitment of the neighboring MIP . We did not observe addi - tional differences in densities corresponding to other MIPs . Together , these results suggest that Tektin homologs could partially re\ufb01ll docking sites of Tektin 5 in the mutant sperm . DISCUSSION Our in situ cryo - ET studies have provided high - resolution recon - structions of native microtubule structures within mouse and hu - man sperm , enabling us to identify Tektin 5 , SPACA9 , and CCDC105 as novel components in sperm doublets . Alignment of gold beads has traditionally been the method of choice to align tomographic tilt series , but the positions of gold beads undergo heterogeneous motions due to the sample defor - mation induced by the electron beam during imaging . 40 The sig - ni\ufb01cant improvement in resolutions made possible by RELION4 ( Figures S1 and S2A ) highlights the bene\ufb01ts of aligning 3D recon - structions of protein complexes with the individual 2D projection views on different tilt images to re\ufb01ne tilt series alignment . 24 Thus , new methods of aligning the molecular features directly from tilt series while considering their local motions have the po - tential to further improve the initial alignment of tilt series and set an even better starting point for subtomogram averaging . 41 The application of AlphaFold2 has facilitated structural modeling based on cryo - EM reconstructions below 10 A \u02da , where secondary structures are resolved . 42 However , most studies have focused on protein complexes with known components ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 , November 9 , 2023 5047 Article identi\ufb01ed through other approaches such as MS analyses of puri\ufb01ed complexes . However , this information may not be readily available in other less - studied cell biology systems . Our studies provided the \ufb01rst demonstration that high - resolution cellular cryo - ET combined with unbiased proteome - wide searches could identify previously unknown components of cellular complexes in their native context . This integrative struc - tural modeling approach offers a powerful alternative to the con - ventional genetic and cell biology approaches to identify protein components and localize them inside cells . The comparison of MIPs in WT and mutant sperm doublets with bovine tracheal doublets suggests that the assembling of MIPs is modular , and novel MIPs discovered in this study ( SPACA9 , CCDC105 , Tektin 5 , and DUSP ) are recruited after the commonly shared MIPs ( common MIPs ) . First , homologs or orthologs of MIPs identi\ufb01ed from bovine tracheal doublets Figure 5 . Characterization of mutant Tekt5 (cid:1) / (cid:1) sperm ( A ) The percentages of motile sperm from wild - type and Tekt5 knockout mice ( > 200 cells were counted for each mouse , three knockout (cid:1) / (cid:1) mice and two wild - type mice were analyzed , and the pool percentage and 95 % con\ufb01dence intervals [ Wilson / Brown method ] were shown ) . ( B ) The percentages of bent sperm from wild - type and Tekt5 knockout mice ( > 200 cells were counted for each mouse , three knockout (cid:1) / (cid:1) mice and two wild - type mice were analyzed , and the pool percentage and 95 % con\ufb01dence intervals by Wilson / Brown method were shown ) . Two examples of bent sperm are shown . ( C ) An overlay of wild - type models with the densities of Tekt5 (cid:1) / (cid:1) sperm around the slanted bundles . The continuous 3 - helix bundle assigned as Tektin 5 ( high occupancies ) and slanted helical bundles ( low occupancies ) are shown . The densities corresponding to the DUSP proteins are barely resolved . Note that there are substantially less densities for these models compared with Figure 3C . ( D ) An overlay of wild - type models with the densities of Tekt5 (cid:1) / (cid:1) sperm around the curved bundles . The occupancies of the curved bundles are lower than the other MIPs and tubulins . Note that there are substantially less densities for these models compared with Figure 3D . ( E ) The twobroken 3 - helix bundles have lower occupancies compared withthesurroundingMIPs and tubulins . Notethatthere aresubstantially less densities for these models compared with Figure 3A . ll OPEN ACCESS 5048 Cell 186 , 5041 \u2013 5053 , November 9 , 2023 Article are all present in WT and mutant mouse sperm , and they adopt indistinguishable conformations ( Figure S4 ) . 13 Second , SPACA9 and CCDC105 crosslink the tubulin dimers only at the exposed lumen sites observed in tracheal doublets , without displacing any common MIPs ( Figures 2G , 2I , and S4 ) . Interest - ingly , the remarkable conformational plasticity of Tektin 5 is partially molded by the doublets and common MIPs , as shown by the bent single helix that would otherwise clash with the microtubule wall ( Figures 3B and 3C ) , as well as the missing single helix and curved 2 - helix segment that would otherwise clash with NME7 , a conserved common MIP ( Figure 3D ) . Lastly , knockout of Tektin 5 decreases the occupancies of the bent Tektin 5 and the nearby DUSP densities ( Figure 5C ) . Such de - pendency or modularity could be further tuned during evolu - tion , as we observed that both the bent Tektin 5 and DUSP densities are absent in human sperm doublets . Bending of the axonemes would stress the nine microtubule \ufb01laments in nine different directions , and mammalian axonemes have to bend to various directions to generate the 3D beating waveforms . 43 The non - uniform arrangement of helical bundles in sperm doublets could be built to reinforce the doublets to with - stand mechanical stress from different directions ( Figures 4 and 6 ) . From a structural perspective , the intramolecular and inter - molecular coiled - coil interaction interfaces of Tektins are parallel to the microtubule axis so that bending would not expand the gap of these interfaces ( Figure 6A ) . Instead , the bending force would distort the straight helical bundles and the ideal bond an - gles / lengths . The transition of releasing such molecular strains could provide a restoring force that allows the curved \ufb01lament to return to the straight conformation . By contrast , the interface between tubulin dimers along the proto\ufb01laments is at a plane perpendicular to the \ufb01lament axis , and bending would open the interface and lower the af\ufb01nity and the potential restoring force . Therefore , the helical bundles are arranged at different an - gles to provide an effective means to bear the bending force , sta - bilizing the axoneme and \ufb02agella ( Figure 6B ) . From a functional perspective , we discovered that mutant mice lacking Tektin 5 have more deformed sperm \ufb02agella , yet they remain fertile . This observation is consistent with previous studies on knockout mouse models lacking Tektin - 3 or Tektin - 4 , which also showed deformed sperm but normal fertility , 44 , 45 underscoring the functional redundancy of Tektins . Our structural analyses of Tekt5 (cid:1) / (cid:1) sperm doublets uncov - ered the compensatory mechanisms of Tektin 5 at the molecu - lar level . Moreover , we discovered that different Tektin 5 con - formers were compensated to varying extents , likely due to the differential dependencies of the distinctive interaction interfaces . The human sperm doublets revealed that the densities formed by Tektin 5 are signi\ufb01cantly less resolved compared with the other MIPs ( Figure 1 ) , adopt different conformations ( Figure S3 F ) , or are completely absent compared with mouse sperm dou - blets ( Figure 1C ) . This is also consistent with the idea that the bundles formed by Tektin 5 are partially redundant and plastic during evolution so that some extent of degeneration is tolerable . Still , such degeneration would be selected against in the wild , particularly in species where sperm from multiple males in the fe - male reproductive tracts competing for fertilization . Our studies combined high - resolution in situ cryo - ET and AlphaFold2 modeling and de\ufb01ned a visual proteomic approach of precisely placing proteins in their native cellular environment without the need for labeling , cellular disruption , or puri\ufb01cation . This work\ufb02ow has allowed us to uncover the cellular locations and interaction networks of several MIPs , providing insights into how they contribute to the mechanics of \ufb02agellar bending . Moreover , this visual proteomic work\ufb02ow could potentially be applied to other cell biology problems , such as membrane re - modeling by viruses and identi\ufb01cation of their in situ interactors . Limitations of the study The majority of the densities in sperm doublets feature well - de\ufb01ned domains that could be isolated and identi\ufb01ed using Figure 6 . Coiled - coil interfaces are suitable to withstand mechanical stress from orthogonal directions ( A ) A model of how 3 - helix bundles would be able to bear mechanical stress differently compared with the microtubules . The bending curvatures and gaps are exaggerated for illustration purposes . ( B ) A schematic of wild - type and mutant sperm doublet structures highlighting the Tektin 5 bundles and the partial redundancy . ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 , November 9 , 2023 5049 Article our visual proteomics approach . However , it is possible that there are unknown MIPs that are composed of coiled coils without substantial intramolecular interactions . These proteins may adopt conformations that are determined by intermolec - ular interactions in the context of native complexes , which are not accounted for by AlphaFold2 predictions . Furthermore , map segregation for these types of proteins could be chal - lenging at (cid:4) 6 \u2013 10 - A\u02da resolutions . Improvement of resolutions for cellular cryo - ET is also desirable to resolve the side - chain densities in the EM reconstructions and distinguish the proteins with similar tertiary structures . While this manuscript was under revision , two single - particle cryo - EM studies of splayed mammalian sperm doublets were reported . 46 , 47 The single - par - ticle cryo - EM reconstructions of microtubule doublets from mouse and bovine sperm are indistinguishable from our cryo - ET reconstruction at our resolutions , suggesting that the sperm doublets are robust enough to withstand gentle biochemical treatments . Importantly , their assignments of protein identity based on side - chain densities are consistent with our identi\ufb01 - cation of CCDC105 , SPACA9 , and the different forms of Tektin 5 , validating our cellular visual proteomics approach for de novo protein discovery . In the future , a more systematic characterization of MIPs using transgenic mice will be needed to elucidate the functions of individual MIPs . Additionally , ge - netic analyses of patients with infertility are likely to identify more essential and redundant components relevant to sperm functions . STAR + METHODS Detailed methods are provided in the online version of this paper and include the following : d KEY RESOURCES TABLE d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS B Mouse models B Human sample d METHOD DETAILS B Sample preparation B Grid preparation B Cryogenic focused ion beam ( cryoFIB ) milling B Image acquisition and tomogram reconstruction B Subvolume averaging B Model building and unbiased matching of density maps to protein candidates B Sequence alignment and search for homologous proteins B Biochemical extractions of mouse sperm B Mass spectrometry ( MS ) - based global protein abun - dance of mouse sperm B Generation of Tektin - 5 knockout mice and functional / - structural analyses of mutant sperm d QUANTIFICATION AND STATISTICAL ANALYSIS SUPPLEMENTAL INFORMATION Supplementalinformationcanbefoundonlineathttps : / / doi . org / 10 . 1016 / j . cell . 2023 . 09 . 017 . ACKNOWLEDGMENTS We are grateful to members of the Agard and Vale laboratories for the discus - sions and critical reading of the manuscript . We thank Xiaowei Zhao , Shixin Yang , and Rui Yan from the CryoEM Facility at the Janelia Research Campus for their assistance with data collection . We thank Zanlin Yu and Hao Wu at UCSF for their suggestions on sample processing and model building . We thank Garrett Greenan , Shawn Zheng , and Sam Li at UCSF for discussions on cryo - ET data processing . We thank Willy Wrigger from Old Dominion Uni - versity for his input and suggestions on the SITUS package . EM data process - ingutilizedcomputingresourcesatboththeworkstationsattheCryoEMFacil - ity at the Janelia Research Campus and the UCSF HPC Wynton cluster . We also thank David Bulkley , Glenn Gilbert , and Matt Harrington from the UCSF cryo - EM facility for their discussion on data collection and processing . We also thank Colin Morrow , Gillian Harris , Crystall Lopez , and Catherine Lindsey from Janelia Vivarium for mouse experiments . Z . C . was supported by the Helen Hay Whitney Foundation Postdoctoral Fellowship . W . M . S . was sup - ported by the National Science Foundation Graduate Research Fellowship ProgramundergrantnumbersDGE1752814andDGE2146752 . Anyopinions , \ufb01ndings , and conclusions or recommendations expressed in this material are those of the author ( s ) and do not necessarily re\ufb02ect the views of the National Science Foundation . P . V . L . received funding from a Pew Biomedical Scholars award and a GCRLE grant from the Global Consortium for Reproductive Longevity and Equality made possible by the Bia - Echo Foundation . D . A . A . received funding from NIH R35GM118099 . R . D . V . received funding from NIH R35GM118106 and the Howard Hughes Medical Institute . The UCSF cryo - EM facility was supported by NIH instrumentation grants 1S10OD026881 , 1S10OD020054 , and 1S10OD021741 . AUTHOR CONTRIBUTIONS Conceptualization , Z . C . , R . D . V . , and D . A . A . ; mouse sample preparation , S . Z . , C . G . , and W . M . S . ; cryo - EM sample preparation , Z . C . and M . S . ; data process - ing , Z . C . ; biochemical extraction , Z . C . ; mass spectroscopy , K . M . H . , R . M . K . , B . J . P . , and N . J . K . ; sperm analyses , W . M . S . , Z . C . , and P . V . L . ; writing \u2013 original draft , Z . C . , K . M . H . , and R . M . K . ; writing \u2013 review & editing , Z . C . , R . D . V . , and D . A . A . DECLARATION OF INTERESTS We declare that one or more authors have a competing interest as de\ufb01ned by Nature Portfolio . The Krogan Laboratory has received research support from Vir Biotechnology , F . Hoffmann - La Roche , and Rezo Therapeutics . N . J . K . has previously held \ufb01nancially compensated consulting agreements with the Icahn School of Medicine at Mount Sinai , New York , and Twist Bioscience Corp . He currently has \ufb01nancially compensated consulting agreements with Maze Therapeutics , Interline Therapeutics , Rezo Therapeutics , and GEn1E Lifesciences , Inc . He is on the Board of Directors of Rezo Therapeutics and is a shareholder in Tenaya Therapeutics , Maze Therapeutics , Rezo Therapeu - tics , and Interline Therapeutics . INCLUSION AND DIVERSITY We support inclusive , diverse , and equitable conduct of research . Received : October 19 , 2022 Revised : August 2 , 2023 Accepted : September 16 , 2023 Published : October 20 , 2023 ll OPEN ACCESS 5050 Cell 186 , 5041 \u2013 5053 , November 9 , 2023 Article REFERENCES 1 . Sironen , A . , Shoemark , A . , Patel , M . , Loebinger , M . R . , andMitchison , H . M . ( 2020 ) . Sperm defects in primary ciliary dyskinesia and related causes of male infertility . Cell . Mol . Life Sci . 77 , 2029 \u2013 2048 . https : / / doi . org / 10 . 1007 / s00018 - 019 - 03389 - 7 . 2 . Ferna\u00b4ndez - Lo\u00b4pez , P . , Garriga , J . , Casas , I . , Yeste , M . , and Bartumeus , F . ( 2022 ) . Predicting fertility from sperm motility landscapes . Commun . Biol . 5 , 1027 . https : / / doi . org / 10 . 1038 / s42003 - 022 - 03954 - 0 . 3 . Ishikawa , T . ( 2017 ) . Axonemestructurefrommotilecilia . ColdSpringHarb . Perspect . Biol . 9 , a028076 . https : / / doi . org / 10 . 1101 / cshperspect . a028076 . 4 . Linck , R . W . , Chemes , H . , andAlbertini , D . F . ( 2016 ) . Theaxoneme : thepro - pulsive engine of spermatozoa and cilia and associated ciliopathies lead - ing to infertility . J . Assist . Reprod . Genet . 33 , 141 \u2013 156 . https : / / doi . org / 10 . 1007 / s10815 - 016 - 0652 - 1 . 5 . Fawcett , D . W . ( 1975 ) . The mammalian spermatozoon . Dev . Biol . 44 , 394 \u2013 436 . https : / / doi . org / 10 . 1016 / 0012 - 1606 ( 75 ) 90411 - x . 6 . Beeby , M . , Ferreira , J . L . , Tripp , P . , Albers , S . V . , and Mitchell , D . R . ( 2020 ) . Propulsive nanomachines : the convergent evolution of archaella , \ufb02agella and cilia . FEMS Microbiol . Rev . 44 , 253 \u2013 304 . https : / / doi . org / 10 . 1093 / femsre / fuaa006 . 7 . Pitnick , S . , Hosken , D . J . , and Birkhead , T . R . ( 2009 ) . Sperm morphological diversity . In In Sperm Biology\u2014An Evolutionary Perspective , Second Edi - tion ( Academic Press ) , pp . 69 \u2013 149 . 8 . Lindemann , C . B . , and Lesich , K . A . ( 2021 ) . The many modes of \ufb02agellar and ciliary beating : insights from a physical analysis . Cytoskeleton ( Hobo - ken ) 78 , 36 \u2013 51 . https : / / doi . org / 10 . 1002 / cm . 21656 . 9 . Civetta , A . ( 2003 ) . Positive selection within sperm - egg adhesion domains offertilin : anADAMgenewithapotentialroleinfertilization . Mol . Biol . Evol . 20 , 21 \u2013 29 . https : / / doi . org / 10 . 1093 / molbev / msg002 . 10 . Lindemann , C . B . ( 1996 ) . Functional signi\ufb01cance of the outer dense \ufb01bers ofmammalianspermexaminedbycomputersimulationswiththegeomet - ric clutch model . Cell Motil . Cytoskeleton 34 , 258 \u2013 270 . https : / / doi . org / 10 . 1002 / ( SICI ) 1097 - 0169 ( 1996 ) 34 : 4 < 258 : : AID - CM1 > 3 . 0 . CO ; 2 - 4 . 11 . Lindemann , C . B . , and Lesich , K . A . ( 2016 ) . Functional anatomy of the mammalian sperm \ufb02agellum . Cytoskeleton ( Hoboken ) 73 , 652 \u2013 669 . https : / / doi . org / 10 . 1002 / cm . 21338 . 12 . Ma , M . , Stoyanova , M . , Rademacher , G . , Dutcher , S . K . , Brown , A . , and Zhang , R . ( 2019 ) . Structure of the decorated ciliary doublet microtubule . Cell 179 , 909 \u2013 922 . e12 . https : / / doi . org / 10 . 1016 / j . cell . 2019 . 09 . 030 . 13 . Gui , M . , Farley , H . , Anujan , P . , Anderson , J . R . , Maxwell , D . W . , Whitchurch , J . B . , Botsch , J . J . , Qiu , T . , Meleppattu , S . , Singh , S . K . , et al . ( 2021 ) . De novo identi\ufb01cation of mammalian ciliary motility proteins using cryo - EM . Cell 184 , 5791 \u2013 5806 . e19 . https : / / doi . org / 10 . 1016 / j . cell . 2021 . 10 . 007 . 14 . Khalifa , A . A . Z . , Ichikawa , M . , Dai , D . , Kubo , S . , Black , C . S . , Peri , K . , McAlear , T . S . , Veyron , S . , Yang , S . K . , Vargas , J . , et al . ( 2020 ) . The inner junction complex of the cilia is an interaction hub that involves tubulin post - translational modi\ufb01cations . eLife 9 , e52760 . https : / / doi . org / 10 . 7554 / eLife . 52760 . 15 . Grossman - Haham , I . ( 2023 ) . Towards an atomic model of a beating ciliary axoneme . Curr . Opin . Struct . Biol . 78 , 102516 . https : / / doi . org / 10 . 1016 / j . sbi . 2022 . 102516 . 16 . Kubo , S . , Black , C . S . , Joachimiak , E . , Yang , S . K . , Legal , T . , Peri , K . , Kha - lifa , A . A . Z . , Ghanaeian , A . , McCafferty , C . L . , Valente - Paterno , M . , et al . ( 2023 ) . Native doublet microtubules from Tetrahymena thermophila reveal theimportanceofouterjunctionproteins . Nat . Commun . 14 , 2168 . https : / / doi . org / 10 . 1038 / s41467 - 023 - 37868 - 0 . 17 . Ichikawa , M . , Khalifa , A . A . Z . , Kubo , S . , Dai , D . , Basu , K . , Maghrebi , M . A . F . , Vargas , J . , and Bui , K . H . ( 2019 ) . Tubulin lattice in cilia is in a stressed form regulated by microtubule inner proteins . Proc . Natl . Acad . Sci . USA 116 , 19930 \u2013 19938 . https : / / doi . org / 10 . 1073 / pnas . 1911119116 . 18 . Leung , M . R . , Roelofs , M . C . , Ravi , R . T . , Maitan , P . , Henning , H . , Zhang , M . , Brom\ufb01eld , E . G . , Howes , S . C . , Gadella , B . M . , Bloom\ufb01eld - Gade\u02c6lha , H . , and Zeev - Ben - Mordehai , T . ( 2021 ) . Themulti - scalearchitectureofmammalian sperm \ufb02agella and implications for ciliary motility . EMBO J . 40 , e107410 . https : / / doi . org / 10 . 15252 / embj . 2020107410 . 19 . Chen , Z . , Greenan , G . A . , Shiozaki , M . , Liu , Y . , Skinner , W . , Zhao , X . , Zhao , S . , Yan , R . , Yu , Z . , Lishko , P . V . , et al . ( 2023 ) . In situ cryo - electron tomog - raphy reveals the asymmetric architecture of mammalian sperm axo - nemes . Nat . Struct . Mol . Biol . 30 , 360 \u2013 369 . 20 . Gadadhar , S . , Alvarez Viar , G . , Hansen , J . N . , Gong , A . , Kostarev , A . , Ialy - Radio , C . , Leboucher , S . , Whit\ufb01eld , M . , Ziyyat , A . , Toure\u00b4 , A . , et al . ( 2021 ) . Tubulin glycylation controls axonemal dynein activity , \ufb02agellar beat , and male fertility . Science 371 , eabd4914 . https : / / doi . org / 10 . 1126 / science . abd4914 . 21 . Bouchard , P . , Penningroth , S . M . , Cheung , A . , Gagnon , C . , and Bardin , C . W . ( 1981 ) . erythro - 9 - [ 3 - ( 2 - Hydroxynonyl ) ] adenine is an inhibitor of sperm motility that blocks dynein ATPase and protein carboxylmethylase activities . Proc . Natl . Acad . Sci . USA 78 , 1033 \u2013 1036 . https : / / doi . org / 10 . 1073 / pnas . 78 . 2 . 1033 . 22 . Nicastro , D . , Schwartz , C . , Pierson , J . , Gaudette , R . , Porter , M . E . , and McIntosh , J . R . ( 2006 ) . The molecular architecture of axonemes revealed by cryoelectron tomography . Science 313 , 944 \u2013 948 . https : / / doi . org / 10 . 1126 / science . 1128618 . 23 . Li , S . , Fernandez , J . J . , Fabritius , A . S . , Agard , D . A . , and Winey , M . ( 2022 ) . Electron cryo - tomography structure of axonemal doublet microtubule from Tetrahymena thermophila . Life Sci . Alliance 5 , e202101225 . https : / / doi . org / 10 . 26508 / lsa . 202101225 . 24 . Zivanov , J . , Oto\u00b4n , J . , Ke , Z . , vonKu\u00a8gelgen , A . , Pyle , E . , Qu , K . , Morado , D . , Castan\u02dco - D\u0131\u00b4ez , D . , Zanetti , G . , Bharat , T . A . M . , et al . ( 2022 ) . A Bayesian approach to single - particle electron cryo - tomography in RELION - 4 . 0 . eL - ife 11 , e83724 . https : / / doi . org / 10 . 7554 / eLife . 83724 . 25 . Song , K . , Shang , Z . , Fu , X . , Lou , X . , Grigorieff , N . , and Nicastro , D . ( 2020 ) . In situ structure determination at nanometer resolution using TYGRESS . Nat . Methods 17 , 201 \u2013 208 . https : / / doi . org / 10 . 1038 / s41592 - 019 - 0651 - 0 . 26 . Linck , R . , Fu , X . , Lin , J . , Ouch , C . , Schefter , A . , Steffen , W . , Warren , P . , and Nicastro , D . ( 2014 ) . Insights into the structure and function of ciliary and \ufb02agellar doublet microtubules : tektins , Ca2 + - binding proteins , and stable proto\ufb01laments . J . Biol . Chem . 289 , 17427 \u2013 17444 . https : / / doi . org / 10 . 1074 / jbc . M114 . 568949 . 27 . Jumper , J . , Evans , R . , Pritzel , A . , Green , T . , Figurnov , M . , Ronneberger , O . , Tunyasuvunakool , K . , Bates , R . , (cid:1) Z\u0131\u00b4dek , A . , Potapenko , A . , et al . ( 2021 ) . Highly accurate protein structure prediction with AlphaFold . Nature 596 , 583 \u2013 589 . https : / / doi . org / 10 . 1038 / s41586 - 021 - 03819 - 2 . 28 . Wriggers , W . , Milligan , R . A . , and McCammon , J . A . ( 1999 ) . Situs : A pack - age for docking crystal structures into low - resolution maps from electron microscopy . J . Struct . Biol . 125 , 185 \u2013 195 . https : / / doi . org / 10 . 1006 / jsbi . 1998 . 4080 . 29 . Uhle\u00b4n , M . , Fagerberg , L . , Hallstro\u00a8m , B . M . , Lindskog , C . , Oksvold , P . , Mar - dinoglu , A . , Sivertsson , A\u02da . , Kampf , C . , Sjo\u00a8stedt , E . , Asplund , A . , et al . ( 2015 ) . Proteomics . Tissue - based map of the human proteome . Science 347 , 1260419 . https : / / doi . org / 10 . 1126 / science . 1260419 . 30 . Baker , M . A . , Hetherington , L . , Reeves , G . M . , and Aitken , R . J . ( 2008 ) . The mouse sperm proteome characterized via IPG strip prefractionation and LC - MS / MS identi\ufb01cation . Proteomics 8 , 1720 \u2013 1730 . https : / / doi . org / 10 . 1002 / pmic . 200701020 . 31 . Firat - Karalar , E . N . , Sante , J . , Elliott , S . , and Stearns , T . ( 2014 ) . Proteomic analysis of mammalian sperm cells identi\ufb01es new components of the centrosome . J . Cell Sci . 127 , 4128 \u2013 4133 . https : / / doi . org / 10 . 1242 / jcs . 157008 . 32 . Mirdita , M . , Schu\u00a8tze , K . , Moriwaki , Y . , Heo , L . , Ovchinnikov , S . , and Stei - negger , M . ( 2022 ) . ColabFold : makingproteinfoldingaccessibletoall . Nat . Methods 19 , 679 \u2013 682 . https : / / doi . org / 10 . 1038 / s41592 - 022 - 01488 - 1 . ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 , November 9 , 2023 5051 Article 33 . Madeira , F . , Pearce , M . , Tivey , A . R . N . , Basutkar , P . , Lee , J . , Edbali , O . , Madhusoodanan , N . , Kolesnikov , A . , and Lopez , R . ( 2022 ) . Search and sequence analysis tools services from EMBL - EBI in 2022 . Nucleic Acids Res . 50 , W276 \u2013 W279 . https : / / doi . org / 10 . 1093 / nar / gkac240 . 34 . Han , L . , Rao , Q . , Yang , R . , Wang , Y . , Chai , P . , Xiong , Y . , and Zhang , K . ( 2022 ) . Cryo - EM structure of an active central apparatus . Nat . Struct . Mol . Biol . 29 , 472 \u2013 482 . https : / / doi . org / 10 . 1038 / s41594 - 022 - 00769 - 9 . 35 . Rupp , G . , O\u2019Toole , E . , and Porter , M . E . ( 2001 ) . The Chlamydomonas PF6 locus encodes a large alanine / proline - rich polypeptide that is required for assembly of a central pair projection and regulates \ufb02agellar motility . Mol . Biol . Cell 12 , 739 \u2013 751 . https : / / doi . org / 10 . 1091 / mbc . 12 . 3 . 739 . 36 . Sillitoe , I . , Bordin , N . , Dawson , N . , Waman , V . P . , Ashford , P . , Scholes , H . M . , Pang , C . S . M . , Woodridge , L . , Rauer , C . , Sen , N . , et al . ( 2021 ) . CATH : increased structural coverage of functional space . Nucleic Acids Res . 49 , D266 \u2013 D273 . https : / / doi . org / 10 . 1093 / nar / gkaa1079 . 37 . Cox , J . , and Mann , M . ( 2008 ) . MaxQuant enables high peptide identi\ufb01ca - tion rates , individualized p . p . b . - range mass accuracies and proteome - wide protein quanti\ufb01cation . Nat . Biotechnol . 26 , 1367 \u2013 1372 . https : / / doi . org / 10 . 1038 / nbt . 1511 . 38 . Choi , M . , Chang , C . Y . , Clough , T . , Broudy , D . , Killeen , T . , MacLean , B . , and Vitek , O . ( 2014 ) . MSstats : an R package for statistical analysis of quantitative mass spectrometry - based proteomic experiments . Bioinfor - matics 30 , 2524 \u2013 2526 . https : / / doi . org / 10 . 1093 / bioinformatics / btu305 . 39 . Yu , G . , Wang , L . G . , Han , Y . , and He , Q . Y . ( 2012 ) . clusterPro\ufb01ler : an R package for comparing biological themes among gene clusters . Omics 16 , 284 \u2013 287 . https : / / doi . org / 10 . 1089 / omi . 2011 . 0118 . 40 . Fernandez , J . J . , Li , S . , Bharat , T . A . M . , and Agard , D . A . ( 2018 ) . Cryo - to - mography tilt - series alignment with consideration of the beam - induced sample motion . J . Struct . Biol . 202 , 200 \u2013 209 . https : / / doi . org / 10 . 1016 / j . jsb . 2018 . 02 . 001 . 41 . Zheng , S . , Wolff , G . , Greenan , G . , Chen , Z . , Faas , F . G . A . , Ba\u00b4rcena , M . , Koster , A . J . , Cheng , Y . , and Agard , D . A . ( 2022 ) . AreTomo : an integrated software package for automated marker - free , motion - corrected cryo - electron tomographic alignment and reconstruction . J . Struct . Biol . X 6 , 100068 . https : / / doi . org / 10 . 1016 / j . yjsbx . 2022 . 100068 . 42 . Fontana , P . , Dong , Y . , Pi , X . , Tong , A . B . , Hecksel , C . W . , Wang , L . , Fu , T . M . , Bustamante , C . , and Wu , H . ( 2022 ) . Structure of cytoplasmic ring of nuclear pore complex by integrative cryo - EM and AlphaFold . Science 376 , eabm9326 . https : / / doi . org / 10 . 1126 / science . abm9326 . 43 . Gade\u02c6lha , H . , Herna\u00b4ndez - Herrera , P . , Montoya , F . , Darszon , A . , and Cor - kidi , G . ( 2020 ) . Human sperm uses asymmetric and anisotropic \ufb02agellar controls to regulate swimming symmetry and cell steering . Sci . Adv . 6 , eaba5168 . https : / / doi . org / 10 . 1126 / sciadv . aba5168 . 44 . Roy , A . , Lin , Y . N . , Agno , J . E . , DeMayo , F . J . , and Matzuk , M . M . ( 2007 ) . Absence of tektin 4 causes asthenozoospermia and subfertility in male mice . FASEB J . 21 , 1013 \u2013 1025 . https : / / doi . org / 10 . 1096 / fj . 06 - 7035com . 45 . Roy , A . , Lin , Y . N . , Agno , J . E . , DeMayo , F . J . , andMatzuk , M . M . ( 2009 ) . Tek - tin 3 is required for progressive sperm motility in mice . Mol . Reprod . Dev . 76 , 453 \u2013 459 . https : / / doi . org / 10 . 1002 / mrd . 20957 . 46 . Zhou , L . , Liu , H . , Liu , S . , Yang , X . , Dong , Y . , Pan , Y . , Xiao , Z . , Zheng , B . , Sun , Y . , Huang , P . , et al . ( 2023 ) . Structures of sperm \ufb02agellar doublet mi - crotubulesexpandthegeneticspectrum ofmaleinfertility . Cell 186 , 2897 \u2013 2910 . e19 . https : / / doi . org / 10 . 1016 / j . cell . 2023 . 05 . 009 . 47 . Leung , M . R . , Zeng , J . , Wang , X . , Roelofs , M . C . , Huang , W . , Zenezini Chiozzi , R . , Hevler , J . F . , Heck , A . J . R . , Dutcher , S . K . , Brown , A . , et al . ( 2023 ) . Structural specializations of the sperm tail . Cell 186 , 2880 \u2013 2896 . e17 . https : / / doi . org / 10 . 1016 / j . cell . 2023 . 05 . 026 . 48 . Mastronarde , D . N . ( 2005 ) . Automated electron microscope tomography using robust prediction of specimen movements . J . Struct . Biol . 152 , 36 \u2013 51 . https : / / doi . org / 10 . 1016 / j . jsb . 2005 . 07 . 007 . 49 . Kremer , J . R . , Mastronarde , D . N . , and McIntosh , J . R . ( 1996 ) . Computer visualization of three - dimensional image data using IMOD . J . Struct . Biol . 116 , 71 \u2013 76 . https : / / doi . org / 10 . 1006 / jsbi . 1996 . 0013 . 50 . Ferna\u00b4ndez , J . J . , Li , S . , and Crowther , R . A . ( 2006 ) . CTF determination and correction in electron cryotomography . Ultramicroscopy 106 , 587 \u2013 596 . https : / / doi . org / 10 . 1016 / j . ultramic . 2006 . 02 . 004 . 51 . Agulleiro , J . I . , and Fernandez , J . J . ( 2015 ) . Tomo3D 2 . 0\u2014exploitation of advanced vector extensions ( AVX ) for 3D reconstruction . J . Struct . Biol . 189 , 147 \u2013 152 . https : / / doi . org / 10 . 1016 / j . jsb . 2014 . 11 . 009 . 52 . Pettersen , E . F . , Goddard , T . D . , Huang , C . C . , Couch , G . S . , Greenblatt , D . M . , Meng , E . C . , and Ferrin , T . E . ( 2004 ) . UCSF Chimera \u2013 a visualization system for exploratory research and analysis . J . Comput . Chem . 25 , 1605 \u2013 1612 . https : / / doi . org / 10 . 1002 / jcc . 20084 . 53 . Goddard , T . D . , Huang , C . C . , Meng , E . C . , Pettersen , E . F . , Couch , G . S . , Morris , J . H . , and Ferrin , T . E . ( 2018 ) . UCSF ChimeraX : meeting modern challenges in visualization and analysis . Protein Sci . 27 , 14 \u2013 25 . https : / / doi . org / 10 . 1002 / pro . 3235 . 54 . Emsley , P . , Lohkamp , B . , Scott , W . G . , and Cowtan , K . ( 2010 ) . Features and development of coot . Acta Crystallogr . D Biol . Crystallogr . 66 , 486 \u2013 501 . https : / / doi . org / 10 . 1107 / S0907444910007493 . 55 . Jimenez - Morales , D . , Rosa Campos , A . , Von Dollen , J . , Krogan , N . , and Swaney , D . ( 2023 ) . MS : Analytical R tools for mass spectrometry ( R pack - age version 1 . 18 . 0 ) . 56 . Perez - Riverol , Y . , Csordas , A . , Bai , J . , Bernal - Llinares , M . , Hewapathir - ana , S . , Kundu , D . J . , Inuganti , A . , Griss , J . , Mayer , G . , Eisenacher , M . , et al . ( 2019 ) . The PRIDE database and related tools and resources in 2019 : improving support for quanti\ufb01cation data . Nucleic Acids Res . 47 , D442 \u2013 D450 . https : / / doi . org / 10 . 1093 / nar / gky1106 . 57 . van der Spoel , A . C . , Jeyakumar , M . , Butters , T . D . , Charlton , H . M . , Moore , H . D . , Dwek , R . A . , and Platt , F . M . ( 2002 ) . Reversible infertility in male mice after oral administration of alkylated imino sugars : a nonhormonal approach to male contraception . Proc . Natl . Acad . Sci . USA 99 , 17173 \u2013 17178 . https : / / doi . org / 10 . 1073 / pnas . 262586099 . 58 . Hagen , W . J . H . , Wan , W . , and Briggs , J . A . G . ( 2017 ) . Implementation of a cryo - electrontomographytilt - schemeoptimizedforhighresolutionsubto - mogram averaging . J . Struct . Biol . 197 , 191 \u2013 198 . https : / / doi . org / 10 . 1016 / j . jsb . 2016 . 06 . 007 . 59 . Zheng , S . Q . , Palovcak , E . , Armache , J . P . , Verba , K . A . , Cheng , Y . , and Agard , D . A . ( 2017 ) . MotionCor2 : anisotropic correction of beam - induced motion for improved cryo - electron microscopy . Nat . Methods 14 , 331 \u2013 332 . https : / / doi . org / 10 . 1038 / nmeth . 4193 . 60 . Bharat , T . A . , and Scheres , S . H . ( 2016 ) . Resolving macromolecular struc - tures from electron cryo - tomography data using subtomogram averaging in RELION . Nat . Protoc . 11 , 2054 \u2013 2065 . https : / / doi . org / 10 . 1038 / nprot . 2016 . 124 . 61 . Pettersen , E . F . , Goddard , T . D . , Huang , C . C . , Meng , E . C . , Couch , G . S . , Croll , T . I . , Morris , J . H . , and Ferrin , T . E . ( 2021 ) . UCSF ChimeraX : structure visualization for researchers , educators , and developers . Protein Sci . 30 , 70 \u2013 82 . https : / / doi . org / 10 . 1002 / pro . 3943 . 62 . UniProt Consortium ( 2021 ) . UniProt : the universal protein knowledgebase in 2021 . Nucleic Acids Res . 49 , D480 \u2013 D489 . https : / / doi . org / 10 . 1093 / nar / gkaa1100 . 63 . Sayers , E . W . , Bolton , E . E . , Brister , J . R . , Canese , K . , Chan , J . , Comeau , D . C . , Connor , R . , Funk , K . , Kelly , C . , Kim , S . , et al . ( 2022 ) . Database re - sourcesofthenationalcenterforbiotechnologyinformation . NucleicAcids Res . 50 , D20 \u2013 D26 . https : / / doi . org / 10 . 1093 / nar / gkab1112 . 64 . Wriggers , W . ( 2012 ) . Conventions and work\ufb02ows for using situs . Acta Crystallogr . D Biol . Crystallogr . 68 , 344 \u2013 351 . https : / / doi . org / 10 . 1107 / S0907444911049791 . 65 . Waterhouse , A . M . , Procter , J . B . , Martin , D . M . , Clamp , M . , and Barton , G . J . ( 2009 ) . Jalview Version 2\u2014a multiple sequence alignment editor and analysis workbench . Bioinformatics 25 , 1189 \u2013 1191 . https : / / doi . org / 10 . 1093 / bioinformatics / btp033 . 66 . Zimmermann , L . , Stephens , A . , Nam , S . Z . , Rau , D . , Ku\u00a8bler , J . , Lozajic , M . , Gabler , F . , So\u00a8ding , J . , Lupas , A . N . , and Alva , V . ( 2018 ) . A completely ll OPEN ACCESS 5052 Cell 186 , 5041 \u2013 5053 , November 9 , 2023 Article reimplemented MPI bioinformatics toolkit with a new HHpred server at its core . J . Mol . Biol . 430 , 2237 \u2013 2243 . https : / / doi . org / 10 . 1016 / j . jmb . 2017 . 12 . 007 . 67 . Cruz - Gonza\u00b4lez , I . , Gonza\u00b4lez - Ferreiro , R . , Freixa , X . , Gafoor , S . , Shakir , S . , Omran , H . , Berti , S . , Santoro , G . , Kefer , J . , Landmesser , U . , et al . ( 2020 ) . Left atrial appendage occlusion for stroke despite oral anticoagulation ( resistant stroke ) . Results from the Amplatzer Cardiac Plug registry . Rev . Esp . Cardiol . ( Engl Ed ) . 73 , 28 \u2013 34 . https : / / doi . org / 10 . 1016 / j . rec . 2019 . 02 . 013 . 68 . Gomez - Roman , N . , Felton - Edkins , Z . A . , Kenneth , N . S . , Goodfellow , S . J . , Athineos , D . , Zhang , J . , Ramsbottom , B . A . , Innes , F . , Kantidakis , T . , Kerr , E . R . , et al . ( 2006 ) . Activation by c - Myc of transcription by RNA polymer - ases I , II and III . Biochem . Soc . Symp , 141 \u2013 154 . ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 , November 9 , 2023 5053 Article STAR + METHODS KEY RESOURCES TABLE REAGENT or RESOURCE SOURCE IDENTIFIER Biological samples Mouse ( Mus musculus ) sperm Gene Targeting and Transgenics Center , Janelia Research Campus N / A Human ( Homo sapiens ) sperm Lishko Laboratory , UC Berkeley N / A Chemicals , peptides , and recombinant proteins NaCl Sigma - Aldrich Cat # 71376 KH 2 PO 4 Sigma - Aldrich Cat # P0662 MgSO 4 , 7H 2 O Sigma - Aldrich Cat # M3409 Dextrose Sigma - Aldrich Cat # D9434 CaCl 2 Sigma - Aldrich Cat # 21097 KCl Sigma - Aldrich Cat # 60128 NaHCO 3 Sigma - Aldrich Cat # S5761 DTT Sigma - Aldrich Cat # DTT - RO TCEP Sigma - Aldrich Cat # C4706 Erythro - 9 - ( 2 - hydroxy - 3 - nonyl ) adenine Santa Cruz Biotechnology Cat # sc - 201184 KCSN Sigma - Aldrich Cat # 60178 Urea Sigma - Aldrich Cat # U5128 SDS Sigma - Aldrich Cat # 62862 Triton Sigma - Aldrich Cat # X100PC Deposited data Cryo - EM map of the 16 nm - repeating A - tubule ( mouse ) This paper EMDB : EMD - 41315 Cryo - EM map of the 16 nm - repeating B - tubule ( mouse ) This paper EMDB : EMD - 41316 Cryo - EM map of the 48 nm - repeating doublets ( mouse ) , the composite map This paper EMDB : EMD - 41431 Cryo - EM map of the 48 nm - repeating doublets ( mouse ) , submap 1 of EMDB : EMD - 41431 This paper EMDB : EMD - 41450 Cryo - EM map of the 48 nm - repeating doublets ( mouse ) , submap 2 of EMDB : EMD - 41431 This paper EMDB : EMD - 41451 Cryo - EM map of the 48 nm - repeating doublets ( human ) This paper EMDB : EMD - 41317 Cryo - EM map of the 48 nm - repeating doublets ( Tekt5 - / - ) This paper EMDB : EMD - 41320 Model of the mouse sperm doublets This paper PDB : 8TO0 PRIDE partner repository for MS data This paper PXD036885 R package source materials for MSstats from Krogan Lab This paper https : / / github . com / kroganlab Experimental models : Organisms / strains Mouse : C57BL / 6J The Jackson Laboratory https : / / www . jax . org / strain / 000664 Tekt5 - / - mouse This paper N / A Software and algorithms Prism v8 GraphPad https : / / www . graphpad . com / SerialEM 3 . 8 Mastronarde 48 https : / / bio3d . colorado . edu / SerialEM / Etomo Kremer et al . 49 https : / / bio3d . colorado . edu / imod / doc / UsingEtomo . html TOMOCTF Fernandez et al . 50 https : / / sites . google . com / site / 3demimageprocessing / tomoctf TOMO3D 2 . 0 Agulleiro and Fernandez 51 https : / / sites . google . com / site / 3demimageprocessing / tomo3d ( Continued on next page ) ll OPEN ACCESS e1 Cell 186 , 5041 \u2013 5053 . e1 \u2013 e6 , November 9 , 2023 Article RESOURCE AVAILABILITY Lead contact Further information and requests for resources and reagents should be directed to the lead contact , David A . Agard ( david @ agard . ucsf . edu ) . Materials availability Experimental reagents generated in this study are available from the lead contact with a completed material transfer agreement . Data and code availability Cryo - EM maps of 48 nm - repeating structures of doublets from wildtype mouse , Tekt5 - / - mouse and human sperm have been deposited in the Electron Microscopy Data Bank ( EMDB ) with accession codes : EMD - 41431 , EMD - 41320 and EMD - 41317 , respec - tively . The EMD - 41431 is a composite map with its two submaps deposited with accession codes : EMD - 41450 and EMD - 41451 . Maps of focused re\ufb01nement of 16 nm - repeating structures of A - and B - tubules from wildtype mouse have been deposited also : EMD - 41315 and EMD - 41316 . The atomic model of the 48 - nm repeat of the mouse sperm doublets has been deposited in the Protein Data Bank ( PDB ) with accession codes 8TO0 . MS data are shared and available through the ProteomeXchange Consortium via the PRIDE partner repository under the dataset identi\ufb01er : PXD036885 ( username : reviewer _ pxd036885 @ ebi . ac . uk ; password : tMEZ90MC ) . 56 R package source materials for MSstats ( version 3 ) are publicly available through the Krogan Lab GitHub : https : / / github . com / kroganlab . After downloading the AlphaFold2 library of the mouse proteome , this code is used to distribute PDB \ufb01les into subdirectories . i = 0 ; for f in * ; do # # Splitting 50 PDBs in each subdirectory d = dir _ $ ( printf % 03d $ ( ( i / 50 + 1 ) ) ) ; mkdir - p $ d ; mv \" $ f \" $ d ; let i + + ; done This code is used to unbiasedly match all PDBs with the target densities in each subdirectory : for \ufb01le in * do echo $ \ufb01le # # CCDC105 _ \ufb02ipped _ b150 . mrc is the target densities , the options could be found in the situs website colores . . / CCDC105 _ \ufb02ipped _ b150 . mrc $ { \ufb01le } - res 6 . 0 - cutoff 0 . 0048 - deg 15 . 0 mkdir . . / output / $ { \ufb01le } _ out mv col _ * . . / output / $ { \ufb01le } _ out / . mv Continued REAGENT or RESOURCE SOURCE IDENTIFIER Chimera Pettersen et al . 52 https : / / www . cgl . ucsf . edu / chimera / ChimeraX Goddard et al . 53 https : / / www . rbvi . ucsf . edu / chimerax / RELION - 4 . 0 Zivanov et al . 24 https : / / relion . readthedocs . io / en / release - 4 . 0 / AlphaFold2 Jumper et al . 27 https : / / alphafold . ebi . ac . uk / Situs Wriggers et al . 28 https : / / situs . biomachina . org / fguide . html Coot 0 . 9 . 8 . 1 Emsley et al . 54 http : / / www2 . mrc - lmb . cam . ac . uk / personal / pemsley / coot MaxQuant 1 . 6 . 3 . 3 Cox and Mann 37 https : / / www . maxquant . org / R Bioconductor package artMS 1 . 14 . 0 Jimenez - Morales et al . 55 https : / / doi . org / 10 . 18129 / B9 . bioc . artMS Other Quantifoil holey carbon grids ( R2 / 2 , 200 - mesh gold ) Quantifoil MicroTools GmbH https : / / www . quantifoil . com / products EM GP2 Automatic Plunge Freezer Leica Microsystems https : / / www . leica - microsystems . com / products ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 . e1 \u2013 e6 , November 9 , 2023 e2 Article done The cross - correlation scores could then be extracted using the following script : for f in * _ outt do echo $ f grep structure $ f / * . pdb > > TheResultFile grep Unnormalized $ f / * . pdb > > TheResultFile done grep \" correlation \" TheResultFile > JustCCResults The \ufb01nal output could then be sorted based on the cross - correlation scores in Excel . Note each PDB would be matched to the target densities with multiple orientations , resulting in multiple entries with the same PDB but different cross - correlation scores . The duplicate items for each PDB could be deleted in Excel . Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request . EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS Mouse models Wild - type and transgenic male C57BL / 6J mice at ages of 10 to 16 weeks were used for this study . All mice were cared for in compli - ance with the guidelines outlined in the Guide for the Care and Use of Laboratory Animals . All experiments were approved by the Janelia Research Campus ( JRC ) IACUC . JRC is an AAALAC - accredited institution . Mice were maintained under SPF conditions . Human sample A man aged 25 - 39 years old was recruited and consented to participate in this study . We did not bias ancestry , race or ethnicity throughout the recruitment process . We only checked the samples under the microscope to make sure the sperm were normozoo - spermicwithacellcountofatleast30millionspermcellspermilliliter . Allexperimentalproceduresusinghuman - derivedsampleswere approved by the Committee on Human Research at the University of California , Berkeley , under IRB protocol number 2013 - 06 - 5395 . METHOD DETAILS Sample preparation Mouse sperm were collected from 10 to 16 - week - old C57Bl / 6J mice based on the published protocol . 19 , 57 Brie\ufb02y , the sperm were extracted from vasa deferentia by applying pressure to cauda epididymides in 1x Krebs buffer ( 1 . 2 mM KH 2 PO 4 , 120 mM NaCl , 1 . 2 mM MgSO 4 , 7H 2 O , 14 mM dextrose , 1 . 2 mM CaCl 2 , 5 mM KCl , 25 mM NaHCO 3 ) . The sperm were washed and resuspended in (cid:4) 100 m L Krebs buffer for the following experiments . For human sperm samples , freshly ejaculated semen samples were obtained by masturbation . Grid preparation EM grids ( Quantifoil R 2 / 2 Au 200 mesh ) were glow discharged to be hydrophilic using an easiGlow system ( Pelco ) . The grid was then loaded onto a Leica GP2 plunge freezer ( pre - equilibrated to 95 % relative humidity at 25 (cid:3) C ) . The mouse sperm suspension was then mixed with 10 nm gold beads ( Electron Microscopy Science , cat # 25487 ) to achieve \ufb01nal concentrations at 2 - 6 million cells / mL . EHNA ( erythro - 9 - ( 2 - hydroxy - 3 - nonyl ) adenine ) ( Santa Cruz Biotechnology , CAS 51350 - 19 - 7 ) was added to a \ufb01nal concentration of 10 mM . Next , 3 . 5 m L of the sperm mixture was loaded onto each grid , followed by a 15 - second incubation period . The grids were then blotted for 4 sec and plunge - frozen in liquid ethane . Cryogenic focused ion beam ( cryoFIB ) milling CryoFIB was performed using an Aquilos II cryo - FIB / SEM microscope ( Thermo Fisher Scienti\ufb01c ) . A panorama SEM map of the whole grid was \ufb01rst taken at 377x magni\ufb01cation using an acceleration voltage of 5 kV with a beam current of 13 pA , and a dwell time of 1 m s . Targets with appropriate thickness for milling were selected on the grid . A platinum layer ( (cid:4) 10 nm ) was sputter coated and a gas injection system ( GIS ) was used to deposit the precursor compound trimethyl ( methylcyclopentadienyl ) platinum ( IV ) . The stage was tilted to 15 - 20 (cid:3) , corresponding to a milling angle of 8 - 13 (cid:3) relative to the plane of grids . FIB milling was performed using stepwise decreasing current as the lamellae became thinner ( 1 . 0 nA to 30 pA , \ufb01nal thickness : (cid:4) 300 nm ) . The grids were then stored in liquid nitrogen before data collection . Image acquisition and tomogram reconstruction Tilt series of mouse sperm were collected on a 300 - kV Titan Krios transmission electron microscope ( Thermo Fisher Scienti\ufb01c ) equip - ped with a high brightness \ufb01eld emission gun ( xFEG ) , a spherical aberration corrector , a Bioquantum energy \ufb01lter ( Gatan ) , and a K3 ll OPEN ACCESS e3 Cell 186 , 5041 \u2013 5053 . e1 \u2013 e6 , November 9 , 2023 Article Summit detector ( Gatan ) . The images were recorded at a nominal magni\ufb01cation of 26 , 000x in super - resolution counting mode using SerialEM . 48 After binning over 2 x 2 pixels , the calibrated pixel size was 2 . 612 A\u02da on the specimen level . For each tilt series , images were acquired using a modi\ufb01ed dose - symmetric scheme between - 48 (cid:3) and 48 (cid:3) relative to the lamella with 4 (cid:3) increments and grouping of two images on either side ( 0 (cid:3) , 4 (cid:3) , 8 (cid:3) , - 4 (cid:3) , - 8 (cid:3) , 12 (cid:3) , 16 (cid:3) , - 12 (cid:3) , - 16 (cid:3) , 20 (cid:3) . ) . 58 At each tilt angle , the image was recorded as movies divided into fourteen subframes . The total electron dose applied to a tilt series was 100 e - / A\u02da 2 . The defocus target was set to be - 2 to - 5 m m . All movie frames were corrected with a gain reference collected in the same EM session . Movement between frames was corrected using MotionCor2 without dose weighting . 59 All tilt series were aligned using AreTomo and the tomograms were inspected to identify good tilt series . 41 Tilt series with crystalline ice and big ice blocks , or possessing less than \ufb01ve doublets of the axonemes were dis - carded . Alignment of the good tilt series was then performed in Etomo using the gold beads as \ufb01ducial markers . 49 The AreTomo is less labor - intensive for screening purposes , while the Etomo work\ufb02ow allowed us to achieve high - resolution reconstructions . The aligned tilt series were then CTF - corrected using TOMOCTF 50 and the tomograms were generated using TOMO3D 51 ( bin4 , pixel size : 10 . 448 A\u02da ) . In total , we started with eight milling grids of mouse sperm and obtained 159 lamellae . Ultimately , the \ufb01nal reconstruc - tions of the consensus averages were based on 77 usable tomograms . Subvolume averaging Subvolume extraction , classi\ufb01cation and re\ufb01nement were \ufb01rst performed using RELION3 as reported previously . 60 Brie\ufb02y , subvo - lumes from the doublets were manually picked every 24 nm and extracted at binning of 6 ( pixel size : 15 . 672 A\u02da , box size : 80 pixels , dimension : 125 . 376 nm ) . These subvolumes were aligned to a map of non - treated mouse sperm doublet structure ( EMDB : EMD - 27444 ) lowpass \ufb01ltered to 80 A\u02da and the resulting map was used as the reference for further processing . Supervised 3D classi\ufb01cation on radial spokes gave rise to four class averages of the 96 - nm repeating units at four different registers . All four class averages were recentered at the base of Radial spoke 2 and re - extracted at the same point at the binning of 4 ( pixel size : 10 . 448 A\u02da , box size : 120 pixels , dimension : 125 . 376 nm ) . All subvolumes were combined and aligned to one reference and duplicate subvolumes were removed based on minimum distance ( < 40 nm ) . The remaining subvolumes were aligned to yield the consensus average for all nine doublets . Subvolumes of the 96 - nm repeating units were recentered on MIP features that repeat every 48 nm or 16 nm to obtain the coordinates of these subvolumes . The tomograms and coordinates were imported in RELION4 without binning . 24 Pseudo - sub - tomograms were extracted ( pixel size : 2 . 612 A\u02da , box size : 220 pixels , dimension : 57 . 464 nm ) and the \ufb01rst round of Re\ufb01ne3D jobs yield the initial reference for the following re\ufb01nement of geometric and optical parameters of the tilt series . TomoFrameAlign and CtfRe\ufb01neTomo jobs were executed alternatively for two rounds and new pseudo - subtomograms were extracted ( see FSC curves in Figure S2A ) . The same \u2018\u2018Re\ufb01ne3D - TomoFrameAlign - CtfRe\ufb01neTomo - TomoFrameAlign - CtfRe\ufb01neTomo\u2019\u2019 process was repeated again and new pseudo - subtomograms were extracted . The \ufb01nal Re\ufb01ne3D job yields the reported maps after three iterations . Further re\ufb01nement did not improve the resolutions and quality of maps . In order to generate a map covering the entire 48 nm - repeating struc - ture of the doublets , the run _ data . star \ufb01le from the \ufb01nal Re\ufb01ne3D job was shifted along the longitudinal axis and another round of re\ufb01nement yielded averages with the shifted register . These reconstructions were aligned to the reported 48 nm repeating structure of doublets from bovine trachea cilia and a composite map was generated to match the register of the periodic structure . The resolutions of the maps were estimated based on the FSC of two independently re\ufb01ned half datasets ( FSC = 0 . 143 ) . Local resolution maps for doublets of both human and mouse sperm were calculated by RELION4 and displayed in UCSF Chimera . 52 These local - resolution maps represent relative differences in resolution across the maps but the absolute values may not be precise . IMOD was used to visualize the tomographic slices . 49 UCSF Chimera was used to manually segment the maps for various structural fea - tures and these maps were colored individually to prepare the \ufb01gures using UCSF ChimeraX . 52 , 53 , 61 Model building and unbiased matching of density maps to protein candidates Model building was performed in Coot v0 . 9 . 8 . 1 54 and rigid body \ufb01tting was achieved using UCSF Chimera . The interpretation of the mouse sperm doublet map started with the atomic model of the bovine trachea doublet ( PDB : 7RRO ) . 13 Densities matching tubulins and 29 bovine MIPs in the bovine trachea doublet were found in the mouse sperm doublet map so all of these densities were consid - ered to be formed by M . musculus orthologs ( Figure S4 ) . We cannot exclude the possibilities that they are formed by sperm - speci\ufb01c homologs with similar tertiary structures . These orthologs were identi\ufb01ed using UniProt 62 or the NCBI protein database 63 based on the sequences of bovine proteins . The atomic models of bovine trachea MIPs were mutated to match the sequence of the mouse proteins using the Chainsaw plugin in Coot . The resulting models of individual proteins were then \ufb01t into the mouse sperm doublet map as rigid bodies in Chimera . MIP densities that are unique in sperm doublets were segmented from the corresponding maps using UCSF Chimera manually . At 6 - 8 A \u02da resolutions , a - helix is well - resolved and b - strands appear as curved sheets and we focused on the unassigned densities with well - de\ufb01ned tertiary structures or domains ( as the various colored densities shown in Figure 1 ) . Meanwhile , the PDB library of 21 , 615 mouse proteins based on AlphaFold2 prediction was downloaded . 27 The unbiased matching was carried out using the COLORES program ( Situs package , see the Data and code availability section in the STAR Methods ) . 28 , 64 The matching was scored and ranked by the cross - correlation scores and the top 200 hits were inspected individually with the target densities in UCSF Chimera . We noticed that matching of densities to much larger PDBs could lead to unrealistic cross correlations ( > 1 ) . Setting the box sizes of the maps to be two or three times larger compared to the isolated densities does not solve the issue since COLORES would cut ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 . e1 \u2013 e6 , November 9 , 2023 e4 Article off the zero valued edges by default to reduce computational loads . We thus used \u2018\u2018voledit\u2019\u2019 command from SITUS to edit the voxels at the corner of the cubic map . Speci\ufb01cally , we edited the \ufb01rst and the last values in the sit \ufb01le to be slightly larger than the threshold values to avoid the cropping . To test the unbiased proteome - wide search approach , we used densities corresponding to known MIPs identi\ufb01ed in bovine tracheal doublets , including PACRG , CFAP20 and NME7 as controls for the method . Indeed , PDB models cor - responding to the respective mouse orthologs were identi\ufb01ed as the top hits ( Data S1 ) , suggesting this visual proteomics approach can identify protein with matching tertiary structures . We then applied this method to identify candidates for the mouse sperm - spe - ci\ufb01c densities ( Data S2 \u2013 S4 ) . For the 4 - helix bundle densities , the CATH library , which curated non - redundant PDBs of published structural domains , 36 was also used and no homologous proteins of SPACA9 were found . The AlphaFold2 predicted PDBs were then used as starting models and initial \ufb01tting pose discovered by COLORES were inspected in Chimera . The various Tektin 5 and CCDC105 models were built in Coot to match the corresponding densities . Unresolved loops were deleted . We observed densities corresponding to 24 copies of SPACA9 for the 48 nm - repeating units of the doublet microtu - bules , albeit with varied occupancies . We rigid - body \ufb01tted all 24 SPACA9 in these densities in Coot to re\ufb02ect the stacking oligomer - ization in the deposited model . We also performed rigid - body \ufb01tting of DUSP3 ( although other DUSP proteins could be \ufb01tted too ) into the globular densities next to the slanted Tektin 5s . These models were combined in ChimeraX and all side chains were stripped using the phenix . pdbtools command . Sequence alignment and search for homologous proteins Sequence alignment was performed using Clustal Omega 33 server and displayed in Jalview . 65 M . musculus Tektin 1 sequence was used as input to search for Tektin homologs using the HHpred server . 66 Biochemical extractions of mouse sperm For each of the three biological replicates , sperm from two mice were washed with PBS and pelleted at 2000x g for 5 min . Then , the E1 - E5 buffers were used to extract proteins from the pellets [ 0 . 1 % Triton in PBS ( E1 ) , 0 . 6 M NaCl in PBS ( E2 ) , 0 . 6 M KCSN in PBS ( E3 ) , 8 M urea ( E4 ) and 10 % SDS ( E5 ) ] . For E1 to E4 , 100 m L of the buffer was added to the pellets and the resuspension was mixed by pipetting up and down using a p200 pipette . Then the solution was incubated at room temperature for 10 min and the pellet was spun down at 21 , 000x g for 10 min . For E5 , after 10 % SDS was added and mixed , the resuspension was heated at 95 (cid:3) for 5 min . After the pellets were spun down , the supernatant was taken as the extraction . 20 m L and 2 m L of the extractions were used for SDS - PAGE analyses , either stained with AcquaStain ( Fisher Scienti\ufb01c , NCO170988 ) and blotted with an antibody against a - tubulins ( ThermoFisher Scienti\ufb01c , DM1A , # 62204 ) . The remaining extractions were used for mass spectrometry analysis . Mass spectrometry ( MS ) - based global protein abundance of mouse sperm Proteins in biochemical fractions E1 , E2 , E3 , E4 and E5 from three biological replicates were reduced and alkylated in 4 mM \ufb01nal concentration tris ( 2 - carboxyethyl ) phosphine ( TCEP ) and 10 mM \ufb01nal concentration iodoacetamide by 20 - minute incubation in the dark , after which excess iodoacetamide was quenched with 10 mM \ufb01nal concentration dithiothreitol ( DTT ) . Proteins were then subjected to methanol chloroform precipitation . Brie\ufb02y , 1 part sample was combined and vortexed sequentially with 4 parts meth - anol , 1 part chloroform , and 3 parts water for phase separation , after which samples were spun for 2 minutes at top speed ( 14 , 000 g ) in a bench - top centrifuge ( Centrifuge 5424R , Eppendorf ) . The upper phase was removed and discarded , and 4 parts methanol were combined and vortexed with the interphase and lower phase and subsequently centrifuged for 3 minutes at 14 , 000 g . The superna - tant was removed and discarded , and the pellet was washed three times in 80 % ice - cold acetone followed by centrifugation for 3 minutes at 14 , 000 g . Extracted proteins were air dried , resuspended in 8 M urea buffer ( 8 M urea , 150 mM NaCl , 50 mM NH 4 HCO 3 , cOmplete Mini EDTA - free protease inhibitor ( Roche , 11836170001 ) ) , and quanti\ufb01ed using Bradford reagent ( Sigma , B6916 ) following Coomassie ( Bradford ) Protein Assay Kit\u2019s protocol ( Thermo Fisher , 23200 ) . Following quanti\ufb01cation , protein sam - ples were diluted 4 - fold to 2 M urea concentration with 0 . 1 M NH 4 HCO 3 pH 8 , digested with trypsin ( Promega , V5111 ) at a protea - se : protein ratio of 1 : 100 ( weight / weight ) , and incubated overnight at 37 (cid:3) C in a thermomixer at 750 rpm . After tryptic digest , samples were acidi\ufb01ed to pH < 3 with 1 % \ufb01nal concentration formic acid , and desalted for MS analysis using HPLC - grade reagents and 100 m L OMIX C18 tips ( Agilent Technologies , A57003100 ) according to the manufacturer\u2019s protocol with the following adjustments . Brie\ufb02y , OMIX tips were conditioned by sequential washes of 100 % acetonitrile and 50 % acetonitrile , 0 . 1 % formic acid , and equilibrated with two washes of 0 . 1 % formic acid . Peptides were bound to the C18 polymer by repeated pipetting , subsequently washed three times with 0 . 1 % formic acid , and sequentially eluted in 50 % acetonitrile , 0 . 1 % formic acid followed by 90 % acetonitrile , 0 . 1 % formic acid . Peptides were dried by vacuum centrifugation ( CentriVap Cold Trap , Labconco ) and stored at - 80 (cid:3) C until MS analysis . Digested , desalted peptides were resuspended to 0 . 125 - 2 m g / m L \ufb01nal concentration in 2 % acetonitrile , 0 . 1 % formic acid . 1 - 2 m L were injected in technical singlet onto an Easy - nLC 1200 ( Thermo Fisher Scienti\ufb01c ) interfaced via a nanoelectrospray source ( Nano - spray Flex ) coupled to an Orbitrap Fusion Lumos Tribrid mass spectrometer ( Thermo Fisher Scienti\ufb01c ) . Peptides were separated on a PepSep reverse - phase C18 column ( 1 . 9 m m particles , 1 . 5 m m x 15 cm , 150 m m ID ) ( Bruker ) with a gradient of 5 - 88 % buffer B ( 0 . 1 % formic acid in acetonitrile ) over buffer A ( 0 . 1 % formic acid in water ) over a 100 - minute data acquisition . Spectra were acquired contin - uously in a data - dependent manner . One full scan in the Orbitrap ( scan range 350 - 1350 m / z at 120 , 000 resolution in pro\ufb01le mode with a custom AGC target and maximum injection time of 50 milliseconds ) was followed by as many MS / MS scans as could be acquired ll OPEN ACCESS e5 Cell 186 , 5041 \u2013 5053 . e1 \u2013 e6 , November 9 , 2023 Article on the most abundant ions in 2 seconds in the dual linear ion trap ( rapid scan type with \ufb01xed HCD collision energy of 32 % , custom AGC target , maximum injection time of 50 milliseconds , and isolation window of 0 . 7 m / z ) . Singly and unassigned charge states were rejected . Dynamic exclusion was enabled with a repeat count of 1 , an exclusion duration of 25 seconds , and an exclusion mass width of \u00b1 10 ppm . Liquid chromatography 67 and MS acquisition parameters are reported in ( Table S1 ) . Raw MS \ufb01les were searched using MaxQuant ( version 1 . 6 . 3 . 3 ) against a database of the mouse proteome ( SwissProt Mus mus - culus reviewed protein sequences , downloaded 07 May 2022 ) with a manual addition to include mouse piercer of microtubule wall 2 protein ( protein sequence from NCBI Reference Sequence NP _ 001185718 . 1 , manually assigned the UniProt identi\ufb01er \u2018\u2018ZCC15orf65\u2019\u2019 in our database after its bovine homolog ) . 37 MaxQuant settings were left at default , with the following exceptions : LFQ was enabled with skip normalization enabled ; and match between runs was enabled with a 1 . 5 - minute matching time window and 20 - minute align - ment window . Trypsin ( KR | P ) was selected and allowed up to two missed cleavages , and variable and \ufb01xed modi\ufb01cations were as - signed for protein acetylation ( N - terminal ) , methionine oxidation and carbamidomethylation . Statistical analysis of protein quantitation was completed with R Bioconductor package artMS ( version 1 . 14 . 0 ) 55 and its function artmsQuanti\ufb01cation , which is a wrapper around the R Bioconductor package Mass Spectrometry Statistics and Quanti\ufb01cation ( MSstats ) ( version 4 . 4 . 0 ) as follows 38 ( Table S2 ) . Peptide intensities from the MaxQuant evidence \ufb01le were summarized to protein intensities using the MSstats function dataProcess with default settings . The differences in log 2 - transformed intensity between biochemical fractions were scored using the MSstats function groupComparison , which \ufb01ts a single linear model for each protein with a single categorical variable for condition , or fraction in our case . From these models , MSstats reports pairwise differences in means between conditions as log 2 fold change ( log2FC ) with a p - value based on a t - test assuming equal variance across all con - ditions , and reports adjusted p - values using the false discovery rate ( FDR ) estimated by the Benjamini - Hochberg procedure . Proteins with signi\ufb01cant changes in abundance between fractions were de\ufb01ned as : ( 1 ) absolute ( log 2 FC ) > 1 ; and ( 2 ) adjusted p - value < 0 . 05 . Proteins with signi\ufb01cant changes in abundance were tested for enrichment of Gene Ontology terms ( Table S3 ) . The over - represen - tation analysis was performed using the enricher function from R package clusterPro\ufb01ler ( version 4 . 4 . 1 ) . 39 Gene Ontology ( GO Bio - logical Process , Molecular Function and Cellular Component ) terms and annotations were obtained from the R annotation package org . Mm . eg . db ( version 3 . 15 . 0 ) . From among all signi\ufb01cantly enriched terms , we selected a set of non - redundant terms following a clustering procedure . We \ufb01rst constructed a term tree based on distances ( 1 - Jaccard Similarity Coef\ufb01cients of shared genes in KEGG or GO ) between the signi\ufb01cant terms . The term tree was cut at a speci\ufb01c level ( h = 0 . 99 ) to identify clusters of non - redundant gene sets ( Table S4 ) . For results with multiple signi\ufb01cant terms belonging to the same cluster , we selected the most signi\ufb01cant ( lowest adjusted p - value ) term . Generation of Tektin - 5 knockout mice and functional / structural analyses of mutant sperm To create an easily detected frameshift mutation in the Tekt5 gene , we used two gRNAs located in exon 1 of the gene . There is a pseudogene located on chromosome 19 with 85 . 8 % homology to the Tektin5 coding region but has a truncated C termini so that the 2 - helix segment is incomplete . Three gRNAs were carefully selected to avoid cutting the pseudogene . They are gRNA1 ( CGCTGGGTCTCCACGCGTTCAGG ) , gRNA2 ( AGTTTCTGTGGCCCCAAGAAAGG ) , and gRNA3 ( CCGAGGAATGCTCAGGC ATCCGG ) . The gRNAs were in vitro transcribed using the MEGA shortscript T7 kit ( Life Tech Corp AM1354 ) . Two combinations of the gRNAs were used : gRNA1 + gRNA2 , and gRNA2 + gRNA3 . The gRNAs and Cas9 protein ( Invitrogen True - cut Cas9 protein V2 , cat # A36498 ) with a concentration of 125 ng each were co - electroporated into 1 - cell C57Bl / 6J embryos using a BEX Genome Editor . A total of 75 pups were weaned and genotyped by PCR . Thirty - nine of them with an obviously shorter PCR band were sequenced , and nine of them with a frameshift deletion were selected for germline testing . Germline transmission was found in 7 out of 9 , and 2 lines were selected for further breeding : line 6 with a 166 bp deletion and line 7 with a 181 bp deletion . Both lines were then mated with wild - type females to generate F1 animals . Mating of heterozygous F1 male and female led to F2 homozygotes , which were further con\ufb01rmed by sequencing . Sperm motility was recorded at 37 (cid:3) C on a Hamilton Thorne IVOS II CASA machine using a Zeiss 10x NH objective , at a frame rate of 60 Hz , in the presence of 1 % polyvinyl alcohol to prevent cell adhesion to glass . The motility and sperm morphologies were inspected and counted manually . The structural analyses of the mutants were performed using the same work\ufb02ow described for wild - type sperm , except omitting the EHNA treatment . Sperm from two different Tektin5 - knockout lines were processed , imaged and analyzed independently . The two independent reconstructions show low occupancies of speci\ufb01c Tektin 5 densities and only the higher - resolution reconstruction is shown . QUANTIFICATION AND STATISTICAL ANALYSIS Statistical analyses of Tekt5 - / - mouse were prepared using Prism9 ( GraphPad ) . We performed six mating trials by using six Tekt5 - / - males ( 3 from each line of two lines ) and six wildtype females . The average and standard deviation are presented ( 7 . 3 \u00b1 1 . 4 , n = 6 ) . For the sperm analyses , we counted enough videos so the number of sperm from each mouse is > 200 . Sperm from three Tekt5 - / - knockout mutants and two wild - type mice were analyzed functionally . ll OPEN ACCESS Cell 186 , 5041 \u2013 5053 . e1 \u2013 e6 , November 9 , 2023 e6 Article Supplemental \ufb01gures ( legend on next page ) ll OPEN ACCESS Article Figure S1 . Work\ufb02ow of data processing , related to Figures 1 , 2 , and 3 and STAR Methods ( A ) Tilt series composed of 25 2D projections were recorded ( scale bar : 100 nm ) . The image shows the midpiece of sperm \ufb02agella that contains mitochondria around the axoneme . Gold beads on the tilt images are indicated ( green arrowhead ) . The \ufb01ducial gold beads were used to align the tilt images . ( B ) 3D tomograms were reconstructed , and subvolumes were picked along the microtubules ( scale bar : 100 nm ) . ( C ) 3Dclassi\ufb01cationandre\ufb01nementwereperformedtoalignandaveragethesubtomogramforthe96nm - repeatingstructuresofthemousespermdoublets . Four views of the 96 nm - repeating structure of doublets from EHNA - treated sperm are shown for the 3D reconstruction generated using RELION3 as reported previously . 19 Gold - standardFouriershellcorrelation ( FSC ) curvecalculatedbetweenhalfmapsofmousespermdoubletsisshown . Theresolutionwasestimated as 26 A\u02da ( FSC = 0 . 143 ) . ( D ) Two slices of the 96 nm - repeating structure of doublets looking along and perpendicular to the \ufb01lament axis . Note the red line in the top panel indicates the planeofthebottomslice , andperiodicstructuresareobservedinsidethemicrotubules . Thecoordinateswererecenteredonthe48 - nmrepeatsandimportedinto RELION4 . In the top panel , note that the features further away from the microtubules are blurrier , suggesting that there are conformational heterogeneities and that they are resolved at lower resolutions . ( E ) The initial Re\ufb01ne3D job of the 48 nm - repeating structures was performed using RELION4 . 24 ( F ) The 3D reconstructions were matched to the 2D projections of individual particles in the raw tilt images , and this step re\ufb01ned both the geometric and optical parameters of the tilt series ( i \u2013 iii ) . ( G ) Another round of subtomogram averaging was performed based on re\ufb01ned tilt series . No additional improvement was observed after 3 rounds of re\ufb01nement and Re\ufb01ne3D as shown in ( F ) \u2013 ( G ) . ll OPEN ACCESS Article ( legend on next page ) ll OPEN ACCESS Article Figure S2 . Characterization of the 48 nm - repeating structure of doublets from mouse sperm , related to Figures 1 , 2 , and 3 ( A ) Gold - standard Fourier shell correlation ( FSC ) curves were calculated between half maps of mouse sperm doublets . The resolutions were reported as FSC = 0 . 143 . Note the FSC curves resulting from the iterative frame alignment and CTF re\ufb01nement between the second and third Re\ufb01ne3D jobs were not shown for the clarity of the \ufb01gure . Further re\ufb01nement after the third Re\ufb01ne3D did not improve the resolution or the quality of the map . ( B ) The local resolution map of mouse sperm doublets was calculated by RELION4 . The ribbon region has the highest resolutions . Densities in the A - tubule have higher resolutions than the ones from the B - tubule . ( C ) Equivalentlongitudinalcross - sectionviewsofdoubletsfrommousespermandbovinetracheacilia ( EMDB : EMD - 24664 ) areshown . 13 Thelatterwaslow - pass \ufb01ltered to 7 . 5 A\u02da , and comparable details of the secondary and tertiary structures of the MIPs are observed . ( D ) The reconstruction of mouse sperm doublet ( gray ) is overlaid with the bovine trachea doublets ( yellow ) . The mouse sperm - speci\ufb01c densities are highlighted ( dashedovals ) . ThebrokenhelicalbundlesandthecurvedhelicalbundlesinsidetheA - tubuleofmousespermdoubletsalongthemicrotubuleaxisareshown . The discontinuous parts of the broken helical bundles are indicated ( dashed rectangles ) . Note that the curved bundles have one straight and two curved groups of densities in every 48 - nm repeat ( outlined using dashed shapes ) . ll OPEN ACCESS Article ( legend on next page ) ll OPEN ACCESS Article Figure S3 . Characterization of the 48 nm - repeating structure of doublets from human sperm , related to Figure 1 ( A ) The gold - standard Fourier shell correlation ( FSC ) curve was calculated between half maps of mouse sperm doublets . The resolution was estimated as 10 . 3 A\u02da ( FSC = 0 . 143 ) . ( B ) The local resolution map of human sperm doublets was calculated by RELION4 . The ribbon region has the highest resolutions . Densities in the A - tubule have higher resolutions than the ones from the B - tubule . ( C ) Equivalent views of doublets from human sperm and bovine trachea cilia ( EMDB : EMD - 24664 ) are shown . 13 The latter was low - pass \ufb01ltered to 10 A\u02da , and comparable details of the secondary and tertiary structures of the MIPs are observed . ( D ) The reconstruction of human sperm doublet ( blue ) is overlaid with the bovine trachea doublets ( yellow ) at low and high thresholds . ( E ) The two broken bundles inside the A - tubule in human sperm are shown at a low threshold ( see the corresponding mouse densities in Figure S2D ) . ( F ) ThecurvedhelicalbundlescontainonestraightandtwocurvedgroupsofdensitiesinsidetheA - tubuleofhumansperm . Humansperm - speci\ufb01cdensitieswere observed to connect one curved bundle to the lumen of A - tubule . ( G ) The human sperm doublets overlaid with mouse sperm doublets are shown . The inconsistent densities are outlined ( dashed line ) ( also see Figures S2D and S3F ) . ll OPEN ACCESS Article ( legend on next page ) ll OPEN ACCESS Article FigureS4 . Rigid - body\ufb01ttingof29identi\ufb01ed MIPsfrombovinetrachea ciliaintothedensitymapofmousespermdoublet , relatedtoFigures2 and 3 and STAR Methods ( A \u2013 F ) Modelsof29knownMIPsfrombovinetrachea cilia ( PDB : 7RRO ) 13 are\ufb01ttedintothedensitymapofmousespermdoublet . Theviewing anglesforallpanels are shown . For proteins that have multiple a - helices ( CFAP161 , RIBC2 , CFAP53 , MNS1 , CFAP21 , NME7 , CFAP141 , EFHC1 , EFHC2 , ENKUR , CFAP210 , EFCAB6 , CFAP45 , PACRG , andTEKTIN1 \u2013 4 ) , thearrangementofsecondarystructuresmatchesdensitiesinspermdoublets . Theoverallshapesof b - sheetsin b - sheet - richproteins ( CFAP52 and CFAP20 ) matchthedensities , andtheseproteinsarehighly conservedinaxonemes . Fortheproteins thatcontainrandom coils , we did observe matching features in the maps and were able to trace some of the main chains at the current resolution ( CFAP95 , SPAG8 , CFAP107 , FAM166B , Pierce1 , Pierce2 , CFAP126 , CFAP276 , and TEKTIP1 ) . ll OPEN ACCESS Article Figure S5 . Characterization of the 16 nm - repeating structures of doublets from mouse sperm , related to Figure 2 ( A and B ) Gold - standard Fourier shell correlation ( FSC ) curves were calculated using half maps of 16 nm - repeating structures of A - tubule and B - tubule . The resolution was estimated as 6 . 0 and 6 . 7 A\u02da , respectively ( FSC = 0 . 143 ) . The Nyquist limit is 5 . 26 A\u02da . ( C \u2013 E ) The local resolution map was calculated from the two half maps of 16 nm - repeating structures of A - tubule using RELION4 . The viewing angles for ( D ) and ( E ) are shown in ( C ) ( black arrow ) . These viewing angles are similar to Figures 1A , 1D , and 1E , respectively . ( F \u2013 H ) The local resolution map was calculated using the two half maps of 16 nm - repeating structures of B - tubule using RELION4 . The viewing angles for ( G ) and ( H ) are shown in ( F ) ( black arrow ) . The viewing angles of ( F ) and ( G ) are similar to Figures 1A and 1F , respectively . ll OPEN ACCESS Article ( legend on next page ) ll OPEN ACCESS Article Figure S6 . Tektin 5 and CCDC105 likely form sperm - speci\ufb01c 3 - helix bundles associated with the A - tubule , related to Figures 2 and 3 ( A ) After unbiased matching , Tektin 5 was scored as the # 5 hit of the predicted structures out of 21 , 615 proteins from the mouse proteome , ranked by cross - correlation scores ( the top 10 are shown ) . Tektin 1 \u2013 4 were ranked at # 7 \u2013 10 due to their similar tertiary structures . ( B ) Typicalfalsepositives ( # 1 \u2013 4and # 6 ) fromthesamesearch . Usually , theseareproteinswithlongsinglehelicesthatdonotmatchthegapsobservedinthemap . Also , they do not explain the 3 - helix bundles . The \ufb01tting of Tektin 5 into the same densities is shown for comparison . ( C ) ThestructureofCCDC105directlypredictedbyAlphaFold2 ( left ) iscomparedwiththepredictedcomplexformedbytwoCCDC105molecules ( right ) . Thefull - length CCDC105 molecule in the complex is colored based on the per - residue con\ufb01dence scores ( predicted local distance difference test [ pLDDT ] , red : high con\ufb01dence ) from the AlphaFold2 prediction . The three P - loops have medium con\ufb01dence scores ( green ) , suggesting the exact conformations of these loops may not be accurately predicted . However , the presence of these structured loops is conceivably con\ufb01dent based on the conserved proline residues ( see the sequence alignment in D ) and the matching protrusion densities observed in our maps ( Figure 2G ) . Note that the conformations of the three proline - rich loops differ in these two predictions . These differences could be caused by the presence of neighboring molecules during AlphaFold2 analyses . 27 ( D ) The sequence alignment of CCDC105 from \ufb01ve mammals ( H . sapiens , M . musculus , B . taurus , S . scrofa , and F . catus ) , zebra\ufb01sh ( D . rerio ) , and sea urchins ( S . purpuratus ) . The three proline - rich loops are marked above the sequences . ( E ) ThemodelsofCCDC105andTektin5are\ufb01ttedintothedensitiesofthe3 - helixbundleattheribbon , wheretheformermodelexplainstheextraprotrusionsand orientation / lengths of helices of the densities , but the latter does not . ll OPEN ACCESS Article Figure S7 . DUSP proteins in the A - tubule , related to Figure 3 ( A ) At a lower threshold compared with Figure 3C , densities connecting the N - terminal residues of the slanted Tektin 5s ( magenta models ) and the DUSPs ( blue models ) are observed . ( B ) The DUSP3 is \ufb01tted into the globular domain , and three orthogonal views are shown . Other homologous DUSP proteins \ufb01t well into the density because of similar tertiary structures ( DUSP 3 , 13 , 14 , 18 , 21 , and 29 ) . ll OPEN ACCESS Article Figure S8 . Biochemical extractions of proteins from mouse sperm , related to Figure 3 ( A ) SDS - PAGE analyses of protein extractions from mouse sperm using 0 . 1 % Triton in PBS ( E1 ) , 0 . 6 M NaCl in PBS ( E2 ) , 0 . 6 M KCSN in PBS ( E3 ) , 8 M urea ( E4 ) , and 10 % SDS ( E5 ) . ( B ) Westernblotanalysesofproteinextractionsfrommousespermusinganantibodyagainst a - tubulins . NotethatstrongbandsweredetectedonlyinE3andE4 , suggesting the microtubule structures were stable in Triton and high NaCl buffers and dissembled completely in KCSN / urea solutions . ( C ) Bar chart of the number of proteins identi\ufb01ed by MS ( protein count ) in each fraction ( E1 \u2013 E5 ) and biological replicate . We identi\ufb01ed a total of 1 , 677 mouse proteins , with a range of 772 to 1 , 326 proteins identi\ufb01ed in each individual fraction and replicate . ( D ) Heatmap of proteins with signi\ufb01cant changes between any two fractions ( absolute log 2 fold change [ log 2 FC ] > 1 , adjusted p value < 0 . 05 ) , listed by fractions ( E1 \u2013 E5 ) and biological replicate , and clustered by correlation of intensity pro\ufb01le . Proteins are colored by the log 2 FC in protein intensity normalized to the row median ( red , increased intensity ; blue , decreased intensity ; gray , not detected ) . Cluster identi\ufb01cation numbers ( cluster ID ) are labeled ( left ) . ( E ) Heatmapofgeneontology ( GO ) enrichmentsamongthesigni\ufb01cantlychangingproteinsidenti\ufb01edineachclusterfrom ( D ) ( lefttoright : clusterID1 \u2013 6 , aslabeled in D ) . GO terms were curated from the top 4 enrichment terms per cluster , and non - redundant terms were selected by an automated clustering procedure ( see STAR Methods ) . Increased shading re\ufb02ects increased signi\ufb01cance of the enrichment term . The number of proteins per enrichment term is shown in white if signi\ufb01cant ( adjusted p value < 0 . 05 ) and gray if not signi\ufb01cant ( adjusted p value > 0 . 05 ) . A bar chart plotting the number of total genes in each cluster ID is included . 68 ( F ) Log 2 protein intensities ( y axis ) for eight mouse proteins as quanti\ufb01ed by MS in each fraction ( E1 \u2013 E5 ) and biological replicate ( colored dots ; maximum n = 3 ) . ll OPEN ACCESS Article",
    "mikolovEfficientEstimationWord2013": "Ef\ufb01cient Estimation of Word Representations in Vector Space Tomas Mikolov Google Inc . , Mountain View , CA tmikolov @ google . com Kai Chen Google Inc . , Mountain View , CA kaichen @ google . com Greg Corrado Google Inc . , Mountain View , CA gcorrado @ google . com Jeffrey Dean Google Inc . , Mountain View , CA jeff @ google . com Abstract We propose two novel model architectures for computing continuous vector repre - sentations of words from very large data sets . The quality of these representations is measured in a word similarity task , and the results are compared to the previ - ously best performing techniques based on different types of neural networks . We observe large improvements in accuracy at much lower computational cost , i . e . it takes less than a day to learn high quality word vectors from a 1 . 6 billion words data set . Furthermore , we show that these vectors provide state - of - the - art perfor - mance on our test set for measuring syntactic and semantic word similarities . 1 Introduction Many current NLP systems and techniques treat words as atomic units - there is no notion of similar - ity between words , as these are represented as indices in a vocabulary . This choice has several good reasons - simplicity , robustness and the observation that simple models trained on huge amounts of data outperform complex systems trained on less data . An example is the popular N - gram model used for statistical language modeling - today , it is possible to train N - grams on virtually all available data ( trillions of words [ 3 ] ) . However , the simple techniques are at their limits in many tasks . For example , the amount of relevant in - domain data for automatic speech recognition is limited - the performance is usually dominated by the size of high quality transcribed speech data ( often just millions of words ) . In machine translation , the existing corpora for many languages contain only a few billions of words or less . Thus , there are situations where simple scaling up of the basic techniques will not result in any signi\ufb01cant progress , and we have to focus on more advanced techniques . With progress of machine learning techniques in recent years , it has become possible to train more complex models on much larger data set , and they typically outperform the simple models . Probably the most successful concept is to use distributed representations of words [ 10 ] . For example , neural network based language models signi\ufb01cantly outperform N - gram models [ 1 , 27 , 17 ] . 1 . 1 Goals of the Paper The main goal of this paper is to introduce techniques that can be used for learning high - quality word vectors from huge data sets with billions of words , and with millions of words in the vocabulary . As far as we know , none of the previously proposed architectures has been successfully trained on more 1 a r X i v : 1301 . 3781v3 [ c s . C L ] 7 S e p 2013 than a few hundred of millions of words , with a modest dimensionality of the word vectors between 50 - 100 . We use recently proposed techniques for measuring the quality of the resulting vector representa - tions , with the expectation that not only will similar words tend to be close to each other , but that words can have multiple degrees of similarity [ 20 ] . This has been observed earlier in the context of in\ufb02ectional languages - for example , nouns can have multiple word endings , and if we search for similar words in a subspace of the original vector space , it is possible to \ufb01nd words that have similar endings [ 13 , 14 ] . Somewhat surprisingly , it was found that similarity of word representations goes beyond simple syntactic regularities . Using a word offset technique where simple algebraic operations are per - formed on the word vectors , it was shown for example that vector ( \u201dKing\u201d ) - vector ( \u201dMan\u201d ) + vec - tor ( \u201dWoman\u201d ) results in a vector that is closest to the vector representation of the word Queen [ 20 ] . In this paper , we try to maximize accuracy of these vector operations by developing new model architectures that preserve the linear regularities among words . We design a new comprehensive test set for measuring both syntactic and semantic regularities 1 , and show that many such regularities can be learned with high accuracy . Moreover , we discuss how training time and accuracy depends on the dimensionality of the word vectors and on the amount of the training data . 1 . 2 Previous Work Representation of words as continuous vectors has a long history [ 10 , 26 , 8 ] . A very popular model architecture for estimating neural network language model ( NNLM ) was proposed in [ 1 ] , where a feedforward neural network with a linear projection layer and a non - linear hidden layer was used to learn jointly the word vector representation and a statistical language model . This work has been followed by many others . Another interesting architecture of NNLM was presented in [ 13 , 14 ] , where the word vectors are \ufb01rst learned using neural network with a single hidden layer . The word vectors are then used to train the NNLM . Thus , the word vectors are learned even without constructing the full NNLM . In this work , we directly extend this architecture , and focus just on the \ufb01rst step where the word vectors are learned using a simple model . It was later shown that the word vectors can be used to signi\ufb01cantly improve and simplify many NLP applications [ 4 , 5 , 29 ] . Estimation of the word vectors itself was performed using different model architectures and trained on various corpora [ 4 , 29 , 23 , 19 , 9 ] , and some of the resulting word vectors were made available for future research and comparison 2 . However , as far as we know , these architectures were signi\ufb01cantly more computationally expensive for training than the one proposed in [ 13 ] , with the exception of certain version of log - bilinear model where diagonal weight matrices are used [ 23 ] . 2 Model Architectures Many different types of models were proposed for estimating continuous representations of words , including the well - known Latent Semantic Analysis ( LSA ) and Latent Dirichlet Allocation ( LDA ) . In this paper , we focus on distributed representations of words learned by neural networks , as it was previously shown that they perform signi\ufb01cantly better than LSA for preserving linear regularities among words [ 20 , 31 ] ; LDA moreover becomes computationally very expensive on large data sets . Similar to [ 18 ] , to compare different model architectures we de\ufb01ne \ufb01rst the computational complex - ity of a model as the number of parameters that need to be accessed to fully train the model . Next , we will try to maximize the accuracy , while minimizing the computational complexity . 1 The test set is available at www . fit . vutbr . cz / \u02dcimikolov / rnnlm / word - test . v1 . txt 2 http : / / ronan . collobert . com / senna / http : / / metaoptimize . com / projects / wordreprs / http : / / www . fit . vutbr . cz / \u02dcimikolov / rnnlm / http : / / ai . stanford . edu / \u02dc ehhuang / 2 For all the following models , the training complexity is proportional to O = E \u00d7 T \u00d7 Q , ( 1 ) where E is number of the training epochs , T is the number of the words in the training set and Q is de\ufb01ned further for each model architecture . Common choice is E = 3 \u2212 50 and T up to one billion . All models are trained using stochastic gradient descent and backpropagation [ 26 ] . 2 . 1 Feedforward Neural Net Language Model ( NNLM ) The probabilistic feedforward neural network language model has been proposed in [ 1 ] . It consists of input , projection , hidden and output layers . At the input layer , N previous words are encoded using 1 - of - V coding , where V is size of the vocabulary . The input layer is then projected to a projection layer P that has dimensionality N \u00d7 D , using a shared projection matrix . As only N inputs are active at any given time , composition of the projection layer is a relatively cheap operation . The NNLM architecture becomes complex for computation between the projection and the hidden layer , as values in the projection layer are dense . For a common choice of N = 10 , the size of the projection layer ( P ) might be 500 to 2000 , while the hidden layer size H is typically 500 to 1000 units . Moreover , the hidden layer is used to compute probability distribution over all the words in the vocabulary , resulting in an output layer with dimensionality V . Thus , the computational complexity per each training example is Q = N \u00d7 D + N \u00d7 D \u00d7 H + H \u00d7 V , ( 2 ) where the dominating term is H \u00d7 V . However , several practical solutions were proposed for avoiding it ; either using hierarchical versions of the softmax [ 25 , 23 , 18 ] , or avoiding normalized models completely by using models that are not normalized during training [ 4 , 9 ] . With binary tree representations of the vocabulary , the number of output units that need to be evaluated can go down to around log 2 ( V ) . Thus , most of the complexity is caused by the term N \u00d7 D \u00d7 H . In our models , we use hierarchical softmax where the vocabulary is represented as a Huffman binary tree . This follows previous observations that the frequency of words works well for obtaining classes in neural net language models [ 16 ] . Huffman trees assign short binary codes to frequent words , and this further reduces the number of output units that need to be evaluated : while balanced binary tree would require log 2 ( V ) outputs to be evaluated , the Huffman tree based hierarchical softmax requires only about log 2 ( Unigram perplexity ( V ) ) . For example when the vocabulary size is one million words , this results in about two times speedup in evaluation . While this is not crucial speedup for neural network LMs as the computational bottleneck is in the N \u00d7 D \u00d7 H term , we will later propose architectures that do not have hidden layers and thus depend heavily on the ef\ufb01ciency of the softmax normalization . 2 . 2 Recurrent Neural Net Language Model ( RNNLM ) Recurrent neural network based language model has been proposed to overcome certain limitations of the feedforward NNLM , such as the need to specify the context length ( the order of the model N ) , and because theoretically RNNs can ef\ufb01ciently represent more complex patterns than the shallow neural networks [ 15 , 2 ] . The RNN model does not have a projection layer ; only input , hidden and output layer . What is special for this type of model is the recurrent matrix that connects hidden layer to itself , using time - delayed connections . This allows the recurrent model to form some kind of short term memory , as information from the past can be represented by the hidden layer state that gets updated based on the current input and the state of the hidden layer in the previous time step . The complexity per training example of the RNN model is Q = H \u00d7 H + H \u00d7 V , ( 3 ) where the word representations D have the same dimensionality as the hidden layer H . Again , the term H \u00d7 V can be ef\ufb01ciently reduced to H \u00d7 log 2 ( V ) by using hierarchical softmax . Most of the complexity then comes from H \u00d7 H . 3 2 . 3 Parallel Training of Neural Networks To train models on huge data sets , we have implemented several models on top of a large - scale distributed framework called DistBelief [ 6 ] , including the feedforward NNLM and the new models proposed in this paper . The framework allows us to run multiple replicas of the same model in parallel , and each replica synchronizes its gradient updates through a centralized server that keeps all the parameters . For this parallel training , we use mini - batch asynchronous gradient descent with an adaptive learning rate procedure called Adagrad [ 7 ] . Under this framework , it is common to use one hundred or more model replicas , each using many CPU cores at different machines in a data center . 3 New Log - linear Models In this section , we propose two new model architectures for learning distributed representations of words that try to minimize computational complexity . The main observation from the previous section was that most of the complexity is caused by the non - linear hidden layer in the model . While this is what makes neural networks so attractive , we decided to explore simpler models that might not be able to represent the data as precisely as neural networks , but can possibly be trained on much more data ef\ufb01ciently . The new architectures directly follow those proposed in our earlier work [ 13 , 14 ] , where it was found that neural network language model can be successfully trained in two steps : \ufb01rst , continuous word vectors are learned using simple model , and then the N - gram NNLM is trained on top of these distributed representations of words . While there has been later substantial amount of work that focuses on learning word vectors , we consider the approach proposed in [ 13 ] to be the simplest one . Note that related models have been proposed also much earlier [ 26 , 8 ] . 3 . 1 Continuous Bag - of - Words Model The \ufb01rst proposed architecture is similar to the feedforward NNLM , where the non - linear hidden layer is removed and the projection layer is shared for all words ( not just the projection matrix ) ; thus , all words get projected into the same position ( their vectors are averaged ) . We call this archi - tecture a bag - of - words model as the order of words in the history does not in\ufb02uence the projection . Furthermore , we also use words from the future ; we have obtained the best performance on the task introduced in the next section by building a log - linear classi\ufb01er with four future and four history words at the input , where the training criterion is to correctly classify the current ( middle ) word . Training complexity is then Q = N \u00d7 D + D \u00d7 log 2 ( V ) . ( 4 ) We denote this model further as CBOW , as unlike standard bag - of - words model , it uses continuous distributed representation of the context . The model architecture is shown at Figure 1 . Note that the weight matrix between the input and the projection layer is shared for all word positions in the same way as in the NNLM . 3 . 2 Continuous Skip - gram Model The second architecture is similar to CBOW , but instead of predicting the current word based on the context , it tries to maximize classi\ufb01cation of a word based on another word in the same sentence . More precisely , we use each current word as an input to a log - linear classi\ufb01er with continuous projection layer , and predict words within a certain range before and after the current word . We found that increasing the range improves quality of the resulting word vectors , but it also increases the computational complexity . Since the more distant words are usually less related to the current word than those close to it , we give less weight to the distant words by sampling less from those words in our training examples . The training complexity of this architecture is proportional to Q = C \u00d7 ( D + D \u00d7 log 2 ( V ) ) , ( 5 ) where C is the maximum distance of the words . Thus , if we choose C = 5 , for each training word we will select randomly a number R in range < 1 ; C > , and then use R words from history and 4 w ( t - 2 ) w ( t + 1 ) w ( t - 1 ) w ( t + 2 ) w ( t ) SUM INPUT PROJECTION OUTPUT w ( t ) INPUT PROJECTION OUTPUT w ( t - 2 ) w ( t - 1 ) w ( t + 1 ) w ( t + 2 ) CBOW Skip - gram Figure 1 : New model architectures . The CBOW architecture predicts the current word based on the context , and the Skip - gram predicts surrounding words given the current word . R words from the future of the current word as correct labels . This will require us to do R \u00d7 2 word classi\ufb01cations , with the current word as input , and each of the R + R words as output . In the following experiments , we use C = 10 . 4 Results To compare the quality of different versions of word vectors , previous papers typically use a table showing example words and their most similar words , and understand them intuitively . Although it is easy to show that word France is similar to Italy and perhaps some other countries , it is much more challenging when subjecting those vectors in a more complex similarity task , as follows . We follow previous observation that there can be many different types of similarities between words , for example , word big is similar to bigger in the same sense that small is similar to smaller . Example of another type of relationship can be word pairs big - biggest and small - smallest [ 20 ] . We further denote two pairs of words with the same relationship as a question , as we can ask : \u201dWhat is the word that is similar to small in the same sense as biggest is similar to big ? \u201d Somewhat surprisingly , these questions can be answered by performing simple algebraic operations with the vector representation of words . To \ufb01nd a word that is similar to small in the same sense as biggest is similar to big , we can simply compute vector X = vector ( \u201d biggest \u201d ) \u2212 vector ( \u201d big \u201d ) + vector ( \u201d small \u201d ) . Then , we search in the vector space for the word closest to X measured by cosine distance , and use it as the answer to the question ( we discard the input question words during this search ) . When the word vectors are well trained , it is possible to \ufb01nd the correct answer ( word smallest ) using this method . Finally , we found that when we train high dimensional word vectors on a large amount of data , the resulting vectors can be used to answer very subtle semantic relationships between words , such as a city and the country it belongs to , e . g . France is to Paris as Germany is to Berlin . Word vectors with such semantic relationships could be used to improve many existing NLP applications , such as machine translation , information retrieval and question answering systems , and may enable other future applications yet to be invented . 5 Table 1 : Examples of \ufb01ve types of semantic and nine types of syntactic questions in the Semantic - Syntactic Word Relationship test set . Type of relationship Word Pair 1 Word Pair 2 Common capital city Athens Greece Oslo Norway All capital cities Astana Kazakhstan Harare Zimbabwe Currency Angola kwanza Iran rial City - in - state Chicago Illinois Stockton California Man - Woman brother sister grandson granddaughter Adjective to adverb apparent apparently rapid rapidly Opposite possibly impossibly ethical unethical Comparative great greater tough tougher Superlative easy easiest lucky luckiest Present Participle think thinking read reading Nationality adjective Switzerland Swiss Cambodia Cambodian Past tense walking walked swimming swam Plural nouns mouse mice dollar dollars Plural verbs work works speak speaks 4 . 1 Task Description To measure quality of the word vectors , we de\ufb01ne a comprehensive test set that contains \ufb01ve types of semantic questions , and nine types of syntactic questions . Two examples from each category are shown in Table 1 . Overall , there are 8869 semantic and 10675 syntactic questions . The questions in each category were created in two steps : \ufb01rst , a list of similar word pairs was created manually . Then , a large list of questions is formed by connecting two word pairs . For example , we made a list of 68 large American cities and the states they belong to , and formed about 2 . 5K questions by picking two word pairs at random . We have included in our test set only single token words , thus multi - word entities are not present ( such as New York ) . We evaluate the overall accuracy for all question types , and for each question type separately ( se - mantic , syntactic ) . Question is assumed to be correctly answered only if the closest word to the vector computed using the above method is exactly the same as the correct word in the question ; synonyms are thus counted as mistakes . This also means that reaching 100 % accuracy is likely to be impossible , as the current models do not have any input information about word morphology . However , we believe that usefulness of the word vectors for certain applications should be positively correlated with this accuracy metric . Further progress can be achieved by incorporating information about structure of words , especially for the syntactic questions . 4 . 2 Maximization of Accuracy We have used a Google News corpus for training the word vectors . This corpus contains about 6B tokens . We have restricted the vocabulary size to 1 million most frequent words . Clearly , we are facing time constrained optimization problem , as it can be expected that both using more data and higher dimensional word vectors will improve the accuracy . To estimate the best choice of model architecture for obtaining as good as possible results quickly , we have \ufb01rst evaluated models trained on subsets of the training data , with vocabulary restricted to the most frequent 30k words . The results using the CBOW architecture with different choice of word vector dimensionality and increasing amount of the training data are shown in Table 2 . It can be seen that after some point , adding more dimensions or adding more training data provides diminishing improvements . So , we have to increase both vector dimensionality and the amount of the training data together . While this observation might seem trivial , it must be noted that it is currently popular to train word vectors on relatively large amounts of data , but with insuf\ufb01cient size 6 Table 2 : Accuracy on subset of the Semantic - Syntactic Word Relationship test set , using word vectors from the CBOW architecture with limited vocabulary . Only questions containing words from the most frequent 30k words are used . Dimensionality / Training words 24M 49M 98M 196M 391M 783M 50 13 . 4 15 . 7 18 . 6 19 . 1 22 . 5 23 . 2 100 19 . 4 23 . 1 27 . 8 28 . 7 33 . 4 32 . 2 300 23 . 2 29 . 2 35 . 3 38 . 6 43 . 7 45 . 9 600 24 . 0 30 . 1 36 . 5 40 . 8 46 . 6 50 . 4 Table 3 : Comparison of architectures using models trained on the same data , with 640 - dimensional word vectors . The accuracies are reported on our Semantic - Syntactic Word Relationship test set , and on the syntactic relationship test set of [ 20 ] Model Semantic - Syntactic Word Relationship test set MSR Word Relatedness Architecture Semantic Accuracy [ % ] Syntactic Accuracy [ % ] Test Set [ 20 ] RNNLM 9 36 35 NNLM 23 53 47 CBOW 24 64 61 Skip - gram 55 59 56 ( such as 50 - 100 ) . Given Equation 4 , increasing amount of training data twice results in about the same increase of computational complexity as increasing vector size twice . For the experiments reported in Tables 2 and 4 , we used three training epochs with stochastic gradi - ent descent and backpropagation . We chose starting learning rate 0 . 025 and decreased it linearly , so that it approaches zero at the end of the last training epoch . 4 . 3 Comparison of Model Architectures First we compare different model architectures for deriving the word vectors using the same training data and using the same dimensionality of 640 of the word vectors . In the further experiments , we use full set of questions in the new Semantic - Syntactic Word Relationship test set , i . e . unrestricted to the 30k vocabulary . We also include results on a test set introduced in [ 20 ] that focuses on syntactic similarity between words 3 . The training data consists of several LDC corpora and is described in detail in [ 18 ] ( 320M words , 82K vocabulary ) . We used these data to provide a comparison to a previously trained recurrent neural network language model that took about 8 weeks to train on a single CPU . We trained a feed - forward NNLM with the same number of 640 hidden units using the DistBelief parallel training [ 6 ] , using a history of 8 previous words ( thus , the NNLM has more parameters than the RNNLM , as the projection layer has size 640 \u00d7 8 ) . In Table 3 , it can be seen that the word vectors from the RNN ( as used in [ 20 ] ) perform well mostly on the syntactic questions . The NNLM vectors perform signi\ufb01cantly better than the RNN - this is not surprising , as the word vectors in the RNNLM are directly connected to a non - linear hidden layer . The CBOW architecture works better than the NNLM on the syntactic tasks , and about the same on the semantic one . Finally , the Skip - gram architecture works slightly worse on the syntactic task than the CBOW model ( but still better than the NNLM ) , and much better on the semantic part of the test than all the other models . Next , we evaluated our models trained using one CPU only and compared the results against publicly available word vectors . The comparison is given in Table 4 . The CBOW model was trained on subset 3 We thank Geoff Zweig for providing us the test set . 7 Table 4 : Comparison of publicly available word vectors on the Semantic - Syntactic Word Relation - ship test set , and word vectors from our models . Full vocabularies are used . Model Vector Training Accuracy [ % ] Dimensionality words Semantic Syntactic Total Collobert - Weston NNLM 50 660M 9 . 3 12 . 3 11 . 0 Turian NNLM 50 37M 1 . 4 2 . 6 2 . 1 Turian NNLM 200 37M 1 . 4 2 . 2 1 . 8 Mnih NNLM 50 37M 1 . 8 9 . 1 5 . 8 Mnih NNLM 100 37M 3 . 3 13 . 2 8 . 8 Mikolov RNNLM 80 320M 4 . 9 18 . 4 12 . 7 Mikolov RNNLM 640 320M 8 . 6 36 . 5 24 . 6 Huang NNLM 50 990M 13 . 3 11 . 6 12 . 3 Our NNLM 20 6B 12 . 9 26 . 4 20 . 3 Our NNLM 50 6B 27 . 9 55 . 8 43 . 2 Our NNLM 100 6B 34 . 2 64 . 5 50 . 8 CBOW 300 783M 15 . 5 53 . 1 36 . 1 Skip - gram 300 783M 50 . 0 55 . 9 53 . 3 Table 5 : Comparison of models trained for three epochs on the same data and models trained for one epoch . Accuracy is reported on the full Semantic - Syntactic data set . Model Vector Training Accuracy [ % ] Training time Dimensionality words [ days ] Semantic Syntactic Total 3 epoch CBOW 300 783M 15 . 5 53 . 1 36 . 1 1 3 epoch Skip - gram 300 783M 50 . 0 55 . 9 53 . 3 3 1 epoch CBOW 300 783M 13 . 8 49 . 9 33 . 6 0 . 3 1 epoch CBOW 300 1 . 6B 16 . 1 52 . 6 36 . 1 0 . 6 1 epoch CBOW 600 783M 15 . 4 53 . 3 36 . 2 0 . 7 1 epoch Skip - gram 300 783M 45 . 6 52 . 2 49 . 2 1 1 epoch Skip - gram 300 1 . 6B 52 . 2 55 . 1 53 . 8 2 1 epoch Skip - gram 600 783M 56 . 7 54 . 5 55 . 5 2 . 5 of the Google News data in about a day , while training time for the Skip - gram model was about three days . For experiments reported further , we used just one training epoch ( again , we decrease the learning rate linearly so that it approaches zero at the end of training ) . Training a model on twice as much data using one epoch gives comparable or better results than iterating over the same data for three epochs , as is shown in Table 5 , and provides additional small speedup . 4 . 4 Large Scale Parallel Training of Models As mentioned earlier , we have implemented various models in a distributed framework called Dis - tBelief . Below we report the results of several models trained on the Google News 6B data set , with mini - batch asynchronous gradient descent and the adaptive learning rate procedure called Ada - grad [ 7 ] . We used 50 to 100 model replicas during the training . The number of CPU cores is an 8 Table 6 : Comparison of models trained using the DistBelief distributed framework . Note that training of NNLM with 1000 - dimensional vectors would take too long to complete . Model Vector Training Accuracy [ % ] Training time Dimensionality words [ days x CPU cores ] Semantic Syntactic Total NNLM 100 6B 34 . 2 64 . 5 50 . 8 14 x 180 CBOW 1000 6B 57 . 3 68 . 9 63 . 7 2 x 140 Skip - gram 1000 6B 66 . 1 65 . 1 65 . 6 2 . 5 x 125 Table 7 : Comparison and combination of models on the Microsoft Sentence Completion Challenge . Architecture Accuracy [ % ] 4 - gram [ 32 ] 39 Average LSA similarity [ 32 ] 49 Log - bilinear model [ 24 ] 54 . 8 RNNLMs [ 19 ] 55 . 4 Skip - gram 48 . 0 Skip - gram + RNNLMs 58 . 9 estimate since the data center machines are shared with other production tasks , and the usage can \ufb02uctuate quite a bit . Note that due to the overhead of the distributed framework , the CPU usage of the CBOW model and the Skip - gram model are much closer to each other than their single - machine implementations . The result are reported in Table 6 . 4 . 5 Microsoft Research Sentence Completion Challenge The Microsoft Sentence Completion Challenge has been recently introduced as a task for advancing language modeling and other NLP techniques [ 32 ] . This task consists of 1040 sentences , where one word is missing in each sentence and the goal is to select word that is the most coherent with the rest of the sentence , given a list of \ufb01ve reasonable choices . Performance of several techniques has been already reported on this set , including N - gram models , LSA - based model [ 32 ] , log - bilinear model [ 24 ] and a combination of recurrent neural networks that currently holds the state of the art performance of 55 . 4 % accuracy on this benchmark [ 19 ] . We have explored the performance of Skip - gram architecture on this task . First , we train the 640 - dimensional model on 50M words provided in [ 32 ] . Then , we compute score of each sentence in the test set by using the unknown word at the input , and predict all surrounding words in a sentence . The \ufb01nal sentence score is then the sum of these individual predictions . Using the sentence scores , we choose the most likely sentence . A short summary of some previous results together with the new results is presented in Table 7 . While the Skip - gram model itself does not perform on this task better than LSA similarity , the scores from this model are complementary to scores obtained with RNNLMs , and a weighted combination leads to a new state of the art result 58 . 9 % accuracy ( 59 . 2 % on the development part of the set and 58 . 7 % on the test part of the set ) . 5 Examples of the Learned Relationships Table 8 shows words that follow various relationships . We follow the approach described above : the relationship is de\ufb01ned by subtracting two word vectors , and the result is added to another word . Thus for example , Paris - France + Italy = Rome . As it can be seen , accuracy is quite good , although there is clearly a lot of room for further improvements ( note that using our accuracy metric that 9 Table 8 : Examples of the word pair relationships , using the best word vectors from Table 4 ( Skip - gram model trained on 783M words with 300 dimensionality ) . Relationship Example 1 Example 2 Example 3 France - Paris Italy : Rome Japan : Tokyo Florida : Tallahassee big - bigger small : larger cold : colder quick : quicker Miami - Florida Baltimore : Maryland Dallas : Texas Kona : Hawaii Einstein - scientist Messi : mid\ufb01elder Mozart : violinist Picasso : painter Sarkozy - France Berlusconi : Italy Merkel : Germany Koizumi : Japan copper - Cu zinc : Zn gold : Au uranium : plutonium Berlusconi - Silvio Sarkozy : Nicolas Putin : Medvedev Obama : Barack Microsoft - Windows Google : Android IBM : Linux Apple : iPhone Microsoft - Ballmer Google : Yahoo IBM : McNealy Apple : Jobs Japan - sushi Germany : bratwurst France : tapas USA : pizza assumes exact match , the results in Table 8 would score only about 60 % ) . We believe that word vectors trained on even larger data sets with larger dimensionality will perform signi\ufb01cantly better , and will enable the development of new innovative applications . Another way to improve accuracy is to provide more than one example of the relationship . By using ten examples instead of one to form the relationship vector ( we average the individual vectors together ) , we have observed improvement of accuracy of our best models by about 10 % absolutely on the semantic - syntactic test . It is also possible to apply the vector operations to solve different tasks . For example , we have observed good accuracy for selecting out - of - the - list words , by computing average vector for a list of words , and \ufb01nding the most distant word vector . This is a popular type of problems in certain human intelligence tests . Clearly , there is still a lot of discoveries to be made using these techniques . 6 Conclusion In this paper we studied the quality of vector representations of words derived by various models on a collection of syntactic and semantic language tasks . We observed that it is possible to train high quality word vectors using very simple model architectures , compared to the popular neural network models ( both feedforward and recurrent ) . Because of the much lower computational complexity , it is possible to compute very accurate high dimensional word vectors from a much larger data set . Using the DistBelief distributed framework , it should be possible to train the CBOW and Skip - gram models even on corpora with one trillion words , for basically unlimited size of the vocabulary . That is several orders of magnitude larger than the best previously published results for similar models . An interesting task where the word vectors have recently been shown to signi\ufb01cantly outperform the previous state of the art is the SemEval - 2012 Task 2 [ 11 ] . The publicly available RNN vectors were used together with other techniques to achieve over 50 % increase in Spearman\u2019s rank correlation over the previous best result [ 31 ] . The neural network based word vectors were previously applied to many other NLP tasks , for example sentiment analysis [ 12 ] and paraphrase detection [ 28 ] . It can be expected that these applications can bene\ufb01t from the model architectures described in this paper . Our ongoing work shows that the word vectors can be successfully applied to automatic extension of facts in Knowledge Bases , and also for veri\ufb01cation of correctness of existing facts . Results from machine translation experiments also look very promising . In the future , it would be also interesting to compare our techniques to Latent Relational Analysis [ 30 ] and others . We believe that our comprehensive test set will help the research community to improve the existing techniques for estimating the word vectors . We also expect that high quality word vectors will become an important building block for future NLP applications . 10 7 Follow - Up Work After the initial version of this paper was written , we published single - machine multi - threaded C + + code for computing the word vectors , using both the continuous bag - of - words and skip - gram archi - tectures 4 . The training speed is signi\ufb01cantly higher than reported earlier in this paper , i . e . it is in the order of billions of words per hour for typical hyperparameter choices . We also published more than 1 . 4 million vectors that represent named entities , trained on more than 100 billion words . Some of our follow - up work will be published in an upcoming NIPS 2013 paper [ 21 ] . References [ 1 ] Y . Bengio , R . Ducharme , P . Vincent . A neural probabilistic language model . Journal of Ma - chine Learning Research , 3 : 1137 - 1155 , 2003 . [ 2 ] Y . Bengio , Y . LeCun . Scaling learning algorithms towards AI . In : Large - Scale Kernel Ma - chines , MIT Press , 2007 . [ 3 ] T . Brants , A . C . Popat , P . Xu , F . J . Och , and J . Dean . Large language models in machine translation . In Proceedings of the Joint Conference on Empirical Methods in Natural Language Processing and Computational Language Learning , 2007 . [ 4 ] R . Collobert and J . Weston . A Uni\ufb01ed Architecture for Natural Language Processing : Deep Neural Networks with Multitask Learning . In International Conference on Machine Learning , ICML , 2008 . [ 5 ] R . Collobert , J . Weston , L . Bottou , M . Karlen , K . Kavukcuoglu and P . Kuksa . Natural Lan - guage Processing ( Almost ) from Scratch . Journal of Machine Learning Research , 12 : 2493 - 2537 , 2011 . [ 6 ] J . Dean , G . S . Corrado , R . Monga , K . Chen , M . Devin , Q . V . Le , M . Z . Mao , M . A . Ranzato , A . Senior , P . Tucker , K . Yang , A . Y . Ng . , Large Scale Distributed Deep Networks , NIPS , 2012 . [ 7 ] J . C . Duchi , E . Hazan , and Y . Singer . Adaptive subgradient methods for online learning and stochastic optimization . Journal of Machine Learning Research , 2011 . [ 8 ] J . Elman . Finding Structure in Time . Cognitive Science , 14 , 179 - 211 , 1990 . [ 9 ] Eric H . Huang , R . Socher , C . D . Manning and Andrew Y . Ng . Improving Word Representations via Global Context and Multiple Word Prototypes . In : Proc . Association for Computational Linguistics , 2012 . [ 10 ] G . E . Hinton , J . L . McClelland , D . E . Rumelhart . Distributed representations . In : Parallel dis - tributed processing : Explorations in the microstructure of cognition . Volume 1 : Foundations , MIT Press , 1986 . [ 11 ] D . A . Jurgens , S . M . Mohammad , P . D . Turney , K . J . Holyoak . Semeval - 2012 task 2 : Measuring degrees of relational similarity . In : Proceedings of the 6th International Workshop on Semantic Evaluation ( SemEval 2012 ) , 2012 . [ 12 ] A . L . Maas , R . E . Daly , P . T . Pham , D . Huang , A . Y . Ng , and C . Potts . Learning word vectors for sentiment analysis . In Proceedings of ACL , 2011 . [ 13 ] T . Mikolov . Language Modeling for Speech Recognition in Czech , Masters thesis , Brno Uni - versity of Technology , 2007 . [ 14 ] T . Mikolov , J . Kopeck\u00b4y , L . Burget , O . Glembek and J . \u02c7Cernock\u00b4y . Neural network based lan - guage models for higly in\ufb02ective languages , In : Proc . ICASSP 2009 . [ 15 ] T . Mikolov , M . Kara\ufb01\u00b4at , L . Burget , J . \u02c7Cernock\u00b4y , S . Khudanpur . Recurrent neural network based language model , In : Proceedings of Interspeech , 2010 . [ 16 ] T . Mikolov , S . Kombrink , L . Burget , J . \u02c7Cernock \u00b4 y , S . Khudanpur . Extensions of recurrent neural network language model , In : Proceedings of ICASSP 2011 . [ 17 ] T . Mikolov , A . Deoras , S . Kombrink , L . Burget , J . \u02c7Cernock\u00b4y . Empirical Evaluation and Com - bination of Advanced Language Modeling Techniques , In : Proceedings of Interspeech , 2011 . 4 The code is available at https : / / code . google . com / p / word2vec / 11 [ 18 ] T . Mikolov , A . Deoras , D . Povey , L . Burget , J . \u02c7Cernock\u00b4y . Strategies for Training Large Scale Neural Network Language Models , In : Proc . Automatic Speech Recognition and Understand - ing , 2011 . [ 19 ] T . Mikolov . Statistical Language Models based on Neural Networks . PhD thesis , Brno Univer - sity of Technology , 2012 . [ 20 ] T . Mikolov , W . T . Yih , G . Zweig . Linguistic Regularities in Continuous Space Word Represen - tations . NAACL HLT 2013 . [ 21 ] T . Mikolov , I . Sutskever , K . Chen , G . Corrado , and J . Dean . Distributed Representations of Words and Phrases and their Compositionality . Accepted to NIPS 2013 . [ 22 ] A . Mnih , G . Hinton . Three new graphical models for statistical language modelling . ICML , 2007 . [ 23 ] A . Mnih , G . Hinton . A Scalable Hierarchical Distributed Language Model . Advances in Neural Information Processing Systems 21 , MIT Press , 2009 . [ 24 ] A . Mnih , Y . W . Teh . A fast and simple algorithm for training neural probabilistic language models . ICML , 2012 . [ 25 ] F . Morin , Y . Bengio . Hierarchical Probabilistic Neural Network Language Model . AISTATS , 2005 . [ 26 ] D . E . Rumelhart , G . E . Hinton , R . J . Williams . Learning internal representations by back - propagating errors . Nature , 323 : 533 . 536 , 1986 . [ 27 ] H . Schwenk . Continuous space language models . Computer Speech and Language , vol . 21 , 2007 . [ 28 ] R . Socher , E . H . Huang , J . Pennington , A . Y . Ng , and C . D . Manning . Dynamic Pooling and Unfolding Recursive Autoencoders for Paraphrase Detection . In NIPS , 2011 . [ 29 ] J . Turian , L . Ratinov , Y . Bengio . Word Representations : A Simple and General Method for Semi - Supervised Learning . In : Proc . Association for Computational Linguistics , 2010 . [ 30 ] P . D . Turney . Measuring Semantic Similarity by Latent Relational Analysis . In : Proc . Interna - tional Joint Conference on Arti\ufb01cial Intelligence , 2005 . [ 31 ] A . Zhila , W . T . Yih , C . Meek , G . Zweig , T . Mikolov . Combining Heterogeneous Models for Measuring Relational Similarity . NAACL HLT 2013 . [ 32 ] G . Zweig , C . J . C . Burges . The Microsoft Research Sentence Completion Challenge , Microsoft Research Technical Report MSR - TR - 2011 - 129 , 2011 . 12",
    "mund2023clathrin": "ARTICLE Clathrin coats partially preassemble and subsequently bend during endocytosis Markus Mund 1 , 2 * \ue840 , Aline Tschanz 1 , 3 * \ue840 , Yu - Le Wu 1 , 3 \ue840 , Felix Frey 4 \ue840 , Johanna L . Mehl 1 \ue840 , Marko Kaksonen 2 , 5 \ue840 , Ori Avinoam 1 , 6 \ue840 , Ulrich S . Schwarz 7 , 8 \ue840 , and Jonas Ries 1 \ue840 Eukaryotic cells use clathrin - mediated endocytosis to take up a large range of extracellular cargo . During endocytosis , a clathrin coat forms on the plasma membrane , but it remains controversial when and how it is remodeled into a spherical vesicle . Here , we use 3D superresolution microscopy to determine the precise geometry of the clathrin coat at large numbers of endocytic sites . Through pseudo - temporal sorting , we determine the average trajectory of clathrin remodeling during endocytosis . We find that clathrin coats assemble first on flat membranes to 50 % of the coat area before they become rapidly and continuously bent , and this mechanism is confirmed in three cell lines . We introduce the cooperative curvature model , which is based on positive feedback for curvature generation . It accurately describes the measured shapes and dynamics of the clathrin coat and could represent a general mechanism for clathrin coat remodeling on the plasma membrane . Introduction Endocytosis is an essential function of eukaryotic cells to in - ternalize molecules from their surface . The major endocytic pathway is clathrin - mediated endocytosis ( CME ) , which sup - ports the uptake of many diverse cargos including nutrients , signaling molecules , membrane proteins , and pathogens . During CME , a dense coat of proteins self - assembles on the inner leaflet of the plasma membrane . The membrane itself is bent into an \u03a9 - shaped invagination that is eventually pinched off to form a coated vesicle , completing endocytosis ( Kaksonen and Roux , 2018 ) . The major component of the coat is clathrin triskelia , which comprises three heavy and three light chains ( Fotin et al . , 2004 ) . When triskelia are bound to the plasma membrane by adaptor proteins , they form ordered clathrin lattices . The structural flexibility of triskelia allows these lattices to adapt variable ra - tios of predominantly pentagonal and hexagonal faces , forming both flat and curved geometries . Both geometries have been observed in vivo and in vitro , and their structure has been well characterized in vitro from a structural biology perspective ( Cheng et al . , 2007 ; Dannhauser and Ungewickell , 2012 ; Fotin et al . , 2004 ; Heuser and Kirchhausen , 1985 ; Morris et al . , 2019 ; Pearse , 1976 ; Smith et al . , 1998 ; Takei et al . , 1998 ; Ungewickell and Branton , 1981 ) . However , it remains elusive how clathrin coat formation and membrane bending are temporally and causally related during endocytosis in cells . In EM micrographs , it was observed early on that clathrin lattices can assume many different curvatures in cells ( Heuser , 1980 ) . Since then , two main models of clathrin coat formation during endocytosis have been put forward . In the constant area model ( CAM ) , clathrin grows to its final surface area as a flat coat , which then becomes continuously more curved until ves - icle formation is complete . This model assumes that all the dif - ferently curved clathrin structures are endocytic intermediates . Early observations suggested the presence of pentagonal and hexagonal faces in isolated coated vesicles ( Kanaseki and Kadota , 1969 ) . Combined with the observation of flat lattices being en - riched in hexagons , it was suggested that the integration of at least 12 pentagonal faces is a prerequisite for the formation of a spherical structure ( Heuser , 1980 ) . However , this would require extensive lattice remodeling , which was deemed thermody - namically and structurally unfavorable and thus unlikely to occur ( Kirchhausen , 1993 ) . The constant curvature model ( CCM ) was therefore formulated , which assumes that flat clathrin structures are not endocytic precursors . Instead , it was proposed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Cell Biology and Biophysics , European Molecular Biology Laboratory , Heidelberg , Germany ; 2 Department of Biochemistry , University of Geneva , Geneva , Switzerland ; 3 Candidate for Joint PhD Programme of EMBL and University of Heidelberg , Heidelberg , Germany ; 4 Kavli Institute of Nanoscience , Department of Bionanoscience , Delft University of Technology , Delft , Netherlands ; 5 NCCR Chemical Biology , University of Geneva , Geneva , Switzerland ; 6 Department of Biomolecular Sciences , Weizmann Institute of Science , Rehovot , Israel ; 7 Institute for Theoretical Physics and Bioquant , Heidelberg University , Heidelberg , Germany ; 8 Bioquant , Heidelberg University , Heidelberg , Germany . * M . Mund and A . Tschanz contributed equally to this paper . Correspondence to Jonas Ries : Jonas . ries @ embl . de J . L . Mehl \u2019 s current affiliation is ETH Zurich , Switzerland . F . Frey \u2019 s current affiliation is Institute of Science and Technology Austria , Klosterneuburg , Austria . \u00a9 2023 Mund , Tschanz , et al . This article is available under a Creative Commons License ( Attribution 4 . 0 International , as described at https : / / creativecommons . org / licenses / by / 4 . 0 / ) . Rockefeller University Press https : / / doi . org / 10 . 1083 / jcb . 202206038 1 of 14 J . Cell Biol . 2023 Vol . 222 No . 3 e202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 that the endocytic clathrin coat assumes its final curvature , i . e . , the curvature of the vesicle , directly from the start , while continuously growing in surface area over time . The constant curvature model had been prevalent in the field but was recently challenged by reports that flat clathrin coats can indeed change curvature during endocytosis ( Avinoam et al . , 2015 ; Bucher et al . , 2018 ; Scott et al . , 2018 ) . This has once again boosted interest in clathrin remodeling ( Chen and Schmid , 2020 ; Kaksonen and Roux , 2018 ; Sochacki and Taraska , 2018 ) . Recent years have seen numerous studies based on diverse methods that did not converge on a common result . Reports supported either the constant area model ( Avinoam et al . , 2015 ; Sochacki et al . , 2021 ) , the constant curvature model ( Willy et al . , 2021 ) , a combination of both models ( Bucher et al . , 2018 ; Tagiltsev et al . , 2021 ; Yoshida et al . , 2018 ) , or the simultaneous existence of both models within cells ( Scott et al . , 2018 ) . To conclusively understand this complex process , it would be de - sirable to directly visualize the 3D nanoscale structure of the endocytic clathrin coat in living cells . Unfortunately , however , to date , no method offers the necessary spatial and temporal resolution to do that . Here , we aim to circumvent this limitation by densely sampling the entire endocytic process in fixed cells , subsequently reconstructing the dynamic information . We developed a superresolution microscopy approach to quantitatively study the 3D clathrin coat architecture at endo - cytic sites . We used a novel model fitting framework to extract geometric parameters that allowed us to sort static images of clathrin lattices according to their progression along the endo - cytic timeline . The inferred endocytic dynamics allowed us to reconstruct the stereotypic remodeling of clathrin during en - docytosis at the nanoscale . In summary , we found that a clathrin coat first partially as - sembles on a flat membrane and then simultaneously grows in surface area and coat curvature . While initial bending occurs rapidly , it later slows down and endocytic sites are eventually paused in a state of high curvature before vesicle scission . This trend is conserved across cell lines , suggesting it is a common attribute that is not affected by cell type . Based on this data , we developed a new kinetic growth model , the cooperative curva - ture model ( CoopCM ) . It describes coat area growth as the ad - dition of clathrin to the lattice edge with a constant rate and assumes that the curvature of the coat increases toward a pre - ferred curvature , driven by the cooperative interplay of clathrin triskelia within the lattice . The CoopCM predicts a fast initial curvature increase that is slowed down progressively as the coat becomes spherical and shows excellent agreement with experimental data . Results Quantitative 3D superresolution imaging of clathrin structures Here , we used 3D single - molecule localization microscopy ( SMLM ) to systematically determine the precise geometry of individual clathrin - coated structures at the plasma membrane . For this , we optimized the sample preparation to label clathrin at endocytic sites as densely as possible using indirect immuno - fluorescence with polyclonal antibodies against clathrin light and heavy chains , which was crucial to achieve high - quality SMLM ( Mund and Ries , 2020 ) . We then localized sparsely ac - tivated single fluorophores by fitting an experimentally derived model of the astigmatic point - spread function using a method we developed previously ( Li et al . , 2018 ) . This improved the resolution to about 10 nm in x / y and 30 nm in z ( based on modal values of the localization precision at 3 . 9 nm in x / y and 12 . 5 nm in z ; see Materials and methods ) and reduced typically observed image distortions . The achieved image quality allowed us to clearly visualize and quantify the 3D clathrin coat shape at dif - ferent stages of endocytic site maturation ( Fig . 1 , A \u2013 C ) . The large majority of sites were single structures that were well - isolated from each other and exhibited great structural diversity . In addition , we noted several clusters of closely jux - taposed endocytic sites ( Fig . S1 A ) . In the isolated sites , we ob - serve a variety of shapes including flat , curved , dome - shaped , and spherical structures of different sizes ( Fig . 1 C ) , indicating that endocytosis has been arrested by fixation at different time points during endocytic progression . To quantify the size and shape of individual clathrin coats , we used LocMoFit , a computational pipeline based on a maximum - likelihood model fitting framework that we developed recently ( Wu et al . , 2023 ) . This framework directly fits 3D geometric models to the 3D point cloud of localizations ( Fig . 1 , D and E ) instead of rendered images and thereby harnesses the full in - formation contained in the SMLM data , including , for instance , the localization precision , rather than just spatial coordinates . We describe the clathrin coat as a spherical cap that is defined by a radius r and a closing angle \u03b8 ( Fig . 1 E ) . Our model also accounts for coat thickness , antibody size , and blurring due to the localization precision ( Materials and methods ) . Hence , it describes both shallow and highly curved structures equally well . Moreover , because \u03b8 increases during endocytosis from 0\u00b0 at flat sites to 180\u00b0 at complete vesicles , we could use this pa - rameter to sort the individual images according to endocytic progress ( Fig . 1 F ) . The clathrin coat grows in area and becomes more curved during endocytosis We first imaged immunostained clathrin in chemically fixed SK - MEL - 2 cells , where we focused on the bottom plasma membrane that was densely covered by endocytic sites . These cells have been extensively studied and are established model cells for clathrin - mediated endocytosis with well - characterized endo - cytic dynamics ( Doyon et al . , 2011 ; Aguet et al . , 2013 ; Avinoam et al . , 2015 ; Kaplan et al . , 2022 ) . Using the 3D model fitting pipeline , we determined radius R , closing angle \u03b8 , position , and rotation of 1 , 798 endocytic sites from 13 cells ( Fig . 2 A ) with high accuracy ( Fig . S2 ) . We found that two structural intermediates were enriched , while others were rare and only a small fraction of sites were completely flat ( Fig . 2 B ) . Slightly curved sites with \u03b8 \u2248 70 \u00b0 and strongly curved sites with \u03b8 \u2248 130 \u00b0 were enriched , indicating that those structural states are more frequently found in cells . We only rarely obtained sites with \u03b8 \u2248 180 \u00b0 , which would be expected for complete spherical coats , even though fully formed vesicles are found in our data ( Fig . S1 B ) . This indicates that at the time point of scission , the clathrin coat of nascent Mund et al . Journal of Cell Biology 2 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 vesicles is still incomplete at the neck or that the kinetics of scission are too transient for detection . Deeply invaginated clathrin - coated pits could further be sheared off during sample preparation , thus evading our detection . Curvatures H = 1 / R ranged from 0 to 0 . 022 nm \u2212 1 with a median of 0 . 011 nm \u2212 1 , corresponding to a median radius of 87 nm , and the surface areas A ranged from 9 , 000 to 140 , 000 nm 2 with a median of 54 , 000 nm 2 . These values agree well with previous measure - ments of the vesicular coat using EM ( Avinoam et al . , 2015 ) , where a median curvature of 0 . 015 nm \u2212 1 and median surface area of 54 , 500 nm 2 was measured . We then wanted to understand how the clathrin coat geom - etry changes during endocytosis . To this end , we used the closing angle parameter \u03b8 to sort endocytic sites relative to each other in time . Irrespective of whether endocytosis follows a constant curvature or constant area model , \u03b8 monotonically increases from early to late endocytic time points ( Fig . 2 C ) , and can thus be used as an unbiased proxy for endocytic progression ( Avinoam et al . , 2015 ) . The curvature H was strongly correlated with \u03b8 , indicating that coat curvature increases continuously during endocytosis ( Fig . 2 D ) . Similarly , the surface area A increased from 32 , 000 nm 2 ( median of 5 % of sites with smallest \u03b8 ) to 50 , 000 nm 2 ( median of 5 % of sites with highest \u03b8 ) . The projected area A p decreased from 31 , 000 to 13 , 000 nm 2 ( median of 5 % of sites with lowest and highest \u03b8 respectively ) , which is in close agreement with previous EM measurements ( Bucher et al . , 2018 ) . It is readily obvious that our data are incompatible with the constant curvature model , as the curvature is not constant , but increases monotonically with \u03b8 . Just as clearly , our data do not support the constant area model because the coat surface also increases during endocytosis . Almost all data points are part of one continuous point cloud ( Fig . 2 D ) , indicating a continuous transition between structural states during endocytosis . We noticed an additional small , dis - connected set of data points representing 8 . 5 % of all sites that correspond to endocytic sites with curvatures above 0 . 016 nm \u2212 1 and \u03b8 of 80\u00b0 \u2013 180\u00b0 ( Fig . 2 D , and example structures in Fig . S3 , A and B ) . In a control experiment to check whether these are endocytic structures , we selectively analyzed clathrin structures that colocalized with AP - 2 , a bona fide marker for CME . We did not observe AP - 2 in any of the disconnected sites , from which we conclude that they indeed do not belong to CME . These small structures could represent coated vesicles from the trans Golgi ( Fig . S3 ) . The cooperative curvature model of clathrin coat remodeling Our data clearly showed that clathrin coats grow and become more curved as the plasma membrane gets bent to produce a vesicle . Since the data quantitatively describes all 3D geometries that the endocytic clathrin coat assumes , it allowed us to move toward a mathematical model of clathrin coat formation during endocytosis . Here , we introduce the Cooperative Curvature Model ( CoopCM ) , which describes coat growth based on known struc - tural and dynamical properties of clathrin coats ( Fig . 3 A ) . First , Figure 1 . 3D single - molecule localization microscopy ( SMLM ) of clathrin - coated structures . ( A ) 3D SMLM image of clathrin immunolabeled with AF647 at the bottom membrane of a SK - MEL - 2 cell . ( B ) Enlarged view of the region indicated in A . ( C ) All structures found in B are shown in top view ( xy ) and as 50 nm - thick z - slices ( xz ) , and orientation of slice is indicated by a dotted line . Scale bars are 10 \u00b5m ( A ) ; 1 \u00b5m ( B ) ; and 100 nm ( C ) . ( D ) Geometric analysis pipeline . All clathrin coats are segmented , and their localizations are directly fitted with a spherical cap model using the fitting framework LocMoFit ( Wu et al . , 2023 ) . ( E ) InLocMoFit , individualclathrincoats are parametrizedbytheirsize ( radius R ) , closing angle ( \u03b8 ) , position ( x 0 , y 0 , z 0 ) , and 3D orientation . ( F ) Using \u03b8 asa proxy for endocytic progression , the relative endocytic time point for each structure can be determined . Mund et al . Journal of Cell Biology 3 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 we assume that the clathrin coat starts growing on a flat mem - brane . While triskelia have been shown to exchange dynamically during endocytosis ( Avinoam et al . , 2015 ) , net growth of the area A occurs via the addition of triskelia at the lattice edge \u0190 with a constant growth rate k on ( Eq . 1 ) . d dtA (cid:2) \u02d9 A (cid:2) k on \u0190 . ( 1 ) Moreover , we assume that the intrinsic pucker angle of in - dividual triskelia and their interactions in a lattice together translate into an overall preferred curvature H 0 of the clathrin Figure 2 . Quantitative analysis of clathrin - coated structures in SK - MEL - 2 cells . ( A ) Clathrin coat geometry is quantified using LocMoFit . The fitted spherical cap model is drawn as a circle with its outer radius ( top row , xy ) , as cross - sectional profile ( 50 nm xz slices , middle row ) , and as surface with the corresponding closing angle \u03b8 ( bottom row ) . Scale bar , 100 nm . ( B ) Distributions of closing angle \u03b8 ( median = 108\u00b0 ) , curvature H ( median = 0 . 011 nm \u2212 1 ) , and surface area A ( median = 54 , 000 nm 2 ) of endocytic sites in SK - MEL - 2 cells ( n = 1 , 798 sites , N = 13 cells ) as determined from the model fit . ( C ) Two previously proposed mechanistic models for clathrin coat assembly during endocytosis ( for details , see text ) . In both scenarios , \u03b8 increases monotonically and is thus a proxy for endocytic progression . ( D ) Development of curvature , surface area , and projected area of the clathrin coat during endocytosis . Color indicates point density . Mund et al . Journal of Cell Biology 4 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 coat as a whole . However , the initially flat assembly suggests that this preferred curvature cannot be realized in an immature lattice . We hypothesize that cooperative processes in the coat are required to increase curvature . Coat curvature H then increases asymptotically toward H 0 at an initial rate \u03b3 , slowing down to - ward zero when the preferred curvature is reached ( Eq . 2 ) . d d \u03b8 H (cid:2) \u03b3 1 \u2212 H 2 H 20 (cid:1) (cid:3) . ( 2 ) Choosing a quadratic dependence of the rate of curvature change on curvature is a simple way to represent cooperativity in the lattice , which recently has been demonstrated in experi - ments ( Sochacki et al . , 2021 ; Zeno et al . , 2021 ) and fits our data more accurately than the linear relationship , which would cor - respond to a less cooperative process ( Appendix ) . Eq . 2 can be solved analytically to yield an expression of the curvature H depending on \u03b8 ( Eq . 3 ) , which can then be fitted to our data . H \u03b8 ( ) (cid:2) H 0 tanh \u03b3 \u03b8 H 0 (cid:1) (cid:3) . ( 3 ) Analogous expressions for other geometric parameters like surface area can be derived straightforwardly . The fit of this CoopCM to the data of curvature H in relation to the closing angle \u03b8 shows excellent agreement ( Fig . 3 B ) . In comparison , the CAM fitted the curvature data slightly worse than the CoopCM , and the CCM agreed poorly with the data . The improved fit of the CoopCM compared with the CAM or CCM is not a result of the additional fitting parameter , as we showed using the Bayesian Information Criterion ( see Appendix ) . From the curvature fits , we calculated the corresponding curves for surface area ( Fig . 3 C ) and the edge length of the clathrin coat ( Fig . 3 D ) . The edge length decreases monotonically and approaches zero at \u03b8 = 180 \u00b0 , thereby stopping growth ac - cording to Eq . 1 . Both graphs again highlight the very close agreement of the model prediction with the experimental data . From the fit , we determined that invagination occurs when about half of the total coat area has grown ( A 0 = 0 . 51 ) , and that the preferred curvature of a clathrin coat is H 0 = 0 . 014 nm \u2212 1 , corresponding to a radius of R 0 = 72 nm . The model yields nearly identical parameters when the surface area or edge length is fitted instead of curvature ( Table 1 ) , highlighting its robustness . We then decided to test if the observed mode of clathrin re - modeling is specific for SK - MEL - 2 cells . For this , we analyzed the endocytic clathrin ultrastructure also in U2OS cells and 3T3 mouse fibroblasts ( Fig . S4 ; and Tables 2 and 3 ) . In these cell lines , just like in SK - MEL - 2 cells , the curvature as well as the surface area continuously increase throughout endocytosis . We ob - served that the preferred curvature of the clathrin coat is smaller in U2OS ( H 0 = 0 . 012 nm \u2212 1 , R 0 = 85 nm ) and 3T3 cells ( H 0 = 0 . 011 nm \u2212 1 , R 0 = 89 . 7 nm ) compared with SK - MEL - 2 ( H 0 = 0 . 014 nm \u2212 1 , R 0 = 72 nm ) . This suggests that the average size of vesicles formed in CME is cell - line specific . The fraction of the surface area acquired on the flat membrane is very similar for all three cell lines , with U2OS - derived sites initiating curvature at A 0 = 0 . 52 and 3T3 sites at A 0 = 0 . 45 ( Tables 2 and 3 ) . Taken together , we have shown in several cell lines that clathrin coats neither grow via constant curvature or constant area pathways , but rather first grow flat and then acquire curvature and more surface area simultaneously , with a nonlinear mode of curvature generation . Temporal reconstruction of structural dynamics in endocytosis We systematically segmented all endocytic sites in the super - resolution images and thereby obtained a large dataset where all endocytic time points were sampled homogeneously . The dis - tribution of structural states within the dataset is thus repre - sentative of their lifetime , with long - lived , stable curvature states being overrepresented in the data compared to transient states . This opens up the possibility to reconstruct the temporal progression of clathrin remodeling during endocytosis and to ask whether clathrin coats acquire their curvatures at a constant rate or if there are certain curvature transitions that occur more rapidly than others . For this , we sorted all endocytic sites by \u03b8 . The rank of an endocytic site thus corresponds to its pseudotime , which de - scribes its relative time point between 0 and 1 along the endo - cytic trajectory ( Fig . 4 A ) . As our model ( Eq . 1 ) describes the dynamic growth of the clathrin coat , we can solve it to derive an expression for \u03b8 over time t ( Eq . 4 ) , where the coat starts out flat and then starts to generate curvature , increasing \u03b8 over time . Figure 3 . Model for clathrin coat growth . ( A ) Schematic of the cooperative curvature model , where clathrin lattices grow by the addition of triskelia to the edge at a constant growth rate k on . Curvature generation is driven toward a preferred curvature , ultimately creating a spherical vesicle . ( B ) Distinct clathrin growth models and rolling median ( window width = 82 sites ) fitted to curvature over \u03b8 . ( C and D ) The resulting fitting parameters are then used to map the same models also over ( C ) surface area and ( D ) edge length . ( n = 1 , 645 sites , N = 13 cells ) . Mund et al . Journal of Cell Biology 5 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 \u03b8 t ( ) (cid:2) \ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03 24 \u03b3 k on 8 \u03b3 2 H \u2212 20 \u2212 1 t s . ( 4 ) The square root dependence of \u03b8 on time t reflects the slowing down of curvature generation as the clathrin coat ap - proaches its preferred curvature . This expression fits the pseudotime - resolved data remarkably well ( Fig . 4 B ) . Consistent with our previous reasoning , a linear model did not agree well with the data ( Appendix ) , since it gives a linear invagination speed for small t , emphasizing the validity of our cooperative curvature model , which leads to the characteristic square root dependence . The data slightly deviates from the model in the early phase and more notably in the very end for pseudotimes close to 1 . This potentially indicates that clathrin geometry is influenced by other factors besides the coat itself close to vesicle scission . Since this pseudotime - resolved dataset was generated from a large number of endocytic sites in many cells , we effectively generated the average trajectories of how curvature , surface area , projected area , and lattice edge change during endocytosis in SK - MEL - 2 cells ( Fig . 4 , C \u2013 F ) . We observe comparatively few flat clathrin coats . This shows that clathrin lattices are only transiently flat at the beginning of endocytosis and might rep - resent an energetically unfavorable conformation ( Fig . 4 B ) . A fast transition from a flat to a curved structure could further be mediated independently of clathrin coat ( Zhao et al . , 2017 ) , e . g . , by the action of BAR domain proteins ( Henne et al . , 2010 ) or at the base of filopodial projections . We further find comparatively many structures with a curvature of \u03b8 \u2248 70 \u00b0 and \u03b8 \u2248 130 \u00b0 , indi - cating more long - lived endocytic stages . As the surface area is constantly increasing over pseudotime , we assume that the en - richment at \u03b8 \u2248 130 \u00b0 represents a stalling point , where vesicles are formed by the addition of the final clathrin triskelia and potentially other factors that enable the recruitment and me - chanic function of dynamin during vesicle scission . Similarly , the enrichment at \u03b8 \u2248 70 \u00b0 could be indicative of further re - cruitment of regulatory components , potentially supporting a previously suggested endocytic checkpoint ( Loerke et al . , 2009 ) . Table 1 . Summary of clathrin coat growth model fits in SK - MEL - 2 . Curvature H ( \u03b8 ) Surface area A ( \u03b8 ) Edge length \u0190 ( \u03b8 ) Projected area A p ( \u03b8 ) CAM A 58 , 300 \u00b1 400 nm 2 56 , 600 \u00b1 400 nm 2 49 , 400 \u00b1 500 nm 2 48 , 800 \u00b1 500 nm 2 CCM R 94 . 5 \u00b1 0 . 7 nm 84 . 8 \u00b1 0 . 4 nm 97 . 3 \u00b1 0 . 8 nm 92 . 8 \u00b1 0 . 7 nm CoopCM \u03b3 ( 9 . 4 \u00b1 0 . 1 ) 10 - 3 nm \u2212 1 ( 9 . 0 \u00b1 0 . 1 ) 10 - 3 nm \u2212 1 ( 9 . 4 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 ( 9 . 1 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 H 0 ( 13 . 9 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 ( 13 . 8 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 ( 13 . 4 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 ( 13 . 6 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 R 0 72 . 0 \u00b1 0 . 6 nm 72 . 2 \u00b1 0 . 6 nm 74 \u00b1 0 . 8 nm 73 . 4 \u00b1 0 . 9 nm A 0 0 . 51 0 . 55 0 . 49 0 . 52 k on 78 . 1 nm \u02dc s \u2212 1 72 . 2 nm \u02dc s \u2212 1 83 . 4 nm \u02dc s \u2212 1 78 nm \u02dc s \u2212 1 Fitted parameter values for the constant area model ( CAM ) , the constant curvature model ( CCM ) and the cooperative curvature model ( CoopCM ) when fitting curvature H ( \u03b8 ) , surface area A ( \u03b8 ) , edge length \u0190 ( \u03b8 ) , and projected area A p ( \u03b8 ) . A : Surface area fitted with CAM ; R : Radius fitted with CCM ; \u03b3 : Constant rate of curvature increase fitted with CoopCM ; H 0 : Preferred curvature of the clathrin coat fitted with CoopCM ; R 0 : preferred radius of the clathrin coat fitted with CoopCM ; A 0 : Fraction of surface area growing as a flat lattice before curvature initiation , defined as A 0 = A ( \u03b8 = 0 . 01 ) / A ( \u03b8 = 180\u00b0 ) ; k on : local growth rate obtained from \u03b8 ( t ) fitted with CoopCM and measured in nm per pseudotime units \u02dc s - 1 . N = 13 cells ; n = 1 , 645 sites . Table 2 . Summary of clathrin coat growth model fits in U2OS . Curvature H ( \u03b8 ) Surface area A ( \u03b8 ) Edge length \u0190 ( \u03b8 ) Projected area A p ( \u03b8 ) CAM A 80 , 800 \u00b1 1 , 600 nm 2 76 , 900 \u00b1 1 , 700 nm 2 65 , 900 \u00b1 1 , 600 nm 2 65 , 400 \u00b1 1 , 600 nm 2 CCM R 119 . 1 \u00b1 2 . 5 nm 101 . 6 \u00b1 1 . 5 nm 122 . 8 \u00b1 2 . 7 nm 115 \u00b1 2 . 4 nm CoopCM \u03b3 ( 7 . 9 \u00b1 0 . 2 ) 10 \u2212 3 nm - 1 ( 7 . 7 \u00b1 0 . 2 ) 10 \u2212 3 nm \u2212 1 ( 8 . 1 \u00b1 0 . 2 ) 10 \u2212 3 nm \u2212 1 ( 7 . 9 \u00b1 0 . 2 ) 10 \u2212 3 nm \u2212 1 H 0 ( 11 . 8 \u00b1 0 . 2 ) 10 \u2212 3 nm \u2212 1 ( 11 . 5 \u00b1 0 . 2 ) 10 \u2212 3 nm - 1 ( 11 . 0 \u00b1 0 . 3 ) 10 \u2212 3 nm - 1 ( 11 . 1 \u00b1 0 . 3 ) 10 \u2212 3 nm - 1 R 0 85 . 0 \u00b1 1 . 8 nm 87 . 1 \u00b1 1 . 7 nm 90 . 5 \u00b1 2 . 7 nm 90 . 2 \u00b1 2 . 6 nm A 0 0 . 52 0 . 52 0 . 45 0 . 47 k on 89 . 3 nm \u02dc s \u2212 1 91 . 5 nm \u02dc s \u2212 1 110 . 3 nm \u02dc s \u2212 1 105 . 5 nm \u02dc s \u2212 1 Fittedparameter valuesfor the constantareamodel ( CAM ) , the constantcurvature model ( CCM ) , and the cooperative curvaturemodel ( CoopCM ) whenfitting curvature H ( \u03b8 ) , surface area A ( \u03b8 ) , edge length \u0190 ( \u03b8 ) , and projected area A p ( \u03b8 ) . A : Surface area fitted with CAM ; R : Radius fitted with CCM ; \u03b3 : Constant rate of curvature increase fitted with CoopCM ; H 0 : Preferred curvature of the clathrin coat fitted with CoopCM ; R 0 : preferred radius of the clathrin coat fitted with CoopCM ; A 0 : Fraction of surface area growing as a flat lattice before curvature initiation , defined as A 0 = A ( \u03b8 = 0 . 01 ) / A ( \u03b8 = 180\u00b0 ) ; k on : local growth rate obtained from \u03b8 ( t ) fitted with CoopCM and measured in nm per pseudotime units \u02dc s \u2212 1 . N = 3 cells ; n = 241 sites . Mund et al . Journal of Cell Biology 6 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 This functional interpretation is supported by the observation of similar peaks in U2OS ( at \u03b8 \u2248 60 \u00b0 and \u2248 130 \u00b0 ) as well as 3T3 cells ( at \u03b8 \u2248 50 \u00b0 and \u2248 130 \u00b0 ) . Within the first 10 % of pseudotemporal progression , the clathrin coat rapidly acquires shallow curvature of H = 0 . 007 nm \u2212 1 ( R = 134 nm ) . It then becomes gradually more bent up to Table 3 . Summary of clathrin coat growth model fits in 3T3 mouse fibroblasts . Curvature H ( \u03b8 ) Surface area A ( \u03b8 ) Edge length \u0190 ( \u03b8 ) Projected area A p ( \u03b8 ) CAM A 90 , 100 \u00b1 1 , 000 nm 2 83 , 700 \u00b1 1 , 000 nm 2 65 , 800 \u00b1 1 , 000 nm 2 66 , 200 \u00b1 1 , 000 nm 2 CCM R 115 . 8 \u00b1 1 . 5 nm 98 . 3 \u00b1 0 . 7 nm 121 . 1 \u00b1 2 nm 108 . 5 \u00b1 1 . 5 nm CoopCM \u03b3 ( 8 . 2 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 ( 7 . 8 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 ( 7 . 9 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 ( 7 . 8 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 H 0 ( 11 . 3 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 ( 11 . 2 \u00b1 0 . 1 ) 10 \u2212 3 nm \u2212 1 ( 11 . 2 \u00b1 0 . 2 ) 10 \u2212 3 nm \u2212 1 ( 11 . 2 \u00b1 0 . 2 ) 10 \u2212 3 nm \u2212 1 R 0 88 . 8 \u00b1 0 . 7 nm 89 . 3 \u00b1 0 . 7 nm 89 . 1 \u00b1 1 . 4 nm 89 . 5 \u00b1 1 . 3 nm A 0 0 . 45 0 . 49 0 . 47 0 . 49 k on 128 . 4 nm \u02dc s \u2212 1 120 . 8 nm \u02dc s \u2212 1 123 . 9 nm \u02dc s \u2212 1 121 nm \u02dc s \u2212 1 Fittedparameter valuesfor the constantareamodel ( CAM ) , the constantcurvature model ( CCM ) , and the cooperative curvaturemodel ( CoopCM ) whenfitting curvature H ( \u03b8 ) , surface area A ( \u03b8 ) , edge length \u0190 ( \u03b8 ) , and projected area A p ( \u03b8 ) . A : Surface area fitted with CAM ; R : Radius fitted with CCM ; \u03b3 : Constant rate of curvature increase fitted with CoopCM ; H 0 : Preferred curvature of the clathrin coat fitted with CoopCM ; R 0 : preferred radius of the clathrin coat fitted with CoopCM ; A 0 : Fraction of surface area growing as a flat lattice before curvature initiation , defined as A 0 = A ( \u03b8 = 0 . 01 ) / A ( \u03b8 = 180\u00b0 ) ; k on : local growth rate obtained from \u03b8 ( t ) fitted with CoopCM and measured in nm per pseudotime units \u02dc s \u2212 1 . N = 7 cells ; n = 688 sites . Figure 4 . Temporal reconstruction of clathrin coat remodeling . ( A ) Endocytic sites are sorted by \u03b8 to reconstruct pseudotime . Enriched \u03b8 states , for example the peak at 135\u00b0 , represent long - lived states that remain for extended periods in pseudotime . Color code represents the same number of sites in both histograms . ( B ) The square - root dependencebetween \u03b8 and pseudotimeapproximatedby the cooperative curvature model ( redline ) . ( C ) Curvaturegeneration over pseudotime is approximated by the cooperative curvature model . ( D \u2013 F ) Fit results in C were used to describe ( D ) edge length , ( E ) surface area , and ( F ) projected area change over pseudotime . A rolling median ( window of 82 sites ) is plotted alongside ( black line ) . ( G ) Superresolution averages for distinct endocytic stages , resulting from all collected snapshots . Each bin contains the same number of snapshots of clathrin - coated structures sorted along their pseudotime ( n = 163 per bin ) , so that all bins represent equally long pseudotime intervals . Individual sites were rescaled to the median bin radius and aligned by their center coordinates as well as rotation angles . Scale bar is 100 nm . ( n = 1 , 645 sites , N = 13 cells ) . Mund et al . Journal of Cell Biology 7 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 \u2248 60 % of the endocytic timeline , when the sites reach an average curvature of 0 . 012 nm \u2212 1 ( R = 83 nm ) . During the last \u2248 40 % of progression , the curvature increase almost stops , until ves - icle scission occurs . Interestingly , the earliest sites in our dataset already contain \u2248 50 % of the surface area of the final vesicles , which is also reflected in the fitting results ( A 0 = 0 . 51 ; Fig . 4 E ) . This indicates that the initial assembly of the clathrin coat occurs very rapidly or is in part below our detection limit . Finally , we made a 3D nanoscale movie , which vividly illus - trates the assembly and remodeling of the clathrin coat during endocytosis from pseudotime - resolved averages of our data ( Fig . 4 G ) . This yielded a nanoscale pseudo - temporal movie of endocytosis in SK - MEL - 2 cells ( Video 1 ) . We performed identical analyses for U2OS - and 3T3 - derived clathrin sites and obtained highly similar results as for SK - MEL - 2 cells , indicating that this trajectory represents a general pathway for clathrin coat remodeling during endocytosis ( Fig . S4 ; and Tables 2 and 3 ) . Also , these data can be represented as high - resolution averages ( Fig . 5 ) . Discussion Quantitative description of clathrin coat ultrastructure The nature of clathrin assembly during endocytosis has become a classical question in cell biology that remains unresolved ( re - viewed in Chen and Schmid , 2020 ; Sochacki and Taraska , 2018 ) . There have been two main competing mechanistic models of how triskelia form a coat : the constant curvature model predicts that the clathrin coat assumes its final curvature directly from the start , while continuously increasing in surface area over time . In contrast , the constant area model predicts that clathrin grows to its final surface area as a flat coat , which then becomes continuously more curved until vesicle formation is complete . Each model is supported by a number of studies , which mostly relied on electron microscopy techniques . Among them , 3D correlative electron tomography yields exquisite 3D endo - cytic membrane shapes ( Avinoam et al . , 2015 ) but can be tedious to extend toward large numbers of cells . Platinum replica elec - tron microscopy offers large fields of view and thus superb throughput , but gives limited 3D resolution ( Bucher et al . , 2018 ; Sochacki et al . , 2017 ) . We reason that the trajectory of clathrin coat assembly dur - ing endocytosis could best be determined by systematically imaging large numbers of endocytic sites with high 3D resolu - tion . Therefore , we used SMLM , which combines high 3D res - olution with high throughput . Although lower than in electron microscopy , the resolution we achieved allowed us to resolve the precise 3D coat ultrastructure of clathrin - coated pits . Impor - tantly , due to the molecular specificity for clathrin , we were able to segment all endocytic sites at the bottom surface of the cells in our images in an unbiased way , thus ensuring homogenous sampling of the structural variety of clathrin coats . We applied a novel maximum - likelihood - based fitting framework that we developed recently ( Wu et al . , 2023 ) , which allows fitting complex geometric models to localization point clouds . Since the fit is applied directly to the localizations , it considers all quantitative parameters of all localizations , most notably localization uncertainty . This results in the precise and reliable quantification of the underlying structures ( Wu et al . , 2020 ) , even when taking linkage error through indirect im - munolabeling into account ( Fig . S5 , A \u2013 L ) . We described the clathrin coat as a spherical cap , which matched the large majority of sites very well . Additionally , in our data , we observed that some sites are asymmetric , ellipsoi - dal , or deformed more irregularly ( Fig . S1 C ) . We do not currently evaluate this asymmetry . Rather , we introduced rotational symmetry during the averaging ( Fig . 4 G ) , where we aligned Figure 5 . Temporal reconstruction of clathrin coat remodeling across three different cell lines . ( A and B ) Superresolution averages for distinct en - docytic stages , resulting from all collected snapshots for ( A ) U2OS ( n = 241 sites , N = 3 cells ) and ( B ) 3T3 ( n = 688 sites , N = 7 cells ) . Each bin contains the same number of snapshots of clathrin - coated structures sorted along their pseudotime ( n U2OS = 24 per bin , n 3T3 = 68 ) and represent equally long pseudotime intervals . Mund et al . Journal of Cell Biology 8 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 each site based on their model parameters and thus averaged out many irregularities . A deviation of the cross - sectional profile from a circle is nevertheless preserved in the averag - ing ( Fig . S2 , C \u2013 G ) , and in future studies , more complex geo - metric models will enable analyzing the structure and asymmetries of the coat in more detail . Flat clathrin - coated structures In the first step of our analysis pipeline , we segmented all clathrin - coated structures in the superresolution images . We thereby aimed to achieve homogenous sampling to capture all endocytic intermediates with the same probability irrespective of time point , size , and orientation . Interestingly , the surfaces of the earliest detectable endocytic sites already contained half the surface area of sites in the latest CME stages . How do these earliest sites assemble ? We cannot sort the very flat sites , which are comparatively rare in our dataset , in time using \u03b8 because their \u03b8 is close to zero . However , their relative fraction among all sites is still informative . As short - lived states are underrepre - sented , the relative absence of small flat sites in our data in - dicates that the initial assembly occurs rapidly . In addition to the short lifetimes of small flat sites , two technical reasons could potentially contribute to their rare oc - currence in our data . Due to their small size and potentially sub - stoichiometric labeling , which is also noticeable as holes in larger structures ( Fig . S1 D ) , they might not have sufficient immunostaining signal and thus they might be hard to be differentiated from small clusters of primary and secondary antibodies that are routinely observed in SMLM images . Ad - ditionally , the very first small clathrin assemblies might be meta - stable with mostly loosely bound triskelia , and thus might not be stabilized well by chemical fixation . However , it is im - portant to note that very small sites are also rarely found in previously published electron microscopy datasets ( Avinoam et al . , 2015 ; Bucher et al . , 2018 ; Sochacki et al . , 2021 ) . Since EM should not suffer from the same technical limitations , it seems likely that these states are indeed short - lived . Generally , completely flat clathrin coats are rare in our data . Instead , we observed that clathrin coats quickly acquire curva - ture already early in endocytosis . This agrees with the early observation made using EM ( Heuser , 1980 ) that even flat structures already had some small degree of curvature . Inter - estingly , an enrichment of pentagons was observed toward the edges of flat lattices , leading to the hypothesis that even without additional formation of pentagons these flat lattices have the built - in capacity to become curved and are not very stable . This notion is in agreement with recent demonstrations that flat clathrin coats store energy , which can drive coat bending without additional factors ( Sochacki et al . , 2021 ; Tagiltsev et al . , 2021 ) . The importance of curvature generation for clathrin coat maturation is supported by previous studies that suggest failure to initiate significant curvature as a hallmark of abortive events ( Loerke et al . , 2009 ; Wang et al . , 2020 ) . Observed in live - cell studies , these dim and short - lived events fail to significantly increase in intensity and do not undergo scission ( Loerke et al . , 2009 ; Mettlen et al . , 2009 ) . As structures in our data set mostly contain at least shallow curvatures at minimally half the final coat area , we are likely not capturing these abortive events , and they have negligible impact on our analysis . The cooperative curvature model Here , we visualized the shapes of clathrin coats with high 3D resolution . This was crucial to precisely measure the clathrin curvature throughout the endocytic timeline ranging from flat to highly curved structures . Especially shallow curvatures are hard to accurately assess with 2D imaging . Thus , our data en - abled us to robustly test different growth models for clathrin during endocytosis . The constant curvature model predicts a continuous increase in surface area , constant coat curvature , as well as a edge length that increases until a half - spherical geometry is reached and then decreases again . Our experimental data , most notably , the increase in curvature and monotonical decrease in edge length are incompatible with these predictions , thus ruling out the constant curvature model ( Fig . 3 , green lines ) . The constant area model , in a strict interpretation , on the other hand , predicts the coat to assemble completely on a flat membrane , after which its area stays constant and only curva - ture increases to form a spherical vesicle . These predictions agree reasonably well with the data for curvature propagation during CME but fail to explain the monotonic increase in coat surface over time . Thus , the continuous increase of surface area that we observed here rules out the constant area model as well ( Fig . 3 , blue lines ) . Interestingly , earlier work compatible with the constant area model already suggested the presence of shallow curvatures in even the flattest clathrin - coated struc - tures ( Avinoam et al . , 2015 ; Bucher et al . , 2018 ; Heuser , 1980 ) . We found that around half of the clathrin coat area has preassembled when plasma membrane invagination begins ( Fig . 2 D and Fig . 4 C ) , agreeing well with previous reports ( Bucher et al . , 2018 ; Scott et al . , 2018 ) , after which the coat keeps growing gradually . Based on this observation , we developed a mathematical model that for the first time considers that cur - vature is generated in a positive ( nonlinear ) feedback loop . Our cooperative curvature model assumes that net area growth of the clathrin coat occurs via triskelia addition to the many available binding sites at the coat edge . We proposed that this process depends on the number of free binding sites , which scale with the edge length and can be described by a single kinetic constant k on . Triskelia addition within the coat is still likely to occur at a lower rate , as clathrin lattices can be unsaturated and have defects ( Frey et al . , 2020 ; Sochacki et al . , 2021 ) , and tris - kelia can exchange ( Avinoam et al . , 2015 ) , which is energetically favorable for curvature changes within these lattices ( Frey and Schwarz , 2020 ; Sochacki et al . , 2021 ; Tagiltsev et al . , 2021 ) . While growth at the edge accounts for an increase in surface area , the curvature is most likely generated in the bulk of the coat . Here , we assumed a nonlinear relation between the rate of curvature increase and curvature , which reflects cooperativity in the lattice , e . g . , due to rearrangements of neighboring tris - kelia or larger regions thereof . This assumption of cooperativity is supported by recent experiments , which suggest that clathrin exhibits curvature - sensing properties , preferentially assembling Mund et al . Journal of Cell Biology 9 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 at prebent membranes ( Sochacki et al . , 2021 ; Zeno et al . , 2021 ) . The model described above shows an excellent fit between theory and experiment and further predicts a square root de - pendence of \u03b8 over time . This describes curvature generation during clathrin coat maturation as a nonlinear mechanism driven by a positive feedback of multiple triskelia , slowing down once the coat approaches a preferred degree of curvature . Even though our model agrees very well with the data , it still exhibits certain limitations worth mentioning : First , it only considers the coat itself and ignores the plethora of other proteins within the endocytic machinery . These include numerous BAR domain proteins that are dynamically recruited and disassembled during endocytosis and thus are bound to influence membrane curvature at endocytic sites , as well as a variety of clathrin adaptor proteins , whose presence or absence could explain the cell - type specific differences in average vesicle sizes that we observed ( Fig . S4 ) . Taken together , these factors could explain the imperfection of our model in the very begin - ning and the final part of the timeline ( Fig . 4 A ) , where vesicle scission is driven by the fast - acting mechanoenzyme dynamin . Additionally , we only modeled clathrin recruitment and ignored clathrin disassembly , which could be mediated by adaptor un - binding ( Taylor et al . , 2011 ) and uncoating factors , including auxilin ( He et al . , 2020 ) , that are recruited before the end of vesicle scission . We also assumed that the clathrin coat has constant properties , most notably that the intrinsic coat - driven curvature generation toward its preferred curvature occurs unidirectionally and remains the same throughout endocytosis ( Eq . 3 ) . It is however likely that the properties of the clathrin coat change during endocytosis , e . g . , by coat stiffening or in - creasingly tight interactions between triskelia ( Frey and Schwarz , 2020 ; Frey et al . , 2020 ; Sochacki et al . , 2021 ) . Second , we reconstructed an average trajectory of clathrin coat remodeling generated from many individual static snapshots , thereby averaging conceivable different pathways of CME . How - ever , different pathways with substantially changed relationships between the parameters like curvature and \u03b8 would be visible in the corresponding plots as separate point clouds . This is not the case , rather weobserved acontinuous correlationbetweencurvatureand \u03b8 following a single trajectory , indicating that CME follows a single , stereotypic pathway . We did identify a small , disconnected popu - lation of sites in our data set that most likely originate from a dis - tinct cellular mechanism ( Fig . S3 ) . While this indicates that our approach should capture potentially different pathways of clathrin - coated vesicle formation , we cannot exclude minor mechanistic variations that are included in the average trajectory . Clathrin recruitment has been quantified extensively using live - cell microscopy , resolving the sequential recruitment and dynamic characteristics of many important endocytic compo - nents ( Aguet et al . , 2013 ; Cocucci et al . , 2012 ; Doyon et al . , 2011 ; He et al . , 2017 ; Jin et al . , 2021 Preprint ; Loerke et al . , 2009 ; Mettlen et al . , 2009 ; Saffarian et al . , 2009 ; Sch\u00a8oneberg et al . , 2018 ; Taylor et al . , 2011 ) . We wondered if it is possible to cor - relate our pseudotime reconstruction with previously reported real - time dynamics of clathrin , which shows an initial fluores - cence intensity increase , a plateau , and finally a sharp intensity decrease after vesicle scission . While we observed a correlation between the number of clathrin localizations and surface area ( Fig . S5 , M \u2013 O ) , we note that indirect immunolabeling is not al - ways quantitative , which complicates a direct comparison of previously reported live - cell fluorescence intensity over time and our pseudotime trajectory data . Nevertheless , we speculate that the initial fast intensity increase in live - cell studies likely corresponds to the initial growth of the flat clathrin coat that escapes our detection due to the fast dynamics and small size . The slower fluorescence increase and subsequent plateauing then coincide with curvature generation and the final addition of triskelia at the coat edge , as resolved in detail in our pseu - dotime data . To better understand the correlation between changes in the nanoscale architecture of clathrin coats and dy - namic consequences , we would ultimately require a method that combines high structural with temporal resolution . In a recent publication , this was attempted using a live - cell 2D super - resolution microscopy approach ( Willy et al . , 2021 ) . This study reported an increasing projected area of clathrin over time , suggesting that curvature is present in the earliest detectable clathrin structures , and concluded that CME follows the CCM . Although we also find that completely flat structures are rare and curvature is initiated before the final surface is acquired , our data is entirely incompatible with the CCM . This is espe - cially true in the first half of endocytosis with shallow but progressively increasing curvatures ( Fig . 3 ) , which is challeng - ing to measure using 2D imaging . This shows that it remains highly desirable to ultimately image the 3D nanoscale architec - ture of the clathrin coat in living cells in real - time . In summary , we characterized the dynamics and geometries of the clathrin coat in endocytosis by combining 3D super - resolution microscopy , quantitative analysis , and mathematical modeling . We found that clathrin bending and assembly occur simultaneously along a precisely defined trajectory . We anticipate that this work will be foundational to further study the structural mechanism of endocytosis , both under physiological conditions , and in diseases , where changes in CME likely occur very fre - quently and have just recently been shown to have profound functional consequences ( Moulay et al . , 2020 ; Xiao et al . , 2018 ) . Materials and methods Cell culture SK - MEL - 2 cells ( gift from David Drubin , UC Berkeley , described in Doyon et al . [ 2011 ] ) were cultured adherently in DMEM ( Gibco , no . 10565 - 018 ) , supplemented with 10 % ( v / v ) FBS ( Gibco , no . 10270 - 106 ) , and ZellShield ( Biochrom AG ) at 37\u00b0C , under an atmosphere with 5 % CO 2 and 100 % humidity . U2OS cells ( # 300174 ; Cell Line Services ) were cultured ad - herently as described previously ( Thevathasan et al . , 2019 ) in DMEM ( Gibco , no . 11880 - 028 ) supplemented with 10 % FBS ( Gibco , no . 10270 - 106 ) , 1x GlutaMAX ( Gibco , no . 35050 - 038 ) , nonessential amino acids ( Gibco , no . 11140 - 035 ) , and ZellShield ( Biochrom AG ) at 37\u00b0C under an atmosphere with 5 % CO 2 and 100 % humidity . 3T3 mouse fibroblasts ( gift from Alba Diz - Mu\u00f1oz , EMBL Heidelberg ) were cultured adherently in DMEM ( 4 . 5 g / l D - Glucose ) supplemented with 1\u00d7 MEM NEAA ( Gibco , no . 11140 - 035 ) , 1\u00d7 GlutaMAX ( Gibco , no . 35050 - 038 ) , and 10 % Mund et al . Journal of Cell Biology 10 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 ( v / v ) FBS ( Gibco , no . 10270 - 106 ) at 37\u00b0C under an atmos - phere with 5 % CO 2 , 100 % humidity . Sample preparation for superresolution imaging of clathrin - coated pits Cells were seeded onto high - precision 24 - mm round glass cov - erslips ( No . 1 . 5H , catalog no . 117640 ; Marienfeld ) . Coverslips were previously cleaned by incubating them overnight in a methanol / hydrochlorid acid ( 50 : 50 ) solution while stirring . The coverslips were then repeatedly rinsed with water until a neu - tral pH was reached . They were then placed into a laminar flow cell culture hood overnight to dry . In a final cleaning step , the coverslips are irradiated with ultraviolet light for 30min . Cells were fixed as described previously ( Li et al . , 2018 ) using 3 % ( w / v ) formaldehyde , 10 mM MES pH 6 . 1 , 150 mM NaCl , 5 mM EGTA , 5 mM glucose , and 5 mM MgCl 2 for 20 min . Fix - ation was quenched in 0 . 1 % ( w / v ) NaBH 4 for 7 min . The sample was washed three times with PBS and permeabilized for 15 min with 0 . 01 % ( w / v ) digitonin ( Sigma - Aldrich ) in PBS . The sample was then washed twice with PBS and blocked for 1 h with 2 % ( w / v ) BSA / PBS , washed with PBS , and stained for 3 \u2013 12 h with anti - clathrin light chain ( sc - 28276 ; Santa Cruz Biotechnology ) and anti - clathrin heavy chain rabbit polyclonal antibodies ( ab21679 ; Abcam ) in 1 % ( w / v ) BSA / PBS . After three washes with PBS , the sample was incubated for 3 \u2013 4 h with a secondary donkey anti - rabbit antibody ( 711 - 005 - 152 ; Jackson ImmunoResearch ) that was conjugated to Alexa Fluor 647 \u2013 NHS at an average degree of labeling of 1 . 5 . The sample was then washed three times and mounted for imaging in blinking buffer ( 50 mM Tris / HCl pH 8 , 10 mM NaCl , 10 % ( w / v ) D - glucose , 500 \u00b5g ml \u2212 1 glucose oxidase , 40 \u00b5g ml \u2212 1 glucose catalase , and 35 mM MEA in H 2 O ) . For the analysis of the disconnected population of sites , 3T3 cells were transfected with a plasmid encoding the sigma 2 sub - unit , fused to GFP ( gift from Steeve Boulant , University of Florida , # 53610 ; Addgene ) , to obtain cells transiently expressing AP2 - GFP . The transfection was performed using a Lipofectamine 2 , 000 re - agent ( Life Technologies ) according to the manufacturer \u2019 s rec - ommendations : 1 \u00b5g DNA was mixed with 50 \u00b5l OptiMEM I ( Thermo Fisher Scientific ) . The same was done for 3 \u00b5l Lipo - fectamin in 50 \u00b5l OptiMEM I . Both solutions were incubated for 5 min and then mixed together and incubated for an additional 10 min at room temperature . The media of previously seeded cells was exchanged to prewarmed OptiMEM I , to which the DNA - Lipofectamin solution ( 100 \u00b5l ) was added dropwise . After \u223c 24 h of incubation ( at 5 % CO 2 , 37\u00b0C ) , the medium was exchanged with fresh growth medium . After additional incubation for \u223c 16 h , cells were fixed according to the same protocol described above . Superresolution microscopy SMLM images were acquired at room temperature ( 24\u00b0C ) on a custom - built microscope ( Mund et al . , 2018 ) with a 160\u00d7 NA 1 . 43 oil - immersion objective ( Leica ) . The sample was illumi - nated using a LightHub laser combiner ( Omicron - Laserage La - serprodukte ) with Luxx 405 , 488 , and 638 and Cobolt 561 lasers , which were triggered using a Mojo FPGA ( Embedded Micro ) for microsecond pulsing control of lasers . The lasers were guided through a speckle reducer ( LSR - 3005 - 17S - VIS ; Optotune , Dietikon , Switzerland ) and coupled into a multimode fiber ( M105L02S - A ; Thorlabs ) . The output of the fiber was magni - fied and imaged in the sample . Fiber - generated fluorescence was removed using a laser clean - up filter ( 390 / 482 / 563 / 640 HC Quad ; AHF ) . The focus was stabilized using a closed - loop system based on reflecting the signal of a near - infrared laser on the coverslip and detecting the resulting signal on a quadrant photodiode , re - sulting in focus stabilities of \u00b1 10 nm over several hours . Fluo - rescence emission was filtered using a 676 / 37 or a 700 / 100 bandpass filter ( AHF ) and recorded by an EMCCD camera ( Evolve512D ; Photometrics ) . Typically , 100 , 000 \u2013 300 , 000 frames were acquired using 15 - or 30 - ms exposure times and laser power densities of \u223c 15 kW / cm 2 . 405 - nm laser intensity was adjusted automatically by changing pulse duration to keep the number of localizations per frame constant during the acquisition . For the analysis of the disconnected population of sites , one diffraction - limited image was additionally acquired before SMLM imaging . For this , a 488 laser at 1 . 4 kW / cm 2 was used to take a single frame at 30 ms exposure time . Emission was fil - tered using a 525 / 50 bandpass filter ( AHF ) . The microscope hardware and data acquisition was handled via Micro - Manager 1 . 4 . 22 using custom - written software ( Edelstein et al . , 2010 , 2014 ; Deschamps and Ries , 2020 ) . Data analysis All data analysis was conducted in SMAP ( [ Ries , 2020 ] based on MATLAB and available as open source under https : / / github . com / jries / SMAP ) . Superresolution image reconstruction For fitting the localizations , peaks were detected as maxima in raw camera images after applying a difference - of - Gaussian fil - ter . At those positions , cropped images of 13 \u00d7 13 pixels were fitted with an experimentally derived PSF model ( free fitting parameters : x , y , z , photons per localization , background per pixel ) using an MLE ( Maximum likelihood estimation ) fitter ( Li et al . , 2018 ) . The x , y , and z positions were corrected for residual drift by a custom algorithm based on redundant cross - correlation . In short , the data were distributed into 10 - time bins of equal length . For each bin , a superresolution image was reconstructed . We then calculated the image cross - correlations among all superresolution images and extracted the relative displacements in x and y from the position of the maximum in the cross - correlation images . We then calculated the drift tra - jectory that best describes the relative displacements . In a second step , the z - drift was measured in an analogous way using in - tensity profiles in z instead of images . Localizations persistent over consecutive frames ( detected within 35 nm from one an - other and with a maximum gap of one dark frame ) were merged into one localization by calculating the weighted average of x , y , and z positions and the sums of photons per localization and background . Localizations were filtered by the localization pre - cision in x , y ( 0 \u2013 20 nm ) and z ( 0 \u2013 30 nm ) to exclude dim local - izations . The modal value for the localization precision \u03c3 was 3 . 9 nm in x / y and 12 . 5 nm in z , leading to a resolution estimate ( calculated using the Full Width Half Maximum ( FWHM ) using FWHM = 2 \u221a ( 2 ln 2 ) \u03c3 ) of 9 . 2 nm in x / y and 29 . 4 nm in z ( typical Mund et al . Journal of Cell Biology 11 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 values based on the representative image ) . Superresolution images were constructed with every localization rendered as a two - dimensional spherical Gaussian with a sigma of 3 nm . The red hot color map used represents the density of localizations and is scaled in a way that 0 . 03 % of the pixel values are saturated . Quantitative geometric analysis of clathrin - coated structures Clathrin - coated structures were segmented semiautomatically . First , we manually defined a region of interest excluding the edges of the cell . Then , the image was blurred using a Gaussian filter with a sigma of 100 nm and peaks were detected using a manually set threshold . This typically yielded several hundreds of sites in a region of 30 \u00d7 30 \u00b5m . These candidate sites were curated manually , and only single , well - isolated clathrin - coated structures were retained in the dataset . Next , these structures were analyzed using LocMoFit , an MLE - based model fitting framework that we developed recently ( Wu et al . , 2023 ) . LocMoFit directly fits localization coordinates with the probability density function ( PDF ) of a parametrized geometric model . In this study , we modeled clathrin - coated structures with a hollow spherical cap parameterized by the surface area A and the closing angle \u03b8 . \u03b8 is defined as the angle between the two vectors that point to the pole and the edge , respectively , from the center of the sphere . The position of the model is defined as the center of mass of the cap . In practice , we discretized the cap by placing spiral points ( Saff and Kuijlaars , 1997 ) or the spherical Fibonacci lattice on the surface of the cap to approximate an even distribution . In LocMoFit , these points were treated as discrete fluorophore coordinates when con - structing the PDF of the model . During fitting , additional pa - rameters including the center position and the orientation of the model were determined with respect to the fluorophore coor - dinates , and an extra uncertainty and the background weight were applied to the PDF . After fitting , the sphere radius is de - rived as r (cid:2) \ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03 A / 2 \u03c0 ( 1 \u2212 cos \u03b8 ) p , projected area as A p = \u03c0 sin 2 \u03b8 , and edge length as \u03b5 = 2 \u03c0 sin \u03b8 . For some flat sites where the fit pro - duced slightly negative curvature values , curvature H and \u03b8 were set to 0 nm \u2212 1 and 0 , \u00b0 respectively , to approximate them as completely flat . After model fitting , a second curation step was performed . With this , we ensured that only well - fitted sites are included in the final data set . Sites were excluded if they were upside down ( clearly originating from an upper membrane ) , double - sites with two sites clearly connected to each other or an adjacent flat structure , and large plaques or structures nondistinguishable from an antibody cluster . Pseudo - temporal reconstruction of clathrin remodeling during endocytosis To sort endocytic sites in pseudotime , they were sorted by the closing angle \u03b8 , assigning each site a rank index . Flat sites with a manually assigned H = 0 nm \u2212 1 and \u03b8 = 0 were all assigned an index of 0 . As pseudo - temporal sorting assumes that all sites are part of the same endocytic trajectory , endocytic sites with cur - vatures above a cell line \u2014 specific threshold ( H > 0 . 016 nm \u2212 1 for SK - MEL - 2 ; H > 0 . 013 nm \u2212 1 for U2OS ; H > 0 . 014 nm \u2212 1 for 3T3 ) that form a visibly disconnected point cloud ( Figs . S3 and S4 ) were excluded for this analysis ( Fig . S5 ) . To compute pseudo - temporal averages , the sites were spatially and rotationally aligned and rescaled by the average radius of all sites within the respective bin . As the geometric model is rotationally symmet - ric , we then performed rotational averaging by generating 72 duplicates of each structure , rotating them by 5\u00b0 with respect to each other , and averaging them . Video 1 was computed using a sliding window approach , where each frame shows an average of 30 sites , and the frame - to - frame increment is 20 sites . The median pseudotime of those 30 sites is indicated . Further data analysis All data that resulted from the quantitative geometric descrip - tion of clathrin - coated structures were further analyzed in R ( R Core Team , 2020 ) . Fitting of the growth models ( Tables 1 , 2 , and 3 ) was performed according to the equations described in the Appendix using a nonlinear least square fit . All analyses were then performed on filtered data sets , excluding disconnected sites above a cell line \u2014 specific threshold ( H > 0 . 016 nm \u2212 1 for SK - MEL - 2 ; H > 0 . 013 nm \u2212 1 for U2OS ; H > 0 . 014 nm \u2212 1 for 3T3 ) , and sites of negative curvature . During fitting , for sites with \u03b8 = 0\u00b0 , we set \u03b8 = 0 . 0001\u00b0 to avoid division by 0 . For depicting the growth models in Figs . 3 and 4 , as well as Fig . S5 , parameters resulting from the H ( \u03b8 ) fit were used and mapped to the A , \u0190 , and A p data . Simulations Simulations were performed using the simulation engine for SMLM data implemented in SMAP and LocMoFit as described in Thevathasan et al . ( 2019 ) . The realistic simulations were based on a two - state ( bright and dark ) fluorophore model plus bleaching ( Sage et al . , 2019 ) , and parameters ( number of pho - tons , background photons , and fluorophore on - time t l ) were extracted from our experiment . ( 1 ) First , we defined an equally distributed closing angle \u03b8 from 0 to 180\u00b0 and calculated the surface area A ( \u03b8 ) = 2 \u03c0 ( 1 \u2212 cos \u03b8 ) / H ( \u03b8 ) 2 ( see Appendix for details ) . Here , H ( \u03b8 ) is defined as in Eq . 3 , with fitting parameters de - termined in SK - MEL - 2 ( see Table 1 ) . ( 2 ) With the defined model parameters , we generated protein positions for each simulated structure by taking randomly drawn N } A samples from the PDF of the hollow spherical cap with no uncertainty . ( 3 ) With a probability p label = 0 . 6 , a fluorescent label was created at a pro - tein position . ( 4 ) Linkage displacements in x , y , and z were added to a label and were determined as normally distributed random variables with a variance corresponding to the linkage error of 5 nm . The fluorophore is assumed to be freely rotating three - dimensionally between different blinks . ( 5 ) Each fluo - rophore appeared at a random time and lived for a time t l , de - termined as a random variable from an exponential distribution with a mean of 1 . 6 frames . ( 6 ) A label had a probability p react = 0 . 5 to be reactivated and then appeared at a random later time point , otherwise it was bleached . ( 7 ) When it was on , a fluorophore had a constant brightness . Thus , the brightness in each frame was proportional to the fraction of the time the fluorophore was on in each frame . ( 8 ) The emitted photons in each frame were determined as a random Poisson variable with a mean value corresponding to the average brightness in the frame . ( 9 ) For Mund et al . Journal of Cell Biology 12 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 each frame , we calculated the CRLB ( Cram\u00b4er - Rao lower bound ) in x , y , and z from the number of photons ( with a mean of 11 , 000 ) and the background photons ( 130 per pixel ) based on the theoretical Gaussian PSF ( Mortensen et al . , 2010 ) or a 3D cspline PSF model derived from beads calibrations ( Li et al . , 2018 ) . This error was added to the true x , y , and z positions of the fluo - rophores as normally distributed random values with a variance corresponding to the respective calculated CRLB . Online supplemental material Fig . S1 provides examples of diverse clathrin coat structures . Fig . S2 provides simulations . Fig . S3 shows non - endocytic clathrin structures . Fig . S4 shows clathrin coat remodeling in three dif - ferent cell lines . Fig . S5 provides linkage error investigation and number of localizations per clathrin coat . Video 1 shows a pseudo - temporal movie of clathrin coat remodeling during en - docytosis in SK - MEL - 2 cells . Appendix shows a detailed de - scription of the cooperative curvature model . All presented data are available in the BioStudies database ( https : / / www . ebi . ac . uk / biostudies / ) under accession number S - BIAD566 . Acknowledgments We thank the entire Ries and Kaksonen labs for fruitful dis - cussions and support . This work was supported by the European Research Council ( ERC CoG - 724489 to J . Ries ) , the National Institutes of Health Common Fund 4D Nucleome Program ( Grant U01 to J . Ries ) , the Human Frontier Science Program ( RGY0065 / 2017 to J . Ries ) , the EMBL Interdisciplinary Postdoc Programme ( EIPOD ) under Marie Curie Actions COFUND ( Grant 229597 to O . Avinoam ) , the European Molecular Biology Laboratory ( M . Mund , A . Tschanz , Y . - L . Wu and J . Ries ) , and the Swiss National Sci - ence Foundation ( grant 310030B _ 182825 and NCCR Chemical Biology to M . Kaksonen ) . O . Avinoam is an incumbent of the Miriam Berman Presidential Development Chair . Author contributions : M . Mund , O . Avinoam , M . Kaksonen and J . Ries conceived the study , M . Mund , A . Tschanz , J . L . Mehl and O . Avinoam performed experiments , M . Mund , A . Tschanz , Y . - L . Wu , J . L . Mehl , O . Avinoam and J . Ries analyzed superresolution data ; F . Frey and U . S . Schwarz developed the cooperative curvature model ; J . Ries supervised the study . M . Mund , A . Tschanz , F . Frey , J . Ries wrote the manuscript with input from all authors . Disclosure : The authors declare no competing financial interests . Submitted : 10 June 2022 Revised : 29 November 2022 Accepted : 27 December 2022 References Aguet , F . , C . N . Antonescu , M . Mettlen , S . L . Schmid , and G . Danuser . 2013 . Advances in analysis of low signal - to - noise images link dynamin and AP2 to the functions of an endocytic checkpoint . Dev . Cell . 26 : 279 \u2013 291 . https : / / doi . org / 10 . 1016 / j . devcel . 2013 . 06 . 019 Avinoam , O . , M . Schorb , C . J . Beese , J . A . G . Briggs , and M . Kaksonen . 2015 . Endocytic sites mature by continuous bending and remodeling of the clathrin coat . Science . 348 : 1369 \u2013 1372 . https : / / doi . org / 10 . 1126 / science . aaa9555 Bucher , D . , F . Frey , K . A . Sochacki , S . Kummer , J . - P . Bergeest , W . J . Godinez , H . - G . Kr\u00a8ausslich , K . Rohr , J . W . Taraska , U . S . Schwarz , and S . Boulant . 2018 . Clathrin - adaptorratio and membrane tensionregulatethe flat - to - curved transition of the clathrin coat during endocytosis . Nat . Commun . 9 : 1109 . https : / / doi . org / 10 . 1038 / s41467 - 018 - 03533 - 0 Chen , Z . , andS . L . Schmid . 2020 . Evolving models for assembling and shaping clathrin - coated pits . J . Cell Biol . 219 : e202005126 . https : / / doi . org / 10 . 1083 / jcb . 202005126 Cheng , Y . , W . Boll , T . Kirchhausen , S . C . Harrison , and T . Walz . 2007 . Cryo - electron tomography of clathrin - coated vesicles : Structural im - plicationsforcoat assembly . J . Mol . Biol . 365 : 892 \u2013 899 . https : / / doi . org / 10 . 1016 / j . jmb . 2006 . 10 . 036 Cocucci , E . , F . Aguet , S . Boulant , and T . Kirchhausen . 2012 . The first five seconds in the life of a clathrin - coated pit . Cell . 150 : 495 \u2013 507 . https : / / doi . org / 10 . 1016 / j . cell . 2012 . 05 . 047 Dannhauser , P . N . , and E . J . Ungewickell . 2012 . Reconstitution of clathrin - coated bud and vesicle formation with minimal components . Nat . Cell Biol . 14 : 634 \u2013 639 . https : / / doi . org / 10 . 1038 / ncb2478 Deschamps , J . , and J . Ries . 2020 . EMU : Reconfigurable graphical user inter - faces for micro - manager . BMC Bioinf . 21 : 456 . https : / / doi . org / 10 . 1186 / s12859 - 020 - 03727 - 8 Doyon , J . B . , B . Zeitler , J . Cheng , A . T . Cheng , J . M . Cherone , Y . Santiago , A . H . Lee , T . D . Vo , Y . Doyon , J . C . Miller , et al . 2011 . Rapid and efficient clathrin - mediated endocytosis revealed in genome - edited mammalian cells . Nat . Cell Biol . 13 : 331 \u2013 337 . https : / / doi . org / 10 . 1038 / ncb2175 Edelstein , A . D . , M . A . Tsuchida , N . Amodaj , H . Pinkard , R . D . Vale , and N . Stuurman . 2014 . Advanced methods of microscope control using \u03bc Manager software . J . Biol . Methods . 1 . https : / / doi . org / 10 . 14440 / jbm . 2014 . 36 Edelstein , A . , N . Amodaj , K . Hoover , R . Vale , and N . Stuurman . 2010 . Com - puter control of microscopes using \u00b5Manager . Curr . Protoc . Mol . Biol . Chapter 14 : Unit 14 . 20 . https : / / doi . org / 10 . 1002 / 0471142727 . mb1420s92 Fotin , A . , Y . Cheng , P . Sliz , N . Grigorieff , S . C . Harrison , T . Kirchhausen , and T . Walz . 2004 . Molecular model for a complete clathrin lattice from electron cryomicroscopy . Nature . 432 : 573 \u2013 579 . https : / / doi . org / 10 . 1038 / nature03079 Frey , F . , andU . S . Schwarz . 2020 . Competingpathwaysforthe invaginationof clathrin - coated membranes . Soft Matter . 16 : 10723 \u2013 10733 . https : / / doi . org / 10 . 1039 / D0SM01375G Frey , F . , D . Bucher , K . A . Sochacki , J . W . Taraska , S . Boulant , and U . S . Schwarz . 2020 . Eden growth models for flat clathrin lattices with vacancies . New J . Phys . 22 : 073043 . https : / / doi . org / 10 . 1088 / 1367 - 2630 / ab99e1 Fr\u00fch , S . M . , U . Matti , P . R . Spycher , M . Rubini , S . Lickert , T . Schlichthaerle , R . Jungmann , V . Vogel , J . Ries , and I . Schoen . 2021 . Site - specifically - la - beled antibodies for super - resolution microscopy reveal In Situ linkage errors . ACS Nano . 15 : 12161 \u2013 12170 . https : / / doi . org / 10 . 1021 / acsnano . 1c03677 He , K . , R . Marsland III , S . Upadhyayula , E . Song , S . Dang , B . R . Capraro , W . Wang , W . Skillern , R . Gaudin , M . Ma , and T . Kirchhausen . 2017 . Dy - namics of phosphoinositide conversion in clathrin - mediated endocytic traffic . Nature . 552 : 410 \u2013 414 . https : / / doi . org / 10 . 1038 / nature25146 He , K . , E . Song , S . Upadhyayula , S . Dang , R . Gaudin , W . Skillern , K . Bu , B . R . Capraro , I . Rapoport , I . Kusters , et al . 2020 . Dynamics of Auxilin 1 and GAK in clathrin - mediated traffic . J . Cell Biol . 219 : e201908142 . https : / / doi . org / 10 . 1083 / jcb . 201908142 Henne , W . M . , E . Boucrot , M . Meinecke , E . Evergren , Y . Vallis , R . Mittal , and H . T . McMahon . 2010 . FCHo proteins are nucleators of clathrin - mediated endocytosis . Science . 328 : 1281 \u2013 1284 . https : / / doi . org / 10 . 1126 / science . 1188462 Heuser , J . 1980 . Three - dimensional visualization of coated vesicle formation in fibroblasts . J . Cell Biol . 84 : 560 \u2013 583 . https : / / doi . org / 10 . 1083 / jcb . 84 . 3 . 560 Heuser , J . , and T . Kirchhausen . 1985 . Deep - etch views of clathrin assemblies . J . Ultrastruct . Res . 92 : 1 \u2013 27 . https : / / doi . org / 10 . 1016 / 0889 - 1605 ( 85 ) 90123 - 5 Jin , M . , C . Shirazinejad , B . Wang , A . Yan , J . Sch\u00a8oneberg , S . Upadhyayula , K . Xu , and D . G . Drubin . 2021 . Asymmetric Arp2 / 3 - mediated actin as - sembly facilitates clathrin - mediated endocytosis at stalled sites in genome - edited human stem cells . bioRxiv . ( Preprint posted July 16 , 2021 ) . https : / / doi . org / 10 . 1101 / 2021 . 07 . 16 . 452693 Kaksonen , M . , and A . Roux . 2018 . Mechanisms of clathrin - mediated endo - cytosis . Nat . Rev . Mol . Cell Biol . 19 : 313 \u2013 326 . https : / / doi . org / 10 . 1038 / nrm . 2017 . 132 Mund et al . Journal of Cell Biology 13 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 Kanaseki , T . , and K . Kadota . 1969 . The \u201c vesicle in a basket \u201d . A morphological study of the coated vesicle isolated from the nerve endings of the Guinea pig brain , with special reference to the mechanism of mem - brane movements . J . Cell Biol . 42 : 202 \u2013 220 . https : / / doi . org / 10 . 1083 / jcb . 42 . 1 . 202 Kaplan , C . , S . J . Kenny , X . Chen , J . Sch\u00a8oneberg , E . Sitarska , A . Diz - Mu\u00f1oz , M . Akamatsu , K . Xu , and D . G . Drubin . 2022 . Load adaptation by endocytic actin networks . Mol Biol Cell . 33 : ar50 . https : / / doi . org / 10 . 1091 / mbc . E21 - 11 - 0589 Kirchhausen , T . 1993 . Coated pits and coated vesicles \u2014 sorting it all out . Curr . Opin . Struct . Biol . 3 : 182 \u2013 188 . https : / / doi . org / 10 . 1016 / S0959 - 440X ( 05 ) 80150 - 2 Li , Y . , M . Mund , P . Hoess , J . Deschamps , U . Matti , B . Nijmeijer , V . J . Sabinina , J . Ellenberg , I . Schoen , and J . Ries . 2018 . Real - time 3D single - molecule localization using experimental point spread functions . Nat . Methods . 15 : 367 \u2013 369 . https : / / doi . org / 10 . 1038 / nmeth . 4661 Loerke , D . , M . Mettlen , D . Yarar , K . Jaqaman , H . Jaqaman , G . Danuser , and S . L . Schmid . 2009 . Cargo and dynamin regulate clathrin - coated pit matu - ration . PLoS Biol . 7 : e57 . https : / / doi . org / 10 . 1371 / journal . pbio . 1000057 Mettlen , M . , M . Stoeber , D . Loerke , C . N . Antonescu , G . Danuser , and S . L . Schmid . 2009 . Endocytic accessory proteins are functionally distin - guished by their differential effects on the maturation of clathrin - coated pits . Mol . Biol . Cell . 20 : 3251 \u2013 3260 . https : / / doi . org / 10 . 1091 / mbc . e09 - 03 - 0256 Morris , K . L . , J . R . Jones , M . Halebian , S . Wu , M . Baker , J . - P . Armache , A . Avila Ibarra , R . B . Sessions , A . D . Cameron , Y . Cheng , and C . J . Smith . 2019 . Cryo - EM of multiple cage architectures reveals a universal mode of clathrin self - assembly . Nat . Struct . Mol . Biol . 26 : 890 \u2013 898 . https : / / doi . org / 10 . 1038 / s41594 - 019 - 0292 - 0 Mortensen , K . I . , L . S . Churchman , J . A . Spudich , and H . Flyvbjerg . 2010 . Op - timized localization analysis for single - molecule tracking and super - resolution microscopy . Nat . Methods . 7 : 377 \u2013 381 . https : / / doi . org / 10 . 1038 / nmeth . 1447 Moulay , G . , J . Lain\u00b4e , M . Lema \u02c6 \u0131 tre , M . Nakamori , I . Nishino , G . Caillol , K . Mamchaoui , L . Julien , F . Dingli , D . Loew , et al . 2020 . Alternative splicing of clathrin heavy chain contributes to the switch from coated pits to plaques . J . Cell Biol . 219 : e201912061 . https : / / doi . org / 10 . 1083 / jcb . 201912061 Mund , M . , and J . Ries . 2020 . How good are my data ? Reference standards in superresolution microscopy . Mol . Biol . Cell . 31 : 2093 \u2013 2096 . https : / / doi . org / 10 . 1091 / mbc . E19 - 04 - 0189 Mund , M . , J . A . van der Beek , J . Deschamps , S . Dmitrieff , P . Hoess , J . L . Monster , A . Picco , F . N\u00b4ed\u00b4elec , M . Kaksonen , and J . Ries . 2018 . Sys - tematic nanoscale analysis of endocytosis links efficient vesicle for - mation to patterned actin nucleation . Cell . 174 : 884 \u2013 896 . e17 . https : / / doi . org / 10 . 1016 / j . cell . 2018 . 06 . 032 Pearse , B . M . 1976 . Clathrin : A unique protein associated with intracellular transfer of membrane by coated vesicles . Proc . Natl . Acad . Sci . USA . 73 : 1255 \u2013 1259 . https : / / doi . org / 10 . 1073 / pnas . 73 . 4 . 1255 R Core Team . 2020 . R : A Language and Environment for Statistical Com - puting . R Foundation for Statistical Computing , Vienna , Austria Ries , J . 2020 . SMAP : A modular super - resolution microscopy analysis plat - form for SMLM data . Nat . Methods . 17 : 870 \u2013 872 . https : / / doi . org / 10 . 1038 / s41592 - 020 - 0938 - 1 Saff , E . B . , and A . B . J . Kuijlaars . 1997 . Distributing many points on a sphere . Math . Intell . 19 : 5 \u2013 11 . https : / / doi . org / 10 . 1007 / BF03024331 Saffarian , S . , E . Cocucci , and T . Kirchhausen . 2009 . Distinct dynamics of endocytic clathrin - coated pits andcoated plaques . PLoS Biol . 7 : e1000191 . https : / / doi . org / 10 . 1371 / journal . pbio . 1000191 Sage , D . , T . - A . Pham , H . Babcock , T . Lukes , T . Pengo , J . Chao , R . Velmurugan , A . Herbert , A . Agrawal , S . Colabrese , et al . 2019 . Super - resolution fight club : Assessment of 2D and 3D single - molecule localization microscopy software . Nat . Methods . 16 : 387 \u2013 395 . https : / / doi . org / 10 . 1038 / s41592 - 019 - 0364 - 4 Sch\u00a8oneberg , J . , D . Dambournet , T . - L . Liu , R . Forster , D . Hockemeyer , E . Betzig , and D . G . Drubin . 2018 . 4D cell biology : Big data image analytics and lattice light - sheet imaging reveal dynamics of clathrin - mediated endocytosis in stem cell derived intestinal organoids . Mol . Biol . Cell . 29 : 2959 \u2013 2968 . https : / / doi . org / 10 . 1091 / mbc . E18 - 06 - 0375 Scott , B . L . , K . A . Sochacki , S . T . Low - Nam , E . M . Bailey , Q . Luu , A . Hor , A . M . Dickey , S . Smith , J . G . Kerkvliet , J . W . Taraska , and A . D . Hoppe . 2018 . Membrane bending occurs at all stages of clathrin - coat assembly and defines endocytic dynamics . Nat . Commun . 9 : 419 . https : / / doi . org / 10 . 1038 / s41467 - 018 - 02818 - 8 Smith , C . J . , N . Grigorieff , and B . M . F . Pearse . 1998 . Clathrin coats at 21 A resolution : A cellular assembly designed to recycle multiple membrane receptors . EMBO J . 17 : 4943 \u2013 4953 . https : / / doi . org / 10 . 1093 / emboj / 17 . 17 . 4943 Sochacki , K . A . , and J . W . Taraska . 2018 . From flat to curved clathrin : Con - trolling a plastic ratchet . Trends Cell Biol . 29 : 241 \u2013 256 . https : / / doi . org / 10 . 1016 / j . tcb . 2018 . 12 . 002 Sochacki , K . A . , A . M . Dickey , M . - P . Strub , and J . W . Taraska . 2017 . Endocytic proteins arepartitionedat the edge of the clathrinlattice in mammalian cells . Nat . Cell Biol . 19 : 352 \u2013 361 . https : / / doi . org / 10 . 1038 / ncb3498 Sochacki , K . A . , B . L . Heine , G . J . Haber , J . R . Jimah , B . Prasai , M . A . Alfonzo - M\u00b4endez , A . D . Roberts , A . Somasundaram , J . E . Hinshaw , and J . W . Tar - aska . 2021 . The structure and spontaneous curvature of clathrin lattices at the plasma membrane . Dev . Cell . 56 : 1131 \u2013 1146 . e3 . https : / / doi . org / 10 . 1016 / j . devcel . 2021 . 03 . 017 Tagiltsev , G . , C . A . Haselwandter , and S . Scheuring . 2021 . Nanodissected elastically loaded clathrin lattices relax to increased curvature . Sci . Adv . 7 : eabg9934 . https : / / doi . org / 10 . 1126 / sciadv . abg9934 Takei , K . , V . Haucke , V . Slepnev , K . Farsad , M . Salazar , H . Chen , and P . De Camilli . 1998 . Generation of coated intermediates of clathrin - mediated endocytosis on protein - free liposomes . Cell . 94 : 131 \u2013 141 . https : / / doi . org / 10 . 1016 / S0092 - 8674 ( 00 ) 81228 - 3 Taylor , M . J . , D . Perrais , and C . J . Merrifield . 2011 . A high precision survey of the molecular dynamics of mammalian clathrin - mediated endocytosis . PLoS Biol . 9 : e1000604 . https : / / doi . org / 10 . 1371 / journal . pbio . 1000604 Thevathasan , J . V . , M . Kahnwald , K . Cie\u00b4sli\u00b4nski , P . Hoess , S . K . Peneti , M . Re - itberger , D . Heid , K . C . Kasuba , S . J . Hoerner , Y . Li , et al . 2019 . Nuclear pores as versatile reference standards for quantitative superresolution microscopy . Nat . Methods . 16 : 1045 \u2013 1053 . https : / / doi . org / 10 . 1038 / s41592 - 019 - 0574 - 9 Ungewickell , E . , and D . Branton . 1981 . Assembly units of clathrin coats . Na - ture . 289 : 420 \u2013 422 . https : / / doi . org / 10 . 1038 / 289420a0 Wang , X . , Z . Chen , M . Mettlen , J . Noh , S . L . Schmid , and G . Danuser . 2020 . DASC , a sensitive classifier for measuring discrete early stages in clathrin - mediated endocytosis . Elife . 9 : e53686 . https : / / doi . org / 10 . 7554 / eLife . 53686 Willy , N . M . , J . P . Ferguson , A . Akatay , S . Huber , U . Djakbarova , S . Silahli , C . Cakez , F . Hasan , H . C . Chang , A . Travesset , etal . 2021 . Denovoendocytic clathrin coats develop curvature at early stages of their formation . Dev . Cell . 56 : 3146 \u2013 3159 . e5 . https : / / doi . org / 10 . 1016 / j . devcel . 2021 . 10 . 019 Wu , Y . - L . , A . Tschanz , L . Krupnik , and J . Ries . 2020 . Quantitative data analysis in single - molecule localization microscopy . Trends Cell Biol . 30 : 837 \u2013 851 . https : / / doi . org / 10 . 1016 / j . tcb . 2020 . 07 . 005 Wu , Y . - L . , P . Hoess , A . Tschanz , U . Matti , M . Mund , and J . Ries . 2023 . Maximum - likelihood model fitting for quantitative analysis of SMLM data . Nat . Methods . 20 : 139 \u2013 148 . https : / / doi . org / 10 . 1038 / s41592 - 022 - 01676 - z Xiao , G . - Y . , A . Mohanakrishnan , and S . L . Schmid . 2018 . Role for ERK1 / 2 - dependent activation of FCHSD2 in cancer cell - selective regulation of clathrin - mediated endocytosis . Proc . Natl . Acad . Sci . USA . 115 : E9570 \u2013 E9579 . https : / / doi . org / 10 . 1073 / pnas . 1810209115 Yoshida , A . , N . Sakai , Y . Uekusa , Y . Imaoka , Y . Itagaki , Y . Suzuki , and S . H . Yoshimura . 2018 . Morphological changes of plasma membrane and protein assembly during clathrin - mediated endocytosis . PLoS Biol . 16 : e2004786 . https : / / doi . org / 10 . 1371 / journal . pbio . 2004786 Zeno , W . F . , J . B . Hochfelder , A . S . Thatte , L . Wang , A . K . Gadok , C . C . Hayden , E . M . Lafer , and J . C . Stachowiak . 2021 . Clathrin senses membrane cur - vature . Biophys . J . 120 : 818 \u2013 828 . https : / / doi . org / 10 . 1016 / j . bpj . 2020 . 12 . 035 Zhao , W . , L . Hanson , H . - Y . Lou , M . Akamatsu , P . D . Chowdary , F . Santoro , J . R . Marks , A . Grassart , D . G . Drubin , Y . Cui , and B . Cui . 2017 . Nanoscale manipulation of membrane curvature for probing endocytosis in live cells . Nat . Nanotechnol . 12 : 750 \u2013 756 . https : / / doi . org / 10 . 1038 / nnano . 2017 . 98 Mund et al . Journal of Cell Biology 14 of 14 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 Supplemental material Mund et al . Journal of Cell Biology S1 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 Figure S1 . Examples of diverse clathrin coat structures . ( A ) Large clusters of clathrin molecules excluded from further analysis . Shown in top view ( xy ) with the dotted line indicating a 50 nm - thick z - slice ( xz ) shown below . ( B ) Vesicular structures are sometimes fitted with a lower \u03b8 than expected . ( C ) While a spherical model describes the structure of most endocytic clathrin coats faithfully , there are few cases , as exemplified here , where the elliptical and irregular shape of an assembling coat is difficult to approximate with a simple geometric model . Two orthogonal 50 - nm - thick z - slices are shown here in xz and yz , and the respective spherical model fit is plotted as a dotted line . ( D ) Non - continuous labeling of clathrin manifests itself as holes in the coat , indicated with a blue arrow . All scale bars are 100 nm . Mund et al . Journal of Cell Biology S2 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 FigureS2 . Simulations . ( Aand B ) Estimation error of the closingangle \u03b8 ( N = 1 , 800 sites ) . ( A ) The comparison offitted \u03b8 of simulated clathrin coat structures to corresponding ground - truth \u03b8 shows no systematic bias and a narrow spread of the error across the ground truth although the spread increases when \u03b8 < 20 \u00b0 or \u03b8 > 160 \u00b0 . We reasoned that in the earlier range , where the structures are flat , the slightly increased error corresponds to the insensitivity of \u03b8 to flat structures . In the later range , where the vesicles are almost closed , the error was caused by indistinguishable tiny holes corresponding to the real vesicle openings and unlabeled clathrin in the coat . The fitted structures were simulated to have similar quality as the experimental data and to distribute evenly across \u03b8 . ( B ) The distribution of \u03b8 shows no significant error compared to the expectation corresponding to the evenly distributed \u03b8 , except for the small potential underestimation of entirely closed coats . ( C \u2013 G ) Averaging preserves U - shapes in simulated endocytic sites . ( C ) A ground - truth structure was used for simulating U - shaped clathrin coats . The model for simulation was built by combining a hemisphere with a radius of R = 97 nm and a cylinder with a height D = 0 . 5 R . We choose the radius value according to the median radius of bin 6 in Fig . 4 G . This bin has a closing angle slightly larger than 90\u00b0 . 20 - nm - thick cross - sections of the averages ( N = 100 sites ) registered based on ( D ) the ground truth and ( E ) the spherical fit are shown with the ground - truth U - shaped model ( dotted line ) . Histograms of normalized estimation errors are shown for the parameters ( F ) surface area and ( G ) radius , with mean values of \u2212 2 . 6 and 2 . 5 % respectively . The scale bar is 100 nm . Mund et al . Journal of Cell Biology S3 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 FigureS3 . Non - endocytic clathrinstructures . Results ofLocMoFitanalysis of clathrinstructures inSK - MEL - 2cells ( N = 13cells , n = 1 , 798 sites ) . ( A ) A strong correlation between curvature and \u03b8 can be observed for most structures ( n = 1 , 645 sites , black ) . A disconnected point cloud ( n = 153 sites , 8 . 5 % , red ) indicates the presence of endocytosis - unrelated clathrin structures . ( B \u2013 D ) The same distinct population of data points can be observed for the ( B ) surface area , ( C ) edge length , and ( D ) projected area . ( B ) Example structures from the disconnected population of sites in top view ( xy ) and 50 nm - thick z - slices ( xz ) and their respective fitted \u03b8 values . Scale bar , 100 nm . ( C \u2013 F ) Analysis of clathrin coats not following general trajectory of curvature generation . ( C ) 3T3 cell transiently overexpressing AP2 - GFP . Left : Single - molecule localization microscopy image of immunolabeled clathrin . Middle : Diffraction - limited image of the AP2 - GFP signal . Right : Overlay of the two targets ( scale bar , 10 \u00b5m ) . ( D ) Enlarged image of the section indicated in C ( scale bar , 1 \u00b5m ) . ( E ) Example sites indicated in B ( scale bar , 100 nm ) . 1 : Example for a structure annotated as \u201c GFP positive . \u201d 2 : Example for an \u201c inconclusive \u201d GFP signal . 3 and 4 : Example of \u201c GFP negative \u201d structures . \u201d ( F ) Analysis results when estimating \u03b8 and curvature from clathrin structures and annotating them depending on their AP2 signal ( N = 3 cells and n = 277 sites ) . No AP2 - GFP positive structures are found in the disconnected population of sites , suggesting that they are most likely not generated via clathrin - mediated endocytosis . Mund et al . Journal of Cell Biology S4 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 Figure S4 . Clathrin coat remodeling in three different cell lines . ( A \u2013 C ) Results of LocMoFit analysis for clathrin structures . Different growth models are fitted to curvature H over \u03b8 . The resulting fitting parameters are then used to map the same models also over the surface area , edge length , and projected area ( left to right ) . Purple : Completely flat sites with H = 0 nm \u2212 1 and \u03b8 = 0 . Red : Disconnected sites that were excluded from the fitting . Black line : Rolling median ( window width = 5 % of total number of sites ; A ) U2OS ( N = 3 cell , n grey = 241 sites , n red = 53 disconnected sites , n purple = 1 completely flat sites ) , ( B ) 3T3 mouse fibroblasts ( N = 7 cells , n grey = 688 sites , n red = 51 disconnected sites , n purple = 8 completely flat sites ) , and ( C ) SK - MEL - 2 cells ( N = 13 cells , n grey = 1 , 631 sites , n red = 153 disconnected sites , n purple = 14 completely flat sites ) . ( D \u2013 F ) Temporal reconstruction of clathrin coat remodeling . ( i ) Distribution of \u03b8 slightly differs between cell lines , especially in the earlier states . Median \u03b8 shown as dotted lines correspond to 99 . 6\u00b0 for U2OS ; 121 . 4\u00b0 for 3T3 ; and 108 . 5\u00b0 for SK - MEL - 2 cells . ( ii ) The cooperative curvature model ( red line ) highlights the square - root dependence between \u03b8 and pseudotime . ( iii ) The cooperative curvature model is used to describe the curvature H propagation over pseudotime . Resulting fitting parameters are then used to map the same model to surface area A , edge length \u0190 , and projected area A p . A rolling median is plotted in black ( window width = 5 % of total number of sites ) . Mund et al . Journal of Cell Biology S5 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 Figure S5 . Linkage error investigation and number of localizations per clathrin coat . ( A \u2013 L ) Impact of linkage error on geometric model fit . Indirect immunolabeling displaces the label from the target molecule by two antibodies . This generates a so - called linkage error of on average \u00b1 10 nm , and resulting localizations might not accurately represent the underlying structure of interest ( Fr\u00fch et al . , 2021 ) . ( A ) An ideal case of no linkage error , where the model ( dotted black line ) with radius R M accurately represents the underlying clathrin coat ( red ) with a radius R C . ( B ) Histogram of quantified curvature with a median of 0 . 011 nm \u2212 1 . ( C ) Models fitted to curvature propagation over \u03b8 . ( D ) Uniform displacement of localizations ( light red , \u00b1 10 nm ) due to unbiased labeling by the antibodies . The radius R M still accurately represents the true radius R C . ( E ) Histogram of the curvature , assuming no needed correction of R M . Median : 0 . 011 nm \u2212 1 . ( F ) Models fitted to curvature propagation over \u03b8 remain the same as for C . ( G ) Biased labeling of the antibodies ( light red , + 10 nm ) could result in an overestimation of R M by 10 nm . ( H ) Histogram of the curvature corrected by subtracting 10 nm from quantified R M . Median : 0 . 013 nm \u2212 1 . ( I ) Models fitted to corrected curvature propagation over \u03b8 . ( J ) Biased labeling could result in an underestimation of R M by 10 nm . ( K ) Histogram of the curvature corrected for the overestimation in radius by adding 10 nm from quantified R M . Median : 0 . 010 nm \u2212 1 . ( L ) Models fitted to corrected curvature propagation over \u03b8 . ( C , F , I , and L ) Fitting parameters are A 0 : Fraction of surface area growing as a flat lattice before curvature initiation , defined as A 0 = A ( \u03b8 = 0 . 01 ) / A ( \u03b8 = \u03c0 ) ; R 0 : Preferred radius of the clathrin coat fitted with CoopCM ; H 0 : Preferred curvature of the clathrin coat fitted with CoopCM . Analysis of N = 13 SK - MEL - 2 cells , n = 1 , 645 sites . While the fitting parameters scale with the error in radius estimation , the relationships among the parameters and thus our mechanistic interpretation by the cooperative curvature model still holds true . ( M \u2013 O ) Number of localizations versus surface area . For N = 6 SK - MEL - 2 cells , n = 700 sites the number of localizations found in one clathrin - coated structure was extracted . This is plotted against the quantified surface area determined for each coat . Mund et al . Journal of Cell Biology S6 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023 Video 1 . Pseudo - temporal movie ofclathrin coatremodelingduring endocytosis inSK - MEL - 2 cells . Eachframe correspondsto a slidingwindowaverage of 30 sites , with a frame - to - frame increment of 20 sites . The median pseudotime of those 30 sites is indicated . Provided online is an Appendix , containing a detailed description of the cooperative curvature model . Further , all data shown here are also available in the BioStudies database ( https : / / www . ebi . ac . uk / biostudies / ) under accession number S - BIAD566 . Mund et al . Journal of Cell Biology S7 Remodeling of the clathrin coat during endocytosis https : / / doi . org / 10 . 1083 / jcb . 202206038 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 3 / e202206038 / 1447430 / j c b _ 202206038 . pd f b y gue s t on 09 A ugu s t 2023",
    "linEffectivenessTeachingScience1996": "Research in Science Education , 1996 , 26 ( 4 ) , 495 - 511 The Effectiveness of Teaching Science with Pictorial Analogies Huann - shyang Lin and Bih - ruh Shiau National Kaohsiung Normal University Frances Lawrenz University of Minnesota Abstract This study used a conceptual problem solving test to investigate the effect of a series of pictorial analogies on the concepts of density , pressure , and atmospheric pressure in Year 8 classrooms . The analogies were taught following Glynn ' s teaching with analogies model . It was found that the students taught with the pictorial analogies scored significantly higher than their counterparts ( p < . 01 ) . In addition , the low achievers benefited more from this teaching strategy than did the high achievers . Further , qualitative analysis revealed that most of the students ' alternative conceptions were from preexisting naive intuitions rather than arising from analog instruction . In recent years , teaching science with analogies has attracted attention from science educators . Part of the reason for the interest in analogies is that most theoretical scientific concepts are creative inventions or from scientists ' imaginations . The invention may or may not actually exist in the real world . When students need examples in order to acquire understanding , it is believed that analogies can play a role in helping them get an initial sense of the concepts ( Lawson , 1993 ) . Moreover , analogies can be used to help students develop further understanding through applications in different situations . Analogies can be classified in many different ways . Dagher ( 1995 ) grouped them according to whether the analogies were included in text ( text - based ) , or presented by teachers ( teacher - based ) . Gabel and Sherwood ( 1980 ) divided analogies according to whether they were physical objects and manipulated by students ( physical ) , or descriptions of phenomena to which one related mentally ( verbal ) . This study investigated the effect of using teacher - based verbal analogies ( most of them were pictures ) on the physical science achievement of Year 8 students . The effects of teacher - based verbal analogies have been reported by many researchers . However , the results are not consistent . Most researchers found that analogies are capable of enhancing students ' conceptual understanding ( Brown , 1994 ; Clement , 1993 ; Dupin & Johsua , 1989 ; Raven & Cole , 1978 ; Stavy , 1991 ) . But some studies concluded that there were no significant effects at all ( Dowell , 1968 ; Drugge & Kass , 1978 ; Friedel , Gabel , & Samuel , 1990 ) . The reason why students have not profited from analog teaching has been investigated by researchers . It is concluded that both types of analogy ( Drugge & Kass , 1978 ) and students ' levels of understanding of analogies ( Gabel & Sherwood , 1980 ) are two of the key factors affecting students ' learning performance . Additional research was conducted by Thiele and Treagust ( 1994 ) to understand how teachers use analogies in natural classrooms . The two researchers indicated that there was little evidence that teachers preplanned their analogies . In addition , teachers tended to draw upon their experiences or their own professional readings as a source of analogs , which may account for students ' difficulty in understanding the relationship between analog and target concept . Furthermore , when students encounter the work of mapping from an analog to a target concept , if the mapping support from teachers is not sufficient , the potential fbr misrepresentation is created ( Duit , 1991 ; Zook , 1991 ; Zook & Maier , 1994 ) . 496 LIN , SHIAU AND LAWRENZ In contrast to the above studies where analogies were presented by teachers , Wong ( 1993 ) asked students to create , apply , and modify their own analogies for a given scientific phenomenon . In the process of conducting self - generated analogies , students were asked to specify similarities and differences between the analogies they created and the scientific phenomenon . Although factors related to the effectiveness of using teacher - based verbal analogies are many and varied , researchers have reached consensus that the students ' thorough understanding of the analogy used is one of the critical factors . Friedel , Gabel and Samuel ( 1990 ) indicated that teachers had always assumed that their students understood the analogies being taught , saw the relationship between the analog and the corresponding scientific concept , and were able to transfer their problem solving skills from one problem to another . Actually , it is not necessarily true that students understand the meaning of the analogies teachers use . For instance , in the study by Gabel and Sherwood ( 1980 ) , the two researchers were surprised that there was a large number ( 48 % ) of students who did not understand 90 % of the analogies they used in an experimental treatment . A further analysis revealed that students who understood the analogies scored significantly higher than those who did not . In order to increase students ' understanding of the analogies used in teaching , Glynn ( 1989 ) developed the teaching - with - analogies ( TWA ) model which included the following six operations : 1 . Introduce target 2 . Cue retrieval of analog 3 . Identify relevant features of target and analog 4 . Map similarities 5 . Draw conclusions about target 6 . Indicate where analogy breaks down Harrison and Treagust ( 1993 ) followed the TWA model and used it successfully in a Year 10 optics class on refraction of tight . Glynn , Harrison , and Treagust all stressed the importance of helping students understand the meaning of analogies . With a purpose similar to the TWA model , Brown ( 1994 ) and Clement ( 1993 ) developed bridging analogies to facilitate students ' conceptual change ~ They saw these bridging analogies as conceptually intermediate between situations drawing out valid anchoring intuitions and target problems drawing out naive conceptions . For example , in the teaching of a unit on force , a sequence of analogies was used . The anchor analogy is a hand pressing down on a spring , which is used to let students believe that there is an upward tbrce . The intermediate analogy is putting a book on a flexible board between two sawhorses , in which students still can see the bending of the board . The target problem is putting the same book on a table and asking whether the table pushes up on the book . This teaching strategy of bridging analogies has been used with apparent success ( Clement & Brown , 1984 ; Brown , 1992 ) . A review of the inconsistent results of the research studies described above indicates that more research is needed to determine the effect of using teacher - based verbal analogies . A closer look reveals that most of the studies focused on students ' \" algorithmic problem solving ability . \" Students ' \" conceptual problem solving ability ' , was rarely measured . Science educators have assumed that success in solving mathematical problems indicates mastery of a scientific concept . Consequently , more algorithmic manipulation problems than conceptual problems were included in school tests and research studies . Recently , however , researchers have become more aware of the situation and found that students ' conceptual problem solving ability lagged tar behind their algorithmic problem solving ability ( Nakhleh , ~ \" 1993 ; Nakhleh & Mitchell , 1993 ) . Nakhleh and Mitchell concluded that science teaching must take on a much more concept - based approach . Teacher - based verbal analog teaching is much more closely related to concept - based teaching than algorithm - based teaching . Therefore , the measurement of students ' achievement in these settings should be focused on their conceptual problem solving ability . TEACHING SCIENCE WITH PICTORIAL ANALOGIES 497 Since Brown ' s ( 1994 ) and Clement ' s ( I 993 ) bridging analogy and Glynn ' s ( 1989 ) TWA model have already shown positive results , they were used in this study . Pictorial analogies on the topic of density , pressure , water pressure and atmospheric pressure were developed and used with junior high school students who were struggling to grasp the meaning of these theoretical physical science concepts . The primary purpose of this study was to determine the effect of using teacher - based verbal analogies on students ' conceptual problem solving ability . A secondary purpose of the study was to identify whether high or low achievers profited more from the treatment . The final purpose of the study was to find out what prevents students from correctly solving conceptual problems . Methodology Instrument The test used to monitor students ' conceptual problem solving ability consisted of tour open - ended , essay - type items ( see Appendix ) . For each item , students applied appropriate concepts of density , pressure , water pressure , and atmospheric pressure to explain or predict a phenomenon . Content validity of the test was assured by asking four experts ( two faculty chemists and two science educators ) and two high school science teachers to rate the degree of the representativeness of the content covered , the readability , and the clarity of the items . After rating the students ' responses , the Cronbach alpha reliability of the test was found to be document . 76 . Item I was based on the principle that the boiling point of water depends on the magnitude of the atmospheric pressure on the system . The lowered air pressure caused by the cooling of the ice on the tube , is the main reason that the water boils again . Item 2 asked students in which direction a drop of mercury inside a glass tube connected to a flask would move when the flask was submitted to varying temperatures . The students with only rote learning about the nature of gases could get the first and second parts of the item correct . However , on the third part , they needed a deeper understanding . When the surrounding pressure is kept constant , the volume of the air inside the flask sealed by the mercury ( i . e . , the volume of the system ) will expand or shrink depending on the temperature of the air in the flask . Item 3 dealt with the balance of atmospheric pressure and water pressure . While the bottle is closed by the hand , the atmospheric pressure overcomes the water pressure and prevents the water flowing out the small hole . When the bottle is open , the top of the bottle and the small hole have equal air pressure . Consequently , the water pressure itself will push the water out of the bottle from the small hole . Item 4 examined students ' ability to apply the concept of density in a practical situation . Since the density of gold is given and the two pieces have the same size and shape , the piece with a density of 19 . 3 gcm \" 3 would be the real gold . Scoring Scheme of the Instrument The students ' answers were graded by three science educators with the following scheme for each item : answers with irrelevant statements or misconceptions were given 0 points ; answers with partial misconceptions but that indicated some degree of relevance toward the target concept were assigned 1 point ; answers with sound arguments but minor mistakes of using apparatus or concepts were assigned 2 points ; answers with correct statements and use of target concepts were assigned as 3 points . 498 LIN , SHIAU AND LAWRENZ ( a ) ( b ) ( c ) 1 C ~ ~ Figure 1 . Pictorial analogy of density . TEACHING SCIENCE WITH PICTORIAL ANALOGIES 499 Treatment The pictorial analogies were developed by the authors of this paper . For the concept of density , the analog used to enhance students ' understanding is shown in Figure 1 . In Figure 1 , there are three parts . Part A : Suppose a volley ball has the same mass ( 100g ) as a baseball . Now we have two weightless containers with the same volume . One can be filled with 5 volley balls and the other one can be filled with 30 baseballs . The first one weighs 500g and the other one 3000g . The two containers are then compared to wood and iron . Part B : If we can evenly cut 1 cm 3 from each of the two containers , then the weight of the wood piece ( 0 . 6g ) would be much less than the iron piece ( 7 . 9g ) . At this time , the definition of density was introduced . Part C : When iron is manufactured in different shapes , a 1 cm 3 cut from each of them would have the same mass and the same density . This situation is analogous to that when a small cup of sugar water is taken from a pitcher , it still has the same sweetness . I w I / / / / / I / I ( A ) ( B ) w / I / I / I W / I I I I I I I I Figure 2 . Pictorial analogy of pressure . 500 LIN , SHIAU AND LAWRENZ In order to make sense of the difference between force and pressure , the picture in Figure 2 was shown to students . They were asked : Which way would be easier for lifting the wood ? How much more weight should the people in ( B ) push up than the people in ( A ) ? Subsequently , the bridging case of the spring replacing the person was introduced . Students were asked for the similarities and differences between it and the people case . Finally , students discussed how putting the same piece of wood on a piece of foam in different ways would cause it to sink to different depths in the foam . Toward the end of the lesson , students were reminded , in the above three cases , the pi ' essure that the people , the spring , and the foam experienced are related to the area and the force pushing on that area . In the three cases , the force which resulted from gravity was equal to the weight of the wood . After in ~ oducing the equation P = F / A , students discussed some applications in small groups : for example , why does it hurt more when stepped on by a high heel shoe than by a flat sports shoe ? Students were asked to figure out and explain the relationship of force , pressure and area in this question . latin ( a ) 1033 wa 76cm U ~ ( b ) ( c ) 0 i 0 0 0 t , O Figure 3 . Torricetli ' s experiment , pictorial analogy , and demonstration of atmospheric pressure . TEACHING SCIENCE WITH PICTORIAL ANALOGIES 501 It is common in textbooks to introduce the concept of atmospheric pressure by presenting Torricelli ' s 1643 experiment ( as shown in Figure 3 part a ) . However , there is no guarantee that this will help students understand the existence of atmospheric pressure . In this study , after showing the Torricelli experiment , in order to stress the existence of air pressure , an analog like Figure 3 part b was introduced . The teacher showed and explained that the weight of air pushes down on the surface of the mercury ( like the compressed spring ) in the container , but not on the top of the mercury in the tube ( like the uncompressed spring ) . Toward the end of the lesson , students were asked whether the mercury or the water would stay higher in the tube of Torricelli ' s experiment . They predicted the height of different liquids with various densities . At the end of the discussion , a demonstration was used to let students see the \" push action \" of atmospheric pressure . Shortly after burning a piece of paper inside a I000 mL flask as in part C of Figure 3 , a cooked and shell - peeled egg was put on the mouth of the flask . Students were able to see the egg was quickly pushed by the atmospheric pressure into the flask , because the pressure inside the flask was lower than the outside atmospheric pressure . They were asked why it happened and how the egg could be taken out of the flask . For the concept of water pressure , the pictorial analogy in Figure 4 was shown to students . In part a , the water is referred to as a well - sequenced pile of people . \" The deeper the water , the higher the pressure \" is analogised so that the people at the lowest level of the pile feel the most pressure . Their pressure is also related to the average weight of people above them . After students had an initial understanding , the meaning of the formula P ( pressure ) = H ( height ) x D ( density of the liquid ) was explained . The demonstration of part b was followed by an explanation to stress the relationship between height of the water and its pressure . ( a ) ( b ) Figure 4 . Pictorial analogy and demonstration of water pressure . 502 LIN , SHIAU AND LAWRENZ Procedure Three physical science teachers and six classes of Year 8 students ( N = 225 ) participated in this study . All of them were from a typical school located in central Kaohsiung City of Taiwan . In this school , students were from families representing a broad range of socioeconomic backgrounds . The three teachers were recommended by the school principal and were interested in trying out new teaching strategies . Two of them had 12 years and one had eight years of teaching experience . Originally , it was intended to select two teachers only . However , when the two teachers were asked if they would teach a class without using the pictorial analogies , they refused because they felt morally obligated to use the materials once they had learned about them . They simply could not keep these materials away from their students . Therefore , the authors were forced to use another teacher ' s two classes as the control group . Two classes were selected from each of the two teachers ' sets of classes as the experimental group . In the selection of the classes , an attempt was made to minimise the initial differences between them . Classes were matched on students ' learning achievement ( grade point average ) for each class , and the students ' grades on the first midterm examination of physical science . Each student was classified either as a high achiever or a low achiever based on her / his performance on the first term examination . The cut off point for forming these groups within the experimental and control groups was the median score for each group . After the selection and assignment of the groups was completed , the two teachers of the experimental group participated in a half day workshop . The TWA model was discussed and the pictorial analogies were reviewed until the teachers were comfortable and confident in their use . The teachers were asked strongly to remind their students about the unshared attributes between an analog and its target concept . For example , in Figure 1 of the pictorial analogy for density , students were to be reminded that neither the shape nor the structure of the particles ( molecules ) of wood or iron were like volleyballs or baseballs . All of the pictorial analogies were produced as transparencies for the teachers ' convenience . The teacher in the control group taught the same way as before , which did not include pictorial analogies . It should be noted that in this one and one - half month long quasi - experimental study , all of the six classes devoted the same amount of time to the learning of these concepts . In addition , classroom transactions of each teacher were all video - taped and analysed to assure that the experimental group students were taught with the pictorial analogies . The control group teacher rarely used analog teaching . Instead , her students engaged in extra problem solving . Although both of the control and the experimental group students are all in the same school , because the recess time for each period is as short as 10 minutes , it is not likely that there was cross - talk between students who received the verbal - pictorial analogies and those who did not . At the end of the experiment , one class was randomly selected from each of the three teachers ' classes to take the conceptual problem solving test . In order to assure that every student understood the meaning of each test item , the authors set up the equipment and demonstrated the test , item by item for each class . Students were encouraged to state freely their answers on the answer sheet . It took one hour for each class test . Data Analysis Three science educators were asked to grade the conceptual problem solving test . The first author explained the scoring scheme and gave examples for each criterion . Then they pilot - graded ten students ' answersfor the first item of the test and together checked their gradings . If a student ' s score differed by more than one point , a discussion tbllowed and continued until a consensus was needed about how to assign an appropriate score tbr that situation . Finally , they separately graded the rest of the students ' item one answers . The second , third , and the fourth item of the test were graded using TEACHING SCIENCE WITH PICTORIAL ANALOGIES 503 the same procedure as for the first item . The correlational analysis on the three science educators ' grading revealed a high consistency . The inter - rater correlational coefficients ranged from . 93 to . 95 . Each student ' s total score was counted as the average of the three graders ' scores . Using the students ' scores on the school ' s first term examination ( well - developed and conducted by the school authority ) as the covariate , and the students ' conceptual problem solving scores as the dependent variable , a two - way analysis of covariance ( ANCOVA ) was conducted to investigate whether the experimental group was significantly different from the control group and if the teaching strategy benefited certain ability level students . In addition , qualitative analysis of students ' answers was used to understand students ' alternative conceptions and the barriers that prevented them from correctly solving a problem . The first term examination score was used as the covariate because it was the most highly correlated with conceptual understanding and there was a homogeneity of the slopes of regression on the covariate and the dependent variable . This meant that the slopes of the control group and the experimental group were parallel and their initial difference on the covariate could be adjusted for the comparison of the dependent variable ( Howell , 1982 ) . Quantitative Results and Discussion The two - way analysis of covariance revealed that both of the two independent variables had statistically significant effects . First , there was a significant effect associated with the instructional approach of the teacher - based pictorial analogies ( F = 9 . 11 ; df = 1 ; p < . 01 ) ; and second , there was a significant difference between various learning ability levels ( F = 6 . 22 ; df = 1 ; p < . 05 ) . There was , however , no interaction between the two factors ( F = 2 . 08 ; dr = I ; p > . 05 ) . The result also revealed that five percent of the variance was accounted for by the instructional approach , three percent by learning ability level , and 29 % by the covariate ( i . e . , the first term examination ) . Table 1 shows the means and standard deviations of the control and experimental group for students ' achievement on the conceptual problem solving test and the covariate ( Science 1 ) . It can be seen that the adjusted mean of the experimental group ( 3 . 94 ) is significantly higher than the adjusted mean of the control group ( 2 . 55 ) . In order to examine the effect on low and high achievers , mean analysis and t - tests were conducted . Table 2 shows the high and low achievers ' means and standard deviations on the covariate and the conceptual problem solving test . For the high achievers the adjusted mean score of the experimental group is higher than the adjusted mean of its counterpart ( 5 . 04 vs . 4 . 08 , p = . 20 ) , but the difference is not statistically significant . Table I Means and Standard Deviations of the Conceptual Problem Solving Test for Experimental and Control Groups Mean scores ( SD ) Group N Science 1 Adjusted Experiment 69 49 . 07 3 . 94 \" * ( 22 . 43 ) ( 0 . 25 ) Control 36 59 . 80 2 . 55 ( 18 . 56 ) ( 0 . 36 ) * * t - test for difference in adjusted means , p < . 01 504 LIN , SHIAU AND LAWRENZ Table 2 The High and Low Achievers ' Means and Standard Deviations the Conceptual Problem Solving Test Mean scores ( SD ) Ability Group N Science I Adjusted High Experiment 35 67 . 31 5 . 04 ( 13 . 81 ) ( 0 . 34 ) High Control 18 74 . 83 4 . 08 ( 9 . 77 ) ( 0 . 48 ) Low Experiment 34 30 . 29 2 . 70 \" * ( 11 . 13 ) ( 0 . 37 ) Low Control 18 44 . 78 1 . 16 ( 11 . 03 ) ( 0 . 54 ) * * t - test for difference in adjusted means , p < . 01 In contrast , the comparison of low achievers on the control and experimental groups showed a statistically significant difference . The low achievers in the experimental group outperformed their counterparts ( 2 . 70 vs . 1 . 16 , p < . 01 ) . This indicates that the main effect is due to the low achievers . As mentioned earlier , the design of the study assigned different teachers for the control and experimental groups . A major concern arises from this design ; that is , how can we be sure that the difference between the experimental and control groups is really from the treatment instead of from the different teacher ' s teaching ' ? . There is no easy way to answer this question . However , in order to be more confident , we observed the control and the experimental group students ' learning progress pattern to check if there were any difference in the patterns . The results of term examinations conducted by the school authority were used for comparison . Since the correlational analysis revealed that the scores on the term examinations are highly related to the study ' s conceptual problem solving scores ( r ranged from . 68 to . 70 , p < . 0001 ) , it is reasonable to analyse the term examination scores for reference . As can be seen from Figure 5 , among the first , second , and third term examinations , the comparison of the control and the experimental groups ' achievement reveals that the second one has the narrowest margin ( mean difference = 6 ) , while the first and the third term examinations both have an 11 point margin . This result indicates that the treatment ( which occurred between the first and the second term examinations ) resulted in an effect on students ' performance on the school examinations . Apparently , without the treatment , the students ' performance returned to the original level ( mean difference = 11 ) . Two further comparisons between the control and the experimental groups on the second and the third term examinations provide more evidence on the effect of the treatment . The first comparison uses the first term examination score as a covariate , and the second term examination score as the dependent variable . It is apparent that there was a significant treatment effect ( F = 5 . 87 ; df = 1 ; p < . 05 ) . Looking at the adjusted means of the two groups in Table 3 reveals that the experimental group outperformed its c ' 6unterpart ( 46 . 78 vs . 42 . 79 , p < . 05 ) . The same procedure was used for the second comparison except that the third term examination score was used as the dependent variable . TEACHING SCIENCE WITH PICTORIAL ANALOGIES 505 Mean 80orr 60 , o 55 ' 50 ' x 45 ' o : control group x : experimental group Ist 2nd 3rd Term examination Figure 5 . The learning progress pattern of the control and the experimental group . The ANCOVA result indicates that there was no statistically significant difference on the third term examination score between the two groups . Again , Table 3 provides the adjusted means of the two groups . The adjusted mean score of the experimental group ( 52 . 10 ) is very close to the adjusted mean score of the control group ( 51 . 87 ) . The above two comparisons and the pattern of Figure 5 provides more confidence in the treatment effect for the following reasons . If the experimental group ' s performance on the second term exam resulted from teacher effects , then why didn ' t it happen on the third term examination which dealt with different content without using analogies ? In addition , how would one account for the consistent differences between the two groups on the first and the third term examinations , while the difference on the second term examination is unique ? Table 3 Means and Standard Deviations of the Second and Third Term Examinations for Experimental and Control Groups Mean scores ( SD ) Group N Science 1 Adjusted - second Adjusted - third examination examination Experiment 69 49 . 07 46 . 78 * 52 . I0 ( 22 . 43 ) ( 1 . 37 ) ( 1 . 61 ) Control 36 59 . 80 42 . 79 51 . 87 ( 18 . 56 ) ( 1 . 89 ) ( 2 . 26 ) * t - test for difference in adjusted means , p < . 05 Results and Discussion Table 4 shows the main alternative conceptions about boiling and atmospheric pressure , pressure , and density as well as the problematic aspects of understanding held by the students . As can be seen , most of the alternative conceptions are from students ' naive intuitions instead of instruction . For example , in Item 1 , even if students were taught that the boiling point of water will change when the water is heated at different altitudes , some students still answered that the boiling of water at a 506 LIN , SHIAU AND LAWRENZ temperature under 100 \" C was not really boiling . Despite this alternative conception , some students correctly performed the lbllowing algorithmic problem solving item that appeared in the school term examination : What is the atmospheric pressure at 2000 metres up on a mountain ? Would the boiling point of the water heated on the mountain be higher or lower than 100 \" C ? The students successfully used the mathematical formula and calculated the atmospheric pressure at that height : P = 1033 . 6 gcm ~ - - 10 . 9gcm z x H / t 00 , where the H is the height of the mountain . They could solve the problem algorithmically but had alternative conceptions . This result seems to support the ' conclusion of Anderson and Smith ( 1987 ) , Gussarsky and Gorodetsky ( 1990 ) and Helm and Novak ( 1983 ) . These researchers concluded that students bring to instruction a variety of naive alternative conceptions that are often at odds with scientific ideas , In a problem solving situation , the students are likely to make their own associations based on concepts used in their daily living to justify an answer . For Item 2 , it is also apparent that students ' predictions were based on their daily experiences rather than scientific reasoning : \" Heating the empty flask wilt create steam inside the flask to push the mercury outward , \" and \" The moving motion of the mercury will be like the moving of the mercury in a working body thermometer . \" It can be seen that students ' familiarity with cooking food and reading thermometers influenced their explanations . In addition , students had difficulty in identifying the variables involved in the problem . Most students answered that \" The mercury won ' t move , since both the left side and the right side of the mercury have the same temperature . \" They regarded the situation of part three of the item as different from part one or part two , in which , the flask has been submitted to a different temperature . In fact , the \" system \" is the air inside the flask sealed by the mercury . Since the system ' s volume is not a constant ( i . e . , the mercury is moveable ) , when the temperature around the \" system \" is decreased at the condition of n ( number of moles of air ) and p ( pressure of the surrounding ) being fixed ( p = I atm ) , the volume of the \" system \" will be decreased . In other words , it is cooling the \" system \" at constant pressure . Table 4 Summary of Students ' Alternative Conceptions and Difficulties About Density , Atmospheric Pressure , and Pressure 1 . Water can only be boiled at 100 \" C . Any building happened below 100 \" C is a fake phenomenon . 2 . A turbulence caused by a confrontation between a cold front and a hot front will result in building of the liquid . 3 . When heat is trapped inside a test tube , the water inside the test tube will boil , even without providing any additional heat . 4 . When water flows out from the bottle , the space originally occupied by the water must be replaced by air , otherwise the water can ' t flow . 5 . Heating the empty flask in item 2 will create steam inside the flask to push the mercury outward . 6 . The moving motion of the mercury in item 2 will like the moving of the mercury in a working body thermometer . 7 . General problematic aspects of problem solving ( a ) verbalisation and language difficulties ( b ) improper handling of the density definition ( c ) lack of , mastery of the pressure concept ( d ) difficulty of identifying variables ( factors ) involved in problem solving situation TEACHING SCIENCE WITH PICTORIAL ANALOGIES 507 In Item 3 , the students ' alternative conception number 4 in Table 4 deserves a special note . Again , this is a student ' s spontaneous naive intuition and coincidentatly , this conception is much like the seventeenth century doctrine of \" horror vacui . \" At that time the Aristotelian science tbllowers believed that nature abhors a vacuum . Because today students are experienced in the use of straws to drink and the use of a pipette to transfer fluids , it is easy for them to believe that the space originally occupied by the water in a bottle must be replaced by air . They simply believe that when the water flows out from the bottle , it is like the use of a straw . The air sucked from the straw is soon replaced by the drink . The finding of this alternative conception provides support for the argument of Matthews ( 1994 ) who pointed out the similarity between the students ' naive intuition and ancient scientists ' beliefs about air pressure . Theretbre , the addition of history of science in science teaching may help more students construct scientific concepts from naive concepts . For Item 4 , the majority of the students have difficulty in making hypotheses and designing experimental procedures to identify the real gold . These two scientific processing skills are rarely practised in most classrooms and laboratories . Most experiments were carried out by students as recipes like pertbrming cooking with the required procedures and directions provided . If students do not have the chance to explore inquiry procedures , they may remember the knowledge successfully from their teacher ' s instruction . However , it would be difficult for them to apply this knowledge in a different situation , because they are not equipped with the two critical scientific processing skills . Although quite a few students still held naive conceptions and did not construct the scientific concepts successfially even after the treatment with pictorial analogies , the significant progress of the experimerltal group should not be ignored . Especially , it is encouraging that there is almost no indication of negative results such as those described by Zook and Di Vesta ( 1991 ) . They pointed out that some students ' misconceptions were the products of teaching because of the analogies used by both teachers and textbooks . The results of this study show that if the analogies are appropriately presented , such as the systematic use of analogies in the TWA model and the bridging analogies , the negative effect can be avoided . While looking at the positive effect of the study resulting from the treatment of analogies , readers are reminded that the novelty effect ( Gay , 1981 ) should be considered . As mentioned in the methodology part , the experimental group teachers refused to teach a class for comparison without using the pictorial analogies . This ethical issue indicates that these teachers might quickly recognise the power of the pictorial analogies to enhance understanding . This raises an important question in the study : How much did this enthusiasm affect their teaching ? Although the teaching of analogies lasted for one and one - half months and enthusiasm may have diminished , this study did not investigate whether the gains were due to \" cold \" rationalism or \" hot \" motivational factors . This can be investigated in future studies by changing the experimental design or extending the treatment to a longer period of time ( Gay , 1981 ) . Conclusion and Implications for Science Education The results of the study appear to confirm the research report of Dreistadt ( 1969 ) , Dupin and Johsua ( 1989 ) and Raven and Cole ( I 978 ) . However , the above researchers concluded that the effect of analogies was mostly on students ' algorithmic problem solving ability . This study added the improvement of students ' conceptual problem solving ability as one more of the benefits of using teacher - based pictorial analogies in science teaching . A further analysis revealed that the low achievers benefited more from the teaching strategy than did the high achievers . The reason could be as indicated by Gabel and Sherwood ( 1980 ) : that is , since the capable students are able to think at the formal level , it is not necessary to make the theoretical concepts concrete for them . 508 LIN , SHIAU AND LAWRENZ The quantitative and qualitative results together indicate the fruitful efficacy of using teacher - based verbal analogies . In the quantitative analysis , the statistically significant difference between the experimental and the control group students ' achievement on the conceptual problem solving test revealed the effectiveness of this teaching strategy . The additional analysis on the students ' progress pattern on the three term examinations confirmed that the difference of achievement between the two groups was mainly attributable to the treatment and not to a teacher effect . Finally , the qualitative results indicated that no negative effect was found from the analog teaching61 It is apparent that if analogies are presented as research studies recommend , students ' learning outcomes can be enhanced . Unfortunately , analogies are not used by science teachers as often as expected ( Treagust , Duit , Joslin , & Lindauer , 1992 ) . In addition , Treagust ( 1993 ) points out that the majority of science teachers have no formal training in the use of analogies . In order to provide empirical evidence , this study followed the recommendations of former analog studies in developing pictorial analogies and providing a halt - day inservice training for the teachers involved with the experimental group . The fruitfulness of the study can be used to encourage the implementation of teacher - based pictorial verbal analogies in science classrooms . Despite the positive et - ti ~ ct of teacher - based verbal analogies , the students still performed poorly on the conceptual problem solving test . The tact that students ' conceptual problem solving ability lags tar behind algorithmic problem solving ability has been also noted by Nakhleh ( 1993 ) . In this study , the students scored only a mean of four points out of a possible total of 12 points on the conceptual problem solving test . Their difficulties were varied in many ways , ranging from various alternative conceptions to improper handling of a concept ' s definition . It may not be easy for students to describe , predict , and explain phenomena in the natural world , since they were rarely asked to do so in previous examinations . Most examinations , including standardised tests , centre on computational skills and recalling definitions . Questions that require students to synthesise information and apply concepts are not very common in such examinations . The students ' low achievement on the conceptual problem solving test may serve to remind science educators that more efforts are needed to focus on conceptual - teaching pedagogies . Goswami ( 1991 ) indicated that even young children can reason analogically in both classical and problem analogy tasks as long as they have knowledge of the relations used in the analogies . However , even a simple analog leaves great scope for students to arrive at their own conclusion about the related science content . Most analogies in textbooks are provided with limited extent of mapping between analog and target ( Thiele , Venville , & Treagust , 1995 ) . In addition , not only do teachers tend to use analogies spontaneously rather than in a pre - planned way , but they always draw on their own experiences ( not students ' prior knowledge ) as a source of those analogies ( Thiele & Treagust , 1994 ) . This appears to result in difficulties tbr some students . If positive effects are expected from the use of analogies in science teaching , based on the results of this study , then pre - planned and systematically presented bridging analogies are strongly recommended . Acknowledgement The authors gratefully thank the two anonymous reviewers of this article for their constructive contributions and helpful comments . In addition , this research was made possible by the financial aid from the National Science Council ( NSC 84 - 2511 - S - 017 - 005 ) , Taiwan , Republic of China . Correspondence : Huann - shyang Lin , Department of Chemistry , National Kaohsiung Normal University , Ho - ping 1st Road , Kaohsiung , Taiwan . Internet email : t 1666 @ nknucc . nknu . edu . tw TEACHING SCIENCE WITH PICTORIAL ANALOGIES 509 References Anderson , C . W . , & Smith , E . L . ( 1987 ) . Teaching science ( Report Series No . 169 ) . East Lansing , MI : Michigan State University Institute for Research on Teaching . ( ERIC Document Reproduction Service No . ED 274 541 ) . Brown , D . E . ( 1992 ) . Using examples to remediate misconceptions in physics : Factors influencing conceptual change . Journal of Research in Science Teaching , 29 ( 1 ) , 17 - 34 . Brown , D . E . ( 1994 ) . Facilitating conceptual change using analogies and explanatory models . hzternational Journal of Science Education , 16 ( 2 ) , 201 - 214 . Clement , J . ( 1993 ) . Using bridging analogies and anchoring intuitions to deal with students ' preconception in physics . Journal of Research in Science Teaching , 30 ( 10 ) , 1241 - 1257 . Clement , J . , & Brown , D . E . ( 1984 ) . Using analogical reasoning to deal with deep misconceptions in physics ( Technical Report No . 95 ) . Amherst , University of Massachusetts : Scientific Reasoning Research Institute . Dagher , Z . R . ( 1995 ) . Review of studies on the effectiveness of instructional analogies in science education . Science Education , 79 ( 3 ) , 295 - 312 . Dowell , R . E . ( 1968 ) . The relations between the use of \" analogies and their effect on student achievement in teaching selected concepts in high school biology . Dissertation Abstract International 29 ( 10 ) , 3519 - A ( University Microfilms No . 69 - 6732 ) . Dreistadt , R . ( 1969 ) . The use of analogies and incubation in obtaining insight in creative problem solving . Journal of Psychology , 71 , 159 - 175 . Drugge , N . L . , & Kass , H . ( 1978 , March ) . The effect of selected analogies on understanding of scientific explanations . Paper presented at the annual meeting of the National Association for Research in Science Teaching . ( ERIC Document Reproduction Service No . ED 152 - 537 ) Duit , R . ( 1991 ) . On the role of analogies and metaphors in learning science . Science Education , 75 , 649 - 672 . Dupin , J . J . , & Johsua , S . ( 1989 ) . Analogies and modelling analogies in teaching : Some examples in basic electricity . Science Education , 73 , 207 - 224 . Friedel , A . W . , Gabel , D . L . , & Samuel , J . ( 1990 ) . Using analogies for chemistry problem solving . School Science and Mathematics , 90 , 674 - 682 . Gabel , D . L . , & Sherwood , R . D . ( 1980 ) . Effect of using analogies on chemistry achievement according to Piagetian level . Science Education , 64 ( 5 ) , 709 - 716 . Gay , L . R . ( 1981 ) . Educational Research ( p . 218 ) . Columbus , Ohio : Merrill . Glynn , S . M . ( 1989 ) . Explaining science concepts : A teaching - with - analogies model . In S . M . Glynn , R . H . Yeany , & B . K . Britton ( Eds . ) , The psychology of learning science ( pp . 219 - 240 ) . Hillsdale , NJ : Lawrence Erlbaum . Goswami , U . ( 1991 ) . Analogical reasoning : What develops ? A review of research and theory . Child Development , 62 ( 1 ) , 1 - 22 . Gussarsky , E . , & Gorodetsky , M . ( 1990 ) . On the concept \" Chemical equilibrium \" : The associate framework . Journal of Research in Science Teaching , 27 , 197 - 204 . Harrison , A . G . , & Treagust , D . F . ( 1993 ) . Teaching with analogies : A case study in grade - 10 optics . Journal of Research in Science Teaching , 30 ( 10 ) , 1291 - 1307 . Helm , H . , & Novak , J . D . ( Eds . ) . ( 1983 ) . Misconceptions in science and mathematics . Proceedings of the international seminar . Ithaca , New York : Cornell University . Howell , D . C . ( ~ 982 ) . Statistical methods of psychology ( 2nd ed . ) . Boston : Prindle , Weber and Schmidt . Lawson , A . E . ( 1993 ) . The importance of analogy : A prelude to the special issue . Journal of Research in Science Teaching , 30 ( I 0 ) , 1213 - 1214 . 510 LIN , SHIAU AND LAWRENZ Matthews , M . R . ( 1994 ) . Science teaching : The role of history and philosophy of science . New York : Routledge . Nakhleh , M . B . ( 1993 ) . Are our students conceptual thinkers or algorithmic problem solvers ? Journal of Chemical Education , 70 ( 1 ) , 52 - 55 . Nakhleh , M . B . , & Mitchell , R . C . ( 1993 ) . Concept learning versus problem solving , there is a difference . Journal of Chemical Education , 70 ( 3 ) , 190 - 192 . Raven , R . J . , & Cole , R . ( 1978 ) . Relationships between Piaget ' s operative coml ~ rehension and physiology modelling process of community college students . Science Education , 62 ( 4 ) , 481 - 489 . Stavy , R . ( 1991 ) . Using analogy to overcome misconceptions about conservation of matter . Journal of Research in Science Teaching , 28 , 305 - 313 . Thiele , R . B . , & Treagust , D . F , ( 1994 ) . An interpretive examination of high school chemistry teachers ' analogical explanations . Journal of Research in Science Teaching , 31 ( 3 ) , 227 - 242 . Thiele , R . B . , Venville , G . J . , & Treagust , D . F . ( 1995 ) . A comparative analysis of analogies in secondary biology and chemistry textbooks used in Australian schools . Research in Science Education , 25 ( 2 ) , 221 - 230 . Treagust , D . F . , Duit , R . , Joslin , P . , & Lindauer , I . ( 1992 ) . Science teachers ' use of analogies : Observations from classroom practice . International Journal of Science Education , 14 , 413 - 422 . Treagust , D . F . ( 1993 ) . The evolution of an approach for using analogies in teaching and learning science . Research in Science Education , 23 , 293 - 301 . Wong , E . D . ( 1993 ) . Self - generated analogies as a tool for constructing and evaluating explanations of scientific phenomena . Journal of Research in Science Teaching , 30 , 367 - 380 . Zook , K . B . ( 1991 ) . Effects of analogical processes on learning and misrepresentation . Educational Psychology Review , 3 , 41 - 72 . Zook , K . B . , & Di Vesta , F . J . ( 1991 ) . Instructional analogy and conceptual misrepresentations . Journal of Educational Psychology , 83 ( 2 ) , 246 - 252 . Zook , K . B . , & Maier , J . M . ( 1994 ) . Systematic analysis of variables that contribute to the formation of analogical misrepresentation . Journal of Educational Psychology , 86 ( 4 ) , 589 - 600 . Appendix The Conceptual Problem Solving Test Give reasons supporting each answer or describe your answer to the following questions . 1 . As shown in the following figure , a test tube is filled with ~ / 5 volume of water and heated to boiling point ( 100 \" C ) . The test tube is then removed from the burner and a rubber stopper ( equipped with a thermometer ) is inserted in the end of the test tube . Why does the water boil again if a sack of ice is put around the upper end of the test tube ( the reading of the thermometer is below 100 ~ 4 f TEACHING SCIENCE WITH PICTORIAL ANALOGIES 511 . As shown in the following figure , an empty flask is sealed with a rubber stopper which includes a glass tube ( at the end of the glass tube , there is a drop of mercury ) . ( a ) Can you predict and explain the movement of the mercury if the flask is immersed in the beaker filled with water of 3 ~ ( b ) Can you predict and explain the movement of the mercury if the flask is immersed in the beaker filled with water of 80 \" C ? ( c ) Can you predict and explain the movement of the mercury if the flask ( not including the beaker ) is put inside the refrigerator of 5 ~ . As shown in the Ibllowing figure , a plastic bottle with a small hole on the side is filled with water . The water flows out from the small hole . As soon as you put your hand on the top of the bottle , the water stops flowing . Why does the water stop flowing ? 4 . In a jewelry store , there are two pieces of golden metal with the same size , shape , and appearance . Knowing that the density of pure gold is 19 . 3gcm 3 , how can you use the following equipment to identify which piece is the real gold ? e e",
    "rhyscoxDirectedDiversityLeveraging2021a": "Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation Samuel Rhys Cox \u2217 National University of Singapore , Singapore samuel . cox @ u . nus . edu Yunlong Wang \u2217 National University of Singapore , Singapore yunlong . wang @ nus . edu . sg Ashraf Abdul National University of Singapore , Singapore ashrafabdul @ u . nus . edu Christian von der Weth National University of Singapore , Singapore vonderweth @ nus . edu . sg Brian Y . Lim \u2020 National University of Singapore , Singapore brianlim @ comp . nus . edu . sg ABSTRACT Crowdsourcing can collect many diverse ideas by prompting ideators individually , but this can generate redundant ideas . Prior methods reduce redundancy by presenting peers\u2019 ideas or peer - proposed prompts , but these require much human coordination . We introduce Directed Diversity , an automatic prompt selection ap - proach that leverages language model embedding distances to maxi - mize diversity . Ideators can be directed towards diverse prompts and away from prior ideas , thus improving their collective creativity . Since there are diverse metrics of diversity , we present a Diver - sity Prompting Evaluation Framework consolidating metrics from several research disciplines to analyze along the ideation chain \u2014 prompt selection , prompt creativity , prompt - ideation mediation , and ideation creativity . Using this framework , we evaluated Di - rected Diversity in a series of a simulation study and four user studies for the use case of crowdsourcing motivational messages to encourage physical activity . We show that automated diverse prompting can variously improve collective creativity across many nuanced metrics of diversity . CCS CONCEPTS \u2022 Human - centered computing ; \u2022 Collaborative and social computing ; \u2022 Collaborative and social computing theory , concepts and paradigms ; \u2022 Computer supported cooperative work ; KEYWORDS Diversity , Collective Creativity , Crowdsourcing , Ideation , Motiva - tional messaging , Collective Intelligence , Creativity Support Tool \u2217 Co - first authors , ordered alphabetically \u2020 Corresponding author This work is licensed under a Creative Commons Attribution International 4 . 0 License . CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan \u00a9 2021 Copyright held by the owner / author ( s ) . ACM ISBN 978 - 1 - 4503 - 8096 - 6 / 21 / 05 . https : / / doi . org / 10 . 1145 / 3411764 . 3445782 ACM Reference Format : Samuel Rhys Cox , Yunlong Wang , Ashraf Abdul , Christian von der Weth , and Brian Y . Lim . 2021 . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation . In CHI Conference on Human Factors in Computing Systems ( CHI \u201921 ) , May 08 \u2013 13 , 2021 , Yokohama , Japan . ACM , New York , NY , USA , 34 pages . https : / / doi . org / 10 . 1145 / 3411764 . 3445782 1 INTRODUCTION Crowdsourcing has been used to harness the power of human creativity at scale to perform creative work such as text editing [ 7 , 21 , 78 ] , iterating designs [ 27 ] , information synthesis [ 54 ] , and motivational messaging [ 4 , 50 , 95 ] . In such tasks , empowering crowd workers to ideate effectively and creatively is key to achiev - ing high - quality results . Different prompting techniques have been proposed to stimulate creativity and improve the diversity of ideas [ 2 , 27 , 50 , 95 ] , but they suffer from ideation redundancy , where multiple users express identical or similar ideas [ 10 , 48 , 76 , 80 ] . Current efforts to avoid redundancy include iterative or adaptive task workflows [ 99 ] , constructing a taxonomy of the idea space [ 40 ] , and visualizing a concept map of peer ideas [ 80 ] , but these re - quire much manual effort and are not scalable . Instead , we propose an automatic prompt selection mechanism \u2014 Directed Diversity \u2014 to scale crowd ideation diversity . Directed Diversity composes prompts with one or more phrases to stimulate ideation . It helps to direct workers towards new ideas and away from existing ideas with the workflow : 1 ) extract phrases from text corpuses in a target domain , 2 ) embed phrases into a vector embedding , and 3 ) automat - ically select phrases for maximum diversity . These phrases are then shown as prompts to ideators to stimulate ideation . The phrase em - bedding uses the Universal Sentence Encoder ( USE ) [ 14 ] to position phrases within an embedding vector space . Using the embedding vectors , we calculated distances between phrases to optimally select phrases that are farthest apart from one another ; this maximizes the diversity of the selected phrases . Hence , Directed Diversity guides ideators towards under - utilized phrases or away from existing or undesirable phrases . The embedding space provides a basis to calculate quantitative , distance - based metrics to estimate diversity in selected phrases and prompts , and subsequently ideated messages . These met - rics can complement empirical measurements from user studies evaluate prompts and ideations . We curate multiple measures CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . and evaluation techniques and propose a Diversity Prompting Evaluation Framework to evaluate perceived and subjective cre - ativity and objective , computed creativity , and diversity of crowd ideations . We demonstrate the framework with experiments on Di - rected Diversity to 1 ) evaluate its efficacy to select diverse prompts in a simulation study , 2 ) measure the perceived diversity of selected prompts and effort to generate ideas in an ideation study , and 3 ) evaluate the creativity and diversity of generated ideas in valida - tion studies using quantitative and qualitative analyses . The exper - iments were conducted with the application use case of writing motivational messages to encourage physical activity [ 2 , 3 , 50 , 95 ] , though we discuss how Directed Diversity can apply to other crowd ideation tasks . In summary , our contributions are : 1 . We present Directed Diversity , a corpus - driven , automatic approach that leverages embedding distances based on a language model to select diverse phrases by maximizing a diversity metric . Using these constrained prompts , crowd - workers are directed to generate more diverse ideas . This results in improved collective creativity and reduced redun - dancy . 2 . A Diversity Prompting Evaluation Framework to evaluate the efficacy of diversity prompting along an ideation chain . This draws constructs from creativity and diversity literature , metrics computed from a language model embedding , and is validated with statistical and qualitative analyses . 3 . We applied the evaluation framework in a series of four experiments to evaluate Directed Diversity for prompt selec - tion , and found that it can improve ideation diversity without compromising ideation quality , but at a cost of higher user effort . 2 BACKGROUND AND RELATED WORK We discuss related research on supporting crowd ideation with the cognitive basis for creative ideation , how creativity support tools help crowd ideation , and how artificial intelligence can help collective intelligence . 2 . 1 Cognitive Psychology of Creative Ideation Different cognitive models of creativity have been proposed to explain how ideation works . Memory - based explanation models describe how people retrieve information relevant to a cue ( prompt ) from long - term memory and process it generate ideas [ 1 , 26 , 53 , 67 , 68 ] . Since retrieval is dependent on prompts , they need to be sufficiently diverse to stimulate diverse ideation [ 68 ] , otherwise people may fixate on a few narrow ideas [ 42 ] . Ideation - based models [ 64 ] explain how individuals can generate many ideas through complex thinking processes , including analogical reasoning [ 36 , 46 , 61 ] , problem constraining [ 84 ] , and vertical or lateral thinking [ 35 ] . We focus on prompting to promote memory - based retrieval than these other reasoning processes . Besides cue - based retrieval and thinking strategies , other factors influence ideation creativity , such as personal traits , motivation to perform the task , and domain - relevant skills that can affect individual creativity [ 90 ] . We provide technological support to improve the creative mental process , rather than to select creative personalities , recruit domain experts , or improve task motivation . Next , we discuss how different cognitive factors have been leveraged at scale to support creative ideation with the crowd . 2 . 2 Creativity Support Tools for Crowd Ideation Creativity Support Tools have been widely studied in HCI to en - able crowdworkers to ideate more effectively and at scale [ 31 , 32 ] . Showing workers ideas from their peers has been very popular [ 18 , 33 , 79 , 81 ] , but can have limited benefit to creativity if peer ideas are too distant from the ideators\u2019 own ideas [ 18 ] . Other ap - proaches include employing contextual framing to prompt ideators to imagine playing a role for the task [ 69 ] or using avatars for vir - tual interactions while brainstorming [ 57 ] . While these methods focus on augmenting individual creativity , they do not coordinate the crowd , so multiple new ideations may be redundant . More re - cent approaches apply provide more explicit guidance to workers . IdeaHound [ 80 ] visualizes an idea map to encourage workers to focus on gaps between peer ideas , but does not inform what ideas or topics will fill the gaps . BlueSky [ 40 ] and de Vries et al . [ 95 ] use crowd or expert annotators to construct taxonomies to constrain the sub - topics for ideation , but these taxonomies require significant manual effort to construct and are difficult to scale . Chan et al . [ 16 ] employed latent Dirichlet allocation ( LDA ) to automatically identify topics , but this still requires much manual curation which does not scale to many topics . With Directed Diversity , we automatically extract a phrase corpus and embed the phrases as vectors , and se - lect diverse phrases for focused prompting . We employ pre - trained language model to provide crowd ideation support , thus we next dis - cuss how artificial intelligence can support collective intelligence . 2 . 3 Supporting Collective Intelligence with Artificial Intelligence Collective Intelligence is defined as groups of individuals ( the collec - tive ) working together exhibiting characteristics such as learning , judgement and problem solving ( intelligence ) [ 56 ] . Crowdsourcing is a form of collective intelligence exhibited when crowdworkers work towards a task mediated by the crowdsourcing platform . How - ever , managing crowdwork to ensure data quality and maximize efficiency is difficult because of the nature and volume of the tasks , and varying abilities and skills of workers [ 97 ] . HCI research has contributed much towards this with interfaces to improve crowd - worker efficiency , designing incentives for workers , and workflows to validate work quality [ 9 , 38 , 62 , 89 , 97 ] . Furthermore , recent de - velopments in artificial intelligence ( AI ) provides opportunities to complement human intelligence to improve the quality and effi - ciency of crowd work [ 47 , 97 ] , optimize task allocation [ 22 , 28 ] , adhere to budget constraints [ 45 ] , and dynamically control quality [ 11 ] . With Directed Diversity , we used AI to optimize ideation di - versity by shepherding the crowd towards more desired and diverse ideation with diverse prompt selection . 3 TECHNICAL APPROACH We aim to improve the collective diversity of crowdsourced ideas by presenting crowdworker ideators , with carefully selected prompts that direct them towards newer ideas and away from existing ones . The prompts presented to the ideators consist of one or more phrases Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Figure 1 : Pipeline of the overall technical approach to extract , embed , and select phrases to generate diverse prompts . a ) Phrase extraction by collecting phrases from online articles and discussion forums ( shown as pages ) , filtering phrases to select a clean subset ( shown as the black dash for each phrase ) ; b ) Phrase embedding using the Universal Sentence Encoder [ 14 ] to compute the embedding vector of each phrase ( shown as scatter plot ) ; c ) Phrase Selection by constructing the minimal spanning tree to select optimally spaced phrases ( see Figure 2 for more details ) . that represent ideas that are distinct and different from prior ideas . Prompts can have one or more phrases . As a running example throughout the technical discussion and experiments , we apply our approach to the application of motivational messages for healthy physical activity , where it is important to collect diverse motiva - tional messages [ 50 , 95 ] . Figure 1 shows the 3 - step overall approach to extract , embed , and select phrases . We next describe each of these steps in detail . 3 . 1 Phrase Extraction We extracted phrases from selected sources of documents with the following semi - automatic data - driven process : 1 ) collect a corpus of documents , 2 ) tokenize documents into sentences , 3 ) extract phrases as constituent structures , 4 ) filter for length , slang , emoti - cons . We collected documents about exercising , weight loss , and healthy living from two types of sources : i ) credible , authoritative health news articles 1 to obtain texts relevant to the domain ( health and fitness ) , and ii ) discussion posts from popular subreddits 2 of online health communities related to fitness and physical activity to obtain texts relevant to the task ( motivational messaging [ 50 , 95 ] ) . Together , the combined corpus contained 3 , 235 articles and 32 , 721 user posts . To extract phrases , we tokenized each document into sentences and performed Part - of - Speech ( POS ) tagging using Python Spacy to select phrases that form syntactic constituents [ 13 ] . From each sentence ( e . g . , \" Regular exercising helps to improve people\u2019s health at any age . \" , we extract verb phrases ( e . g . , \" helps to improve \" ) , noun phrases ( e . g . , \" regular exercising , \" \" people\u2019s health , \u201d \u201cage\u201d ) and prepo - sitional phrases ( e . g . , \u201cat any age\u201d ) . To provide more context to each phrase , we combined adjoining verb and noun phrases to generate noun - verb phrases ( e . g . \u201cregular exercising helps to improve\u201d ) and verb - noun phrases ( e . g . , \u201chelps to improve people\u2019s health\u201d ) . After extracting the phrases , we filtered phrases for length , quality , and relevance . We kept phrases that were 3 to 5 words long , since short phrases may not sufficiently stimulate creativity and long prompts may restrict creativity . Since user posts often contain typographical errors , slang , or other stylistic devices ( e . g . , emoticons ) , we kept phrases that only contain words from a dictionary 3 of American 1 Source of three authoritative websites on health : www . health . harvard . edu , www . medicinenet . com , www . webmd . com . 2 Source of 20 subreddits from www . reddit . com : 90daysgoal , advancedfitness , advance - drunning , bodyweightfitness , c25k , crossfit , fitness , gainit , getmotivated , ketogains , kettlebells , leangains , loseit , motivation , powerlifting , running , selfimprovement , swim - ming , weightroom , xxfitness . 3 Debian Wordlist pagkage . packages . debian . org / es / sid / wordlist and British words . To reduce repetition of phrases , we removed shorter phrases that overlapped with longer phrases ( e . g . , excluded \u201cfederal exercise recommendations\u201d , kept \u201cfederal exercise recom - mendations and guidelines\u201d ) . The final corpus contained clean 3 , 666 phrases . We next describe the construction of the multi - dimensional idea space to characterize how the phrases are separated or similar . 3 . 2 Phrase Embedding The corpus of extracted phrases provides a large set of poten - tial phrases for prompting , but we seek to select phrases that are least similar to one another . For each phrase , we obtain a multi - dimensional vector representation , called an embedding , so that the phrase is a data point in an idea space . Similar work by Siangliu - lue et al . [ 79 ] obtained embeddings of N = 52 ideas by training a Crowd Kernel model [ 91 ] from 2 , 818 triplet annotations is not scalable to our corpus of N = 3 , 666 phrases , since that would need N ( N \u2212 1 ) ( N \u2212 2 ) / 3 = 16 . 4 million triplets . Instead , similar to Chan et al . \u2019s [ 18 ] use of GloVE [ 71 ] , we use pre - trained language models based on deep learning to encode each word or sentence as a vector representation . Specifically , we use the more recent Universal Sen - tence Encoder ( USE ) [ 14 ] to obtain embeddings for phrases in our corpus , compute their pairwise distances , and selected a maximally diverse subset of phrases . Our approach is generalizable to other language embedding techniques [ 98 ] . To obtain the phrase embedding presentation , we use a pre - trained USE model 4 to obtain embedding vectors for each phrase . With USE , all embeddings are 512 - dimensional vectors are located on the unit hypersphere , i . e . , all vectors are unit length , and only their angles are different . Hence , the dissimilarity between two phrase embeddings x i and x j is calculated as the angular distance arccos ( x i , x j ) , which is between 0 and \u03c0 . For our phrase corpus , the pairwise distance between phrases ranged from Min = 0 . 06 to Max = 0 . 58 , Median = 0 . 4 , inter - quartile range 0 . 39 to 0 . 46 , SD = 0 . 043 ; see Appendix Figure 10 . We use the same USE model to compute embeddings and distances for ideated messages . For a dataset of 500 motivational messages ideated in a pilot study with no prompting , the pairwise distance between ideations ranged from Min = 0 . 169 to Max = 0 . 549 , Median = 0 . 405 , inter - quartile range 0 . 376 to 0 . 432 , SD = 0 . 043 ; see Appendix Figure 11 . Table 1 shows example phrases 4 Pre - trained Universal Sentence Encoder model ( https : / / tfhub . dev / google / universal - sentence - encoder / 4 ) , which was trained using both unsupervised learning on Wikipedia , web news , web question - answer pages , and discussion forums , and super - vised learning on Stanford Natural Language Inference ( SNLI ) corpus . CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Table 1 : Demonstration of pairwise embedding angular distances between an example text items ( first data row ) and neighbor - ing text items . Text items with semantically similar words have smaller distances . For interpretability , we highlighted words to indicate darker color with higher cosine similarity to the first phrase . a ) Example extracted Phrases Distance Phrase tofirstPhrase app with yoga poses 0 ( self ) yoga really taking off 0 . 284 popular form of yoga today 0 . 304 yoga pants or sweats 0 . 351 of handstand push - ups 0 . 406 on the road to diabetes 0 . 475 b ) Example Ideations from Ideation User Study Distance Ideated Message tofirstIdeation Exercise will release endorphins and you will feel good for a while after doing it . 0 ( self ) Exercise releases endorphins and makes you feel better ! 0 . 171 Exercise relieves stress in both the mind and the body . It\u2019s the best way to get your mental health in check . 0 . 301 We are the leading country in obesity . Do you want to be part of ? 0 . 509 and messages and their corresponding pairwise dissimilarity dis - tances . With the embedding vectors and pairwise distances for all phrases , the next step selects diverse phrases with which to prompt ideators . 3 . 3 Phrase Selection Given the embeddings of the curated phrases , we want to select the subset of phrases with maximum diversity . Mathematically , this is the dispersion problem or diversity maximization problem of \u201carranging a set of points as far away from one another as pos - sible\u201d . Among several diversity formulations [ 20 ] , we choose the Remote - MST diversity formulation [ 37 ] ( also called Remote - tree [ 20 ] or functional diversity [ 72 ] ) that defines diversity as the sum of edge weights of a minimum spanning tree ( MST ) over a set of vertices . It is robust against nonuniformly distributed data points ( e . g . , with multiple clusters , see Table 4 ) . We construct the minimum spanning tree by performing agglomerative hierarchical clustering on the data points with single linkage [ 82 ] . Next , we describe how we select phrases as prompts to direct ideators towards diverse phrases , or away from prior ideas . Figure 2 illustrates the technical approach . 3 . 3 . 1 Directing towards Diverse Phrases . For phrase selection , we aim to select a fixed number of points n from the corpus with maximum diversity . This is equivalent to finding a maximal edge - weighted clique in a fully connected weighted graph , which is known to be NP - hard [ 39 ] . Hence , we propose a scalable greedy approach that uses the dendrogram representation of the MST re - sulting from the hierarchical clustering . Starting from the root , we set the number of clusters to the desired number of phrases n . For each cluster C r , we select the phrase that is most distant from other points , with largest minimum pairwise distance from all points from outside the cluster , i . e . , x r = argmax i \u2208 C r (cid:18) min j (cid:60) C r d (cid:0) x i , x j (cid:1)(cid:19) where x r is the diverse phrase selected in cluster C r , x i is a point in cluster C r and x j is a point in the corpus not in C r , and d is the pairwise distance between x i and x j . This method has O ( n 2 ) time complexity and runs in less than one second on a desktop PC for 3 . 6k phrases ; it is generalizable and can be substituted for other approximate algorithms to select most diverse points [ 20 , 41 ] . Figure 2 ( top row ) illustrates the phrase selection method to direct towards areas without ideations : a ) Start with all phrases in a corpus represented as USE embed - ding points . b ) Construct a dendrogram ( MST ) from all points , using single - linkage hierarchical clustering . c ) Set # clusters equal to desired number of diverse phrases . For each cluster , find the most distant phrase . d ) Selected phrases are the approximately most diverse from the corpus , for the desired number of phrases . 3 . 3 . 2 Directing Away from Prior Ideas . Other than directing ideators towards new ideas with diverse prompts , it is important to help them to avoid prior ideas written by peers . We further propose a method to remove corpus phrases that are close to prior ideas so that ideators do not get prompted to write ideas similar to prior ones . The method , illustrated in Figure 2 ( bottom row ) , is similar as before , but with some changes : a ) Add the embedding points of prior ideas to the corpus . b ) Calculate phrase - ideation distance d ( x Pi , x Ij ) for each phrase x Pi and ideation x Ii and exclude phrases too close to the ideas , i . e . , d < \u03b4 , where \u03b4 is an application - dependent threshold , \u03b4 = 0 . 29 in our case . c ) Same as step ( c ) , but different clusters , since fewer points are clustered . d ) Same as step ( d ) , but different prompts would be selected , even if the number of phrases is the same . 3 . 3 . 3 Directing with Prompts of Grouped Phrases . Instead of prompting with only one phrase , prompting with multiple related terms can help ideators to better understand the concept being prompted and generate higher quality ideas [ 17 , 67 , 83 ] . We extend the phrase selection method to group multiple phrases in a sin - gle prompt using the following greedy algorithm . After step ( a ) , we i ) sorted phrases by descending order of minimum pairwise distance for each phrase to produce a list of seed candidates , ii ) for each seed phrase , perform a nearest neighbors search to re - trieve a specified prompt size ( number of phrases \u0434 in a prompt ) Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Figure 2 : Procedure to direct ideation towards diverse phrases ( top ) and away from prior or redundant ideas ( bottom ) . To attract ideation with diverse prompts : a ) start with embeddings of corpus - extracted phrases ; b ) construct minimum spanning tree ( MST ) ; c ) traverse tree to select distant prompts from clusters ( most distant points as green dots , in clustered phrases as green ellipses ) ; d ) selected prompts are the most diverse . To repel ideation from prior ideas , e ) compute embeddings of prior ideas ( red hollow dots ) ; f ) compute prompt - ideation pairwise distances of all prompts from each prior ideation , exclude phrases ( dotted black circles ) with pairwise distance less than a user - defined threshold ( red bubble ) , and construct the MST with remaining phrases ; g ) traverse MST to select a user - defined number of prompts ; h ) selected prompts are diverse , yet avoids prior ideas . and remove the selected neighbors from the seed list , iii ) repeat seed neighbor selection until n seed phrases have been processed . We grouped the phrases into a prompt and calculate its embed - ding point x Pri as the angular average of all phrases x Pk in the prompt , i . e . , x Pri = \u0434 (cid:205) k = 1 x Pk / Z , where Z = \u0434 (cid:205) k = 1 x Pk 2 is the magni - tude of the vector sum and x Pri is also a unit vector . We then perform steps ( b ) to ( d ) with the prompts x Pri instead of individual phrases . Note that the corpus of prompts will be smaller than the corpus of phrases . This approach has disjoint prompts that do not share phrases , but there can be alternative approaches to group phrases 5 . 4 DIVERSITY PROMPTING EVALUATION FRAMEWORK To evaluate the effectiveness of the Directed Diversity prompt se - lection technique to improve the collective creativity of generated ideas , we define an ideation chain as a four step process ( 3 top ) : 1 ) setting the prompt selection technique will influence 2 ) the creativ - ity of selected prompts ( prompt creativity ) , 3 ) the ideation process 5 Analternativeapproachis , afterstep ( c ) , tosimplygroupnearestneighbors . However , this will cause the prompt embeddings to be shifted after the diversity is maximized , so it may reduce the diversity of the selected prompts . of the ideators ( prompt - ideation mediation ) , and 4 ) the creativ - ity of their ideation ( ideation creativity ) . We propose a Diversity Prompting Evaluation Framework , shown in 3 , to measure and track how creative and diverse information propagates along this ideation chain to evaluate how and whether a creativity prompting technique improves various measures of creativity and diversity in outcome ideas . Note that our proposed framework is descriptive to curate many useful metrics , but not prescriptive to recommend best metrics . 4 . 1 Research Questions and Experiments Prompt stimuli act along the ideation chain to increase ideation diversity , but it is unclear how well they work and at which point along the chain they may fail . We raise three research questions between each step in the ideation chain , which we answer in four experiments ( Section 2 ) with various measures and factors . RQ1 . How do the prompt techniques influence the perceived di - versity of prompts ? ( RQ1 . 1 ) How do they affect diversity in prompts ? ( RQ1 . 2 ) How well can users perceive differences in creativity and diver - sity in these prompts ? These questions relate to the prompt selection technique effectiveness and serve as a manipulation check . We an - swer them in a Characterization Simulation Study ( Section 2 . 1 ) with objective diversity measures , and an Ideation User Study ( 2 . 2 ) with subjective measures perceived prompt diversity measures . CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Figure 3 : Diversity prompting evaluation framework to evaluate prompting to support diverse ideation along the ideation chain . We pose research questions ( RQ1 - 3 ) between each step to validate the ideation diversification process . For each step , we manipulate or measure various experiment constructs to track how well ideators are prompted to generate creative ideas . Except for prompt selection , each construct refers to a statistical factor determined from factor analyses of multiple dependent variables . Constructs are grouped in colored blocks indicating different data collection method ( \u25a1 Computed embedding - based metric , \u25a1 ratings from ideators , \u25a1 ratings from validators , \u25a1 thematic coding of ideations ) . RQ2 . How does diversity in prompts affect the ideation process for ideators ? ( RQ2 . 1 ) Do differences in diversity affect ideation ef - fort ? ( RQ2 . 2 ) How well do ideators adopt and apply the content of the prompts ? ( RQ2 . 3 ) How does prompt creativity affect diversity in ideations ? We answer these questions as a mediation analysis in the Ideation User Study ( Section 5 . 2 ) with objective measures of task time and similarity between ideations and stimulus prompts , the - matically coded creativity metrics , and perceived ease of ideation . RQ3 . How do prompt selection techniques affect diversity in ideations ? Having validated the manipulation checks , we evalu - ate the effectiveness of prompt selection techniques in questions in the Ideation User Study ( Section 5 . 2 ) with subjective measures self - assessed creativity and thematically coded creativity metrics , and two Validation User Studies ( Section 2 . 3 ) with subjective mea - sures of perceived creativity . 4 . 1 . 1 Independent variables of Prompt Specifications . We manip - ulated prompt selection technique , prompt count , and prompt size as independent variables ; these are detailed in Appendix Table 6 . We chose Random prompt selection as a key baseline where selec - tion is non - trivial and data - driven based on our corpus , but not intelligently selected for diversity . 4 . 2 Diversity and Creativity Measures of Prompting and Ideation We measured diversity and creativity for selected prompts and generated ideas with embedding - based and human rated metrics . We color code variable names based on data collection method as in 3 4 . 2 . 1 Embedding - based Diversity Metrics for Prompts and Ideations . Although crowd creativity research has focused on the mean pair - wise distance as a metric for idea diversity , our literature review has revealed many definitions and metrics . Here , we describe com - putational metrics calculated from the embedding - based distances . Inspired by Stirling\u2019s general framework diversity framework [ 87 ] , we collect definitions from crowd ideation [ 15 , 27 , 40 , 79 , 80 ] , ecol - ogy [ 24 , 73 , 94 ] , recommender systems [ 29 , 44 , 60 , 93 ] , and theoreti - cal computer science [ 20 , 37 ] . These cover many aspects of diversity to characterize the mean distance and minimum Chamfer distance between points , MST - based dispersion , sparseness of points around the median , span from the centroid , and entropy to indicate the even - ness of points in the embedding vector space . Table 2 and Table 3 describe distance metrics for individual and collective text items , respectively . These metrics describe nuances of diversity , which we illustrate with example distributions in Table 4 . Other measures of Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Table 2 : Metrics of distances between two points in a multi - dimensional vector space . Each metric can be calculated for an individual text item . These metrics can apply to the embedding of phrases or ideations . Metric Definition Interpretation MeanPairwise Distance 1 N \u2212 1 N (cid:205) j = 1 d ( x i , x j ) Average distance of all other points to the current point . MinimumPairwiseDistance min j (cid:60) i d ( x i , x j ) Distance of closest neighbor to current point . This focuses on redundancy and ignores points that are very far from the current point . Table 3 : Metrics of diversity of phrases or ideation embeddings in a vector space . These capture more characteristics of diversity than average distances in Table 2 . Each metric can only be calculated collectively for multiple items . Metric Definition Interpretation Remote - Clique 1 N 2 (cid:205) i , j d ( x i , x j ) Average of mean pairwise distances . While commonly used in crowd ideation studies [ 27 , 44 , 80 ] , it is insensitive to highly clustered points . ChamferDistance 1 N N (cid:205) i = 1 min j (cid:60) i d ( x i , x j ) Average of minimum pairwise distances . Chamfer distance [ 43 ] ( or Remote - pseudoforest [ 20 ] ) measures the distance to the nearest neighbor . However , it is biased when points are clustered . MSTDispersion Mean of MST edge distances 1 | E MST | (cid:205) ( x i , x j ) \u2208 E MST d ( x i , x j ) Popular in ecology research as functional diversity [ 72 ] , and called Remote - tree or Remote - MST [ 20 , 37 ] , this learns a minimum spanning tree ( MST ) of the points , and calculates the sum of edge weights . Span percentile P % d ( x i , \u00af x M ) P th percentile distance to centroid ( \u00af x M = N (cid:205) i = 1 x Mi / N ) ; i . e . , \u201cradius\u201d of distribution [ 12 , 65 ] . We calculate 90 th percentile to centroid ( vs . medoid ) to be robust against outliers and skewed distributions , respectively . Sparseness Mean distance to medoid 1 N N (cid:205) i = 1 d ( x Mi , \u02dc x M ) Sparsity of points positioned around the medoid ( \u02dc x M = argmin x i { N (cid:205) j = 1 d ( x i , x j ) } ) [ 51 , 52 , 77 ] . If points cluster around the medoid , then this metric will be small ( i . e . , not sparse ) . Entropy Shannon - Wiener index for points in a grid partition (cid:205) b f b log ( f b ) This index [ 75 , 86 ] indicates how evenly points are distributed ; more even is more diverse . We calculated entropy for a 2D projection of the USE feature space to avoid high time complexity 6 and divided the space into a 5 \u00d7 5 grid 7 , and counted the frequency f b of points in each bin b . diversity and divergence [ 20 ] can be included in the framework , which we defer to future work . Next , we describe human - subjects ratings to validate these embedding - based metrics with measures that do not depend on the embeddings to avoid circular dependency . 4 . 2 . 2 Creativity Measures for Ideations . Along with the computed diversity metrics , we evaluate with qualitative characteristics of creativity . From creativity literature , we draw from Torrance\u2019s [ 92 ] description of several measures for creativity , including quality , flexibility and originality . Quality measures whether an ideation is \u201cusable , practical , or appropriate\u201d [ 66 ] . We asked ideators to self - assess on a 5 - point Likert scale their message\u2019s effectiveness ( towards motivation ) and creativity . We ask validator crowdworkers to rate each individual ideation on a 7 - point Likert scale whether it is effective ( motivating [ 95 ] ) , helpful 8 [ 50 , 88 ] , and informative [ 50 ] 8 Note that a message could be helpful but written with negative impressions and thus not motivating . towards encouraging physical activity ; rank collections of ideations on effectiveness , informativeness and unrepetitiveness . ; and rate the pairwise difference between ideation pairs from each collection . Note that Directed Diversity was not designed to improve quality , since these metrics were not explicitly modeled . Flexibility [ 85 ] measures how many unique ideas were generated , and originality [ 100 ] measures how infrequently each idea occurs . These require expert annotation to identify distinct categories . We conducted a thematic analysis on the messages using open coding [ 34 ] to derive categories and affinity diagramming [ 8 ] to consolidate categories to themes ( see details in Appendix Table 19 ) . We calculate the flexibility and originality measures based on the coded categories ( fine - grained ) and themes ( coarser ) described in Appendix Table 7 4 . 2 . 3 Creativity Measures for Prompts . As a manipulation check , it is important to verify that prompts that are computed as more diverse , are perceived by ideators as more creative . Since perceived CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Table 4 : Comparison of diversity metrics for canonical examples of distributed points in a 2D space . Points farther apart mean higher diversity . Here , we calculate Euclidean instead of angular distance , but intuitions are similar . creativity encompasses more qualitative effects , computed diver - sity may not be correlated with creativity . Thus , we measure the creativity and usefulness of prompts by asking about prompt under - standability , relevance to domain topic ( physical activity ) , relevance to task 9 ( motivation ) , helpfulness to inspire ideation , and unexpect - edness [ 66 ] along 7 - point Likert scales . 4 . 2 . 4 Mediating Variables for Prompt - Ideation Process . Even if more diverse prompts can facilitate more creative ideation , it is important to understand whether this requires more effort and time , how the consistency of phrases within prompts affect ideation , and how well ideators adopt words and concepts from the phrases into their ideations . We measure effort as ease of ideation with a 7 - point Likert scale survey question . For individual creativity , fluency [ 30 ] is defined as the number of ideas an individual writes within a fixed time . Chan et al . had also measured fluency for an 8 - minute crowd ideation task [ 18 ] . In contrast , we asked ideators to only write one idea per prompt without time constraint , so we measure the in - verse relation of ideation task time to generate one ideation [ 5 ] . Specifically , since task time is skewed , we use \u2013 Lo\u0434 ( ideation time ) to represent fluency . For prompts with more than one phrase , the similarity between phrases can affect their perceived consistency . Therefore , we measure the intra - prompt mean phrase and prompt average phrase Chamfer distances ( Appendix Table 8 ) to indicate the similarity between intra - prompt phrases . We measure the adoption of the prompt ideas by calculating the proportion of words from phrases in the ideations as prompt recall and prompt precision , and computing the prompt - ideation distance between the embeddings of the prompt and ideation ( Appendix Table 9 ) . 4 . 3 Factor analyses to draw constructs from experiment variables With the numerous variables from our experiments , we observed some may be correlated since they measure similar notions or participants may confound questions to have similar meanings . We employed an iterative design - analytical process to organize and consolidate variables into factors with the following steps . 9 Note that a prompt could be relevant to the domain , but not motivating . \u2022 Identify metrics of creativity and diversity from a literature review from various research domains , such as ecology , cre - ativity , crowdsourcing , theoretical computer science , recom - mender systems ( Section 1 . 8 . 1 ) . Ideate additional measures and questions to capture user behavior and opinions when generating and validating ideas . We refine and reduce mea - sures based on survey pilots and usability testing . \u2022 Collect measurements of each metric with different meth - ods : a ) Compute embedding - based metrics from prompts shown and messages written . This was computed individ - ually for each text item ( e . g . , mean pairwise distance ) and collectively for all text items in each prompt technique ( e . g . , Remote - MST diversity ) . b ) Measure perception ratings and behavioral measures regarding reading prompts and ideating messages and rating messages . We asked text ra - tionale to help with interpretations . c ) Measure subjective thematic measures to qualitatively assess the collective creativity with thematic analysis and idea counting . \u2022 Perform factor analysis on quantitative data to organize correlated variables into fewer factors . Variables are first grouped by data collection method 10 and analyzed together . To determine the number of factors , we examined scree plots and verified grouped variables as consistent with constructs from literature . The final number of factors are statistically significant by the Bartlett Test of Sphericity ( all p < . 0001 ) . See Appendix Tables 10 - 17 for the results of the factor analysis , including factor loadings and statistical significance . Table 5 summarizes the learned factors from 42 variables that we developed . \u2022 Perform statistical hypothesis testing using these learned factors to answer our research questions . 5 EVALUATION : APPLYING FRAMEWORK TO STUDY DIRECTED DIVERSITY We have described a general descriptive framework for evaluat - ing diversity prompting . We applied it to evaluate our proposed 10 E . g . , individual text item metrics , collective text items metrics , ratings of text item from ideators , ratings of text item from validators , ratings of collection of text items from validators . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Table 5 : Constructs from factor analyses of variables along ideation chain . Factor loadings in Appendix Tables 10 - 17 . Chain Factor Construct Interpretation P r o m p t C r e a t i v i t y Prompt Distance How distant and isolated the prompt is from other prompts . Prompt Consistency How similar ( consistent ) the phrases in a prompt are . Prompt Dispersion How spread out the selected prompts are from one another . Prompt Evenness How evenly spaced the selected prompts are among themselves . Prompt Unexpectedness Ideator rating of how unexpected a prompt was on a 5 - pt Likert scale . Prompt Understandability Ideator rating of how understandable a prompt was on 5 - pt Likert scale Prompt Relevance Ideator rating of prompt relevant to the domain ( i . e . exercise ) on 5 - pt Likert scale Prompt Quality Ideator rating of the overall quality of prompt on 5 - pt Likert scale . P r o m p t - I d e a t i o n M e d i a t i o n Ideation Fluency Ideator speed to ideate ( reverse of time taken ) . Ideation Ease Ideator ease of ideating based on multiple 5 - point Likert scale ratings . Phrase Adoption Measures the extent of phrase usage from the prompts in the ideation . I d e a t i o n C r e a t i v i t y Ideation Distance How distant and isolated the ideation is from other ideations . Ideation Dispersion How spread out the ideations are from one another . Ideation Evenness How evenly spaced the ideations are among themselves . Ideation Flexibility Count of unique categories / themes across all ideations . Ideation Originality How rare each category / theme is across all ideations . Ideation Self - Quality Ideator self - rating of the overall quality of the ideation on 5 - pt Likert scale . Ideation Quality Validator rating of overall quality of individual ideation on 7 - pt Likert scale . IdeationInformative - Helpfulness Validator rating of informativeness and helpfulness of individual ideation on 7 - pt Likert scale . Ideations Unrepetitive Validator cumulative rating of non - redundancy in collection of ideations . Ideations Informative Validator cumulative rating of informativeness in collection of ideations . Ideations Motivating Validator cumulative rating of overall quality of collection of ideations . Ideations Pairwise Difference Validator rating of difference between a pair of ideations in collection . Directed Diversity prompt selection technique against baseline ap - proaches ( no prompting , random prompt selection ) in a series of experiments ( characterization , ideation , individual validation , col - lective validation ) , for the use case of crowd ideating motivational messages for physical activity . Here , we describe the procedures for each experiment and their results . 5 . 1 Characterization Simulation Study The first study uses computational methods to rapidly and scalably evaluate prompt selection techniques . This helps us to fine tune prompt parameters to maximize their potential impact in later human experiments . 5 . 1 . 1 Experiment Treatments and Method . We varied three inde - pendent variables ( prompt selection , prompt count , prompt size ) to measure the impact on 7 dependent variables of distance and diver - sity metrics . We varied Prompt Selection technique ( None , Random , or Directed ) to investigate how much Directed Diversity improves prompt diversity with respect to baseline techniques . For None prompt selection , we simulated ideation with 500 ideas collected from a pilot study where crowd ideators wrote messages without prompts . We simulated Random selection by randomly selecting phrases from the phrase corpus ( Section 1 . 4 ) and Directed selec - tion with our technical approach ( Sections 1 . 4 to 1 . 6 ) . If we assume that prompt embeddings are an unbiased estimator for ideation embeddings , then this gives an approximation of ideation diver - sity due to prompting . We conducted experiments for directing towards diverse prompts and for directing away from the 500 pilot prior ideations . We varied the number of prompts ( Prompt Count , n = 50 , 150 , . . . , 950 ) to simulate how diversity increases with the number of ideation tasks performed . This investigates how diversity increases as the budget for crowd tasks increases . To investigate how well Directed selection avoids prior ideations , we varied the number of repeller prior ideations ( Repeller Prior Ideations Count , n R = 50 , 100 , 150 , 200 ) . We varied the number of phrases in prompts ( Prompt Size , \u0434 = 1 to 5 ) to simulate ideating on one or more phrases in each prompt . We computed the prompt embedding as the average of all phrases in the prompt . For Random selection , we randomly chose phrases to group together for each prompt . This random neighbor selection will lead to variation in prompt consistency , but does not bias the prompt embedding on average . For Directed selection , phrases in each prompt were chosen as described in Sec - tion 1 . 6 . 3 . 5 . 1 . 2 Results on Manipulation Efficacy Analysis ( RQ1 . 1 ) . We visu - alized ( Figure 4 ) the phrase embeddings to help to interpret how the selected prompts are distributed , whether they are well spread out , clustered , etc . We used Uniform Manifold Approximation and Projection ( UMAP ) [ 59 ] to reduce the 512 dimensions of USE to a 2D projection . Hyperparameters were selected such that the 2D CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Figure 4 : 2D UMAP projection showing how diversely selected prompts and resulting ideation messages are distributed based on Directed or Random prompt selection technique and prompt size ( number in brackets ) . Each point represents the embed - ding of a text item . Light grey points represent all phrases in the extracted corpus , dark grey points represent selected phrases from the simulation study ( Section 5 . 1 ) and blue dots represent the ideated messages written by crowdworkers in the ideator user study ( Section 5 . 2 ) . Gradient lines connect ideation messages to their stimulus prompts . points in UMAP had pairwise distances correlated with that of the 512 - dimension USE embeddings . We can see that Directed prompt selection led to prompts that were more spread out , and less redundant from prior ideation . This is more pronounced for higher prompt size ( \u0434 = 3 ) . Random ( 3 ) had lower diversity than None with tighter clustering of prompts ( grey points in middle - bottom graph ) than of messages ( blue points in left graph ) . This was because Random ( 3 ) prompts averaged their embeddings from multiple phrases , such that this variance of means of points is smaller than the variance of points 11 . We further conducted a characterization study with 50 simulations for each prompt configuration to confirm that Directed Diversity improves diversity and reduces redundancy from prior ideations for various embedding - based metrics ( see Appendix E and Figure 12 ) . 5 . 2 Ideation User Study The Ideation User Study serves as a manipulation check that higher prompt diversity can be perceived by ideators , and as an initial evaluation of ideation diversity based on computed and thematically coded metrics . 5 . 2 . 1 Experiment Treatment and Procedure . We conducted a between - subjects experiment with two independent variables prompt selection technique ( None , Random , Directed ) and prompt size ( \u0434 = 1 and 3 ) , and kept constant prompt count n = 250 . The None condition ( no prompt ) allows us to measure if the quality of ideations become worse due to the undue influence of phrases in prompts . The Random condition provides a strong baseline since it also leverages the extracted phrases in the first step of Directed Diversity . A prompt size of \u0434 > 1 can provide more contexts to help 11 This is analogous to standard error is to standard deviation ideators understand the ideas in the phrases , but may also lead to more confusion if the phrases are not consistent ( too dissimilar ) . Figure 5 shows example prompts that ideator participants see in different conditions . The experiment apparatus and survey ques - tions were implemented in Qualtrics ( see Appendix Figures 13 - 19 for instructions and question interface ) . 5 . 2 . 2 Experiment Task and Procedure . Participants were tasked to write motivational messages and answer questions with the following procedure : Read the introduction to describe the experiment objective and consent to the study . 1 . Complete a 4 - item word associativity test [ 19 ] to screen for English language skills . 2 . Write 5 messages to motivate for physical activity for a fitness mobile app . For each message , one at a time , a ) On the first page , depending on condition , see no prompt or a prompt with one or three phrases selected randomly or by Directed Diversity ( see Figure 5 ) , then write a motivational message in one to three sentences . This page is timed to measure ideation task time . b ) Rate on a 5 - point Likert scale the experience of ideating the current message : ease of ideation ( described in Section 4 . 2 . 4 ) , self - assessed success in writing motivationally , and success in writing creatively ( Section 4 . 2 . 2 ) ; perception of the prompt on : understandability , relevance to domain topic ( physical activity ) , relevance to task ( motivation ) , helpfulness for inspiration , and unexpectedness ( Section 4 . 2 . 3 ) . c ) Reflect and describe in free text on their rationale , thought process , phrase word usage , and ideation effort . We analyze Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Figure 5 : Example prompts shown to participants in different conditions : None ( left , \u0434 = 0 ) , Directed ( 1 ) ( center , \u0434 = 1 ) , and Directed ( 3 ) ( right , \u0434 = 3 ) . Phrase texts would be different for Random ( 1 ) and Random ( 3 ) selection techniques . these quotes to verify our understanding of the collected quantitative data . \u2022 Answer demographics questions , and end the survey by receiving a completion code . 5 . 2 . 3 Experiment Data Collection and Statistical Analyses . We re - cruited participants from Amazon Mechanical Turk with high qual - ification ( \u2265 5000 completed HITs with > 97 % approval rate ) . Of 282 workers who attempted the survey , 250 passed the test to complete the survey ( 88 . 7 % pass rate ) . They were 45 . 2 % female , between 21 and 70 years old ( M = 38 . 6 ) ; 76 . 4 % of participants have used fitness apps . Participants were compensated after screening and were ran - domly assigned to one prompt selection technique . Participants in the None condition were compensated with US $ 1 . 80 , while others with US $ 2 . 50 due to more time needed to answer the additional sur - vey questions about prompts . Participants completed the survey in median time 15 . 4 minutes and were compensated > US $ 8 / hour . We collected 5 messages per participant , 50 participants per condition , 250 ideations per condition , and 1 , 250 total ideations . For all response variables , we fit linear mixed effects models de - scribed in Appendix Tables 20 - 23 . To allow a 2 - factor analysis , we divided responses in the None ( 0 ) condition ( no prompt , 0 phrases ) randomly and evenly to None ( 1 ) and None ( 3 ) . Results are shown in Figure 6 . We performed post - hoc contrast tests for specific dif - ferences identified . Due to the large number of comparisons in our analysis , we consider differences with p < . 001 as significant and p < . 005 as marginally significant . Most significant results reported are p < . 0001 . This is stricter than a Bonferroni correction for 50 comparisons ( significance level = . 05 / 50 ) . We next describe the sta - tistically significant results for prompt mediation check ( RQ1 . 2 ) , mediation analysis ( RQ2 . 1 , 2 . 2 ) , and ideation evaluation ( RQ3 . 1 , 3 . 2 ) . We include participant quotes from their rationale text response where available and relevant . 5 . 2 . 4 Results of Manipulation Check on Creativity and Mediation on Ideation Effort ( RQ1 . 2 , 2 . 1 ) . We discuss findings on how ideators perceived creativity factors in prompts and how prompt config - urations affected their ideation effort . Figure 6 ( Top ) shows that compared to Random , Directed Diversity selected prompts that were more unexpected ( good for diversity ) ; but were slightly more difficult to understand ( by half unit on 5 - point Likert scale ) , very slightly less relevant ( 1 / 4 unit ) , and of slightly lower quality ( 1 / 2 unit ) . However , the relevance of the selected diverse prompts was not explicitly controlled . P173 in Directed ( 1 ) felt that the phrase \u201cfirst set of challenges is\u201d was \u201cstraightforward and gave me the idea of what to write . It was very easy\u201d ; whereas P157 in felt that the phrase \u201creview findings should be\u201d \u201cdidn\u2019t really have anything I could think to tie towards a motivational message . I tried to think of it as looking back to see progress in terms of reviewing your journey . \u201d Random prompts with more phrases were harder to understand , perhaps , because they were randomly grouped and are less seman - tically similar . P128 in Random ( 3 ) found that \u201cthese [ phrases ] were hard to combine since they deal with different aspects of exercise . Also the weight lifting seems to be not the best thing for addressing obesity , so that was hard to work in . \u201d We found that ideation effort was mediated by prompt factors . Figure 6 ( Bottom ) shows that Directed prompts were least easy to use for ideation , and less adopted than Random selected prompts . This is consistent with Directed prompts being less understandable than Random . Ideating with 1 - phrase prompts increased ideation time from 44 . 1s by 21 . 6s ( 48 . 9 % ) compared to None , and viewing 3 phrase increased time further by 11 . 9s . In summary , Directed Diver - sity may improve diversity by selecting unexpected prompts , but at some cost of ideator effort and confusion . This cost compromises prompt adoption and suggests that directing diversity may not work . Yet , as we will show later , Directed Diversity does improve ideation creativity . We analyzed the confound of understandability further in Appendix Section K . Next , we investigate if prompts characteristics mediate more ideation creativity . 5 . 2 . 5 Results of Mediation Analysis of Diversity Propagation from Prompt to Ideation ( RQ2 . 3 ) . We found that prompt configuration and perceived prompt creativity mediated the individual diver - sity of ideated messages ( RQ2 . 2 ) . Appendix Table 21a ( in ) shows that Ideation Mean ( or Min ) Pairwise Distance increased with Prompt Mean ( or Min ) Pairwise Distance by + 0 . 176 ( or + 0 . 146 ) , and marginally with Intra - Prompt Phrase Mean Distance by + 0 . 021 ( or + 0 . 020 ) . This means that farther Prompts stimulated farther Ideations , and higher variety of Phrases within each prompt drove slightly farther Ideations too . Hence , prompt diversity ( mean pair - wise distance ) influenced ideation diversity , and prompt redundancy ( minimum pairwise distance ) influenced ideation redundancy . Ap - pendix Table 21b shows that as Prompt Relevance decreased by one Likert unit ( on 5 - point scale ) , ideation mean pairwise distance decreased by 0 . 0034 ( 7 . 9 % of ideation pairwise distance SD of 0 . 043 ) CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Figure 6 : Results of ideators\u2019 perceived prompt creativity ( Top ) and ideation effort ( Bottom ) for different Prompt Selection technique and Prompt Size . All factors values on a 5 - point Likert scale ( \u2013 2 = \u201cStrongly Disagree\u201d to 2 = \u201cStrongly Agree\u201d ) . Dotted lines indicate extremely significant p < . 0001 comparisons , otherwise very significant with p - value stated ; solid lines indicate no significance at p > . 01 . Error bars indicate 90 % confidence interval . and ideation minimum pairwise distance decreases by 0 . 0056 ( 13 % of SD ) . This suggests that prompting with irrelevant phrases slightly reduced diversity , since users had to have to conceive their own inspiration ; e . g . , P165 in Directed ( 1 ) \u201ccouldn\u2019t make sense of the given messages , so I tried my best to make something somewhat mo - tivational and correct from them . \u201d . Prompt understandability and quality did not influence ideation individual diversity ( p = n . s . ) . In summary , selecting and presenting computationally diverse and less redundant prompts increased the likelihood of crowdworkers ideating messages that are more computationally diverse and less redundant . 5 . 2 . 6 Results on Evaluating Individual , Collective Objective , The - matic Ideation Diversity ( RQ3 ) . Having shown the mediating effects of diverse prompts , we now evaluate how prompt selection tech - niques affect self - assessed creativity ratings , objective diversity metrics of ideations , and thematically coded diversity metrics of ideations . To carefully distinguish between the commonly used mean pairwise distance with the less used minimum pairwise dis - tance , we performed our analyses on them separately . We calculated one measurement of each collective diversity metric in Table 3 for all messages in each prompt selection condition , and computed uncertainty estimations from synthesized 50 bootstrap samples 12 to generate 50 readings of each diversity metric . We performed factor analyses on the metrics as described in Section 1 . 9 , and per - formed statistical analyses on these factors as described in Appendix Table 22 . Analyses on both individual diversity and collective diver - sity measures had congruent results ( Figure 7 ) , though results for collective diversity had more significant differences ( p < . 001 ) . For collective diversity , our factor analysis found that Ideation Disper - sion was most correlated with mean pairwise distance , and Ideation Evenness with entropy and mean of Chamfer distance . Directed ( 3 ) 12 For each dataset , randomly sample with replacement from the original dataset until the same dataset size is reached . improved Ideation Dispersion from None , while Random reduced Dispersion ( even more for 3 vs . 1 phrases ) . Directed prompts im - proved Ideation Evenness more than Random with respect to None . There was no significant difference for self - assessed Ideation Qual - ity ( p = n . s . , Table 23a in Appendix ) . The previous ideation diversity metrics were all computational . We next assess diversity with human judgement based on the - matic analysis . To conserve manpower to evaluate ideations , we limited thematic coding and crowdworker validation to ideations from three conditions of prompts with 1 phrase , i . e . , None , Ran - dom ( 1 ) , and Directed ( 1 ) . From the results of computational met - rics , we expect bigger differences between Directed ( 3 ) and Ran - dom ( 3 ) for this analysis too . From our thematic analysis , we coded 239 categories 13 which we consolidated to 53 themes ( see Table 19 in Appendix ) . Figure 8 shows results from our statistical anal - ysis . We found that ideations generated with Directed prompts had higher Flexibility and Originality in categories and themes than with Random or None . Ideations from Random prompts mostly had higher Flexibility and Originality compared to None , but the theme Originality was significantly lower . This could be because Random prompts primed ideators to fixate on fewer broad ideas ( themes ) , instead of the higher number of fine - grained idea categories . In summary , despite lower ideation ease and understandabil - ity with Directed prompts ( Section 2 . 2 . 4 ) , we found objective and thematic evidence that Directed Diversity improved ideation di - versity compared to Random and None . Next , we describe how crowdworkers would rate these ideations . 13 Example categories ( in themes ) : Pull - ups ( Exercise Suggestion ) , Strong immune system ( Health Benefits ) , Set daily exercise goal ( Goals ) . See Appendix 8 . 7 for full list of categories and themes . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Figure 7 : Results of computed individual and collective diversity from ideations for different prompt configurations . See Figure 6 caption for how to interpret charts . Figure 8 : Results of diversity in categories and themes derived from thematic analysis of ideations . 5 . 3 Validation User Studies The third and fourth studies employed third - party crowdworkers to assess the creativity of ideated messages from the Ideation User Study , to answer ( RQ3 ) How do prompt selection techniques affect diversity in ideations ? This provides a less biased validation than asking ideators to self - assess . We conducted three experiments with different questioning format to strengthen the experiment design . Appendix Figures 20 - 24 details the questionnaires . 5 . 3 . 1 Individual Validation : Experiment Treatment and Procedure . For the individual validation study , we conducted a within - subjects experiment with prompt selection technique ( None , Random , Di - rected ) as independent variable , and controlled prompt size ( \u0434 = 1 ) . Each participant assessed 25 ideation messages chosen randomly from the three conditions . Participants went through the same pro - cedure as in the Ideation user study , but with a different task in step 3 : 3 . Assess 25 messages regarding how well they motivate for physical activity . For each message , a ) Read a randomly chosen message . b ) Rate on a 7 - point Likert scale , whether the message is moti - vating ( effective ) , informative , and helpful ( as described in Section 4 . 2 . 4 ) . c ) Reflect and write the rationale in free text on why they rated the message as effective or ineffective . This was only asked randomly two out of 25 times , to avoid fatigue . As we discuss later , we found that participants confounded the three ratings questions and answered them very similarly ( responses were highly correlated ) , thus , we designed collective validation user studies to pose different questions and distinguish between the measures . 5 . 3 . 2 Collective Ranking Validation : Experiment Treatment and Pro - cedure . The collective validation study had the same experiment design as before , but different procedure step 3 : 1 . Complete 5 trials to rate collections of ideation messages , where for each trial , a ) Study three groups of 5 messages each ( 3 \u00d7 5 messages ) to b ) Rank message groups as most , middle or least motivating , informative , and unrepetitive ( Section 4 . 2 . 4 ) . Instead of rating messages individually , participants viewed grouped messages from each condition side - by - side and answered ranking questions . Messages in each group were selected from those ideated with the same prompt selection technique . By asking participants to assess collections rather than individual messages , we explicitly measured perceived diversity , since the user perceived the differences between all ideations in the collection ; this is more direct than asking them about the \u201cinformativeness\u201d of an ideation , since this could be confounded with \u201chelpfulness\u201d , \u201cteaching some - thing new\u201d , \u201ctelling something different from other messages\u201d , etc . This approach differs from the triplet similarity comparison [ 55 , 91 ] employed by Siangliulue et al . [ 79 ] , and benefits from requiring fewer assessments . We asked participants to rank groups rather than rate them relatively to obtain a forced choice [ 25 ] . Another method to assess diversity involves longitudinal exposure ( e . g . , [ 50 ] ) , but this is expensive and difficult to scale . 5 . 3 . 3 Collective Pairwise Rating Validation : Experiment Treatment and Procedure . The collective pairwise rating validation study fur - ther validates our results with an existing , commonly used measure to rate the difference between pairs of messages , both from the same prompt selection technique [ 27 , 79 ] . We randomly selected 200 message - pairs from None , Random ( 1 ) and Directed ( 1 ) , yielding CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Figure 9 : Results of perceived individual and collective creativity from the three validation user studies . a pool of 600 message - pairs . All steps in the procedure are identical as before except for Step 3 : 3 . Rate 30 message - pairs randomly selected from the message - pair pool , where for each message - pair , a ) Read the two messages b ) Rate their difference on a 7 - point Likert scale : 1 \u201cNot at all different ( identical ) \u201d to 7 \u201cVery different\u201d This complements the previous study by having participants focus on two messages to compare , which is more manageable than assessing 5 messages , but is limited to a less holistic impression on multiple messages . 5 . 3 . 4 Experiments Data Collection and Statistical Analysis . For all validation studies , we recruited participants from Amazon Mechan - ical Turk with the same high qualification as the ideation study . Of 348 workers who attempted the surveys , 290 passed the screening tests to complete the surveys ( 83 . 3 % pass rate ) . They were 50 . 2 % female , between 22 and 71 years old ( M = 38 . 1 ) ; 67 . 5 % of participants have use fitness apps . For the individual validation study , Partici - pants completed the survey in median time 14 . 7 minutes and were compensated US $ 1 . 50 ; for the collective ranking validation study , participants completed the survey in median time 12 . 7 minutes and were compensated US $ 1 . 80 ; for the collective pairwise rating validation study , participants completed the survey in median time 8 . 4 minutes and were compensated US $ 1 . 00 . In total , 740 messages were individually rated 3 , 375 times ( M = 4 . 56x per message ) , 450 message groups were ranked 1 , 350 times ( M = 3 . 00x per message group ) , and 600 message pairs were rated 2 , 430 times ( M = 4 . 05x per message pair ) . To assess inter - rater agreement , we calculated the average aggregate - judge correlations [ 18 ] as r = . 59 , . 62 , . 63 for mo - tivation , informativeness and helpfulness for individual validation ratings , respectively ; these were comparable to Chan et al . \u2019s r = . 64 for idea novelty [ 18 ] . We performed the same statistical analyses as in the Ideation User Study ( see Section 2 . 2 . 3 ) , report the linear mixed effects models in Appendix Table 23 , and include participant quotes from their rationale text response where relevant . For the collective ranking validation study , we counted how often each Prompt Selection technique was ranked first or last across the 5 trials , performed factor analyses on the counts for best and worst ranks for the three metrics ( motivating , informative , unrepetitive ) to derive three orthogonal factors ( Ideations Unrepetitive , Ideations Informative , Ideations Motivating ) , and performed the statistical analysis on the factors ( see Table 23b in Appendix ) . 5 . 3 . 5 Results on Evaluating Individual and Collective Ideation Cre - ativity ( RQ3 ) . We investigated whether Directed prompts stimulate the highest ideation diversity and whether 3 rd - party validations agree with our computed and thematic results . For illustration , Ap - pendix Table 25 shows examples of message - groups with high and low factor values . Figure 9 shows results of our statistical analysis . We found that ideations from Directed prompts were most different and least repet - itive , ideations from Random were no different and as repetitive as None . Ideations generated with prompts were more informative and helpful than without prompts , but there was no difference whether the prompts were Directed or Random . For example , P4 reviewed the message \u201cExercise and live longer , and prosper more ! \u201d ideated with None , and felt that \u201cit\u2019s basically telling you what you already know . It\u2019s a rather generic message . \u201d ; P63 reviewed the message \u201cWaking up early and working out will help you get into shape , and is a great way to have more energy and better sleep . \u201d from the Directed ( 1 ) prompt \u201cinto a habit of sleep\u201d and felt \u201cit\u2019s effec - tive because it gives me a goal and tells me why this is a good goal\u201d . There were no significant differences in ideation quality or moti - vation , though there was a marginal effect that Random prompts could hurt quality compared to None . Therefore , Directed Diversity helped to reduce ideation redundancy compared to Randomly se - lected prompts , improved informativeness , and did not compromise quality . 5 . 4 Summary of Answers to Research Questions We summarize our findings to answer our research questions with results from multiple experiments . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan RQ1 . How did prompt selection techniques affect diversity in prompts ? Compared to Random , Directed Diversity : a ) selected more diverse prompts , b ) with less redundancy from prior ideation , c ) that ideators perceived as more unexpectedness , but d ) of poorer quality and understandability . RQ2 . How did diversity in prompts affect the ideation process for ideators ? Compared to Random , prompts selected with Directed Diversity were : a ) harder to ideate with , b ) less applied for ideation , c ) but their higher prompt diversity somewhat drove higher ideation diversity . RQ3 . How did prompt selection techniques affect diversity in ideations ? Compared to None and Random , Directed Diversity : a ) improved ideation diversity and reduced redundancy , b ) increased the flexibility and originality of ideated categories , c ) without com - promising ideation quality . 6 DISCUSSION We discuss the generalization of our technical approach , evaluation framework , and experiment findings . 6 . 1 Need for Sensitive and Mechanistic Measures of Creativity We have developed an extensive evaluation framework for two key reasons : 1 ) to precisely detect effects on diversity , and 2 ) to track the mechanism of diversity prompting . We have sought to be very diverse in our evaluation of prompt technique to carefully identify any benefits or issues . We have found that some popu - lar metrics ( e . g . , mean pairwise distance ) were less sensitive than others ( e . g . , MST Dispersion / Remote - tree ) . Therefore , a null re - sult in one metric ( e . g . , [ 79 ] ) may not mean that diversity was not changed ( if measured by another metric ) . Instead of only depend - ing on the \u201cblack box\u201d experimentation of prompt treatment on ideation ( e . g . , [ 18 , 40 , 79 , 80 ] ) , investigating along the ideation chain is interpretable and helpful for us to identify potential issues or breakdowns in the diversity prompting mechanism . Had our evalu - ation results on ideation diversity been non - significant , this would be helpful to debug the lack of effectiveness . Conversely , we may find that an ideation diversity effect may be due to contradictory or confounding effects . Indeed , we found that Directed Diversity improved diversity , despite poorer prompt understandability and adoption . Ideators could not directly use the selected prompts , but still managed to conceive ideas that were more diverse than not having seen prompts or seeing random ones . This suggests that they generated ideas sufficiently near the prompts . The findings also suggested that the increased effort helped to im - prove diverse ideation [ 5 , 6 , 96 ] , but the ideator user experience should be improved . Future work is needed to improve Directed Diversity to reduce ideator effort and improve the relevance of selected prompts , such as by limiting the distance of new prompts from prior ideations , or using idea - based embeddings [ 79 , 80 ] in - stead of language models , as discussed next . 6 . 2 Generalization of Directed Diversity to other Domains The full process of Directed Diversity ( Figure 1 ) allows us to gen - eralize its usage to other domains , such as text creativity tasks beyond motivational messages ( e . g . , birthday greetings [ 79 ] ) by changing the document sources in the phrase extraction step . In the phrase embedding step , we used the Universal Sentence En - coder [ 14 ] , but other text embedding models ( e . g . , word2vec [ 63 ] , GloVe [ 71 ] , ELMo [ 74 ] , BERT [ 23 ] ) could be used that model lan - guages slightly differently . In the third step , we selected phrases based on the Remote - tree diversity formulation using an efficient greedy algorithm that approximates the diversity maximization . Other diversity criteria and maximization algorithms could be used ( see review [ 20 ] ) . Note that since USE and similar language mod - els are domain - independent , which do not model the semantics of specific domains and semantic quality , Directed Diversity cannot guarantee improving quality . A domain - specific model trained with human - annotated labels of quality could be used to improve both diversity and quality . Furthermore , instead of representing text with language models , the idea space could be explicitly modelled to obtain embeddings from annotated semantic similarity [ 55 , 79 ] . Finally , since Directed Diversity operates on a vector representation of prompting and ideations , it can also be used for ideation tasks beyond text as long as they can be represented in a feature vector by feature engineering or with deep learning approaches , such as furniture [ 58 ] , mood boards [ 49 ] , and emojis [ 101 ] . 6 . 3 Generalization of Evaluation Framework Our Evaluation Framework is a first step towards the goal of stan - dardizing the evaluation of crowd ideation . This requires further validation and demonstration on existing methods of supporting crowd ideation . Due to the costs of engineering effort , set - up prepa - ration , and recruitment , we defer it to future work . Just as the Directed Diversity pipeline is generalizable , we discuss how the Diversity Prompt Evaluation Framework is generalizable . We had identified many diversity metrics , but only measured some of them ; see [ 20 ] for a review of other mathematical metrics . If applying the framework to non - text domains , the vector - based distance metrics should still be usable if the concepts can be embedded with a do - main model . While we analyzed diversity in terms of mathematical metrics [ 20 ] and several measures for creativity [ 92 ] , other criteria may be important to optimize , such as serendipity for recommender systems to avoid boredom [ 44 ] . To measure creativity , just as in prior research [ 50 ] , we had used several Likert scale ratings ( e . g . , helpfulness and informativeness ) and found evidence that participants confound them . Furthermore , it may be excessive to apply all our measures , therefore the re - searcher is advised to use them judiciously . For example , we found that individually rating ideations tends to lead to poor statistical significance , so this data collection method should be avoided . The thematic analysis coding is also very labor intensive for the re - search team , but provides rich insights into the ideas generated . We had proposed using ranking and pairwise rating validations of collections of ideations as a scalable way to measure collective diversity . While our evaluations based on generating motivational mes - saging for physical activity helped to provide a realistic context , it was limited to measuring preliminary impressions of validators . The social desirability effect may have limited how accurately par - ticipants rated the effectiveness of the messages . While our focus CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . was on evaluating diversity , future work that also seeks to improve and evaluate motivation towards behavior change should conduct longitudinal trials with stronger ecological validity [ 50 ] . 7 CONCLUSION In this paper , we presented Directed Diversity to direct ideators to generate more collectively creative ideas . This is a generalizable pipeline to extract prompts , embed prompts using a language model , and select maximally diverse prompts . We further proposed a gener - alizable Diversity Prompting Evaluation Framework to sensitively evaluate how Directed Diversity improves ideation diversity along the ideation chain \u2014 prompt selection , prompt creativity , prompt - ideation mediation , and ideation creativity . We found that Directed Diversity improved collective ideation diversity and reduce redun - dancy . With the generalizable prompt selection mechanism and evaluation framework , our work provides a basis for further devel - opment and evaluations of prompt diversity mechanisms . ACKNOWLEDGMENTS This work was carried out in part at NUS Institute for Health In - novation and Technology ( iHealthtech ) and with funding support from the NUS ODPRT and Ministry of Education , Singapore . REFERENCES [ 1 ] Leonard Adelman , James Gualtieri , and Suzanne Stanford . 1995 . Examining the effect of causal focus on the option generation process : An experiment using protocol analysis . Organizational Behavior and Human Decision Processes . https : / / doi . org / 10 . 1006 / obhd . 1995 . 1005 [ 2 ] Elena Agapie , Bonnie Chinh , Laura R Pina , Diana Oviedo , Molly C Welsh , Gary Hsieh , and Sean Munson . 2018 . Crowdsourcing Exercise Plans Aligned with Expert Guidelines and Everyday Constraints . In CHI 2018 , 324 . https : / / doi . org / 10 . 1145 / 3173574 . 3173898 [ 3 ] Elena Agapie , Lucas Colusso , Sean A Munson , and Gary Hsieh . 2016 . PlanSourc - ing : Generating Behavior Change Plans with Friends and Crowds . In CSCW 2016 , 119 \u2013 133 . https : / / doi . org / 10 . 1145 / 2818048 . 2819943 [ 4 ] Faez Ahmed , Sharath Kumar Ramachandran , Mark Fuge , Samuel Hunter , and Scarlett Miller . 2019 . Interpreting Idea Maps : Pairwise comparisons reveal what makes ideas novel . Journal of Mechanical Design 141 , 2 . https : / / doi . org / 10 . 1115 / 1 . 4041856 [ 5 ] Baptiste Barbot . 2018 . The dynamics of creative ideation : Introducing a new assessment paradigm . Frontiers in Psychology 9 , DEC : 1 \u2013 8 . https : / / doi . org / 10 . 3389 / fpsyg . 2018 . 02529 [ 6 ] Roger E . Beaty and Paul J . Silvia . 2012 . Why do ideas get more creative across time ? An executive interpretation of the serial order effect in divergent thinking tasks . Psychology of Aesthetics , Creativity , and the Arts 6 , 4 : 309 \u2013 319 . https : / / doi . org / 10 . 1037 / a0029171 [ 7 ] Michael S Bernstein , Greg Little , Robert C Miller , Bj\u00f6rn Hartmann , Mark S Ackerman , David R Karger , David Crowell , and Katrina Panovich . 2010 . Soylent : A Word Processor with a Crowd Inside . In Proceedings of the 23nd Annual ACM Symposium on User Interface Software and Technology , 313 \u2013 322 . https : / / doi . org / 10 . 1145 / 1866029 . 1866078 [ 8 ] H . Beyer and K Holtzblatt . 1998 . Contextual design : defining customer - centered systems . Morgan Kaufmann . [ 9 ] JeffreyP . Bigham , MichaelS . Bernstein , andEytanAdar . 2015 . Human - Computer Interaction and Collective Intelligence . In Handbook of Collective Intelligence . [ 10 ] Osvald M Bjelland and Robert Chapman Wood . 2008 . An Inside View of IBM\u2019s \u201cInnovation Jam . \u201d MIT Sloan management review 50 , 1 : 32 . [ 11 ] Jonathan Bragg , Mausam , and Daniel S . Weld . 2016 . Optimal testing for crowd workers . In Proceedings of the International Joint Conference on Autonomous Agents and Multiagent Systems , AAMAS . [ 12 ] Nathan Brown , Stavros Tseranidis , and Caitlin Mueller . 2015 . Multi - objective optimization for diversity and performance in conceptual structural design . In Proceedings of IASS Annual Symposia , IASS 2015 Amsterdam Symposium : Future Visions \u2013 Computational Design , 1 \u2013 12 . [ 13 ] Andrew Carnie . 2010 . Constituent structure . Oxford University Press . [ 14 ] Daniel Cer , Yinfei Yang , Sheng yi Kong , Nan Hua , Nicole Limtiaco , Rhomni St . John , Noah Constant , Mario Guajardo - C\u00e9spedes , Steve Yuan , Chris Tar , Yun Hsuan Sung , Brian Strope , and Ray Kurzweil . 2018 . Universal sentence encoder forEnglish . In Proceedingsofthe2018ConferenceonEmpiricalMethodsinNatural Language Processing ( System Demonstrations ) , 169 \u2013 174 . https : / / doi . org / 10 . 18653 / v1 / d18 - 2029 [ 15 ] Joel Chan , Steven Dang , and Steven P Dow . 2016 . Comparing Different Sense - making Approaches for Large - Scale Ideation . In CHI 2016 . https : / / doi . org / 10 . 1145 / 2858036 . 2858178 [ 16 ] Joel Chan , Steven P . Dow , and Christian D . Schunn . 2018 . Do the Best De - sign Ideas ( Really ) Come from Conceptually Distant Sources of Inspiration ? In Engineering a Better Future . 111 \u2013 139 . [ 17 ] Joel Chan and Christian Schunn . 2015 . The impact of analogies on creative con - cept generation : Lessons from an in vivo study in engineering design . Cognitive Science 39 , 1 : 126 \u2013 155 . https : / / doi . org / 10 . 1111 / cogs . 12127 [ 18 ] Joel Chan , Pao Siangliulue , Denisa Qori McDonald , Ruixue Liu , Reza Moradinezhad , SafaAman , ErinTSolovey , KrzysztofZGajos , andStevenPDow . 2017 . Semanticallyfarinspirationsconsideredharmful ? accountingforcognitive statesincollaborativeideation . In Proceedingsofthe2017ACMSIGCHIConference on Creativity and Cognition , 93 \u2013 105 . https : / / doi . org / 10 . 1145 / 3059454 . 3059455 [ 19 ] Jesse Chandler , Cheskie Rosenzweig , Aaron J . Moss , Jonathan Robinson , and Leib Litman . 2019 . Online panels in social science research : Expanding sampling methods beyond Mechanical Turk . Behavior Research Methods 51 , 5 : 2022 \u2013 2038 . https : / / doi . org / 10 . 3758 / s13428 - 019 - 01273 - 7 [ 20 ] Barun Chandra and Magn\u00fas M . Halld\u00f3rsson . 2001 . Approximation Algorithms for Dispersion Problems . Journal of Algorithms 38 , 2 : 438 \u2013 465 . https : / / doi . org / 10 . 1006 / jagm . 2000 . 1145 [ 21 ] Elizabeth Clark , Anne Spencer Ross , Chenhao Tan , Yangfeng Ji , and Noah A Smith . 2018 . Creative writing with a machine in the loop : Case studies on slogans and stories . In 23rd International Conference on Intelligent User Interfaces , 329 \u2013 340 . https : / / doi . org / 10 . 1145 / 3172944 . 3172983 [ 22 ] Peng Dai , Christopher H . Lin , Mausam , and Daniel S . Weld . 2013 . POMDP - based control of workflows for crowdsourcing . Artificial Intelligence 202 : 52 \u2013 85 . https : / / doi . org / 10 . 1016 / j . artint . 2013 . 06 . 002 [ 23 ] JacobDevlin , MingWeiChang , KentonLee , andKristinaToutanova . 2018 . BERT : Pre - training of deep bidirectional transformers for language understanding . arXiv preprint : arXiv : 1810 . 04805 . [ 24 ] Sandra Diazz and Marcelo Cabido . 2001 . Vive la diff \u00e9 rence : plant functional diversity matters to ecosystem processes . Trends in Ecology and Evolution 16 , 11 : 646 \u2013 655 . https : / / doi . org / 10 . 1016 / S0169 - 5347 ( 01 ) 02283 - 2 [ 25 ] JohnR . Douceur . 2009 . Paperratingvs . paperranking . OperatingSystemsReview ( ACM ) 43 , 2 : 117 \u2013 121 . https : / / doi . org / 10 . 1145 / 1531793 . 1531816 [ 26 ] Michael R . P . Dougherty , Charles F . Gettys , and Eve E . Ogden . 1999 . MINERVA - DM : A memory processes model for judgments of likelihood . Psychological Review . https : / / doi . org / 10 . 1037 / 0033 - 295X . 106 . 1 . 180 [ 27 ] Steven P . Dow , Alana Glassco , Jonathan Kass , Melissa Schwarz , Daniel L . Schwartz , and Scott R . Klemmer . 2010 . Parallel prototyping leads to better design results , more divergence , and increased self - efficacy . ACM Transactions on Computer - Human Interaction 17 , 4 . https : / / doi . org / 10 . 1145 / 1879831 . 1879836 [ 28 ] Zipei Fan , Xuan Song , and Ryosuke Shibasaki . 2014 . CitySpectrum : A non - negative tensor factorization approach . UbiComp 2014 - Proceedings of the 2014 ACM International Joint Conference on Pervasive and Ubiquitous Computing : 213 \u2013 233 . https : / / doi . org / 10 . 1145 / 2632048 . 2636073 [ 29 ] Daniel M . Fleder and Kartik Hosanagar . 2007 . Recommender systems and their impact on sales diversity . EC\u201907 - Proceedings of the Eighth Annual Conference on Electronic Commerce : 192 \u2013 199 . https : / / doi . org / 10 . 1145 / 1250910 . 1250939 [ 30 ] Nancy A Fontenot . 1993 . Effects of training in creativity and creative problem finding upon business people . The Journal of Social Psychology 133 , 1 : 11 \u2013 22 . [ 31 ] Jonas Frich , Lindsay MacDonald Vermeulen , Christian Remy , Michael Mose Biskjaer , and Peter Dalsgaard . 2019 . Mapping the Landscape of Creativity Sup - port Tools in HCI . In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems - CHI \u201919 , 1 \u2013 18 . https : / / doi . org / 10 . 1145 / 3290605 . 3300619 [ 32 ] Jonas Frich , Michael Mose Biskjaer , and Peter Dalsgaard . 2018 . Twenty Years of Creativity Research in Human - Computer Interaction : Current State and Future Directions . In Proceedings of the 2018 Designing Interactive Systems Conference ( DIS \u201918 ) , 1235 \u2013 1257 . https : / / doi . org / 10 . 1145 / 3196709 . 3196732 [ 33 ] Victor Girotto , Erin Walker , and Winslow Burleson . 2017 . The effect of periph - eral micro - tasks on crowd ideation . In Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems , 1843 \u2013 1854 . https : / / doi . org / 10 . 1145 / 3025453 . 3025464 [ 34 ] Barney G . Glaser and Anselm L . Strauss . 2017 . Discovery of grounded theory : strategies for qualitative research . Routledge . [ 35 ] Vinod Goel . 2010 . Neural basis of thinking : Laboratory problems versus real - world problems . Wiley Interdisciplinary Reviews : Cognitive Science . https : / / doi . org / 10 . 1002 / wcs . 71 [ 36 ] AdamE . Green , DavidJ . M . Kraemer , JonathanA . Fugelsang , JeremyR . Gray , and Kevin N . Dunbar . 2012 . Neural correlates of creativity in analogical reasoning . Journal of Experimental Psychology : Learning Memory and Cognition . https : / / doi . org / 10 . 1037 / a0025764 [ 37 ] Magn\u00fas M . Halld\u00f3rsson , Kazuo Iwano , Naoki Katoh , and Takeshi Tokuyama . 1999 . Findingsubsetsmaximizingminimumstructures . SIAMJournalonDiscrete Mathematics . https : / / doi . org / 10 . 1137 / S0895480196309791 Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan [ 38 ] F He , Y Pan , Q Lin , X Miao , and Z Chen . 2019 . Collective Intelligence : A Taxon - omy and Survey . IEEE Access 7 : 170213 \u2013 170225 . https : / / doi . org / 10 . 1109 / ACCESS . 2019 . 2955677 [ 39 ] Seyedmohammadhossein Hosseinian , Dalila B M M Fontes , Sergiy Butenko , Marco Buongiorno Nardelli , Marco Fornari , and Stefano Curtarolo . 2017 . The MaximumEdgeWeightCliqueProblem : FormulationsandSolutionApproaches . In Optimization Methods and Applications . 217 \u2013 237 . [ 40 ] GaopingHuangandAlexanderJQuinn . 2017 . BlueSky : Crowd - PoweredUniform Sampling of Idea Spaces . In Proceedings of the 2017 ACM SIGCHI Conference on Creativity and Cognition ( C & C \u201917 ) , 119 \u2013 130 . https : / / doi . org / 10 . 1145 / 3059454 . 3059481 [ 41 ] Piotr Indyk , Sepideh Mahabadi , Mohammad Mahdian , and Vahab S . Mirrokni . 2014 . Composablecore - setsfordiversityandcoveragemaximization . Proceedings of the ACM SIGACT - SIGMOD - SIGART Symposium on Principles of Database Systems : 100 \u2013 108 . https : / / doi . org / 10 . 1145 / 2594538 . 2594560 [ 42 ] David G . Jansson and Steven M . Smith . 1991 . Design fixation . Design Studies . https : / / doi . org / 10 . 1016 / 0142 - 694X ( 91 ) 90003 - F [ 43 ] MarkW . Jones , J . AndreasB\u00e6rentzen , andMilosSramek . 2006 . 3Ddistancefields : A survey of techniques and applications . IEEE Transactions on Visualization and Computer Graphics 12 , 4 : 581 \u2013 599 . https : / / doi . org / 10 . 1109 / TVCG . 2006 . 56 [ 44 ] Marius Kaminskas and Derek Bridge . 2016 . Diversity , serendipity , novelty , and coverage : A survey and empirical analysis of beyond - Accuracy objectives in recommender systems . ACM Transactions on Interactive Intelligent Systems 7 , 1 : 1 \u2013 42 . https : / / doi . org / 10 . 1145 / 2926720 [ 45 ] David R . Karger , Sewoong Oh , and Devavrat Shah . 2014 . Budget - optimal task allocation for reliable crowdsourcing systems . Operations Research 62 , 1 : 1 \u2013 24 . https : / / doi . org / 10 . 1287 / opre . 2013 . 1235 [ 46 ] L . Robin Keller and Joanna L . Ho . 1988 . Decision Problem Structuring : Gen - erating Options . IEEE Transactions on Systems , Man and Cybernetics . https : / / doi . org / 10 . 1109 / 21 . 21599 [ 47 ] Aniket Kittur , Jeffrey V . Nickerson , Michael S . Bernstein , Elizabeth M . Ger - ber , Aaron Shaw , John Zimmerman , Matthew Lease , and John J . Horton . 2013 . The future of crowd work . In Proceedings of the ACM Conference on Computer Supported Cooperative Work , CSCW . https : / / doi . org / 10 . 1145 / 2441776 . 2441923 [ 48 ] Ana Cristina Bicharra Klein , Mark , and Garcia . 2015 . High - speed idea filtering with the bag of lemons . Decision Support Systems 78 : 39 \u2013 50 . https : / / doi . org / 10 . 1016 / j . dss . 2015 . 06 . 005 [ 49 ] Janin Koch , Nicolas Taffin , Michel Beaudouin - Lafon , Markku Laine , Andr\u00e9s Lucero , and Wendy E . MacKay . 2020 . ImageSense : An Intelligent Collaborative Ideation Tool to Support Diverse Human - Computer Partnerships . Proceedings of the ACM on Human - Computer Interaction 4 , CSCW1 : 1 \u2013 27 . https : / / doi . org / 10 . 1145 / 3392850 [ 50 ] Rafal Kocielnik and Gary Hsieh . 2017 . Send Me a Different Message : Utilizing Cognitive Space to Create Engaging Message Triggers . In CSCW , 2193 \u2013 2207 . https : / / doi . org / 10 . 1145 / 2998181 . 2998324 [ 51 ] Etienne Lalibert\u00e9 and Pierre Legendre . 2010 . A distance - based framework for measuring functional diversity from multiple traits . Ecology 91 , 1 : 299 \u2013 305 . [ 52 ] Joel Lehman and Kenneth O . Stanley . 2011 . Abandoning objectives : Evolution through the search for novelty alone . Evolutionary Computation 19 , 2 : 189 \u2013 222 . https : / / doi . org / 10 . 1162 / EVCO _ a _ 00025 [ 53 ] Todd I Lubart . 2001 . Models of the creative process : Past , present and future . Creativity research journal 13 , 3 \u2013 4 : 295 \u2013 308 . [ 54 ] Kurt Luther , Nathan Hahn , Steven P Dow , and Aniket Kittur . 2015 . Crowdlines : Supporting synthesis of diverse information sources through crowdsourced outlines . In Third AAAI Conference on Human Computation and Crowdsourcing . [ 55 ] Laurens Van Der Maaten and Kilian Weinberger . 2012 . Stochastic triplet em - bedding . In 2012 IEEE International Workshop on Machine Learning for Signal Processing , 1 \u2013 6 . https : / / doi . org / 10 . 1109 / MLSP . 2012 . 6349720 [ 56 ] Thomas W . Malone , Robert Laubacher , and Chrysanthos N . Dellarocas . 2009 . Harnessing Crowds : Mapping the Genome of Collective Intelligence . https : / / doi . org / 10 . 2139 / ssrn . 1381502 [ 57 ] Manon Marinussen and Alwin de Rooij . 2019 . Being Yourself to Be Creative : How Self - Similar Avatars Can Support the Generation of Original Ideas in Virtual Environments . In Proceedings of the 2019 on Creativity and Cognition ( C & C \u201919 ) , 285 \u2013 293 . https : / / doi . org / 10 . 1145 / 3325480 . 3325482 [ 58 ] Justin Matejka , Michael Glueck , Erin Bradner , Ali Hashemi , Tovi Grossman , and George Fitzmaurice . 2018 . Dream lens : Exploration and visualization of large - scale generative design datasets . Conference on Human Factors in Computing Systems - Proceedings 2018 - April : 1 \u2013 12 . https : / / doi . org / 10 . 1145 / 3173574 . 3173943 [ 59 ] LelandMcInnes , JohnHealy , andJamesMelville . 2018 . UMAP : UniformManifold Approximation and Projection for Dimension Reduction . Retrieved from http : / / arxiv . org / abs / 1802 . 03426 [ 60 ] Cai - nicolas Ziegler Sean M Mcnee , Joseph A Konstan , and Georg Lausen . 2005 . Improving Recommendation Lists Through Topic Diversification . In Proceedings of the 14th international conference on World Wide Web , 22 \u2013 32 . [ 61 ] JokeMeheus . 2000 . AnalogicalReasoninginCreativeProblemSolvingProcesses : Logico - Philosophical Perspectives . In Metaphor and Analogy in the Sciences . https : / / doi . org / 10 . 1007 / 978 - 94 - 015 - 9442 - 4 _ 2 [ 62 ] P . Michelucci and J . L . Dickinson . 2016 . The power of crowds . Science 351 , 6268 : 32 \u2013 33 . https : / / doi . org / 10 . 1126 / science . aad6499 [ 63 ] Tomas Mikolov , Ilya Sutskever , Kai Chen , Greg Corrado , and Jeffrey Dean . 2013 . Distributed Representations of Words and Phrases and their Compositionality . In NIPS\u201913 : Proceedingsofthe26thInternationalConferenceonNeuralInformation Processing Systems , 3111 \u2013 3119 . https : / / doi . org / 10 . 5555 / 2999792 . 2999959 [ 64 ] Fabio Del Missier , Mim\u00ec Visentini , and Timo M\u00e4ntyl\u00e4 . 2015 . Option generation in decision making : Ideation beyond memory retrieval . Frontiers in Psychology . https : / / doi . org / 10 . 3389 / fpsyg . 2014 . 01584 [ 65 ] Caitlin T . Mueller and John A . Ochsendorf . 2015 . Combining structural perfor - mance and designer preferences in evolutionary design space exploration . Au - tomation in Construction 52 : 70 \u2013 82 . https : / / doi . org / 10 . 1016 / j . autcon . 2015 . 02 . 011 [ 66 ] Michael D Mumford , Wayne A Baughman , K Victoria Threlfall , Elizabeth P Supinski , and David P Costanza . 1996 . Process - based measures of creative problem - solving skills : I . Problem construction . Creativity Research Journal 9 , 1 : 63 \u2013 76 . [ 67 ] Bernard A . Nijstad and Wolfgang Stroebe . 2006 . How the Group Affects the Mind : A Cognitive Model of Idea Generation in Groups . Personality and Social Psychology Review 10 , 3 : 186 \u2013 213 . https : / / doi . org / 10 . 1207 / s15327957pspr1003 _ 1 [ 68 ] Bernard A Nijstad , Wolfgang Stroebe , and Hein F M Lodewijkx . 2002 . Cognitive stimulation and interference in groups : Exposure effects in an idea generation task . Journal of Experimental Social Psychology 38 , 6 : 535 \u2013 544 . https : / / doi . org / 10 . 1016 / S0022 - 1031 ( 02 ) 00500 - 0 [ 69 ] Jonas Oppenlaender and Simo Hosio . 2019 . Design Recommendations for Aug - menting Creative Tasks with Computational Priming . In Proceedings of the 18th International Conference on Mobile and Ubiquitous Multimedia ( MUM \u201919 ) . https : / / doi . org / 10 . 1145 / 3365610 . 3365621 [ 70 ] Rebecca Passonneau . 2006 . Measuring agreement on set - valued items ( MASI ) for semantic and pragmatic annotation . Proceedings of the 5th International Conference on Language Resources and Evaluation , LREC 2006 : 831 \u2013 836 . [ 71 ] Jeffrey Pennington , Richard Socher , and Christopher D . Manning . 2014 . GloVe : Global Vectors for Word Representation Jeffrey . In Proceedings ofthe 2014 Confer - ence on Empirical Methods in Natural Language Processing ( EMNLP ) , 1532 \u2013 1543 . https : / / doi . org / 10 . 3115 / v1 / D14 - 1162 [ 72 ] Owen L . Petchey and Kevin J . Gaston . 2002 . Functional diversity ( FD ) , species richness and community composition . Ecology Letters 5 , 3 : 402 \u2013 411 . https : / / doi . org / 10 . 1046 / j . 1461 - 0248 . 2002 . 00339 . x [ 73 ] OwenL . PetcheyandKevinJ . Gaston . 2002 . Extinctionandthelossoffunctional diversity . Proceedings of the Royal Society B : Biological Sciences . https : / / doi . org / 10 . 1098 / rspb . 2002 . 2073 [ 74 ] Matthew E . Peters , Mark Neumann , Mohit Iyyer , Matt Gardner , Christopher Clark , Kenton Lee , and Luke Zettlemoyer . 2018 . Deep contextualized word rep - resentations . In Proceedingsofthe2018ConferenceoftheNorthAmericanChapter of the Association for Computational Linguistics : Human Language Technologies , 2227 \u2013 2237 . https : / / doi . org / 10 . 18653 / v1 / n18 - 1202 [ 75 ] Carlo Ricotta and Laszlo Szeidl . 2006 . Towards a unifying approach to diversity measures : bridging the gap between the Shannon entropy and Rao\u2019s quadratic index . Theoretical population biology 70 , 3 : 237 \u2013 243 . [ 76 ] C . Riedl , I . Blohm , J . M . Leimeister , and H . Krcmar . 2010 . Rating scales for collective intelligence in innovation communities : Why quick and easy deci - sion making does not get it right . In Thirty First International Conference on Information Systems . [ 77 ] SebastianRisi , SandyD . Vanderbleek , CharlesE . Hughes , andKennethO . Stanley . 2009 . How Novelty Search Escapes the Deceptive Trap of Learning to Learn . In GECCO\u201909 . [ 78 ] Pararth Shah , Dilek Hakkani - T\u00fcr , Gokhan T\u00fcr , Abhinav Rastogi , Ankur Bapna , Neha Nayak , and Larry Heck . 2018 . Building a conversational agent overnight with dialogue self - play . arXiv preprint arXiv : 1801 . 04871 . [ 79 ] Pao Siangliulue , Kenneth C Arnold , Krzysztof Z Gajos , and Steven P Dow . 2015 . Toward Collaborative Ideation at Scale : Leveraging Ideas from Others to Generate More Creative and Diverse Ideas . In Proceedings of the 18th ACM ConferenceonComputerSupportedCooperativeWork & SocialComputing ( CSCW \u201915 ) , 937 \u2013 945 . https : / / doi . org / 10 . 1145 / 2675133 . 2675239 [ 80 ] Pao Siangliulue , Joel Chan , Steven P Dow , and Krzysztof Z Gajos . 2016 . Idea - Hound : Improving Large - scale Collaborative Ideation with Crowd - Powered Real - time Semantic Modeling . In UIST 2016 - Proceedings of the 29th Annual Symposium on User Interface Software and Technology ( UIST \u201916 ) , 609 \u2013 624 . https : / / doi . org / 10 . 1145 / 2984511 . 2984578 [ 81 ] Pao Siangliulue , Joel Chan , Krzysztof Z Gajos , and Steven P Dow . 2015 . Pro - viding timely examples improves the quantity and quality of generated ideas . In Proceedings of the 2015 ACM SIGCHI Conference on Creativity and Cognition , 83 \u2013 92 . https : / / doi . org / 10 . 1145 / 2757226 . 2757230 [ 82 ] R . Sibson . 1973 . Slink : an optimally efficient algorithm for the single - link cluster method . The Computer Journal 16 , 1 : 30 \u2013 34 . [ 83 ] UtNaSio , KennethKotovsky , andJonathanCagan . 2015 . Fixationorinspiration ? A meta - analytic review of the role of examples on design processes . Design Studies 39 : 70 \u2013 99 . https : / / doi . org / 10 . 1016 / j . destud . 2015 . 04 . 004 CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . [ 84 ] Steven M . Smith . 2010 . The Constraining Effects of Initial Ideas . In Group Creativity : Innovation through Collaboration . https : / / doi . org / 10 . 1093 / acprof : oso / 9780195147308 . 003 . 0002 [ 85 ] Paul T Sowden , Lucie Clements , Chrishelle Redlich , and Carine Lewis . 2015 . Improvisation facilitates divergent thinking and creativity : Realizing a benefit of primary school arts education . Psychology of Aesthetics , Creativity , and the Arts 9 , 2 : 128 . https : / / doi . org / 10 . 1037 / aca0000018 [ 86 ] Ian F . Spellerberg and Peter J . Fedor . 2003 . A tribute to Claude - Shannon ( 1916 - 2001 ) and a plea for more rigorous use of species richness , species diversity and the \u201cShannon - Wiener\u201d Index . Global Ecology and Biogeography 12 , 3 : 177 \u2013 179 . https : / / doi . org / 10 . 1046 / j . 1466 - 822X . 2003 . 00015 . x [ 87 ] Andy Stirling . 2007 . A general framework for analysing diversity in science , technology and society . Journal of the Royal Society Interface 4 : 707 \u2013 719 . https : / / doi . org / 10 . 1098 / rsif . 2007 . 0213 [ 88 ] Victor J . Strecher , Saul Shiffman , and Robert West . 2005 . Randomized con - trolled trial of a web - based computer - tailored smoking cessation program as a supplement to nicotine patch therapy . Addiction 100 , 5 : 682 \u2013 688 . https : / / doi . org / 10 . 1111 / j . 1360 - 0443 . 2005 . 01093 . x [ 89 ] Shweta Suran , Vishwajeet Pattanaik , and Dirk Draheim . 2020 . Frameworks for collective intelligence : A systematic literature review . ACM Computing Surveys 53 , 1 : 1 \u2013 36 . https : / / doi . org / 10 . 1145 / 3368986 [ 90 ] Simon Taggar . 2002 . INDIVIDUAL CREATIVITY AND GROUP ABILITY TO UTILIZE INDIVIDUAL CREATIVE RESOURCES : A MULTILEVEL MODEL . Academy of Management Journal 45 , 2 : 315 \u2013 330 . [ 91 ] Omer Tamuz , Ce Liu , Serge Belongie , Ohad Shamir , and Adam Tauman Kalai . 2011 . Adaptively learning the crowd kernel . arXiv preprint arXiv : 1105 . 1033 . [ 92 ] E Paul Torrance . 2018 . Guiding creative talent . Pickle Partners Publishing . [ 93 ] Sa\u00fal Vargas , Linas Baltrunas , Alexandros Karatzoglou , and Pablo Castells . 2014 . Coverage , redundancy and size - awareness in genre diversity for recommender systems . RecSys 2014 - Proceedings of the 8th ACM Conference on Recommender Systems : 209 \u2013 216 . https : / / doi . org / 10 . 1145 / 2645710 . 2645743 [ 94 ] Sebastien Villeger , Norman W . H . Mason , and David Mouillot . 2008 . New Mul - tidimensional Functional Diversity Indices for a Multifaceted Framework in Functional Ecology . Ecology 89 , 8 : 2290 \u2013 2301 . https : / / doi . org / 10 . 1890 / 07 - 1206 . 1 [ 95 ] Roelof A J de Vries , Khiet P Truong , Sigrid Kwint , Constance H C Drossaert , and Vanessa Evers . 2016 . Crowd - Designed Motivation : Motivational Messages for Exercise Adherence Based on Behavior Change Theory Roelof . In CHI 2016 , 297 \u2013 308 . https : / / doi . org / 10 . 1145 / 2858036 . 2858229 [ 96 ] Meijuan Wang , Ning Hao , Yixuan Ku , Roland H . Grabner , and Andreas Fink . 2017 . Neural correlates of serial order effect in verbal divergent thinking . Neu - ropsychologia 99 : 92 \u2013 100 . https : / / doi . org / 10 . 1016 / j . neuropsychologia . 2017 . 03 . 001 [ 97 ] Daniel S Weld , Christopher H Lin , and Jonathan Bragg . 2015 . Artificial Intelli - gence and Collective Intelligence . In Handbook of Collective Intelligence . [ 98 ] Tom Young , Devamanyu Hazarika , Soujanya Poria , and Erik Cambria . 2018 . Recent trends in deep learning based natural language processing . IEEE Com - putational Intelligence Magazine 13 , 3 : 55 \u2013 75 . https : / / doi . org / 10 . 1109 / MCI . 2018 . 2840738 [ 99 ] Lixiu Yu and Jeffrey V . Nickerson . 2011 . Cooks or cobblers ? Crowd Creativity through Combination . 1393 . https : / / doi . org / 10 . 1145 / 1978942 . 1979147 [ 100 ] F Zenasni and T I Lubart . 2009 . Perception of emotion , alexithymia and creative potential . Personality and Individual Differences 46 , 3 : 353 \u2013 358 . https : / / doi . org / 10 . 1016 / j . paid . 2008 . 10 . 030 [ 101 ] Xianda Zhou and William Yang Wang . 2018 . MojiTalk : Generating Emotional Responses at Scale . In Proceedings ofthe 56th Annual Meeting ofthe Association for Computational Linguistics , 1128 \u2013 1137 . https : / / doi . org / 10 . 18653 / v1 / P18 - 1104 A DEFINITIONS OF PROMPT SELECTION VARIABLES Table 6 : Independent variables used in the simulation and user studies to manipulate how prompts are shown to ideators . Variable Definition Interpretation PromptSelection None : no prompt , other than task instructions Random : randomly selected phrase from corpus Directed : prioritized phrase from corpus Selection algorithm for selecting phrases to include in prompts . PromptCount Number of prompts { 50 , 100 , 150 , 200 , 250 , . . . , n prompts } Indicates how many prompts shown to generate new messages . A prompt may contain \u2265 1 phrases . This was only tested in the simulation study . PromptSize Number of phrases per prompt { 1 , 2 , 3 , 4 , 5 } Prompts selected depends on Prompt Selection . B ADDITIONAL DEFINITIONS OF DIVERSITY METRICS B . 1 Thematic Analysis Method for Flexibility and Originality Metrics Flexibility [ 85 ] measures how many unique ideas ( conceptual cat - egories ) was generated , and originality [ 100 ] measures how infre - quently each conceptual category occurs . These require expert annotation to identify distinct categories . We conducted a thematic analysis of ideated messages using open coding of grounded theory [ 34 ] to derive categories . These categories were added , reduced , merged , and refined by iteratively assessing the messages . We then consolidated the categories into themes using affinity diagram - ming [ 8 ] . This was done separately for different prompt techniques . The thematic analysis was primarily performed by one co - author researcher with regular discussion with co - authors who are ex - perienced HCI researchers with experience in Amazon Mechan - ical Turk experiments and research on health behavior change . We calculated inter - rater reliability on a random 10 % subset of messages was coded independently by another co - author to ob - tain a Krippendorff\u2019s alpha with MASI distance [ 70 ] of \u03b1 = 0 . 82 , which indicated good agreement . Note that while thematic anal - yses and affinity diagramming are popular methods to interpret qualitative data , we use them here for data pre - processing . Finally , we calculate the flexibility and originality measures based on the coded categories ( fine - grained ) and themes ( coarser ) described in Table 7 . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Table 7 : Metrics of creativity of ideation based on categories and themes derived from a thematic analysis of generated ideas . Metrics are shown for categories , but are the same for themes . Metric Definition Interpretation MessagesFlexibility Number of categories coded (cid:205) c [ f c > 0 ] This counts how many unique categories / themes were observed in messages for each Prompt Technique . A higher count indicates qualitatively more diversity . MessagesOriginality Category originality o c = ( 1 \u2212 f c / N p ) How original o c each theme is , where f c is the frequency for the category c , N p is the number of messages with the Prompt Technique p . B . 2 Intra - Prompt Diversity Metrics based on Embedding Distances Table 8 : Metrics of prompt diversity for all phrases in a single prompt . Metric Definition Interpretation Intra - Prompt MeanPhrase Distance Intra - prompt mean phrase - phrase distance 1 \u0434 (cid:205) i , j \u2208 Prompt d ( x Pi , x Pj ) Indicates how similar ( consistent ) all phrases are to one another in the same prompt . Prompts with better consistency would be easier to understand and use . Prompt Phrase Chamfer distance 1 \u0434 (cid:205) i \u2208 Prompt min j (cid:60) i d ( x Pi , x Pj ) Average distinctiveness of phrases in prompt . C DEFINITIONS OF PROMPT ADOPTION METRICS Table 9 : Metrics indicating how much of prompt text and concepts are adopted into the ideations . Metric Definition Interpretation Prompt Recall 1 \u0434 (cid:205) Phrase \u2208 Prompt n word \u2208 Ideation \u2227 word \u2208 Phrase n word \u2208 Phrase The proportion of words from phrases that were used in the ideated message . Prompt Precision (cid:205) phrase \u2208 prompt n word \u2208 Ideation \u2227 word \u2208 Phrase n word \u2208 Ideation The proportion of ideated message words that were from phrases in the shown prompt . Prompt - Ideation Distance Prompt - Ideation distance d ( x Pri , x Ij ) Indicates how similar the written ideation message is to the prompt , as a measure of how the phrase ( s ) ideas were adopted . D PAIRWISE EMBEDDING DISTANCES OF PHRASES AND MESSAGES These figures show the distribution of pairwise distances based on the embeddings of phrases and messages . Figure 10 : Distribution of pairwise distances between the extracted phrases ( N = 3 , 666 ) . The pairwise distances ranged from Min = 0 . 057 to Max = 0 . 586 , Median = 0 . 430 , inter - quartile range 0 . 394 to 0 . 460 , SD = 0 . 047 . Figure 11 : Distribution of pairwise distances between the messages ( N = 250 ) ideated in the pilot study with no prompt - ing ( None ) . The pairwise distances ranged from Min = 0 . 169 to Max = 0 . 549 , Median = 0 . 405 , inter - quartile range 0 . 376 to 0 . 432 , SD = 0 . 043 . CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . E RESULTS OF CHARACTERIZATION SIMULATION STUDY We created 50 simulations for each prompts configuration to get a statistical estimate of the performance of each prompt selection technique . Figure 12 shows the results from the simulation study . Error bars are extremely small , and not shown for simplicity . Span and Sparseness results not shown , but are similar to Mean Dis - tance . Note that we computed the mean of MST edge distances instead of sum , which is independent of number of prompts . In general , Directed Diversity selects prompts to be more diverse for fewer prompts ( smaller prompt count ) , but after a threshold , Ran - dom selection can provide for better diversity . This demonstrates directing is useful for small crowd budgets . Note that the actual threshold depends on corpus and application domain . We found an interaction effect where single - phrase prompts benefit most with Directed Diversity , since for low prompt count , and Directed ( 1 - phrase ) has highest diversity , followed by Directed ( 3 ) , Random ( 3 ) , and Random ( 1 ) with lowest diversity . Figure 12 : Influence of prompt selection technique , prompt size , and prompt count on various distance and diversity metrics . Higher values for all metrics indicate higher diversity . Span and Sparseness results are not shown , but are similar to Mean Distance . Note that we computed the mean of MST edge distances instead of sum , which is independent of number of prompts . Error bars are extremely small , and not shown for simplicity . F FACTOR LOADINGS FROM FACTOR ANALYSIS IN USER STUDIES Table 10 : The rotated factor loading of factor analysis on metrics of prompt distance and consistency . Factors explained 73 . 6 % of the total variance . Bartlett\u2019s Test for Sphericity to indicate common factors was significant ( \u03c7 2 = 5810 , p < . 0001 ) . PromptDistance Prompt Consistency Phrase Minimum Pairwise Distance 0 . 95 0 . 08 Prompt Minimum Pairwise Distance 0 . 95 0 . 05 Intra - Prompt Mean Phrase Distance - 0 . 05 - 0 . 69 Table 11 : The rotated factor loading of factor analysis on metrics of perceived helpfulness of prompts . Factors explained 68 . 9 % of the total variance . Bartlett\u2019s Test for Sphericity to indicate common factors was significant ( \u03c7 2 = 2575 , p < . 0001 ) . PromptQuality Prompt Unexpectedness PromptRelevance Prompt Understandability Phrase Helpfulness rating 0 . 85 - 0 . 13 0 . 2 0 . 15 Phrase Relevance to Task ( Motivation ) rating 0 . 89 - 0 . 2 0 . 23 0 . 14 Phrase Understanding rating 0 . 62 - 0 . 21 0 . 27 0 . 47 Phrase Relevance to Domain ( Exercise ) rating 0 . 59 - 0 . 14 0 . 54 0 . 18 Phrase Unexpectedness rating - 0 . 11 0 . 63 - 0 . 06 - 0 . 05 Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Table 12 : The rotated factor loading of factor analysis on metrics of prompt adoption . Factors explained 65 . 1 % of the total variance . Bartlett\u2019s Test for Sphericity to indicate common factors was significant ( \u03c7 2 = 1315 , p < . 0001 ) . PhraseAdoption Prompt Precision 0 . 82 Prompt Recall 0 . 67 Prompt - Ideation Distance - 0 . 92 Table 13 : The rotated factor loading of factor analysis on diversity metrics of generated messages . Factors explained 75 . 2 % of the total variance . Bartlett\u2019s Test for Sphericity to indicate common factors was significant ( \u03c7 2 = 2676 , p < . 0001 ) . IdeationDispersion IdeationEvenness Message Remote - clique 0 . 99 0 . 16 Message Sparseness 0 . 99 0 . 16 Message Span 0 . 77 - 0 . 05 Message MST Dispersion 0 . 29 0 . 96 Message Chamfer Distance - 0 . 02 0 . 91 Message Entropy 0 . 01 0 . 3 Table 14 : The rotated factor loading of factor analysis on metrics of perceived quality of the generated messages . Factors ex - plained 80 . 9 % of the total variance . Bartlett\u2019s Test for Sphericity to indicate common factors was significant ( \u03c7 2 = 5810 , p < . 0001 ) . Ideation Informative - Helpfulness IdeationQuality Informativeness rating 0 . 8 0 . 39 Helpfulness rating 0 . 66 0 . 65 Motivation rating 0 . 39 0 . 79 Table 15 : The rotated factor loading of factor analysis on metrics of group ranking of the generated messages . Factors explained 93 . 9 % of the total variance . Bartlett\u2019s Test for Sphericity to indicate common factors was significant ( \u03c7 2 = 366 , p < . 0001 ) . For usability , \u201cunrepetitive\u201d was measured with the word \u201crepetitive\u201d in the survey . IdeationsUnrepetitive Ideations Informative IdeationsMotivating Sum ( Most Unrepetitive ( Rank = 1 ) ) 1 . 32 0 . 40 0 . 10 Sum ( Most Informative ( Rank = 1 ) ) 0 . 48 0 . 72 0 . 05 Sum ( Least Unrepetitive ( Rank = 3 ) ) - 0 . 73 - 0 . 50 0 . 02 Sum ( Least Informative ( Rank = 3 ) ) - 0 . 26 - 0 . 89 - 0 . 03 Sum ( Most Motivating ( Rank = 1 ) ) 0 . 17 - 0 . 04 1 . 00 Sum ( Least Motivating ( Rank = 3 ) ) 0 . 05 - 0 . 06 - 0 . 54 Table 16 : The rotated factor loading of factor analysis on metrics of message distinctness . Factors explained 74 . 8 % of the total variance . Bartlett\u2019s Test for Sphericity to indicate common factors was significant ( \u03c7 2 = 1022 , p < . 0001 ) . IdeationDistance Ideation Min Pairwise Distance 0 . 86 Ideation Mean Pairwise Distance 0 . 86 CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Table 17 : The rotated factor loading of factor analysis on metrics of ideation effort . Factors explained 59 . 0 % of the total variance . Bartlett\u2019s Test for Sphericity to indicate common factors was significant ( \u03c7 2 = 1008 , p < . 0001 ) . Ideation Self - Quality IdeationEase Message Creativity Self - Rating 0 . 76 0 . 19 Message Motivation Self - Rating 0 . 63 0 . 55 Message Writing Ease 0 . 76 0 . 19 G SURVEY SCREENSHOTS IN USER STUDIES G . 1 Ideation User Study Figure 13 : The instructions in the Ideation User Study for the None condition . Figure 14 : For the None , users are asked to write a message that is at least one to three sentences long . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Figure 15 : The instructions of Ideation User Study for the Random ( 1 ) and Directed ( 1 ) conditions . Figure 16 : Random ( 1 ) and Directed ( 1 ) prompts consisted of one phrase per prompt . Note that selected phrase for each trial will be different . CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Figure 17 : The instructions of Ideation User Study for the Random ( 3 ) and Directed ( 3 ) conditions . Figure 18 : Random ( 3 ) and Directed ( 3 ) prompts consist of three phrases per prompt . Note that selected phrases for each trial will be different . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Figure 19 : Ideators are asked to evaluate the message they wrote by providing Likert scale ratings for many different fac - tors along with a short reflection about the message writing process . The screenshot above shows the evaluation screen for Directed ( 3 ) . G . 2 Validation User Studies Figure 20 : The instruction for individual message rating tasks . CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Figure 21 : Validators rated a randomly selected message on a Likert scale and gave a justification . Figure 22 : The instruction for group message ranking tasks . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Figure 23 : Validators were asked to rank groups of messages for motivation , informativeness and repetitiveness . Note that while we used the word \u201crepetitive\u201d for usability in the survey , we analyzed this dependent variable as \u201cunrepetitive\u201d to be consistent with other diversity metrics . CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Figure 24 : Validators were asked to rate the difference of two messages in a message - pair . H EXAMPLES OF PROMPTS AND MESSAGES WRITTEN BY IDEATORS Table 18 : Messages generated in our study and the phrase prompt ( s ) that were shown to ideators . PromptSelection Phrase ( s ) Shown Message Written Random ( 1 ) daily club swim workout Do you want a way to train your whole body ? Try a swim workout ! You can even join a club to help challenge you to reach your goals ! Random ( 1 ) like a barrier of insecurity Get out and try a new exercise today . Don\u2019t let not doing it be a barrier or insecurity . Even pro athletes have to try new exercises for the first time . Directed ( 1 ) snooze button repeatedly isn\u2019t exercise Reminder that hitting the snooze button repeatedly is NOT considered an exercise ! Make sure to wake up first thing , and get your legs moving ! Directed ( 1 ) next set of stats Not happy with what you see on the scale or the number of calories you burned ? Don\u2019t let one day\u2019s data ruin your mood . Give it time and you\u2019ll see better results if you keep at it ! Random ( 3 ) ( 1 ) hard workout may feel ( 2 ) multiple exercise interventions in terms ( 3 ) exercise program for clients plagued Hard workouts may feel uncomfortable . However , those carry the most enjoyment and success for you ! Random ( 3 ) ( 1 ) religious institution offers exercise classes ( 2 ) workout program because people ( 3 ) other forms of water aerobics Your religious institution offers exercise classes and your local pool offers water aerobics . Exercise with people for motivation ! Directed ( 3 ) ( 1 ) in the risk of diabetes ( 2 ) for the development of diabetes ( 3 ) from complications of diabetes Exercising will help you stay in shape . It will prevent health issues in the future and it can stop the risk of developing diabetes . Directed ( 3 ) ( 1 ) book and workout videos ( 2 ) mechanics and workout plans ( 3 ) exercise tapes or videos Watching tapes and videos are good ways to try out new exercises . Follow along and impress your loved ones with your new moves ! Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan I THEMATIC ANALYSIS OF MESSAGES Table 19 : Themes and categories identified with the qualitative coding of ideated messages . Theme Categories Ambiguous Benefits Ambiguous benefits Anecdote Anecdote Appeal to \" Obvious \" Knowledge Appeal to \" obvious \" knowledge Appeal by Cohorts Appeal to children | Appeal to older ages | Appeal to overweight Appeal to Fear Appeal to fear Appeal to Guilt Appeal to guilt Appeal to Shame Appeal to shame Appeal to Social Approval Appeal to social approval Barrier to Ability Barrier to ability Barrier to Boredom Encourage exercise variety | Prompt to try something new | Tips to make exercise less boring / more fun Barrier to Comfort Barrier to comfort Barrier to Cost Cheap exercises | Lower healthcare / insurance costs Barrier to Effort It will get easier | Recommending less effortful routines or exercises | Take a short break then carry on Barrier to Energy Barrier to energy Barrier to Enjoyment Prompt to research fun exercise | Recommending enjoyable activity Barrier to Motivation Barrier to motivation Barrier to Resources At home exercises | No equipment needed | No gym available Barrier to Self - Efficacy Don\u2019t feel bad if confused | Don\u2019t need certificate / qualifications | Improving self - confidence | Recommending exercises within ability Barrier to Time Barrier to time Call to Action Call to action Call to Authority Citing health experts | Unspecified authority Collective Societal Benefits Collective societal benefits Equipment Bench press | Exercise machine | Exercise machines ( unspecified | Rubber exercise tubing / band | Swimming gear | Treadmill | Vertical or horizontal press | Work - standing desk Exercise Suggestion Aerobic exercises | Aerobics | Anaerobic exercise | Biking | Body weight exercises | Cardiovascular exercise | Climbing | Competitive cycling | Dance | Diving | Double clean | Exercise through chores | Handstand | High intensity exercise | Hot yoga | Internal rotation workouts | Iron Yoga | Jump on bed | Jumping jacks | Lift weights | Lifting luggage | Meditation | Pull - ups | Pushing kids on swings | Push - ups | Resistance exercises | Ring Pull - ups | Running | Seated leg - raises | Sit - ups | Snatches | Sports | Squats | Strength training | Strenuous / moderate / vigorous exercise | Stretching | Swimming | Tennis | Using stairs | Vertical and horizontal presses | Volleyball | Walk your dog | Walking | Water exercises | Work / desk exercise | Yoga Fear of Injury Don\u2019t overexert yourself | Recommending exercises to avoid injury | Research good techniques to avoid injury | Take breaks | Tips to avoid injury for outdoor activities Food and Drink Avoid steroids / pills / drugs | Avoid unhealthy food | Exercise supplements | Exercise to avoid medication / drugs | Food recommendation | Staying hydrated | Stress eating advice Future Life Improve quality of life | Live longer Goals Journaling to track your goal | Set Actionable goals | Set daily exercise goal | Set goals based on health recommendations | Set unspecified goal | Set weight goal | Tips to reach goals | Visualizing meeting goals Health Advice Advice for diabetics | See a doctor if you are worried Health Benefits Better mobility | Bone health | Cardio Health | Fluid regulation | Help with foot problems | Lowers blood pressure | More stamina | Pain / strain relief | Slows aging process | Strong immune system | Unspecified health benefit CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Health Risks Arthritis | Breathing difficulty | Cancer | Depression | Diabetes | Heart disease | Hernias | Obesity | Unspecified health risk Improving Appearance \" Look better \" | Beach ready | Chest and back | Improved posture | Look more appealing to potential partners | Nice butt | Six - pack abs Inspirational Phrase Inspirational phrase Lack of Knowledge Research exercise routines | Research nutrition | Study exercise form Lack of Social Support Exercise with an expert | Exercise with friends | Family want you to be healthy | Find places to support you | Impress your doctor | Interacting with others | Join a health club | Join an exercise class | Meeting new people | Playing / exercising with children | Social competition Mental Health Benefits \" Feeling \" better | Happier | Improved sleep | Improves cognitive abilities | Lower depression | Lowers anxiety | Relaxing | Release endorphins | Self - esteem | Stress reduction | Unspecified mental health benefits Muscle Building Biceps | Core strength | Leg Muscles | Physically stronger | Shoulder muscles | Triceps | Unspecified Muscle building | Upper body strength Overcoming Beliefs Changing mindset Overcoming Family Obligations Overcoming family obligations OvercomingSelf - Consciousness Overcoming self - consciousness Push to Do More Prompt to increase | Prompt to steadily increase Push to Start Prompt to start exercising | Prompt to stop sedentary lifestyle Rewards Prizes from exercise competitions | Reward with food | Reward with new clothes | Reward with new exercise equipment | Unspecified reward Self - Empowerment Self - empowerment Self - Forgiving Self - forgiving Self - Reflection Self - reflection Social Comparison Avoid social comparisons | Downwards Social comparison | Upwards Social Comparison Specific Locations Around the neighbourhood | At desk exercises | At home | At school | Beach | Church / community centre | Front yard | Gym | Outdoors | Park | Travelling / airport | Walk to train station Specificity Appropriate Exercises ( e . g . \" Try what\u2019s best for you \" | Developing habits | Even small amounts of exercise | Exercise daily | Exercise regularly | Follow exercise plan / routine | Specific amount / distance to exercise | Specific days a week to exercise | Specific minutes to exercise Time to See Results Dedicate time | Fast results | Promise of results | Tips to progress faster Time to Exercise Anytime | End of the day | Morning exercise | Spring | Summer / hot weather Use of Technology Exergame | Experts review your exercises from an app | Follow videos | Listen to music / podcast | Reflect on progress | Use apps for exercise tips | Use apps for workout schedules | Use apps to track progress | Watch TV while exercising Weight Loss Aid digestion | Boost metabolism | Burning calories | Burning cellulite | Maintaining weight | Slimming down Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan J LINEAR MIXED MODELS AND STATISTICAL ANALYSIS RESULTS OF PROMPT CREATIVITY , PROMPT - IDEATION MEDIATION , AND IDEATION DIVERSITY Table 20 : Statistical analysis of responses due to effects ( one per row ) , as linear mixed effects models , all with Participant as random effect , Prompt Selection and Prompt Size as fixed effects , their interaction effect . a ) model for manipulation check analysis of how prompt configurations affect perceived prompt creativity ( RQ1 . 2 ) ; b ) model for mediation analysis of how prompt configurations affect ideation effort ( RQ2 . 2 ) . clr23n . s . means not significant at p > . 01 . p > F is the significance level of the fixed effect ANOVA . R 2 is the model\u2019s coefficient of determination to indicate goodness of fit . a ) Prompt Creativity Manipulation Check ( RQ1 . 2 ) Response Linear Effects Model ( Participant random effect ) p > F R 2 PromptUnexpected - ness Prompt Selection + Prompt Size + Prompt Selection \u00d7 Size < . 0001 n . s n . s . 523 PromptUnderstand - ability Prompt Selection + Prompt Size + Prompt Selection \u00d7 Size . 0008 < . 0001 . 0316 . 500 PromptRelevance Prompt Selection + Prompt Size + Selection \u00d7 Size < . 0001 < . 0001 n . s . . 450 Prompt Quality Prompt Selection + Prompt Size + Prompt Selection \u00d7 Size < . 0001 < . 0001 n . s . . 572 b ) Prompt - Ideation Effort Mediation Analysis ( RQ2 . 2 ) Response Linear Effects Model ( Participant as random effect ) p > F R 2 IdeationFluency Prompt Selection + Prompt Size + Prompt Selection \u00d7 Size < . 0001 < . 0001 . 0042 . 542 Ideation Ease Prompt Selection + Prompt Size + Selection \u00d7 Size < . 0001 n . s n . s . 546 PromptAdoption Prompt Selection + Prompt Size + Prompt Selection \u00d7 Size < . 0001 n . s n . s . 575 Table 21 : Statistical analysis and results of mediation effects ( RQ2 . 3 ) of how prompt configurations ( a ) and perceived prompt creativity ( b ) affect ideation diversity . See Table 20 caption to interpret tables . Positive and negative numbers in second column represent estimated model coefficients indicating how much each fixed effect influences the response . a ) Prompt Distance to Ideation Mediation Response Linear Mixed Effects Model ( Participant as random effect ) p > F R 2 IdeationMeanPairwiseDistance + 0 . 18 + 0 . 06 + 0 . 01 + 0 . 02 Prompt Mean Distance + Prompt Min Distance + Pr . P . Chamfer Dist . + Intra - Pr . P . Mean Dist . < . 0001 . 0205 n . s . < . 0001 . 399 IdeationMinimumPairwiseDistance + 0 . 10 + 0 . 15 + 0 . 06 + 0 . 02 Prompt Mean Distance + Prompt Min Distance + Pr . P . Chamfer Dist . + Intra - Pr . P . Mean Dist . . 0241 < . 0001 . 0115 . 0041 . 315 b ) Prompt Creativity to Ideation Mediation Response Linear Mixed Effects Model ( Participant as random effect ) p > F R 2 IdeationMeanPairwiseDistance \u2013 0 . 0014 + 0 . 0001 \u2013 0 . 0034 + 0 . 0018 Pr . Unexpectedness + Pr . Understandability + Prompt Relevance + Prompt Quality . 0315 n . s . 0020 n . s . 367 IdeationMinimumPairwiseDistance + 0 . 0024 \u2013 0 . 0026 \u2013 0 . 0056 + 0 . 0052 Pr . Unexpectedness + Pr . Understandability + Prompt Relevance + Prompt Quality . 0087 n . s . 0003 . 0431 . 272 CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . Table 22 : Statistical analysis of how prompt selection influences ideation diversity defined by different metrics ( RQ3 ) : a ) indi - vidual diversity , b ) collective diversity , and c ) thematic diversity . See Table 20 caption for how to interpret tables . a ) Ideation Individual Diversity Response Linear Effects Model ( Participant as random effect ) p > F R 2 IdeationMeanPairwisedistance Prompt Selection + Prompt Size + Selection \u00d7 Size < . 0001 n . s . 0005 . 361 IdeationMinPairwisedistance Prompt Selection + Prompt Size + Selection \u00d7 Size < . 0001 n . s n . s . 296 IdeationSelf - Quality Prompt Selection + Prompt Size + Selection \u00d7 Size n . s . 0292 . 0152 . 570 b ) Ideation Collective Diversity Response Linear Effects Model ( Sample as random effect ) p > F R 2 IdeationDispersion Prompt Selection + Prompt Size + Prompt Selection \u00d7 Size < . 0001 n . s < . 0001 . 873 IdeationEvenness Prompt Selection + Prompt Size + Prompt Selection \u00d7 Size < . 0001 . 0030 . 0061 . 984 c ) Ideation Collective Diversity ( Thematic Coding ) Response Linear Effects Model ( Sample as random effect ) p > F R 2 Category Flexibility Prompt Selection < . 0001 . 979 Category Originality Prompt Selection < . 0001 . 933 Theme Flexibility Prompt Selection < . 0001 . 911 Theme Originality Prompt Selection < . 0001 . 396 Table 23 : Statistical analysis of how prompt selection influences ideation creativity as validated by different methods ( RQ3 . 1 ) : a ) individual rating , b ) collective ranking , and c ) collective pairwise rating . See Table 20 caption for how to interpret tables . a ) Individual Rating Validation Response Linear Effects Model ( Participant + Ideation as random effects ) p > F R 2 Ideation Informative Helpfulness Prompt Selection < . 0001 . 559 Ideation Quality Prompt Selection n . s . . 467 b ) Collective Ranking Validation Response Linear Effects Model p > F R 2 Ideations Unrepetitive Prompt Selection < . 0001 . 284 Ideations Informative Prompt Selection < . 0001 . 340 Ideations Motivating Prompt Selection . 0426 . 028 c ) Collective Pairwise Rating Validation Response Linear Effects Model p > F R 2 DifferenceRating Prompt Selection < . 0001 . 279 Figure 25 : Results of computed individual diversity from ideations for different prompt configurations for ( left ) prompts that users understood ( > 0 ) or did not and ( right ) ideations that were fast or slow . Directed Diversity : Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideation CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan K INVESTIGATING CONFOUND OF PROMPT UNDERSTANDABILITY ON IDEATION DIVERSITY Having found that prompt understanding difficulty is correlated with ideation diversity , we investigated the alternative hypothesis that the difficulty of interpreting the prompts was a key reason for improved ideation because of increased ideation determination , rather than the content diversity in phrases due to the prompt selec - tion technique . We argue that the increase in ideation diversity due to Directed Diversity is evidenced by increased perceived diversity ratings from validators and the higher number of idea categories from the thematic analysis . This shows that Directed Diversity did stimulate more diverse ideas due to some knowledge transfer from prompt to ideations , albeit with difficulty . We identify three more sources of evidence next . First , we qualitatively analyzed ideation rationales and found that while prompts could be rated hard to understand or irrelevant , participants still adopted some ideas . Ideators cherry - picked parts that were usable or conceived tangential ideas : e . g . , P1 read \u201cor - thopaedic surgeons and exercise specialists\u201d and decided to \u201ccut out the bit about surgeons . . . I focused on the idea of specialists . . . \u201d ; P2 read \u201cballistic stretch uses vigorous momentum\u201d , commented that \u201cthis isn\u2019t a phrase that I\u2019m familiar with\u201d , yet could write about stretching : \u201cStretch , breathe , and feel mindful . \u201d Second , we quantitatively analyzed the Ideation Mean Pairwise Distance for prompts that participants understood ( Phrase Under - standing factor > 0 ) . Table 24a describes the statistical analysis of the linear mixed effects model . We found that although dis - tance was slightly higher ( i . e . , less diverse ) when ideators under - stood phrases less , regardless of understanding , ideations from Directed ( 1 ) prompts had higher distances than ideations from Ran - dom ( 1 ) prompts ( Figure 25 , left ) . The effect due to Prompt Type was larger than due to Phrase Understanding . Furthermore , we an - alyzed whether the difficulty to understand may manifest as slower ideation speed due to more thinking time to ideate to lead to better diversity , but did not find a correlation between phrase understand - ing and ideation speed ( \u03c1 = . 046 , p = n . s . ) , and found the opposite effect that slower ideations led to lower distances ( Table 24a and Figure 25 right ) . These suggest that prompt selection is a primary factor . Third , we investigated if Directed Diversity helped to stimulate ideas closer to prompts than would be done naturally without prompts ( None ) or accidentally with Random prompts . We analyzed this by calculating the prompt - ideation distance between Directed prompts and their corresponding ideated message , and their closest None and Random messages . Table 24b describes the statistical analysis of the linear mixed effects model . Figure 26 shows that the directed ideations were closest to the prompts , indicating the efficacy of Directed Diversity to transfer knowledge for ideation diversity . Table 24 : Statistical analysis of a ) how ideators\u2019 understanding of phrases influences ideation diversity and b ) how similar Directed ideations are to their prompts compared to other None and Random Messages . a ) Ideation Individual Diversity Response Linear Effects Model ( Participant as random effect ) p > F R 2 Ideation Mean Pairwisedistance Prompt Selection + Prompt Size + Selection \u00d7 Size + Phrase Understanding > 0 + Selection \u00d7 ( Understanding > 0 ) + Log ( Ideation Speed ) > Median + Selection \u00d7 ( Speed > Median ) < . 0001 n . s . 0001 . 0074 n . s . . 0043 . 0213 . 289 b ) Prompt - Ideation Closeness Response Linear Effects Model ( Participant random effect ) p > F R 2 Prompt - MessageDistance Message Type < . 0001 . 047 Figure 26 : Results of prompt - message distance ( how dissimilar a prompt is from a message ) comparing different messages with respect to Directed ( 3 ) prompts . CHI \u201921 , May 08 \u2013 13 , 2021 , Yokohama , Japan Samuel Cox et al . L EXAMPLES OF MESSAGE - GROUP RANKING The factors of message - group ranking were derived from the sum of rankings ( for each of the three condition ) per validator for his five ranking trials ( see the factor loadings in Table 15 The rotated factor loading of factor analysis on metrics of group ranking of the generated messages . ) . Therefore , these factors reflect the probabil - ity of how a validator ranked the message - groups of each condition . The following table shows examples of the factors and the corre - sponding message - group samples . Table 25 : Examples of the factors with low ( < Median ) and high ( \u2265 Median ) scores for \u201cIdeations Unrepetitive\u201d and \u201cIdeations Informative\u201d . I d e a t i o n s U n r e p e t i t i v e I d e a t i o n s I n f o r m a t i v e Example Message - Group of 5 Ideations High High \u2022 Exercise can help you have really good sleep . \u2022 Why don\u2019t you try something new ? Shake it up a little ? Maybe lift a few small weights , or add in some squats - variety keeps things interesting . \u2022 You have 24 hours in a day \u2013 think about how much time you spend on social media or doing something that\u2019s not going to benefit you in the long run and use that time to workout by prioritizing your health ! \u2022 Go for the goal , do not stop , do not think you cannot do it . YOU CAN ! \u2022 Summer is coming up and you want to look good when you are outside . Exercising at a health club is a good way to meet other people . Have a friend to work out with you and have each other motivate each other . High Low \u2022 Just get moving . It\u2019s that simple . \u2022 Your dog is bored . Take him for a walk ! It\u2019s good for both of you and he\u2019ll be thrilled ! \u2022 Work out more . You will feel and look better . You will get more toned . \u2022 Exercising can improve your cardio health , thus helping you to live a more fulfilling life . \u2022 Start exercising more ! You\u2019ll improve your mood and boost your self confidence . You\u2019ll feel great ! Low High \u2022 Switch off an air conditioner while working out . Let the sweat out , and burn some calories . \u2022 Not happy with what you see on the scale or the number of calories you burned ? Don\u2019t let one day\u2019s data ruin your mood . Give it time and you\u2019ll see better results if you keep at it ! \u2022 Sleep is when the body recovers and is very important . Rest early and run tomorrow ! \u2022 Overcome your anger and your fear by going to the gym and working out ! \u2022 Always stretch so that you perform at your best . You can do it ! Low Low \u2022 Exercise helps build strong muscles as well as well as making your body more flexible . You will reduce your risk of disease and injury by keeping up with your program . \u2022 The first page of every book is the hardest to grasp , the first drink tastes the most sour and the first minute of every exercise is the hardest . All things get easier as you press on . \u2022 Walking to the train station is better as it gets you more active . Avoid lifts to the train station . \u2022 Keep exercising to keep your mind off difficult personal issues , like college admissions . \u2022 By using the proper squat position , you can train muscles that take pressure off of your knee and back to help with pain in both areas .",
    "kaplan2022load": "Molecular Biology of the Cell \u2022 33 : ar50 , 1 \u2013 16 , May 15 , 2022 33 : ar50 , 1 MBoC | ARTICLE Load adaptation by endocytic actin networks ABSTRACT Clathrin - mediated endocytosis ( CME ) robustness under elevated membrane ten - sion is maintained by actin assembly \u2013 mediated force generation . However , whether more actin assembles at endocytic sites in response to increased load has not previously been investigat - ed . Here actin network ultrastructure at CME sites was examined under low and high mem - brane tension . Actin and N - WASP spatial organization indicate that actin polymerization initi - ates at the base of clathrin - coated pits and that the network then grows away from the plasma membrane . Actin network height at individual CME sites was not coupled to coat shape , rais - ing the possibility that local differences in mechanical load feed back on assembly . By manipu - lating membrane tension and Arp2 / 3 complex activity , we tested the hypothesis that actin assembly at CME sites increases in response to elevated load . Indeed , in response to elevated membrane tension , actin grew higher , resulting in greater coverage of the clathrin coat , and CME slowed . When membrane tension was elevated and the Arp2 / 3 complex was inhibited , shallow clathrin - coated pits accumulated , indicating that this adaptive mechanism is especially crucial for coat curvature generation . We propose that actin assembly increases in response to increased load to ensure CME robustness over a range of plasma membrane tensions . INTRODUCTION Actin networks produce force for a wide variety of cellular processes through a Brownian ratchet mechanism ( Mogilner and Oster , 1996 , 2003 ; Pollard , 2016 ) . Live - cell studies of lamellipodia ( Mueller et al . , 2017 ) , biochemical reconstitutions ( Bieling et al . , 2016 ; Funk et al . , 2021 ; Li et al . , 2021 ) , and modeling studies ( Akamatsu et al . , 2020 ) have shown that actin networks nucleated by the Arp2 / 3 complex respond to increased load by becoming more dense , which en - hances force production . This phenomenon has been demonstrated in the context of actin networks producing force on the lamelli - podum , whose membrane is essentially flat on the length scale of an individual actin filament branch . However , actin networks can also produce force on membranes that change shape over time , for ex - ample during vesicle formation . Whether actin networks in the latter context show load adaptation has not been investigated . From yeast to humans , transient actin assembly is associated with the formation of clathrin - coated endocytic vesicles . In yeast cells , actin assembly is required to generate forces to invaginate the plasma membrane against a high intrinsic turgor pressure for clath - rin - mediated endocytosis ( CME ) ( Kaksonen et al . , 2005 ; Aghamo - hammadzadeh and Ayscough , 2009 ; Idrissi et al . , 2012 ; Kukulski et al . , 2012 ) . When actin assembly is perturbed in mammalian cells , CME typically slows in a manner that depends on cell type ( Fujimoto et al . , 2000 ; Merrifield et al . , 2005 ; Yarar et al . , 2005 , 2007 ; Grassart et al . , 2014 ; Dambournet et al . , 2018 ; Sch\u00f6neberg et al . , 2018 ) . A potential cause of this reported variation between cell types might Monitoring Editor Alex Mogilner New York University Received : Nov 29 , 2021 Revised : Mar 28 , 2022 Accepted : Mar 29 , 2022 This article was published online ahead of print in MBoC in Press ( http : / / www . molbiolcell . org / cgi / doi / 10 . 1091 / mbc . E21 - 11 - 0589 ) on April 7 , 2022 . Author contributions : C . K . and D . G . D . conceived the study and experiments . C . K . performed live cell data acquisition , SRM data analysis and live cell data analysis . S . J . K . , X . C . , and K . X . performed SRM and super - resolution data reconstruction and supervised SRM imaging . J . S . supported the SRM data analysis . E . S . and A . D . - M . performed membrane tether pulling experiments by atomic force microscopy and data analysis and supervised AFM tether pulling experiments . C . K . , and M . A . prepared the plot layouts and figures . C . K . , M . A . , and D . G . D . wrote the manuscript with feedback from the other authors . \u2020 Present address : Department of Biology , University of Washington , Seattle , WA 98195 . * Address correspondence to : David G . Drubin ( drubin @ berkeley . edu ) , Matthew Akamatsu ( akamatsm @ uw . edu ) , or Ke Xu ( xuk @ berkeley . edu ) . \u00a9 2022 Kaplan et al . This article is distributed by The American Society for Cell Biol - ogy under license from the author ( s ) . Two months after publication it is available to the public under an Attribution \u2013 Noncommercial - Share Alike 4 . 0 International Cre - ative Commons License ( http : / / creativecommons . org / licenses / by - nc - sa / 4 . 0 ) . \u201cASCB\u00ae , \u201d \u201cThe American Society for Cell Biology\u00ae , \u201d and \u201cMolecular Biology of the Cell\u00ae\u201d are registered trademarks of The American Society for Cell Biology . Abbreviations used : CCP , clathrin - coated pit ; CCS , clathrin - coated structure ; CME , clathrin - mediated endocytosis ; SI , shape index . Charlotte Kaplan a , Sam J . Kenny b , Xuyan Chen b , Johannes Sch\u00f6neberg a , c , Ewa Sitarska d , e , Alba Diz - Mu\u00f1oz d , Matthew Akamatsu a , \u2020 , * , Ke Xu b , f , * , and David G . Drubin a , * a Department of Molecular and Cell Biology and b Department of Chemistry , University of California , Berkeley , Berkeley , CA 94720 - 3220 ; c Department of pharmacology and Department of chemistry and biochemistry , University of California , San Diego , La Jolla , CA 92093 ; d Cell Biology and Biophysics Unit , European Molecular Biology Laboratory Heidelberg , 69117 Heidelberg , Germany ; e Collaboration for joint PhD degree between EMBL and Heidelberg University , Faculty of Biosciences ; f Chan Zuckerberg Biohub , San Francisco , CA 94158 2 | C . Kaplan et al . Molecular Biology of the Cell be differences in plasma membrane tension ( Pontes et al . , 2017 ; Djakbarova et al . , 2021 ) . Indeed , in mammalian cells , actin assembly becomes increasingly critical as plasma membrane tension increases ( Yarar et al . , 2005 ; Liu et al . , 2009 ; Batchelder and Yarar , 2010 ; Boulant et al . , 2011 ) . Actin perturbation results in the accumulation of \u201cU - shaped\u201d membrane invaginations , reflecting difficulty in pro - gressing to the subsequent \u201c \u03a9 - shaped\u201d membrane stage ( Fujimoto et al . , 2000 ; Yarar et al . , 2005 ; Boulant et al . , 2011 ; Almeida - Souza et al . , 2018 ) . In total , these findings suggest that actin assembly improves the efficiency of CME in mammalian cells , potentially com - pensating for changes in plasma membrane tension . Despite the fact that actin assembly appears to be associated with CME in all eukaryotes , and over a large range of membrane tension values , whether it adapts to changes in membrane tension is not known . Knowing whether the actin cytoskeleton at CME sites adapts to changes in membrane tension will facilitate elucidation of the fundamental mechanisms by which cytoskeletal complexes pro - duce force . Platinum replica electron microscopy of cultured cells led to the proposal that actin networks assemble in a collar - like arrangement around the vesicle neck ( Collins et al . , 2011 ) . This actin organization would imply that a constricting force is generated toward the neck of the pit , supporting fission . However , actin filaments interact not only with the vesicle neck but also with the bud surface , given that the clathrin coat is impregnated with actin - binding linker proteins like Hip1R and Epsin ( Engqvist - Goldstein et al . , 2001 ; Messa et al . , 2014 ; Sochacki et al . , 2017 ; Clarke and Royle , 2018 ) . Such an ar - rangement and evidence from photobleaching studies imply that actin filaments also apply a force that pulls the clathrin - coated pit ( CCP ) into the cell interior ( Kaksonen et al . , 2003 ; Akamatsu et al . , 2020 ) . In yeast , such a pulling mechanism is likely . Actin filaments are nucleated in a ring surrounding the pit and grow from the plasma membrane toward the cell interior . The resulting actin filaments are coupled to the clathrin coat surface , generating an inward force orthogonal to the plane of the plasma membrane ( Kaksonen et al . , 2003 ; Carroll et al . , 2012 ; Skruzny et al . , 2012 , 2015 ; Picco et al . , 2015 ; Mund et al . , 2018 ) . Because the endocytic machinery is highly conserved from yeast to mammals , a similar mechanism for actin force generation during CME seems likely ( Akamatsu et al . , 2020 ) . However , ultrastructural evidence for actin organization through dif - ferent stages of mammalian CME , and for how this organization might respond to changing membrane tension , is lacking . Several competing models for actin organization at CME sites in mammalian cells have been proposed , so it is important to distin - guish between these models ( Engqvist - Goldstein et al . , 2001 ; Boulant et al . , 2011 ; Messa et al . , 2014 ; Sochacki et al . , 2017 ; Clarke and Royle , 2018 ) . Recent advances in superresolution microscopy ( SRM ) permit examination of cellular ultrastructure with large sample sizes , low invasiveness , and high molecular specificity to reveal the ultrastructure of membrane cytoskeletal systems in mammalian cells ( Xu et al . , 2012 , 2013 ; Hauser et al . , 2018 ) . Here , we combined two - color , three - dimensional stochastic opti - cal reconstruction microscopy ( 2c - 3D STORM ) with live - cell fluores - cence imaging to determine how filamentous actin is organized at CME sites with different coat geometries under various values of plasma membrane tension . These measurements led to the conclu - sion that the actin network grows from the base of the pit inward , supporting a pulling mechanism for mammalian endocytic actin fila - ments . The size of the actin network is not tightly coupled to the geometry of the endocytic coat , implying that assembly is modu - lated by factors uncoupled from coat shape . Importantly , under el - evated membrane tension , the actin network grows higher relative to the pit at all endocytic coat geometries . These observations sup - port a mechanism in which the actin network adapts to load by in - creasing in size to enhance force production and ensure robust CME across a wide range of membrane tension values . RESULTS Actin organization at CME sites suggests that force generation can be either parallel or orthogonal to the axis of CCP formation , or both , at different sites in the same cell We used 2c - 3D STORM ( Rust et al . , 2006 ; Huang et al . , 2008 ) to determine the ultrastructural organization of the actin cytoskeleton at sites of CME . Henceforth , we refer to clathrin - coated structures ( CCSs ) as relatively flat clathrin structures and CCPs as curved , in - vaginating clathrin structures . Our 2c - 3D STORM method preserves cellular ultrastructure by chemical fixation of intact cells , provides high molecular specificity due to immunofluorescence labeling , and allows large numbers of sites to be imaged ( Xu et al . , 2012 ) . We conducted our experiments on a skin - melanoma cell line ( SK - MEL - 2 ) wherein \u223c 87 % of dynamin2 - eGFP EN ( DNM2 - eGFP EN ) spots coaccu - mulate with actin ( Grassart et al . , 2014 ) . We used this SK - MEL - 2 cell line endogenously expressing DNM2 - eGFP EN and clathrin light chain A - tagRFP - T ( CLTA - TagRFP - T EN ) for both live - cell and super - resolution experiments ( Doyon et al . , 2011 ) . In these cells , we resolved CF680 - labeled CCSs in the X - Y plane as discrete , round or elliptical shapes on the ventral surface ( Figure 1A ) . Most of the CCSs appeared connected to filamentous actin vi - sualized by Alexa Fluor 647 \u2013 tagged phalloidin . These superresolu - tion reconstructions resolve the association between clathrin coats and actin networks for hundreds of pits with high resolution in all three dimensions . The SDs of positions of single fluorophores were 10 nm in plane for XY and 19 nm in depth for the Z dimension ( Supplemental Figure S1 ; Materials and Methods ) . Knowing how actin networks are organized spatially in three di - mensions at CME sites provides insights into its force generation mechanisms . It was important to show that we could distinguish ac - tin specifically associated with CCSs from actin in the cell cortex . In Supplemental Figure S2 , we show STORM images to compare actin at CCSs with actin at randomly selected regions of the cell cortex . We found examples of actin that specifically accumulates at the CCP ( Supplemental Figure S2 , D and I ) . Here , actin builds up higher into the cell interior compared with the actin extending horizontally away from the CCP and at randomly selected sites on the cell cortex ( Sup - plemental Figure S2 , E and J ) . This observation is consistent with recent cryoelectron tomograms in this cell type distinguishing branched actin networks around CCPs from the largely unbranched cortical actin network ( Serwas et al . , 2021 ) . To then investigate actin organization at multiple CCSs , we ren - dered the CCSs in three dimensions by cropping a tight area around each clathrin and actin structure in x - y STORM image pro - jections and generated an x - z STORM image projections from the selected regions of interest ( Figure 1 , B and C ) . To our surprise , we observed strikingly different actin filament spatial organizations in the clathrin coats that we examined , even when they were near each other in the same cell and even when they were at the same morphological stage of CME ( see below ) ( Figure 1C ) . In the first example shown , a thin layer of actin filaments resided along the base of the clathrin coat ( Figure 1C , inset 1 ) , reminiscent of struc - tures observed in electron micrograph by Collins et al . ( 2011 ) . In contrast , in the second example , actin filaments covered the CCP completely ( Figure 1C , inset 2 , and Supplemental Figure S2D ) . This organization resembles actin interacting with the entire clathrin Volume 33 May 15 , 2022 Actin load adaptation in endocytosis | 3 coat as in yeast ( Mulholland , 1994 ; Idrissi et al . , 2008 ; Ferguson et al . , 2009 ; Kukulski et al . , 2012 ; Buser and Drubin , 2013 ; Mund et al . , 2018 ) and in recent cryoelectron tomograms ( Akamatsu et al . , 2020 ) . These micrographs indicate that distinct CME - associ - ated actin structures can coexist in the same cell , consistent with models for force generation both parallel and orthogonal to the invagination axis . CME site \u2013 associated actin networks grow from the CCP base to the tip of the coat , and their organization is not coupled to clathrin coat geometry We used quantitative analysis of these STORM reconstructions to determine how actin is organized at CME sites and how actin orga - nization relates to coat geometry . We first selected 989 high - resolu - tion clathrin coats by visual inspection based on quality control cri - teria explained in Materials and Methods and determined which ones had associated actin filaments . When cropping CCSs for analy - sis , we could clearly distinguish single sites from double CCSs . In Supplemental Figure S3 we present views of 20 sites identified as singles and 20 sites identified as doubles in this manner . We ex - cluded 86 % of the visually inspected images ( in total 6858 ) that ac - counted for double CCSs and CCSs of noncircular shape , as well as very large and very small clathrin antibody \u2013 positive immunostained structures in the XY plane ( Supplemental Figure S4A ) . We detected actin associated with 74 % of the clathrin coats selected for further analysis , which is comparable to previous measurements for the same cell type that we made from endocytic traces for events in live cells in which dynamin2 - GFP associated with actin - RFP , considering that our superresolved images are snapshots of time - lapse events ( Grassart et al . , 2014 ) . We classified the coats by stage based on their shape in the XZ dimension , similar to earlier analyses of electron micrographs ( Avinoam et al . , 2015 ) . For presentation purposes , clathrin coat shapes were binned into three geometrical categories based on their aspect ratio ( which we call the shape index [ SI ] ) : shallow ( flat - ter coat ) , U - shaped ( intermediate aspect ratio ) , or \u03a9 - shaped ( coat with height similar to its width ) ( Figure 1D ; Supplemental Figure S4 , B and C ) . This measurement of SI was consistent whether we carried out the measurement on XZ or YZ projections of the coat ( r 2 = 0 . 92 ) ( Supplemental Figure S5 ) . We found a surprisingly wide variety of actin organizations in each of the three geometries ( Figure 1E and Supplemental Figure S4D ) . To quantify the relation - ship between actin organization and endocytic coat geometry , the SI of the coat was plotted as a function of the extent of actin cover - ing clathrin . When the actin network is larger in size than the clath - rin coat , we define this state as 100 % coverage ( see Materials and Methods ; Supplemental Figure S4B ) . There was no significant cor - relation between actin / coat coverage and coat shape for any of the three geometries ( Figure 1F ) . We also observed a lateral mean displacement between the peak signals of clathrin and actin of 74 \u00b1 42 nm , indicating an asymmetry in actin localization around the pits ( Yarar et al . , 2005 ; Collins et al . , 2011 ; Jin et al . , 2021 ; Serwas et al . , 2021 ) . This asymmetry value did not significantly change between CME geometries ( Supplemental Figure S4 , E and F ) . We conclude that irrespective of coat geometry , some pits have a thin actin network at the base of the pit , others have an interme - diate level of coverage around the clathrin coat , and others have actin completely covering the pit . Thus , actin assembly at CCSs is not coupled to a particular stage of coat curvature development but might be recruited to promote curvature development . In our reconstructions , we observed that whenever actin only partially covers the clathrin coat , the network is always located at the base of the pit ( Figure 1E ) . This observation suggests that actin polymerization is nucleated at the base of the pit and that the net - work then grows inward around and over the tip of the clathrin coat . To test the generality of this observation , we calculated the differ - ence in average axial ( Z ) position between the clathrin and actin signals for each site . We define this difference , D z , such that nega - tive D z corresponds to actin residing nearer the base of the pit ( Figure 1G ) . To determine whether actin grows from the base or tip of the pit , we plotted D z as a function of the extent of coverage between actin and the clathrin coat . Increasing values of D z would indicate that the network grows from the base of the pit toward the cell interior ( Figure 1G ) . Indeed , as a function of actin / coat cover - age , D z increased from negative values to near zero ( Figure 1H ) . We conclude that actin polymerization is initiated at the base of clathrin coats . Given our finding that actin growth originates from the base of the pit , we next investigated the spatial distribution of the actin nu - cleation factor N - WASP at clathrin coats by 2c - 3D STORM . Consis - tent with our conclusions about where actin assembly occurs at CCPs , N - WASP localized to the base of both shallow and highly curved clathrin coats ( Supplemental Figure S6A ) . More unexpect - edly , at some CME sites , N - WASP covered the entire clathrin coat irrespective of coat geometry ( Supplemental Figure S6B ) . In summary , we conclude that actin polymerization is nucleated at the base of CCPs and grows toward the coat\u2019s tip ( Figure 1I ) . Unexpectedly , actin nucleation is not coupled to coat geometry . A possible explanation for the variety of actin organizations that we observed associated with clathrin coats is that actin network organi - zation responds to differences in load at CME sites . Dynamics of CME slow down under elevated membrane tension We next combined osmotic shock with live - cell fluorescence micros - copy and 2c - 3D STORM to determine how actin - mediated force generation contributes to CCP formation under elevated mem - brane tension ( Figure 2A ) . Previous EM studies identified a require - ment for actin filaments at the \u201cU\u201d to \u201c \u03a9 \u201d transition ( Boulant et al . , 2011 ) . However , for a mechanistic understanding , the quantitative relationship between membrane tension and endocytic dynamics must be elucidated ( Akamatsu et al . , 2020 ) . Our quantitative light microscopy \u2013 based analysis of a large number of sites at different CME stages provided the necessary sensitivity to detect effects throughout the process . We first needed to establish conditions un - der which CME dynamics are affected by elevated membrane ten - sion in these cells . To determine how membrane tension is affected by changes in media osmolarity , we performed membrane tether pulling experi - ments by atomic force microscopy ( AFM ) on SK - MEL - 2 cells cultured under isotonic and hypotonic conditions ( 75 mOsm ) ( Figure 2B ) . In isotonic media , the force required to maintain a pulled membrane tether was 33 . 0 \u00b1 7 . 4 pN . Under hypotonic conditions the tether force increased to 48 . 0 \u00b1 17 . 1 pN ( Figure 2B ) . These measurements allowed us to quantitatively relate a given hypotonic environment in these cells to changes in membrane tension . To decipher the relationship between CME dynamics and membrane tension , we used total internal reflection fluorescence ( TIRF ) microscopy to image SK - MEL - 2 cells under isotonic and hy - potonic conditions ( Figure 2C ) . CLTA - TagRFP - T EN and DNM2 - eGFP EN fluorescence lifetimes were determined by single - particle tracking ( Hong et al . , 2015 ) . In isotonic media most clathrin tracks were relatively short ( 47 \u00b1 32 s ) with a burst of DNM2 - eGFP EN signal peaking at the end ( DNM2 lifetime 39 \u00b1 32 s ) ( Figure 2C , 4 | C . Kaplan et al . Molecular Biology of the Cell Volume 33 May 15 , 2022 Actin load adaptation in endocytosis | 5 Supplemental Movie 1 , and Supplemental Table S1 ) . Neither iso - tonic media nor exchange to slightly hypotonic 225 mOsm media noticeably affected DNM2 - eGFP EN and CLTA - TagRFP - T EN life - times or CME initiation and completion rates ( Supplemental Figure S7 , A \u2013 D ) . Only 1 \u2013 2 % of these CLTA - TagRFP - T EN and DNM2 - eGFP EN fluorescence tracks persisted over the entire 4 . 8 min movie under these conditions ( Supplemental Figure S7 , B and D ) . Upon treatment with moderately hypotonic ( 150 mOsm ) media for 10 min , CLTA - TagRFP - T EN lifetime in cells increased by 20 % ( 62 \u00b1 51 vs . 49 \u00b1 31 s ) ( Supplemental Figure S7E ) . This moderate treatment also had mild effects on the CME initiation rate ( 15 \u00b1 5 \u00d7 10 \u2013 2 \u03bcm \u2013 2 min \u2013 1 vs . 25 \u00b1 6 \u00d7 10 \u2013 2 \u03bcm \u2013 2 min \u2013 1 ) , completion rate ( 11 \u00b1 4 \u00d7 10 \u2013 2 \u03bcm \u2013 2 min \u2013 1 vs . 19 \u00b1 6 \u00d7 10 \u2013 2 \u03bcm \u2013 2 min \u2013 1 ) , and stalling rate ( events that persist the entire length of our 4 . 8 min movies ) ( 3 \u00b1 1 % vs . 0 . 8 \u00b1 0 . 9 % ( Supplemental Figure S7F ) . Treatment of cells with more strongly hypotonic ( 75 mOsm ) me - dia dramatically perturbed CME dynamics . After 2 min in 75 mOsm hypotonic media , tracks elongated and very often lasted the entire duration of the 4 . 8 min movie ( Figure 2C , Supplemental Movie 2 , and Supplemental Table S1 ) . The mean lifetimes of CLTA - TagRFP - T EN and DNM2 - eGFP EN tracks were 128 \u00b1 112 s ( Figure 2D ) and 125 \u00b1 112 s ( Figure 2E ) , respectfully . We also observed a substantial decrease in the CME initiation rate ( 10 . 3 \u00b1 1 . 5 \u00d7 10 \u2013 2 \u03bcm \u2013 2 min \u2013 1 vs . 22 . 2 \u00b1 3 . 5 \u00d7 10 \u2013 2 \u03bcm \u2013 2 min \u2013 1 ) and completion rate ( 5 . 3 \u00b1 1 . 7 \u00d7 10 \u2013 2 \u03bcm \u2013 2 min \u2013 1 vs . 17 . 8 \u00b1 3 . 4 \u00d7 10 \u2013 2 \u03bcm \u2013 2 min \u2013 1 ) and an increase in track stalling ( 19 \u00b1 9 % vs . 0 . 7 \u00b1 0 . 3 % ) ( Figure 2 , F \u2013 H ; Supplemental Figure S7H ) . After 10 min of culturing , the CLTA - TagRFP - T EN and DNM2 - eGFP EN lifetimes began to recover , most likely reflecting cellular adaptation to the hypotonic treatment ( Figure 2 , C \u2013 H , and Supplemental Movie 3 , 4 , and 5 ) . We did not detect effects of hypo - tonic media treatment on lifetimes of tracks containing only CLTA - TagRFP - T EN , characteristic of clathrin structures not associated with CME ( Supplemental Figure S7 , A , C , E , and G ) ( Hong et al . , 2015 ) . DNM2 - eGFP EN - only events , presumably representing non \u2013 clathrin mediated endocytosis events , showed a moderate response to el - evated membrane tension with 75 mOsm hypotonic media ( Supple - mental Figure S7G ) . We conclude that elevating plasma membrane tension with hypotonic shock perturbs CME dynamics in a dose - dependent manner . Actin force generation assists clathrin coat remodeling under elevated membrane tension Next , we determined the role of branched actin filament assembly on endocytic outcome under elevated membrane tension . Branched actin networks are generated by activated Arp2 / 3 complex ( Mullins et al . , 1998 ) . We inhibited Arp2 / 3 complex \u2013 mediated actin polym - erization using the small molecule CK666 ( Nolen et al . , 2009 ; Hetrick et al . , 2013 ) . We reasoned that inhibition of the Arp2 / 3 complex might have less effect on membrane tension than global actin in - hibitors such as latrunculin and jasplakinolide often used in previous CME studies . Nevertheless , because Arp2 / 3 complex inhibition can affect membrane tension ( Diz - Mu\u00f1oz et al . , 2016 ) , we first carefully established experimental conditions in which effects of high mem - brane tension or CK666 would not mask one another . Treatment of cells with 100 \u00b5M CK666 did not affect the membrane tether force in these cells ( p > 0 . 5 ) during the time the experiment was per - formed ( Supplemental Figure S8A ) . Under these optimized condi - tions , we performed live - cell fluorescence microscopy and STORM to learn more about the clathrin coat remodeling steps at which branched actin network assembly is required . We first titrated CK666 and monitored the effect on CME dy - namics after 2 min of treatment . We aimed to identify a minimal CK666 concentration that induced a rapid effect on CME dynamics . CLTA - TagRFP - T EN lifetimes to 79 \u00b1 66 s after 2 min of 100 \u00b5M CK666 treatment compared with 56 \u00b1 40 s for the dimethyl sulfoxide ( DMSO ) control ( Supplemental Figure S8B and Supplemental Table S2 ) . Similarly , DNM2 - EGFP EN lifetimes increased to 65 \u00b1 64 s upon CK666 treatment , compared with 49 \u00b1 38 s for the DMSO control ( Supplemental Figure S8B ; Supplemental Table S2 ) . CK666 at 100 \u00b5M did not affect CME completion frequency or the percentage of persistent CLTA - TagRFP - T EN tracks ( Supplemental Figure S8 , C and D ) , though we observed a small decrease in the CME initiation rate ( Supplemental Figure S8E ) . The elongation of DNM2 - eGFP EN and CLTA - TagRFP - T EN lifetimes upon 100 \u00b5M CK666 treatment was exacerbated upon simultaneous incubation in moderately hypotonic media ( 150 mOsm ) . Compared with controls , the combination of 100 \u00b5M CK666 and 150 mOsm hyopotonic media markedly lengthened the clathrin lifetimes to 96 \u00b1 86 s ( compared with DMSO control , 59 \u00b1 52 s ) and dynamin2 to FIGURE 1 : 2c - 3D STORM resolves clathrin structures highly connected to actin networks at different stages of endocytosis . ( A ) STORM image of the ventral surface of an SK - MEL - 2 cell immunolabeled with the CF - 680 antibody ( clathrin coats in red ) and phalloidin - AF647 ( actin in cyan ) . Orange squares are areas shown in panel B . Color bar shows the z position of actin . Scale bar : 5 \u00b5m . ( B ) Magnification of highlighted areas 1 and 2 in panel A . Magenta squares are shown in panel C . Scale bars : 250 nm . ( C ) X - Z projections of the regions highlighted in panel B . Scale bars : 100 nm . ( D ) Illustration of binning clathrin coats ( red ) into three geometric stages based on their aspect ratio ( shape index SI ) . Shallow : SI < 0 . 7 ; U - shape : 0 . 7 < SI > 0 . 9 and \u03a9 : SI > 0 . 9 . ( E ) X - Z projections of representative STORM images showing clathrin coats ( red ) with different actin ( cyan ) coverages around clathrin . Calculated SI of shallow CCSs from left to right image : 0 . 56 , 0 . 53 , 0 . 51 , 0 . 55 ; for U - shaped CCPs from left image to right image : 0 . 87 , 0 . 89 , 0 . 86 , 0 . 82 ; for \u03a9 - shaped CCPs from left image to right image : 1 . 31 , 1 . 06 , 1 . 31 , 1 . 52 . Scale bars : 100 nm . ( F ) Graph of endocytic coat SI as a function of actin coverage for shallow ( black dots ) , U - shaped ( blue dots ) , and \u03a9 - shaped ( gray dots ) pits . Categories of shape indices are chosen similar to E . Pits with actin coverage > 5 % are shown . R = \u2013 0 . 04 , n = 719 . Events accumulated from six cells . ( G ) Cartoon depicting the clathrin coat with actin either at the tip of the coat ( top ) , covering the clathrin coat completely ( middle ) , or at the base of the clathrin coat ( bottom ) . Dashed black lines indicate the average Z position of actin and clathrin . D z is the difference between average actin and clathrin Z positions . D z < 0 is defined as the average actin Z position nearer the base of the pit . Schematic is a hypothetical plot of D z vs . actin coverage for scenarios in which actin grows from the tip of the coat ( red line ) or the base of the pit ( black line ) . ( H ) D z as a function of actin coverage ( for actin coverage > 5 % , R = 0 . 66 , n = 719 , N = 6 cells ) . ( I ) Cartoon of actin ( blue ) growing from the base of the pit ( black lines ) to cover clathrin coat ( red ) from a shallow membrane invagination to a fully formed membrane vesicle . X - Z projection ( side profile ) is shown . Dashed arrows indicate that growth of the actin network is not tightly coupled to the endocytic coat geometry and is variable in extent . 6 | C . Kaplan et al . Molecular Biology of the Cell 84 \u00b1 85 s ( compared with DMSO control , 47 \u00b1 49 s ) ( Figure 3 , A \u2013 F , and Supplemental Table S3 ) . We conclude that in these cells , Arp2 / 3 complex \u2013 mediated actin assembly is required to maintain normal CME dynamics under elevated membrane tension . Next , we used STORM to determine which endocytic coat ge - ometries are enriched upon a combination of Arp2 / 3 complex inhi - bition and osmolarity treatment . Cells were cultured under the CK666 and osmolarity conditions described above , chemically fixed , and then immunolabeled for clathrin . Clathrin coat height served as a metric for the membrane invagination stage . As in the above 2c - 3D STORM experiments , the full progression from a flat clathrin coat to a rounded vesicle could be clearly resolved in 3D ( Figure 3G , bottom image panel ) . Control cells treated with DMSO showed an average clathrin coat height of 98 \u00b1 21 nm ( Figure 3H and Supplemental Figure S8F ) . The average height increased to 106 \u00b1 27 nm when cells were treated with 100 \u00b5M CK666 ( Figure 3H and Supplemental Figure S8F ) . Thus , when Arp2 / 3 complex activity was inhibited in these cells , clathrin pits accumulated at a greater height . This observation suggests that more clathrin pits may stall at a later stage of progression , consistent with observations of accu - mulated coat geometries from actin inhibitor studies ( Yarar et al . , 2005 ; Boulant et al . , 2011 ; Yoshida et al . , 2018 ) . Interestingly , when Arp2 / 3 - mediated actin polymerization was inhibited in cells with elevated membrane tension , the average clathrin coat height decreased to 96 \u00b1 24 nm ( Figure 3 , G and H , and Supplemental Figure S8F ) . This height decrease was also reflected in FIGURE 2 : Quantitative analysis of CME mechanosensitivity under elevated membrane tension . ( A ) Schematic of cells in isotonic media ( top ) or hypotonic media ( bottom ) , which causes water influx and stretches the cell membrane . In this figure , hypotonic treatment is 75 mOsm . ( B ) Mean membrane tether force values measured by AFM of cells in isotonic media ( n = 18 ) or in hypotonic media ( n = 17 ) . Mean values were obtained by pulling at least three tethers ( three independent experiments ) . In hypotonic treatment , circles are mean tether values from 2 to 10 min after hypotonic media exchange and triangles are mean tether values obtained between 10 and 16 min after hypotonic media exchange . Bars are mean \u00b1 SD . p = 0 . 002 by two - tailed Mann \u2013 Whitney test . ( C ) Kymographs of TIRF micrographs of live SK - MEL - 2 cells endogenously expressing CLTA - TagRFP - T EN ( magenta ) and DNM2 - eGFP EN ( green ) . Time is on the X axis . Kymographs are 4 . 8 min long . Cells were imaged in isotonic media ( top ) or hypotonic media for 2 min ( middle ) or 10 min ( bottom ) . ( D ) Cumulative distribution plot of clathrin lifetimes marked by CLTA - TagRFP - T EN in isotonic media ( red ) , hypotonic media for 2 min ( violet ) , and hypotonic media for 10 min ( orange ) . These tracks were associated with DNM2 - eGFP EN . ( E ) Cumulative distribution plot of dynamin2 lifetime marked by DNM2 - eGFP EN in isotonic media ( light green ) , hypotonic media for 2 min ( blue ) , and hypotonic media for 10 min ( dark green ) . These tracks were associated with CLTA - TagRFP - T EN . n = 5831 tracks in 17 cells across four experiments for D \u2013 H . ( D , E ) Detailed statistics in Supplemental Table S1 . ( F ) Plot of endocytic initiation rate for the three conditions . p < 0 . 05 by two - tailed Mann \u2013 Whitney test for both comparisons . ( G ) Endocytic completion rate in the three conditions . p < 0 . 05 by two - tailed Mann \u2013 Whitney for both comparisons . ( H ) Percentage of persistent tracks ( defined as tracks lasting the entirety of the image acquisition ) for the three conditions . p < 0 . 05 by two - tailed Mann \u2013 Whitney for both comparisons . ( F \u2013 H ) Barplots show mean \u00b1 SD . Statistics to plots in Supplemental Figure S7H . Volume 33 May 15 , 2022 Actin load adaptation in endocytosis | 7 FIGURE 3 : Importance of Arp2 / 3 complex \u2013 mediated actin polymerization during CME increases under elevated membrane tension . In this figure , CK666 ( Arp2 / 3 complex inhibitor ) treatment is 100 \u00b5M and hypotonic shock is 150 mOsm . ( A ) Kymographs of cells expressing CLTA - TagRFP - T EN ( magenta ) and DNM2 - eGFP EN ( green ) imaged by TIRF . Cells were imaged in isotonic media . The media was then exchanged to include CK666 ( bottom panel ) and imaged after 4 min . ( B ) Cumulative distribution plots of clathrin lifetimes in control , DMSO - treated conditions ( orange , n = 4219 ) , and CK666 - treated ( red , n = 3124 ) conditions . ( C ) Cumulative distribution plots for control , DMSO - treated ( dark green , n = 4219 ) , and CK666 - treated ( light green , n = 3124 ) dynamin2 lifetimes associated with clathrin lifetimes in B . ( BC ) Control , N = 10 cells , and CK666 treatment , N = 10 cells , measured in three independent experiments . Complete statistics in Supplemental Table S3 . ( D ) Kymographs of cells in hypotonic media . In the top panel , cells were placed in hypotonic media and imaged after 4 min . In the bottom panel , cells were treated with CK666 in hypotonic media and imaged after 4 min . ( E ) Cumulative distribution plots of clathrin lifetimes for control , DMSO - treated conditions ( magenta , n = 1405 ) , and CK666 - treated conditions ( blue , n = 2783 ) in hypotonic media . ( F ) Cumulative distribution plots of DMSO - treated ( black , n = 1405 ) and CK666 - treated ( olive , n = 2783 ) dynamin2 lifetimes in hypotonic media associated with clathrin lifetimes in E . ( E , F ) Control , N = 9 cells , and CK666 treatment , N = 10 cells , measured in three independent experiments . Complete statistics in Supplemental Table S3 . ( G ) Representative STORM images of immunolabeled clathrin - coated structures in control cells arranged by coat height . Top panel shows the x - y projections and bottom panel the corresponding x - z projections . The white square in the x - y projections shows the area that was cropped to generate the x - z projections . The heights of clathrin coats in the x - y projection from left to right image are 61 , 96 , 124 , 158 , and 177 nm . Scale bars : 100 nm . ( H ) Clathrin coat heights when cells were treated with DMSO ( n = 154 ) or CK666 ( n = 158 ) in isotonic media or CK666 in hypotonic media ( n = 159 ) . Clathrin coat images for quantitative analysis were collected from at least three different cells for each condition from a single experiment . Statistics are given in Supplemental Figure S8F . p < 0 . 05 in both comparisons by Mann \u2013 Whitney test . 8 | C . Kaplan et al . Molecular Biology of the Cell an accumulation of smaller shape indices and a very mild effect on clathrin coat width ( Supplemental Figure S8H ) . This result , along with our observation of slowed CME dynamics , suggests that upon membrane tension elevation in response to hypotonic conditions in SK - MEL - 2 cells , inhibition of Arp2 / 3 complex activity may cause an enrichment of shallow endocytic coat geometries . Actin organization adapts to elevated membrane tension by increasing clathrin coat coverage Finally , we used STORM to determine the relationship between el - evated membrane tension and actin cytoskeleton organization at CME sites ( Figure 4 ) . We treated cells with strong ( 75 mOsm ) hypo - tonic shock for 5 min and then chemically fixed them for 2c - 3D STORM . When we superresolved clathrin and actin by 2c - 3D STORM in these cells , the actin cytoskeleton remained intact ( Pan et al . , 2019 ) and associated with CCSs ( Supplemental Figure S9 , A \u2013 D ) . Strikingly , in response to hypotonic media treatment , the aver - age actin height increased for all endocytic geometries ( Figure 4 , B , C , and E ; Supplemental Figure S9 , E \u2013 G ) . When we quantified actin cell cortex height in randomly selected regions of the cell cortex , we did not observe any increase in cortical actin height , reinforcing the conclusion that any observed effects are specific for CME sites ( Sup - plemental Figure S10 , A \u2013 J and P \u2013 R ) . Overall , the average actin height at CME sites increased from 130 \u00b1 30 nm under isotonic me - dia conditions to 160 \u00b1 40 nm ( Figure 4C ) . This increase corre - sponded to an increase of actin growth over the clathrin coat from covering 66 \u00b1 23 % of the coat to covering 73 \u00b1 21 % under 75 mOsm hypotonic media ( Figure 4D ) . The increase in actin height was ob - served for all coat shapes ( e . g . , shallow or curved pits ) ( Figure 4E ) . Similarly , for different extents of clathrin coverage by actin , the aver - age actin height was greater in hypotonic conditions across different levels of coat coverage ( Figure 4F ) . However , the extent of asym - metry between actin and clathrin signals did not significantly change compared with that in the isotonic condition ( Supplemental Figure S9H ) . These observations of greater average actin height and cover - age over the clathrin coat suggest that the force contribution of ac - tin at CME sites is increased when membrane tension is elevated such that a greater load is carried by the actin network . Overall , these observations led us to conclude that under ele - vated plasma membrane tension , actin grows higher in the z - dimen - sion around clathrin coats ( Figure 4G ) . Such an adaptive mechanism for actin organization presumably generates the required forces to ensure the efficient progression of mammalian CME under various levels of membrane tension . DISCUSSION By combining 2c - 3D STORM imaging and quantitative live - cell TIRF microscopy for thousands of CME sites , with AFM tether pulling membrane tension measurements , and manipulations of membrane tension and Arp2 / 3 - mediated actin assembly , we showed that actin assembly adapts to changes in load at individual CME sites . This mechanism likely ensures that sufficient forces are generated for ro - bust endocytic progression over a range of membrane tension re - gimes . While STORM of individual CCSs cannot attain the resolution of EM , our approach had several advantages that allowed us to gain new mechanistic insights : 1 ) it allowed us to sample much larger numbers of CME sites with a preserved actin network than is possi - ble by EM , thus permitting rigorous quantitative analysis , 2 ) we im - aged CME sites and associated actin in intact cells that had not been subjected to unroofing or extraction protocols , and 3 ) we were able to use antibodies and fluorescent phalloidin to unambiguously identify specific proteins at CME sites . Actin assembly and organization adapt to elevated membrane tension Elevating membrane tension can have a dramatic impact on CME progression in mammalian cells ( Raucher and Sheetz , 1999 ; Boulant et al . , 2011 ; Ferguson et al . , 2016 , 2017 ; Willy et al . , 2017 ; Bucher et al . , 2018 ) . However , how cells adapt to compensate for increased load at CME sites has not been elucidated . Our results provide criti - cal mechanistic insights into how the CME machinery adapts to el - evated membrane tension to maintain robust CME progression . We showed quantitatively that actin assembly and organization adapt to changes in membrane tension , which we measured by AFM mem - brane tether pulling ( Figure 2 ) . Changes in membrane tension can in principle occur globally ( entire cells ) or locally ( in different regions of one cell , or even within different regions of a single endocytic site ) ( Houk et al . , 2012 ; Rangamani et al . , 2014 ; Shi et al . , 2018 ) . Because we detect different actin organizations at individual CME sites with similar clathrin coat geometries within a single cell , such differences might reflect subcellular , local load variation resulting from variance in such factors as membrane tension and cell adhesion . The obser - vation in other studies that actin assembles late in CME progression ( Merrifield et al . , 2002 ; Grassart et al . , 2014 ; Akamatsu et al . , 2020 ; Jin et al . , 2021 ) , coupled with the observation here that actin can associate with clathrin coats of different geometries , is consistent with the possibility that clathrin lattices of low curvature can develop high curvature under the \u201cconstant area\u201d model ( Avinoam et al . , 2015 ; Sochacki et al . , 2021 ) . Moreover , our data provide evidence that the low to high coat curvature transition becomes more depen - dent on actin assembly forces as load increases . Measurement of membrane tension changes using AFM . Using AFM , we measured the membrane tether force for SK - MEL - 2 cells in isotonic and hypotonic conditions ( Figure 2 ) . We measured tension on the dorsal cell surface while imaging endocytic events on the ventral surface . Local differences in membrane tension have been observed in some cell types ( Shi et al . , 2018 ) but not others ( Houk et al . , 2012 ) . Nevertheless , AFM remains the gold standard for mea - suring apparent membrane tension in cells ( Sitarska and Diz - Mu\u00f1oz , 2020 ) . In isotonic conditions , the tether force was 33 \u00b1 7 pN ( Figure 2B ) . This tether force is within an intermediate range measured for other cell types such as NIH3T3 cells , HeLa cells , and macrophages ( Sitarska and Diz - Mu\u00f1oz , 2020 ; Roffay et al . , 2021 ) . Assuming a 100 pN * nm bending rigidity of the plasma membrane , this value corresponds to a membrane tension of 0 . 14 \u00b1 0 . 04 pN / nm ( Diz - Mu\u00f1oz et al . , 2013 ) . In hypotonic conditions , the membrane tether force increased to 48 \u00b1 17 pN , which corresponds to a dou - bling of membrane tension to 0 . 29 \u00b1 0 . 04 pN / nm . Higher mem - brane tether values have been reported for other cell types ( Sitarska and Diz - Mu\u00f1oz , 2020 ) . Below , we describe implications of earlier results and our new results reported here on actin\u2019s role in CME un - der three different plasma membrane tension regimes ( Figure 4G ) . Low - membrane - tension regime . When membrane tension is low , clathrin coat assembly provides sufficient energy to bend the under - lying plasma membrane into a full spherical shape , similar to the constant - curvature model ( Figure 4G ) ( Saleem et al . , 2015 ; Willy et al . , 2021 ) . We indeed found in our STORM data that 26 % of clath - rin coats lack actin in shallow , U - shape and \u03a9 - shape CME geome - tries . This observation is consistent with mathematical modeling , which indicates that the coat can provide sufficient energy to bend the plasma membrane when membrane tension is low ( 0 . 002 pN / nm ) ( Hassinger et al . , 2017 ) . Here , we consider low membrane tension to be of a value lower than those we measured for SK - MEL - 2 cells in Volume 33 May 15 , 2022 Actin load adaptation in endocytosis | 9 FIGURE 4 : The actin network at CME sites increases in size in response to elevated membrane tension . In this figure , hypotonic refers to 75 mOsm media . ( A ) Schematic of cells in hypotonic media , which increases plasma membrane tension . The response of the actin network ( blue ) to elevated plasma membrane tension ( purple ) was previously unknown . ( B ) Representative STORM images of clathrin ( red ) and actin ( cyan ) in x - z projections for cells fixed after treatment in the hypotonic media for 5 min ( bottom ) . Coated pits are classified as shallow , U - shaped , or \u03a9 - shaped based on the aspect ratio of the coat . Scale bars : 100 nm . ( C ) Plots of actin Z height at CCPs from cells in the isotonic ( n = 736 ) and hypotonic ( n = 527 ) media measured from STORM x - z projections . Lines are median \u00b1 interquartile range . p < 0 . 0001 determined by Mann \u2013 Whitney test . ( D ) Plots of actin coverage over the clathrin coat in pits found in STORM x - z projection images in isotonic ( n = 719 ) and hypotonic ( n = 509 ) conditions . Pits with actin coverage > 1 % are plotted . Lines are median \u00b1 interquartile range . p < 0 . 0001 determined by Mann \u2013 Whitney test . ( E ) Actin Z height as a function of coat shape in isotonic ( gray , n = 736 ) and hypotonic ( purple , n = 527 ) conditions . ( F ) Actin Z height as a function of actin coverage over the clathrin coat in isotonic ( gray , n = 719 ) and hypotonic ( purple , n = 509 ) conditions . The data for isotonic conditions were also used to generate the plots in Figure 1 . Three independent STORM experiments with N _ = 6 cells in isotonic and N _ = 7 cells in hypotonic media . ( G ) Cartoon depicting an adaptive actin force \u2013 generating mechanism that counteracts elevated membrane tension to ensure robust CME progression . This schematic describes three scenarios in which membrane tension is low , intermediate , or high and how CME is accomplished efficiently by adaptive actin network organization . Under low tension ( bottom ) , the clathrin coat provides sufficient force to advance CME . At intermediate tension ( middle ) , actin polymerization is required for the transition from U to \u03a9 shape . At high tension ( top ) , endocytic progression slows . More pits stall at the shallow conformation . In response to increased resistance , the actin network grows to envelop the coat and provide additional force against high membrane tension . 10 | C . Kaplan et al . Molecular Biology of the Cell standard media ( Figure 2 ) . Given that actin polymerization appears to be dispensable for CME in some cell types , even though it still might make CME more efficient , we hypothesize that the basal membrane tension may be lower in those cell types . However , we urge caution with the interpretation of experiments with harsher actin drug treat - ments , as prolonged actin inhibitor treatment of cells can perturb the actin cortex and reduce the tether force measurements ( membrane tension ) by a factor of \u223c 50 % ( Pontes et al . , 2017 ) . In future studies , strategies should be developed to relate local membrane tension to specific endocytic site geometry and actin organization . Intermediate tension regime . In the intermediate tension regime , defined as the resting tension of SK - MEL - 2 cells in isotonic condi - tions ( Figure 2 ) , clathrin coat assembly and membrane curvature \u2013 in - ducing proteins still appear to provide sufficient energy to drive clathrin coats to adopt the U shape ( Figure 4G , intermediate mem - brane tension ) . When we inhibited Arp2 / 3 - mediated actin polymer - ization using CK666 , clathrin coat progression stalled at an aspect ratio most likely reflecting the U - shaped stage , consistent with the effects of actin assembly inhibition reported for other cell types ( Yarar et al . , 2005 ; Boulant et al . , 2011 ; Almeida - Souza et al . , 2018 ) . Thus , at intermediate membrane tension it appears that actin force generation is primarily required for the U - to \u03a9 - shaped clathrin coat transition . The actin network observed near the base of the pit may reflect a role driving plasma membrane neck constriction and scis - sion by generating forces orthogonal to the direction of membrane invagination as proposed previously ( Bucher et al . , 2018 ; Scott et al . , 2018 ; Mund et al . , 2021 ) . High - membrane - tension regime . Our live - cell and STORM obser - vations indicate that as membrane tension is elevated further , coat deformation and membrane invagination become increasingly de - pendent on actin force generation ( Figure 4 ) . When membrane ten - sion was elevated to an intermediate level ( e . g . , 150 mOsm media ) , CME lifetimes slowed modestly ( Supplemental Figure S7 ) . When we inhibited Arp2 / 3 - mediated actin polymerization using CK666 in cells treated with 150 mOsm media , the average clathrin coat height was lower than in CK666 - treated cells cultured under isotonic condi - tions , likely reflecting an enrichment of shallow pits ( Figure 3H ) . Un - der elevated membrane tension , we observed no change in actin cortex thickness , while others reported that the actually actin cortex tends to become thinner and less dense ( Supplemental Figure S10 ) ( Houk et al . , 2012 ; Chugh et al . , 2017 ; Roffay et al . , 2021 ) . However , our study shows that the actin network associated with CCPs instead increases in height under elevated membrane tension ( Figure 4 ) . Actin assembly from the base of the CCP continues until the net - work covers the clathrin coat completely , allowing it to interact with proteins linking the actin network to the clathrin coat ( Engqvist - Goldstein et al . , 2001 ; Messa et al . , 2014 ; Sochacki et al . , 2017 ) . Actin - binding linker proteins such as Hip1R and Epsin1 cover the clathrin coat completely and are thus positioned to provide high internalization efficiency by harnessing actin assembly forces per - pendicular to the plasma membrane ( Akamatsu et al . , 2020 ; Serwas et al . , 2021 ) . It may be that the \u201cconstant area\u201d model ( Avinoam et al . , 2015 ; Sochacki et al . , 2021 ) , in which a flat or shallow coat grows to full size and then bends , applies under the high load re - gime ( Bucher et al . , 2018 ; Scott et al . , 2018 ; Mund et al . , 2021 ) , and it is here that actin assembly forces are most important for remodel - ing the coat into a curved vesicle . An important question for future studies concerns the nature of the adaptive mechanism that increases actin assembly in response to elevated membrane tension . Possible mechanisms include the following : 1 ) stalling works passively to allow actin to assemble lon - ger ; or 2 ) high load works actively through mechanisms documented for other contexts , for example by increasing contact of filaments associated with the coated pit with membrane - associated N - WASP - Arp2 / 3 complex , leading to more assembly , alongside load - depen - dent decreases in the rate of filament capping ( Bieling et al . , 2016 ; Akamatsu et al . , 2020 ; Funk et al . , 2021 ; Li et al . , 2021 ) . As endocy - tosis and other actin - mediated trafficking events operate within a membrane geometry and local protein and lipid environment differ - ent from that in lamellipodia , a high priority is now to determine whether similar adaptive mechanisms operate during such local membrane bending processes . When the actin network fully covers the clathrin coat , it resem - bles the radial organization described by mathematical modeling for mammalian CME and the actin network organization described for budding yeast ( Ferguson et al . , 2009 ; Hassinger et al . , 2017 ; Mund et al . , 2018 ; Akamatsu et al . , 2020 ) . In yeast this actin organi - zation drives endocytic membrane invagination against the high resistance that results from turgor pressure . Mathematical modeling showed that this organization produces high forces perpendicular to the plasma membrane ( Hassinger et al . , 2017 ) . Actin - generated forces parallel and orthogonal to the membrane invagination at high tension may coexist to drive membrane invagination followed by neck constriction and scission . CME dynamics dramatically slow down when cells are in this high - membrane - tension regime , resulting in only \u223c 40 % of endocytic lifetimes shorter than 50 s compared with 80 % at low membrane tension ( Figure 2D ) . We also detected \u223c 19 % of lifetimes being lon - ger than the 4 . 8 min movies that we captured , a 20 - fold increase in stalled ( persistent ) tracks ( Figure 2H ) . We suggest that this tension regime pushes this adaptive mechanism to the limit . N - WASP spatial organization suggests an actin force generation control mechanism N - WASP spatial organization at CCSs and CCPs provides valuable mechanistic insight into how actin network assembly contributes to force generation during CME . We found that N - WASP localizes at the base of early and late CCPs , where it likely interacts with SH3 domain \u2013 containing proteins present at the endocytic membrane neck ( Sch\u00f6neberg et al . , 2017 ; Sochacki et al . , 2017 ; Almeida - Souza et al . , 2018 ) . This organization is similar to that of the ho - mologous nucleation - promoting factor Las17 in budding yeast ( Mund et al . , 2018 ) . Filaments nucleated at the base of CCPs would be able to interact with coat proteins such as Hip1R and Epsin1 / 2 / 3 to generate forces to invaginate the plasma membrane ( Hassinger et al . , 2017 ; Mund et al . , 2018 ; Akamatsu et al . , 2020 ; Joseph et al . , 2020 ) . Intriguingly , we also sometimes observed a strikingly different N - WASP spatial organization in which it was distributed over the full clathrin coat . The type II nucleation factors Abp1 and cortactin bind to actin filaments and to the Arp2 / 3 complex and could serve as binding partners for N - WASP around the CCP ( Le Clainche et al . , 2007 ; Pinyol et al . , 2007 ; Helgeson and Nolen , 2013 ; Guo et al . , 2018 ) . WASP located on the clathrin coat might reflect a distinct path for filament nucleation , originating from the coat rather than the base , that is potentially important to generate higher forces when actin already surrounds the CCP . A late burst of actin assembly was shown previously to often accompany CME ( Merrifield et al . , 2002 ) and to facilitate CME progression , especially when membrane tension is high ( Fujimoto et al . , 2000 ; Yarar et al . , 2005 ; Batchelder and Yarar , 2010 ; Boulant et al . , 2011 ; Grassart et al . , 2014 ; Li et al . , 2015 ; Ferguson et al . , 2017 ; Volume 33 May 15 , 2022 Actin load adaptation in endocytosis | 11 Yoshida et al . , 2018 ; Akamatsu et al . , 2020 ; Jin et al . , 2021 ) . In this study we observed that load - adapted , CME - associated actin assem - bly buffers changes in plasma membrane tension , thereby ensuring clathrin - coated vesicle ( CCV ) formation over a range of membrane tension regimes . We expect that the ability of actin assembly to re - spond to load is a common feature critical to many membrane re - modeling processes . MATERIALS AND METHODS Request a protocol through Bio - protocol . Cell culture SK - MEL - 2 cells from clone Ti13 ( hCLTA EN \u2013 1 / hDNM2 EN \u2013 1 ) were cul - tured in DMEM / F12 with GlutaMax supplement ( 10565 - 018 ; Thermo Fisher Scientific ) containing 10 % fetal bovine serum ( FBS ) and a 1000 U / ml penicillin \u2013 streptomycin mix ( 15140122 ; Thermo Fisher Scientific ) and kept in a 37\u00b0C humidified incubator with 5 % CO 2 ( cell source information in Doyon et al . , 2011 ) . After each cell vial was thawed , cells were checked after two passages for myco - plasma contamination . Cell line authentication was performed by short tandem repeat validation . Antibodies and reagents The primary antibodies used were mouse anti - clathrin light chain ( AB CON . 1 , MA5 - 11860 ; Thermo Fisher Scientific ) , mouse anti - clathrin heavy chain ( AB X - 22 , MA1 - 065 ; Thermo Fisher Scientific ) , and rabbit anti \u2013 N - WASP ( ab126626 ; Abcam ) . The secondary anti - bodies used were Alexa Fluorophore 647 chicken anti - rabbit ( A21443 ; Thermo Fischer Scientific ) , goat anti - mouse ( 115 - 005 - 205 ; Jackson ImmunoResearch ) conjugated to CF680 - NHS ester ( Bio - tium ; 92139 ) . Reagents and small molecule inhibitors used were DMSO ( D2650 ; Sigma Aldrich ) , CK666 ( SML0006 , batch # 0000012761 ; Sigma Aldrich ) , and phalloidin - AF647 ( A22287 ; Fisher Scientific ) . Preparation of CF680 - labeled secondary goat anti - mouse antibody CF680 NHS ester was dissolved at a concentration of 3 mM in anhy - drous DMSO . One microliter of dye solution , 80 \u00b5l of a 1 . 25 mg / ml suspension of unlabeled goat anti - mouse immunoglobulin G1 sec - ondary antibody ( 115 - 005 - 205 ; Jackson ImmunoResearch Labora - tories ) , and 10 \u00b5l of 1 M sodium bicarbonate solution were mixed and allowed to react for 15 min at room temperature . The reaction mixture was added to an equilibrated NAP - 5 column ( Sigma ; GE17 - 0853 - 01 ) and flushed with phosphate - buffered saline ( PBS ) . The dye - conjugated antibody was collected from the first colored eluent fraction , and a concentration of 0 . 12 mg / ml was determined with a NanoDrop spectrophotometer . Sample preparation for two - color clathrin and actin imaging Round coverslips ( 18 mm ) were cleaned 20 min in 70 % ethanol ( Electron Microscopy Science ; Cat # 72222 - 01 ) . Cells were detached with 500 \u00b5l 0 . 05 % trypsin ( 25300 - 054 ; Life Technologies ) , washed once in DMEM / F12 , and collected by centrifugation . Cells were counted using a hemocytometer , and 20 , 000 cells / ml were seeded onto 18 mm round coverslips in 12 - well plates . Cells were incubated for 16 \u2013 24 h in culture media before preparation for imaging . Cells were fixed first for 1 \u2013 2 min in 0 . 3 % ( vol / vol ) glutaraldehyde ( GA ) solution containing 0 . 25 % ( vol / vol ) Triton in cytoskeleton buffer ( CB : 10 mM MES , 150 mM NaCl , 5 mM ethylene glycol - bis ( 2 - aminoethylether ) - N , N , N \u2032 , N \u2032 - tetraacetic acid ( EGTA ) , 5 mM glucose , 5 mM MgCl 2 , 0 . 005 % NaN 3 , pH 6 . 1 ) and then immediately fixed for 10 min in 2 % ( vol / vol ) GA solution in CB . Both solutions were pre - pared fresh from a 70 % GA stock ( Electron Microscopy Science ; cat # 16365 ) ( protocol follows Xu et al . , 2012 ) . After fixation , samples were washed once in CB and then incubated for 7 min in freshly prepared CB containing 0 . 1 % ( wt / vol ) NaBH 4 . Subsequently , sam - ples were washed three times for 10 min in CB with gentle agitation on a shaker . Samples were then blocked for 30 min in 5 % ( wt / vol ) bovine serum albumin ( BSA ) in CB ( Sigma Aldrich ; A3733 ) . For dense clathrin labeling , light ( diluted 1 : 200 ) and heavy ( diluted 1 : 200 ) chain antibodies were used together in a 1 % ( wt / vol ) BSA CB solution . Primary antibody immunostaining was performed over - night at 4\u00b0C . The next day , samples were washed twice in 1 % ( wt / vol ) BSA CB for 5 min . The mouse secondary antibody CF680 was used at a final concentration of 0 . 40 \u2013 0 . 60 \u00b5g / ml in a 1 % BSA \u2013 1 \u00d7 CB solution . Samples were stained for 30 min at room temperature in the dark and washed twice for 5 min in 1 % ( wt / vol ) BSA CB solution and then for 10 min in CB solution . Samples were then placed into a solution of CB containing 0 . 5 \u00b5M phalloidin - AF647 and kept at room temperature in the dark for a minimum of 2 h . Samples were washed once with PBS before STORM imaging . Sample preparation for single - color clathrin and dual - color N - WASP imaging Cells were prepared as for the two - color sample preparation on cov - erslips and then fixed for 20 min in 3 % ( vol / vol ) paraformaldehyde ( PFA ; 15710 ; Electron Microscopy Sciences ) in CB ( protocol follows Li et al . , 2018 ) . Samples were washed quickly in CB and subse - quently were incubated for 7 min in freshly prepared 0 . 1 % ( wt / vol ) NaBH 4 in CB solution . Subsequently , samples were washed three times for 10 min in CB with gentle agitation on a shaker and per - meabilized afterward in a 0 . 1 % Triton - PBS solution for 1 \u2013 2 min . For single - antibody clathrin staining , subsequent washing , blocking , and antibody incubation steps were similar to those of the two - color clathrin and actin sample preparation protocol . Dual - color immunolabeling was performed with primary anti - body against N - WASP ( diluted 1 : 200 ) and clathrin heavy and clath - rin light chain ( diluted 1 : 600 \u2013 1 : 1000 ) in 1 % ( wt / vol ) BSA in PBS over - night at 4\u00b0C . Samples were washed the next day twice for 5 min in 1 % ( wt / vol ) BSA in PBS . Secondary antibody staining was first per - formed with Alexa Fluorophore 647 anti - rabbit antibody ( diluted 1 : 200 ) in 1 % BSA ( wt / vol ) in PBS for 30 min at room temperature and kept in the dark . After two 10 - min - long washes in PBS containing 1 % ( wt / vol ) BSA , secondary antibody staining was performed with CF680 anti - mouse antibody ( diluted 1 : 600 ) . The samples were given three final washes in PBS for 10 min each . SRM imaging Dye - labeled cell samples were mounted on glass slides in a standard STORM imaging buffer consisting of 5 % ( wt / vol ) glucose , 100 mM cysteamine , 0 . 8 mg / ml glucose oxidase , and 40 \u00b5g / ml catalase in 1 M Tris - HCl ( pH 7 . 5 ) ( Rust et al . , 2006 ; Huang et al . , 2008 ) . Cover - slips were sealed using Cytoseal 60 . STORM imaging was per - formed on a homebuilt setup ( Wojcik et al . , 2015 ) based on a modi - fied Nikon Eclipse Ti - U inverted fluorescence microscope using a Nikon CFI Plan Apo \u03bb 100 \u00d7 oil immersion objective ( NA 1 . 45 ) . Dye molecules were photoswitched to the dark state and imaged using a 647 - nm laser ( MPB Communications ) ; this laser was passed through an acousto - optic tunable filter and introduced through an optical fiber into the back focal plane of the microscope and onto the sample at an intensity of \u223c 2 kW cm 2 . A translation stage was used to shift the laser beam toward the edge of the objective so the light reached the sample at incident angles slightly smaller than the 12 | C . Kaplan et al . Molecular Biology of the Cell critical angle of the glass \u2013 water interface . A 405 - nm laser was used concurrently with the 647 - nm laser to reactivate fluorophores into the emitting state . The power of the 405 - nm laser ( typical range 0 \u2013 1 W cm \u2013 2 ) was adjusted during image acquisition so that at any given instant , only a small , optically resolvable fraction of the fluorophores in the sample was in the emitting state . For 3D STORM imaging , a cylindrical lens was inserted into the imaging path so that images of single molecules were elongated in opposite directions for mole - cules on the proximal and distal sides of the focal plane ( Huang et al . , 2008 ) . The raw STORM data were analyzed according to pre - viously described methods ( Rust et al . , 2006 ; Huang et al . , 2008 ) . Data were collected at a frame rate of 110 Hz for a total of \u223c 80 , 000 frames per image . Single - and two - color imaging was performed on cells labeled with Alexa Fluor 647 only or Alexa Fluor 647 and CF680 with 647 - nm excitation based on a ratiometric detection scheme ( Bossi et al . , 2008 ; Testa et al . , 2010 ; Gorur et al . , 2017 ) . In the two - color imaging scheme , light emitted from the AF647 and CF680 fluorophores was collected concurrently and split into two light paths using a long - pass dichroic mirror ( T685lpxr ; Chroma ) . Each light path was projected onto half of an Andor iXon Ultra 897 EM - CCD camera . Dye assignment was performed by localizing and recording the intensity of each single molecule in each channel . Conventional imaging of 560 - and 488 - nm dyes was performed im - mediately before STORM imaging using the appropriate laser and filter set . Emission data were collected through the short - wave - length reflected path of the aforementioned optical setup and over - laid directly onto the final STORM image . Details of selection and analysis of SRM images are found in the Supplemental Information . Our 3D - STORM setup was based on the same design as in Huang et al . ( 2008 ) and so we achieved comparable resolutions . The experimental STORM resolution was measured by repeatedly measuring the position of a single fluorophore and determining the SD of the localization distribution ( Huang et al . , 2008 ; Xu et al . , 2015 ) . We accordingly examined our STORM data in this work and overlaid the localization distributions of 24 representative single molecules from three different samples , as shown ( Supplemental Figure S1 ) . Gaussian fits ( red curves ) gave SDs of 10 nm in plane for the XY directions and 19 nm in depth for the Z direction . These re - sults are similar to those reported in Figure 1C of Huang et al . ( 2008 ) , where SDs are 9 nm in X , 11 nm in Y , and 22 nm in Z . TIRF microscopy TIRF imaging was carried out on a Nikon Eclipse Ti2 inverted micro - scope with a CFI60 60 \u00d7 Apo TIRF objective and a Hamamatsu Orca - Flash 4 . 0 V2 sCMOS camera . eGFP and Tag . RFP - T fluorescence was excited using 488 and 561 nm lasers and detected using a Chroma HC TIRF Quad Dichroic ( C - FL TIRF Ultra Hi S / N 405 / 488 / 561 / 638 ) and Chroma HC Quad emission filters BP 525 / 50 and BP600 / 50 , respectively ( Bellows Falls , VT ) . Unless mentioned specifically , chan - nels were acquired sequentially at a 1 . 2 s interval and 400 ms expo - sure time over 4 . 8 to 6 min . Real - time acquisition was achieved by a National Instruments ( PXI 1033 ; Austin , TX ) controller . The system was controlled with NIS - Elements software and maintained at 37 \u00b0 C by an OkoLab environmental chamber ( Burlingame , CA ) . Hypo - osmotic media treatment SK - MEL - 2 cells were plated on glass coverslips 1 d before osmotic treatment and imaging : 20 , 000 cells / m were seeded 16 \u2013 24 h before the experiment on 25 mm round # 1 . 5 glass coverslips that had been cleaned with 70 % ethanol ( Warner Instruments ; 64 - 0715 ) . Isotonic imaging media contained DMEM and Ham\u2019s F - 12 medium ( DMEM / F12 ) without phenol red ( 11039 ; Thermo Fisher Scientific ) with 5 % vol / vol FBS . The media was diluted with an inorganic salt solution containing 10 mM CaCl 2 0 . 3 mM MgCl 2 and 0 . 1 mM MgSO 4 ( CMM ) to maintain concentrations of critical ions , while obtaining hypo - os - motic conditions by diluting the media containing components such as d - glucose . The 225 mOsm hypotonic imaging media contained 1 : 4 vol / vol CMM solution in DMEM / F12 , the 150 mOsm hypotonic imaging media contained 1 : 1 vol / vol CMM solution in DMEM / F12 , and the 75 mOsm hypotonic imaging media contained 4 : 1 vol / vol CMM solution in DMEM / F12 . Five percent vol / vol FBS was present in all hypotonic solutions . For live - cell fluorescence microscopy CLTA - TagRFP - T EN and DNM2 - eGFP EN fluorescence in SK - MEL - 2 cells was acquired first in isotonic media over a course of 4 . 8 min . Subsequently , media was exchanged on the stage to hypotonic media ( either 225 , 150 , or 75 mOsm ) and movies were acquired for 4 . 8 min , starting 2 and 10 min after media exchange . Media exchange on the stage did not affect CME initiation rates or fluores - cence lifetimes beyond the existing experimental intrinsic variability ( Supplemental Figure S7 , A and B ) . For STORM imaging , 75 mOsm hypotonic buffer treatment was performed in the cell culture dish for 5 min . Cells were immediately chemically fixed after the 5 min treatment and further treated with the STORM sample preparation protocol as described above . Methods to analyze TIRF data can be found in the Image analysis section . CK666 concentration titration SK - MEL - 2 cells ( 20 , 000 / ml ) were seeded in eight - well chambers 16 h\u201424 h before the experiment ( 80826 ; ibidi , Fitchburg , WC ) . A CK666 ( SML0006 , batch # 0000012761 ; Sigma Aldrich ) stock solu - tion was prepared at 50 mM in DMSO and kept at \u2013 20\u00b0C . CK666 ( 25 , 50 , and 100 \u00b5M ) and equivalent 0 . 5 % vol / vol DMSO , 1 % vol / vol DMSO , and 2 % DMSO vol / vol solutions for controls were prepared fresh in DMEM / F12 containing 5 % FBS and kept at 37\u00b0C until used . Cells were first imaged in DMEM / F12 containing 5 % FBS solution as a baseline control for 4 . 8 min . Subsequently , imaging solution was exchanged on the microscopy stage to CK666 - or DMSO - contain - ing imaging solution and another 4 . 8 - min movie was acquired after 2 min of treatment . Each treatment was repeated twice , and an area of 1024 pixels \u00d7 1024 pixels was used to record 3 \u2013 6 cells per experiment . CK666 in combination with hypo - osmotic media Cells were prepared as for the CK666 concentration titration experi - ment described above . Solutions of 2 % vol / vol DMSO in DMEM / F12 , 100 \u00b5M CK666 in DMEM / F12 , 2 % vol / vol DMSO in 1 : 1 vol / vol CMM solution in DMEM / F12 ( 150 mOsm hypotonic media ) , and 100 \u00b5M CK666 1 : 1 vol / vol CMM solution in DMEM / F12 ( 150 mOsm hypotonic media ) were prepared fresh and kept at 37\u00b0C until used . All solutions contained 5 % FBS . Cells were first imaged in DMEM / F12 - 5 % FBS solution as a baseline control for 6 min . Subsequently , the imaging solution was exchanged on the microscopy stage to the desired experimental solutions and a 6 - min movie was recorded af - ter 4 min of incubation . Tether pulling experiments using AFM Custom - cut 35 - mm glass - bottom dishes ( Greiner Bio - One ; # 627860 ) were coated with fibronectin ( 50 \u00b5g / ml ; Corning # 356008 ) for 30 min and washed with Dulbecco\u2019s phosphate buffered saline ( DPBS ) shortly before use . SK - MEL - 2 cells were seeded at a density of 0 . 15 \u2013 0 . 20 \u00d7 10 5 cells / ml in DMEM / F12 GlutaMax supplement Volume 33 May 15 , 2022 Actin load adaptation in endocytosis | 13 with 1 % FBS and a penicillin \u2013 streptomycin mix ( Gibco ; # 15140 - 122 ) in a 37\u00b0C humid incubator with 5 % CO 2 for 2 \u2013 4 h and used directly for membrane tether pulling experiments . OBL - 10 cantilevers ( Bruker ) were mounted on a CellHesion 200 AFM ( Bruker ) integrated into an Eclipse Ti inverted light microscope ( Nikon ) , calibrated using the thermal noise method and coated with 2 . 5 mg / ml concanavalin A ( C5275 ; Sigma ) for 1 h at 30\u00b0C . After the cantilever was rinsed with DPBS , it was positioned at any location over the cell for tether pull - ing using bright - field imaging . The approach velocity was set to 1 \u00b5m / s , contact force to 100 \u2013 300 pN , contact time to 300 ms \u2013 10 s , and retraction speed to 10 \u00b5m / s . After a 10 \u00b5m tether was pulled , the cantilever position was held constant until the moment of tether breakage and at least 2 s afterward . The sampling rate was set to 2000 Hz . After measurements of tether forces in control conditions , an inorganic salt solution containing 10 mM CaCl 2 , 0 . 3 mM MgCl 2 , and 0 . 1 mM MgSO 4 was added to the medium ( 4 : 1 vol / vol ) to achieve 75 mOsm hypotonic treatment . Tether forces were mea - sured after media dilution for 2 \u2013 16 min . Tether forces per cell are the average of at least three tethers . Cells were not used longer than 1 h for data acquisition . Force - time curve analysis was performed us - ing JPKSPM Data Processing Software . Data analysis , statistical analysis , and data plotting For statistical analysis , data plotting , and image analysis , Prism ver - sion 7 . 0e and python module matplotlib , numpy , pandas modules , and scikit library in Jupyter notebook ( Python 3 . 8 . 3 ) were used . For some parts of the image analysis we used FIJI / ImageJ ( version 2 . 1 . 0 / 1 . 53c ) . Image analysis Selection of clathrin - coated superresolved structures for image analysis . Clathrin - containing structures were extracted from pro - cessed STORM images using a custom MATLAB routine . Briefly , a kernel convolution with a disk of \u223c 80 nm radius was performed on the superresolved clathrin image channel to accentuate possible CCPs . The resulting image was median filtered , and peaks were de - tected by sequentially identifying and cropping out regions corre - sponding to the local image maxima . Regions 310 nm \u00d7 310 nm wide square centered at these peaks were cropped from all color channels and aligned . Subsequently , we selected superresolved clathrin coats by visual inspection for further quantitative image analysis based on the following criteria : We excluded clathrin - la - beled structures for image analysis that appeared deformed , that covered nearly the entire 310 nm \u00d7 310 nm wide square , or that contained small punctuated structures that were not distinguishable from background noise or small clathrin seeds ( Supplemental Figure S4A ) . Clathrin - coated structures selected for analysis were first in - spected to determine whether they appeared round - shaped , ellipti - cal - shaped , or triangle - shaped in the x - y projected superresolved image . These images were 3D rendered to determine whether the x - z projection resulted in the typical wide - crescent shape , U - shape , or round - shape appearance for the clathrin coat . If we could identify the stereotypical clathrin - coat shapes in both projections , we in - cluded the clathrin coat in the pool for further image analysis . We define stereotypical shapes of individual clathrin coats based on platinum replica EM data of CCPs in the same cell type ( Supplemen - tal Figure S3 ) ( courtesy of Justin Taraska ) . We then assigned the clathrin coats as being actin negative when the number of localiza - tions in the actin channel in the 310 nm \u00d7 310 nm region of interest ( ROI ) was below 50 , which is close to the background level . Clathrin - coated structures were classified as actin positive when the number of localizations in the actin channel in the 310 nm \u00d7 310 nm ROI was above 50 and actin signal overlapped with the clathrin signal ( Supplemental Figure S9 , E \u2013 G ) . STORM image data display in figures . Reconstructed superreso - lution images were visualized using the \u201cinsight\u201d software package and saved in a . png file format ( Huang et al . , 2008 ) . These images were loaded into ImageJ , converted from an RGB image into an 8 bit image , pseudocolored for clathrin = red hot lookup table , actin = cyan lookup table , and then converted back into an RGB image . Image analysis of reconstructed clathrin and actin side views . Se - lected reconstructed superresolved images of clathrin and actin were treated as digital images for further analysis . A custom - written Jupyter notebook script was used to project reconstructed side views of clathrin and actin onto their x - and z - axes to obtain histo - grams of the clathrin and actin pixel intensity signals normalized to their respective maximum intensity value ( Supplemental Figure S4 , B and C ) . From these normalized z - axis and x - axis pixel intensity histograms , we read out the height and width of clathrin and actin at the 30th percentile . This process resulted in more robust results than the readout at the full width at half maximum ( Supplemental Figure S4 , B and C ) . Actin and clathrin coat height : The z - axis intensity histograms were used to report the actin and clathrin coat heights in the x - z projections . Before extracting the values , a mean filter set to 50 was run over the histograms to reduce signal fluctuations that interfere with the size measurement ( Supplemental Figure S4 , B and C ) . Actin - clathrin overlap calculation : We calculated the total over - lap between clathrin and actin using the indices obtained at the upper position of clathrin and lower position of actin at the 30th percentile of the respective z axis \u2013 projected intensity histograms ( Supplemental Figure S4B ) . We then reported the overlap relative to the clathrin height in units of percentage . Clathrin - coat width : We used the 30th percentile of x - axis inten - sity histogram to report the clathrin coat width in the x - z projection . Before extracting the values , a median filter set to100 was run over the histogram to smooth out signal fluctuations that interfere with the correct size determination ( Supplemental Figure S4 , B and C ) . Shape index calculation : Shape indices of individual clathrin structures are defined as the ratio between clathrin coat height and clathrin coat width . This value allowed us to sort flat ( low shape in - dex ) and rounded ( high shape index ) clathrin coats from each other . Asymmetry of actin signal around clathrin coat : To evaluate the asymmetry of the spatial actin organization around the clathrin coat , we determined the difference in the positions of the peak actin and clathrin signals on both x - z and y - z projections of our images ( Sup - plemental Figure S4E ) . We obtained the center of the clathrin coat in nanometers by rendering the clathrin superresolved image into a diffraction - limited image , as explained in the section Selection of clathrin - coated superresolved structures for image analysis . We first identified the position of the actin maximum intensity in the x - z pro - jection profile and y - z projection profile in nanometers . Then we measured the distance of the obtained position to the middle posi - tion of the clathrin intensity profile . This distance measurement is proportional to the asymmetry of the actin position with respect to clathrin ; namely , a low distance corresponds to high symmetry and a high distance corresponds to high asymmetry in the position of the actin signal . Image analysis of TIRF live - cell microscopy data . Fluorescent dif - fraction - limited spots of DNM2 - eGFPEN and CLTA - TagRFP - TEN in SK - MEL - 2 cells were tracked in each channel over the course of a 14 | C . Kaplan et al . Molecular Biology of the Cell movie using the detection and tracking features of the cmeAnalysis software package in Matlab ( Aguet et al . , 2013 ) . Subsequently , we cleaned up the background signal and , on the basis of the x and y positions of associated CLTA - TagRFP - TEN and DNM2 - eGFPEN flu - orescent tracks , separated them from CLTA - TagRFP - TEN and DNM2 - eGFPEN that were not associated with each other , using a custom - written Matlab script ( 10 , 36 , 42 ) . We measured fluores - cence lifetimes for DNM2 - eGFPEN and CLTA - TagRFP - TEN tracks that were associated and not associated with each other and that appeared and disappeared within the duration of a movie . We clas - sified the tracks as \u201cpersistent\u201d when they were already present in the first movie frame and lasted longer than the movie . CME initiation rate and completion rate measurement . We de - fined a track as initiated if it appeared within the course of the acqui - sition , excluding the first frame . We classified a track as complete when it appeared and disappeared within the course of the acquisi - tion . To calculate rates of initiation and completion , we used these values along with measurement of the area of the tracked region , which was obtained from the binary cell mask image generated dur - ing the detection step of the cmeAnalysis software that highlights the cell area in which particle tracking was performed . ImageJ was then used to calculate the area of the cell from the binary mask image . The final metric is defined as the number of initiations ( or completions ) per area per time . . Actin cortex height quantification . To measure the height of the actin cortex in our 2c - 3D STORM images , we cropped an approxi - mately 6 \u00b5m \u00d7 6 \u00b5m ROI of the superresolved clathrin image in a cell in which both actin and clathrin were imaged . Using this mask we removed all coordinates from the actin X , Y , and Z localization fluo - rophore list that overlapped with clathrin in that ROI . Then , we re - ported the SD of those Z coordinates not associated with clathrin and their median as a metric for actin cortex height . The clathrin mask was generated as follows : An approximately 4 \u00b5m \u2013 6 \u00b5m \u00d7 4 \u00b5m \u2013 6 \u00b5m ROI of the superresolved clathrin image of a cell was convolved with Gaussian distribution ( sigma = 5 ) ( Supple - mental Figure S2K ) . We ran a local maxima finder on the Gaussian - rendered image ( Scikit Image , Python ) . We generated a 2D histo - gram array of the local maxima and saved these as an image file . To this image we applied an outline finding algorithm ( FIJI ) , followed by a maximum filter ( px = 3 , FIJI ) , and finally with a hole - filling algo - rithm ( FIJI ) . The hole - filled image was inverted such that all clathrin areas had pixel value zero and all nonclathrin areas had pixel value 1 ( Supplemental Figure S2L ) . This mask was subsequently applied to a 2D histogram \u2013 generated array of actin X , Y , and Z superre - solved coordinates from the same ROI ( Supplemental Figure S2 , M and O ) . Both the clathrin mask and the 2D histogram actin array were adjusted to the same pixel size and rescaled by a factor of four . We then calculated the SDs of binned Z coordinates . Bins , respec - tively pixels , with fewer than four localizations were rejected for the analysis . The SDs of binned Z coordinates were converted from pix - els into nanometers for data visualization and plotted in histograms in Supplemental Figure S2 , P and Q . In Supplemental Figure S2R we report the median SD of the actin Z coordinates as a metric for actin cortex height . ACKNOWLEDGMENTS C . K . was funded by the German Research Foundation ( DFG KA4305 / 1 - 1 ) . D . G . D . was funded by National Institutes of Health ( NIH ) grant R35GM118149 . K . X . was funded by the National Sci - ence Foundation under CHE - 1554717 and the Pew Biomedical Scholars Award . A . D . M . was funded by the European Molecular Bi - ology Laboratory ( EMBL ) , Human Frontiers Science Program ( HFSP ) grant number RGY0073 / 2018 , and Deutsche Forschungsgemein - schaft ( DFG ) grant numbers DI 2205 / 2 - 1 and DI 2205 / 3 - 1 . M . A . was funded by NIH grant 1 K99 GM132551 - 01 . E . S . was funded by the EMBL and the Joachim Herz Stiftung Add - on Fellowship for Inter - disciplinary Science . We thank Justin Taraska and Kem Sochacki ( National Heart , Lung , and Blood Institute , NIH ) for use of the plati - num replica EM images for comparison in Supplemental Figure S3 . We thank Yidi Sun , Padmini Rangamani , and Ross T . A . Pedersen for critical reading and discussions on the manuscript . We thank Sung - min Son and Daniel A . Fletcher for valuable input and discussion on the manuscript . REFERENCES Aghamohammadzadeh S , Ayscough KR ( 2009 ) . Differential requirements for actin during yeast and mammalian endocytosis . Nat Cell Biol 11 , 1039 \u2013 1042 . Akamatsu M , Vasan R , Serwas D , Ferrin MA , Rangamani P , Drubin DG ( 2020 ) . Principles of self - organization and load adaptation by the actin cytoskel - eton during clathrin - mediated endocytosis . eLife 9 , e49840 . Almeida - Souza L , Frank RAW , Grac\u00eda - Nafr\u00eda J , Colussi A , Gunawardana N , Johnson CM , Yu M , Howard G , Andrews B , Vallis Y , et al . ( 2018 ) . A flat BAR protein promotes actin polymerization at the base of clathrin - coated pits . Cell 174 , 325 \u2013 337 . e14 . Avinoam O , Schorb M , Beese CJ , Briggs JAG , Kaksonen M ( 2015 ) . ENDOCYTOSIS . Endocytic sites mature by continuous bending and remodeling of the clathrin coat . Science 348 , 1369 \u2013 1372 . Batchelder EM , Yarar D ( 2010 ) . Differential requirements for clathrin - depen - dent endocytosis at sites of cell - substrate adhesion . Mol Biol Cell 21 , 3070 \u2013 3079 . Bieling P , Li T - D , Weichsel J , McGorty R , Jreij P , Huang B , Fletcher DA , Mullins RD ( 2016 ) . Force feedback controls motor activity and mechani - cal properties of self - assembling branched actin networks . Cell 164 , 115 \u2013 127 . Bossi M , F\u00f6lling J , Belov VN , Boyarskiy VP , Medda R , Egner A , Eggeling C , Sch\u00f6nle A , Hell SW ( 2008 ) . Multicolor far - field fluorescence nanoscopy through isolated detection of distinct molecular species . Nano Lett 8 , 2463 \u2013 2468 . Boulant S , Kural C , Zeeh JC , Ubelmann F , Kirchhausen T ( 2011 ) . Actin dynamics counteract membrane tension during clathrin - mediated endo - cytosis . Nat Cell Biol 13 , 1124 \u2013 1131 . Bucher D , Frey F , Sochaki KA , Kummer S , Bergeest JP , Godinez WJ , Kr\u00e4usslich HG , Rohr K , Taraska JW , Schwarz US , et al . ( 2018 ) . Clathrin - adaptor ratio and membrane tension regulate the flat - to - curved transi - tion of the clathrin coat during endocytosis . Nat Commun 9 , 1 \u2013 13 . Buser C , Drubin DG ( 2013 ) . Ultrastructural imaging of endocytic sites in Saccharomyces cerevisiae by transmission electron microscopy and im - munolabeling . Microsc Microanal 19 , 381 \u2013 392 . Carroll SY , Stimpson HEM , Weinberg J , Toret CP , Sun Y , Drubin DG ( 2012 ) . Analysis of yeast endocytic site formation and maturation through a regulatory transition point . Mol Biol Cell 23 , 657 \u2013 668 . Chugh P , Clark AG , Smith MB , Cassani DAD , Dierkes K , Ragab A , Roux PP , Charras G , Salbreux G , Paluch EK ( 2017 ) . Actin cortex architecture regulates cell surface tension . Nat Cell Biol 19 , 689 \u2013 697 . Clarke NI , Royle SJ ( 2018 ) . FerriTag is a new genetically - encoded inducible tag for correlative light - electron microscopy . Nat Commun 9 , 1 \u2013 10 . Collins A , Warrington A , Taylor KA , Svitkina T ( 2011 ) . Structural organization of the actin cytoskeleton at sites of clathrin - mediated endocytosis . Curr Biol 21 , 1167 \u2013 1175 . Dambournet D , Sochacki KA , Cheng AT , Akamatsu M , Taraska JW , Hockemeyer D , Drubin DG ( 2018 ) . Genome - edited human stem cells expressing fluorescently labeled endocytic markers allow quantitative analysis of clathrin - mediated endocytosis during differentiation . J Cell Biol 2 , jcb . 201710084 - 11 . Diz - Mu\u00f1oz A , Fletcher DA , Weiner OD ( 2013 ) . Use the force : membrane tension as an organizer of cell shape and motility . Trends Cell Biol 23 , 47 \u2013 53 . Diz - Mu\u00f1oz A , Thurley K , Chintamen S , Altschuler SJ , Wu LF , Fletcher DA , Weiner OD ( 2016 ) . Membrane tension acts through PLD2 and mTORC2 to limit actin network assembly during neutrophil migration . PLoS Biol 14 , e1002474 - 30 . Volume 33 May 15 , 2022 Actin load adaptation in endocytosis | 15 Djakbarova U , Madraki Y , Chan E , Kural C ( 2021 ) . Dynamic interplay be - tween cell membrane tension and clathrin - mediated endocytosis . Biol Cell 113 , 344 \u2013 373 . Doyon JB , Zeitler B , Cheng J , Cheng AT , Cherone JM , Santiago Y , Lee AH , Vo TD , Doyon Y , Miller JC , et al . ( 2011 ) . Rapid and efficient clathrin - mediated endocytosis revealed in genome - edited mammalian cells . Nat Cell Biol 13 , 331 \u2013 337 . Engqvist - Goldstein AE , Warren RA , Kessels MM , Keen JH , Heuser J , Drubin DG ( 2001 ) . The actin - binding protein Hip1R associates with clathrin dur - ing early stages of endocytosis and promotes clathrin assembly in vitro . J Cell Biol 154 , 1209 \u2013 1223 . Ferguson JP , Huber SD , Willy NM , Ayg\u00fcn E , Goker S , Atabey T , Kural C ( 2017 ) . Mechanoregulation of clathrin - mediated endocytosis . J Cell Sci 130 , 3631 \u2013 3636 . Ferguson JP , Willy NM , Heidotting SP , Huber SD , Webber MJ , Kural C ( 2016 ) . Deciphering dynamics of clathrin - mediated endocytosis in a liv - ing organism . J Cell Biol 282 , jcb . 201604128 - 12 . Ferguson SM , Raimondi A , Paradise S , Shen H , Mesaki K , Ferguson A , Destaing O , Ko G , Takasaki J , Cremona O , et al . ( 2009 ) . Coordinated actions of actin and BAR proteins upstream of dynamin at endocytic clathrin - coated pits . Dev Cell 17 , 811 \u2013 822 . Fujimoto LM , Roth R , Heuser JE , Schmid SL ( 2000 ) . Actin assembly plays a variable , but not obligatory role in receptor - mediated endocytosis in mammalian cells . Traffic 1 , 161 \u2013 171 . Funk J , Merino F , Schaks M , Rottner K , Raunser S , Bieling P ( 2021 ) . A barbed end interference mechanism reveals how capping protein promotes nucleation in branched actin networks . Nat Commun 12 , 1 \u2013 17 . Gorur A , Yuan L , Kenny SJ , Baba S , Xu K , Schekman R ( 2017 ) . COPII - coated membranes function as transport carriers of intracellular procollagen I . J Cell Biol 216 , 1745 \u2013 1759 . Grassart A , Cheng AT , Hong SH , Zhang F , Zenzer N , Feng Y , Briner DM , Davis GD , Malkov D , Drubin DG ( 2014 ) . Actin and dynamin2 dynam - ics and interplay during clathrin - mediated endocytosis . J Cell Biol 205 , 721 \u2013 735 . Guo S , Sokolova OS , Chung J , Padrick S , Gelles J , Goode BL ( 2018 ) . Abp1 promotes Arp2 / 3 complex - dependent actin nucleation and stabilizes branch junctions by antagonizing GMF . Nat Commun 9 , 1 \u2013 14 . Hassinger JE , Oster G , Drubin DG , Rangamani P ( 2017 ) . Design principles for robust vesiculation in clathrin - mediated endocytosis . Proc Natl Acad Sci USA 114 , E1118 \u2013 E1127 . Hauser M , Yan R , Li W , Repina NA , Schaffer DV , Xu K ( 2018 ) . The spectrin - actin - based periodic cytoskeleton as a conserved nanoscale scaffold and ruler of the neural stem cell lineage . Cell Rep 24 , 1512 \u2013 1522 . Helgeson LA , Nolen BJ ( 2013 ) . Mechanism of synergistic activation of Arp2 / 3 complex by cortactin and N - WASP . eLife 2 , e00884 . Hetrick B , Han MS , Helgeson LA , Nolen BJ ( 2013 ) . Small molecules CK - 666 and CK - 869 inhibit actin - related protein 2 / 3 complex by blocking an activating conformational change . Chem Biol 20 , 701 \u2013 712 . Hong SH , Cortesio CL , Drubin DG ( 2015 ) . Machine - learning - based analysis in genome - edited cells reveals the efficiency of clathrin - mediated endo - cytosis . Cell Rep 12 , 2121 \u2013 2130 . Houk AR , Jilkine A , Mejean CO , Boltyanskiy R , Dufresne ER , Angenent SB , Altschuler SJ , Wu LF , Weiner OD ( 2012 ) . Membrane tension maintains cell polarity by confining signals to the leading edge during neutrophil migration . Cell 148 , 175 \u2013 188 . Huang B , Wang W , Bates M , Zhuang X ( 2008 ) . Three - dimensional super - resolution imaging by stochastic optical reconstruction microscopy . Science 319 , 810 \u2013 813 . Idrissi FZ , Blasco A , Espinal A , Geli MI ( 2012 ) . Ultrastructural dynamics of proteins involved in endocytic budding . Proc Natl Acad Sci USA 109 , E2587 \u2013 2594 . Idrissi F - Z , Gr\u00f6tsch H , Fern\u00e1ndez - Golbano IM , Presciatto - Baschong C , Riezman H , Geli MI ( 2008 ) . Distinct acto / myosin - I structures associ - ate with endocytic profiles at the plasma membrane . J Cell Biol 180 , 1219 \u2013 1232 . Jin M , Shirazinejad C , Wang B , Yan A , Sch\u00f6neberg J , Upadhyayula S , Xu K , Drubin DG ( 2021 ) . Rescue of stalled clathrin - mediated endocytosis by asymmetric Arp2 / 3 - mediated actin assembly . bioRxiv , https : / / doi . org / 10 . 1101 / 2021 . 07 . 16 . 452693 . Joseph JG , Osorio C , Yee V , Agrawal A , Liu AP ( 2020 ) . Complimentary ac - tion of structured and unstructured domains of epsin supports clathrin - mediated endocytosis at high tension . Commun Biol 3 , 743 . Kaksonen M , Sun Y , Drubin DG ( 2003 ) . A pathway for association of recep - tors , adaptors , and actin during endocytic internalization . Cell 115 , 475 \u2013 487 . Kaksonen M , Toret CP , Drubin DG ( 2005 ) . A modular design for the clathrin - and actin - mediated endocytosis machinery . Cell 123 , 305 \u2013 320 . Kukulski W , Schorb M , Kaksonen M , Briggs JAG ( 2012 ) . Plasma membrane reshaping during endocytosis is revealed by time - resolved electron tomography . Cell 150 , 508 \u2013 520 . Le Clainche C , Pauly BS , Zhang CX , Engqvist - Goldstein \u00c5EY , Cunningham K , Drubin DG ( 2007 ) . A Hip1R - cortactin complex negatively regulates actin assembly associated with endocytosis . EMBO J 26 , 1199 \u2013 1210 . Li D , Shao L , Chen BC , Zhang X , Zhang M , Moses B , Milkie DE , Beach JR , Hammer JA 3rd , Pasham M , et al . ( 2015 ) . ADVANCED IMAGING . Extended - resolution structured illumination imaging of endocytic and cytoskeletal dynamics . Science 349 , aab3500 . Li T - D , Bieling P , Weichsel J , Dyche Mullins R , Fletcher DA ( 2021 ) . The molecular mechanism of load adaptation by branched actin networks . bioRxiv , https : / / doi . org / 10 . 1101 / 2021 . 05 . 24 . 445507 . Li Y , Mund M , Hoess P , Deschamps J , Matti U , Nijmeijer B , Sabinina VJ , Ellenberg J , Schoen I , Ries J ( 2018 ) . Real - time 3D single - molecule localization using experimental point spread functions . Nat Methods 15 , 367 \u2013 369 . Liu AP , Loerke D , Schmid SL , Danuser G ( 2009 ) . Global and local regula - tion of clathrin - coated pit dynamics detected on patterned substrates . Biophys J 97 , 1038 \u2013 1047 . Merrifield CJ , Feldman ME , Wan L , Almers W ( 2002 ) . Imaging actin and dynamin recruitment during invagination of single clathrin - coated pits . Nat Cell Biol 4 , 691 \u2013 698 . Merrifield CJ , Perrais D , Zenisek D ( 2005 ) . Coupling between clathrin - coated - pit invagination , cortactin recruitment , and membrane scission observed in live cells . Cell 121 , 593 \u2013 606 . Messa M , Fern\u00e1ndez - Busandiego R , Sun EW , Chen H , Czapla H , Wrasman K , Wu Y , Ko G , Ross T , Wendland B , et al . ( 2014 ) . Epsin deficiency impairs endocytosis by stalling the actin - dependent invagination of endocytic clathrin - coated pits . eLife 3 , e03311 . Mogilner A , Oster G ( 1996 ) . Cell motility driven by actin polymerization . Biophys J 71 , 3030 \u2013 3045 . Mogilner A , Oster G ( 2003 ) . Force generation by actin polymerization II : the elastic ratchet and tethered filaments . Biophys J 84 , 1591 \u2013 1605 . Mueller J , Szep G , Nemethova M , Vries ID , Lieber AD , Winkler C , Kruse K , Small JV , Schmeiser C , Keren K , et al . ( 2017 ) . Load adaptation of lamel - lipodial actin networks . Cell 171 , 188 \u2013 200 . e16 . Mulholland J ( 1994 ) . Ultrastructure of the yeast actin cytoskeleton and its association with the plasma membrane . J Cell Biol 125 , 381 \u2013 391 . Mullins RD , Heuser JA , Pollard TD ( 1998 ) . The interaction of Arp2 / 3 complex with actin : Nucleation , high affinity pointed end capping , and formation of branching networks of filaments . PNAS 95 , 6181 \u2013 6186 . Mund M , Tschanz A , Wu Y - L , Frey F , Mehl JL , Kaksonen M , Avinoam O , Schwarz US , Ries J ( 2021 ) . Superresolution microscopy reveals partial preassembly and subsequent bending of the clathrin coat during endo - cytosis . bioRxiv , https : / / doi . org / 10 . 1101 / 2021 . 10 . 12 . 463947 . Mund M , van der Beek JA , Deschamps J , Dmitrieff S , Hoess P , Monster JL , Picco A , N\u00e9d\u00e9lec F , Kaksonen M , Ries J ( 2018 ) . Systematic nanoscale analysis of endocytosis links efficient vesicle formation to patterned actin nucleation . Cell 174 , 884 \u2013 896 . e17 . Nolen BJ , Tomasevic N , Russell A , Pierce DW , Jia Z , McCormick CD , Hartman J , Sakowicz R , Pollard TD ( 2009 ) . Characterization of two classes of small molecule inhibitors of Arp2 / 3 complex . Nature 460 , 1031 \u2013 1034 . Pan L , Zhang P , Hu F , Yan R , He M , Li W , Xu J , Xu K ( 2019 ) . Hypotonic stress induces fast , reversible degradation of the vimentin cytoskeleton via intracellular calcium release . Adv Sci ( Weinh ) 6 , 1900865 . Picco A , Mund M , Ries J , N\u00e9d\u00e9lec F , Kaksonen M ( 2015 ) . Visualizing the functional architecture of the endocytic machinery . eLife 4 , e04535 . Pinyol R , Haeckel A , Ritter A , Qualmann B , Kessels MM ( 2007 ) . Regulation of N - WASP and the Arp2 / 3 complex by Abp1 controls neuronal mor - phology . PLoS One 2 , e400 . Pollard TD ( 2016 ) . Actin and actin - binding proteins . Cold Spring Harb Perspect Biol 8 , a018226 . Pontes B , Monzo P , Gauthier NC ( 2017 ) . Membrane tension : a challenging but universal physical parameter in cell biology . Semin Cell Dev Biol 71 , 30 \u2013 41 . Rangamani P , Mandadap KK , Oster G ( 2014 ) . Protein - induced membrane curvature alters local membrane tension . Biophys J 107 , 751 \u2013 762 . Raucher D , Sheetz MP ( 1999 ) . Membrane expansion increases endocytosis rate during mitosis . J Cell Biol 144 , 497 \u2013 506 . Roffay C , Molinard G , Kim K , Roux A ( 2021 ) . Passive coupling of membrane tension and cell volume during active response of cells to osmosis . Proc Natl Acad Sci USA 118 , e2103228118 . 16 | C . Kaplan et al . Molecular Biology of the Cell Rust MJ , Bates M , Zhuang X ( 2006 ) . Sub - diffraction - limit imaging by stochastic optical reconstruction microscopy ( STORM ) . Nat Methods 3 , 793 \u2013 796 . Saleem M , Morlot S , Hohendahl A , Manzi J , Lenz M , Roux A ( 2015 ) . A bal - ance between membrane elasticity and polymerization energy sets the shape of spherical clathrin coats . Nat Commun 6 , 6249 . Sch\u00f6neberg J , Dambournet D , Liu T - L , Forster R , Hockemeyer D , Betzig E , Drubin DG ( 2018 ) . 4D cell biology : big data image analytics and lattice light - sheet imaging reveal dynamics of clathrin - mediated endocytosis in stem cell - derived intestinal organoids . Mol Biol Cell 29 , 2959 \u2013 2968 . Sch\u00f6neberg J , Lehmann M , Ullrich A , Posor Y , Lo W - T , Lichtner G , Schmo - ranzer J , Haucke V , No\u00e9 F ( 2017 ) . Lipid - mediated PX - BAR domain recruitment couples local membrane constriction to endocytic vesicle fission . Nat Commun 8 , 15873 . Scott BL , Sochacki KA , Low - Nam ST , Bailey EM , Luu QA , Hor A , Dickey AM , Smith S , Krekvliet JG , Taraska JW , et al . ( 2018 ) . Membrane bending occurs at all stages of clathrin - coat assembly and defines endocytic dynamics . Nat Commun 9 , 419 . Serwas D , Akamatsu M , Moayed A , Vegesna K , Vasan R , Hill JM , Sch\u00f6ne - berg J , Davies KM , Rangamani P , Drubin DG ( 2021 ) . Actin force generation in vesicle formation : mechanistic insights from cryo - electron tomography , bioRxiv , https : / / doi . org / 10 . 1101 / 2021 . 06 . 28 . 450262 . Shi Z , Graber ZT , Baumgart T , Stone HA , Cohen AE ( 2018 ) . Cell membranes resist flow . Cell 175 , 1769 \u2013 1779 . e13 . Sitarska E , Diz - Mu\u00f1oz A ( 2020 ) . Pay attention to membrane tension : mecha - nobiology of the cell surface . Curr Opin Cell Biol 66 , 11 \u2013 18 . Skruzny M , Desfosses A , Prinz S , Dodonova SO , Gieras A , Uetrecht C , Jakobi AJ , Abella M , Hagen WJH , Schulz J , et al . ( 2015 ) . An organized co - assembly of clathrin adaptors is essential for endocytosis . Dev Cell 33 , 150 \u2013 162 . Skruzny M , Brach T , Ciuffa R , Rybina S , Wachsmuth M , Kaksonen M ( 2012 ) . Molecular basis for coupling the plasma membrane to the actin cyto - skeleton during clathrin - mediated endocytosis . Proc Natl Acad Sci USA 109 , E2533 \u2013 E2542 . Sochacki KA , Dickey AM , Strub M - P , Taraska JW ( 2017 ) . Endocytic proteins are partitioned at the edge of the clathrin lattice in mammalian cells . Nat Cell Biol 12 , 517 \u2013 541 . Sochacki KA , Heine BL , Haber GJ , Jimah JR , Prasai B , Alfonzo - M\u00e9ndez MA , Roberts AD , Somasundaram A , Hinshaw JE , Taraska JW ( 2021 ) . The structure and spontaneous curvature of clathrin lattices at the plasma membrane . Dev Cell 56 , 1131 \u2013 1146 . e3 . Testa I , Wurm CA , Medda R , Rothermel E , von Middendorf C , F\u00f6lling J , Ja - kobs S , Sch\u00f6nle A , Hell SW , Eggeling C ( 2010 ) . Multicolor fluorescence nanoscopy in fixed and living cells by exciting conventional fluorophores with a single wavelength . Biophys J 99 , 2686 \u2013 2694 . Willy NM , Ferguson JP , Akatay A , Huber S , Djakbarova U , Silahli S , Cakez C , Hasan F , Chang HC , Travesset A , et al . ( 2021 ) . De novo endocytic clathrin coats develop curvature at early stages of their formation . Dev Cell 56 , 3146 \u2013 3159 . Willy NM , Ferguson JP , Huber SD , Heidotting SP , Ayg\u00fcn E , Wurm SA , Johnston - Halperin E , Poirier MG , Kural C ( 2017 ) . Membrane mechanics govern spatiotemporal heterogeneity of endocytic clathrin coat dynam - ics . Mol Biol Cell 28 , 3480 \u2013 3488 . Wojcik M , Hauser M , Li W , Moon S , Xu K ( 2015 ) . Graphene - enabled elec - tron microscopy and correlated super - resolution microscopy of wet cells . Nat Commun 6 , 1 \u2013 6 . Xu K , Babcock HP , Zhuang X ( 2012 ) . Dual - objective STORM reveals three - dimensional filament organization in the actin cytoskeleton . Nat Methods 9 , 185 \u2013 188 . Xu K , Shim S - H , Zhuang X ( 2015 ) . Super - resolution imaging through stochastic switching and localization of single molecules : an overview . In : Far - Field Optical Nanoscopy , ed . P Tinnefeld , C Eggeling , SW Hell , Berlin : Springer , 27 \u2013 64 . Xu K , Zhong G , Zhuang X ( 2013 ) . Actin , spectrin , and associated proteins form a periodic cytoskeletal structure in axons . Science 339 , 452 \u2013 456 . Yarar D , Waterman - Storer CM , Schmid SL ( 2005 ) . A dynamic actin cytoskel - eton functions at multiple stages of clathrin - mediated endocytosis . Mol Biol Cell 16 , 964 \u2013 975 . Yarar D , Waterman - Storer CM , Schmid SL ( 2007 ) . SNX9 couples actin assembly to phosphoinositide signals and is required for membrane remodeling during endocytosis . Dev Cell 13 , 43 \u2013 56 . Yoshida A , Sakai N , Uekusa Y , Imaoka Y , Itagaki Y , Suzuki Y , Yoshimura SH ( 2018 ) . Morphological changes of plasma membrane and protein assem - bly during clathrin - mediated endocytosis . PLoS Biol 16 , e2004786 - 28 .",
    "doherty2011endocytic": "4380 | G . J . Doherty et al . Molecular Biology of the Cell MBoC | ARTICLE The endocytic protein GRAF1 is directed to cell - matrix adhesion sites and regulates cell spreading ABSTRACT The rho GTPase - activating protein GTPase regulator associated with focal adhe - sion kinase - 1 ( GRAF1 ) remodels membranes into tubulovesicular clathrin - independent carri - ers ( CLICs ) mediating lipid - anchored receptor endocytosis . However , the cell biological func - tions of this highly prevalent endocytic pathway are unclear . In this article , we present biochemical and cell biological evidence that GRAF1 interacted with a network of endocytic and adhesion proteins and was found enriched at podosome - like adhesions and src - induced podosomes . We further demonstrate that these sites comprise microdomains of highly or - dered lipid enriched in GRAF1 endocytic cargo . GRAF1 activity was upregulated in spreading cells and uptake via CLICs was concentrated at the leading edge of migrating cells . Depletion of GRAF1 , which inhibits CLIC generation , resulted in profound defects in cell spreading and migration . We propose that GRAF1 remodels membrane microdomains at adhesion sites into endocytic carriers , facilitating membrane turnover during cell morphological changes . INTRODUCTION Cells interact with their immediate environments through the liga - tion of plasma membrane - anchored or transmembrane receptors for soluble molecules , such as growth factors , extracellular matrix components , and proteins presented on the surface of neighboring cells . Cell - matrix adhesions are local , dynamic attachments of cell - surface proteins , including integrins and glycophosphatidylinositol - anchored proteins ( GPI - APs ) , to extracellular matrix components that allow indirect bridging of this matrix to the internal cytoskele - ton . These dynamic anchor points determine the position of the cell in space and allow cells to undergo shape changes , including those required during cell division , spreading , and migration ( Doherty and McMahon , 2008 ) . Initial sites of such adhesion can produce focal complexes , which can 1 ) be disassembled ( if conditions so dictate ) ; 2 ) become stabilized ; or 3 ) grow into larger , more mature ( and less dynamic ) adhesion sites , known as focal adhesions , that allow strong connection of the matrix to actin stress fibers . The diversity and dy - namics of these integrin - based adhesions , including podosomes and invadopodia , are coordinated by the activity of rho - family small G proteins that transduce internal and external cues into signals for the development , maintenance , growth , and disassembly of these anchor points . The turnover of adhesion - associated lipid domains and proteins has the potential to regulate adhesion sites , and several studies have suggested an important role for endocytosis in their dynamics ( reviewed in Caswell et al . [ 2009 ] ) . Adhesive sites strongly affect lipid order and promote the formation of microdomains necessary for certain endocytic events . Interestingly , loss of adhesion correlates with rapid endocytosis of molecules enriched in microdomains , such as cholera toxin B subunit ( CTxB ; del Pozo et al . , 2004 ; Schlunck et al . , 2004 ; Gaus et al . , 2006 ) . It has been shown that focal adhesion Monitoring Editor Yu - Li Wang Carnegie Mellon University Received : Dec 2 , 2010 Revised : Aug 16 , 2011 Accepted : Sep 20 , 2011 * These authors contributed equally to this work . The authors declare no competing interests . Address correspondence to : Richard Lundmark ( Richard . lundmark @ medchem . umu . se ) . \u201cASCB \u00ae , \u201c \u201cThe American Society for Cell Biology \u00ae , \u201d and \u201cMolecular Biology of the Cell \u00ae \u201d are registered trademarks of The American Society of Cell Biology . \u00a9 2011 Doherty et al . This article is distributed by The American Society for Cell Biology under license from the author ( s ) . Two months after publication it is avail - able to the public under an Attribution \u2013 Noncommercial \u2013 Share Alike 3 . 0 Unport - ed Creative Commons License ( http : / / creativecommons . org / licenses / by - nc - sa / 3 . 0 ) . This article was published online ahead of print in MBoC in Press ( http : / / www . molbiolcell . org / cgi / doi / 10 . 1091 / mbc . E10 - 12 - 0936 ) on September 30 , 2011 . Abbreviations used : BAR , Bin / amphiphysin / Rvs ; CLIC , clathrin - independent car - rier ; CTxB , cholera toxin B subunit ; DA , dominant - active ; DTT , dithiothreitol ; FAK , focal adhesion kinase ; GFP , green fluorescent protein ; GIT1 , G protein \u2013 coupled receptor kinase \u2013 interacting ArfGAP 1 ; GPI - AP , glycophosphatidylinositol - an - chored protein ; GRAF1 , GTPase regulator associated with focal adhesion ki - nase - 1 ; GST , glutathione S - transferase ; MEF , mouse embryonic fibroblast ; PBS , phosphate - buffered saline ; pFAK , phosphorylated FAK ; PH , pleckstrin homology ; PIP2 , phosphatidyl inositol 4 , 5 bisphosphate ; PLA , podosome - like adhesion ; RFP , red fluorescent protein ; rhoGAP , rho GTPase - activating protein ; ROCK , rho ki - nase ; siRNA , small interfering RNA ; Tfn , transferrin . Gary J . Doherty a , * , Monika K . \u00c5hlund b , * , Mark T . Howes c , Bj\u00f6rn Mor\u00e9n b , Robert G . Parton c , Harvey T . McMahon a , and Richard Lundmark b a MRC Laboratory of Molecular Biology , Cambridge CB2 0QH , United Kingdom ; b Medical Biochemistry and Biophys - ics , Ume\u00e5 University , 901 87 Ume\u00e5 , Sweden ; c Institute for Molecular Bioscience and Centre for Microscopy and Microanalysis , University of Queensland , Brisbane QLD 4072 , Australia Volume 22 November 15 , 2011 GRAF1 trafficking directs cell spreading | 4381 turnover can be induced by regrowth of microtubules after their de - polymerization in a process that requires the endocytic protein dy - namin and focal adhesion kinase ( FAK ; Ezratty et al . , 2005 , 2009 ) . Dynamin is also found at podosomes . These are rather small and highly active adhesion sites comprising adhesion proteins with an actin core . Here , dynamin is thought to function in the interplay be - tween membrane and actin dynamics ( Ochoa et al . , 2000 ) . The best understood endocytic process is mediated by the coat protein clathrin , but cells use a variety of alternative , clathrin - inde - pendent endocytic routes ( Doherty and McMahon , 2009 ) . The most prevalent of these in fibroblasts is mediated by clathrin - inde - pendent carriers ( CLICs ; reviewed in Mayor and Pagano [ 2007 ] ; Hansen and Nichols [ 2009 ] ; Howes et al . [ 2010b ] ) . We have recently shown that the membrane remodeling protein GTPase regulator associated with focal adhesion kinase - 1 ( GRAF1 ) regulates GPI - APs and CTxB uptake into CLICs , as well as a large proportion of fluid - phase uptake ( Lundmark et al . , 2008 ) . These pleiomorphic , tubulo - vesicular carriers lack an obvious protein coat and form in the ab - sence of clathrin and caveolin ( Kirkham et al . , 2005 ) . Internalization of GPI - APs and other microdomain - enriched molecules via these carriers depends upon the activity of cdc42 , GRAF1 , Arf1 , ordered lipid microdomains , and actin polymerization ( Sabharanjak et al . , 2002 ; Chadda et al . , 2007 ; Kumari and Mayor , 2008 ; Lundmark et al . , 2008 ) . GRAF1 has a rho GTPase - activating protein ( rhoGAP ) domain that stimulates the GTPase activity of cdc42 in vitro ( Hildebrand et al . , 1996 ) , as well as Bin / amphiphysin / Rvs ( BAR ) and pleckstrin homology ( PH ) domains required for membrane binding and CLIC formation . We describe a link between the en - docytic activity of GRAF1 and adhesion modulation . GRAF1 was found enriched at podosome - like adhesions induced in HeLa cells . Our investigations of the activity of these structures support a central role for GRAF1 and CLIC production in cell spreading and migration through modulation of the dynamics of these adhe - sion sites . RESULTS The GRAF1 interactome links endocytosis and cell adhesion Monomers of the dimeric protein GRAF1 are composed of BAR , PH , rhoGAP , and SH3 domains ( Figure 1A ) . Our previous studies have shown that GRAF1 is necessary for the generation of clathrin - inde - pendent endocytic carriers that endocytose a large amount of extra - cellular fluid and cargoes that include CTxB and GPI - APs ( Lundmark et al . , 2008 ) . GRAF1 has also been implicated in regulation of focal adhesions and cytoskeletal rearrangements ( Hildebrand et al . , 1996 ) . To address how these seemingly distinct roles might be linked , we first sought to identify additional proteins that function together with GRAF1 . We immunoprecipitated GRAF1 from rat brain cytosol and identified dynamin , FAK , and G protein \u2013 coupled receptor kinase \u2013 interacting ArfGAP 1 ( GIT1 ) as interacting partners using mass spectrometry and immunoblotting ( Figure 1B ) . We also tested whether GRAF1 could be isolated by immunoprecipitating dynamin , FAK , or GIT1 . Under these conditions , GRAF1 coimmuno - precipitated with dynamin , but not with either FAK or GIT1 , sug - gesting that these interactions were weaker or that the antibodies perturbed the binding ( Figure 1C , top panel ) . GRAF1 could , how - ever , be coimmunoprecipitated with the active , phosphorylated form of FAK , pFAK ( Figure 1C , bottom panel ) . Pulldown studies us - ing the GRAF1 SH3 domain as bait in rat brain cytosol verified that the interaction to dynamin and FAK was dependent on this domain ( Figure 1D ) . We then showed that GRAF1 , GIT1 , and dynamin are found together at a subset of basal plasma membrane structures in HeLa cells , suggesting that GRAF1 also interacts with these proteins in vivo ( Figure 1E ) . GIT1 is an Arf GTPase - activating protein that lo - calizes to cell - matrix adhesions ; it has been proposed that GIT1 pro - motes adhesion down - regulation and cell spreading ( Zhao et al . , 2000 ; Manabe et al . , 2002 ) and influences endocytosis and recycling of G protein \u2013 coupled receptors ( Premont et al . , 1998 ; Claing et al . , 2000 ; Lahuna et al . , 2005 ) . Interestingly , the endocytic membrane fission protein dynamin is also implicated in cell - matrix adhesion site turnover and is present at podosomes and focal complexes ( Ochoa et al . , 2000 ; Ezratty et al . , 2005 ) . The known interactions of GRAF1 are illustrated in an interactome in Figure 1F . Taken together with previous results , our observations suggest that the N - terminal BAR and PH domains endow GRAF1 with the ability to generate / stabilize highly curved endocytic membranes from ordered , phosphatidyl inositol 4 , 5 - bisphosphate ( PIP2 ) - enriched plasma membrane re - gions , while the C - terminal rhoGAP and SH3 domains modulate rho - family small G - protein activity and interact with dynamin and FAK . GRAF1 is not a general component of focal adhesions but localizes to podosome - like adhesions Given the nature of the GRAF1 interactome , we examined whether GRAF1 localization / function was linked to adhesion sites . When GRAF1 was expressed at very high levels , cells underwent a drastic morphological rearrangement , developing long protrusions and collapsing around the nucleus ( Taylor et al . , 1999 ; Supplemental Figure S1 , A and B ) . In these cells , the typical focal adhesions and actin stress fibers seen in low - level , overexpressing cells or control cells were absent , and only a few , small punctate adhesions contain - ing vinculin and actin could be detected . When cells with endoge - nous or low levels of overexpressed GRAF1 were costained with markers for adhesions such as vinculin , we found that GRAF1 was not detected at large , elongated adhesions known as mature focal adhesions ( see Figure S1 , C and D ) . Instead , both endogenous and overexpressed GRAF1 were found enriched at small , round , vincu - lin - positive structures in a subset of cells ( see Figure S1 , C and D ) . Our data suggested that GRAF1 might be recruited to a specific intermediate state of adhesive structures . To determine whether its interacting partner GIT1 , which is normally found at cell - matrix ad - hesion sites , could promote the localization of GRAF1 to adhesion sites , we co - overexpressed these proteins . This resulted in fewer mature focal adhesions and a large increase in the number of vincu - lin - and GRAF1 - positive structures ( Figure 2 , A and B ) . This is consis - tent with previous observations that GIT1 promotes turnover of focal adhesions ( Zhao et al . , 2000 ) and promotes podosome forma - tion ( Wang et al . , 2009 ) . GRAF1 - positive adhesive structures closely resembled podo - some - like adhesions ( PLAs ) previously described to be induced by expression of dominant - active cdc42 in HeLa cells ( Dutartre et al . , 1996 ) . Therefore , to further verify that GRAF1 localized to cdc42 - induced adhesions , we coexpressed dominant active cdc42 ( cdc42 Q61L ; herein named cdc42 - DA ) with GRAF1 and costained these cells for vinculin . Indeed , we found that co - overexpressing cells ex - hibited GRAF1 in cdc42 - and vinculin - positive PLAs ( Figure 2C ) . Interestingly , these structures were found to contain punctate actin and dynamin , consistent with their similarity to podosomes and the fact that dynamin has a role at adhesions ( see Figure S1 , E and F ) . To verify that it is prolonged local cdc42 activity that promotes ad - hesion localization of GRAF1 , we expressed the GRAF1 arginine finger mutant , GRAF1 R412D , in cells . This mutant , which is inca - pable of stimulating GTP hydrolysis by cdc42 , significantly in - creased the proportion of cells with abundant PLAs containing both vinculin and GRAF1 ( Figures 2B and S1G ) . The R412D mutation 4382 | G . J . Doherty et al . Molecular Biology of the Cell FIGURE 1 : GRAF1 interacts with proteins involved in membrane remodeling and cell adhesion . ( A ) Domain model of GRAF1 . ( B ) Immunoprecipitation of GRAF1 from rat brain cytosol showing that GRAF1 interacts with dynamin , GIT1 , and FAK as identified by mass spectrometry and confirmed by Western blotting with the indicated antibodies . ( C ) Immunoprecipitates of GIT1 , FAK , dynamin , and pFAK from rat brain cytosol were analyzed by SDS \u2013 PAGE and immunoblotting with antibodies against the indicated proteins . ( D ) Coomassie - stained gel of pulldown experiments against brain lysates with beads bound to GST or GST - tagged GRAF1 SH3 domain ( GST - SH3 ) . Indicated proteins were identified by mass spectrometry . ( E ) Confocal micrograph of a HeLa cell expressing GFP - GIT1 and myc - GRAF1 , costained for dynamin . Merged image shows section views ( as indicated by yellow lines ) . Note that structures with all three proteins colocalizing are found at the basal surface , while dynamin is also found at the top surface . Scale bar : 5 \u03bcm . ( F ) Schematic representation of the GRAF1 interactome , showing the interactions that link cell adhesion , small G - protein regulation , and GRAF1 - mediated membrane trafficking ( Hildebrand et al . , 1996 ; Zamir and Geiger , 2001a , 2001b ; Hoefen and Berk , 2006 ; Lundmark et al . , 2008 ) . Dotted lines show interactions known to be directly activating ( arrowheads ) or inhibiting ( no arrowheads ) the active state of depicted small G protein . Volume 22 November 15 , 2011 GRAF1 trafficking directs cell spreading | 4383 also significantly abrogated GRAF1\u2019s ability to cause profound cellular morphological changes ( Figure S1B ) . GRAF1 localizes to src - induced podosomes Podosome formation is highly stimulat ed by src kinase through activation of cdc42 , and active src has been used to promote podosome formation in a number of cell lines ( Moreau et al . , 2006 ) . To assay whether src - induced podosomes contained GRAF1 , we expressed the dominant active version of src ( src Y527F ) in HeLa cells . This clearly induced both small and rosette - like basal adhesions that , similarly to podosomes , were positive for vinculin , cortactin , and pFAK ( Figures 2D and S2 ) . When costained for GRAF1 , the majority ( 76 % ) of cells with such podosomes were found to have GRAF1 highly enriched in these structures ( Figure 2D ) . These results verified the similarity be - tween podosomes and GRAF1 - positive adhesive structures and indicated that GRAF1 might have a role at bona fide po - dosomes . On the basis of size and appear - ance ; the presence of dynamin , punctate actin , and cortactin ; and the dependence on cdc42 and src activity , we propose that GRAF1 is present at a subset of adhesions most accurately described to date as PLAs . Acute adhesion site modulation by small molecules results in clustering of CTxB in GRAF1 - positive PLAs Cell - matrix adhesions are dependent upon the contractive force mediated by myosin and actin filaments . This process is controlled by cdc42 , rhoA , and rac1 , and induction of PLAs is thought to rely upon cdc42 activity ( Nobes and Hall , 1995 ; Moreau et al . , 2006 ) . To favor cdc42 activ - ity over rhoA activity acutely , we inhibited the rhoA effector protein rho kinase ( ROCK ) , a key regulator of adhesions ( Totsukawa et al . , 2000 ) . Inhibition of ROCK using the small molecule Y - 27632 led to a dramatic loss of actin stress fibers and focal adhesions in favor of PLAs ( Figure S3A ) . In GRAF1 - expressing cells , we observed a profound increase in the proportion of cells exhibiting PLA - local - ized GRAF1 on their basal surfaces ( Figures 2 , B and E , and S3 , B \u2013 D ) . Similarly , the drug blebbistatin , which induces loss of stress fibers , resulted in a significantly in - creased number of PLAs positive for GRAF1 ( Figures 2B and S3E ) . The membrane at cell - matrix adhesions is composed of highly ordered lipids that are also necessary for endocytosis via CLICs FIGURE 2 : GRAF1 is not a general component of focal adhesions but localizes to PLAs . ( A ) Fluorescence micrograph of HeLa cells coexpressing myc - tagged GRAF1 and GFP - tagged GIT1 and costained for myc and vinculin . Insets show magnifications of the areas indicated by yellow squares . ( B ) Bar graph showing the percentage of cells in which myc - GRAF1 or myc - GRAF1 R412D was found localized with vinculin in PLAs following the indicated treatments or overexpression of GFP - GIT1 . Cells expressing myc - GRAF1 were treated with Y - 27632 , an inhibitor of rhoA kinase ( 5 min ) , or blebbistatin , an inhibitor of nonmuscle myosin II ( 10 min ) , and the number of cells where GRAF1 colocalized with vinculin was counted . Bars and error bars correspond to mean and SEM calculated from three independent experiments ( n > 250 ; \u03b1 = 0 . 05 ; two - tailed Fisher\u2019s exact test , * * * , p < 0 . 001 ) . ( C ) Fluorescence micrograph of cells coexpressing myc - tagged GRAF1 and GFP - cdc42 - DA and costained for myc and vinculin . ( D ) Merged confocal micrograph of a cell expressing src Y527F and GFP - GRAF1 and costained for vinculin . Inset below shows three - dimensional view of the section indicated by the dotted line . Bar graph depicting the percentage of cells with untagged or GFP - tagged GRAF1 localized to vinculin - defined , src - induced podosomes . Bars and error bars correspond to mean and SEM calculated from six independent experiments , each including 30 cells . ( E ) Fluorescence micrographs of cells treated with the ROCK inhibitor Y - 27632 for 15 min or untreated before fixation and staining for vinculin and overexpressed myc - GRAF1 . Insets in the top panel show magnification and three - dimensional rotation of the area indicated in the vinculin panel . Arrows indicate basal structures in which GRAF1 and vinculin colocalize . Scale bars : 10 \u03bcm . 4384 | G . J . Doherty et al . Molecular Biology of the Cell Membrane remodeling by GRAF1 and uptake of CTxB is stimulated at the leading edge of migrating cells As previously shown ( Lundmark et al . , 2008 ) , CTxB was internalized by tubular GRAF1 - positive structures ( Figure 4A ) . We ob - served that these structures communicated with adhesion sites , supporting a role for GRAF1 in promoting endocytosis of lipids enriched in sites of high membrane order ( Figure 4A ) . Similarly , the long , \u201ctrapped\u201d membrane tubules generated by the over - expression of GRAF1 BARPH truncation mutant ( Lundmark et al . , 2008 ) spanned long stretches along the basal surface of cells , linking mature focal adhesions ( Figure S4C ) . When tracked by live - cell imaging , the formation of GRAF1 R412D - and GRAF1 BARPH - positive tubules ( which are more stable then GRAF1 - positive tubules ) was detected . These tubules were found to both originate from , and specifically communi - cate with , sites of adhesions ( Figure S4 and Supplemental Movies S1 \u2013 S4 ) . This sug - gested that GRAF1 - mediated membrane trafficking is coupled to adhesion modula - tion . To study how GRAF1 - mediated and clathrin - dependent endocytic events are in - fluenced by the movement of cells , we ex - amined the uptake of the CLIC cargo CTxB and the clathrin - mediated endocytosis cargo transferrin ( Tfn ) by migrating fibro - blasts ( which are more migratory than HeLa cells ; Figure 4B ) . We found that the uptake of CTxB was concentrated at the leading edge of the cell . By contrast , Tfn , a clathrin - dependent endocytic cargo , appeared to be endocytosed homogeneously from the plasma membrane , consistent with previous studies ( Howes et al . , 2010a ) . Furthermore , CTxB - enriched endocytic structures were found to colocalize with paxillin , a marker of integrin - based adhesions , to a significantly higher extent than with Tfn ( Figure 4C ) . Given these data , one might therefore expect that GRAF1 - medi - ated endocytic activity would be affected by adhesion and spread - ing of cells . Rac1 promotes cell spreading and lamellopodia / ruffles at the leading edges of cells ( Nobes and Hall , 1995 ) . We therefore examined whether rac1 - induced cell spreading could also induce GRAF1 - mediated endocytosis . Cells expressing dominant - active rac1 ( rac1 Q61L ; herein named rac1 - DA ) had a round , spread - out morphological phenotype with prevalent ruffled structures contain - ing rac1 . Strikingly , when GRAF1 was co - overexpressed in these cells , we found it on large numbers of membrane tubules at the ruffled regions ( Figure 4D and Movie S5 ) . These tubular structures contained CTxB ( Figure 4D , right panel ) . In contrast with both cdc42 and rac1 , rhoA activity is known to promote cell adhesion and contraction . Expression of DA rhoA ( rhoA Q63L ; herein named rhoA - DA ) results in the rounding up of cells and promotion of ma - ture focal adhesions . Co - overexpression of rhoA - DA and GRAF1 likewise led to contracted cells ( owing to actin stress fiber formation and maintenance ; Figure 4E ) . GRAF1 - positive tubular structures ( Sims and Dustin , 2002 ; Gaus et al . , 2006 ) . We therefore exam - ined whether induced PLAs and podosomes contained the GM1 receptor , a glycosphingolipid that is enriched in microdomains . When CTxB , which binds GM1 specifically , was added to Y - 27632 - treated cells , we found that it clustered in vinculin - posi - tive structures on the membrane , overlapping closely with GRAF1 localization ( Figure 3A ) . Treatment with blebbistatin like - wise resulted in clustering of CTxB at PLAs containing GRAF1 ( Supplemental Figure S3E ) . We measured the amount of colocal - ization and found that approximately one - half the area of acutely induced PLAs contained CTxB ( Figure 3B ) . Our data show that acute induction of adhesion turnover affects membrane order and GRAF1 activity , which supports the view that plasma mem - brane dynamics are closely coupled to cell - matrix adhesion turn - over . The presence of GRAF1 at src - induced podosomes indi - cates that these structures might have a similar membrane composition . Indeed , we found that CTxB was clustered in src - induced podosomes together with GRAF1 ( Figure 3C ) . FIGURE 3 : Adhesion reorganization results in clustering of CTxB at the cell surface in GRAF1 - positive structures . ( A ) Fluorescence micrographs of myc - GRAF1 \u2013 expressing HeLa cells incubated with Y - 27632 together with CTxB - Alexa555 for 5 min as indicated before washing , fixation , and costaining for vinculin . Insets are magnifications of the area marked by a rectangle in panel 1 . ( B ) Vinculin - and myc - GRAF1 - positive PLA areas were selected from four cells from two different experiments , and the percent colocalization of CTxB was measured as described in the text . The mean is indicated by a red line and the error bars represent standard deviation ( SD ) above and below the mean . ( C ) Confocal micrograph of cell expressing src Y527F and GFP - GRAF1 and incubated together with CTxB - Alexa555 for 5 min before washing , fixation , and costaining for vinculin . The merged image is rotated to highlight the basal localization of GRAF1 - positive structures . Scale bars ; 10 \u03bcm . Volume 22 November 15 , 2011 GRAF1 trafficking directs cell spreading | 4385 overexpression of the dominant - negative GRAF1 variant GRAF1 BARPH ( Lundmark et al . , 2008 ) , suggesting that the associated membrane trafficking is similarly abrogated . GRAF1 is necessary for efficient spreading and migration of HeLa cells To determine whether GRAF1 activity was necessary for spreading and migration of cells , we depleted cells of GRAF1 using small interfering RNAs ( siRNAs ) previously shown to profoundly abrogate CLIC / endo - cytic membrane manufacture ( Lundmark et al . , 2008 ) . Immunoprecipitation of GRAF1 from cell lysates and detection by immunob - lotting showed that GRAF1 was efficiently depleted using either of two different siRNAs ( Figure 5A ) . When cells were seeded sparsely and allowed to spread after deattachment , we noted that GRAF1 - depleted cells were elongated and unable to spread fully , in con - trast with validated controls ( Figure 5B ) . We measured the area of experimental cells and found that GRAF1 - depleted cells had significantly smaller surface - connected areas compared with control cells . We also found that cells lacking GRAF1 had approximately double the average length : width ratios of controls ( Figure 5C ) . Interestingly , cells lack - ing rac1 and rac2 displayed a similar elon - gated phenotype ( Wheeler et al . , 2006 ) . Two large - scale siRNA screens have identified the gene encoding GRAF1 as being involved in migration and adhesion ( Simpson et al . , 2008 ; Winograd - Katz et al . , 2009 ) . We ob - served that GRAF1 - depleted cells were defi - cient in migrating into an induced wound in a cell monolayer ( Figure 5D ) , and the distri - bution of mature focal adhesions was fre - quently disturbed ( Figure 5E ) . To obtain real - time quantitative data on migration , we assayed the ability of cells to \u201cheal\u201d an elec - trically induced wound . While control cells closed the wound within 5 \u2013 6 h , GRAF1 - de - pleted cells were profoundly deficient in their abilities to spread and migrate into this wound ( Figure 5F ) . Taken together , our data demonstrate a role for GRAF1 in cell migra - tion , due to regulation of cell - matrix adhe - sion sites through endocytic turnover of membranes . DISCUSSION The mechanisms by which plasma mem - brane morphological changes and mem - brane trafficking are coupled in processes such as cell migration are greatly debated . Many studies have high - lighted the importance of specific rho - family , small G proteins in both cytoskeletal dynamics and endocytic events ( Mayor and Pagano , 2007 ; Doherty and McMahon , 2008 ) . Our studies reported here on membrane remodeling and rhoGAP protein GRAF1 have revealed that a prevalent clathrin - independent endocytic pathway is were found only in a small subset of rhoA - DA \u2013 expressing cells ( 14 . 3 \u00b1 3 . 7 % ) . These tubules were much longer than GRAF1 - positive tubules in rac1 - expressing cells and communicated with focal adhe - sions ( Figure 4E ) . However , GRAF1 was not enriched at the abun - dant , mature focal adhesions in rhoA - DA \u2013 expressing cells . The tubular phenotype was strikingly reminiscent of that observed upon FIGURE 4 : Membrane remodeling by GRAF1 and uptake of CTxB are stimulated at the leading edge of migrating cells . ( A ) Fluorescence micrographs of myc - GRAF1 \u2013 expressing HeLa cells incubated with CTxB - Alexa555 for 5 min before washing , fixation , and costaining for vinculin . ( B ) Confluent wild - type MEF monolayers were wounded by scratching , and cells were allowed to migrate into the wound for 4 \u2013 6 h . CTxB - 555 and Tfn - 647 were then added to migrating cells for 2 min of uptake at 37\u00b0C . Cells were acid - stripped and fixed and were then labeled for endogenous paxillin . Arrows indicate colocalization between paxillin and CTxB but not Tfn . ( C ) Twenty - four cells across three independent experiments were treated as in ( B ) , and the percentage of paxillin ( green ) pixels that colocalized with either CTxB ( red ) or Tfn ( blue ) pixels was calculated using Volocity version 3 . 0 . Bars and error bars correspond to mean and SEM ( n = 12 \u2013 15 ; \u03b1 = 0 . 05 ; Student\u2019s t test , * * , p < 0 . 01 ) . ( D ) Fluorescence micrographs of HeLa cells expressing myc - GRAF1 and DA rac1 ( GFP - Rac - DA ; left panels ) and incubated with CTxB - Alexa555 for 5 min ( right panels ) . The length of GRAF1 tubules was measured in fluorescence micrographs of seven different cells ( n = 193 ) as described . The mean length is indicated by a red line and the error bars represent standard deviation ( SD ) above and below the mean . ( E ) Fluorescence micrographs of cells expressing myc - GRAF1 and DA rhoA ( GFP - rhoA - DA ) and costained for vinculin . GRAF1 - positive tubules were found in 14 . 3 \u00b1 3 . 7 % , as determined from three independent experiments ( n = 124 ; error values represent SEM ) . Length of GRAF1 tubules in \u03bcm was measured from nine different cells ( n = 108 ) and depicted as in Figure 4D . Scale bars : 10 \u03bcm . 4386 | G . J . Doherty et al . Molecular Biology of the Cell intimately coupled to adhesion site regula - tion . GRAF1 - dependent CLICs are induced by cell spreading , enriched at the leading edge of cells , and necessary for cell spread - ing and migration . Interestingly , proteomic analysis has recently revealed enrichment for cargo molecules linked to adhesions ( Howes et al . , 2010a ) . Although the exact role of GRAF1 in receptor trafficking is not known , our data illuminate an important cell physiological function for GRAF1 - depen - dent CLICs , and provide further insight into how cells coordinate membrane and protein redistribution with changes in their morphologies . While it has previously been suggested ( Hildebrand et al . , 1996 ) that GRAF1 func - tions at focal adhesions , owing to its interac - tion with FAK , we find no evidence that GRAF1 is a structural component of such ad - hesions . However , we found GRAF1 at what , at present , can be best described as PLAs . These structures could be induced by pro - longed src and cdc42 activity ( through ex - pression of active src , cdc42 - DA , or GRAF1 R412D , or by inhibition of ROCK using small molecules ) and are enriched in cortactin , ac - tin , dynamin , actin , vinculin , FAK , and GIT1 . Although the exact role of GRAF1 - positive adhesions is not clear , the localization to src - induced podosomes indicates a relevant function of GRAF1 at such sites . Podosomes are classically found in cells of the monocytic lineage but are becoming increasingly rec - ognized as being present in a number of cell FIGURE 5 : GRAF1 is necessary for cell spreading and migration . ( A ) Immunoprecipitation and immunodetection of GRAF1 from HeLa cells treated with either a control siRNA or siRNAs against GRAF1 . Tubulin was detected in the cell lysates as an immunoprecipitation control . GRAF1 expression following siRNA treatment ( GRAF1 expr . ( % ) ) was quantified as described from three independent experiments . Values correspond to normalized means \u00b1 SD . ( B ) Micrographs of cells treated with either a control siRNA or siRNAb against GRAF1 for 72 h before fixation and imaging . Inset 1 ( left panel ) shows magnification of the marked square . Inset 1 ( right panel ) exemplifies cell area indicated in blue , and length and width indicated in red and green , respectively . ( C ) Bar graphs showing quantification of the cell area and length : width ratios , as described in the text , of cells treated with control siRNA or siRNAs against GRAF1 , as in ( B ) . In the left panel , bars and error bars correspond to mean and SEM calculated from five or six independent experiments ( n = 50 \u2013 60 ; \u03b1 = 0 . 05 ; Student\u2019s t test , * * * , p < 0 . 001 ) . In the right panel , bars and error bars correspond to mean and SEM calculated from two independent experiments ( n > 100 ; a \u03b1 = 0 . 05 ; Student\u2019s t test , * * * , p < 0 . 001 ) . Length : width ratio was scored as the ratio between length ( longest straight line within a cell ( red line in ( B ) ) and width ( the broadest region perpendicular to the measured length ( green line in ( B ) ) . ( D ) Micrograph showing the regrowth of control siRNA - treated and GRAF1 siRNA - treated cells into an induced wound in the cell monolayer . ( E ) Fluorescence micrograph of control cells and GRAF1 - depleted cells stained for vinculin . ( F ) Principle of electrically induced and monitored wound - healing assay performed as described ( left panel ) . Graph showing the recovery from electrical wound healing of a confluent HeLa cell layer . Cells were previously transfected with a control siRNA or GRAF1 - siRNA , as indicated . An increase in impedance reflects the migration of surrounding healthy cells onto an electrode through which ( at time zero ) a high current has passed to irreversibly injure the cells on the electrode . Note the delay in wound healing observed in GRAF1 - depleted cells and their slowed and incomplete recovery ( even after 9 h ) , compared with control cells . Impedance values were normalized to postwounding nadirs , and the shaded areas surrounding each curve represent one SD above and below the mean values for each condition . Scale bars : 10 \u03bcm . Volume 22 November 15 , 2011 GRAF1 trafficking directs cell spreading | 4387 of cells . Indeed , it may be that it is the permissive state of lipids that allows for the spatiotemporal coupling of adhesion site turnover and membrane trafficking . GRAF1 might be recruited to such domains preferentially , owing to both their distinct membrane composition and local , small G - protein activity . The extent of GRAF1 - mediated endocytosis linked to adhesion sites versus endocytosis at nonadhe - sive microdomains appears high , and it appears that this ratio is co - ordinated with the migratory behavior of the cell . Taken together , our data support a role for GRAF1 at adhesive microdomains , where its membrane remodeling activity promotes the manufacture of CLICs . This activity is promoted at the leading edge of cells , where it is required for membrane spreading and cell migration . MATERIALS AND METHODS Reagents , antibodies , and cDNA contructs Polyclonal antisera against GRAF1 ( RA - 83 and RaZ1 ) were generated as previously described ( Lundmark et al . , 2008 ) . Purchased antibod - ies were : mouse anti - myc clone 9E10 and rabbit anti - myc ( Cell Sig - naling Technology , Danvers , MA ) , mouse anti - dynamin ( Hudy ) , mouse anti - GIT1 ( BD Biosciences ) , and mouse anti - paxillin , mouse anti - vin - culin , rabbit anti - FAK , rabbit anti - pFAK , and rabbit anti - cortactin ( Abcam , Cambridge , MA ) . All secondary antibodies were conjugated to Alexa Fluor 488 , 546 , or 647 ( Invitrogen , Carlsbad , CA ) . cDNA constructs encoding human GRAF1 and derivatives were as de - scribed previously ( Lundmark et al . , 2008 ) , except for the amino acid substitution R412D , which was created using PCR - directed mutagen - esis ( Stratagene , Agilent , Santa Clara , CA ) . Red fluorescent protein ( RFP ) - tagged GRAF1 and untagged GRAF1 was cloned using the Gateway system ( Invitrogen ) . Green fluorescent protein ( GFP ) - tagged cdc42 Q61L ( 12600 ; Nalbant et al . , 2004 ) , rac1 Q61L ( 12981 ) , rhoA Q63L ( 12968 ; Subauste et al . , 2000 ) , and untagged src Y527F ( 13660 ) were purchased from Addgene ( Cambridge , MA ) . GFP - GIT1 and RFP - paxillin were kind gifts from A . F . Horwitz and E . E . Marcantonio , respectively . Y - 27632 and blebbistatin was purchased from Sigma - Aldrich ( St . Louis , MO ) and used at 10 and 20 \u03bcM , respectively . Immunoprecipitation and pulldown For immunoprecipitation experiments , rat brain cytosol was gener - ated by homogenization of rat brains in buffer ( 25 mM HEPES , 150 mM NaCl , 1 mM dithiothreitol [ DTT ] , 0 . 1 % Triton X - 100 , and protease inhibitors ) , before centrifugation at 50 , 000 rpm for 30 min at 4\u00b0C . The supernatant was removed and added to protein A Sep - harose 4B beads ( GE Healthcare ) , to which antibodies had been pre - viously bound , and incubated at 4\u00b0C for 3 h . Beads were washed three times in buffer ( 25 mM HEPES , 150 mM NaCl ) supplemented with 1 % NP - 40 , and once in buffer without NP - 40 , before analysis by SDS \u2013 PAGE combined with immunoblotting or Coomassie Blue stain - ing . Immunoprecipitation of GRAF1 to determine siRNA efficiency was performed as described above , except that HeLa cells were ly - sed in buffer ( 25 mM HEPES , 150 mM NaCl , 1 % NP - 40 , and protease inhibitors ) , before centrifugation at 14 , 000 rpm for 30 min at 4\u00b0C . Recombinant proteins were expressed in a BL21 ( DE3 ) pLysS Escher - ichia coli strain as glutathione S - transferase ( GST ) - fusion proteins and purified using glutathione - Sepharose 4B beads ( Amersham Bio - sciences ) and gel filtration on a sephacryl S - 200 column ( GE Health - care ) . Pulldown experiments against rat brain cytosol using purified proteins and identification by mass spectrometry were performed as previously described ( Lundmark et al . , 2008 ) . Cell culture and transfections HeLa cells and Balb3T3 cells were grown in DMEM media ( Gibco , Invitrogen ) supplemented with l - glutamine , 10 % fetal bovine types . These enigmatic structures are much more dynamic than focal adhesions and resemble the intermediates formed during focal adhesion formation and turnover ( Block et al . , 2008 ) . Ordinarily , GRAF1 - mediated endocytic activity is very dynamic and most preva - lent at the leading edge of cells , suggesting that GRAF1 - positive PLAs enriched with GM1 might represent a stabilized intermediate that occurs during adhesion turnover ( favored by prolonged cdc42 activity or alterations in the integrity of membrane microdomains ) . Truncated GRAF1 , lacking the regulatory GAP and SH3 domains ( GRAF1 BARPH ) , generates long , tubular membrane structures in cells ( Lundmark et al . , 2008 ) . Interestingly , we found that these tu - bules frequently spanned long stretches between mature focal adhe - sions , suggesting that truncated GRAF1 is unable to appropriately regulate the dynamics of membrane carriers at adhesive sites . The real time formation of GRAF1 - positive carriers showed that such structures both originate at and are targeted to adhesion sites , suggesting that CLICs might control bidirectional trafficking to regu - late membrane turnover . Our biochemical analyses identified interacting partners of GRAF1 that include FAK , GIT1 , and dynamin , supporting a role for GRAF1 at the interface between membrane dynamics and adhesion . Both GIT1 and FAK are known to regulate adhesion sites and influ - ence endocytosis ( Premont et al . , 1998 ; Claing et al . , 2000 ) . In addi - tion to the established role of dynamin during fission of endocytic carriers , dynamin also localizes to podosomes and invadopodia ( Ochoa et al . , 2000 ; Lundmark et al . , 2008 ) . Interestingly , a GTPase - deficient mutant of dynamin was found to perturb actin dynamics at such sites ( Ochoa et al . , 2000 ; Baldassarre et al . , 2003 ) and a mutant lacking its proline - rich domain ( and therefore unable to interact with SH3 domains ) was found to disrupt podosome formation ( Lee and De Camilli , 2002 ) . Dynamin and GRAF1 interact avidly , with both re - sulting in plasma membrane invagination . While fission of CLICs from the plasma membrane has been shown to be capable of func - tioning without dynamin , this does not necessarily mean that all clathrin - independent events are dynamin independent . We show that GRAF1 and dynamin colocalize at PLAs , suggesting that these proteins might organize molecular networks and membrane dynam - ics at such sites . The membrane architecture at podosomes / PLAs is also presently unclear , but it has been suggested that formation of both podosomes and invadopodia involves membrane invagina - tions , and our results support this . Uptake of gold - labeled gelatin via narrow membrane tubules has been described as strongly associ - ated with podosome localization , and it has been suggested that dynamin coats membrane tubules in the central region of podo - somes ( Gawden - Bone et al . , 2010 ; Ochoa et al . , 2000 ) . Further ultra - structural studies will assess the precise proteolipid architecture of PLAs , podosomes , and GRAF1 - positive tubules and determine whether GRAF1 - mediated endocytosis is directly coupled with the internalization of any particular adhesion molecule ( s ) . Previous work has shown that integrin - based adhesions influence membrane order and control the recruitment of active rac1 to cho - lesterol - dependent microdomains at the leading edge , where it is found together with the CTxB receptor GM1 ( del Pozo et al . , 2004 ; Gaus et al . , 2006 ) . Furthermore , detachment of cells from their sub - strata triggers the internalization of highly ordered regions of the plasma membrane ( del Pozo et al . , 2004 ) . We show that GRAF1 - positive PLAs include highly ordered membranes . CLIC manufacture is likewise dependent upon the presence of highly ordered mem - branes , and the removal of these lipids from the plasma membrane could promote adhesion turnover through dissipation of adhesion proteins that preferentially accumulate in microdomains . GRAF1 - mediated CLIC generation is promoted by rac1 at the leading edge 4388 | G . J . Doherty et al . Molecular Biology of the Cell serum , and nonessential amino acids ( for MEM ) , and transfected using Lipofectamine 2000 ( Invitrogen ) or Neon transfection system for electroporation ( Invitrogen ) for transient protein expression . Mouse embryonic fibroblasts ( MEFs ) were generated and grown as previously described ( Kirkham et al . , 2005 ) . For GRAF1 depletion , HeLa cells were transfected with stealth siRNA specific against hu - man GRAF1 ( Invitrogen ) , using Lipofectamine 2000 or Neon trans - fection system for electroporation according to the manufacturer\u2019s instructions . Cells were cultured for 72 h for efficient silencing of the GRAF1 expression . Stealth negative control medium GC Block - it siRNA ( Invitrogen ) was used as a control . GRAF1 siRNAa : GUA AU - CUGUGCUGAAUGGGAGAUAA ; GRAF1 siRNAb : CCACUCAU - GAUGUACCAGUUUCAAA . Fixed - sample and real - time imaging For immunofluorescence analysis , HeLa cells were fixed in 3 % para - formaldehyde in phosphate - buffered saline ( PBS ) for 15 min at 37\u00b0C , then washed and blocked in 5 % goat serum with 0 . 05 % saponin in PBS before staining with the appropriate antibodies in 1 % goat se - rum with 0 . 05 % saponin in PBS using standard protocols . Confocal images were taken sequentially using either a TCS SP5 system con - focal laser - scanning microscope ( Leica Microsystems ) or a fully mo - torized A1 R Laser Scanning Confocal Microscope system ( Nikon Instruments , USA ) using a 60 \u00d7 lens ( Plan Apochromat VC Oil DIC N2 , Nikon ) at appropriate excitation and emission wavelengths un - der control of the NIS - Elements Microscope Imaging Software . Epi - fluorescence and phase - contrast images were taken using a Zeiss Axioimager Z1 system with AxioVision software . Images were pro - cessed using Adobe Photoshop CS2 ( San Jose , CA ) . For immuno - fluorescence trafficking assays in HeLa cells , Alexa Fluor 546 / 555 - con - jugated CTxB ( Invitrogen ) was diluted in prewarmed media , added to cells , and incubated for the time periods and temperatures de - scribed in the figure legends . After being washed , cells were fixed and subjected to immunofluorescence analysis as described above . For polarized uptake experiments , confluent monolayers of MEFs , grown on 12 - mm , round , glass coverslips ( Lomb Scientific , Australia ) were wounded by scratching with a 200 \u03bcl pipette tip . Cells were allowed to migrate into the wound for 4 h before addition of 10 \u03bcg / ml CTxB - 555 ( Invitrogen ) and 20 \u03bcg / ml Tfn - 647 ( Invitrogen ) for 2 min at 37\u00b0C . Cells were fixed in 4 % paraformaldehyde and la - beled with anti - paxillin antibodies ( BD Transduction Laboratories ) . An axiovert 200 m SP LSM 510 META confocal laser - scanning micro - scope ( Zeiss ) was used to capture images , which were processed using Volocity , version 3 . 7 . Quantification and image analysis For quantification of the proportion of cells with protrusions , PLA - lo - calized GRAF1 , GRAF1 localization to src - induced podosomes , and the length of GRAF1 - positive structures , HeLa cells were transfected with constructs for overexpression as indicated in the figure legends and processed for immunofluorescence analysis as described in the preceding section . Protrusive structures ( > 20 \u03bcm long and < 10 - \u03bcm wide protrusions ) were measured using Axiovision software ( Zeiss ) , and the proportion of cells with such protrusions was calculated for each condition ( n > 200 for each condition ) . The percentage of cells with PLA - localized GRAF1 was counted in three independent ex - periments for each condition using the Axiovision software ( n > 250 for each condition ) . The percentage of cells with src - induced podo - somes in which GRAF1 was also localized was calculated from six independent experiments ( 30 cells per experiment ) using the NIS Elements software ( Nikon ) . Length of GRAF1 structures was mea - sured using the ImageJ segmented lines tool . To calculate the length : width ratio of cells , phase - contrast images were captured and the length ( longest straight line within a cell ) and width ( the broadest region perpendicular to the measured length ) was measured using ImageJ . For cell area determination , cell borders from 50 \u2013 60 cells from at least five independent experiments were manually defined in captured phase - contrast images , and the area was calculated using the \u201cOutline\u201d function of the AxioVision software . Creation of bar graphs and statistical analysis was performed using Graphpad Prism ( La Jolla , CA ) . For colocalization experiments in migrating cells , 12 \u2013 15 cells in three independent experiments were processed using Volocity , version 3 . 7 , and colocalization was determined using the colocalization function with automatic threshold . Automatic thresh - olds were applied to individual cells , and percentage of overlapping pixels was calculated for CTxB and Tfn channels against paxillin . For quantification of the colocalization between CTxB and PLAs , 3 \u2013 10 separate images of each condition were thresholded for the bright - est areas ; overlapping areas were then transferred into a new chan - nel . GRAF1 - and vinculin - positive PLA areas were manually defined , and the amount of CTxB colocalization in these areas was calculated using Adobe Photoshop . For determining the amount of GRAF1 fol - lowing siRNA treatment , intensity of GRAF1 and tubulin bands iden - tified by immunoblotting from three independent experiments was quantified using ImageJ . GRAF1 intensity was related to tubulin in - tensity for each sample , and the amount of GRAF1 in control cells was set to 100 % . Biophysical cell recordings HeLa cells ( 10 5 ) were transfected with siRNA against GRAF1 or con - trol siRNA for 24 h before being plated into chambers of 8W1E elec - trode arrays ( Applied Biophysics , Troy , NY ) and incubated at 37\u00b0C with 5 % CO 2 . Impedance values between the electrode and coun - terelectrode were recorded continuously from each array at a 15 - kHz oscillator frequency using an ECIS 1600 system with elevated field module ( Electric Cell - substrate Impedance Sensing , Applied Bio - physics ) . Cell attachment , spreading , and layer confluence were verified electrically and microscopically before electrical wounding at 45 kHz , 4 V for 10 s with subsequent recording from electrodes using the same parameters as prewounding . Data were normalized to initial electrode impedance value for each wounding experiment . ACKNOWLEDGMENTS This work was supported by the Swedish Cancer Society ; the Swed - ish Medical Research Council ; Swedish foundation for strategic re - search ; the Medical Faculty , Ume\u00e5 University ; the Royal Swedish Academy of Sciences ; the Magn Bergvall Foundation ; the Harald Jeanssons Foundation ; the \u00c5ke Wibergs Foundation ; and the Na - tional Health and Medical Research Council of Australia ( R . G . P . ) . Gary Doherty was supported by a Trinity College , Cambridge , Internal Graduate Studentship and Research Scholarship and an MRC post - doctoral award . Many thanks to Sven Carlsson and all members of the McMahon lab for help and support . Special thanks to Safa Lucken - Ardjomande for technical assistance regarding molecular cloning and live - cell imaging . Confocal microscopy on migrating cells was performed at the Australian Cancer Research Foundation ( ACRF ) / In - stitute for Molecular Bioscience Dynamic Imaging Facility for Cancer Biology , which was established with the support of the ACRF . REFERENCES Baldassarre M , Pompeo A , Beznoussenko G , Castaldi C , Cortellino S , McNiven MA , Luini A , Buccione R ( 2003 ) . Dynamin participates in focal extracellular matrix degradation by invasive cells . Mol Biol Cell 14 , 1074 \u2013 1084 . Volume 22 November 15 , 2011 GRAF1 trafficking directs cell spreading | 4389 Block MR , Badowski C , Millon - Fremillon A , Bouvard D , Bouin AP , Faurobert E , Gerber - Scokaert D , Planus E , Albiges - Rizo C ( 2008 ) . Podosome - type adhesions and focal adhesions , so alike yet so different . Eur J Cell Biol 87 , 491 \u2013 506 . Caswell PT , Vadrevu S , Norman JC ( 2009 ) . Integrins : masters and slaves of endocytic transport . Nat Rev Mol Cell Biol 10 , 843 \u2013 853 . Chadda R , Howes MT , Plowman SJ , Hancock JF , Parton RG , Mayor S ( 2007 ) . Cholesterol - sensitive Cdc42 activation regulates actin polymerization for endocytosis via the GEEC pathway . Traffic 8 , 702 \u2013 717 . Claing A , Perry SJ , Achiriloaie M , Walker JK , Albanesi JP , Lefkowitz RJ , Premont RT ( 2000 ) . Multiple endocytic pathways of G protein - coupled receptors delineated by GIT1 sensitivity . Proc Natl Acad Sci USA 97 , 1119 \u2013 1124 . del Pozo MA , Alderson NB , Kiosses WB , Chiang HH , Anderson RG , Schwartz MA ( 2004 ) . Integrins regulate Rac targeting by internalization of membrane domains . Science 303 , 839 \u2013 842 . Doherty GJ , McMahon HT ( 2008 ) . Mediation , modulation , and conse - quences of membrane - cytoskeleton interactions . Annu Rev Biophys 37 , 65 \u2013 95 . Doherty GJ , McMahon HT ( 2009 ) . Mechanisms of endocytosis . Annu Rev Biochem 78 , 857 \u2013 902 . Dutartre H , Davoust J , Gorvel JP , Chavrier P ( 1996 ) . Cytokinesis arrest and redistribution of actin - cytoskeleton regulatory components in cells expressing the Rho GTPase CDC42Hs . J Cell Sci 109 , 367 \u2013 377 . Ezratty EJ , Bertaux C , Marcantonio EE , Gundersen GG ( 2009 ) . Clathrin me - diates integrin endocytosis for focal adhesion disassembly in migrating cells . J Cell Biol 187 , 733 \u2013 747 . Ezratty EJ , Partridge MA , Gundersen GG ( 2005 ) . Microtubule - induced fo - cal adhesion disassembly is mediated by dynamin and focal adhesion kinase . Nat Cell Biol 7 , 581 \u2013 590 . Gaus K , Le Lay S , Balasubramanian N , Schwartz MA ( 2006 ) . Integrin - medi - ated adhesion regulates membrane order . J Cell Biol 174 , 725 \u2013 734 . Gawden - Bone C , Zhou Z , King E , Prescott A , Watts C , Lucocq J ( 2010 ) . Dendritic cell podosomes are protrusive and invade the extracellular matrix using metalloproteinase MMP - 14 . J Cell Sci 123 , 1427 \u2013 1437 . Hansen CG , Nichols BJ ( 2009 ) . Molecular mechanisms of clathrin - indepen - dent endocytosis . J Cell Sci 122 , 1713 \u2013 1721 . Hildebrand JD , Taylor JM , Parsons JT ( 1996 ) . An SH3 domain - containing GTPase - activating protein for Rho and Cdc42 associates with focal adhesion kinase . Mol Cell Biol 16 , 3169 \u2013 3178 . Hoefen RJ , Berk BC ( 2006 ) . The multifunctional GIT family of proteins . J Cell Sci 119 , 1469 \u2013 1475 . Howes MT et al . ( 2010a ) . Clathrin - independent carriers form a high capacity endocytic sorting system at the leading edge of migrating cells . J Cell Biol 190 , 675 \u2013 691 . Howes MT , Mayor S , Parton RG ( 2010b ) . Molecules , mechanisms , and cel - lular roles of clathrin - independent endocytosis . Curr Opin Cell Biol 22 , 519 \u2013 527 . Kirkham M , Fujita A , Chadda R , Nixon SJ , Kurzchalia TV , Sharma DK , Pagano RE , Hancock JF , Mayor S , Parton RG ( 2005 ) . Ultrastructural iden - tification of uncoated caveolin - independent early endocytic vehicles . J Cell Biol 168 , 465 \u2013 476 . Kumari S , Mayor S ( 2008 ) . ARF1 is directly involved in dynamin - independent endocytosis . Nat Cell Biol 10 , 30 \u2013 41 . Lahuna O , Quellari M , Achard C , Nola S , Meduri G , Navarro C , Vitale N , Borg JP , Misrahi M ( 2005 ) . Thyrotropin receptor trafficking relies on the hScrib - \u03b2 PIX - GIT1 - ARF6 pathway . EMBO J 24 , 1364 \u2013 1374 . Lee E , De Camilli P ( 2002 ) . Dynamin at actin tails . Proc Natl Acad Sci USA 99 , 161 \u2013 166 . Lundmark R , Doherty GJ , Howes MT , Cortese K , Vallis Y , Parton RG , McMahon HT ( 2008 ) . The GTPase - activating protein GRAF1 regulates the CLIC / GEEC endocytic pathway . Curr Biol 18 , 1802 \u2013 1808 . Manabe R , Kovalenko M , Webb DJ , Horwitz AR ( 2002 ) . GIT1 functions in a motile , multi - molecular signaling complex that regulates protrusive activity and cell migration . J Cell Sci 115 , 1497 \u2013 1510 . Mayor S , Pagano RE ( 2007 ) . Pathways of clathrin - independent endocytosis . Nat Rev Mol Cell Biol 8 , 603 \u2013 612 . Moreau V , Tatin F , Varon C , Anies G , Savona - Baron C , Genot E ( 2006 ) . Cdc42 - driven podosome formation in endothelial cells . Eur J Cell Biol 85 , 319 \u2013 325 . Nalbant P , Hodgson L , Kraynov V , Toutchkine A , Hahn KM ( 2004 ) . Activa - tion of endogenous Cdc42 visualized in living cells . Science 305 , 1615 \u2013 1619 . Nobes CD , Hall A ( 1995 ) . Rho , rac , and cdc42 GTPases regulate the assem - bly of multimolecular focal complexes associated with actin stress fibers , lamellipodia , and filopodia . Cell 81 , 53 \u2013 62 . Ochoa GC et al . ( 2000 ) . A functional link between dynamin and the actin cytoskeleton at podosomes . J Cell Biol 150 , 377 \u2013 389 . Premont RT , Claing A , Vitale N , Freeman JL , Pitcher JA , Patton WA , Moss J , Vaughan M , Lefkowitz RJ ( 1998 ) . \u03b2 2 - Adrenergic receptor regulation by GIT1 , a G protein - coupled receptor kinase - associated ADP ribosyla - tion factor GTPase - activating protein . Proc Natl Acad Sci USA 95 , 14082 \u2013 14087 . Sabharanjak S , Sharma P , Parton RG , Mayor S ( 2002 ) . GPI - anchored proteins are delivered to recycling endosomes via a distinct cdc42 - regulated , clathrin - independent pinocytic pathway . Dev Cell 2 , 411 \u2013 423 . Schlunck G , Damke H , Kiosses WB , Rusk N , Symons MH , Waterman - Storer CM , Schmid SL , Schwartz MA ( 2004 ) . Modulation of Rac localization and function by dynamin . Mol Biol Cell 15 , 256 \u2013 267 . Simpson KJ , Selfors LM , Bui J , Reynolds A , Leake D , Khvorova A , Brugge JS ( 2008 ) . Identification of genes that regulate epithelial cell migration using an siRNA screening approach . Nat Cell Biol 10 , 1027 \u2013 1038 . Sims TN , Dustin ML ( 2002 ) . The immunological synapse : integrins take the stage . Immunol Rev 186 , 100 \u2013 117 . Subauste MC , Von Herrath M , Benard V , Chamberlain CE , Chuang TH , Chu K , Bokoch GM , Hahn KM ( 2000 ) . Rho family proteins modulate rapid apoptosis induced by cytotoxic T lymphocytes and Fas . J Biol Chem 275 , 9725 \u2013 9733 . Taylor JM , Macklem MM , Parsons JT ( 1999 ) . Cytoskeletal changes induced by GRAF , the GTPase regulator associated with focal adhesion kinase , are mediated by Rho . J Cell Sci 112 , 231 \u2013 242 . Totsukawa G , Yamakita Y , Yamashiro S , Hartshorne DJ , Sasaki Y , Matsumura F ( 2000 ) . Distinct roles of ROCK ( Rho - kinase ) and MLCK in spatial regu - lation of MLC phosphorylation for assembly of stress fibers and focal adhesions in 3T3 fibroblasts . J Cell Biol 150 , 797 \u2013 806 . Wang J , Taba Y , Pang J , Yin G , Yan C , Berk BC ( 2009 ) . GIT1 mediates VEGF - induced podosome formation in endothelial cells : critical role for PLC\u03b3 . Arterioscler Thromb Vasc Biol 29 , 202 \u2013 208 . Wheeler AP , Wells CM , Smith SD , Vega FM , Henderson RB , Tybulewicz VL , Ridley AJ ( 2006 ) . Rac1 and Rac2 regulate macrophage morphology but are not essential for migration . J Cell Sci 119 , 2749 \u2013 2757 . Winograd - Katz SE , Itzkovitz S , Kam Z , Geiger B ( 2009 ) . Multiparametric analysis of focal adhesion formation by RNAi - mediated gene knock - down . J Cell Biol 186 , 423 \u2013 436 . Zamir E , Geiger B ( 2001a ) . Components of cell - matrix adhesions . J Cell Sci 114 , 3577 \u2013 3579 . Zamir E , Geiger B ( 2001b ) . Molecular complexity and dynamics of cell - matrix adhesions . J Cell Sci 114 , 3583 \u2013 3590 . Zhao ZS , Manser E , Loo TH , Lim L ( 2000 ) . Coupling of PAK - interacting exchange factor PIX to GIT1 promotes focal complex disassembly . Mol Cell Biol 20 , 6354 \u2013 6363 .",
    "cureton2010length": "The Length of Vesicular Stomatitis Virus Particles Dictates a Need for Actin Assembly during Clathrin - Dependent Endocytosis David K . Cureton 1 , Ramiro H . Massol 2 , Sean P . J . Whelan 3 * , Tomas Kirchhausen 1 * 1 Department of Cell Biology , Harvard Medical School , and Immune Disease Institute at Children\u2019s Hospital , Boston , Massachusetts , United States of America , 2 The Division of Gastroenterology and Nutrition , Children\u2019s Hospital , Boston , Massachusetts , United States of America , 3 Department of Microbiology and Molecular Genetics , Harvard Medical School , Boston , Massachusetts , United States of America Abstract Microbial pathogens exploit the clathrin endocytic machinery to enter host cells . Vesicular stomatitis virus ( VSV ) , an enveloped virus with bullet - shaped virions that measure 70 6 200 nm , enters cells by clathrin - dependent endocytosis . We showed previously that VSV particles exceed the capacity of typical clathrin - coated vesicles and instead enter through endocytic carriers that acquire a partial clathrin coat and require local actin filament assembly to complete vesicle budding and internalization . To understand why the actin system is required for VSV uptake , we compared the internalization mechanisms of VSV and its shorter ( 75 nm long ) defective interfering particle , DI - T . By imaging the uptake of individual particles into live cells , we found that , as with parental virions , DI - T enters via the clathrin endocytic pathway . Unlike VSV , DI - T internalization occurs through complete clathrin - coated vesicles and does not require actin polymerization . Since VSV and DI - T particles display similar surface densities of the same attachment glycoprotein , we conclude that the physical properties of the particle dictate whether a virus - containing clathrin pit engages the actin system . We suggest that the elongated shape of a VSV particle prevents full enclosure by the clathrin coat and that stalling of coat assembly triggers recruitment of the actin machinery to finish the internalization process . Since some enveloped viruses have pleomorphic particle shapes and sizes , our work suggests that they may use altered modes of endocytic uptake . More generally , our findings show the importance of cargo geometry for specifying cellular entry modes , even when the receptor recognition properties of a ligand are maintained . Citation : Cureton DK , Massol RH , Whelan SPJ , Kirchhausen T ( 2010 ) The Length of Vesicular Stomatitis Virus Particles Dictates a Need for Actin Assembly during Clathrin - Dependent Endocytosis . PLoS Pathog 6 ( 9 ) : e1001127 . doi : 10 . 1371 / journal . ppat . 1001127 Editor : John A . T . Young , The Salk Institute for Biological Studies , United States of America Received June 3 , 2010 ; Accepted September 1 , 2010 ; Published September 30 , 2010 Copyright : (cid:2) 2010 Cureton et al . This is an open - access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use , distribution , and reproduction in any medium , provided the original author and source are credited . Funding : This work was supported by NIH ( http : / / www . nih . gov / ) grants U54 AI057159 ( New England Regional Center of Excellence in Biodefense and Emerging Infectious Disease ( NERCE BEID ) ) to TK and AI081842 to SPJW . The funders had no role in study design , data collection and analysis , decision to publish , or preparation of the manuscript . Competing Interests : The authors have declared that no competing interests exist . * E - mail : kirchhausen @ crystal . harvard . edu ( TK ) ; swhelan @ hms . harvard . edu ( SPJW ) Introduction Eukaryotic cells internalize constituents of the plasma mem - brane and extracellular cargos by entrapping them in membrane - bound carriers . The most prominent and well - characterized endocytic carriers are clathrin - coated vesicles ( reviewed in [ 1 \u2013 3 ] ) . Coated vesicles transport lipids , proteins , and other essential macromolecules from the cell surface to endosomal organelles . Extensive biochemical and cell biological research supports the following model for conventional coated vesicle formation in higher eukaryotes . The AP - 2 clathrin adaptor complex recruits clathrin to the cytosolic leaflet of the plasma membrane and sequesters cargos at the endocytic site [ 4 , 5 ] . The continued assembly of clathrin into a lattice - like configuration helps deform the underlying membrane and ultimately creates an invagination , or \u2018pit\u2019 [ 2 ] . Recruitment of the GTPase , dynamin , then facilitates scission of the coated pit from the plasma membrane [ 6 ] , and clathrin is rapidly removed from the cargo - loaded vesicle by the combined action of the heat shock cognate protein 70 ( Hsc70 ) and its co - chaperone auxilin [ 7 , 8 ] . The entire process is typically complete within 30 \u2013 60 s [ 9 , 10 ] . Coated pits incorporate and internalize soluble cargos of various sizes , such as transferrin ( 5 nm ) [ 9 , 11 ] and low density lipoproteins ( 25 nm ) [ 9 , 12 ] . Many viruses and intracellular bacteria are also internalized by the clathrin machinery [ 9 , 13 \u2013 16 ] . We previously evaluated how cells internalize the 70 6 200 nm bullet - shaped vesicular stomatitis virus ( VSV ) . We found that VSV internalization occurs through elongated , partially clathrin - coated structures that have longer lifetimes ( , 2 min . ) than typical endocytic clathrin - coated vesicles and require local actin polymerization for uptake [ 15 ] . During VSV internalization , the clathrin coat first assembles as a partially closed dome at one end of the virion [ 15 , 17 ] , and growth of the coat stalls when it encounters the long particle axis . Actin assembly then drives one or more late stage ( s ) of the internalization process , as recruitment of the actin machinery peaks during completion of clathrin assembly , and pharmacological inhibition of actin polymerization blocks VSV internalization without interfering with clathrin coat assembly [ 15 ] . Relatively small , spherical viruses like dengue virus ( 50 nm ) [ 13 ] and some influenza A viruses ( X - 31 strain , , 120 nm ) [ 14 , 18 ] also enter using a clathrin - dependent route , but it is unclear whether actin function is required for their uptake . Our observations with VSV led us to PLoS Pathogens | www . plospathogens . org 1 September 2010 | Volume 6 | Issue 9 | e1001127 hypothesize that the physical dimensions of the virion block the ongoing polymerization of clathrin during its uptake , and that the stalled structure recruits regulators of actin assembly whose activity is required to complete the internalization process . Defective interfering ( DI ) particles arise spontaneously during virus replication . Such particles depend upon coinfecting helper virus to support their replication but contain all the essential cis - acting regulatory elements for genome replication and assembly . One such well - characterized DI particle of VSV is termed DI - T , which lacks 82 % of the viral genome [ 19 , 20 ] . Since the length of a VSV particle is dictated by the genome size [ 21 ] , DI - T particles are 75 nm long and appear as truncated bullets by electron microscopy [ 22 ] . DI - T particles contain normal proportions of the viral structural proteins [ 23 ] , including the viral surface glycopro - tein ( G ) , which mediates VSV attachment and entry into host cells . Here we took advantage of significant differences in the physical dimensions of VSV and DI - T to investigate how the geometry of a viral cargo influences the actin - dependency of clathrin internal - ization . Using live cell fluorescence microscopic imaging , we compared the uptake mechanisms of VSV and DI - T at the single particle level . We report that in contrast to the clathrin - and actin - dependent uptake of VSV , the shorter DI - T particles enter cells through fully coated clathrin carriers that do not require actin dynamics for vesicle budding . These observations highlight the plasticity of the clathrin endocytic system , where clathrin coats serve as a scaffold to direct actin assembly when the clathrin machinery alone is not sufficient to mediate internalization . Results Biological properties of DI - T particles To generate a clonal population of VSV DI - T particles , we recovered DI - T from cDNA ( Figure 1A ) [ 24 ] and amplified the particles by co - infection of cells with VSV [ 25 ] . We separated DI - T particles from VSV by rate zonal centrifugation in a sucrose density gradient . Electron microscopic analysis ( Figure 1B , C ) confirmed that VSV virions measure 70 + / 2 8 nm by 204 + / 2 14 nm ( n = 114 ) [ 21 ] , while the shorter DI - T particles have a length of 76 + / 2 8 nm ( n = 81 ) [ 22 ] . DI - T particles , like VSV , are covered with spike - like projections that correspond to homotrimers of G protein ( Figure 1B ) , and SDS - PAGE analysis of purified particles confirmed that VSV and DI - T particles contain similar ratios of G protein to core virion components ( Figure 1D ) [ 23 ] . The purified stocks of DI - T lacked full - length virions ( Figure 1B ) , with only a single VSV virion observed amongst more than 3 , 000 DI - T particles . Limited dilutions of the purified DI - T stock contained , 1x10 6 plaque forming units of virus per microgram of total viral protein , or 10 - 5 times fewer infectious particles than for an equivalent protein quantity of VSV particles ( not shown ) . Thus , we have successfully purified relatively homogeneous populations of VSV and DI - T particles , which differ only in their physical dimensions . DI - T particles enter cells by clathrin - dependent endocytosis To visualize VSV and DI - T by fluorescence microscopy , we covalently labeled the G proteins with spectrally separable fluorescent dye molecules ( Alexa Fluor 568 and 647 , respectively ) using conditions that do not reduce viral infectivity [ 15 ] . Spinning disk confocal images of labeled particles adsorbed onto glass coverslips showed diffraction - limited objects with single - peaked distributions of fluorescence intensity values ( Figure 1E ) , indicating that the DI - T and VSV populations primarily consist of individual particles [ 15 ] . We tracked the entry of DI - T particles into BSC1 cells stably expressing an eGFP - tagged s 2 subunit of the AP - 2 adaptor complex ( s 2 - eGFP ) , which incorporates into all clathrin - coated structures that form on the plasma membrane [ 9 , 15 , 26 ] . Single DI - T particles readily attached to cells and progressed through the following set of defined events ( see Figures 2A , B ; Video S1 for examples ) : ( 1 ) membrane - bound DI - T particles diffused slowly ( D = 5 6 10 2 11 \u2013 5 6 10 2 12 cm 2 s 2 1 ) and with the random directionality characteristic of Brownian motion ; ( 2 ) shortly after DI - T attachment , a dim spot of AP - 2 signal arose and remained colocalized with the particle , signifying incorporation of DI - T into an assembling clathrin - coated pit ; ( 3 ) the AP - 2 signal steadily increased over time until it peaked as coat assembly completed , and the DI - T particle then underwent an abrupt movement into the cell , after which the AP - 2 signal disappeared due to clathrin uncoating . This sequence of events is identical to what we previously observed for VSV entering cells by clathrin - dependent endocytosis [ 15 ] . Moreover , the efficiency of DI - T uptake via the clathrin pathway is similar to that of the full - length VSV particles , as 89 % ( 55 / 62 ) of DI - T particles that attached to 3 individual cells during imaging entered by clathrin - dependent endocytosis . These data show that DI - T efficiently enters cells through the clathrin pathway and validate the use of DI - T and VSV as comparative endocytic cargos . Clathrin structures capture VSV and DI - T particles with similar kinetics To directly compare how DI - T and VSV particles engage the clathrin machinery , we simultaneously inoculated BSC1 cells with the two spectrally distinct particle forms and then analyzed their mode of incorporation into AP - 2 containing clathrin structures on a single cell basis ( Figure 2C ) . For each complete virus uptake event , we quantified the kinetics of particle capture by measuring the elapsed time between virion attachment and the appearance of an AP - 2 signal that colocalized with the bound particle . The capture time for DI - T particles was 110 + / 2 80 s ( n = 121 ) , which is statistically indistinguishable ( Student\u2019s t - test , p = 0 . 2 ) to that measured for VSV ( 130 + / 2 125 s ( n = 87 ) ) and agrees well with our prior measurements ( Figure 2D ) [ 15 ] . The AP - 2 structures that captured DI - T or VSV initiated within a , 250 nm zone ( the resolution limit of the optical system ) of the attached particle . We therefore conclude that DI - T and VSV particles engage the clathrin system in an indistinguishable manner , which likely reflects a shared mechanism triggered by the same viral glycoprotein - receptor interactions . Author Summary We present a detailed comparison between the clathrin - dependent entry mechanisms of a parental virus ( VSV ) and its smaller defective interfering particle ( DI - T ) . We used the difference in virion length to probe why actin assembly is required for the uptake of full - length VSV particles by nonpolarized mammalian cells . By imaging the entry of single particles in an unbiased manner , we resolved differences in the maturation kinetics , clathrin content , and actin dependency of clathrin endocytic structures internalizing VSV or DI - T virions . Our principal finding is that , unlike VSV uptake , DI - T internalization does not induce or require robust actin polymerization . We have also established , for the first time , that the geometry of an endocytic cargo can alter the mechanism of clathrin uptake . We propose that VSV - containing clathrin struc - tures display characteristics of \u2018frustrated\u2019 endocytic intermediates that cells resolve by using the force of actin assembly to deform the plasma membrane into a complete endocytic vesicle . Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 2 September 2010 | Volume 6 | Issue 9 | e1001127 Cells internalize DI - T particles using conventional clathrin - coated vesicles To investigate the characteristics of the clathrin coat responsible for DI - T internalization , we imaged the uptake of both particle forms by the same BSC1 cell . We found that DI - T particles are internalized through AP - 2 containing structures significantly faster than full - length virions ( Figure 3A , B ; Video S2 ) . Quantitative analysis of data compiled from 4 cells showed that pits incorporating DI - T ( n = 36 ) form in 43 + / 2 14 s , which is similar to the assembly kinetics of pits that lack virus particles ( n = 212 , 35 + / 2 10 s ) ( Figure 3C , Table 1 ) . As expected , AP - 2 structures that capture VSV ( n = 29 ) require longer ( 75 + / 2 22 s ) to complete ( Figure 3C , Table 1 ) . A similar analysis conducted in cells transiently expressing eGFP - tagged clathrin light chain A1 ( eGFP - LCa ) yielded analogous results ( Figure 3D , E , Table 1 , Videos S3 , S4 ) . The interaction of full - length VSV with a cell had no impact on the uptake kinetics of DI - T into the same cell ( Table 1 ) . The different kinetics of DI - T versus VSV internalization suggests that DI - T enters cells through conventional fully coated clathrin structures and not the partially coated vesicles responsible for VSV uptake . To further investigate this possibility , we measured the maximum fluorescent signal of eGFP - tagged AP - 2 or clathrin molecules , as this peak signal is known to be proportional to the overall size of a clathrin coat [ 9 , 15 , 26 ] . We compared this value for structures that contained DI - T with those that contained VSV or lacked either particle . Similar quantities of coat components were present in structures associated with DI - T to those lacking viral particles ( Figure 3C , E , Table 1 ) . As expected , VSV - containing structures accumulate more AP - 2 and LCa molecules than structures lacking virus ( Figure 3C , E , Table 1 ) . Taken together , the above experiments suggest that DI - T enters cells through pits that acquire a full clathrin coat . Consistent with this , electron micrographs of DI - T particles captured during cell entry show particles present in circular pits entirely surrounded by a clathrin coat ( Figure 4A ) . This is in marked contrast to the partial clathrin coat found at one end of the endocytic carriers that internalize VSV ( Figure 4B ) [ 15 ] . Clathrin structures containing VSV recruit more cortactin that pits that internalize DI - T During the final phase of coat assembly , endocytic clathrin structures associated with VSV show a strong recruitment of cortactin , an F - actin and dynamin binding protein that activates Arp2 / 3 - mediated assembly of branched actin filaments [ 15 , 27 , 28 ] . To determine whether DI - T uptake is associated with an acute recruitment of cortactin , we monitored internalization Figure 1 . Biological properties of DI - T particles . ( A ) Structure of VSV and DI - T genomic RNAs . The single - stranded negative - sense RNA genomes are shown in a 3 9 - 5 9 orientation . The five viral genes are colored as follows : red nucleocapsid ( N ) protein gene ; orange , phosphoprotein ( P ) gene ; yellow , matrix ( M ) protein gene ; green , glycoprotein ( G ) gene ; blue , large ( L ) polymerase protein gene . The noncoding genomic terminal leader ( Le ) and trailer ( Tr ) regions , which serve as promoters for RNA synthesis and genomic RNA encapsidation , are shown in gray . The DI - T genome comprises 2 , 208 nts . The 5 9 terminal 2 , 163 nts derive from the 5 9 terminus of the parental VSV genome ( 11 , 161 nts ) , and the 3 9 terminus contains a 45 nt inverted complement of the wild - type Tr ( TrC ) . ( B ) Electron micrographs of purified VSV and DI - T particles negatively stained with PTA . Middle panel , inset shows an expanded view of virions from the boxed region to facilitate visualization of the viral glycoprotein spikes . Inset scale bar , 100 nm . ( C ) Virion geometry . The dimensions of individual DI - T ( blue ) and VSV ( red ) particles measured from electron micrographs of negatively stained particles . Each open circle represents the measurement for a single particle . Horizontal red lines denote the mean value of each population , and the numerical means ( + / 2 SD ) are provided above each plot . ( D ) Protein composition of purified VSV and DI - T particles . Viral proteins ( L , G , P , N , and M ) were separated by SDS - PAGE and visualized with Coomassie blue staining . The ratio of N protein to G protein was quantified using ImageJ and is displayed below the gel as a comparative measure of average virion surface glycoprotein density in each particle population . ( E ) Fluorescence intensity of virus particles labeled with Alexa Fluor dye molecules . Purified DI - T or VSV particles were labeled with Alexa Fluor 647 or 568 and imaged on separate glass coverslips using a spinning disk confocal microscope . The fluorescence intensity of individual spots in a single field of view was quantified , and the distribution of intensity values ( in arbitrary units , a . u . ) for DI - T ( blue ) and VSV ( red ) particles is shown . doi : 10 . 1371 / journal . ppat . 1001127 . g001 Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 3 September 2010 | Volume 6 | Issue 9 | e1001127 into BSC1 cells transiently co - expressing monomeric Cherry - LCa ( mCherry - LCa ) and low levels of cortactin - eGFP . As previously shown [ 15 , 26 , 29 ] , conventional clathrin - coated pits exhibit minimal cortactin recruitment that typically peaks just before completion of clathrin assembly ( Figure 5A , Video S5 ) . Cortactin recruitment is similarly sparse during the uptake of DI - T particles ( Figure 5B , Video S6 ) . In marked contrast , and as expected [ 15 ] , large bursts of cortactin accompany the internalization of VSV ( Figure 5C , Video S7 ) . Quantitative analysis revealed that the peak fluorescence intensity of cortactin detected in the late phase of VSV uptake averaged 3 - fold higher than the signal associated with pits containing DI - T or pits that did not capture a virus particle ( Figure 5D , Table 1 ) . These data suggest that while formation of short - branched actin filaments is required during the late stages of clathrin - mediated VSV entry , this need is obviated during clathrin - dependent DI - T entry . Actin polymerization is not required for DI - T internalization To directly test whether actin assembly is required for DI - T entry , we treated BSC1 cells with latrunculin B ( latB ) , a chemical inhibitor of actin filament assembly [ 30 ] , and tracked the endocytic fate of DI - T and VSV in the same cells . Treatment of cells with 6 . 3 m M latB did not change the efficiency of DI - T entry ( Figure 6A \u2013 C , Video S8 ) , but it reduced the internalization of VSV by . 75 % ( Figure 6A \u2013 C ) . As expected [ 15 ] , latB treatment did not affect the capture efficiency of either particle type by clathrin ( Figure 6C ) . The lifetimes and AP - 2 content of pits lacking particles or containing DI - T was similarly unaffected by latB ( Figure 6D ) . We conclude that the shorter DI - T particles bypass the actin requirement displayed by the larger VSV for efficient clathrin - based uptake . Figure 2 . Clathrin structures capture VSV and DI - T particles with similar kinetics . ( A ) Schematic of clathrin - dependent virus internalization . 1 . A particle ( blue ) attaches to receptor moieties ( orange ) on the cell surface ( black horizontal line ) , and the virus - receptor complex diffuses in the plane of the membrane . 2 . The virus particle is captured by the clathrin endocytic machinery ( AP - 2 , green ; clathrin , red ) after diffusion into an existing clathrin structure ( e . g . Dengue virus ) or entrapment within a clathrin structure that initiates in close proximity to the virion ( e . g . VSV and influenza A virus ) . 3 . Clathrin assembly completes , and the virus - containing pit is severed from the cell surface in a dynamin - dependent process . Internalization of VSV also requires local actin assembly . Clathrin is rapidly removed from the nascent vesicle , and the vesicle is actively transported further into cell . ( B ) Example of a complete DI - T internalization event . A single DI - T particle ( red ) attaches to a BSC1 cell expressing s 2 - eGFP ( green ) and diffuses on the cell surface . A dim spot of AP - 2 appears beneath the virion , signifying capture of the particle . The AP - 2 fluorescence intensity increases as the clathrin coat assembles , and the virus disappears into the cell shortly after the AP - 2 signal reaches a maximum ( Video S1 ) . Numbered stages correspond to the events described in A . The path of particle motion is depicted as a linear , color - coded trace that progresses with time from blue to red . ( C ) VSV and DI - T particle capture by clathrin structures in the same cell . BSC1 cells stably expressing s 2 - eGFP ( green ) were inoculated with Alexa Fluor 647 - labeled DI - T ( blue , blue arrowheads ) and Alexa Fluor 568 - labeled VSV ( red , red arrowheads ) . Time - lapse images were acquired at 4 s intervals using a spinning disk confocal microscope . Left , snapshot of a cell depicting coated pits lacking ( white arrowheads ) or containing ( blue / red arrowheads ) virus particles . Right , expanded split - channel views of the region within the dashed box at left . ( D ) Kinetics of virus capture . BSC1 cells stably expressing s 2 - eGFP ( 7 cells ) or eGFP - LCa ( 12 cells ) were inoculated with VSV and DI - T particles . Images were acquired at 3 - 4 s intervals as in C , and the time interval between virus attachment and detection of AP - 2 or LCa beneath a virion was quantified for productive internaliza - tion events . The distribution of capture times is shown for DI - T ( blue ) and VSV ( red ) particles , and the mean time to capture ( + / 2 SD ) for each particle population is provided at right . The kinetics of VSV and DI - T capture are not significantly different ( Student\u2019s t - test p value = 0 . 2 ) . doi : 10 . 1371 / journal . ppat . 1001127 . g002 Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 4 September 2010 | Volume 6 | Issue 9 | e1001127 Discussion The major conclusion of our study is that the physical properties of a virus particle dictate the need for engagement of the actin system during its clathrin - dependent uptake into cells . We formulate this conclusion based on tracking the clathrin - depen - dent internalization of VSV and its shorter DI - T particle into live cells . Internalization of VSV is accompanied by the recruitment of Figure 3 . Cells internalize DI - T particles using conventional clathrin - coated vesicles . ( A ) Internalization of VSV and DI - T by the same cell . BSC1 cells stably expressing s 2 - eGFP ( green ) were inoculated with VSV ( red ) and DI - T ( blue ) particles , and confocal images were captured at 4 s intervals . The images show the sequential internalization of single VSV ( red , red arrowheads ) and DI - T ( blue , blue arrowheads ) particles , followed by the formation of a canonical coated vesicle ( white arrowhead ) within a 3 . 5 6 3 . 5 m m 2 area of the plasma membrane ( Video S2 ) . The first acquired frame of the time - lapse series is designated + 0 s , and the capture time of the subsequent images is shown . ( B ) AP - 2 recruitment during the uptake events shown in A . Left , image quadrants depicting AP - 2 accumulation over time ( left panels ) and at the time of maximum AP - 2 signal during each event ( right panels ) . Upper panels in each quadrant show the AP - 2 channel alone , and the lower panels show overlays of the virus and AP - 2 channels . The highlighted pit from A . is indicated with a white arrowhead , and virus particles are colored as in A . Right , fluorescence intensity ( in arbitrary units , a . u . ) of AP - 2 over time for the events shown at left and in A . The time of AP - 2 detection above the local background was set to t = 0 s for each event . ( C ) Kinetics and AP - 2 content of endocytic structures . The plots show the relative lifetime ( left ) and maximum fluorescence intensity ( right ) of AP - 2 during the uptake of coated pits lacking virus ( pits , black ) or structures that internalized DI - T ( blue ) or VSV ( red ) particles ( from 4 cells ) . Values are expressed as percentages to facilitate comparison of viral and nonviral uptake events across multiple cells . Approximately 50 pits lacking virus were analyzed in each cell , and the mean of the measured values was calculated for each parameter . The values for each nonviral and viral uptake event were divided by the mean for pits lacking virus in the same cell , and the resulting values were multiplied by 100 . Each open circle represents a single uptake event , and horizontal red lines demark the mean of the compiled population . The number of events is provided above each plot . Numerical values and statistical analyses are provided in Table 1 . ( D ) Clathrin recruitment during virus entry . Left , kymograph views of internalization events from a single BSC1 cell transiently expressing eGFP - LCa ( Videos S3 , S4 ) . Images were captured as in A . and displayed as described in B . Right , fluorescence intensity of eGFP - LCa over time for the events shown at left . ( E ) Kinetics and clathrin content of endocytic structures . The plots show the relative lifetime ( left ) and maximum fluorescence intensity ( right ) of clathrin during the uptake of coated pits lacking virus ( pits , black ) or structures that internalized DI - T ( blue ) or VSV ( red ) particles ( from 3 cells ) . Data were calculated and plotted as described in C . Numerical values and statistical analyses are provided in Table 1 . doi : 10 . 1371 / journal . ppat . 1001127 . g003 Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 5 September 2010 | Volume 6 | Issue 9 | e1001127 cortactin at a late step of the endocytic process , and chemical inhibition of actin polymerization blocks virus uptake ( this study and [ 15 ] ) . By contrast , internalization of the shorter DI - T particles is insensitive to chemical inhibition of actin polymerization and is not accompanied by the same spike of cortactin recruitment observed for the full - length VSV particles . VSV and DI - T particles differ in their length and not the density of the glycoprotein spike that dictates their entry . We suggest that the shape of the full - length VSV particles presents a physical barrier that frustrates completion of the clathrin - coated pit . The stalled clathrin structure then recruits the actin machinery required to finalize internalization . The shorter DI - T particles no longer provide such a physical barrier during engulfment , which results in their actin - independent internalization through coated vesicles that acquire a full complement of clathrin . The importance of particle length for the clathrin - dependent uptake of VSV Receptor - dependent signaling events lead to actin filament assembly during the clathrin - independent uptake of other viruses , including poliovirus and coxsackie B virus [ 31 , 32 ] . Here we show that the initial interactions of DI - T with the cell are indistinguish - able to those of VSV , as both particle types diffuse slowly on the cell surface and are captured by coated pits with similar kinetics ( Figure 2 ) . Such similar behavior suggests that DI - T and VSV engage an as yet unknown viral receptor in an analogous manner . Since initial receptor interactions appear indistinguishable , it seems remote that VSV G - receptor interactions induce a signaling cascade that leads to actin polymerization at sites of VSV but not DI - T uptake . We therefore conclude that the glycoprotein is not the primary trigger of actin assembly during clathrin - dependent VSV internalization . We previously showed that endocytic structures containing VSV do not acquire a full clathrin coat [ 15 ] , which agreed with earlier electron micrographs depicting virus particles in tubular invagi - nations with clathrin at the cytosolic tip [ 17 ] . The morphology of those structures suggests that cells initially capture one tip of the virus particle , and clathrin assembly stalls when the constricting coat encounters the long axis of the virion ( Figure 7 ) . Here we found that DI - T particles do not alter the process of clathrin - coated vesicle formation ( Figures 3 , 4 ) . This finding implies that clathrin can fully enclose cargos displaying VSV G provided that the particle shape does not physically prohibit clathrin assembly or closure of the plasma membrane ( Figure 7 ) . Consequently , the physical properties of the VSV particle captured by a clathrin - coated pit are what dictate the altered mode of actin - dependent uptake . We therefore propose the model ( Figure 7 ) that it is the incomplete clathrin structure formed during VSV uptake that elicits the actin - based response to rescue the endocytic process and leads to the successful engulfment of the trapped virus particle . Actin - dependent clathrin - mediated endocytosis The importance of actin function during clathrin - mediated endocytosis varies . In yeast cells , actin polymerization drives invagination and endocytosis of long - lived ( . 2 min . ) clathrin assemblies through tubular intermediates with clathrin at the cytosolic tip [ 33 \u2013 35 ] . Mammalian cells also form large ( 0 . 2 \u2013 1 m m ) , relatively flat arrays of clathrin , or \u2018plaques , \u2019 on the adherent cell surface [ 26 , 36 ] . These structures also exhibit long lifetimes ( 2 - 15 + min . ) and require local actin assembly for internalization [ 26 ] . Coated pit internalization from the apical ( but not the basolateral ) surface of polarized mammalian cells [ 37 , 38 ] and lamprey neuronal synapses [ 39 , 40 ] is also actin - dependent . By contrast , actin polymerization is only required for the uptake of clathrin Table 1 . Summary of kinetic and fluorescence intensity data . Particle ( s ) Fluorescentprotein ( s ) Treatment # Expts / # Cells Eventsanalyzed # Events Lifetime % Lifetime % Max fluorescence DI - T s 2 1 / 3 DI - T 62 47 + / 2 13 s ND ND DI - T + VSV s 2 1 / 4 pits 212 35 + / 2 10 s 100 + / 2 24 100 + / 2 23 DI - T 36 43 + / 2 14 s 128 + / 2 37 * * 119 + / 2 38 * * VSV 29 75 + / 2 22 s 224 + / 2 68 * * 148 + / 2 48 * * LCa 1 / 3 pits 209 44 + / 2 19 s 100 + / 2 34 100 + / 2 26 DI - T 48 44 + / 2 14 s 100 + / 2 30 99 + / 2 26 VSV 34 89 + / 2 36 s 222 + / 2 104 * * 130 + / 2 38 * * DI - T + VSV s 2 + latB 2 / 4 pits 193 47 + / 2 14 s 100 + / 2 27 100 + / 2 27 + latB DI - T 46 54 + / 2 21 s 118 + / 2 46 * 133 + / 2 42 * * + latB VSV entry 20 141 + / 2 104 s 298 + / 2 198 * * 207 + / 2 59 * * + latB trapped VSV 31 615 + / 2 216 s 1283 + / 2 430 * * 192 + / 2 73 * * DI - T LCa + cortactin 1 / 3 pits 155 51 + / 2 19 s 100 + / 2 34 100 + / 2 50 DI - T 30 65 + / 2 28 s 129 + / 2 63 * 130 + / 2 63 * 1 / 3 pits 220 52 + / 2 24 s 100 + / 2 36 100 + / 2 66 VSV LCa + cortactin VSV 21 166 + / 2 63 s 335 + / 2 105 * * 300 + / 2 164 * * Kinetic and fluorescence values are provided as the mean + / 2 SD for all events in a given context . The number of experiments and cells analyzed is indicated . The absolute lifetimes are expressed in seconds ( s ) . The % lifetime and % maximum fluorescence intensity values were calculated as described in the legend of Figure 3C . The % maximum fluorescence intensity of cortactin is provided for events analyzed in cells co - expressing mCherry - LCa and cortactin - eGFP . A two - tailed Student\u2019s t - test was used to determine whether data from 2 categories differ in a statistically significantly manner . ( * ) denotes a statistical difference with a p - value , 0 . 005 and . 0 . 00005 when comparing data in a given category to data for pits lacking virus in the same category . ( * * ) similarly denotes a p - value , 0 . 00005 . All values for VSV are significantly different ( p , 0 . 005 for % max AP - 2 fluorescence ; p , 0 . 00005 for all other categories ) from those measured for DI - T particles in the same context . ND , not determined . doi : 10 . 1371 / journal . ppat . 1001127 . t001 Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 6 September 2010 | Volume 6 | Issue 9 | e1001127 plaques and not conventional coated pits in several types of nonpolarized mammalian cells [ 26 , 41 ] . Thus , while actin dynamics play an evolutionarily conserved role in clathrin - dependent endocytosis , mammalian cells regulate the interplay between the clathrin and actin systems . Our analyses of VSV and coated plaque internalization reveal two properties that correlate with their actin - dependent uptake mechanisms . First , plaques and VSV - containing structures remain on the plasma membrane for . 2 - fold longer than standard coated pits [ 15 , 26 ] ( Figure 3C , E ) . The prolonged presence of a clathrin lattice might promote interactions between clathrin - associated proteins and regulators of actin polymerization . Second , clathrin plaques and VSV - containing structures physically differ from standard coated pits . Plaques fail to constrict their outer boundaries during the final phase of clathrin assembly [ 26 ] , and pits containing VSV lack a complete clathrin coat [ 15 , 17 ] . Such unusual structural features might attract proteins that associate with exposed lipids , such as dynamin . Indeed , significantly more dynamin molecules accumulate during the final stages of VSV uptake [ 15 ] . This localized increase in dynamin may enhance the recruitment of dynamin - interacting proteins with the capacity to bind lipids and activate the Arp2 / 3 complex through N - WASP , including endophilin , syndapin , and SNX9 [ 42 \u2013 46 ] . Although these proteins may link the clathrin and actin systems and facilitate localized membrane remodeling during endocytosis ( reviewed in [ 47 ] ) , further studies are needed to determine whether they play a role in clathrin - dependent VSV endocytosis . Comparative studies of VSV and DI - T uptake may provide a useful tool to dissect the mechanisms that regulate actin assembly during clathrin - mediated endocytosis . Implications for the internalization mechanisms of other viruses The dimensions of the actin - dependent VSV particle and the actin - independent DI - T particle fall within the range of shapes present in many pleomorphic viruses . Our work suggests that this may lead to important distinctions in their mode of uptake . For example , some influenza A virus strains , such as X - 31 , produce spherical particles that measure 80 \u2013 120 nm in diameter [ 18 ] . By contrast , the Udorn strain forms filamentous particles that measure 80 \u2013 100 nm wide and up to 30 microns in length [ 48 , 49 ] . Although influenza A virus can enter cells by clathrin - dependent and - independent mechanisms [ 14 ] , the impact of particle shape on the entry pathway remains unknown . It seems likely that remodeling of the cortical actin cytoskeleton will be important for uptake of filamentous influenza particles , and clathrin may facilitate local membrane deformation during the endocytic process . The Arenaviridae generate roughly spherical particles that range in diameter from 40 \u2013 300 nm [ 50 , 51 ] , and it is known that clathrin function is important for efficient infection of cells by some New World arenaviruses [ 52 ] . It will now be of interest to determine whether spherical particles of different diameter employ altered modes of clathrin - based endocytosis . Pseudotyping is often used to study the entry pathway of highly pathogenic viruses , including the long filamentous filoviruses , Ebola and Marburg , as well as several arenaviruses . Such pseudotypes are frequently based on VSV or retroviral virions in which the endogenous entry proteins have been replaced with the surface glycoproteins of the pathogenic virus [ 53 \u2013 56 ] . Although viral pseudotypes are useful for studying the entry process , VSV and the spherical virions of retroviruses ( , 100 nm in diameter ) do not accurately recapitulate the sizes or shapes of the pleomorphic viruses . Our studies of VSV and DI - T clearly show that virion geometry can fundamentally alter aspects of the viral internaliza - tion process . Therefore , it is critically important to study viral endocytosis using pseudotyped or virus - like particles that closely approximate the physical properties of a virus in question . Materials and Methods Cells and viruses African green monkey kidney BS - C - 1 cells ( herein BSC1 , American Type Culture Collection ( ATCC ) CCL - 26 ; Manassas , VA ) and Vero cells ( ATCC ) were maintained at 37 u C and 5 % CO 2 in Dulbecco\u2019s Modified Eagle Medium ( DMEM , Invitrogen Corporation ; Carlsbad , CA ) supplemented with 10 % fetal bovine serum ( Tissue Culture Biologicals ; Tulare , CA ) . BSC1 cells stably expressing rat s 2 adaptin - eGFP ( s 2 - eGFP ) [ 9 ] were maintained as above in the presence of 0 . 4 mg mL 2 1 geneticin ( G418 , Invitrogen ) . Figure 4 . Electron microscopic images of DI - T and VSV particles in clathrin - coated pits . ( A ) Electron micrographs depicting DI - T particles at early ( left ) or late ( right ) stages of clathrin - dependent endocytosis . BSC1 cells were incubated with , 1000 particles per cell for 10 min . at 37 u C . Cells were then fixed , and samples were processed for ultra thin sectioning and viewed as described in the materials and methods . ( B ) Electron micrographs of VSV particles at sequential ( left to right , top to bottom ) stages of clathrin - dependent endocytosis . Vero cells were inoculated with VSV at an MOI of 5 , and samples were prepared for analysis at 6 h post - infection . doi : 10 . 1371 / journal . ppat . 1001127 . g004 Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 7 September 2010 | Volume 6 | Issue 9 | e1001127 Recombinant VSV ( rVSV ) [ 25 ] was amplified and purified as before [ 15 ] . Defective interfering T ( DI - T ) particles of VSV were recovered from a cDNA clone of the DI - T genome [ 24 ] . The DI - T particles were amplified by co - infecting baby hamster kidney cells ( BHK - 21 , ATCC C - 13 ) with rVSV ( multiplicity of infection ( MOI ) 50 ) . A subsequent passage was performed by inoculating cells with filtered , undiluted supernatant from the primary amplification and rVSV ( MOI of 50 ) . Viruses were concentrated by centrifugation at 44 , 000 6 g , and the virus pellet was resuspended in NTE ( 10 mM Tris pH 7 . 4 , 100 mM NaCl , 1 mM EDTA ) . The two particle forms were separated on a 15 \u2013 45 % sucrose gradient prepared in NTE by centrifugation at 77 , 000 6 g for 5 h . The DI - T particles were extracted from the upper virus band , concentrated as before , and resuspended to 1 mg mL 2 1 of total protein in PBS . Dye conjugation to virus particles Purified DI - T and VSV particles were labeled with Alexa Fluor dye molecules ( Molecular Probes , Invitrogen ) as previously described [ 15 ] except that the final dye concentration in the labeling reaction was reduced to 25 m g ml 2 1 . Plaque assays of virus preps before and after labeling showed that dye conjugation did not affect the infectivity of VSV particles or the capacity of DI - T virions to inhibit plaque formation by VSV . The surface density of G protein on VSV or DI - T particles was estimated by measuring the ratio of G protein to N protein in each particle population . To separate and visualize the viral proteins , purified virions were subjected to SDS - PAGE using 10 % polyacrylamide and 0 . 13 % bis - acrylamide and stained with Coomassie blue . The relative amounts of N or G protein were established using ImageJ ( U . S . National Institutes of Health , Bethesda , Maryland ; http : / / rsb . info . nih . gov / ij / ) . Nucleic acid transfection BSC1 cells were seeded into 6 - well plates at , 60 , 000 cells per well 16 \u2013 20 h prior to transfection . Plasmid DNA was introduced into the cells using FuGENE 6 ( Roche Diagnostics ; Indianapolis , Figure 5 . Clathrin structures containing VSV recruit more cortactin that pits that internalize DI - T . ( A ) Cortactin recruitment during coated pit formation . Left , snapshot showing the surface of a BSC1 cell transiently expressing cortactin - eGFP ( green ) and mCherry - LCa ( red ) at 18 h post - transfection . Time - lapse images were acquired at 3 s intervals , and frame 83 is shown . Middle , split channel kymographs of coated pit formation in the cell at left . White arrowheads highlight pits in which cortactin recruitment is clearly visible above the local background . Right , example plot of cortactin and clathrin fluorescence intensity over time during the formation of a single clathrin - coated pit in the cell shown at left ( Video S5 ) . ( B and C ) Examples of cortactin recruitment during DI - T ( B ) and VSV ( C ) uptake . BSC1 cells transiently expressing mCherry - LCa ( red ) and cortactin - eGFP ( green ) were inoculated with Alexa Fluor 647 - labeled DI - T or VSV , and images were acquired as in A . Left , split - channel kymograph views of protein and virion ( blue ) fluorescence intensity over time ( Videos S6 , S7 ) . Images in the right - hand panels show a snapshot of the maximal cortactin or clathrin fluorescence , and white arrowheads highlight the peak cortactin signal . Right , plots of the cortactin and clathrin fluorescence intensity over time for each internalization event . ( D ) Relative peak fluorescence intensity of cortactin in cells co - expressing mCherry - LCa and cortactin - eGFP . At 18 h post - transfection , samples were separately inoculated with DI - T or VSV particles , and images were acquired as in A . For each cell that was imaged , the maximum cortactin fluorescence associated with , 50 pits lacking virus particles ( pits , black ) and all pits that internalized a DI - T ( left , blue , 3 cells ) or VSV ( right , red , 5 cells ) particle was measured . The data are plotted as described in the legend of Figure 3C , and the number of events is shown above each plot . Numerical values and statistical analyses are provided in Table 1 . doi : 10 . 1371 / journal . ppat . 1001127 . g005 Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 8 September 2010 | Volume 6 | Issue 9 | e1001127 IN ) according to the manufacturer\u2019s instructions . Prior to trans - fection , media on the cells was replaced with 1 ml OPTIMEM ( Invitrogen ) . After addition of the transfection mixture , cells were incubated at 37 u C for 5 h , and the existing media was supplemented with 2 ml DMEM containing 10 % FBS . To ensure optimal replacement of endogenous clathrin light chain molecules with rat eGFP - clathrin light chain A1 ( eGFP - LCa ) [ 9 ] , cells were transfected with 0 . 75 m g of plasmid DNA encoding eGFP - LCa . The cells were cultured for , 36 h and seeded onto glass coverslips , 14 h prior to image acquisition . Co - expression of mCherry - LCa ( constructed as described for tomato - LCa [ 8 ] ) and mouse cortactin - eGFP [ 57 , 58 ] was achieved by transfection of cells on glass coverslips with 1 m g of each plasmid , and the cells were imaged , 18 h later . Figure 6 . Actin polymerization is not required for DI - T internalization . ( A ) The endocytic fate of virus particles after inhibition of actin polymerization . BSC1 cells stably expressing s 2 - eGFP ( green ) were treated with 6 . 3 m M latB for 12 min . and inoculated with DI - T and VSV particles in the continued presence of latB . Time - lapse images of a single cell were acquired at 4 s intervals for 692 s , and an 8 . 8 6 5 . 0 m m 2 area of the cell surface is shown . The upper panels show the complete internalization of a DI - T particle ( blue , blue arrowheads ) , where + 0 s indicates the first frame of the time - lapse series . The lower panels show the subsequent capture but failed internalization of 2 VSV ( red , red arrowheads ) particles on the same area of cell membrane ( time scale continued from above ) ( Video S8 ) . ( B ) AP - 2 fluorescence intensity for the events shown in A . Note that the adaptor fluorescence associated with the DI - T particle ( blue ) and a conventional coated pit ( black ) that formed within the same membrane area peak and disappear normally , while the adaptor signal associated with the upper - most VSV particle ( red ) does not , signifying failed internalization . ( C ) Effect of latB on the efficiency of virus capture and internalization . BSC1 cells stably expressing s 2 - eGFP were treated and imaged as described in A . Left , the percentage of virus particles that were captured by a clathrin structure after attachment . Right , the percentage of captured virus particles that were successfully internalized within 300 s of capture ( see D . for details ) . Cumulative data are from 5 cells . ( D ) Effect of latB on the lifetime and peak fluorescence intensity of AP - 2 in clathrin structures . Data were acquired as described in A . and displayed as in the legend of Figure 3C . Left , relative lifetime of AP - 2 in structures that lack ( pits , black ) or capture a virus particle . Inset shows a rescaled distribution of the pit and DI - T internalization events . Right , maximum fluorescence intensity of AP - 2 in the events at left . Data are from 4 of the 5 cells analyzed in C , as thermal drift during imaging prevented accurate fluorescence intensity measurements in one cell . The number of events in each category is shown above the corresponding plots at right . DI - T ( blue ) data consists only of productive internalization events . VSV events are categorized as productive internalizations ( VSV entry , red ) or non - productive captures ( trapped VSV , red ) . A non - productive capture is defined as a stable colocalization between a spot of AP - 2 and a VSV particle that began at least 300 s before the last captured image and did not result in virus uptake before cessation of imaging . The 300 s cutoff was chosen because a majority ( 22 / 24 ) of productive internalizations occurred within 300 s of capture . Captures in which the final image frame was acquired before 300 s elapsed were excluded from the analysis , as the eventual endocytic fate of the particle cannot be predicted . doi : 10 . 1371 / journal . ppat . 1001127 . g006 Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 9 September 2010 | Volume 6 | Issue 9 | e1001127 Live cell imaging Cells on 25 mm coverslips ( No . 1 . 5 , Electron Microscopy Sciences ; Hatfield , PA ) were placed into a perfusion chamber and overlaid with a - MEM lacking phenol red ( Invitrogen ) and supplemented with 20 mM HEPES pH 7 . 4 and 2 % FBS . The chamber was placed in a heated sample holder ( 20 / 20 Technology Inc . ; Wilmington , NC ) mounted on the stage of a Mariana imaging system ( Intelligent Imaging Innovations , Denver , CO ) based on an Axiovert 200M inverted microscope ( Carl Zeiss , Inc . ; Thornwood , NY ) . The microscope stage and objective lenses were maintained at 37 u C within an environmental chamber , and the air above the cells was supplied with 5 % CO 2 . The position of the sample holder with respect to the objective lens was manipulated using a PZ - 2000 automated stage ( Applied Scientific Instrumen - tation ; Eugene , OR ) . Samples were illuminated using 40 \u2013 50 mW solid state lasers ( l = 488 , Coherent , Inc . ; Santa Clara , CA , l = 561 , Cobolt AB ; Solna , Sweden , l = 640 , 40 mW ; Coherent ) directed through an acousto - optic tunable filter ( AOTF ; Gooch and Housego L . L . C . ; Melbourne , FL ) . Laser illumination was imparted on the sample through a CSU - X1 spinning disk confocal unit ( Yokogawa Electric Corporation ; Tokyo , Japan ) and a 63X objective lens ( Plan - Apochromat , NA 1 . 4 , Carl Zeiss ) . Emission spectra were selected using single band width emission filters ( LF405 / 488 / 561 / 635 - A , Semrock ; Rochester , NY ) , and after transmission through a spherical aberration correction device ( Infinity Photo - Optical ; Boulder , CO ) , the emission photons were collected using a Cascade II : 512B back - illuminated EMCCD camera ( Roper Scientific , Photometrics ; Tuscon , AZ ) . Under this configuration , a single pixel corresponds to 0 . 07 6 0 . 07 m m 2 . Slidebook 4 . 2 . 13 ( Intelligent Imaging Innovations , Inc . ( III ) ; Denver , CO ) was used to command the hardware devices and visualize the acquired data . To image virus internalization , Alexa - labeled virus was centrifuged briefly to remove aggregates , and cells were inoculated with virus to result in attachment of , 50 particles per cell within 20 min . ( , 1 6 10 7 p . f . u . of VSV , , MOI 100 , 0 . 005 \u2013 0 . 05 particles per m m 2 of cell surface area ) . Time - lapse acquisitions were typically carried out for 8 \u2013 10 min . per cell , and images were captured at 3 \u2013 4 s intervals after sequentially illuminating the sample with the appropriate laser for 50 \u2013 100 ms per wavelength . To assess the effect of latrunculin B ( latB ) ( Sigma - Aldrich , Inc . ; St . Louis , MO ) on coated vesicle formation and virus entry , cells were treated with 6 . 3 m M of latB for , 10 min . at 37 u C prior to the addition of virus particles . Image analysis Image analysis was performed as previously described [ 15 ] with the following modifications . Slidebook 4 . 2 . 13 ( III ) was used to view , scale , and export images for publication . Movies were generated by exporting a series of continuous TIFF files from a Slidebook time - lapse acquisition and compiling the images into a single AVI file using ImageJ ( NIH ) . SigmaPlot 8 . 0 ( SYSTAT ; Point Richmond , CA ) was used to plot data . Microsoft Excel was used to determine whether data from 2 categories differ in a statistically significant manner using two - tailed Student\u2019s t - tests . An automated image analysis application ( IMAB ) [ 8 ] developed within MATLAB ( Mathworks ; Natick , MA ) was used to track the formation of clathrin - coated structures and the internalization of single virus particles . Images were processed for analysis as previously described , and established criteria were used to exclude incomplete endocytic events and events in which pixels assigned to one clathrin structure merged with or split from an adjacent structure [ 8 ] . For each cell of interest , the first , 50 pits lacking virus particles were analyzed in detail to measure the coat lifetime and protein composition ( see below ) . All complete virus uptake events that occurred in these cells were also analyzed in a similar manner . With the exception of internalization events blocked by latB , incomplete internalization events or events truncated by the start / end of image acquisition were not analyzed . Aggregates of virus were identified as objects with fluorescence intensities greater than that of single particles and were excluded from all analyses . IMAB was used to measure the fluorescence intensity of coat components or virus particles in the following manner . A roughly spherical mask encompassing 69 pixels was centered on the peak fluorescence intensity of the object in all frames that the object was detectable above the local background fluorescence . The contri - bution of the local background fluorescence was estimated by measuring the average intensity of pixels within a ring of single pixel width that extended from the outer boundary of the object mask . The intensity of the pixels within the object mask was then summed , and the average intensity value of pixels in the local background was subtracted from each pixel within the object mask to estimate the fluorescence contributed by proteins ( or dye molecules ) within the object of interest . Electron microscopy Purified virus particles were deposited onto carbon - coated copper grids and stained with 2 % phosphotungstic acid ( PTA ) in H 2 O ( pH 7 . 5 ) . To visualize DI - T particles in clathrin - coated pits , BSC1 cells were inoculated with unlabeled DI - T particles using a dose that yielded , 1000 attached particles per cell after 10 min . Samples were then processed for ultra - thin sectioning as previously described [ 15 , 59 ] . Electron micrographs of VSV particles in clathrin endocytic structures were obtained from cells infected with VSV for 6 h , where the entry of newly released particles could readily be visualized . Virus particles and ultra - thin sections of cells were viewed using a Tecnai G 2 Spirit BioTWIN transmission electron microscope ( FEI , Hillsboro , OR ) . Figure 7 . Model of DI - T and VSV entry . DI - T ( above ) and VSV ( below ) particles engage host cells through interactions between the viral surface glycoproteins and unknown cellular receptor moieties . Following attachment , both particle types undergo slow diffusion ( diffusion coefficient , 5 6 10 2 11 cm 2 s 2 1 ) on the cell surface for an average of , 2 min . before being captured by a clathrin - coated pit . For DI - T , continued clathrin assembly drives complete particle envelopment by the plasma membrane and leads to virus endocytosis . In contrast , the presence of a VSV particle in a coated pit physically prevents complete membrane constriction by the clathrin machinery and causes clathrin assembly to halt prematurely . The force provided by actin polymerization then further remodels the plasma membrane and thereby encloses the virus particle in a partially clathrin - coated vesicle . doi : 10 . 1371 / journal . ppat . 1001127 . g007 Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 10 September 2010 | Volume 6 | Issue 9 | e1001127 Supporting Information Video S1 Clathrin - dependent DI - T internalization . BSC1 cells stably expressing s 2 - eGFP ( green ) were inoculated with DI - T particles , and time - lapse images were acquired at 4 s intervals . A 4 . 2 6 4 . 2 m m 2 area of the cell surface is shown as in Figure 2B . The video depicts a single DI - T particle ( red ) that attaches to the plasma membrane and diffuses slowly . A clathrin - coated pit captures the virus particle when a dim adaptor spot colocalizes with the virus signal . The adaptor fluorescence intensity increases as coated pit assembly proceeds , and particle internalization occurs shortly after the adaptor signal peaks . Disappearance of the adaptor signal signifies clathrin uncoating , and the virus - containing vesicle is then transported toward the cell interior . Found at : doi : 10 . 1371 / journal . ppat . 1001127 . s001 ( 0 . 31 MB AVI ) Video S2 Sequential internalization of single VSV and DI - T particles by the same cell . BSC1 cells stably expressing s 2 - eGFP ( green ) were simultaneously inoculated with wild - type VSV and DI - T particles , and time - lapse images were acquired at 4 s intervals . A 3 . 5 6 3 . 5 m m 2 area of the cell surface is shown as in Figure 3A . At the outset of imaging ( t = 0 ) , one VSV ( red ) particle and one DI - T particle ( blue ) are visible on the cell surface . The VSV particle enters first , followed by the DI - T particle . Found at : doi : 10 . 1371 / journal . ppat . 1001127 . s002 ( 0 . 39 MB AVI ) Video S3 Uptake of a single DI - T particle by a cell expressing eGFP - LCa . BSC1 cells transiently expressing eGFP - LCa were inoculated with DI - T and full - length VSV particles , and images were captured as for Video S2 . At the onset of imaging , a DI - T particle is present within a 3 . 5 6 3 . 5 m m 2 area of the cell surface . Shortly thereafter , the particle briefly colocalizes with a spot of eGFP - LCa but does not enter . As visualized in Figure 3B , a clathrin - coated pit subsequently initiates near the virion , and the particle disappears into the cell after the clathrin signal peaks . Found at : doi : 10 . 1371 / journal . ppat . 1001127 . s003 ( 0 . 77 MB AVI ) Video S4 Uptake of a single VSV particle by a cell expressing eGFP - LCa . The movie depicts a separate area ( of equal size ) of the plasma membrane from the same cell that internalized the particle shown in Video S3 . A VSV particle ( red ) attaches to the cell surface , and a clathrin endocytic structure subsequently internal - izes the particle ( Figure 3D ) . Found at : doi : 10 . 1371 / journal . ppat . 1001127 . s004 ( 0 . 77 MB AVI ) Video S5 Cortactin recruitment during conventional clathrin - coated pit formation . Time - lapse images were acquired at 3 s intervals from a cell transiently co - expressing mCherry - LCa ( red ) and cortactin - eGFP ( green ) ( Figure 5A ) . A 7 . 0 6 7 . 0 m m 2 area of the cell surface is shown . Note that the cortactin signal is nearly indistinguishable from the local background during most clathrin endocytic events . Found at : doi : 10 . 1371 / journal . ppat . 1001127 . s005 ( 3 . 01 MB AVI ) Video S6 Cortactin recruitment during the internalization of a single DI - T particle . BSC1 cells transiently co - expressing mCherry - LCa ( red ) and cortactin - eGFP ( green ) for 18 h were inoculated with DI - T particles ( blue ) , and time - lapse images were acquired at 3 s intervals . The video shows the internalization of a single DI - T particle by a clathrin - coated vesicle . The internaliza - tion event is centered within a 3 . 5 6 3 . 5 m m 2 area of the cellular plasma membrane ( Figure 5B ) . The left panel shows an overlay of the 3 channels , and the right panel shows only the cortactin channel displayed in monochrome . Note that the cortactin fluorescence intensity during DI - T entry is nearly indistinguishable from the local background . Found at : doi : 10 . 1371 / journal . ppat . 1001127 . s006 ( 1 . 42 MB AVI ) Video S7 Cortactin recruitment during the internalization of a wild - type VSV particle . BSC1 cells transiently co - expressing mCherry - LCa ( red ) and cortactin - eGFP ( green ) for 18 h were inoculated with VSV particles ( blue ) , and time - lapse images were acquired at 3 s intervals . The video is displayed as described for Video S3 and shows the clathrin - dependent uptake of a single VSV particle in a 3 . 5 6 3 . 5 m m 2 area of the plasma membrane ( Figure 5C ) . Note the visible burst of cortactin that occurs in the final stages of VSV internalization . Found at : doi : 10 . 1371 / journal . ppat . 1001127 . s007 ( 1 . 51 MB AVI ) Video S8 BSC1 cells stably expressing s 2 - eGFP were treated with 6 . 3 m M latB for 12 min . Cells were inoculated with DI - T ( blue ) and VSV ( red ) particles , and images were acquired at 4 s intervals . The movie shows an 8 . 8 6 5 . 0 m m 2 area of the plasma membrane . The 2 DI - T particles ( the lower particle is already in a coated pit at the onset of imaging ) enter in the presence of latB , while the 2 VSV particles are subsequently captured by coated pits but fail to enter the cell ( Figure 6A , B ) . Found at : doi : 10 . 1371 / journal . ppat . 1001127 . s008 ( 4 . 65 MB AVI ) Acknowledgments We express gratitude to Eric Marino for maintaining the imaging resource used in this study . We also gratefully acknowledge Maria Ericsson and Irene Kim for preparation of samples for electron microscopic analysis and members of the Kirchhausen and Whelan labs for thought - provoking discussions . Author Contributions Conceived and designed the experiments : DKC RHM SPJW TK . Performed the experiments : DKC . Analyzed the data : DKC SPJW TK . Contributed reagents / materials / analysis tools : DKC RHM SPJW TK . Wrote the paper : DKC SPJW TK . References 1 . Conner SD , Schmid SL ( 2003 ) Regulated portals of entry into the cell . Nature 422 : 37 \u2013 44 . 2 . Kirchhausen T ( 2000 ) Clathrin . Annu Rev Biochem 69 : 699 \u2013 727 . 3 . Kirchhausen T ( 2009 ) Imaging endocytic clathrin structures in living cells . Trends Cell Biol 19 : 596 \u2013 605 . 4 . Owen DJ , Evans PR ( 1998 ) A structural explanation for the recognition of tyrosine - based endocytotic signals . Science 282 : 1327 \u2013 1332 . 5 . Honing S , Ricotta D , Krauss M , Spate K , Spolaore B , et al . ( 2005 ) Phosphatidylinositol - ( 4 , 5 ) - bisphosphate regulates sorting signal recognition by the clathrin - associated adaptor complex AP2 . Mol Cell 18 : 519 \u2013 531 . 6 . Damke H , Baba T , Warnock DE , Schmid SL ( 1994 ) Induction of mutant dynamin specifically blocks endocytic coated vesicle formation . J Cell Biol 127 : 915 \u2013 934 . 7 . Lee DW , Zhao X , Zhang F , Eisenberg E , Greene LE ( 2005 ) Depletion of GAK / auxilin 2 inhibits receptor - mediated endocytosis and recruitment of both clathrin and clathrin adaptors . J Cell Sci 118 : 4311 \u2013 4321 . 8 . Massol RH , Boll W , Griffin AM , Kirchhausen T ( 2006 ) A burst of auxilin recruitment determines the onset of clathrin - coated vesicle uncoating . Proc Natl Acad Sci U S A 103 : 10265 \u2013 10270 . 9 . Ehrlich M , Boll W , Van Oijen A , Hariharan R , Chandran K , et al . ( 2004 ) Endocytosis by random initiation and stabilization of clathrin - coated pits . Cell 118 : 591 \u2013 605 . 10 . Loerke D , Mettlen M , Yarar D , Jaqaman K , Jaqaman H , et al . ( 2009 ) Cargo and dynamin regulate clathrin - coated pit maturation . PLoS Biol 7 : e57 . 11 . Hanover JA , Beguinot L , Willingham MC , Pastan IH ( 1985 ) Transit of receptors for epidermal growth factor and transferrin through clathrin - coated pits . Analysis of the kinetics of receptor entry . J Biol Chem 260 : 15938 \u2013 15945 . 12 . Anderson RG , Brown MS , Goldstein JL ( 1977 ) Role of the coated endocytic vesicle in the uptake of receptor - bound low density lipoprotein in human fibroblasts . Cell 10 : 351 \u2013 364 . Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 11 September 2010 | Volume 6 | Issue 9 | e1001127 13 . van der Schaar HM , Rust MJ , Chen C , van der Ende - Metselaar H , Wilschut J , et al . ( 2008 ) Dissecting the cell entry pathway of dengue virus by single - particle tracking in living cells . PLoS Pathog 4 : e1000244 . 14 . Rust MJ , Lakadamyali M , Zhang F , Zhuang X ( 2004 ) Assembly of endocytic machinery around individual influenza viruses during viral entry . Nat Struct Mol Biol 11 : 567 \u2013 573 . 15 . Cureton DK , Massol RH , Saffarian S , Kirchhausen TL , Whelan SP ( 2009 ) Vesicular stomatitis virus enters cells through vesicles incompletely coated with clathrin that depend upon actin for internalization . PLoS Pathog 5 : e1000394 . 16 . Veiga E , Guttman JA , Bonazzi M , Boucrot E , Toledo - Arana A , et al . ( 2007 ) Invasive and adherent bacterial pathogens co - Opt host clathrin for infection . Cell Host Microbe 2 : 340 \u2013 351 . 17 . Simpson RW , Hauser RE , Dales S ( 1969 ) Viropexis of vesicular stomatitis virus by L cells . Virology 37 : 285 \u2013 290 . 18 . Harris A , Cardone G , Winkler DC , Heymann JB , Brecher M , et al . ( 2006 ) Influenza virus pleiomorphy characterized by cryoelectron tomography . Proc Natl Acad Sci U S A 103 : 19123 \u2013 19127 . 19 . Meier E , Harmison GG , Keene JD , Schubert M ( 1984 ) Sites of copy choice replication involved in generation of vesicular stomatitis virus defective - interfering particle RNAs . J Virol 51 : 515 \u2013 521 . 20 . Huang AS , Wagner RR ( 1966 ) Defective T particles of vesicular stomatitis virus . II . Biologic role in homologous interference . Virology 30 : 173 \u2013 181 . 21 . Peng G , Tsao J , Schein S , Green TJ , Luo M , et al . ( 2010 ) Cryo - EM model of the bullet - shaped vesicular stomatitis virus . Science 327 : 689 \u2013 693 . 22 . Huang AS , Greenawalt JW , Wagner RR ( 1966 ) Defective T particles of vesicular stomatitis virus . I . Preparation , morphology , and some biologic properties . Virology 30 : 161 \u2013 172 . 23 . Wagner RR , Schnaitman TA , Snyder RM ( 1969 ) Structural proteins of vesicular stomatitis viruses . J Virol 3 : 395 \u2013 403 . 24 . Pattnaik AK , Ball LA , LeGrone AW , Wertz GW ( 1992 ) Infectious defective interfering particles of VSV from transcripts of a cDNA clone . Cell 69 : 1011 \u2013 1020 . 25 . Whelan SP , Ball LA , Barr JN , Wertz GT ( 1995 ) Efficient recovery of infectious vesicular stomatitis virus entirely from cDNA clones . Proc Natl Acad Sci U S A 92 : 8388 \u2013 8392 . 26 . Saffarian S , Cocucci E , Kirchhausen T ( 2009 ) Distinct dynamics of endocytic clathrin - coated pits and coated plaques . PLoS Biol 7 : e1000191 . 27 . Uruno T , Liu J , Zhang P , Fan Y , Egile C , et al . ( 2001 ) Activation of Arp2 / 3 complex - mediated actin polymerization by cortactin . Nat Cell Biol 3 : 259 \u2013 266 . 28 . McNiven MA , Kim L , Krueger EW , Orth JD , Cao H , et al . ( 2000 ) Regulated interactions between dynamin and the actin - binding protein cortactin modulate cell shape . J Cell Biol 151 : 187 \u2013 198 . 29 . Merrifield CJ , Perrais D , Zenisek D ( 2005 ) Coupling between clathrin - coated - pit invagination , cortactin recruitment , and membrane scission observed in live cells . Cell 121 : 593 \u2013 606 . 30 . Coue M , Brenner SL , Spector I , Korn ED ( 1987 ) Inhibition of actin polymerization by latrunculin A . FEBS Lett 213 : 316 \u2013 318 . 31 . Coyne CB , Bergelson JM ( 2006 ) Virus - induced Abl and Fyn kinase signals permit coxsackievirus entry through epithelial tight junctions . Cell 124 : 119 \u2013 131 . 32 . Coyne CB , Kim KS , Bergelson JM ( 2007 ) Poliovirus entry into human brain microvascular cells requires receptor - induced activation of SHP - 2 . Embo J 26 : 4016 \u2013 4028 . 33 . Kaksonen M , Toret CP , Drubin DG ( 2006 ) Harnessing actin dynamics for clathrin - mediated endocytosis . Nat Rev Mol Cell Biol 7 : 404 \u2013 414 . 34 . Kaksonen M , Sun Y , Drubin DG ( 2003 ) A pathway for association of receptors , adaptors , and actin during endocytic internalization . Cell 115 : 475 \u2013 487 . 35 . Idrissi FZ , Grotsch H , Fernandez - Golbano IM , Presciatto - Baschong C , Riezman H , et al . ( 2008 ) Distinct acto / myosin - I structures associate with endocytic profiles at the plasma membrane . J Cell Biol 180 : 1219 \u2013 1232 . 36 . Heuser J ( 1980 ) Three - dimensional visualization of coated vesicle formation in fibroblasts . J Cell Biol 84 : 560 \u2013 583 . 37 . Gottlieb TA , Ivanov IE , Adesnik M , Sabatini DD ( 1993 ) Actin microfilaments play a critical role in endocytosis at the apical but not the basolateral surface of polarized epithelial cells . J Cell Biol 120 : 695 \u2013 710 . 38 . Da Costa SR , Sou E , Xie J , Yarber FA , Okamoto CT , et al . ( 2003 ) Impairing actin filament or syndapin functions promotes accumulation of clathrin - coated vesicles at the apical plasma membrane of acinar epithelial cells . Mol Biol Cell 14 : 4397 \u2013 4413 . 39 . Shupliakov O , Bloom O , Gustafsson JS , Kjaerulff O , Low P , et al . ( 2002 ) Impaired recycling of synaptic vesicles after acute perturbation of the presynaptic actin cytoskeleton . Proc Natl Acad Sci U S A 99 : 14476 \u2013 14481 . 40 . Bourne J , Morgan JR , Pieribone VA ( 2006 ) Actin polymerization regulates clathrin coat maturation during early stages of synaptic vesicle recycling at lamprey synapses . J Comp Neurol 497 : 600 \u2013 609 . 41 . Boucrot E , Saffarian S , Massol R , Kirchhausen T , Ehrlich M ( 2006 ) Role of lipids and actin in the formation of clathrin - coated pits . Exp Cell Res 312 : 4036 \u2013 4048 . 42 . Qualmann B , Roos J , DiGregorio PJ , Kelly RB ( 1999 ) Syndapin I , a synaptic dynamin - binding protein that associates with the neural Wiskott - Aldrich syndrome protein . Mol Biol Cell 10 : 501 \u2013 513 . 43 . Soulet F , Yarar D , Leonard M , Schmid SL ( 2005 ) SNX9 regulates dynamin assembly and is required for efficient clathrin - mediated endocytosis . Mol Biol Cell 16 : 2058 \u2013 2067 . 44 . Yarar D , Waterman - Storer CM , Schmid SL ( 2007 ) SNX9 couples actin assembly to phosphoinositide signals and is required for membrane remodeling during endocytosis . Dev Cell 13 : 43 \u2013 56 . 45 . Ringstad N , Nemoto Y , De Camilli P ( 1997 ) The SH3p4 / Sh3p8 / SH3p13 protein family : binding partners for synaptojanin and dynamin via a Grb2 - like Src homology 3 domain . Proc Natl Acad Sci U S A 94 : 8569 \u2013 8574 . 46 . Otsuki M , Itoh T , Takenawa T ( 2003 ) Neural Wiskott - Aldrich syndrome protein is recruited to rafts and associates with endophilin A in response to epidermal growth factor . J Biol Chem 278 : 6461 \u2013 6469 . 47 . Dawson JC , Legg JA , Machesky LM ( 2006 ) Bar domain proteins : a role in tubulation , scission and actin assembly in clathrin - mediated endocytosis . Trends Cell Biol 16 : 493 \u2013 498 . 48 . Roberts PC , Compans RW ( 1998 ) Host cell dependence of viral morphology . Proc Natl Acad Sci U S A 95 : 5746 \u2013 5751 . 49 . Roberts PC , Lamb RA , Compans RW ( 1998 ) The M1 and M2 proteins of influenza A virus are important determinants in filamentous particle formation . Virology 240 : 127 \u2013 137 . 50 . Neuman BW , Adair BD , Burns JW , Milligan RA , Buchmeier MJ , et al . ( 2005 ) Complementarity in the supramolecular design of arenaviruses and retroviruses revealed by electron cryomicroscopy and image analysis . J Virol 79 : 3822 \u2013 3830 . 51 . Murphy FA , Webb PA , Johnson KM , Whitfield SG , Chappell WA ( 1970 ) Arenoviruses in Vero cells : ultrastructural studies . J Virol 6 : 507 \u2013 518 . 52 . Rojek JM , Sanchez AB , Nguyen NT , de la Torre JC , Kunz S ( 2008 ) Different mechanisms of cell entry by human - pathogenic Old World and New World arenaviruses . J Virol 82 : 7677 \u2013 7687 . 53 . Chandran K , Sullivan NJ , Felbor U , Whelan SP , Cunningham JM ( 2005 ) Endosomal proteolysis of the Ebola virus glycoprotein is necessary for infection . Science 308 : 1643 \u2013 1645 . 54 . Negrete OA , Levroney EL , Aguilar HC , Bertolotti - Ciarlet A , Nazarian R , et al . ( 2005 ) EphrinB2 is the entry receptor for Nipah virus , an emergent deadly paramyxovirus . Nature 436 : 401 \u2013 405 . 55 . Clemente R , de la Torre JC ( 2009 ) Cell entry of Borna disease virus follows a clathrin - mediated endocytosis pathway that requires Rab5 and microtubules . J Virol 83 : 10406 \u2013 10416 . 56 . Bhattacharyya S , Warfield KL , Ruthel G , Bavari S , Aman MJ , et al . ( 2010 ) Ebola virus uses clathrin - mediated endocytosis as an entry pathway . Virology 401 : 18 \u2013 28 . 57 . Le Clainche C , Pauly BS , Zhang CX , Engqvist - Goldstein AE , Cunningham K , et al . ( 2007 ) A Hip1R - cortactin complex negatively regulates actin assembly associated with endocytosis . Embo J 26 : 1199 \u2013 1210 . 58 . Kaksonen M , Peng HB , Rauvala H ( 2000 ) Association of cortactin with dynamic actin in lamellipodia and on endosomal vesicles . J Cell Sci 113 Pt 24 : 4421 \u2013 4426 . 59 . Maupin P , Pollard TD ( 1983 ) Improved preservation and staining of HeLa cell actin filaments , clathrin - coated membranes , and other cytoplasmic structures by tannic acid - glutaraldehyde - saponin fixation . J Cell Biol 96 : 51 \u2013 62 . Virus Length Alters Clathrin - Dependent Endocytosis PLoS Pathogens | www . plospathogens . org 12 September 2010 | Volume 6 | Issue 9 | e1001127",
    "vasconcelosCopyNotCopy2017": "Design Computing and Cognition DCC\u201916 . J . S . Gero ( ed ) , pp . xx - yy . \u00a9 Springer 2016 1 To copy or not to copy : the influence of instructions in design fixation experiments Luis Vasconcelos , Chih - Chun Chen , Eloise Taysom and Nathan Crilly University of Cambridge , UK Design fixation experiments often require participants to solve a design problem whilst being exposed to an example solution and instructions for how to treat that example . However , little is known about the influence of such instructions , leading to difficulties in interpreting results and under - standing how the introduction of examples affects idea generation . In our experiment , participants were all provided with the same design problem and example solution , but were presented with different instructions , rang - ing from strongly encouraging copying the example to strongly discourag - ing copying . Analyses of participants\u2019 work indicated that only the instruc - tions encouraging copying had an effect . When encouraged to copy , participants tended to only copy the structural features of the example ra - ther than the underlying concept . By contrast , the number of features cop - ied was not reduced when participants were discouraged from copying . These findings suggest that there are subtle interactions between instruc - tions and stimuli that influence design fixation . Introduction Inspiration is vital to creative design . This has driven design researchers to conduct many studies into inspiration , for instance to find out what materi - als inspire designers [ 1 ] , how designers achieve inspiration [ 2 ] , and how inspiration can improve designers\u2019 performance [ 3 ] . Many of these studies have observed the use of external stimuli during idea generation , and have reported that whilst external stimuli can assist idea generation , they can al - so hinder it . This detrimental effect of inspiration is described in the design literature as \u2018design fixation\u2019 [ 4 ] . Design fixation was originally studied in L . A . Vasconcelos , C - C . Chen , E . Taysom and N . Crilly 2 situations where the stimuli represented possible design solutions to the problem that was being addressed . It is often measured in the solutions that the designers propose according to the repetition of features from the stim - uli that the designers see [ 4 - 10 ] . Although reproducing the features of good designs might be beneficial or efficient , the researchers have found that repetition is still present when the stimuli contain flawed inspiration sources , which they had thought designers would identify and avoid [ 4 , 7 , 10 ] . In general , fixation research suggests that blindly copying features from stimuli is harmful to idea generation , and therefore to the design process it - self . As such , some studies have incorporated constraining textual instruc - tions into the stimuli given to participants , to prevent them from copying features . However , the efficacy of the instructions seemed to vary across different accounts : they were effective in some cases [ 11 , 12 ] and ineffec - tive in others [ 4 , 13 ] . Thus , it is still uncertain how textual instructions can influence the copying behaviour , and this uncertainty might be attributed to different factors . For instance , people may tend to overlook instructions presented along with example material [ 14 ] , they can interpret the instruc - tions in different ways if instructions are not strict , or wonder why they are being exposed to the stimuli and change their idea generation process ac - cordingly . Whatever the reason might be , it is important to understand the relevance of the instructions provided to designers as part of the inspiration material . Methodologically , this would help in determining how fixation studies should be conducted and the results interpreted . More practically , it would also help to better understand how stimuli should be presented in software tools that aim to help idea generation by providing external stimuli to designers , and in other contexts in which designers are stimulat - ed with examples ( for such tools , see [ 15 - 17 ] ) . In order to better understand the influence of instructions on idea gen - eration , we conducted an experiment in which participants in different treatment conditions were provided with the same design problem and the same stimuli , but the instructions they were given with respect to that solu - tion differed between the groups . In reporting on that experiment we offer insights that are useful for interpreting the existing inspiration and fixation studies , and for designing new ones . These insights are also relevant for considering how inspirational material might best be presented to design - ers outside of experimental contexts , for example where computer tools are used to index and retrieve inspirational material . To copy or not to copy 3 Inspiration , fixation and the introduction of inspiration sources Design researchers have been studying many different aspects of the inspi - ration process , including properties of the inspiration sources ( e . g . quantity of sources ) and aspects of the design process ( e . g . total time available ) [ 18 ] . One of the stimuli characteristics that has been studied is whether good and bad examples would be copied indiscriminately . Researchers wanted to know whether designers would fixate on the examples they see and would copy them to some extent , irrespective of their quality . Indeed , a series of studies has found that participants still replicated features from previous examples even though they were flawed [ 4 , 7 , 10 , 11 , 19 , 20 ] . To counteract this indiscriminate replication , some studies have tried warning participants about the flaws in the examples [ 4 , 11 , 20 ] , while others have tried instructing participants not to copy the examples [ 13 , 21 , 22 ] . Con - sidering these two approaches , Chrysikou & Weisberg [ 11 ] found that warning participants of flaws in the examples was not enough ; they had to be told to avoid repeating those flaws . Yilmaz et al . [ 22 ] also reported tell - ing participants not to reproduce the examples and found that feature repe - tition was reduced . Conversely , Jansson & Smith [ 4 ] and Perttula & Sipil\u00e4 [ 13 ] found fixation effects even when participants were instructed to avoid using features from the examples provided . As such , surveying the pub - lished literature reveals conflicting results relating to the use of instruc - tions when providing stimuli to designers . Whilst many variables have been manipulated in fixation experiments , and some studies have already tested the effectiveness of using textual in - structions to some extent , the way the stimuli are introduced in such exper - iments has not yet been studied systematically . Such stimuli introductions can typically be divided into two components : a descriptive statement on what the stimulus is ( e . g . \u201chere is an example solution\u201d ) and a prescriptive instruction for how the stimulus should be used ( e . g . \u201cdon\u2019t copy its fea - tures\u201d ) . Currently , the stimuli introductions ( i . e . descriptions and instruc - tions ) given to participants vary from study to study . For instance , partici - pants have been told that the example should be considered a solution for the given problem [ 7 ] , that it was provided to help them get started [ 6 ] , that it was there to raise thoughts [ 23 ] , and that it should be used to awak - en thoughts but should not be reproduced [ 13 ] ( also , sometimes the studies do not report how the stimuli were introduced ) . Such variation in the way the stimuli are provided can be attributed to a lack of agreement across studies about which \u2018real world\u2019 situation is being simulated ( e . g . contexts in which examples are accidentally seen , intentionally searched for , or al - ready known ) . Regardless of the reason , the variation in the way copying L . A . Vasconcelos , C - C . Chen , E . Taysom and N . Crilly 4 is encouraged or discouraged makes it difficult to know to what extent the design work is being influenced by such instructions . It could even be pos - sible that designers will incorporate features from the examples provided irrespective of how constraining the instructions provided to them are . To investigate the influence of stimuli instructions on idea generation , Smith et al . [ 21 ] did an experiment in which participants were asked to ei - ther conform to or diverge from the example solutions provided in a crea - tive task . Some participants ( group 1 ) were told that the examples were great ideas previously created for that task , and that participants should create ideas like those whilst not copying them exactly . Other participants ( group 2 ) were told that the examples restricted people\u2019s creativity , and that participants should create ideas that were different to the examples . Finally , other participants ( control group ) were given the same examples but without any instructions . It was found that when compared to the ideas generated by participants in the control group , those in group 1 and in group 2 generated more ideas containing features from the examples and that groups 1 and 2 were not different in this respect . Based on these re - sults , the researchers proposed that the participants did not conform to ex - amples because they had assumed that they should ; they conformed be - cause they could not forget the examples that they had seen . However , the instructions used in the experiment were not strict ( i . e . they suggested how ideas should be created , but did not forbid or require the use of features from the examples , allowing participants to interpret the information in different ways ) and both the description of the stimuli and the instruction for their use varied between groups , making it difficult to infer the influ - ence of each piece of information in isolation . We believe that a different experimental setup could provide research with more data , complementing the results previously found and helping the field to clarify the influence of the textual instructions used in experimental research and professional de - sign practice . To investigate this , we designed and conducted an experi - ment to find out the effects of instructions in design inspiration and fixa - tion , be the instructions neutral or either \u2013 slightly or very \u2013 encouraging or discouraging Methodology Objective and hypothesis This experiment investigates how textual instructions accompanying ex - ternal inspiration sources can shape the design work of the participants . In To copy or not to copy 5 particular , we hypothesise that instructions provided along with external sources will not influence the ideas that participants generate . Experimental method Participants were randomly allocated to different experimental conditions . They were verbally asked to be creative and to design , individually , as many ideas as possible to a given problem . They were also asked to sketch and describe in writing their ideas on sheets of paper . Except for a control group that designed without any stimuli , all other experimental groups re - ceived a sketch of one example solution and a description of what the sketch represented . These stimulated groups received instructions for the use of features from the example , and these instructions varied with re - spect to how constraining or encouraging they were ( see the materials sec - tion for the complete instructions ) . Finally , the authors assessed the partic - ipants\u2019 ideas to evaluate the influence of the instructions on the level of fixation demonstrated . Participants One hundred and sixty - eight first - year students in Engineering from the University of Cambridge , UK , were assigned to six experimental groups ( n = 28 ) . Participation in the experiment was part of the students\u2019 education , and was aimed at collecting data that could later be used to introduce them to the concept of design fixation . No demographic data was collected from the participants , but as first year undergraduate students they were broadly similar in age and design experience , drawn from a cohort with a male - female ratio of 3 : 1 . Task and problem The participants were told to solve the following problem . \u201c Bicycles are a popular mode of transportation and recreation for many people . While growing up , a person might go through a series of ever - larger bikes , some - times having several models , one after the other . However , having several bikes can be a problem for many reasons . Your task is to generate as many ideas as possible to eliminate the need to have multiple bikes as people grow up . \u201d This problem was selected because it was expected to satisfy the follow - ing three criteria . First , it was unlikely that the participants had designed solutions to it before , although they were likely to have experienced the situation described in the problem previously ( i . e . while growing up , they probably had multiple bikes ) , therefore helping their understanding of it . L . A . Vasconcelos , C - C . Chen , E . Taysom and N . Crilly 6 Second , the problem could be solved in many different ways , with many different underlying principles being applied , thus leaving enough room for creativity . Finally , both the design brief and the potential ideas held a low level of complexity , thus being suitable for a quick experiment fitting with the constraints of the course . Procedure ( overview ) The experiment took place in a large lecture theatre with all the partici - pants present . During the first five minutes , the participants listened to an oral explanation about the activities to follow and received all the material they needed . Participants in the five stimulated groups ( SGs ) received the design brief , the inspiration source , and blank sheets of paper , whilst par - ticipants in the control group ( CG ) received the brief without any inspira - tion source . Then , the participants were asked to think of ideas for three minutes without actually committing any designs to paper ( because differ - ent participants had different materials and content , this ensured they all had enough time to read all the materials and start developing some ideas ) . Finally , for the remaining ten minutes , all participants individually gener - ated as many ideas as possible in silence , ideally including both a sketch and a written description for each idea . Materials All participants received the same design problem written on an A4 sheet , as well as blank A4 sheets to sketch and annotate their own ideas . Except for the participants in the control group , all participants received one addi - tional sheet with an example solution , i . e . an annotated sketch of a bike ( Figure 1 ) . To copy or not to copy 7 Fig1 . Example solution provided to participants along with the following descrip - tion . \u201c A modular bike with parts of various sizes that can be connected and swapped to fit people with very different heights . Apart from the socketing parts and expansible / contractible wheels , the angles between tubes can also be modified in specific joints \u201d . The sketch used is a modification of the ECO 07 Compactable Urban Bicycle [ 24 ] . The example solution was preceded with the description : \u201cBelow is an ex - ample of how you should present your ideas ( i . e . annotated sketches ) \u201d . This description was either immediately followed by an instruction regard - ing the use of features from the examples ( constraining or encouraging ) or by no instruction whatsoever . The instructions for the different experi - mental groups are listed below against a code for each experimental group . \u2022 SG2\u22122 ( strictly forbidding ) : \u201c make sure you do not use features from this example in your own work \u201d . \u2022 SG\u22121 ( constraining ) : \u201c avoid using features from this example in your own work \u201d . \u2022 SG0 ( neutral ) : no instruction was given . \u2022 SG + 1 ( encouraging ) : \u201c consider using features from this example in your own work \u201d . \u2022 SG + 2 ( strictly requiring ) : \u201c make sure you use features from this exam - ple in your own work \u201d . L . A . Vasconcelos , C - C . Chen , E . Taysom and N . Crilly 8 Analysis The assessment of the participant\u2019s ideas was conducted by the first three authors of this study , with backgrounds in design , computer science , and mechanical engineering respectively . Initially , the three evaluators agreed on the metrics to be included in the analysis . After that , the three evalua - tors analysed the design work of a random experimental group together in order to agree on the assessment method , ultimately reaching a consensus with respect to how to interpret and assess the ideas . Finally , each of the evaluators individually judged a subset of the remaining ideas . If any eval - uator had trouble judging an idea , this idea was then discussed collectively . This interactive assessment method was intended to offer a reliable analy - sis of the ideas ; but because there was no redundancy in the assessment , no inter - rater agreement could be calculated . We considered \u2018one idea\u2019 either to be a sketch or a written description ( usually both ) that presented an un - derstandable way to solve the problem . Participants often generated more than one idea , but in some cases all ideas could be considered as one , par - ticularly when the idea was a bike . For instance , if the ideas could all be incorporated onto the same bike without interference , then they were con - sidered as a single idea . Conversely , if there were two or more ideas for the same bike component ( e . g . frame , wheel , handle bar ) , then they were considered to be distinct ideas . The metrics used in the assessment were \u2018idea fluency\u2019 and idea \u2018repeti - tion\u2019 . Idea fluency is the total number of ideas generated , also called \u2018quantity\u2019 elsewhere [ 25 ] . Idea repetition might happen in different levels , for instance , the repetition of idea types , conceptual features , or structural features . With respect to the idea type , we divided the ideas into two broad categories : bike and non - bike ideas , thus by designing a bike the partici - pant would be repeating the idea type . With respect to their conceptual fea - tures , we also divided the ideas into two categories : modular or non - modular ideas , thus by designing a modular idea the participant would be repeating the conceptual feature . Finally , we examined the incorporation of structural features in the participants\u2019 ideas . These features were intention - ally included into the example design in order to permit a measure of fixa - tion . There were five structural features : swappable components to change bike size ; frame joints ( lugs ) that act as sockets for the tubes ; wheels with bendable spokes ; an hourglass - shaped frame ; and a saddle that cannot be adjusted in height directly . Eight participants ( 4 . 8 % ) either did not generate any idea or generated ideas that could not be interpreted by the authors ; the results from such participants were not included in our analysis . The adjusted number of par - ticipants per experimental group is indicated in the following section . To copy or not to copy 9 Results and discussion Idea fluency Instructions had no effect on the number of ideas generated . The data did not satisfy the assumptions for a standard ANOVA , therefore a non - parametic Kruskal \u2013 Wallis test equivalent to a one - way ANOVA was im - plemented instead . The results show that the number of ideas generated did not vary significantly across the five stimulated groups ( SGs ) ( H ( 4 ) = 2 . 63 , p = . 62 ) . However , there is a significant difference in the idea fluency be - tween participants in these groups and those in the control group ( CG ) ( H ( 5 ) = 18 . 27 , p < . 01 ) , with participants in the control group generating a greater number of ideas . Additionally , although the majority of participants in the stimulated groups had an idea fluency of 1 , participants in the non - constraining groups ( SG , SG + 1 , and SG + 2 ) had a higher frequency of flu - encies greater than 1 ( i . e . more participants in those groups generated more than one idea ) . However , this difference was not shown to be significant ( X 2 ( 4 ) = 19 . 85 , p = . 47 ) . A significant difference in the frequency of idea flu - encies was found between the stimulated groups and the control group ( X 2 ( 5 ) = 55 . 38 , p < . 001 ) , with participants in the control group having a higher frequency of fluencies greater than 1 . Table 1 shows summary sta - tistics for these results . Table 1 Summary of ideas generated per participant and ideas frequencies across groups Generated ideas SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 CG Mean ( and SD ) for the number of ideas per participant 1 . 50 ( 1 . 27 ) 1 . 30 ( 0 . 61 ) 1 . 48 ( 0 . 49 ) 1 . 41 ( 0 . 56 ) 1 . 54 ( 1 . 10 ) 2 . 39 ( 1 . 31 ) Range of ideas per participant 1 - 6 1 - 3 1 - 4 1 - 4 1 - 6 1 - 5 Total number of ideas 39 35 40 38 43 67 Participants with 1 idea ( and % ) 21 ( 80 . 8 % ) 21 ( 77 . 8 % ) 16 ( 59 . 3 % ) 19 ( 70 . 4 % ) 19 ( 67 . 9 % ) 11 ( 39 . 3 % ) Participants with 2 ideas ( and % ) 2 ( 7 . 7 % ) 4 ( 14 . 8 % ) 10 ( 37 % ) 6 ( 22 . 2 % ) 7 ( 25 % ) 3 ( 10 . 7 % ) Participants with 3 ideas ( and % ) 1 ( 3 . 9 % ) 2 ( 7 . 4 % ) 0 ( 0 . 0 % ) 1 ( 3 . 7 % ) 0 ( 0 . 0 % ) 7 ( 25 % ) Participants with 4 ideas ( and % ) 0 ( 0 . 0 % ) 0 ( 0 . 0 % ) 1 ( 3 . 7 % ) 1 ( 3 . 7 % ) 1 ( 3 . 6 % ) 6 ( 21 . 4 % ) L . A . Vasconcelos , C - C . Chen , E . Taysom and N . Crilly 10 Participants with 5 ideas ( and % ) 1 ( 3 . 9 % ) 0 ( 0 . 0 % ) 0 ( 0 . 0 % ) 0 ( 0 . 0 % ) 0 ( 0 . 0 % ) 1 ( 3 . 6 % ) Participants with 6 ideas ( and % ) 1 ( 3 . 9 % ) 0 ( 0 . 0 % ) 0 ( 0 . 0 % ) 0 ( 0 . 0 % ) 1 ( 3 . 6 % ) 0 ( 0 . 0 % ) Total number of participants 26 27 27 28 28 28 These results reveal that the idea generation rate was not influenced by how encouraging or constraining the instructions were . However , the pres - ence of an example design affected idea generation : designing without ex - posure to stimuli resulted in more ideas being generated , which we inter - pret as a benefit of isolation from examples . This effect is consistent with other studies in which seeing an example caused reduction in the idea flu - ency [ 7 ] , although studies have also reported an increase in the idea fluen - cy as a result from external stimulation [ 26 ] or even no effect at all [ 4 ] . Fi - nally , we should mention that although all groups were asked to present their ideas with both sketches and textual descriptions , the control group produced many ideas that were presented only in text . Thus , it is possible that the control group created more ideas because they did not spend their time sketching every idea . Repetition of the idea type Instructions had no effect on the number of bike ideas . The data did not satisfy the assumptions for a standard ANOVA , therefore a Kruskal \u2013 Wallis test was is implemented instead . The results show that the number of bike ideas generated did not vary significantly across the five stimulated groups ( SGs ) ( H ( 4 ) = 2 . 98 , p = . 56 ) , nor between these groups and the con - trol group ( CG ) ( H ( 5 ) = 3 . 55 , p = . 62 ) . Consistent with these results , there is also no significant difference in the proportion of bike ideas generated ( compared to non - bike ideas ) across the five stimulated groups ( X 2 ( 4 ) = 2 . 53 , p = . 64 ) . However , there is a significant difference in the pro - portions between these groups and the control group ( X 2 ( 5 ) = 32 . 73 , p = . 001 ) , with participants in the control group having a greater proportion of non - bike ideas , such as other transportation means or policies to dis - courage the use and acquisition of bikes . Table 2 shows summary statistics for these results . Table 2 Summary of bike and non - bike ideas generated across groups Bike Ideas SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 CG Mean ( and SD ) for the number of bike ideas per 1 . 27 ( 0 . 92 ) 1 . 15 ( 0 . 36 ) 1 . 33 ( 0 . 83 ) 1 . 33 ( 0 . 55 ) 1 . 29 ( 0 . 53 ) 1 . 43 ( 0 . 84 ) To copy or not to copy 11 participant Range of bike ideas per participant 0 - 5 1 - 2 0 - 4 1 - 3 0 - 2 0 - 3 Number of bike ideas ( and % ) 33 ( 85 % ) 31 ( 89 % ) 36 ( 90 % ) 36 ( 95 % ) 36 ( 84 % ) 40 ( 60 % ) Number of non - bike ide - as ( and % ) 6 ( 15 % ) 4 ( 11 % ) 4 ( 10 % ) 2 ( 5 % ) 7 ( 16 % ) 27 ( 40 % ) If we adopt a bike and non - bike categorisation for the ideas generated , these results reveal that the type of idea generated was not influenced by the instructions . Additionally , the presence of the bike example did not in - crease the number of bike ideas either . However , whilst the number of bike ideas was roughly the same for all groups , there was a large difference in the proportion of bike ideas between the control and stimulated groups . Only 60 % of all ideas generated by the control group were bikes , whereas the stimulated groups had a much greater proportion of bike ideas ( 89 % on average ) . The results indicate that bike ideas were equally likely to be gen - erated irrespective of the experimental condition , but not seeing the exam - ple allowed participants from the control group to explore different areas of the solution space , again confirming the beneficial isolation effect . This effect is broadly consistent with other studies in which seeing an example caused participants to conform to certain types of solutions , thus reducing the diversity of ideas [ 4 , 5 , 7 ] . Repetition of the conceptual feature Instructions had no effect on the number of modular ideas . The data did not satisfy the assumptions for a standard ANOVA , therefore a Kruskal \u2013 Wallis test was is implemented instead . The results show that the number of modular ideas generated did not vary significantly across the five stimu - lated groups ( SGs ) ( H ( 4 ) = 4 . 21 , p = . 38 ) . However , there appears to be a significant difference in the repetition of modularity between the stimulat - ed groups and the control group ( CG ) ( H ( 5 ) = 11 . 40 , p < . 05 ) , with partici - pants in the control group creating a greater number of modular ideas . When looking at the frequencies , there is no significant difference in the proportion of modular ideas generated ( compared to non - modular ideas ) across the five stimulated groups ( X 2 ( 4 ) = 5 . 12 , p = . 27 ) , nor even when the control group is included in the comparison ( X 2 ( 5 ) = 5 . 68 , p = . 34 ) . Table 3 shows summary statistics for these results . Table 3 Summary of modular and non - modular ideas generated across groups Modular ideas SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 CG L . A . Vasconcelos , C - C . Chen , E . Taysom and N . Crilly 12 Mean ( and SD ) for the number of modular ideas per participant 0 . 08 ( 0 . 27 ) 0 . 3 ( 0 . 47 ) 0 . 19 ( 0 . 40 ) 0 . 19 ( 0 . 40 ) 0 . 18 ( 0 . 39 ) 0 . 43 ( 0 . 50 ) Range of modular ideas per participant 0 - 1 0 - 1 0 - 1 0 - 1 0 - 1 0 - 1 Number of modular ideas ( and % ) 2 ( 5 % ) 8 ( 23 % ) 5 ( 13 % ) 5 ( 13 % ) 5 ( 12 % ) 12 ( 18 % ) Number of non - modular ideas ( and % ) 37 ( 95 % ) 27 ( 77 % ) 35 ( 87 % ) 33 ( 87 % ) 38 ( 88 % ) 55 ( 82 % ) If we adopt a modular and non - modular categorisation for the ideas gener - ated , these results reveal that the type of idea generated was not influenced by the instructions . In fact , modular ideas were extremely rare across all groups . However , the results suggest that the presence of an example de - sign affected idea generation and the control group created on average more modular ideas than the stimulated groups . This apparently surprising result can be attributed to a higher overall number of ideas generated by the control group , which is supported by there being no difference in the proportion of modular ideas generated among all groups . Still , the fixation literature would suggest that the stimulated groups would generate more modular ideas , an effect similar to the repetition of the idea type ( i . e . bikes ) . In particular , participants in SG + 1 and SG + 2 were encouraged to use features from the example in their own work ( and modularity was a visible feature on the example provided ) but the results from those groups do not indicate that they acted accordingly . Similarly , participants in SG\u22122 and SG\u22121 were discouraged from using features from the example but generated as many modular ideas as the other groups . One possible expla - nation for this is that instructions have no influence on idea generation , as we had previously hypothesised . However , as there was no difference in the proportion of modular ideas between control and stimulated groups , we believe that the general principle of modularity ( included in the example as a conceptual feature ) was less obvious than the structural features and thus did not induce fixation effects . Repetition of structural features Instructions had a significant effect on the number of ideas that contained the structural features of the example provided . The data did not satisfy the assumptions for a standard ANOVA , therefore a Kruskal \u2013 Wallis test was implemented instead . The results show that the number of structural fea - tures incorporated into the participants\u2019 ideas varied significantly across the five stimulated groups ( SG ) ( H ( 4 ) = 41 . 62 , p < . 001 ) and between these To copy or not to copy 13 groups and the control group ( CG ) ( H ( 5 ) = 54 . 63 ; p < . 001 ) , with participants in the encouraged groups ( SG + 1 and SG + 2 ) incorporating a greater num - ber of structural features . Consistent with these results , there is also a sig - nificant difference in the relative number of ideas with structural features across the stimulated groups ( X 2 ( 4 ) = 37 . 32 , p < . 001 ) , with a higher number of features being associated with positive instructions to copy . There is al - so a significant difference between these groups and the control group ( X 2 ( 5 ) = 49 . 74 , p < . 001 ) , with the control group\u2019s repetition rate being close to the neutrally stimulated group ( SG0 ) . Table 4 shows summary statistics for these results . Table 4 Summary of features incorporated into the participants\u2019 ideas and fre - quencies of ideas with features across groups Feature SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 CG Mean ( and SD ) for the number features in - corporated per idea 0 . 08 ( 0 . 27 ) 0 . 17 ( 0 . 38 ) 0 . 10 ( 0 . 30 ) 0 . 71 ( 0 . 98 ) 0 . 93 ( 0 . 99 ) 0 . 13 ( 0 . 42 ) Range of features in - corporated 0 - 1 0 - 1 0 - 1 0 - 3 0 - 3 0 - 2 Ideas with 0 features ( and % ) 36 ( 92 % ) 29 ( 83 % ) 36 ( 90 % ) 22 ( 58 % ) 19 ( 44 % ) 60 ( 90 % ) Ideas with 1 features ( and % ) 3 ( 8 % ) 6 ( 17 % ) 4 ( 10 % ) 8 ( 21 % ) 11 ( 26 % ) 5 ( 7 % ) Ideas with 2 features ( and % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) 5 ( 13 % ) 10 ( 23 % ) 2 ( 3 % ) Ideas with 3 features ( and % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) 3 ( 8 % ) 3 ( 7 % ) 0 ( 0 % ) Ideas with 4 features ( and % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) Ideas with 5 features ( and % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) 0 ( 0 % ) Ideas with features in - corporated ( and % ) 3 ( 8 % ) 6 ( 17 % ) 4 ( 10 % ) 16 ( 42 % ) 23 ( 56 % ) 7 ( 10 % ) These results reveal that encouraging instructions influenced the repetition or incorporation of conceptual features , an outcome that does not support our hypothesis . On average , participants in SG + 1 and SG + 2 incorporated more features per idea and generated more ideas that incorporated any fea - ture . Additionally , participants in the strictly forbidden group , SG\u22122 , did not incorporate any structural feature from the example . However , partici - pants in SG\u22121 did not follow this trend and produced results similar to the neutral and control groups . This result is partially consistent with research L . A . Vasconcelos , C - C . Chen , E . Taysom and N . Crilly 14 from Smith et al . [ 21 ] but inconsistent with Chrysikou & Weisberg [ 11 ] , as encouraging instructions increased fixation , whereas constraining instruc - tions did not decrease it . However , contrary to the results from Smith et al . [ 21 ] , our constrained groups did not replicate more features than a control group , contradicting the idea that participants could not forget the example they had seen . It seems that a few structural features were naturally likely to be incorporated into the ideas generated ( supported by the similar results from SG\u22121 , SG0 and CG ) , with SG + 1 and SG + 2 incorporating more fea - tures because instructed to do so . When comparing the results from the repetition of structural features to the repetition of the conceptual feature used in this study , we infer that concrete structural features can be more easily copied or can fixate more than abstract conceptual features from ex - amples . This is similar to what has been suggested in previous studies [ 27 , 28 , 29 ] . Again , we should highlight that the control group produced pro - portionally less sketches than the other groups . This puts the other groups in an unfavourable position with respect to the count of structural features incorporated , since these groups had to represent a shape of the bike in which repetition could be more easily recognised \u2013 it is difficult to identify structural repetition when the idea is represented only by text \u2013 thus possi - bly biasing the results . Study limitations The main limitations of this study involve the duration of the generation session , the pool of participants chosen , the design problem used , and the assumptions for inter - rater agreement . These limitations are discussed next . The idea generation session in this study was ten minutes long , which can be considered short when compared to other fixation studies in which generation sessions lasted for 30 or 60 minutes [ 30 ] , and shorter still com - pared to professional practice [ 31 ] . Also , past research suggests [ 32 ] , novel ideas tend to occur later in the idea generation session . As a result , the short session adopted for this study might have contributed to inflated fixa - tion scores . The participants in this study were undergraduate students and the gen - eration session was part of an ongoing engineering course . This might have resulted in a more diligent participant behaviour when compared to other studies in which participants and experimenters did not have a stu - dent - lecturer relationship . As a result , the setup adopted for this study might have contributed to an increased participant adherence to the in - structions . The design problem used in this study was chosen because it was unlikely that the participants had designed solutions to it before . As To copy or not to copy 15 such , it is possible that using familiar problems will produce different re - sults , although research has demonstrated that fixation effects can be ob - served with both familiar and unfamiliar problems [ 4 ] . Finally , in this study we did not measure the inter - rater agreement for the assessment , a test which is often performed for similar studies . Such measurement is important to demonstrate that the reliability of the evalua - tors\u2019 assessment when working individually . Whereas the assessment per - formed in this study involved many interactions between evaluators ( thus we expected a high agreement between evaluators ) , we cannot quantify how good this agreement might have been , or even if there were varying levels of agreement for the different ideation metrics used in this study . Conclusion and future work In this study we have tested the influence of instructions on idea genera - tion . In particular , we analysed how instructions may affect the number of ideas generated and the repetition or incorporation of the example or its parts into the participants\u2019 ideas . The instructions used were provided along with an external stimulus and its description . It is important to dif - ferentiate the descriptions from instructions because in this study we have controlled the former but manipulated the latter . We found that instructions had some influence on the idea generation of our participants . When asked to use features from the example , participants copied structural features but failed to copy a more abstract conceptual feature . When asked not to use features from the example , however , most participants did not reduce the number of features copied . This result allows us to infer that more concrete features are easier to recognise \u2013 and thus reproduce \u2013 than more abstract features , such as modularity . Also , it might indicate that positive instruc - tions are more effective than negative ones , which can tell us how to frame future instructions , whether that is with respect to experimental stimuli in research or inspirational stimuli in design practice . Irrespective of how constraining the instructions were , participants ex - hibited fixation effects due to their exposure to the example design ( in comparison to the control group , all stimulated groups created more ideas of the same type as the example provided ) . This result is in line with many other design fixation experiments in which participants become stuck on a particular idea type . However , it is important to emphasise that the descrip - tion of the stimulus itself could also be causing the fixation effect as the stimulus was presented to the participants as \u201c an example of how they should present their ideas \u201d . Thus , perhaps there is an implicit suggestion L . A . Vasconcelos , C - C . Chen , E . Taysom and N . Crilly 16 for the participants to produce ideas similar to the example , i . e . a bike . Fu - ture studies could investigate such possibility and complement our under - standing about stimuli introduction with the influence of the descriptions provided to designers . In this study we observed that some results might have been influenced by participants communicating their ideas in different ways ( some of them drawing , some writing , and some doing both ) , an issue that cannot be con - firmed until the analysis of the results . In inspiration and fixation studies , participants are asked to provide their ideas according to a content tem - plate , but that is rarely compulsory . It is fundamental that all ideas contain the same elements ( e . g . text and sketch ) in order to be analysed and com - pared , so studies should make sure every idea has a similar content . Addi - tionally , future studies could investigate the difference between providing written , pictorial or spoken instructions to designers , but also how they as - similate instructions . For instance , it is possible that people need instruc - tions about the instructions ( i . e . \u201cread the instructions carefully\u201d ) , as one way to make sure that they will read and fully understand what is required from them . Finally , when considering design practice , the results reported here are relevant to how inspirational stimuli should be framed when presented to designers . This is particularly important for the development and imple - mentation of computer - aided design tools that provide designers with ex - ternal stimuli . Much has been researched on how such software tools might be structured and interacted with , and what form the inspirational stimuli should take [ 33 , 34 ] . However , it is also important to understand how those stimuli should be introduced , whether by description , instruc - tion , or both . Should designers be steered towards or away from the repeti - tion of structural features , directed to identify conceptual features , or simp - ly left alone to interpret and respond as they see fit ? By developing a better understanding of how stimuli instructions influence idea generation , we will move closer to answering such questions and thereby be more capable of supporting design activities . Acknowledgement The authors would like to thank Carlos Coimbra Cardoso for commenting on earlier drafts of this paper , to all the students for taking part in the ex - periment , and to the three DCC reviewers for improving this work with their feedback . This work was supported by the CAPES Foundation , Min - istry of Education of Brazil , under grant / process : BEX 11468 / 13 - 0 and by the UK\u2019s Engineering and Physical Sciences Research Council To copy or not to copy 17 ( EP / K008196 / 1 ) . Research data supporting this publication is available from the DSpace @ Cambridge repository at the web address given below . It contains data from the participants and from our analysis . The data from the participants consists of all the annotated sketches that participants gen - erated in the experiment . The data from the analysis consists of the evalua - tion for all ideas generated by the participants in the experiment . https : / / www . repository . cam . ac . uk / handle / 1810 / 254702 References 1 . Gon\u00e7alves M , Cardoso C , Badke - Schaub P ( 2014 ) What inspires designers ? Preferences on inspirational approaches during idea generation . Des Stud 35 : 29 \u2013 53 . 2 . Zhao M ( 2013 ) Seek it or let it come : how designers achieve inspirations . In : CHI13 Ext . Abstr . Hum . Factors Comput . Syst . ACM , pp 2779 \u2013 2784 3 . Eckert CM ( 1997 ) Design inspiration and design performance . In : Proc . 78th World Conf . Text . Inst . Citeseer , pp 369 \u2013 387 4 . Jansson DG , Smith SM ( 1991 ) Design fixation . Des Stud 12 : 3 \u2013 11 . 5 . Cardoso C , Badke - Schaub P ( 2011 ) The influence of different pictorial repre - sentations during idea generation . J Creat Behav 45 : 130 \u2013 146 . 6 . Dahl DW , Moreau P ( 2002 ) The Influence and Value of Analogical Thinking during New Product Ideation . J Mark Res 39 : 47 \u2013 60 . 7 . Linsey JS , Tseng I , Fu K , et al . ( 2010 ) A Study of Design Fixation , Its Mitiga - tion and Perception in Engineering Design Faculty . J Mech Des 132 : 041003 . 8 . Purcell AT , Gero JS ( 1996 ) Design and other types of fixation . Des Stud 17 : 363 \u2013 383 . 9 . Youmans RJ ( 2011 ) Design fixation in the wild : Design environments and their influence on fixation . J Creat Behav 45 : 101 \u2013 107 . 10 . Youmans RJ ( 2011 ) The effects of physical prototyping and group work on the reduction of design fixation . Des Stud 32 : 115 \u2013 138 . 11 . Chrysikou EG , Weisberg RW ( 2005 ) Following the Wrong Footsteps : Fixa - tion Effects of Pictorial Examples in a Design Problem - Solving Task . J Exp Psychol Learn Mem Cogn 31 : 1134 \u2013 1148 . 12 . Yilmaz S , Seifert CM , Gonzalez R ( 2010 ) Cognitive heuristics in design : In - structional strategies to increase creativity in idea generation . Artif Intell Eng Des Anal Manuf 24 : 335 \u2013 355 . 13 . Perttula M , Sipil\u00e4 P ( 2007 ) The idea exposure paradigm in design idea genera - tion . J Eng Des 18 : 93 \u2013 102 . 14 . LeFevre J - A , Dixon P ( 1986 ) Do Written Instructions Need Examples ? Cogn Instr 3 : 1 \u2013 30 . 15 . Chakrabarti A , Sarkar P , Leelavathamma B , Nataraju BS ( 2005 ) A functional representation for aiding biomimetic and artificial inspiration of new ideas . Artif Intell Eng Des Anal Manuf 19 : 113 \u2013 132 . L . A . Vasconcelos , C - C . Chen , E . Taysom and N . Crilly 18 16 . Linsey J , Wood K ( 2008 ) Wordtrees : A method for design - by - analogy . ASEE annual conference . 17 . Vattam S , Wiltgen B , Helms M , et al . ( 2011 ) DANE : Fostering Creativity in and through Biologically Inspired Design . In : Taura T , Nagai Y ( eds ) Des . Creat . 2010 . Springer London , pp 115 \u2013 122 18 . Sio UN , Kotovsky K , Cagan J ( 2015 ) Fixation or inspiration ? A meta - analytic review of the role of examples on design processes . Des Stud 39 : 70 \u2013 99 . 19 . Perttula MK , Liikkanen LA ( 2006 ) Exposure effects in design idea genera - tion : unconscious conformity or a product of sampling probability . In : Proc . Nord . pp 42 \u2013 55 20 . Viswanathan V , Atilola O , Esposito N , Linsey J ( 2014 ) A study on the role of physical models in the mitigation of design fixation . J Eng Des 25 : 25 \u2013 43 . 21 . Smith SM , Ward TB , Schumacher JS ( 1993 ) Constraining effects of examples in a creative generation task . Mem Cognit 21 : 837 \u2013 845 . 22 . Yilmaz S , Seifert CM , Gonzalez R ( 2010 ) Cognitive heuristics in design : In - structional strategies to increase creativity in idea generation . Artif Intell Eng Des Anal Manuf 24 : 335 \u2013 355 . 23 . Liikkanen LA , Perttula M ( 2010 ) Inspiring design idea generation : insights from a memory - search perspective . J Eng Des 21 : 545 \u2013 560 . 24 . ECO 07 - Compactable Urban Bicycle . In : Behance . https : / / www . behance . net / gallery / 293563 / ECO - 07 - Compactable - Urban - Bicycle . Accessed 10 Nov 2015 25 . Shah JJ , Smith SM , Vargas - Hernandez N ( 2003 ) Metrics for measuring idea - tion effectiveness . Des Stud 24 : 111 \u2013 134 . 26 . Purcell AT , Gero JS ( 1992 ) Artificial Intelligence in Design Conference 1991 Special Issue Effects of examples on the results of a design activity . Knowledge - Based Systems 5 : 82 \u2013 91 . 27 . Zahner D , Nickerson JV , Tversky B , et al ( 2010 ) A fix for fixa - tion ? Rerepre - senting and abstracting as creative processes in the de - sign of information sys - tems . Artificial Intelligence for Engineering Design , Analysis and Manufac - turing 24 : 231 \u2013 244 . 28 . Cheong H , Hallihan G , Shu LH ( 2014 ) Understanding Analogi - cal Reasoning in Biomimetic Design : An Inductive Approach . In : Gero JS ( ed ) Design Computing and Cognition \u201912 . Springer Nether - lands , Dordrecht , pp 21 \u2013 39 . 29 . Feng T , Cheong H , Shu LH ( 2014 ) Effects of Abstraction on Selecting Rele - vant Biological Phenomena for Biomimetic Design . Journal of Mechanical Design 136 : 111111 . 30 . Vasconcelos LA , Crilly N ( 2016 ) Inspiration and fixation : Questions , methods , findings , and challenges . Design Studies 42 : 1 \u2013 32 . 31 . Crilly N ( 2015 ) Fixation and creativity in concept development : The attitudes and practices of expert designers . Design Studies 38 : 54 \u2013 91 . 32 . Kudrowitz B , Dippo C ( 2013 ) Getting to the novel ideas : exploring the alter - native uses test of divergent thinking . In : ASME 2013 International Design Engineering Technical Conferences and Computers and Information in Engi - neering Conference . American Society of Mechanical Engineers , pp V005T06A013 \u2013 V005T06A013 . To copy or not to copy 19 33 . T\u00f6re Yargin G , Crilly N ( 2015 ) Information and interaction requirements for software tools supporting analogical design . Artificial Intelligence for Engi - neering Design , Analysis and Manufacturing 29 : 203 \u2013 214 . 34 . Shneiderman B ( 2000 ) Creating Creativity : User Interfaces for Supporting In - novation . ACM Trans Comput - Hum Interact 7 : 114 \u2013 138 .",
    "kordeAlternatingIndividualGroup2016": "Alternating individual and group idea generation : Finding the elusive synergy Runa Korde , Paul B . Paulus \u204e The University of Texas at Arlington , United States H I G H L I G H T S \u2022 Alternating individual and group ideation was better than group ideation . \u2022 Alternating individual and group ideation was better than solitary ideation . \u2022 The bene \ufb01 ts of alternation are only evident in the alone sessions . \u2022 The results provide evidence for the synergistic bene \ufb01 ts of group ideation . \u2022 The results support an association model of group ideation . a b s t r a c t a r t i c l e i n f o Article history : Received 3 May 2016 Revised 9 November 2016 Accepted 9 November 2016 Available online 23 November 2016 Three experiments were designed to test theef \ufb01 cacy ofideation procedures that involvedalternation ofindivid - ual and group idea generation sessions ( hybrid brainstorming ) as compared to traditional individual and group ideation . Thehybridconditionledtothebestperformanceintermsofnumberofideasgenerated . Thiseffectwas strongest in comparison to the group condition . A meta - analytic comparison involving all three experiments in - dicated that the hybrid condition outperformed both the alone and the group conditions . Since after each group ideaexchangesession there wasanenhancementinthenumberofideas generatedinthealonesession , thepat - tern of performance in the hybrid condition supported the cognitive perspective of group creativity ( Nijstad & Stroebe , 2006 ; Paulus & Brown , 2007 ) . Social cues in the form of the co - presence of other participants in the alonecondition , theadditionofpracticesessionstoallconditions , andanadditionalphasedidnotchangethepat - tern ofresults . Theresults oftheexperiments supportthe original suggestion by Osborn ( 1953 ) thatthemostef - fective brainstorming process is one that involves a variation in individual and group ideation . \u00a9 2016 Elsevier Inc . All rights reserved . Keywords : Group creativity Brainstorming Innovation Brainwriting Creativity Synergy 1 . Introduction Collaborative idea generation is a frequently used approach in many settings . Itis typicallyassumed thatsharingideasingroupswillincrease thenumberandnoveltyofideas ( Paulus , Dzindolet , Poletes , & Camacho , 1993 ; Sutton & Hargadon , 1996 ) . However , studies that have compared collaborative ideation or brainstorming with the performance of similar numbers of individual performers ( nominal groups ) in various formats have typically found that interactive groups tend not to be better than nominal groups ( cf . , Paulus & Coskun , 2012 ) . It has been suggested that alternating group and individual ideation may be a more optimal procedure than either alone or group ideation . However , research with this paradigm has been very limited thus far , and the results have been mixed . In a series of three experiments we examine more precisely than past research the variation in idea generation over the course of the alone and group alternation process in order to demon - strate the synergistic bene \ufb01 t of group ideation . A review of traditional ideation methods shows that the bene \ufb01 ts of group ideation depend on the type of paradigm used . In face - to - face ( FTF ) settingswhereinideas aresharedverbally , groupstypicallygener - ate fewer ideas than comparable size nominal groups , with the differ - ence increasing with group size ( Bouchard & Hare , 1970 ; Diehl & Stroebe , 1987 ) . However , this performance de \ufb01 cit for groups can disap - pear when ideas are exchanged either electronically ( electronic brain - storming - EBS ) or by writing ( brainwriting - BWr ) . These techniques help overcome the production blocking and evaluation apprehension that are some of the factors responsible for the low performance of FTF groups ( Dennis & Williams , 2003 ; Heslin , 2009 ) . Although the EBS and BWr techniques appear to be useful for group ideation , their bene \ufb01 ts are limited . EBS groups can typically outperform Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 \u204e Corresponding author at : Department of Psychology , The University of Texas at Arlington , Arlington , TX 76019 - 0528 , United States . E - mail address : Paulus @ uta . edu ( P . B . Paulus ) . http : / / dx . doi . org / 10 . 1016 / j . jesp . 2016 . 11 . 002 0022 - 1031 / \u00a9 2016 Elsevier Inc . All rights reserved . Contents lists available at ScienceDirect Journal of Experimental Social Psychology journal homepage : www . elsevier . com / locate / jesp nominals only if groups are size 8 or larger ( Derosa , Smith , & Hantula , 2007 ) . However , these effects are not large ( Paulus , Kohn , & Arditti , 2011 ) . Research has found that BWr groups can generate as many or more ideas than comparable nominal groups ( Goldenberg , Larson , & Wiley , 2013 ; Paulus , Korde , Dickson , Carmeli , & Cohen - Meitar , 2015 : Paulus & Yang , 2000 ) . The positive effects of BWr appear to depend on the exact procedure used ( Goldenberg et al . , 2013 ) . Even though EBS and BWr can eliminate the production losses ob - served in FTFbrainstorming , it is apparentthattheevidence for thesyn - ergistic effect of group collaboration using these paradigms is quite limited . A demonstration of synergy would require that the perfor - mance of the groups exceed those of the nominals ( Larson , 2010 ) . There are a number of potential reasons for this state of affairs . Sharing ideas in a group should be cognitively stimulating ( Nijstad & Stroebe , 2006 ; Paulus & Brown , 2007 ) . There is in fact evidence that exposure to the ideas of others can stimulate additional ideas because of associa - tive processes ( Dugosh , Paulus , Roland , & Yang , 2000 ; Nijstad , Stroebe , & Lodewijkx , 2002 ) and the exposure to different categories of ideas ( cf . , Deuja , Kohn , Paulus , & Korde , 2014 ) . However , the group exchange process can also slow down the rate of exchanging ideas . The ef \ufb01 ciency of the group exchange process is limited by the task or coordination de - mands of this process ( Steiner , 1972 ) . This is most evident in the FTF case since group members have to take turns to present their ideas . Al - though turn taking is not required in EBS and BWr , the time spent read - ing ideas from others takes away from generating one ' s own ideas . Furthermore , although ideas from others can be stimulating , they can also distract from one ' s own train of thought or may lead to free riding ( Kerr & Bruun , 1983 ) in that one may feel less responsibility for gener - ating ideas if others are generating many ideas . One solution to this dilemma is to have individuals participate in both aloneand groupsessions . Infact , Osborn ( 1953 ) noted \u201c an alterna - tion between groupideation and individualideation is desirable , since a combination of these two methods has produced maximum results in almost every case . \u201d ( p . 209 ) . Although Osborn provided no empirical data to support this claim , it seems reasonable to expect that combining individual and group brainstorming into a single paradigm or \u201c hybrid brainstorming \u201d may be an optimal procedure . This would allow both for unconstrained ideation in individual brainstorming and stimulation of additional ideas by exposure to ideas of others . This might be most evident with the EBS and BWr paradigms since these signi \ufb01 cantly re - duce production blocking . There has been very limited work on brainstorming procedures that combine the individual and group paradigms . Some studies have varied theorder of aloneand groupbrainstorming to determinewhichorder is most bene \ufb01 cial . One might expect that group brainstorming would be most bene \ufb01 cial after a period of individual brainstorming . This allows participants to generate their ideas in an unconstrained fashion \ufb01 rst , and then when thinking of new ideas becomes more dif \ufb01 cult , group in - teraction may provide cues to additional ideas or categories of ideas . One study has in fact found that this order is most bene \ufb01 cial ( cf . , Baruah & Paulus , 2008 ) . However , one can also argue that having group interaction before a solitary ideation session would be bene \ufb01 cial since this provides participants a chance to re \ufb02 ect on the shared ideas and build on them while the shared ideas are still salient in the associa - tive network ( Brown & Paulus , 2002 ) . Evidence for the superiority of this order has been obtained in two studies using BWr ( Paulus & Yang , 2000 ; Paulus et al . , 2015 ) and also in several studies with verbal brain - storming ( Dunnette , Campbell , & Jaastad , 1963 ; Leggett , Putman , Roland , & Paulus , 1996 ) . One of the more famous examples of this type of effect is Einstein ' s writing of the theory of relativity . After a longdiscussion session with his friends on this topic , heisolatedhimself in a room to write out the basics of his theory ( Brian , 1996 ) . Given the different \ufb01 ndings on the effect of the order of alone and group brainstorming and the fact that some studies \ufb01 nd no effect of order ( Paulus , Larey , & Ortega , 1995 ; Rotter & Portugal , 1969 ; Taylor , Berry , & Block , 1958 ) , it would seem that the most important consideration might be simply the importance of alternating alone and group brainstorming rather than focusing on the bene \ufb01 ts of one particularorder . However , only asmall number of studies have assessed this possibility and each of these have methodological limitations . A number of studies used different procedures for generating ideas in dif - ferent conditions . Rotter and Portugal ( 1969 ) compared nominal and interactive groups with those that worked both alone and as a group ( in two different orders ) . All of the participants wrote their ideas on a sheet of paper . The participants in the alone condition performed best , with the alternating condition leading to better performance than the groupcondition . However , thealonesessionsweredoneonlywithwrit - ing whereas the group session also involved discussion of the written ideas that took time away from idea generation . Several EBS studies have used a form of asynchronous brainstorming group in which mem - bers submit their ideas at different points in time . That is , ideas are sub - mittedindividuallybuttoacentrallocationorforumthatisaccessibleto all members of the group . This paradigm should allow group members to follow their train of thought without being disrupted , while also being able to gain stimulation from other members ' ideas . Dornburg , Stevens , Hendrickson , and Davidson ( 2009 ) examined asynchronous brainstorming over the course of four days using a large EBS group that was compared to individual brainstorming . There was no differ - ence in the number of ideas generated between the individuals and the group . Individuals outperformed the group in originality , feasibility and effectiveness . But the duration of brainstorming was not controlled , and the study only involved one interactive group of 30 . Ocker , Fjermestad , Hiltz , and Johnson ( 1998 ) compared FTF brainstorming with two differenttypes of EBS procedures ( synchronous andasynchro - nous ) and a combination group . However , the synchronous groups worked atthesame timein the same room , while members of the asyn - chronous groups worked from different locations at different times . In the combination group , participants \ufb01 rst worked FTF and later worked asynchronously . All groups received two weeks to work on the given task . The amount of time allowed for communication among group members in the asynchronous conditions was not controlled . The com - bined groups generated signi \ufb01 cantly more creative solutions and a bet - ter quality solution than the asynchronous , the synchronous , and the FTF groups . Some studies did not have alone or nominal control conditions , Girotra , Terwiesch , and Ulrich ( 2010 ) compared a \u201c hybrid \u201d condition to a group condition . Participants in the hybrid condition were asked to write their ideas individually for the \ufb01 rst 10 min and then share and discuss the ideas as a group ( FTF ) for the remaining time . Partici - pants using the hybrid process generated three times more ideas than those in the real groups . Since the hybrid condition used both BWr and FTF techniques , while the real group only used FTF , it is possible that the bene \ufb01 t of the hybrid condition may simply be due to a differ - ence in techniques across conditions . De Vreede and Reiter - Palmon ( 2010 ) examined the asynchronous brainstorming process using multiple groups . Large EBS groups were comprised of subgroups that either completed the entire brainstorming process on their own ( parallel mode ) or built on the work provided by the previoussubgroups ( serial mode ) . They found that serial processing wasbettersuitedfortasksthatrequirein - depthprocessingandelabora - tion , butparallelprocessingwasappropriatefortasksthatdemandmul - tiple new ideas . However , there was no individual control group to determine the relative effectiveness of group interaction . Although there is evidence that using a hybrid procedure may have some bene \ufb01 t over traditional group brainstorming , the results are not clear because of lack of control groups and lack of comparability of pro - cedures . In thepresentseries of experiments thesame technique ( BWr ) was used in all conditions . Alone and group BWr conditions were com - pared to hybrid conditions that alternated alone and group brainstorm - ing sessions . This design allowed us to determine to what extent the alternating process enhanced the number of ideas relative to the alone and group conditions . Eight - minute sessions were employed since 178 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 prior unpublished research has suggested that bene \ufb01 ts of hybrid brain - storming can be observed with this type of timing ( Korde , 2012 ) . BWr was used since we felt that this procedure was more likely than the FTF and EBS ones to demonstrate a bene \ufb01 t of the hybrid procedure be - cause of the more ef \ufb01 cient group exchange process . There are several demonstrations of such small BWr groups exceeding the performance of nominal groups ( Paulus et al . , 2015 ; Paulus & Yang , 2000 ) . However , small EBS groups of size b 8 tend to show production losses relative to same size nominal groups . These differences in outcome for EBS and BWr groups may be due to that fact that a collaborative stimulation ef - fect can occur only if participants pay attention to the shared ideas ( Dugosh et al . , 2000 ; Paulus & Yang , 2000 ) . In the typical EBS programs there is no way to be sure that participants are reading all the ideas pre - sented to them . In the BWr procedure used in the present experiments , reading shared ideas is part of the process . Recent research has shown that BWrcan producemore ideasthanEBSwhenparticipants are forced to make an effort to read the ideas ( Michinov , 2012 ) . However , there has to be reasonable balance between time spent reading others ideas and generating one ' s own ideas ( Brown & Paulus , 2002 ) . It is expected that hybrid brainstorming will be more effective than both group and alone brainstorming , but this could occur for a variety of reasons . There are least two theoretical perspectives , which suggest that individual ideation will be elevated after group ideation . The cogni - tive models suggest that the cognitive stimulation from group ideation sessions can yield associations that can be the basis for enhanced idea generation in a subsequent solitary ideation session . In fact , Brown and Paulus ( 2002 ) reported simulations based on their cognitive matrix model ( cf . , Brown , Tumeo , Larey , & Paulus , 1998 ) that demonstrated spikes in performance in an alone phase that followed a group phase . A production blocking / constraint perspective suggests that perfor - mance in a group session will be inhibited because the time taken to at - tend to others ' ideas will take away from one ' s own idea generation time . But , in the alone session performance is no longer constrained bythis demand , andthis mayprovidesomeextra motivationtotakead - vantage of this new freedom to generate ideas at will ( Amabile , 1996 ) . We are not aware of any demonstrations of such effects , but there is a literaturesuggestingthatvariouscognitiveandtaskconstraintscanmo - tivate individual creative behavior ( Medeiros , Partlow , & Mumford , 2014 ; Stokes , 2005 ) . However , this research has not addressed the pos - sible positive effects of eliminating the constraint on subsequent creativity . In contrast to the above discussion , a social comparison perspective suggeststhattheperformanceinthealonephasesmaybecriticalforen - hancing performance in the subsequent group phases ( Paulus & Dzindolet , 1993 ) . The typically higher rate of performance in the alone session may become a reference point for participants in a subsequent group session and result in an elevated level of performance relative to a condition in which participants share ideas only as a group . PaulusandDzindolet ( 1993 ) foundthatperformanceinsubsequentses - sions is positively correlated with the performance rate in earlier ses - sions ( Studies 3 & 4 ) . Thus those who started with a task that yielded a higher rate of idea generation performed better on a subsequent dif - ferent idea generation task than participants who started off with a lower rate task . This type of pace entrainment effect was stronger for group ideation than individual ideation , consistent with the entrain - mentperspectiveofKelly ( 1988 ) aswellasasocialcomparisonperspec - tive . Thus according to a social comparison perspective , a preliminary alone session may enhance the performance of a subsequent group ses - sion and this pattern may be entrained during the sessions that follow . In a preliminary experiment we examined the potential bene \ufb01 ts of hybrid ideation and the impact of starting a session with alone or group ideation ( see Korde , 2014 and supplementary materials for de - tails ) . The experiment employed a brainwriting paradigm for 4 eight - minute sessions , either all alone or in an alone - group - alone - group ( AGAG ) or a group - alone - group - alone ( GAGA ) sequence . It was found that the AGAG sequence led to more ideas than the other two conditions . The performance in the alone phases was elevated relative to the other phases , but only the second alone phase in the AGAG con - dition was signi \ufb01 cantly different from same phase in the alone condi - tion . Thus this study provides some support for the perspectives which suggest that the bene \ufb01 t of hybrid ideation will be primarily due to enhancement of performance in the alone phases . However , a possi - ble sampling bias in the AGAG condition and the fact that hybrid condi - tion participants received additional training may have in \ufb02 uenced the results . Thus we conducted two similar experiments to more de \ufb01 nitive - ly assess the effects of hybrid ideation . For these experiments all condi - tions , measures , and exclusions are reported . 2 . Experiment 1 Experiment 1 included the three conditions of the preliminary study and a group condition , with four phases in each condition . Participants either switched back and forth from group brainstorming to working alone or generated ideas alone or as a group for the entire session . There were two hybrid conditions depending on which phase was \ufb01 rst \u2013 alone or group . The four conditions were 1 ) alone - group - alone - group ( AGAG ) 2 ) group - alone - group - alone ( GAGA ) , 3 ) alone - alone - alone - alone ( Alone ) and 4 ) group - group - group - group . Each phase ( alone or group ) within the conditions lasted for 8 min and the total brainstormingsessionforeachconditionwas32minlong . Inallthecon - ditions , participants performed in groups of three . However , in the alone condition they performed the task individually but in the pres - ence of other members of their group . Based on previously discussed research and theory , the hybrid con - ditions should lead to more ideas than the nominal and group conditions . H1 . Participants in the hybrid conditions will generate more ideas than those in the alone and group conditions . Prior researchmight lead one to expect differences between the two different hybrid conditions . Given the \ufb01 ndings that either sequence can be bene \ufb01 cial , there is not a strong empirical basis for making a differen - tial prediction . However , both the cognitive and blocking / constraint perspectives suggest that a condition that begins with a group session may have an advantage . However , the social comparison and entrain - ment perspectives suggest that the condition that begins with an alone session will lead to the better performance . Furthermore , the cog - nitive and blocking / constraint perspectives both suggest that the alone sessions that follow the group sessions will demonstrate elevations in performancein comparison tothe alonecondition . In contrast the social comparison and entrainment perspectives suggest that a group session that follows an alone session will be elevated relative to the group condition . H2a . In accord with the cognitive and blocking / constraint perspectives , the hybrid condition that begins with a group session will lead to a bet - ter performance than one that begins with an alone session since group sessions should be followed by elevations in the performance during alone sessions . H2b . In line with the social comparison and entrainment perspectives the hybrid condition that begins with an alone session should lead to the best performance since this will lead to elevations in the perfor - mance in the subsequent phases and elevations in performance the group sessions that follow the alone sessions . H3a . In accord with the cognitive and constraint perspectives the alone sessions that follow group sessions in both of the hybrid conditions should show enhanced performance relative to the comparable phases in the alone condition . H3b . In accord with the social comparison and entrainment perspec - tives the group sessions that follow an alone session in both of the 179 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 hybrid conditions should show enhanced performance relative to the comparable phases in the group condition . A blocking / constraint perspective would suggest that the more constrained the performance is in the group phase in terms of number of ideas , the greater will be the subsequent enhancement in the alone phase . This could be examined by examining the extent to which the performance in the \ufb01 rst group phase of the hybrid condition is lower than the prior alone phase . The greater the discrepancy , the greater should be the experience of constraint and the subsequent increased motivation to perform in the alone session that follows the group session . H4 . If blocking / constraint plays a role in the elevation of the perfor - mance of the alone phases after the group phases in the hybrid condi - tion , the performance in the second alone session will be positively related to the degree of constraint experienced in the group phase in comparison to the prior alone phase . Studies that have used BWr techniques have obtained mixed results asto the bene \ufb01 t of sharingideas in a group . Paulus and Yang ( 2000 ) and Coskun ( 2005 ) both found that real groups generated more ideas than nominal groups when the BWr technique was used . However , thenom - inal groups generated ideas on a single sheet of paper , and the interacting groups generated ideas on slips of paper . A study by Goldenberg et al . ( 2013 ) found that when participants in the nominal condition are also asked to write their ideas on separate slips of paper , the difference in number of ideas generated by the interacting groups and nominal groups disappears . Since the BWr technique involved sep - arate slips in all conditions , it was anticipated that there would be no difference between the alone and group condition . However , Paulus et al . ( 2015 ) found that group BWr led to more ideas than alone writing foremployeesof atechnologycompanywhenall participants usedindi - vidual slips . This result was only marginally signi \ufb01 cant with the rela - tively small sample . Thus given the mixed \ufb01 ndings in past studies , we did not make a directional prediction for alone and group brainwriting . 2 . 1 . Method 2 . 1 . 1 . Participants Participants were undergraduate students from a large university in theSouthwest U . S . whoparticipated in thestudy in order to ful \ufb01 llintro - ductorypsychologyclassrequirements . 189students participatedinthe study , and data from nine participants were not used as they failed to follow instructions correctly . Six of these participants belonged to the Alone condition and three to the Group condition . Participants were randomly assigned to one of the four experimental conditions , each consisting of 15 groups of three . Data from 45 individual participants were pooled together to create 15 nominal groups . In our previously published studies on brainwriting , we have found effects related to alone and group sequence with 10 or 12 groups per condition ( Paulus et al . , 2015 ; Paulus & Yang , 2000 ) . All data were collected prior to our analyses . The average age of the participants was 20 . 16 years ( SD = 3 . 45 ) . There were 110 females and 70 males . 2 . 1 . 2 . Design and procedure The experiment used a 4 ( Condition ) \u00d7 4 ( Phases ) mixed design . In - formed consent was obtained from the participants before providing them with the instruction packet . Participants were provided instruc - tions about the task , the rules of brainstorming ( Osborn , 1957 ) , and the procedure to be used . All the participants wrote their ideas on col - ored slips of paper and were instructed to write only one idea per slip . Each participant received a different set of colored slips ( pink , white , or yellow ) . In the group phases , participants were seated at a table , fac - ing each other and were asked to pass the ideas to the person on their right and were told to read the ideas that they would receive from their left . Once the participants received their own ideas back , they were asked to place them at the center of the table . During the alone phases , participants were asked to write their ideas and simply place the slip of paper next to them without passing it on . At the end of each phase the slips were collected and placed into separate envelopes and labeled by group and phase . The participants were given different colors of pen for each phase . The colors of pens used were constant across condition and were always used in the same order ( blue , red , purple , black ) regardless of phase type ( alone or group ) . Participants in all of the conditions were given a group practice session with the \u201c al - ternate uses of a paper clip \u201d topic , even though those in the Alone con - dition would be performing alone , to ensure that all conditions had similar training . For the brainstorming session , participants were asked to generate ideas to the \u201c thumbs problem \u201d\u2014 the implications of having an extra thumb on one ' s hand . At the end of the session , partic - ipants completed a questionnaire about the task and their performance . The number of non - redundant ideas for each phase was used as the dependent variable . This was the only performance measure obtained in this experiment . We also obtained responses to a post - experimental questionnaire . This included items that asked participants to rate their performance in the different phases . However , since only the hybrid condition involved different phases , the items were not comparable across the conditionsand thus werenot analyzed . In thesubsequent ex - periment , all of the other measures obtained and all of the conditions employed are reported as well as any exclusion of participants . 2 . 2 . Results The data were analyzed at the group level using a 4 ( Condition ) \u00d7 4 ( Phases ) mixed ANOVA ( refer to Table 1 for descriptive statistics ) . The results showed that there was a signi \ufb01 cant main effect of condition , F ( 3 , 58 ) = 2 . 854 , p = 0 . 045 , \u03b7 p2 = 0 . 129 . Participants in the AGAG con - ditiongeneratedmore ideasthanthoseinthegroupcondition , although this difference was marginally signi \ufb01 cant ( p = 0 . 056 ) . The other condi - tions were not signi \ufb01 cantly different from one another . Thus , there was only partial support for H1 and no support for H2a or H2b . There was also a signi \ufb01 cant main effect of phases , F ( 3 , 174 ) = 99 . 365 , p b 0 . 001 , \u03b7 p2 = 0 . 631 . Signi \ufb01 cantly more ideas were generated in Phase 1 across all conditions as compared to phases 2 , 3 and 4 ( p b 0 . 001 ) . A signi \ufb01 cant interaction effect wasalso observed , F ( 9 , 174 ) = 13 . 008 , p b 0 . 001 , \u03b7 p 2 = 0 . 402 . This is primarily due to the elevation of the performance in the alone sessions of the hybrid conditions that followed the group sessions in support of H3a . ( see Fig . 1 ) . Pairwise comparisons using the Bonferroni adjustment indicated that quantity was signi \ufb01 cantly higher in Phase 1 of the AGAG condition as compared to the GAGA ( mean dif - ference = 10 . 113 , SE = 3 . 247 , p = 0 . 017 ) and Group conditions ( mean difference = 11 . 008 , SE = 3 . 247 , p = 0 . 008 ) . For phase 2 , the perfor - mance in the GAGA condition was signi \ufb01 cantly higher than the AGAG ( meandifference = 8 . 804 , SE = 3 . 145 , p = 0 . 042 ) andGroup ( meandif - ference = 12 . 312 , SE = 3 . 094 , p = 0 . 001 ) conditions . In phase 3 , signif - icantly more ideas were generated in the AGAG condition as compared to the GAGA ( mean difference = 11 . 479 , SE = 3 . 417 , p = 0 . 008 ) and Group ( mean difference = 12 . 729 , SE = 3 . 417 , p = 0 . 003 ) conditions . There were signi \ufb01 cantly more ideas in the phase 4 of the GAGA condi - tion as compared to the same phase of the Group condition ( mean dif - ference = 10 . 938 , SE = 2 . 979 , p = 0 . 003 ) . More ideas were generated in phase 3 of the AGAG condition as compared to the same phase of the Alone condition , but this difference was only marginally signi \ufb01 cant , t ( 58 ) = 2 . 592 , p = 0 . 072 . A speci \ufb01 c planned comparison of the com - bined quantity in the alone phases ( Phases 2 and 4 ) of the GAGA condi - tion was signi \ufb01 cantly higher than the combined quantity of the corresponding phases in the Alone condition F ( 2 , 58 ) = 4 . 860 , p = 0 . 011 , \u03b7 p2 = 0 . 144 , insupportofH3a . Contrarytotheblocking / constraint perspective ( H4 ) , the extent to which performance was lower in the \ufb01 rst group phase of the AGAG condition relative performance in the prior alone phase was negatively related to performance in the 180 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 subsequentalone phase , b = \u2212 0 . 383 ( CI \u2212 0 . 748 , \u2212 0 . 017 ) , SE = 0 . 169 , t ( 13 ) = \u2212 2 . 261 , p = 0 . 042 . 2 . 3 . Discussion As in the preliminary study , there was an elevation in performance during the alone phases that followed the group phases of the hybrid conditions . This pattern of results supports the prediction made on the basis the cognitive / associational models of collaborative ideation ( Brown & Paulus , 2002 ; Nijstad & Stroebe , 2006 ) . However , it could also re \ufb02 ect the enhancement of idea generation after constraint due to production blocking . As a result of the elevated performance during these phases the hybrid conditions performed better than the other two conditions , but these differences were not statistically signi \ufb01 cant . The results did not support the social comparison perspective since there was not an elevation of performance in the group phases after the alone phases in the hybrid conditions in comparison to the compa - rable phases in the group condition . The group condition did not generate signi \ufb01 cantly more ideas than the alone condition . This is consistent with the \ufb01 ndings of Goldenberg et al . ( 2013 ) who also found no differences in quantity when slips of paper were used to write the ideas instead of a single sheet of paper . This result may seem to indicate that generating ideas in a group has no bene \ufb01 t as compared to generating ideas individually . When partici - pants are working in a group , their attention and time is divided be - tween generating their own ideas and reading others ' ideas . This may lead to distraction and some production loss ( Santanen , Briggs , & de Vreede , 2004 ; Valacich , Dennis , & Connolly , 1994 ) . In fact , in the \ufb01 rst alone phase of the AGAG condition was signi \ufb01 cantly higher than the \ufb01 rst group phases in the Group and GAGA condition . However , this re - sult does not necessarily indicate a lack of stimulation in the group phases . The effect of stimulation from group ideation can be seen more clearly when comparing within the conditions rather than across the conditions . For the hybrid conditions , the data clearly show an in - crease in ideas in the alone phases after the group phases , whereas the number of ideas in the nominal condition does not increase at any point during the session . Thus , this may re \ufb02 ect the bene \ufb01 t of group stimulation ( Dugosh et al . , 2000 ; Paulus & Brown , 2007 ; Paulus & Yang , 2000 ) and this bene \ufb01 t is only visible after the group phase . The key is to allow the participants to work individually after a group phase so that they may utilize the group derived stimulation more effectively . Another interesting \ufb01 nding from Experiment 1 is that the number of ideas generated by participants in the alone condition drops consider - ably from phase 1 to phase 2 . But after phase 2 , the rate of idea genera - tion stays almost constant until the end . Prior research with verbal brainstorming has found a more consistent decline in the performance of the alone participants over time so that the difference in quantity of ideas between the nominal and interacting groups almost disappears ( Paulus & Dzindolet , 1993 , Experiment5 ) . More of adeclineinthenum - ber of ideas was expected from the nominals in Experiment 1 because they received no ideas from other participants in the 32 - minute long session . It is possible that participants continued to generate ideas due to the motivating effects of the social cues present . When participants generated ideas and kept the slips to their side , the stacks of slips ( but not the ideas ) were visible to the other participants . Previous research has shown that this social comparison of performance can lead to an in - crease in the number of ideas generated ( Dugosh & Paulus , 2005 ; Paulus , Larey , Putman , Leggett , & Roland , 1996 ) . If these participants were generating ideas without the social cues ( i . e . , without being able to see each other ) , their rate of ideation may have shown a steeper drop . If that is true , the decline should be evident in all of the alone phases regardless of condition \u2013 alone or hybrid . Isolating the group members for this phase in the hybrid condition may help determine if the previously seen increase in quantity of the alonephases that follow - ed the group phase was purely due to stimulation they received from the group phase or was in \ufb02 uenced by the effect of social cues from the alone phase . If participants in the hybrid condition are able to utilize the stimulation from the group phase , quantity in the alone phases of the AGAG condition should not drop as drastically as it might in the alone condition . The hybrid condition was designed to mimic the way in which idea - tion often occurs in the real world . Group meetings can occur a few times a week with time in between for individuals to generate more ideas and discuss them at the next meeting and so on . However , these individual brainstorming phases occur when the individual is away from other group members . The individual phases in Experiment 1 oc - curred in full view of the other group members . Therefore , it may be in - structive to examine the effect of the hybrid condition when the group members are separated or unable to view each other ' s performance during the alone phases . In Experiment 2 social cues were eliminated in the alone condition and alone phases of the hybrid condition , and we added another eight - minute phase to lengthen the session to enhance the chances of seeing a decline in the alone condition . This could result in the group condition matching or exceeding the performance of the alone condi - tion since alone participants could begin to run out of ideas without ex - ternal stimulation ( Dennis et al . , 2005 ; Nijstad , Stroebe , & Lodewijkx , 1999 ) . In Experiment 1 we only measured the number of ideas generat - ed since that has been the main measure of choice in this type of re - search . However , in the second experiment we added additional performance measures to obtain a richer set of data on the effects of hy - brid brainwriting and to allow for discrimination between the cognitive and blocking / constraint perspectives . Table 1 Experiment 1 : means and standard deviations of quantity by conditions and phases . Phase Alone AGAG GAGA Group Overall Phase 1 41 . 267 ( 11 . 931 ) 45 . 133 ( 9 . 257 ) 35 . 000 ( 7 . 874 ) 34 . 125 ( 6 . 386 ) 38 . 742 ( 9 . 924 ) Phase 2 27 . 267 ( 9 . 924 ) 26 . 133 ( 11 . 370 ) 34 . 938 ( 8 . 029 ) 22 . 625 ( 4 . 365 ) 27 . 774 ( 9 . 693 ) Phase 3 26 . 667 ( 10 . 661 ) 35 . 667 ( 12 . 715 ) 24 . 188 ( 8 . 727 ) 22 . 938 ( 4 . 057 ) 27 . 242 ( 10 . 527 ) Phase 4 24 . 133 ( 10 . 302 ) 22 . 333 ( 8 . 558 ) 30 . 438 ( 8 . 981 ) 19 . 500 ( 5 . 151 ) 24 . 129 ( 9 . 186 ) Overall 119 . 333 ( 39 . 776 ) 129 . 267 ( 35 . 692 ) 124 . 563 ( 29 . 003 ) 99 . 188 ( 15 . 285 ) \u2013 Note . Standard deviations ( SD ) are listed inparentheses . There were 15 groups in each condition . 15 20 25 30 35 40 45 50 Phase 1 Phase 2 Phase 3 Phase 4 U n i qu e I d e a s AGAG GAGA ALONE GROUP Fig . 1 . Number of unique of ideas generated in the different conditions across phases in Experiment 1 . 181 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 3 . Experiment 2 Experiment 2 compared a hybrid and a group condition to an alone condition , and group members were isolated in the alone condition as well as all alone phases . The overall procedure remained similar to the prior two experiments . The session time was increased from 32 min to 40 min by adding a \ufb01 fth phase . To prevent the participants from seeing each other , partitions were placed between participants during all the individual phases , across all the conditions . So only the group phases occurred in full view off all the group members without the partition . Since the participants would be isolated during the individual phases to eliminate both cogni - tive and social cues , it was expected that fewer ideas would be generat - ed in the alone condition as compared to the hybrid condition . The hybrid condition would receive some stimulation from the group phases , whereas the alone condition would not . The \ufb01 rst two hypotheses are similar to those of the prior experiments . H1 . More ideas will be generated in the hybrid condition as compared to the alone and group conditions . H2 . The alone phases that come after the group exchange in the hybrid condition will demonstrate an elevated level of idea generation com - pared to the same phases in the alone condition . For thisexperimentwe also developed hypotheses for theadditional measures of the novelty of ideas , their variety , and the depth of catego - ries explored . The number of categories ( variety ) is often but not neces - sarily correlated with number of ideas ( Deuja et al . , 2014 ) , average novelty is typically fairly independent of the number of ideas ( Paulus et al . , 2011 ) , but depth of categories ( number of ideas per category ) can be related both to enhanced novelty and quantity ( cf . , Nijstad , De Dreu , Rietzschel , & Baas , 2010 ; Rietzschel , Nijstad , & Stroebe , 2007 ) . For example , a number of studies have found that a focusingon onecat - egoryatatimeasanindividualoragroupenhances thenumber ofideas generated ( Baruah & Paulus , 2011 ; Coskun , Paulus , Brown , & Sherwood , 2000 ; Deuja et al . , 2014 ) . The dual pathway model of Nijstad et al . ( 2010 ) suggests that idea - tional creativity ( novelty ) involves both persistence ( category depth ) and \ufb02 exibility ( category variety ) pathways . 1 Both persistence and \ufb02 ex - ibility are important when generating ideas in a group . Since the hybrid condition has both individual and group phases , these factors become important in understanding cognitive processes involved in the hybrid condition . Persistence , thatis , focusingonacategoryanddiggingdeeper may be more feasiblewhen workingalone . The stimulation fromshared ideaswhenworkinginagroup should increasethe \ufb02 exibility ornumber of categories explored in the subsequent alone phases . Prior , research has typically found that as the number of ideas generated increases so do thenumberof categories ( e . g . , Kohn & Smith , 2010 ) . Based on there - sultsfromExperiment1 , theDeDreu , Nijstad , Baas , Wolsink , andRoskes ( 2012 ) \ufb01 ndings , and the Nijstad et al . ( 2010 ) dual pathway model , the following hypotheses are proposed : H3 . The hybrid condition should demonstrate enhanced depth and va - riety because . of the bene \ufb01 t of both group stimulation and private re \ufb02 ection time . H4 . The hybrid condition will have the highest novelty since partici - pants should beable toutilizeboth \ufb02 exibility and persistence pathways . The exposure to the ideas of others in the group phase of the hybrid condition should increase the awareness of a broader range of catego - ries and increase the extent to which participants generate ideas in a broader range of categories in the subsequent alone phase . H5 . The alone phases that follow the group sessions in the hybrid con - dition will show more category variety ( \ufb02 exibility ) than the corre - sponding phases of the alone condition . The generation of ideas is also related to the extent to which to which categories are tapped more deeply ( Baruah & Paulus , 2011 ; Deuja et al . , 2014 ) . H6a . We expect that the ideas in the alone phases that follow group sessions in the hybrid conditions will demonstrate greater depth than the same phases in the alone condition . Persistence or focusing on a category and digging deeper may be more feasible when working alone since the train of thought is not disrupted by exposure to the ideas of others . H6b . The alone phases will show more category depth than the group phases , regardless of condition . We will also analyze the extent to which the performance in one phase predicts performance in a subsequent phase . For the alone condi - tion intersession correlations should primarily re \ufb02 ect individual differ - ences in ability or motivation . These factors also should play a role in the other conditions . However , in the group condition the tendency for social comparison and normative in \ufb02 uence may lead to even stron - ger correlations among the sessions as in the Paulus and Dzindolet ( 1993 ) study . In the hybrid conditions , the correlations may be some - what lower since the contiguous sessions involve a change in the idea - tion process . Of primary interest will be the patterns in the hybrid condition and their theoretical implications . We would expect the quantity , variety , depth and novelty of the group phase to in \ufb02 uence thesamemeasures inthealonephasebecause of related cognitive stim - ulation and the carry - over of social in \ufb02 uence and entrainment effects from one phase to the other . If the enhanced performance in the alone phases that follow group phases were due to various cognitive process - es , one would expect a number of other relationships . The greater the variety of categories ( \ufb02 exibility ) experienced in the group phase , the better should be the performance in the alone phases because the cate - gories can prime new semantic domains ( e . g . , Coskun et al . , 2000 ) . The greater category depth in the group phase may increase the novelty of the ideas in the subsequent alone phase as well the number of ideas . Category depth has been associated increased novelty in alone and dy - adic conditions ( Nijstad et al . , 2010 ; Rietzschel et al . , 2007 ) . However , category depth has also been associated with an increase in the number of ideas generated since focusing on one idea at a time is an ef \ufb01 cient cognitive strategy ( Deuja et al . , 2014 ) . The novelty of ideas in the group phase may not enhance the number of ideas generated in the subsequent alone session since more common ideas may yield more as - sociations ( Dugosh & Paulus , 2005 ) . H7 . The performance in the alone phases after the group phases will be predicted by the quantity , variety and depth of ideas in the prior group phases . The blocking / constraint perspective is not well suited for making predictions about the different measures of performance since it in - volves a non - speci \ufb01 c motivational process . However , as suggested for Experiment 1 , the degree of constraint experienced in the \ufb01 rst group phase of AGAG condition could be related to enhanced performance in the following group phase . This did not occur in Experiment 1 , but we again assessed this possibility in Experiment 2 . 1 DeDreuetal . ( 2012 ) assessedtheserelationshipsusingindividualbrainstormingand the role of working memory . In experiment 3 , the effects of working memory capacity were tested on the alone , group , and hybrid conditions . Working memory was assessed using three different measures \u2013 OSPAN , RSPAN and WCST . Several mediation models and moderated regression analyses were conducted to examine the effects based on the dualpathwaymodel . TheresultsshowedthatonlylowerscoresonOSPAN ( lowerworking memory capacity ) signi \ufb01 cantly predicted higher novelty in the hybrid condition via the \ufb02 exibilitypathway . Scoresonpersonalityvariableslikeopennesstoexperience , extraver - sion , andperspectivetakingalsodidnothavesigni \ufb01 canteffectsonquantity , varietydepth , or novelty ( see Korde , 2014 ) . 182 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 H8 . If blocking / constraint plays a role , the performance in the second alone session will be positively related to the degree of constraint expe - rienced in the group phase in comparison to the prior alone phase . 3 . 1 . Method 3 . 1 . 1 . Participants ParticipantswererecruitedfromalargeuniversityintheSouthwest - ern U . S . and all participants received research credits to help complete requirements for their courses . If a complete group of three students ar - rived for the session , they were randomly assigned to either the group or hybrid brainstorming conditions . If fewer than three people arrived for the session , they were automatically assigned to the individual brainstorming condition . Data from three individual participants were pooled together to form a nominal group . A total of 270 students partic - ipated in the idea generation and data from all the participants were used for the analyses . The number of groups per condition was doubled in number compared to Experiment 1 since we anticipated needing ad - ditional power for planned path analyses . All of the data were gathered before the analyses of the data were begun . The mean age was 20 . 03 ( SD = 3 . 664 ) ; the youngest participant was 17 and the oldest was 62 years of age . 54 . 1 % of the sample consisted of females . 30 % of the sample was Hispanic , 28 . 5 % were Caucasian , 21 . 1 % Asian , 15 . 2 % African American , and the remaining 1 % comprised of Native Americans , Paci \ufb01 c Islanders and those who reported more than one race . Only one partic - ipant did not disclose their race . 3 . 1 . 2 . Design and procedure The experiment used a 3 ( Condition ) \u00d7 5 ( Phases ) mixed design . Participants were recruited to come tothe labingroups of threetocom - plete the brainstorming task . Similar to the prior experiments , brainwriting was the procedure of choice and slips of paper were used to write ideas . Each participant received a different set of colored of slips ( pink , white , or yellow ) to easily identify as their own . The ideas were written using a different colored pen for each phase ( blue , red , purple , black , and orange ) . Participants in all the conditions received a three - minute group practice session and topic was \u201c alternate uses of a paper clip \u201d . The thumbs problem was used again for the brainstorming task . Participants sat at a table , facing each other . A partition was placed betweenthoseparticipants thatwere starting the \ufb01 rst phaseindividual - ly . Those in the alone condition continued the remaining phases with the partitions in place , stopping every 8 min for a change of pens . Each phase lasted for 8 min just as the previous experiments . The parti - tionwasremovedfortheparticipantsinthehybridconditionjustbefore every group phase . All the phases for the group condition took place without the partition . At the end of the session the participants were asked to complete a post - experiment questionnaire and were debriefed . 3 . 1 . 3 . Dependent variables The dependent variables obtained for this experiment were quanti - ty , novelty , category depth , and category variety of the ideas . The scores on all of these variables were calculated for each phase within each group . The operational de \ufb01 nitions of these variables are as follows : \u2022 Quantity : Total number of non - redundant ideas . Ideas that were re - peated across phases and within a phase were considered as redun - dant and were not counted . Ideas were counted and checked for redundancy by two trained raters . Both raters counted all the ideas for all the sessions and removed redundant ideas by consensus . \u2022 Novelty : Novelty was determined based on the infrequency of the ideas . Ideas that were produced frequently by many participants were considered less novel . Whereas ideas that were rarely produced by participants were rated as novel . Two trained raters rated novelty of the ideas independently . The ideas were rated on a scale of 1 to 5 , where one meant \u201c very common \u201d and \ufb01 ve stood for \u201c very uncom - mon \u201d . \u2022 Categoryvariety ( Flexibility ) : Ideaswerealsosortedintodifferentcat - egories . Using previous data , 24 categories were developed for the thumbs problem ( Korde , 2014 ) . Some of those categories were so - cial / discrimination , sports , gestures , etc . The total number of catego - ries explored by each participant and by the group was recorded . \u2022 Category depth ( Persistence ) : For each category that was explored by aparticipant , thetotalnumberof ideasgeneratedwithinthatcategory andtheaveragenumberof ideaspercategorywascalculatedbydivid - ing quantity for the phase by the corresponding category variety score . An overall category depth score was also calculated for each group . As inthecase ofExperiment1 , thequestionnaire responses were not analyzed since the items were not designed to tap the processes rele - vant to the hybrid condition . All of the other measures obtained in this experiment and all conditions employed are reported as well as any ex - clusion of participants . 3 . 1 . 4 . Coding of ideas Data from the brainstorming session were screened to eliminate re - dundant ideas by consensus . The ideas were coded for novelty and cat - egories by two independent raters . Both raters coded the ideas blind to the conditions and phases . Each rater coded 25 % of the ideas for novelty on a scale of oneto \ufb01 ve as described earlier . Scoresthat were within one point of each other were considered as indicators of inter - rater agree - ment and the ratings were adjusted for analysis ( Diehl & Stroebe , 1991 ; Kohn , Paulus , & Choi , 2011 ) . Inter - rater reliability for the novelty ratings on 25 % of the data was calculated using ICCs resulting in a value of 0 . 834 . One rater then coded the remainder of the data for novelty . Bothratersalsocoded 25 % of thedata for categories . Theratersinde - pendently assigned one of 24 categories to each idea . Cohen ' s Kappa was used to test inter - rater reliability and the resulting value was 0 . 808 . The second rater coded the remaining 75 % of data for categories . To calculate category variety , the number of non - redundant categories explored within each phase was counted . Additionally , the number of non - redundant categories explored throughout the entire session was counted . Therefore , each group had six separate category variety scores \u2013 one for each phase and one overall for the entire session . Category depth was computed by dividing the number of non - redundant ideas by category variety for the corresponding phase and / or session . 3 . 2 . Results The data were analyzed at the group level . Examination of the corre - lations among the variables for the three conditions combined showed that all four measures were signi \ufb01 cantly correlated with one another . Quantity was signi \ufb01 cantly correlated with the remaining three mea - sures ( variety : 0 . 509 , p b 0 . 001 ; depth : 0 . 957 , p b 0 . 001 ; novelty : 0 . 494 , p b 0 . 001 ) . Variety was signi \ufb01 cantly correlated with depth ( 0 . 251 , p = 0 . 017 ) and novelty ( 0 . 407 , p b 0 . 001 ) , and depth was signif - icantly correlated with novelty ( 0 . 427 , p b 0 . 001 ) . Even though quantity and depth were highly correlated , given their distinct theoretical impli - cations , we included both of these measures in our various analyses . 3 . 2 . 1 . Quantity Four separate 3 ( Condition ) \u00d7 5 ( Phases ) mixed ANOVAs were used to test the differences among conditions and phases on each of the four dependent variables \u2013 quantity , novelty , category variety , and category depth ( refer to Table 2 for descriptive statistics ) . The ANOVA for quanti - ty yielded a signi \ufb01 cant main effect of condition , F ( 2 , 87 ) = 3 . 430 , p = 0 . 037 , \u03b7 p 2 = 0 . 073 . Pairwise comparisons using the Bonferroni adjust - ment indicated that the hybrid condition had the highest mean number of ideas ( M = 37 . 407 , SE = 1 . 962 ) , followed by the alone condition 183 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 ( M = 34 . 980 , SE = 1 . 962 ) , and then the group condition ( M = 30 . 260 , SE = 1 . 962 ) . However , the hybrid condition was only signi \ufb01 cantly dif - ferent from the group condition , p = 0 . 035 . The other differences be - tween conditions were not signi \ufb01 cant . H1 was only partially supported . There was also a main effect of phases , F ( 4 , 348 ) = 86 . 179 , p b 0 . 001 , \u03b7 p 2 = 0 . 498 . As expected , the most ideas were generated in the \ufb01 rst phase and quantity declined as the phases progressed . Phase 1wassigni \ufb01 cantlydifferentthanalltheotherphases ( p b 0 . 001 ) . Thein - teraction between condition and phasewasalso signi \ufb01 cant , F ( 8 , 348 ) = 8 . 829 , p b 0 . 001 , \u03b7 p 2 = 0 . 169 ( see Fig . 2 ) . Signi \ufb01 cantlymore ideaswere in generated intheAGAG conditionascompared tothe Group conditionin phase 1 ( mean difference = 9 . 433 , SE = 3 . 000 , p = 0 . 007 ) , phase 3 ( mean difference = 11 . 500 , SE = 2 . 964 , p = 0 . 001 ) and phase 5 ( mean difference = 12 . 067 , SE = 3 . 423 , p = 0 . 002 ) . The within condi - tion comparisons showed enhanced performance in the alone phases after the group phases in the hybrid condition ( see Fig . 2 ) . The com - binedquantityofphase3andphase5wassigni \ufb01 cantlyhigherinthehy - brid condition than the alone condition , F ( 1 , 87 ) = 4 . 590 , p = 0 . 035 , \u03b7 p2 = 0 . 050 . These results support H2 . Quantity in the Group condition was highest in phase 1 ( phase 2 : mean difference = 6 . 0 33 , SE = 1 . 297 , p b 0 . 001 ; phase 3 : mean difference = 9 . 033 , SE = 1 . 360 , p b 0 . 001 ; phase 4 : mean difference = 8 . 800 , SE = 1 . 547 , p b 0 . 001 ; phase 5 : mean difference = 12 . 167 , SE = 1 . 749 , p b 0 . 001 ) . Quantity in phase 2 was also signi \ufb01 cantly higher than the quantity in phase 5 ( mean difference = 6 . 133 , SE = 1 . 609 , p = 0 . 003 ) . 3 . 2 . 2 . Categories Testing for differences in the number of categories explored or cate - gory variety yielded a main effect of condition , F ( 2 , 87 ) = 7 . 849 , p = 0 . 001 , \u03b7 p2 = 0 . 153 . The group condition explored fewer categories than the alone ( mean difference = \u2212 1 . 413 , SE = 0 . 466 , p = 0 . 010 ) and hy - brid conditions ( mean difference = \u2212 1 . 733 , SE = 0 . 466 , p = 0 . 001 ) . However , since the alone and hybrid conditions did not differ signi \ufb01 - cantly from each other , there was only partial support for H3 . There was a main effect of phases , F ( 4 , 348 ) = 19 . 641 , p b 0 . 001 , \u03b7 p2 = 0 . 184 ; category variety decreased from phase 1 to phase 5 . Variety in phase 1 was signi \ufb01 cantly higher than variety in the remaining phases ( Phase 2 : p = 0 . 003 ; Phase 3 , 4 , and 5 : p b 0 . 001 ) . The interaction effect of conditions and phases was signi \ufb01 cant , F ( 8 , 348 ) = 2 . 846 , p = 0 . 004 , \u03b7 p2 = 0 . 061 in support of H5 ( see Fig . 3 ) . Even though category variety Table 2 Experiment 2 : means and standard deviations of quantity , novelty , category variety and category depth by conditions and phases . Condition Quantity ( SD ) Novelty ( SD ) Variety ( SD ) Depth ( SD ) Alone Phase 1 44 . 433 ( 10 . 190 ) 2 . 356 ( 0 . 265 ) 14 . 967 ( 2 . 008 ) 2 . 965 ( 0 . 532 ) Phase 2 35 . 500 ( 10 . 582 ) 2 . 727 ( 0 . 315 ) 13 . 900 ( 2 . 369 ) 2 . 550 ( 0 . 633 ) Phase 3 34 . 633 ( 11 . 263 ) 2 . 777 ( 0 . 258 ) 13 . 067 ( 2 . 728 ) 2 . 646 ( 0 . 687 ) Phase 4 30 . 800 ( 11 . 223 ) 2 . 920 ( 0 . 292 ) 12 . 067 ( 2 . 392 ) 2 . 551 ( 0 . 702 ) Phase 5 29 . 533 ( 12 . 586 ) 2 . 911 ( 0 . 298 ) 11 . 900 ( 2 . 695 ) 2 . 470 ( 0 . 894 ) Overall 174 . 900 ( 49 . 700 ) 2 . 738 ( 0 . 209 ) 21 . 100 ( 1 . 845 ) 8 . 235 ( 2 . 072 ) AGAGA Phase 1 46 . 900 ( 13 . 422 ) 2 . 250 ( 0 . 216 ) 14 . 667 ( 2 . 249 ) 3 . 179 ( 0 . 636 ) Phase 2 33 . 567 ( 9 . 964 ) 2 . 556 ( 0 . 311 ) 13 . 533 ( 2 . 161 ) 2 . 467 ( 0 . 580 ) Phase 3 39 . 933 ( 11 . 756 ) 2 . 893 ( 0 . 343 ) 13 . 967 ( 2 . 646 ) 2 . 857 ( 0 . 654 ) Phase 4 29 . 267 ( 11 . 298 ) 3 . 009 ( 0 . 441 ) 11 . 900 ( 2 . 551 ) 2 . 440 ( 0 . 750 ) Phase 5 37 . 367 ( 14 . 829 ) 3 . 229 ( 0 . 970 ) 13 . 433 ( 2 . 208 ) 2 . 764 ( 0 . 920 ) Overall 187 . 033 ( 55 . 333 ) 2 . 787 ( 0 . 367 ) 21 . 233 ( 1 . 832 ) 8 . 770 ( 2 . 263 ) Group Phase 1 37 . 467 ( 11 . 004 ) 2 . 193 ( 0 . 240 ) 12 . 700 ( 2 . 037 ) 2 . 928 ( 0 . 653 ) Phase 2 31 . 433 ( 11 . 464 ) 2 . 762 ( 0 . 376 ) 12 . 000 ( 3 . 118 ) 2 . 645 ( 0 . 826 ) Phase 3 28 . 433 ( 11 . 416 ) 2 . 929 ( 0 . 387 ) 11 . 333 ( 2 . 412 ) 2 . 465 ( 0 . 752 ) Phase 4 28 . 667 ( 12 . 408 ) 3 . 079 ( 0 . 358 ) 11 . 500 ( 2 . 446 ) 2 . 459 ( 0 . 838 ) Phase 5 25 . 300 ( 12 . 200 ) 3 . 152 ( 0 . 464 ) 11 . 300 ( 2 . 842 ) 2 . 211 ( 0 . 978 ) Overall 151 . 300 ( 56 . 079 ) 2 . 823 ( 0 . 306 ) 20 . 067 ( 2 . 132 ) 7 . 504 ( 2 . 520 ) Phase 1s combined 43 . 202 ( 11 . 968 ) 2 . 269 ( 0 . 248 ) 14 . 157 ( 2 . 281 ) 3 . 036 ( 0 . 604 ) Phase 2s combined 33 . 685 ( 10 . 613 ) 2 . 687 ( 0 . 117 ) 13 . 157 ( 2 . 696 ) 2 . 567 ( 0 . 677 ) Phase 3s combined 34 . 539 ( 12 . 208 ) 2 . 871 ( 0 . 335 ) 12 . 843 ( 2 . 763 ) 2 . 663 ( 0 . 710 ) Phase 4s combined 29 . 753 ( 11 . 504 ) 3 . 004 ( 0 . 372 ) 11 . 865 ( 2 . 427 ) 2 . 491 ( 0 . 758 ) Phase 5s combined 30 . 899 ( 14 . 028 ) 3 . 100 ( 0 . 655 ) 12 . 225 ( 2 . 733 ) 2 . 493 ( 0 . 948 ) Alone phases 37 . 387 ( 6 . 128 ) 2 . 758 ( 0 . 319 ) 13 . 496 ( 1 . 115 ) 2 . 748 ( 0 . 242 ) Group phases 30 . 591 ( 3 . 978 ) 2 . 811 ( 0 . 339 ) 12 . 038 ( 0 . 819 ) 2 . 516 ( 0 . 221 ) Note . Standard deviations ( SD ) are listed inparentheses . There were 30 groups in each condition . 15 20 25 30 35 40 45 50 Phase 1 Phase 2 Phase 3 Phase 4 Phase 5 U n i q u e I d e a s Alone AGAGA Group Fig . 2 . Number of unique of ideas generated in the different conditions across phases in Experiment 2 . 8 10 12 14 16 18 20 Phase 1 Phase 2 Phase 3 Phase 4 Phase 5 C a t e g o r y V a r i e t y Alone Hybrid Group Fig . 3 . Number of categories explored in the different conditions across phases in Experiment 2 . 184 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 declined over time , the pattern was different in the hybrid condition as compared to the alone and group conditions ( see Fig . 3 ) . There was a sharp increase in category variety for the hybrid condition in the third and \ufb01 fth phase . These were the individual brainstorming sessions that occurred right after group brainstorming suggesting that they were ex - posedtonewerormorecategoriesduringthegroupsession . Thedepen - dent t - test yielded a signi \ufb01 cant difference among these phases , t ( 29 ) = 4 . 050 , p b 0 . 001 . Combined variety from phase3 and phase5 wassignif - icantly higher in the hybrid condition than the alone condition , F ( 1 , 87 ) = 4 . 697 , p = 0 . 033 , \u03b7 p2 = 0 . 051 , in support of H5 . 3 . 3 . Category depth The analysis for category depth revealed no main effect of condition , F ( 2 , 87 ) = 0 . 790 , p = 0 . 457 , \u03b7 p2 = 0 . 018 , contrary to H3 . However , there was a main effect of phases , F ( 4 , 348 ) = 20 . 668 , p b 0 . 001 , \u03b7 p2 = 0 . 192 . The category depth decreased from one phase to the next as the session progressed . Category depthin phase1 wassigni \ufb01 cantly higher than cat - egory depth in all the other phases ( p b 0 . 001 ) . The other differences were not signi \ufb01 cant . The interaction between phase and condition on category depth was signi \ufb01 cant , F ( 8 , 348 ) = 3 . 201 , p = 0 . 002 , \u03b7 p2 = 0 . 069 ( see Fig . 4 ) . Similar to the pattern of interaction seen for category variety and consistent with H6a , category depth increased in Phase 3 and Phase 5 of the hybrid condition , after the group brainstorming phase . However , a planned contrast of the combined depth in Phase 3 and Phase 5 showed that those phases in the hybrid condition were not signi \ufb01 cantly different from those in the alone condition , F ( 1 , 87 ) = 1 . 709 , p = 0 . 195 , \u03b7 p2 = 0 . 019 . It was also suggested that category depth could be greater in the alone phases as compared to the group phases , regardless of condition ( H6b ) . Two t - tests were used to test this hypothesis \u2013 one independent t - test and one dependent t - test . The independent t - test compared the alonecondition to the groupcondition , whilethedependent t - test com - paredthealonephasestogroupphaseswithinthehybridcondition . The independent t - test showed that there was no signi \ufb01 cant difference in category depth between the alone and group conditions , t ( 58 ) = 1 . 228 , p = 0 . 224 . The dependent t - test for the hybrid condition showed that the alone phases had signi \ufb01 cantly higher category depth than the group phases , t ( 29 ) = 5 . 776 , p b 0 . 001 . Therefore , H6b was partially supported . Categorydepthwashigherinthealonephase , butonlywith - inthehybridcondition . H3was alsopartiallysupportedsincethehybrid conditiondemonstrated signi \ufb01 cantlyhighervariety butnot asigni \ufb01 cant increase in category depth as compared to the alone condition . 3 . 4 . Novelty AnANOVA tested fordifferences innovelty of theideasacrosscondi - tions and phases . There was no signi \ufb01 cant effect of condition , F ( 2 , 87 ) = 0 . 604 , p = 0 . 549 , \u03b7 p2 = 0 . 014 . Thus H4 was not supported . How - ever , there was a main effect of phases , F ( 4 , 348 ) = 102 . 329 , p b 0 . 001 , \u03b7 p2 = 0 . 540 . Novelty of the ideas increased as the session progressed , leading to highest novelty in the fourth and \ufb01 fth phases . Novelty in Phase 5 was signi \ufb01 cantly higher than novelty in Phase 1 and Phase 2 ( p b 0 . 001 ) , and Phase 3 ( p = 0 . 001 ) . There was an interaction effect of condition and phase on novelty of the ideas , F ( 8 , 348 ) = 3 . 972 , p b 0 . 001 , \u03b7 p2 = 0 . 084 ( see Table 2 ) . In the \ufb01 rst phase , participants in the alone condition generated ideas with signi \ufb01 cantly higher average novelty than participants in the group condition ( mean difference = 0 . 163 , SE = 0 . 062 , p = 0 . 032 ) . By the \ufb01 fth phase , the hybrid and group conditions showed higher novelty than the alone condition , con - sistent with \ufb01 ndings of Baruah and Paulus ( 2016 ) . However , the com - parison was not signi \ufb01 cant ( hybrid \u2013 p = 0 . 176 ; group \u2013 p = 0 . 450 ) . None of the other comparisons were signi \ufb01 cant . 3 . 5 . Mediation model It was expected that the hybrid condition would have the highest novelty since participants would be able to use both \ufb02 exibility and per - sistence ( H4 ) . Based on the previously discussed results of the ANOVA , there was no signi \ufb01 cant main effect of condition on novelty . Even so , the mediation model was examined to test the second part of the hy - pothesis . Since the independent variable ( condition ) was a categorical variable with three levels , two dummy codes ( k - 1 ) were created and the alone condition was used as a reference group ( Hayes & Preacher , 2014 ) . The \ufb01 rst dummy code ( d 1 ) indicated the hybrid condition and the second ( d 2 ) indicated the group condition . Category depth and vari - ety were centered on the mean before they were added to the model . The \u201c create a new estimand \u201d command in AMOS was used to produce each of the speci \ufb01 c indirect effects and the total indirect effect for each dummy coded condition . The chi - square for the base model ( Fig . 5 ) was signi \ufb01 cant , \u03a7 2 ( 3 ) = 12 . 846 , p = 0 . 005 . The GFI ( 0 . 947 ) and AGFI ( 0 . 735 ) were below the cut off , as was the CFI ( 0 . 856 ) . The RMSEA was 0 . 192 ( CI : 0 . 093 , 0 . 305 ) and signi \ufb01 cantly different than zero ( p = 0 . 013 ) . Based on thesevalues the base model was not consid - ered a good \ufb01 t to the data . The modi \ufb01 cation indices suggested that a path was needed from d 2 to novelty , i . e . the direct effect of the group condition on novelty . Theoretically , condition can predict novelty ; but there is no reason to expect that only the group condition should have an effect on novelty . A second model was tested after adding direct paths from both indicator variables for the condition to novelty , and the chi - square was still signi \ufb01 cant , \u03a7 2 ( 1 ) = 3 . 927 , p = 0 . 048 ( Fig . 6 ) . The GFI ( 0 . 983 ) was above the cut off , but the AGFI was not ( 0 . 745 ) . The CFI ( 0 . 957 ) was above the cut off but the RMSEA ( 0 . 181 , CI : 0 . 016 , 0 . 384 ) was marginally signi \ufb01 cant ( p = 0 . 073 ) . The results showed that both models were not a very good \ufb01 t , but the second model had slight better values based on the \ufb01 t indices . Additionally , the AIC of the second model with the direct paths was lower ( 31 . 927 ) than the AIC of the base model ( 36 . 846 ) . Therefore , parameter estimates were examined from the second model . Category depth ( b = 0 . 050 , CI : 0 . 026 , 0 . 068 , SE = 0 . 010 , p = 0 . 001 ) and category variety ( b = 0 . 057 , CI : 0 . 037 , 0 . 081 , SE = 0 . 011 , p b 0 . 001 ) both predicted novelty signi \ufb01 - cantly . The a paths , total paths , direct paths , and indirect paths are de - scribed separately for each indicator variable . 3 . 5 . 1 . Hybrid condition The hybrid condition did not signi \ufb01 cantly predict category depth ( b = 0 . 535 , CI : \u2212 0 . 594 , 1 . 609 , SE = 0 . 561 , p = 0 . 348 ) or variety ( b = 0 . 133 , CI : \u2212 0 . 724 , 1 . 110 , SE = 0 . 470 , p = 0 . 774 ) . The relative total effect of the hybrid condition on novelty was not signi \ufb01 cant ( b = 0 . 049 , CI : \u2212 0 . 090 , 0 . 207 , SE = 0 . 076 , p = 0 . 482 ) , but it was in the expected direction . A one - unit increase ( in this case from the alone condition to the hybrid condition ) was related to an increase in novelty but the relationship was not signi \ufb01 cant . The relative direct ef - fect ( b = 0 . 015 , CI : \u2212 0 . 103 , 0 . 170 , SE = 0 . 069 , p = 0 . 784 ) was also 1 1 . 5 2 2 . 5 3 3 . 5 4 Phase 1 Phase 2 Phase 3 Phase 4 Phase 5 C a t e g o r y D e p t h Alone Hybrid Group Fig . 4 . Number of ideas exploredper category inthe differentconditionsacrossphases in Experiment 2 . 185 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 not signi \ufb01 cant . The relative speci \ufb01 c indirect effects of the hybrid condi - tion ( viadepth : b = 0 . 027 , CI : \u2212 0 . 027 , 0 . 088 , SE = 0 . 028 , p = 0 . 297 ; via category : b = 0 . 008 , CI : \u2212 0 . 047 , 0 . 061 , SE = 0 . 027 , p = 0 . 761 ) and the relative total indirect effect were also not signi \ufb01 cant ( b = 0 . 034 , CI : \u2212 0 . 053 , 0 . 122 , SE = 0 . 044 , p = 0 . 432 ) . The results did not support the 4th hypothesis \u2013 the hybrid condition did not signi \ufb01 cantly predict novelty via the persistence and \ufb02 exibility pathways . 3 . 5 . 2 . Group condition The group did not signi \ufb01 cantly predict category depth ( b = \u2212 0 . 731 , CI : \u2212 1 . 947 , 0 . 435 , SE = 0 . 602 , p = 0 . 220 ) , butitdidpredictcategoryva - riety ( b = \u2212 1 . 033 , CI : - 2 . 014 , \u2212 0 . 012 , SE = 0 . 515 , p = 0 . 048 ) . The rel - ative total effect was not signi \ufb01 cant ( b = 0 . 085 , CI : \u2212 0 . 048 , 0 . 221 , SE = 0 . 068 , p = 0 . 203 ) but the relative direct effect was signi \ufb01 cant ( b = 0 . 181 , CI : 0 . 079 , 0 . 291 , SE = 0 . 053 , p b 0 . 001 ) . A one - unit change ( from the alone condition to the group condition ) signi \ufb01 cantly predict - ed higher novelty when controlling for category depth and category va - riety . The relative speci \ufb01 c indirect effect via depth was not signi \ufb01 cant ( b = \u2212 0 . 036 , CI : \u2212 0 . 112 , 0 . 017 , SE = 0 . 032 , p = 0 . 184 ) , but it was sig - ni \ufb01 cant via variety ( b = \u2212 0 . 059 , CI : \u2212 0 . 128 , \u2212 0 . 003 , SE = 0 . 031 , p = 0 . 039 ) . The relative total indirect effect was also signi \ufb01 cant ( b = \u2212 0 . 096 , CI : \u2212 0 . 192 , \u2212 0 . 004 , SE = 0 . 048 , p = 0 . 041 ) . As com - pared to the control condition , the group condition predicted novelty as a result of the negative effect of the group condition on category variety . 3 . 6 . Additional analyses We also examined the within session correlations to determine whether there were unique patterns in the hybrid condition . Of partic - ular interest is how various measures of performance in the second and fourth group phases predictthe performance in thefollowingaloneses - sions . These results are presented in Tables 3 and 4 . In addition we per - formed regression analyses to determine the relation of the performance in the second and fourth sessions to the \ufb01 nal session ( Table 5 ) . In general the correlations were the strongest for the group condition . For the hybrid condition signi \ufb01 cant correlations and regres - sion effects occurred primarily for the quantity and depth predictors of the four variables , providing partial support for H7 . To examine the potential impact of constraint , we regressed the dif - ference between the \ufb01 rst alone and group session in the hybrid condi - tion on the performance in the third alone phase . The bigger the gap between performance in the alone and subsequent group phase ( lower performance ) , the greater the number of ideas generated in the next phase , b = 0 . 528 , SE = 0 . 232 , t = 2 . 273 , p = 0 . 031 . This same comparison for the other two conditions was not signi \ufb01 cant ( alone condition , b = \u2212 0 . 505 , SE = 0 . 312 , t = \u2212 1 . 616 , p = 0 . 117 ; Fig . 5 . Effect of condition on novelty using the dual - pathway model . The model shows the effect of the hybrid and group conditions as compared to the alone condition . Fig . 6 . Effectofconditiononnoveltyusingthedual - pathwaymodelwithdirectpaths . Themodelshowstheeffectofthehybridandgroupconditionsascomparedtothealonecondition . 186 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 group condition , b = \u2212 0 . 217 , SE = 0 . 408 , t = \u2212 0 . 530 , p = 0 . 600 ) . These results provide some evidence that the performance constraint experienced in the group phase of the hybrid condition motivated in - creased performance in the subsequent alone phase ( H8 ) . 4 . Discussion The results of Experiment 2 are consistent with those of Experiment 1 . The hybridcondition wasthebest in termsof number of ideasbut not signi \ufb01 cantlybetter thanthealonecondition . The mainbene \ufb01 t in thehy - brid condition appears in the alone phases after exposure to ideas of others . This pattern occurred for number of ideas , number of categories and depth of categories but was only signi \ufb01 cant for the within phase analyses in the case of the latter . In terms of number of ideasgenerated , thehybrid condition wassig - ni \ufb01 cantly better than thegroupcondition in the experiments in whicha group condition was employed . However , to demonstrate a synergistic bene \ufb01 t of hybrid ideation the hybrid condition needs to perform better than the alone condition . The trends in the three experiments ( includ - ing the preliminary study ) were in support of such an outcome . A meta - analysis was conducted to examine the difference in the hybrid andaloneconditions acrossthethreeexperiments . Themeansandstan - dard deviations were used to conduct the meta - analysis ( Borenstein , Hedges , Higgins , & Rothstein , 2005 ) . Across the three experiments the total samplesizefor thetwoconditionswas114 . The \ufb01 xedeffects statis - tics revealed a signi \ufb01 cant mean effect of 0 . 402 ( SE = 0 . 187 , 95 % CI = 0 . 036 / 0 . 768 ) , z = 2 . 154 , p = 0 . 013 . The degree of heterogeneity was not signi \ufb01 cant , Q ( 2 ) = 3 . 257 , p = 0 . 196 , indicating that the data from the three experiments were similar to each other . Since the bene \ufb01 cial impact was most evident in the alone phases after group interaction , we also did a meta - analytic comparison across the three experiments for the performance in the third alone phase of the AGAG conditions versus thealoneconditions . The \ufb01 xed effect statisticsshowed a main ef - fect of 0 . 6994 ( SE = 0 . 191 , 95 % CI = 0 . 326 / 1 . 072 ) , z = 3 . 672 , p = 0 . 0002 , providing strong support for the performance enhancement after group idea exchange . Thus a consistent pattern of results has been obtained across the three experiments in support of the hybrid ideation and its synergistic bene \ufb01 t relative to the alone condition . The bene \ufb01 t of hybrid ideation and the elevated performance in the alone phases of the hybrid conditions are consistent with the predic - tions based on the association models of group ideation ( Brown & Paulus , 2002 ; Nijstad & Stroebe , 2006 ) . This perspective suggests that group ideation generates associations that can be the basis for the gen - eration of additional ideas . Only some of these can be expressed in the group session , so the alone session provides an opportunity for the \u201c spreading activation \u201d of the associations to in \ufb02 uence the idea genera - tion in the alone session . The analyses of the relationships between the sessions within the three conditions indicated that the performance was most consistent over time for the group condition , with all four measures of perfor - mance in one phase predicting the performance on all four measures in the subsequent phase in most of the comparisons . This is consistent with the entrainment ( Kelly , 1988 ; Kelly & Karau , 1993 ) and social in - \ufb02 uence perspective ( Paulus & Dzindolet , 1993 ) in that performance Table 3 Experiment 2 . Phase 2 group performance correlations with Phase 3 alone performance . Conditions Alone Hybrid Group Phase 2 quantity Quantity 0 . 828 \u204e\u204e\u204e 0 . 731 \u204e\u204e\u204e 0 . 912 \u204e\u204e\u204e Novelty 0 . 568 \u204e\u204e\u204e 0 . 260 0 . 510 \u204e\u204e Flexibility 0 . 539 \u204e\u204e 0 . 431 \u204e 0 . 634 \u204e\u204e\u204e Depth 0 . 656 \u204e\u204e 0 . 551 \u204e\u204e 0 . 814 \u204e\u204e\u204e Phase 2 novelty Quantity 0 . 305 0 . 050 0 . 717 \u204e\u204e\u204e Novelty 0 . 457 \u204e 0 . 524 \u204e\u204e 0 . 679 \u204e\u204e\u204e Flexibility 0 . 116 0 . 332 0 . 606 \u204e\u204e\u204e Depth 0 . 269 \u2212 0 . 162 0 . 531 \u204e\u204e\u204e Phase 2 variety Quantity 0 . 366 \u204e 0 . 429 \u204e 0 . 559 \u204e\u204e Novelty 0 . 057 0 . 382 0 . 424 \u204e Flexibility 0 . 599 \u204e\u204e\u204e 0 . 269 0 . 481 \u204e\u204e Depth 0 . 054 0 . 314 0 . 453 \u204e Phase 2 depth Quantity 0 . 764 \u204e\u204e\u204e 0 . 618 \u204e\u204e\u204e 0 . 598 \u204e\u204e\u204e Novelty 0 . 642 \u204e\u204e\u204e 0 . 083 0 . 277 Flexibility 0 . 277 0 . 324 0 . 348 Depth 0 . 761 \u204e\u204e\u204e 0 . 494 \u204e\u204e 0 . 595 \u204e\u204e\u204e \u204e\u204e\u204e p b 0 . 001 . \u204e\u204e p b 0 . 01 . \u204e p b 0 . 05 . Table 4 Experiment 2 : Phase 4 group performance correlations with Phase 5 alone performance . Conditions Alone Hybrid Group Phase 4 quantity Quantity 0 . 921 \u204e\u204e\u204e 0 . 700 \u204e\u204e\u204e 0 . 953 \u204e\u204e\u204e Novelty 0 . 466 \u204e\u204e 0 . 099 0 . 625 \u204e\u204e\u204e Flexibility 0 . 348 0 . 290 0 . 661 \u204e\u204e\u204e Depth 0 . 858 \u204e\u204e\u204e 0 . 682 \u204e\u204e\u204e 0 . 732 \u204e\u204e\u204e Phase 4 novelty Quantity 0 . 558 \u204e\u204e 0 . 518 \u204e\u204e 0 . 595 \u204e\u204e\u204e Novelty 0 . 504 \u204e\u204e 0 . 515 \u204e\u204e 0 . 612 \u204e\u204e\u204e Flexibility 0 . 066 0 . 285 0 . 405 \u204e Depth 0 . 699 \u204e\u204e\u204e 0 . 456 \u204e 0 . 426 Phase 4 variety Quantity 0 . 447 \u204e 0 . 491 \u204e\u204e 0 . 543 \u204e\u204e Novelty 0 . 229 \u2212 0 . 085 0 . 292 Flexibility 0 . 167 0 . 198 0 . 568 \u204e\u204e\u204e Depth 0 . 394 \u204e 0 . 463 \u204e\u204e 0 . 326 Phase 4 depth Quantity 0 . 809 \u204e\u204e\u204e 0 . 545 \u204e\u204e 0 . 850 \u204e\u204e\u204e Novelty 0 . 430 \u204e 0 . 071 0 . 573 \u204e\u204e Flexibility 0 . 319 0 . 221 0 . 517 \u204e\u204e Depth 0 . 765 \u204e\u204e\u204e 0 . 549 \u204e\u204e 0 . 709 \u204e\u204e\u204e \u204e\u204e\u204e p b 0 . 001 . \u204e\u204e p b 0 . 01 . \u204e p b 0 . 05 . Table 5 Experiment 2 : Regression results for the relationship of performance in the two alone phases on the \ufb01 nal group phase for the hybrid condition . Conditions Alone Hybrid Group Ph2 & Ph4 quantity Ph5 quantity 78 . 139 \u204e\u204e\u204e 14 . 649 \u204e\u204e\u204e 143 . 522 \u204e\u204e\u204e Ph5 variety 2 . 157 1 . 58 10 . 613 \u204e\u204e\u204e Ph5 novelty 4 . 509 \u204e\u204e 0 . 581 10 . 366 \u204e\u204e\u204e Ph5 depth 41 . 760 \u204e\u204e\u204e 15 . 483 \u204e\u204e\u204e 17 . 520 \u204e\u204e\u204e Ph2 & Ph4 variety Ph5 quantity 4 . 157 \u204e 7 . 243 \u204e 8 . 727 \u204e Ph5 variety 0 . 643 0 . 563 10 . 790 \u204e\u204e\u204e Ph5 novelty 0 . 754 0 . 785 1 . 615 Ph5 depth 3 . 093 8 . 630 \u204e 2 . 001 Ph2 & Ph4 novelty Ph5 quantity 9 . 442 \u204e\u204e 5 . 729 \u204e 14 . 022 \u204e\u204e\u204e Ph5 variety 1 . 202 1 . 224 4 . 031 \u204e Ph5 novelty 4 . 883 \u204e 8 . 340 \u204e 15 . 254 \u204e\u204e\u204e Ph5 depth 11 . 912 \u204e\u204e\u204e 4 . 144 \u204e 5 . 078 \u204e Ph2 & Ph4 depth Ph5 quantity 35 . 905 \u204e\u204e\u204e 8 . 296 \u204e 35 . 672 \u204e\u204e\u204e Ph5 variety 1 . 594 0 . 805 6 . 307 Ph5 novelty 3 . 058 0 . 077 7 . 413 \u204e Ph5 depth 32 . 301 \u204e\u204e\u204e 7 . 956 \u204e 14 . 343 \u204e\u204e\u204e \u204e\u204e\u204e p b 0 . 001 . \u204e\u204e p b 0 . 01 . \u204e p b 0 . 05 . 187 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 becomes more consistent or normative over time in the social context than in solitary sessions . In the hybrid condition the quantity and depthmeasuresinthegroupphasesweretheprimarypredictorsofsub - sequent performance in the subsequent alone phases on all four mea - sures . The number of ideas being shared and the extent to which people dig deeply into speci \ufb01 c categories has been associated with in - creased creativity in groups in a number of studies ( e . g . , Dugosh & Paulus , 2005 ; Rietzschel et al . , 2007 ) . It now appears that these factors alsoaffecttheability tocontinuegeneratingideasinsubsequentsolitary sessions , in support of the cognitive models of group creativity ( Nijstad & Stroebe , 2006 ; Paulus & Brown , 2007 ) . Our research was framed in part in relation to the dual process the - ory of creativity ( Nijstad et al . , 2010 ) , and some of our \ufb01 ndings were consistent that that theory , such as the correlations among \ufb02 exibility , depth and novelty . However , the path models for the hybrid condition did not \ufb01 t with this particularperspective for predictingnovelty . The el - evation in performance in the hybrid condition re \ufb02 ected both enhance - ment of \ufb02 exibility and depth , but this effect does not require an explicit dual process of \ufb02 exibility and persistence . The semantic network model of Brown and Paulus ( 2002 ) has predicted this type of outcome based on a category transition matrix model ( Brown , et al . , 1998 ) . Thus indi - viduals in a group are exposed to ideas and categories of ideas , which will in turn in \ufb02 uence the subsequent ideation process . In generating ideas there is a tendency to stick in the same category until it has been somewhat depleted before moving on to other categories . So this model predicts that group exposure can increase both divergent ( \ufb02 exi - bility ) and convergentthinking ( depth ) . This model has been supported in a number of simulations of existing data ( Brown et al . , 1998 ; Coskun et al . , 2000 ) , and several simulations found evidence of a strong spike in performance of an alone session that followed a group session , includ - ing an increase in the number of categories in comparison to an alone only condition ( Brown & Paulus , 2002 ) . There was some support for the blocking constraint perspective in the hybrid condition as well . In contrast to the \ufb01 ndings of Experiment 1 , the more that the performance in the group phase was lower than that in the prior alone phase , the better was the performance in the alone phase that followed the group phase . These cases should have ex - perienced the most sense of constraint , which may have increased the motivation level when they were again \u201c free \u201d in the third session . How - ever , since the overall high performance of the second and fourth group phases was related to increased performance in the subsequent alone phases , constraint cannot be the sole factor . This pattern \ufb01 ts most clear - ly with the cognitive perspectives . However , our study was not de - signed a priori to clearly differentiate between the cognitive and constraint perspective . Additional research is required to evaluate the potential motivational effects of constraint in these types of paradigms . For example , the constraint effect mightbe observed even if theconsec - utive phases involve different problems . The cognitive perspective would not predict any elevations in performance in this case . Althoughwehavefoundenhancedperformanceinthehybridcondi - tion , the boundary conditionsforthis effectremain to be discovered . For example , how will it be affected by problems varying in dif \ufb01 culty ? The \u201c thumbs problem \u201d and other problems like generating ideas to improve a university are almost unlimited in the ideas that can be generated . They also have many different categories or types of ideas . Thus , ideas from others can be quite bene \ufb01 cial in suggesting new avenues for idea - tion . However , in more constrained problems with limited realistic op - tions ( such as generating ideas for reduced energy consumption ) , the potential stimulatingeffectof sharing ideaswithothers may be lessdra - matic . However , since one may quickly run of ideas on such a task , hav - ing input from others may be even more important . The length of thesession may also in \ufb02 uence the outcomes . As we in - dicated earlier , idea generation tends to slow down rather dramatically in a short period of time for most problems . Brief breaks appear to re - duce the pace of this decline ( Paulus , Nakui , Putman , & Brown , 2006 ) . The alone sessions served not only as breaks for the constraints of the demanding group exchange process but also provided an opportunity for unrestrained expression of ideas . Although there is a theoretical basis for predictingthepattern we observed , there are node \ufb01 nitive the - oretical bases for predicting the optimal length of the sessions . Shorter sessions may be too disruptive , but longer sessions may be too tiring . Having longer group sessions and shorter alone sessions may reduce the bene \ufb01 ts of the group - to - alone carry - over effects . However , if the alone sessions are too long relative to the group sessions , there may not be enough stimulation to have a strong impact on the alone sessions . The hybrid effect has implications for collaborative creativity in real world settings . It suggests that some balance of alone and group idea - tion will be ideal . Although we cannot provide precise suggestions as to the ideal length of the sessions , it is likely that having the alone ses - sion follow the group sessions immediately is best so as to not lose the associational potentialof thepriorgroupsession whichshould dissipate or decay over time ( Brown & Paulus , 2002 ) . EventhoughFTFsessions tend to reduce the numberof ideas generat - edingroupsrelativetootherwaysofsharingideas , thisisstillaverycom - monandpopularapproach . Thehybridapproachshouldalsobebene \ufb01 cial for enhancing the bene \ufb01 ts of FTF brainstorming . This could involve inter - spersing the verbal exchange sessions with ones in which participants generate ideas privately ( either in written form or electronically ) . Ideas generated in the alone sessions can be shared subsequently with the group and provide cues for additional ideas . This type of paradigm should increase the number of ideas generated relative to using only FTF brain - storming . However , for this type of paradigm to yield more ideas than nominal control condition may require longer sessions ( cf . , Nijstad et al . , 1999 ) in which those in the control conditions begin running out of ideas before those in the hybrid condition . Hybrid FTF sessions may also havesomeadvantagesoversolitarysessionsinotherareassuchasthede - velopment of group based skills ( Sutton & Hargadon , 1996 ) , facilitating subsequent decisions about which ideas to select for implementation , and producing higher levels of task enjoyment ( cf . , Paulus et al . , 1993 ) . The modi \ufb01 cations in Experiment 2 of reducing social cues in the alone condition and adding another phase did not appear to change the pattern of results among the conditions . Whether the participants in the alone condition work in full view of each other or in isolation doesnotseemtochangetheperformanceofthealoneconditionrelative to the other conditions ( in comparison to Experiment 1 ) . It was antici - patedthatprovidingextratimemightbene \ufb01 tthegroupandhybridcon - ditions relative to the alone condition . Nijstad et al . ( 1999 ) found that given enough time , group performance can achieve the level of individ - ual performance . That is , individuals and groups should start running out of ideas over time , but the cues provided in the group setting ( both the group and the hybrid conditions ) should enable them to con - tinue coming up with new ideas for a longer period of time . We did not \ufb01 nd results consistent with that expectation since the pattern of results in Experiment 2 wassimilarto that of Experiment1 . A substantially lon - gersessionmightberequireddemonstrateanenhancedperformanceof the group condition relative to the alone one and a stronger effect for the hybrid condition relative to the alone condition . There were no differences in novelty across the three conditions for Experiment 2 , but there was an effect of time . Average novelty of the ideas increased as the session progressed across all three conditions . These results are consistent with previous studies suggesting that the more common ideas are generated earlier in the session ( Kohn & Smith , 2010 ; Paulus , Kohn , Arditti , & Korde , 2013 ) . The interaction be - tween condition and time was also signi \ufb01 cant . The alone condition had the highest novelty at the beginning of the session and lowest nov - elty at the end of the session . But the patterns were different for the groupandhybrid conditions . Thus there is someevidencethatgroup in - teraction can enhance the novelty of idea generation , especially in the later phases of brainstorming sessions ( see also Baruah & Paulus , 2016 ) . The results from the mediation models echo the correlation and re - gressions patterns seen in Tables 3 - 5 . The correlations are stronger for 188 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 the group condition , and predictions based on the dual pathway model apply better to group condition when usingthe alone condition as com - parison . Although it was expected that participants in the hybrid condi - tion would be able to utilizeboth pathways to attain novelty , the results indicate that the dual pathway model may not apply as well to the hy - brid condition since this condition involves alternatingbetween solitary and group ideation and results in generally lower correlations in perfor - mance from one phase to the other . The \ufb01 ndings , however , highlight that engaging in hybrid brainstorming entails processes that are differ - entfromaloneorgroupbrainstorminganddonot \ufb01 twiththedualpath - way model but allow for higher levels of performance . The fact that the \ufb02 exibility and persistence pathways were not related to novelty in the hybrid condition is consistent with that conclusion . In sum , the experiments have demonstrated the bene \ufb01 ts of hybrid brainwriting compared to alone and group brainwriting . This bene \ufb01 t was re \ufb02 ected in number of ideas and the number of categories sur - veyed . These effects were mostly due to the elevated performance dur - ing the alone sessions that followed the group sessions in the hybrid conditions . Apparently , the sharing of ideas in the group phases en - hanced both the \ufb02 exibility and but not the persistence and novelty for individuals in the subsequent alone phases . This series of studies is the \ufb01 rst to demonstrate the synergistic bene \ufb01 t of alternating alone and group ideation . Our research suggests in both short - term sessions and longer - term team interactions , the effective integration of group and solitary ideation is important for effectively tappinggroup ' s creative po - tential . Although hybrid ideation did not increase the novelty of the ideas , past research has found that an increase in the number of ideas is related to an increase in the number of ideas that are both novel and feasible ( e . g . , Paulus et al . , 2011 ) . Thus this pool of good ideas can then be the basis for developing the best ideas for potential implemen - tation ( Kohn et al . , 2011 ) . Acknowledgements The research reported in this paper was supported by collaborative grants to the second author from the National Science Foundation ( CreativeIT 0855825and INSPIREBCS 1247971 ) . Any opinions , \ufb01 ndings , and conclusions or recommendations expressed in this material are those of the authors and do not necessarily re \ufb02 ect the views of the Na - tional Science Foundation . This paper is based in part on a dissertation by the \ufb01 rst author under the direction of the second author . We would like to thank Tyler Hunt and Jaclyn West for their assistance in this research . Appendix A . Supplementary data Supplementary data to this article can be found online at http : / / dx . doi . org / 10 . 1016 / j . jesp . 2016 . 11 . 002 . References Amabile , T . M . ( 1996 ) . Creativity in context . Boulder , CO : West - view Press . Baruah , J . , & Paulus , P . B . ( 2008 ) . Effects of training on idea generation . Small Group Research , 39 ( 5 ) , 523 \u2013 541 . http : / / dx . doi . org / 10 . 1016 / j . jesp . 2011 . 04 . 007 . Baruah , J . , & Paulus , P . B . ( 2011 ) . Category assignment and relatedness in the group idea - tion process . Journal of Experimental Social Psychology , 47 ( 6 ) , 1070 \u2013 1077 . http : / / dx . doi . org / 10 . 1016 / j . jesp . 2011 . 04 . 007 . Baruah , J . , & Paulus , P . B . ( 2016 ) . The role of time and category relatedness in electronic brainstorming . Small Group Research , 47 ( 3 ) , 333 \u2013 342 . Borenstein , M . , Hedges , L . , Higgins , J . , & Rothstein , H . ( 2005 ) . Comprehensivemeta - analysis version 2 . Englewood , NJ : Biostat , 104 . Bouchard , T . J . , Jr . , & Hare , M . ( 1970 ) . Size , performance , and potential in brainstorming groups . Journal of Applied Psychology , 54 ( 1 ) , 51 \u2013 55 . Brian , D . ( 1996 ) . Einstein : A life . New York : Wiley . Brown , V . R . , & Paulus , P . B . ( 2002 ) . Making group brainstorming more effective : Recom - mendations from an associative memory perspective . Current Directions in Psychological Science , 11 , 208 \u2013 212 . http : / / dx . doi . org / 10 . 1111 / 1467 - 8721 . 00202 . Brown , V . , Tumeo , M . , Larey , T . S . , & Paulus , P . B . ( 1998 ) . Modeling cognitive interactions during group brainstorming . Small Group Research , 29 , 495 \u2013 526 . Coskun , H . ( 2005 ) . Cognitive stimulation with convergent and divergent thinking exercises inbrainwriting : Incubation , sequencepriming , andgroupcontext . SmallGroupResearch , 36 ( 4 ) , 466 \u2013 498 http : / / dx . doi . org . ezproxy . uta . edu / 10 . 1177 / 1046496405276475 . Coskun , H . , Paulus , P . B . , Brown , V . , & Sherwood , J . J . ( 2000 ) . Cognitive stimulation and problem presentation in idea generation groups . Group Dynamics : Theory , Research , and Practice , 4 , 307 \u2013 329 . http : / / dx . doi . org / 10 . 1037 / 1089 - 2699 . 4 . 4 . 307 . DeDreu , C . K . W . , Nijstad , B . A . , Baas , M . , Wolsink , I . , & Roskes , M . ( 2012 ) . Working mem - ory bene \ufb01 ts creative insight , musical improvisation , and original ideation through maintained task - focused attention . Personality and Social Psychology Bulletin , 38 ( 5 ) , 656 \u2013 669 . http : / / dx . doi . org / 10 . 1177 / 0146167211435795 . De Vreede , G . , Briggs , R . O . , & Reiter - Palmon , R . ( 2010 ) . Exploring asynchronous brain - storming in large groups : A \ufb01 eld comparison of serial and parallel subgroups . Human Factors , 52 ( 2 ) , 189 \u2013 202 . http : / / dx . doi . org / 10 . 1177 / 0018720809354748 . Dennis , A . R . , & Williams , M . L . ( 2003 ) . In P . B . Paulus , & B . A . Nijstad ( Eds . ) , Electronic brainstorming : Theory , research , and future directions ( pp . 160 \u2013 178 ) . New York , NY , US : Oxford University Press . Dennis , A . R . , Pinsonneault , A . , Hilmer , K . , Barki , H . , Gallupe , B . , Huber , M . , & Beallavance , F . ( 2005 ) . Pattern in electronic brainstorming . International Journal of e - Collaboration , 1 ( 4 ) , 38 \u2013 57 . http : / / dx . doi . org / 10 . 4018 / jec . 2005100103 . Derosa , D . M . , Smith , C . L . , & Hantula , D . A . ( 2007 ) . Themediummatters : Miningthelong - promisedmerit of group interaction increativeideageneration tasks ina meta - anal - ysis of the electronic group brainstorming literature . Computers in Human Behavior , 23 ( 3 ) , 1549 \u2013 1581 . http : / / dx . doi . org / 10 . 1016 / j . chb . 2005 . 07 . 003 . Deuja , A . , Kohn , N . W . , Paulus , P . B . , & Korde , R . M . ( 2014 ) . Taking a broadperspective be - fore brainstorming . Group Dynamics ; Theory , Research , and Practice , 18 ( 3 ) , 222 \u2013 236 . http : / / dx . doi . org / 10 . 1037 / gdn0000008 . Diehl , M . , & Stroebe , W . ( 1987 ) . Productivitylossinbrainstorminggroups : Towardtheso - lution of a riddle . Journal of Personality and Social Psychology , 53 ( 3 ) , 497 \u2013 509 . http : / / dx . doi . org / 10 . 1037 / 0022 - 3514 . 53 . 3 . 497 . Diehl , M . , & Stroebe , W . ( 1991 ) . Productivity loss in idea - generating groups : Tracking down the blocking effect . Journal of Personality and Social Psychology , 61 ( 3 ) , 392 \u2013 403 . http : / / dx . doi . org / 10 . 1037 / 0022 - 3514 . 53 . 3 . 497 . Dornburg , C . C . , Stevens , S . M . , Hendrickson , S . M . L . , & Davidson , G . S . ( 2009 ) . Improving extreme - scale problem solving : Assessing electronic brainstorming effectiveness in an industrial setting . Human Factors , 51 , 519 \u2013 527 . http : / / dx . doi . org / 10 . 1177 / 0018720809343587 . Dugosh , K . L . , & Paulus , P . B . ( 2005 ) . Cognitive and social comparison processes in brain - storming . Journalof Experimental SocialPsychology , 41 , 313 \u2013 320 . http : / / dx . doi . org / 10 . 1016 / j . jesp . 2004 . 05 . 009 . Dugosh , K . L . , Paulus , P . B . , Roland , E . J . , & Yang , H . ( 2000 ) . Cognitive stimulation in brain - storming . Journal of Personality and Social Psychology , 79 , 722 \u2013 735 . http : / / dx . doi . org / 10 . 1037 / 0022 - 3514 . 79 . 5 . 722 . Dunnette , M . D . , Campbell , J . , & Jaastad , K . ( 1963 ) . The effect of group participation on brainstorming effectiveness for 2 industrial samples . Journal of Applied Psychology , 47 ( 1 ) , 30 \u2013 37 . http : / / dx . doi . org / 10 . 1037 / h0049218 . Girotra , K . , Terwiesch , C . , & Ulrich , K . T . ( 2010 ) . Ideagenerationandthequalityofthebest idea . Management Science , 56 ( 4 ) , 591 \u2013 605 . http : / / dx . doi . org / 10 . 1287 / mnsc . 1090 . 1144 . Goldenberg , O . , Larson , J . R . , & Wiley , J . ( 2013 ) . Goal instructions , response format , and idea generation in groups . Small Groups Research , 44 ( 3 ) , 227 \u2013 256 . http : / / dx . doi . org / 10 . 1177 / 1046496413486701 . Hayes , A . F . , & Preacher , K . J . ( 2014 ) . Statistical mediation analysiswith a multicategorical independent variable . British Journal of Mathematical and Statistical Psychology , 67 , 451 \u2013 470 . http : / / dx . doi . org / 10 . 1111 / bmsp . 12028 . Heslin , P . A . ( 2009 ) . Betterthanbrainstorming ? Potentialcontextualboundaryconditions to brainwiritng for idea generation in organizations . Journal of Occupational and Organizational Psychology , 82 ( 1 ) , 129 \u2013 145 . http : / / dx . doi . org / 10 . 1348 / 096317908X285642 . Kelly , J . R . ( 1988 ) . Entrainmentinindividualandgroupbehavior . InJ . E . McGrath ( Ed . ) , The social psychology of time : New perspectives . Thousand Oaks , CA : Sage PublicationsInc . Kelly , J . R . , & Karau , S . J . ( 1993 ) . Entrainment of creativity in small groups . Small Group Research , 24 , 179 \u2013 198 . http : / / dx . doi . org / 10 . 1177 / 1046496493242002 . Kerr , N . L . , & Bruun , S . E . ( 1983 ) . Dispensability of member effort and group motivation losses : Free - rider effects . Journal of Personality and Social Psychology , 44 , 78 \u2013 94 . http : / / dx . doi . org / 10 . 1037 / 0022 - 3514 . 44 . 1 . 78 . Kohn , N . W . , & Smith , S . M . ( 2010 ) . Collaborative \ufb01 xation : Effectsofothers ' ideasonbrain - storming . Applied Cognitive Psychology , 25 ( 3 ) , 359 \u2013 371 . http : / / dx . doi . org / 10 . 1002 / acp . 1699 . Kohn , N . W . , Paulus , P . B . , & Choi , Y . ( 2011 ) . Building on the ideas of others : An examina - tion of the idea combination process . Journal of Experimental Social Psychology , 47 , 554 \u2013 561 . http : / / dx . doi . org / 10 . 1016 / j . jesp . 2011 . 01 . 004 . Korde , R . M . ( 2012 ) . Asynchronousidea generation . Unpublished Master ' s ThesisTheUni - versity of Texas at Arlington . Korde , R . M . ( 2014 ) . Hybrid brainwriting : The ef \ufb01 cacyofalternating between individual and groupbrainstorming and the effectofindividual differences . Unpublished DoctoralDis - sertation The University of Texas at Arlington . Larson , J . R . ( 2010 ) . Insearchofsynergy in smallgroupperformance . NewYork : Psychology Press . Leggett , K . L . , Putman , V . L . , Roland , E . J . , & Paulus , P . B . ( 1996 ) . The effects of training on performance in group brainstorming . Presented at the Southwestern Psychological Asso - ciation , Houston , Texas . Medeiros , K . E . , Partlow , P . J . , & Mumford , M . D . ( 2014 ) . Not too much , not too little : The in \ufb02 uence of constraints on creative problem solving . Psychology of Aesthetics , Creativity , and the Arts , 8 ( 2 ) , 198 \u2013 210 . http : / / dx . doi . org / 10 . 1037 / a0036210 . Michinov , N . ( 2012 ) . Iselectronic brainstorming orbrainwriting the best wayto improve creative performance in groups ? An overlooked comparison of two idea - generation 189 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190 techniques . Journal of Applied Psychology , 42 ( S1 ) , E222 \u2013 E243 . http : / / dx . doi . org / 10 . 1111 / j . 1559 - 1816 . 2012 . 01024 . x . Nijstad , B . A . , & Stroebe , W . ( 2006 ) . Howthegroupaffectsthemind : Acognitivemodelof idea generation in groups . Personality and Social Psychology Review , 10 ( 3 ) , 186 \u2013 213 . http : / / dx . doi . org / 10 . 1207 / s15327957pspr1003 _ 1 . Nijstad , B . A . , De Dreu , C . K . , Rietzschel , E . F . , & Baas , M . ( 2010 ) . The dual pathway to cre - ativity model : Creative ideation as a function of \ufb02 exibility and persistence . European Review of Social Psychology , 21 ( 1 ) , 34 \u2013 77 . http : / / dx . doi . org / 10 . 1080 / 10463281003765323 . Nijstad , B . A . , Stroebe , W . , & Lodewijkx , H . F . M . ( 1999 ) . Persistence of brainstorming groups : How do people know when to stop ? Journal of Experimental Social Psychology , 35 , 165 \u2013 185 . http : / / dx . doi . org / 10 . 1006 / jesp . 1998 . 1374 . Nijstad , B . A . , Stroebe , W . , & Lodewijkx , H . F . M . ( 2002 ) . Cognitive stimulation and inter - ference in groups : Exposure effects in an idea generation task . Journal of Experimental Social Psychology , 38 , 535 \u2013 544 . http : / / dx . doi . org / 10 . 1016 / S0022 - 1031 ( 02 ) 00500 - 0 . Ocker , R . , Fjermestad , J . , Hiltz , S . R . , & Johnson , K . ( 1998 ) . Effects of four modes of group communication on the outcomes of software requirements determination . Journal of Management Information Systems , 15 ( 1 ) , 99 \u2013 118 . http : / / dx . doi . org / 10 . 1080 / 07421222 . 1998 . 11518198 . Osborn , A . F . ( 1953 ) . Applied imagination : Principles and procedures of creative thinking . New York , NY : Scribner . Osborn , A . F . ( 1957 ) . Applied imagination : Principles and procedures of creative thinking . New York , NY : Scribner . Paulus , P . B . , & Brown , V . R . ( 2007 ) . Towardmorecreativeandinnovativegroupideagen - eration : A cognitive - social - motivational perspective of brainstorming . Social and Personality Psychology Compass , 1 ( 1 ) , 248 \u2013 265 . http : / / dx . doi . org / 10 . 1111 / j . 1751 - 9004 . 2007 . 00006 . x . Paulus , P . B . , & Coskun , H . ( 2012 ) . Group creativity . In J . M . Levine ( Ed . ) , Group processes ( pp . 215 \u2013 239 ) . Amsterdam : Elsevier . Paulus , P . B . , & Dzindolet , M . ( 1993 ) . Social in \ufb02 uence processes in group brainstorming . Journal of Personality and Social Psychology , 64 ( 5 ) , 575 \u2013 586 . http : / / dx . doi . org / 10 . 1037 / 0022 - 3514 . 64 . 4 . 575 . Paulus , P . B . , & Yang , H . ( 2000 ) . Idea generation ifgroups : A basis for creativity inorgani - zations . Organizational Behavior and Human Decision Processes , 82 , 76 \u2013 87 . http : / / dx . doi . org / 10 . 1006 / obhd . 2000 . 2888 . Paulus , P . B . , Dzindolet , M . T . , Poletes , G . W . , & Camacho , L . M . ( 1993 ) . Perception of per - formance in group brainstorming : The illusion of group productivity . Personality and Social Psychology Bulletin , 19 , 78 \u2013 89 . http : / / dx . doi . org / 10 . 1177 / 0146167293191009 . Paulus , P . B . , Kohn , N . W . , & Arditti , L . E . ( 2011 ) . Effectsofquantityandqualityinstructions on brainstorming . Journal of Creative Behavior , 45 , 38 \u2013 46 . http : / / dx . doi . org / 10 . 1002 / j . 2162 - 6057 . 2011 . tb01083 . x . Paulus , P . B . , Kohn , N . W . , Arditti , L . E . , & Korde , R . M . ( 2013 ) . Understanding the group size effect in electronic brainstorming . Small Groups Research , 44 , 332 \u2013 352 . http : / / dx . doi . org / 10 . 1177 / 10464 96413479674 . Paulus , P . B . , Korde , R . M . , Dickson , J . J . , Carmeli , A . , & Cohen - Meitar , R . ( 2015 ) . Asynchro - nous brainstorming in an industrial setting : Exploratory studies . Human Factors , 57 ( 6 ) , 1076 \u2013 1094 . http : / / dx . doi . org / 10 . 1177 / 0018720815570374 . Paulus , P . B . , Larey , T . S . , & Ortega , A . H . ( 1995 ) . Performance and perceptions of brain - stormers in an organizational setting . Basic and Applied Social Psychology , 17 , 249 \u2013 265 . http : / / dx . doi . org / 10 . 1207 / s15324834basp1701 & 2 _ 15 . Paulus , P . B . , Larey , T . S . , Putman , V . L . , Leggett , K . L . , & Roland , E . J . ( 1996 ) . Socialin \ufb02 uence processing in computer brainstorming . Basic and Applied Social Psychology , 18 ( 1 ) , 3 \u2013 14 . http : / / dx . doi . org / 10 . 1207 / s15324834basp1801 _ 2 . Paulus , P . B . , Nakui , T . , Putman , V . L . , & Brown , V . R . ( 2006 ) . Effectsoftaskinstructionsand brief breaks on brainstorming . Group Dynamics : Theory , research , and practice , 10 , 206 \u2013 219 . http : / / dx . doi . org / 10 . 1037 / 1089 - 2699 . 10 . 3 . 206 . Rietzschel , E . F . , Nijstad , B . A . , & Stroebe , W . ( 2007 ) . Relative accessibility of domain spe - ci \ufb01 c knowledge and creativity : The effects of knowledge activation on the quantity and originality of generated ideas . Journal of Experimental Social Psychology , 43 , 933 \u2013 946 . http : / / dx . doi . org / 10 . 1016 / j . jesp . 2006 . 10 . 014 . Rotter , G . S . , & Portugal , S . M . ( 1969 ) . Group and individual effects in problem solving . Journal of Applied Psychology , 53 ( 4 ) , 338 \u2013 341 . http : / / dx . doi . org / 10 . 1037 / h0027771 . Santanen , E . L . , Briggs , R . O . , & de Vreede , G . ( 2004 ) . Causal relationship in creative prob - lem solving : Comparing facilitation interventionsforideation . Journal ofManagement Information Systems , 20 ( 4 ) , 167 \u2013 197 . Steiner , I . D . ( 1972 ) . Group processes and productivity . New York : Academic Press . Stokes , P . D . ( 2005 ) . Creativityfromconstraints : Thepsychologyofbreakthrough . NewYork , NY : Springer . Sutton , R . I . , & Hargadon , A . ( 1996 ) . Brainstorming groups in context : Effectiveness in a product design \ufb01 rm . Administrative Science Quarterly , 41 ( 4 ) , 685 \u2013 718 . http : / / dx . doi . org / 10 . 2307 / 2393655 . Taylor , D . W . , Berry , P . C . , & Block , C . H . ( 1958 ) . Does group participation when using brainstorming facilitate or inhibit creative thinking ? Administrative Science Quarterly , 3 , 23 \u2013 47 . http : / / dx . doi . org / 10 . 2307 / 2390603 . Valacich , J . S . , Dennis , A . R . , & Connolly , T . ( 1994 ) . Idea generation in computer - based groups : A new ending to an old story . Organizational Behavior and Human Decision Processes , 57 ( 3 ) , 448 \u2013 467 http : / / dx . doi . org . ezproxy . uta . edu / 10 . 1006 / obhd . 1994 . 1024 . 190 R . Korde , P . B . Paulus / Journal of Experimental Social Psychology 70 ( 2017 ) 177 \u2013 190",
    "lundmark2008gtpaseactivating_supplement": "Current Biology Volume 18 1 Supplemental Data The GTPase - Activating Protein GRAF1 Regulates the CLIC / GEEC Endocytic Pathway Richard Lundmark , Gary J . Doherty , Mark T . Howes , Katia Cortese , Yvonne Vallis , Robert G . Parton , and Harvey T . McMahon Supplemental Experimental Procedures cDNA construct preparation cDNA constructs encoding human GRAF1 ( amino acids 1 - 759 ) , GRAF1 - BAR + PH ( amino acids 1 - 383 ) , GRAF1 PH + GAP ( amino acids 267 - 576 ) , and GRAF1 - SH3 ( 694 - 759 ) were amplified from IMAGE clone 30343863 using PCR and cloned into the pGEX - 4T - 2 vector for bacterial expression ( Amersham Biosciences ) . GRAF2 SH3 domain was amplified similarly from IMAGE clone 6188298 and cloned into the pGEX - 4T - 2 vector . Fragments were also cloned into the pCMVmyc vector with added Not1 site ( a kind gift from JGW Anderson ) or EGFP - C3 ( Clontech ) for mammalian expression . Amino acid substitutions K131E , K132E and R412D were created using PCR directed mutagenesis ( Stratagene ) . The GFP - tagged Cdc42 L61 consrtruct was a kind gift from M . F\u00e4llman . pEGFP - GPI was a generous gift from S . Mayor . caveolin1 - GFP and flotillin1 - GFP were kind gifts from Ben Nichols . Protein expression , protein purification and antibodies Recombinant proteins were expressed in a BL21 ( DE3 ) pLysS E - coli strain as Glutathione S - transferase ( GST ) - fusion proteins and purified using glutathione - Sepharose 4B beads ( Amersham Biosciences ) and gel filtration on a sephacryl S - 200 column ( Amersham ) as previously described [ 1 ] . Polyclonal antisera against GRAF1 were generated by immunising Rabbits ( Ra83 / Ab1 directed against the SH3 domain ; Ra84 / Ab2 directed against the PH + GAP domains ; and RaZ1 / Ab3 directed against the full length protein ; all of these were used for Western blotting analyses , and Ab1 / 3 also used for immunofluorescence analyses ) and a chicken ( CH - 9798 / Ab4 , directed against the full length protein , used for immunofluorescence analyses ) with recombinantly expressed human GRAF1 proteins . Purchased antibodies were : mouse anti - myc clone 9E10 , mouse anti - tubulin ( Sigma - Aldrich ) , Rabbit anti - myc ( Cell Signalling Technology ) , mouse anti - Dynamin , ( BD Transduction Laboratories ) , Rabbit anti - synaptojanin Ra59 [ 1 ] , ( Affinity Bioreagents ) . All secondary antibodies and streptavidins were conjugated to Alexa Fluor 488 , 546 or 647 ( Invitrogen ) . Expression , immunoprecipitation and pull down experiments For analysis of endogenous protein expression , cell lines were grown according to instructions from American Tissue Culture Collection , harvested and lysed in 1 % NP - 40 in PBS supplemented with protease inhibitors . After a 20 , 000g centrifugation the supernatant was analysed by SDS - PAGE and immunoblotting . For immunoprecipitation experiments , rat brain cytosol was generated by homogenization of rat brains in buffer ( 25mM HEPES , 150mM NaCl , 1mM DTT , 0 . 1 % Triton X - 100 and protease inhibitors ) , before centrifugation at 50 , 000rpm for 30 minutes at 4\u00b0C . The supernatant was removed and added to protein A Sephorose 4B beads ( Amersham Biosciensis ) to which antibodies had been previously bound and incubated at 4\u00b0C for 3 hours . Beads were washed three times in buffer ( 25mM HEPES , 150mM NaCl ) supplemented Current Biology Volume 18 2 with 1 % NP - 40 , and once in buffer without NP - 40 before analysis by SDS - PAGE combined with immunoblotting or Coomassie staining . Pull - down experiments against rat brain cytosol using purified proteins and identification by mass - spectrometry were performed as previously described [ 1 ] . Protein and lipid interaction assays Liposomes from total brain lipids ( FOLCH fraction I ) ( Sigma Aldrich ) or synthetic lipids ( Avanti Polar Lipids ) , and liposomes of a specified diameter or phosphoinositide enrichment were generated as previously described [ 2 ] . Liposome binding assays for lipid specificity and curvature sensitivity was performed as previously described [ 2 ] . Briefly , proteins were incubated together with liposomes followed by centrifugation and analysis of the pellet and supernatant by SDS - PAGE and Coomassie staining . In vitro liposome tubulation assays were performed and analysed as previously described [ 2 ] . Isothermal titration calorimetry The binding of synthetic peptides from Dynamin1 to purified GRAF1 SH3 domain was measured by isothermal titration calorimetry ( ITC ) using a VP - ITC ( MicroCal Inc . , USA ) . All experiments were performed in 100mM HEPES / NaOH pH 7 . 4 , 50mM NaCl and 2mM DTT at 10\u00b0C . Protein concentrations were determined by absorbance at 280nm . 1 . 36ml of 51\u00b5M GRAF1 SH3 domain was loaded into the cell . The peptides ( at 1mM which were custom - designed and manufactured at the Institute of Biomolecular Sciences , University of Southampton , UK ) were injected from a syringe in 5\u00b5l steps every 3 . 5 minutes . The heat of the dilution of the ligand was subtracted from the data prior to fitting . Titration curves were fitted to the data using the ORIGIN program ( MicroCal Inc . ) which yielded the stoichiometry ( ~ 1 ) , the binary association constant K a ( = K d - 1 ) and the enthalpy of binding . The entropy of binding ( \u2206 S\u00b0 ) was calculated from the relationship \u2206 G\u00b0 = - RT \u2022 ln ( K a ) and the Gibbs - Helmholtz equation . Cell culture and transfections HeLa cells were grown in RPMI 1640 or MEM media ( GIBCO ) supplemented with LGlutamine , 10 % foetal bovine serum , non - essential amino acids ( for MEM ) , and transfected using Genejuice ( Novagen ) for transient protein expression . For primary cultures , rat hippocampal neurons / astrocytes were prepared by trypsin digestion and mechanical trituration from E18 or P1 Sprague - Dawley rats and plated onto poly L - lysine coated coverslips . Cells were cultured in B27 - supplemented Neurobasal media . For GRAF1 depletion , HeLa cells were transfected with Stealth siRNAs specific against human GRAF1 using Lipofectamine 2000 ( Invitrogen ) according to manufacturers instructions . The Invitrogen siRNA duplex sequences used were siRNAa ( UUA UCU CCC AUU CAG CAC AGA UAU C / GAU AUC UGU GCU GAA UGG GAG AUA A ) , and siRNAb ( UUU GAA ACU GGU ACA UCA UGA GUG G / CCA CUC AUG AUG UAC CAG UUU CAA A ) . Cells were cultured for an additional 48 hours for efficient silencing of the GRAF1 expression . Stealth Block - it siRNA ( Invitrogen ) was used as a control . AP2 siRNA was used as previously described [ 3 ] . Caveolin - 1 knock - out mouse embryonic fibroblasts ( KO MEFs ) were generated and grown as previously described [ 4 ] . NIH 3T3 cells were cultured as per ECACC guidelines . Trafficking assays For immunofluorescence trafficking assays , biotinylated holo - transferrin , ( Sigma Aldrich ) , Alexa Fluor 647 - conjugated transferrin ( Invitrogen ) , Alexa Fluor 546 / 555 - conjugated CTxB ( Invitrogen ) , DiI ( Invitrogen ) , FITC - dextran ( 10kDa MW , used for fluorimetric uptake assay , Invitrogen ) , and biotinylated dextran ( 10kDa MW , used for immunofluorescent uptake assays , Invitrogen ) , were diluted in pre - warmed media , added to cells and incubated for time periods and temperatures as described in figure legends . After washing , cells were fixed and subjected to Current Biology Volume 18 3 immunofluorescence analysis as described below . For quantitative analysis of dextran endocytosis , HeLa cells in 35mm dishes were transfected with siRNAs / control siRNAs 48 hours prior to the experiment . Fluorescein isothiocyanate ( FITC ) - dextran ( Sigma - Aldrich ) was diluted in media to a concentration of 1mg / ml and added to cells before incubation for 15 minutes at the indicated temperature . Cells were washed twice in media and once in PBS before lysis in 1 % NP - 40 in PBS supplemented with protease inhibitors . The lysate was centrifuged at 20 , 000g for 20 minutes at 4\u00b0C and the protein concentration in the supernatant was measured using the BCA Protein Assay Kit ( Pierce ) for normalization . The amount of FITC - dextran in the supernatant was measured as the emission at 515nm after exciting at 488nm using a FP - 6500 spectrofluorometer with Spectra Manager software ( JASCO ) . The MHC Class I uptake assay was performed according to reference [ 5 ] . Briefly , HeLa cells transiently transfected with GFP - tagged GRAF1 or GRAF1 BAR + PH for 16 hours were incubated with W6 / 32 anti - MHC Class I antibody ( American Type Culture Collection ) diluted in culture media for 5 or 15 minutes at 37\u00b0C to allow endocytosis . Cells were washed in PBS and surface - bound antibody was removed by a 30 second acid wash followed by a wash in culture media to re - adjust the pH . Cells were fixed and MHC Class I was visualised using Alexa568 - conjugated antibodies . Fixed sample and real time imaging For immunofluorescence analysis , HeLa cells were fixed in 3 % paraformaldehyde in phosphate - buffered saline ( PBS ) for 15 minutes at 37\u00b0C ( to preserve intracellular tubules which are disrupted by fixation at lower temperatures ) , or 4\u00b0C ( to demonstrate this temperature dependence ) , then washed and blocked in 5 % goat serum , with 0 . 1 % saponin , in PBS before staining with the appropriate antibodies in 1 % goat serum , 0 . 1 % saponin in PBS using standard protocols . Confocal images were taken sequentially using a BioRad Radiance system and LaserSharp software ( BioRad ) . Epifluorescence images were taken using a Zeiss Axioimager Z1 system with AxioVision software . For real time microscopy of the dynamics of GRAF1 - and GRAF1 BAR + PH - positive tubules , transfected cells on glass - bottom Petri dishes ( WillCo Wells BV , Amsterdam ) were washed with buffer ( 125mM NaCl , 5mM KCl , 10mM D - glucose , 1mM MgCl2 , 2mM CaCl2 and 25mM HEPES ) and images were taken using a 5 - live scanning microscope ( Zeiss ) or spinning disc confocal system ( Improvision ) with subsequent analysis in LSM Image Browser ( Zeiss ) , ImageJ ( freeware ) or Volocity ( Improvision ) . Caveolin1 - KO MEFs and control MEFs grown on 12mm coverslips were cotransfected with GFP - GPI and myc - tagged GRAF1 or GRAF1 BAR + PH . Anti - GFP was bound to cells on ice for 30 minutes in unsupplemented CO2 - independent medium ( Gibco ) . Cells were washed in CO2 - independent medium to remove unbound antibody prior to internalization in pre - warmed growth media ( 10 % foetal bovine serum ( Cambrix ) , 2mM L - glutamine in DMEM ( Gibco ) ) at 37\u00b0C , 5 % CO2 for 2 , 10 or 40 minutes . Post internalization , cells were washed in ice - cold CO2 - independent medium and acid stripped using 500mM glycine pH2 . 2 in ice cold PBS for 1 minute . To remove intensive cytosolic labeling , acid - stripped live cells were permeablized in ice - cold PBS containing 0 . 05 % saponin ( Sigma ) for 5 minutes . Cells were washed 2 x 1 minute in ice cold PBS before fixation in 2 % paraformaldehyde . Internalized anti - GFP was labeled with Alexa Fluor - 660 - conjugated goat anti - Rabbit secondary antibody . Myc - tagged GRAF1 FL or BAR + PH was labeled with anti - myc ( 9B11 ) primary and Alexa - Fluor - 555 goat - anti mouse secondary antibodies . Fluorescence microscopy was carried out using an Axiovert 200m SP LSM 510 META confocal laser - scanning microscope ( Zeiss ) . Images were captured under oil with a 63x plan - APOCHROMAT objective , at appropriate excitation and emission wavelengths . Images were processed using Adobe Current Biology Volume 18 4 Photoshop CS2 . Quantification of anti - GFP and GRAF1 / GRAF1 BAR + PH colocalization was carried out using Volocity 3 . 7 . 0 . In brief , target cells were cropped and blue and red channels overlaid to generate a colocalization coefficient based on a percentage of blue voxels that identify with red voxels . Background was subtracted using the automatic threshold feature . Five to seven images across three independent experiments were used to calculate average colocalization and standard error of the mean ( SEM ) . Relative pixel numbers of anti - GFP in each image was calculated in Adobe Photoshop CS2 based on the histogram of each image . For real - time endocytosis experiments , NIH 3T3 cells grown in 35mmM glass bottom dishes were transfected with GFP - tagged GRAF1 or GRAF1 BAR + PH constructs . Cells were washed in ice - cold CO2 - independent medium before binding Alexa Fluor - 555 - conjugated CTxB and / or Alexa Fluor - 647 - conjugated transferrin for 30 minutes on ice . Appropriate cells were identified using an Axiovert 200m SP LSM 510 META confocal laser - scanning microscope in ice - cold CO2 - independent medium . Medium was exchanged for prewarmed CO2 - independent medium plus 10 % heat - inactivated Serum Supreme and images captured using Lasersharp 2000 4 . 0 . Frames were captured every 10 seconds for 50 - 100 frames . Images were processed using Image J v1 . 37 and converted to Quicktime files . Supplemental References 1 . Praefcke , G . J . , and McMahon , H . T . ( 2004 ) . The Dynamin superfamily : universal membrane tubulation and fission molecules ? Nat . Rev . Mol . Cell Biol . 5 , 133 - 147 . 2 . Peter , B . J . , Kent , H . M . , Mills , I . G . , Vallis , Y . , Butler , P . J . , Evans , P . R . , and McMahon , H . T . ( 2004 ) . BAR domains as sensors of membrane curvature : the amphiphysin BAR structure . Science 303 , 495 - 499 . 3 . Motley , A . , Bright , N . A . , Seaman , M . N . , and Robinson , M . S . ( 2003 ) . Clathrin - mediated endocytosis in AP - 2 - depleted cells . J . Cell Biol . 162 , 909 - 918 . 4 . Kirkham , M . , Fujita , A . , Chadda , R . , Nixon , S . J . , Kurzchalia , T . V . , Sharma , D . K . , Pagano , R . E . , Hancock , J . F . , Mayor , S . , and Parton , R . G . ( 2005 ) . Ultrastructural identification of uncoated caveolin - independent early endocytic vehicles . J . Cell Biol . 168 , 465 - 476 . 5 . Caplan , S . , Naslavsky , N . , Hartnell , L . M . , Lodge , R . , Polishchuk , R . S . , Donaldson , J . G . , and Bonifacino , J . S . ( 2002 ) . A tubular EHD1 - containing compartment involved in the recycling of major histocompatibility complex class I molecules to the plasma membrane . EMBO J . 21 , 2557 - 2567 . Current Biology Volume 18 5 Figure S1 | Localisation of GRAF1 to tubular membrane structures is temperature sensitive and dependent on the BAR and PH domains . A , Western blots showing the different forms of GRAF1 detected in adult rat brain and their differential presence / absence in cultured SH - SY5Y ( human neuroblastoma ) , HeLa ( human fibroblast ) , K562 ( human Chronic Myeloid Leukaemia ) , and MEF ( mouse embryonic fibroblast ) , cells . Western blots of purified GRAF1 and myc - tagged GRAF1 ( from lysates of HeLa cells overexpressing this protein ) are shown for comparison . B , Confocal micrograph of an NIH 3T3 cell stained for endogenous GRAF1 distribution . C , Confocal micrographs of HeLa cells fixed either at 4\u00b0C or 37\u00b0C for 10 minutes in 4 % paraformaldehyde and then stained for endogenous GRAF1 . Note the absence of GRAF1 - positive tubules in the 4\u00b0C fixation image . D , Confocal micrographs showing the tubular localization of overexpressed myc - tagged GRAF1 BAR + PH protein in HeLa cells and the cytoplasmic localization of a similarly overexpressed protein with a BAR domain mutation ( KK131 / 132EE ) . E , Confocal micrograph showing the cytoplasmic and punctate localization of overexpressed myc - tagged GRAF1 BAR protein in HeLa cells . Scale bars = 10 \u00b5 m . Current Biology Volume 18 6 Figure S2 A - D | Supplementary biochemical data on the binding between GRAF1 and Dynamin . A , Coomassie - stained gel and confirmatory Western blots of co - immunoprecipitation experiments in rat brain cytosol performed with either control pre - immunization serum ( pre - serum ) or the Ab3 antibody . Bands in the Coomassie - stained gel were identified by mass spectrometry as described . B and C , Coomassie - stained gel and Western blots of pull - down experiments from mouse brain lysate ( B ) or HeLa cell cytosol ( C ) with beads bound to GST ( control ) or GST - tagged GRAF1 BAR + PH , or SH3 proteins . The bands in the Coomassie - stained gel were identified by mass spectrometry as described . Note the major band of Dynamin present in the SH3 lanes , which is not present in the control or BAR + PH condition . \u2018cyt\u2019 marks the HeLa cell lysate ( positive control ) lane . D , The upper panel shows a raw trace from isothermal titration calorimetry performed as described . The lower panel shows the fitting of this data to a one - site binding model from which the affinity ( shown ) can be calculated . GRAF1 SH3 domain and peptide concentrations , as well as injection volumes and times are shown . Current Biology Volume 18 7 Figure S2 E | Dynasore inhibits the uptake of dextran and affects the localization of GRAF1 . E , Confocal micrographs ( maximum projections ) of HeLa cells treated with either DMSO ( vehicle ) or 100\u00b5M dynasore for 1 hour before addition of dextran for 15 minutes , fixation , and immunostaining for dextran and the focal adhesion marker paxillin . Scale bars = 10 \u00b5 m . Current Biology Volume 18 8 Figure S3 | GRAF1 - positive endocytic structures are Clathrin - independent and exclude transferrin A - C , Confocal fluorescent micrographs of HeLa cells stained for endogenous GRAF1 and Clathrin ( A ) , transferrin ( B ) or transferrin receptor ( B ) . The depicted images are used to show some of the different morphologies of GRAF1 - positive structures that are observed in these cells . Scale bars = 10 \u00b5 m . Current Biology Volume 18 9 Figure S4 | GRAF1 BAR + PH overexpression affects CTxB uptake but not transferrin uptake . A and B , Confocal micrographs of HeLa cells transfected with myc - tagged GRAF1 BAR + PH and incubated with CTxB or transferrin for 15 minutes before fixation and staining . C , The graph shows the quantification of images such as depicted in ( A ) . Cells were scored for expression of GRAF1 BAR + PH ( over a threshold corresponding to maximum autofluorescence ) and CTxB internalization ( over an arbitrarily - set threshold above background ) . Note the reduction of the number of transfected cells internalising CTxB compared with controls . D , Live cell microscopy of NIH 3T3 cells expressing GFP - tagged GRAF1 and incubated with CTxB and transferrin ( Tfn ) at 4\u00b0C before chasing their internalization from the time of warming to 37\u00b0C ( time = 00 : 00 ) . Note the internalising GRAF1 - positive tubule containing CTxB . Note also the lack of colocalization of GRAF1 - positive tubules with internalized transferrin . Time is given as minutes : seconds . This sequence is taken from Movie S3 . Scale bars = 10 \u00b5 m . Current Biology Volume 18 10 Figure S5 | Nature of GRAF1 - positive endocytic structures . A and B , Confocal micrographs of HeLa cells overexpressing GFP - tagged flotillin1 ( A ) or GFP - tagged caveolin1 ( B ) and co - stained for endogenous GRAF1 . Note the lack of colocalization . C and D , Confocal micrographs of HeLa cells overexpressing myc - tagged GRAF1 BAR + PH and flotillin1 ( E ) or caveolin1 ( F ) incubated with CTxB for 5 minutes . Note the colocalization of GRAF1 BAR + PH and flotillin1 in CTxB - positive tubular structures . Scale bars = 10 \u00b5 m . Current Biology Volume 18 11 Figure S6 | GRAF1 BAR + PH overexpression does not affect the uptake of MHC Class I which enter GRAF1 - negative compartments . A and B , Epifluorescence micrographs of HeLa cells , transiently transfected with GFP - tagged GRAF1 FL ( A ) or GRAF BAR + PH ( B ) were pulsed with anti - MHC Class I antibody for 5 ( A ) or 15 minutes ( B ) at 37\u00b0C followed by a brief acid wash to remove surface - bound antibody . MHC Class I was visualized using Alexa568 - conjugated secondary antibodies . C , Quantitation of surface GFP - GPI levels in cells overexpressing this protein with myc - tagged GRAF1 FL or GRAF1 BAR + PH . D , Confocal micrographs of HeLa cells treated with siRNA against GRAF1 or a control siRNA and stained for endocgenous GRAF1 . Scale bars = 10 \u00b5 m .",
    "mcfarland2008rna": "RNA Interference - Mediated Knockdown of Dynamin 2 Reduces Endocannabinoid Uptake into Neuronal dCAD Cells Matthew J . McFarland , Tamera K . Bardell , Marla L . Yates , Ekaterina A . Placzek , and Eric L . Barker Department of Medicinal Chemistry and Molecular Pharmacology , Purdue University , West Lafayette , Indiana Received January 3 , 2008 ; accepted April 23 , 2008 ABSTRACT The precise mechanism by which the cellular uptake of the endocannabinoid anandamide ( AEA ) occurs has been the source of much debate . In the current study , we show that neuronal differentiated CAD ( dCAD ) cells accumulate anand - amide by a process that is inhibited in a dose - dependent manner by N - ( 4 - hydroxyphenyl ) arachidonylamide ( AM404 ) . We also show that dCAD cells express functional fatty acid amide hydrolase , the enzyme primarily responsible for anandamide metabolism . Previous data from our laboratory indicated that anandamide uptake occurs by a caveolae - related endocytic mechanism in RBL - 2H3 cells . In the current study , we show that anandamide uptake by dCAD cells may also occur by an endocytic process that is associated with detergent - resistant membrane microdomains or lipid rafts . Nystatin and progester - one pretreatment of dCAD cells significantly inhibited anand - amide accumulation . Furthermore , RNA interference ( RNAi ) - mediated knockdown of dynamin 2 , a protein involved in endocytosis , blocked the internalization of the fluorescently labeled anandamide analog SKM 4 - 45 - 1 ( [ 3 (cid:1) , 6 (cid:1) - bis ( acetyloxy ) - 3 - oxospiro [ isobenzofuran - 1 ( 3 H ) , 9 (cid:1) - [ 9 H ] xanthen - 5 - yl ] - 2 - [ [ 1 - oxo - 5 Z , 8 Z , 11 Z , 14 Z - eicosatetraenyl ] amino ] ethyl ester carbamic acid ) . RNAi - mediated knockdown of the (cid:1) 2 subunit of the clathrin - associated activator protein 2 complex had no effect on SKM 4 - 45 - 1 internalization . We were surprised to find that dynamin 2 knockdown in dCAD cells did not affect [ 3 H ] AEA uptake . However , dynamin 2 knockdown caused a significant increase in the overall levels of intact [ 3 H ] AEA associated with the cells , suggesting that trafficking of [ 3 H ] AEA to FAAH had been disrupted . This finding may be the result of an accumu - lation of the anandamide carrier protein in detergent - resistant membranes after dynamin 2 knockdown . Our studies provide evidence that the cellular uptake of anandamide may occur by a dynamin 2 - dependent , caveolae - related endocytic process in dCAD cells . The endocannabinoid anandamide ( AEA ) is an agonist of the cannabinoid 1 and 2 receptors ( Matsuda et al . , 1990 ; Munro et al . , 1993 ) and some vanilloid type ion channels ( Di Marzo et al . , 2001 ; Voets and Nilius , 2003 ) . AEA is internal - ized by most cell types and is thought to be metabolized primarily by the intracellular enzyme fatty acid amide hy - drolase ( FAAH ) ( Deutsch and Chin , 1993 ; Di Marzo et al . , 1994 ; Cravatt et al . , 1996 ) . Previous studies from our labo - ratory have indicated that AEA uptake potentially occurs by a caveolae - related , or clathrin - independent , endocytic pro - cess ( McFarland et al . , 2004 ) . Interestingly , neurons are not thought to express caveolin - 1 but do exhibit lipid raft - related endocytic processes ( Cameron et al . , 1997 ) . Thus , our studies seek to validate a neuronal - like cell line as a useful model in the study of AEA uptake and to show that AEA uptake can be disrupted by using molecular inhibitors that are specifically targeted to endocytic processes . Cath . a cells display neuronal properties and express pan - neuronal markers but lack classic neuronal morphology ( Suri et al . , 1993 ; Lazaroff , 1996 ) . Cath . a differentiated ( CAD ) cells are derived from the Cath . a cell line and are characterized by the loss of the immortalizing oncogene present in Cath . a cells ( Qi et al . , 1997 ) . CAD cells also undergo a reversible morpho - logical differentiation that is initiated when serum is re - moved from the cell culture media ( Qi et al . , 1997 ) . Differen - tiated CAD ( dCAD ) cells stop proliferating and extend long neuronal processes characteristic of primary neuronal cell cultures ( Qi et al . , 1997 ) . Accompanying the morphological This works supported by National Institutes of Health grants R21 - DA13268 and R21 - DA018844 ( to E . L . B . ) . Article , publication date , and citation information can be found at http : / / molpharm . aspetjournals . org . doi : 10 . 1124 / mol . 108 . 044834 . ABBREVIATIONS : AEA , anandamide ; FAAH , fatty acid amide hydrolase ; CAD , Cath . a differentiated ; cCAD , cycling CAD ; siRNA , short interfering RNA ; AP , activator protein ; Dyn 2 , dynamin 2 ; KRH , Krebs - Ringer - HEPES ; AM404 , N - ( 4 - hydroxyphenyl ) arachidonylamide ; PBS , phosphate - buffered saline ; MAFP , methyl arachidonyl fluorophosphonate ; RNAi , RNA interference ; 488 - CT , cholera toxin B subunit - conjugated Alexa Fluor 488 ; 488 - Tf , transferrin - conjugated Alexa Fluor 488 ; TLC , thin layer chromatography ; CB , cannabinoid ; NP , nystatin / progesterone ; SKM 4 - 45 - 1 , ( [ 3 (cid:1) , 6 (cid:1) - bis ( acetyloxy ) - 3 - oxospiro [ isobenzofuran - 1 ( 3 H ) , 9 (cid:1) - [ 9 H ] xanthen - 5 - yl ] - 2 - [ [ 1 - oxo - 5 Z , 8 Z , 11 Z , 14 Z - eicosatetraenyl ] amino ] ethyl ester carbamic acid ) . 0026 - 895X / 08 / 7401 - 101 \u2013 108 $ 20 . 00 M OLECULAR P HARMACOLOGY Vol . 74 , No . 1 Copyright \u00a9 2008 The American Society for Pharmacology and Experimental Therapeutics 44834 / 3357609 Mol Pharmacol 74 : 101 \u2013 108 , 2008 Printed in U . S . A . 101 a t A SP ET J ou r n a l s on M a r c h 17 , 2015 m o l ph a r m . a s p e tj ou r n a l s . o r g D o w n l o a d e d fr o m differentiation of CAD cells is an observed change in the expression and function of several neuronal proteins . For example , dCAD cells express higher levels of tyrosine hydrox - ylase ( Lazaroff et al . , 1998 ) and somatostatin receptor sub - type 2a ( Hashemi et al . , 2003 ) , exhibit a decrease in sodium current accompanied by an increase in potassium current ( Wang and Oxford , 2000 ) , and display alterations in cyclic AMP - mediated signaling ( Hashemi et al . , 2003 ) . Despite re - ported differences in their cellular signaling and protein ex - pression profiles , both cycling CAD ( cCAD ) and dCAD cells express neuron - specific proteins and display the biochemical characteristics of neuronal cells ( Qi et al . , 1997 ) . In RBL - 2H3 cells , AEA transport occurs via a clathrin - independent , or caveolae - related endocytic process , substan - tiated by studies using pharmacological inhibitors of endocy - tosis ( McFarland et al . , 2004 ) . We tested the hypothesis that the disruption of lipid rafts / detergent - resistant membranes in dCAD cells would also decrease the cellular accumulation of AEA to confirm that endocytosis may play a role in AEA uptake in neuronal cells . To avoid the potential problems associated with biochemical disruption of lipid rafts , we examined AEA uptake in dCAD cells after treatment with short interfering RNA ( siRNA ) to knockdown expression of the (cid:1) 2 subunit of the AP2 complex , a protein necessary for clathrin - mediated endocytosis ( Huang et al . , 2004 ) , as well as dynamin 2 ( Dyn 2 ) , a protein involved in both clathrin - mediated and clathrin - independent endocytic processes ( Altschuler et al . , 1998 ; Maxfield and McGraw , 2004 ) . The results of our studies are consistent with a role for endocy - tosis in the neuronal uptake of endocannabinoids . Materials and Methods Cell Culture . cCAD cells were maintained in a 1 : 1 ratio of Ham\u2019s F - 12 / Dulbecco\u2019s modified Eagle\u2019s medium with 5 % bovine calf serum and 5 % fetal clone 1 supplemented with 1 % penicillin / streptomycin , and 2 mM L - glutamine . Cells were grown in a humidified environ - ment containing 5 % CO 2 and held at the constant temperature of 37\u00b0C . cCAD cells were maintained in serum - free 1 : 1 Ham\u2019s F - 12 / Dulbecco\u2019s modified Eagle\u2019s medium and were allowed to differenti - ate for 36 h before experiments to generate dCAD cells . [ 3 H ] AEA Uptake and SKM 4 - 45 - 1 Internalization Assays . CAD cells were plated at approximately 50 , 000 cells per well in 24 - well culture dishes and then allowed to differentiate in serum - free media for 36 h . The resulting dCAD cells were then transfected with siRNA oligonucleotides as described below . After the transfec - tion period , media were removed and 1 nM [ 3 H ] AEA ( labeled on the arachidonate portion of the molecule ; PerkinElmer Life and Analyt - ical Sciences , Waltham , MA ) in Krebs - Ringer - HEPES ( KRH ) buffer ( 120 mM NaCl , 4 . 7 mM KCl , 2 . 2 mM CaCl 2 , 10 mM HEPES , 1 . 2 mM KH 2 PO 4 , and 1 . 2 mM MgSO 4 , pH 7 . 4 ) was added to all samples and allowed to incubate for 5 min at 37\u00b0C . To determine the effects of cholesterol depletion on AEA accumulation in dCAD cells , uptake assays were performed after a 30 - min incubation in KRH buffer containing nystatin ( 25 (cid:2) g / ml ) and progesterone ( 10 (cid:2) g / ml ) . Control cells were incubated for 30 min in 1 (cid:2) KRH buffer . AM404 ( 100 (cid:2) M ) was used to define nonspecific uptake and was added 10 min before the addition of [ 3 H ] AEA at 37\u00b0C . Cells were then washed once with 1 (cid:2) KRH buffer containing 1 % bovine serum albumin , and MicroScint - 20 ( PerkinElmer Life and Analytical Sciences ) was added to each well . The amount of tritium present was determined using a TopCount microplate scintillation and luminescence counter ( PerkinElmer Life and Analytical Sciences ) . To verify morphological changes , cells were seeded in clear 24 - well tissue culture plates for uptake assays in untransfected dCAD cells . To quantify tritium from the wells of clear plates , cells were solubilized in 1 % SDS and transferred to vials containing 10 ml of EcoLite scintillation fluid ( MP Biomedicals , Irvine , CA ) for subsequent scintillation counting . For experiments examining SKM 4 - 45 - 1 internalization , CAD cells were treated with SKM 4 - 45 - 1 ( 25 (cid:2) M ) and increasing concentra - tions of AM404 for 10 min in prewarmed 37\u00b0C buffer . Fluorescence was quantified using a FUSION microplate reader ( PerkinElmer Life and Analytical Sciences ) . Determination of Plasma Membrane Cholesterol Content . Membrane cholesterol content was determined according to a method described previously ( Millard et al . , 2005 ) . CAD cells were plated in triplicate at 5 (cid:2) 10 4 cells per 35 - mm well in complete culture medium . Twenty - four hours later , cells were washed three times with PBS and then incubated in complete medium that had been supplemented with 10 % lipoprotein - deficient fetal calf serum and 1 (cid:2) Ci / ml [ 3 H ] cholesterol ( 40 Ci / mmol ; GE Healthcare , Chalfont St . Giles , UK ) . Cells were labeled to equilibrium for 48 h . Cells were then washed three times with PBS and then treated with nystatin / progesterone as described above . Membrane cholesterol was ex - tracted by a 10 min incubation with medium containing 4 % methyl - (cid:1) - cyclodextrin . Media was collected , cells were washed three times with PBS , and proteins were extracted . Protein concentrations were determined using the bicinchoninic acid assay . [ 3 H ] Cholesterol ex - tracted into the medium was quantified by scintillation counting and normalized to total protein values . FAAH Activity Assays . FAAH activity was determined by a modification of a previously published method ( Day et al . , 2001 ) . In brief , CAD cells were homogenized in buffer containing 20 mM Tris - HCl , 1 mM EDTA , 0 . 7 (cid:2) g / ml pepstatin A , and 0 . 5 (cid:2) g / ml leupep - tin . Cell homogenates were then incubated with 5 nM [ 3 H ] AEA [ ethanolamine 1 - 3 H ] ( ARC , St . Louis , MO ) in the presence or absence of 500 nM methyl arachidonyl fluorophosphonate ( MAFP ) with a reaction volume of 250 (cid:2) l . After the incubation , the reaction was terminated by adding 500 (cid:2) l of 1 : 1 chloroform / methanol ( 2 (cid:2) assay volume ) to each tube . Samples were vortexed for 30 s and centrifuged at 3900 g for 1 min to separate the aqueous and organic phases . Then , 100 - (cid:2) l aliquots were collected from both the aqueous and organic phases , and radioactivity was quantified using a TopCount Micro - plate scintillation and luminescence counter ( PerkinElmer Life and Analytical Sciences ) . The [ 3 H ] AEA used in these experiments was labeled on the ethanolamine portion of the molecule and not the arachidonate backbone ; therefore , as FAAH activity increased so did the level of radioactivity in the aqueous phase . Western Blot Analysis . Western blot analysis was performed as described previously ( McFarland et al . , 2004 ) . The presence of FAAH , dynamin 2 , and the (cid:1) 2 subunit of the AP2 complex was detected using rabbit polyclonal anti - FAAH , rabbit polyclonal anti - dynamin 2 ( Abcam Inc . , Cambridge , MA ) , and mouse monoclonal anti - (cid:1) 2 subunit ( Novus Biologicals , Inc . , Littleton , CO ) primary an - tibodies , respectively , followed by an incubation with horseradish peroxidase - labeled goat anti - rabbit or goat anti - mouse secondary antibodies ( Bio - Rad Laboratories , Hercules , CA ) , and enhanced chemiluminescence detection reagents . Membranes were then ex - posed to X - ray film . RNAi Transfections . CAD cells were transfected with siRNA oligonucleotides ( Stealth RNAi ; Invitrogen , Carlsbad , CA ) directed against either Dyn2 or the (cid:1) 2 subunit of the AP2 complex mRNA using the Lipofectamine 2000 ( Invitrogen ) RNAi transfection proto - col for mammalian cells provided by Invitrogen ( 16 pmol of each siRNA oligo was added for every 1 (cid:2) l of Lipofectamine 2000 used ) . For later experiments , the siLentFect Lipid Reagent was used fol - lowing the manufacturer\u2019s protocol ( Bio - Rad Laboratories ) . Protein knockdown was achieved by cotransfection of two different siRNA oligonucleotides targeted at each individual protein . The siRNA se - quences were as follows : Dyn 2 : NM _ 007871 _ stealth _ 1410 , GUG - GACCUGGUUAUCCAGGAGCUAA , and NM _ 007871 _ stealth _ 2009 , GGCAGAGAAUGAGGAUGGAGCACAA , (cid:1) 2 / AP2 : NM _ 027915 _ 102 McFarland et al . a t A SP ET J ou r n a l s on M a r c h 17 , 2015 m o l ph a r m . a s p e tj ou r n a l s . o r g D o w n l o a d e d fr o m stealth _ 1591 , GCUGGUCCAACAGGUCUUGAGCUUA , and NM _ 027915 _ stealth _ 2400 , CCAUACAUACUCCACUGAUGCCAAA . Stealth RNAi negative control duplex was purchased from Invitro - gen . Sequences of the negative ( mock ) control contained medium GC content and matched no known mRNA sequence in the vertebrate genome . Confocal Microscopy . dCAD cells plated in a four - well Lab - Tek Chambered Coverglass slide ( Nalge Nunc International , Rochester , NY ) were transfected with the appropriate siRNA oligonucleotides . Then , 72 h after transfection , the cells were treated with the B subunit of cholera toxin conjugated to Alexa Fluor 488 ( 488 - CT ) ( Invitrogen ) , transferrin conjugated to Alexa Fluor 488 ( 488 - Tf ) ( In - vitrogen ) , or SKM 4 - 45 - 1 ( Muthian et al . , 2000 ) for 10 min and imaged using oil immersion confocal microscopy at 60 (cid:2) magnifica - tion with a krypton ( 488 nm ) / argon ( 568 nm ) laser , band pass filter at 522 to 535 nm , and a long pass filter at 588 nm . Thin Layer Chromatography Analysis for Intact [ 3 H ] AEA . dCAD cells transfected with either mock or dynamin 2 - targeted siRNA were treated with [ 3 H ] AEA ( 1 nM ) for 5 min and washed three times with KRH containing 1 % bovine serum albumin . After the removal of the wash buffer , transfected dCAD cells were exposed to 1 ml of lysis buffer ( 0 . 5 % Triton X - 100 , 50 mM Tris - HCl , 150 mM NaCl , and 5 mM EDTA ) on ice for 20 min . The cells were then scraped and added to a 15 - ml centrifuge tube kept on ice that con - tained 1 . 5 ml of 2 : 1 CHCl 3 / MeOH ( v / v ) and 8 (cid:2) l of 1 M HCl . The aqueous and organic phases were mixed by vortexing and then separated by centrifugation ( 3000 g for 15 min at 4\u00b0C ) . The organic layer was removed from each sample , placed in a clean glass tube , and 1 ml of ice - cold CHCl 3 was added to the remaining aqueous layer . The phases were again mixed by vortexing and separated by centrifugation ( 3000 g for 15 min at 4\u00b0C ) . The resulting organic layer was removed and combined with the previously extracted organic layer . The solvent from the sample was evaporated under a stream of argon . The tubes were capped and kept overnight at (cid:3) 20\u00b0C . The concentrated samples were resuspended in 52 (cid:2) l of ice - cold CHCl 3 and then divided into two spots of 20 (cid:2) l each onto 20 - (cid:2) 20 - cm glass - backed silica gel thin layer chromatography ( TLC ) plates . The mobile phase in which the TLC plates were developed consisted of 11 : 5 ethyl acetate / isooctane saturated with H 2 O and acetic acid . After the plates were dried , the lipids were visualized with iodine vapor . The location of the [ 3 H ] AEA spot was identified by comparison with the unlabeled AEA standard . The silica corresponding to the AEA spots was then scraped off of the TLC plate into scintillation vials and allowed to equilibrate for 72 h in scintillation fluid . Radio - activity was then quantified by liquid scintillation counting . Results Endocannabinoid Uptake Properties of dCAD and cCAD Cells . Differentiated and undifferentiated CAD ( dCAD and cCAD , respectively ) cells displayed robust uptake of radio - labeled anandamide . The accumulation of AEA by CAD cells was inhibited by the AEA uptake inhibitor AM404 , with a K i value in the low micromolar to high nanomolar range ( Fig . 1A ) . SKM 4 - 45 - 1 is a fluorescent analog of AEA ( Muthian et al . , 2000 ) . Previous results using AM404 indicate that SKM 4 - 45 - 1 is internalized by the same mechanism as AEA in C6 glioma cells ( Muthian et al . , 2000 ) . SKM 4 - 45 - 1 internalization in CAD cells was also inhibited by AM404 , with K i values similar to those previously reported for inhibition of [ 3 H ] AEA uptake ( Fig . 1B ) ( Muthian et al . , 2000 ) . Furthermore , Western blot analysis of whole cell lysates from dCAD and cCAD cells revealed that both cell types express the AEA - metabolizing enzyme FAAH ( Fig . 2A ) . FAAH activity assays performed on CAD cell homog - enates revealed FAAH activity that was inhibited in the pres - ence of the FAAH inhibitor MAFP ( Fig . 2B ) . FAAH has been implicated in playing an important role in AEA uptake ( Day et al . , 2001 ; Glaser et al . , 2003 ) . We were surprised to find that although FAAH was present in CAD cells , inhibition of FAAH by phenylmethylsulfonyl fluoride did not reduce AEA uptake in CAD cells ( data not shown ) . This finding suggests that FAAH may not contribute to AEA uptake in these neuronal cells . Effect of Nystatin / Progesterone Pretreatment on AEA Uptake . As described above , both cCAD and dCAD cells display endocannabinoid properties . However , dCAD cells are morphologically similar to primary neuronal cell cultures . Thus , dCAD cells were used for our additional stud - ies to assess the role that endocytosis might play in the AEA uptake by a neuronal cell line . dCAD cells were treated with nystatin and progesterone to deplete membrane cholesterol levels and thereby disrupt detergent - resistant lipid raft do - mains . Based on the results from previous studies with RBL - 2H3 cells ( McFarland et al . , 2004 ) , we expected that this treatment would decrease the specific uptake of AEA by dCAD cells . Indeed , the disruption of detergent - resistant membrane microdomains in dCAD cells resulted in an (cid:4) 50 % decrease in the specific uptake of [ 3 H ] AEA ( Fig . 3 ) . Control experiments confirmed that nystatin and progesterone treat - Fig . 1 . Dose - dependent inhibition of AEA uptake and SKM 4 - 45 - 1 inter - nalization by AM404 . Both dCAD ( f ) and cCAD ( circf ) cells were treated with increasing concentrations of AM404 for 10 min at 37\u00b0C . A , after the incubation with AM404 , [ 3 H ] AEA was added at a final concentration of 1 nM as described under Materials and Methods . Data shown are means (cid:5) S . E . M . and are representative of four separate experiments performed in triplicate . K i values ( means (cid:5) S . E . M . ; n (cid:6) 3 ) were 850 (cid:5) 600 nM ( cCAD ) and 930 (cid:5) 150 nM ( dCAD ) . The cpm values for cCAD cells were total (cid:6) 6440 (cid:5) 914 and nonspecific (cid:6) 850 (cid:5) 86 . The cpm values for dCAD cells were total (cid:6) 5124 (cid:5) 1456 and nonspecific (cid:6) 794 (cid:5) 35 . B , cells were treated with SKM 4 - 45 - 1 ( 25 (cid:2) M ) and increasing concentrations of AM404 for 10 min , and fluorescence was determined on a FUSION microplate reader ( PerkinElmer Life and Analytical Sciences ) . IC 50 val - ues were converted to K i values using the Cheng - Prusoff equation ( K m value for SKM 4 - 45 - 1 (cid:6) 16 (cid:2) M ) . K i values ( means (cid:5) S . E . M . ; n (cid:6) 3 ) were 24 (cid:5) 12 (cid:2) M ( cCAD ) and 16 (cid:5) 10 (cid:2) M ( dCAD ) . Data shown are represen - tative of three separate experiments performed in duplicate . Arbitrary fluorescence unit values for cCAD cells were total (cid:6) 13 , 858 (cid:5) 1856 , nonspecific (cid:6) 7521 (cid:5) 573 , and background (cid:6) 4313 (cid:5) 625 . Arbitrary fluorescence unit values for dCAD cells were total (cid:6) 23 , 137 (cid:5) 6590 , nonspecific (cid:6) 11 , 309 (cid:5) 4024 , and background (cid:6) 2006 (cid:5) 74 . AEA Uptake via Endocytosis in dCAD Cells 103 a t A SP ET J ou r n a l s on M a r c h 17 , 2015 m o l ph a r m . a s p e tj ou r n a l s . o r g D o w n l o a d e d fr o m ment reduced membrane cholesterol by approximately 50 % , consistent with the effect on [ 3 H ] AEA uptake ( Fig . 3 , inset ) . RNAi - Mediated Knockdown of Endocytic Machin - ery . The studies described above with cholesterol depletion , along with our previous work ( McFarland et al . , 2004 ) , suggested that AEA uptake occurs via a caveolae - related endocytic process . The nonspecific nature of the cholesterol depletion treatments limits interpretation of our results . Therefore , we assessed the role of clathrin - independent ver - sus clathrin - dependent endocytosis of AEA in dCAD cells using molecular inhibitors . Specifically , we used an RNAi approach to knockdown expression of either the (cid:1) 2 subunit of the AP2 complex or dynamin 2 . Knockdown of the (cid:1) 2 subunit of the AP2 complex should only inhibit clathrin - mediated endocytosis and have no effect on clathrin - independent , or caveolae - related , endocytic processes ( Johannes and Lamaze , 2002 ; Huang et al . , 2004 ) . Dynamin 2 , however , is a GTPase that has been shown to play a role in both clathrin - mediated and clathrin - independent endocytosis ( Altschuler et al . , 1998 ; Maxfield and McGraw , 2004 ) . After transfection with siRNA oligonucleotides using Lipofectamine 2000 ( Invitro - gen ) , the cells were incubated for 72 h to allow for the deg - radation of resident protein . Western blot analysis confirmed that we had successfully reduced the expression of both pro - teins by 70 to 90 % ( Fig . 4 ) . Control transfection experiments using green fluorescent protein ( pEGFP - N1 ; Clontech , Moun - tain View , CA ) confirmed transfection efficiency in CAD cells to be greater than 80 % ( data not shown ) , consistent with our observed knockdown effects . To verify the effects of both (cid:1) 2 subunit and dynamin 2 knock - down , we used the Alexa Fluor 488 - conjugated probes trans - ferrin ( 488 - Tf ) and the B subunit of cholera toxin ( 488 - CT ) , which are endocytosed by clathrin - dependent and lipid raft / caveolae - related endocytic processes , respectively . As was ex - pected , after a 10 - min treatment with either probe at 37\u00b0C , RNAi knockdown of the (cid:1) 2 subunit of the AP2 complex signif - icantly reduced the internalization of 488 - Tf ( Fig . 5 ) but had no effect on 488 - CT endocytosis in dCAD cells compared with con - trol ( Fig . 6 ) . In the case of dynamin 2 , RNAi knockdown inhib - ited internalization of both 488 - Tf ( Fig . 5 ) and 488 - CT ( Fig . 6 ) in dCAD cells . These control experiments confirm that by knock - ing down expression of the (cid:1) 2 subunit of the AP2 complex we are only inhibiting clathrin - mediated endocytosis ( i . e . , 488 - Tf internalization ) , whereas knockdown of dynamin 2 inhibits both clathrin - dependent endocytosis as well as a lipid raft - related endocytic process that is clathrin - independent . AEA Uptake after RNAi Transfection . Our results above showing AM404 inhibition of both [ 3 H ] AEA and SKM 4 - 45 - 1 internalization ( Fig . 1 ) suggest that both compounds are trans - ported by a common process in the CAD cells . Because the fluorescence associated with SKM 4 - 45 - 1 is only observed after internalization of the compound , it is a true marker of actual uptake into the cell ( Muthian et al . , 2000 ) . To assess the effects that RNAi knockdown of endocytic machinery might have on Fig . 2 . Expression of FAAH in CAD cells . A , cCAD and dCAD cells express FAAH . Western blots were performed as described under Mate - rials and Methods . Data are representative of three separate experi - ments . B , cCAD and dCAD cells were homogenized and then incubated at 37\u00b0C for 10 min . Cell homogenates were then incubated with 5 nM [ 3 H ] AEA in the presence or absence of 500 nM MAFP . After the incuba - tion , the reaction was terminated , and the aqueous and organic phases were separated . Bars represent AEA - derived [ 3 H ] ethanolamine present in the aqueous phase ( i . e . , metabolized AEA ) after the termination of the reaction as a direct measure of FAAH activity . Data shown are means (cid:5) S . D . of three separate experiments performed in duplicate . Fig . 3 . Effect of lipid raft disruption on AEA uptake in dCAD cells . dCAD cells were pretreated for 30 min with 25 (cid:2) g / ml nystatin and 10 (cid:2) g / ml progesterone ( NP treatment ) . After NP treatment , AEA uptake assays were performed in 1 (cid:2) KRH at 37\u00b0C in the presence or absence of 100 (cid:2) M AM404 to define nonspecific transport . Transport assays were performed for 5 min with 1 nM [ 3 H ] AEA as described under Materials and Methods . Data are means (cid:5) S . E . M . combined from three separate experiments performed in triplicate . The cpm values for control cells were total (cid:6) 5755 (cid:5) 118 and nonspecific (cid:6) 502 (cid:5) 41 . The cpm values for NP - treated cells were total (cid:6) 3117 (cid:5) 40 and nonspecific (cid:6) 729 (cid:5) 61 . Inset , effect of NP treatment on membrane cholesterol content . Cholesterol content in CAD cell membranes was determined as described under Materials and Methods . Data are means (cid:5) S . E . M . from two separate experiments per - formed in triplicate . (cid:1) , p (cid:7) 0 . 05 . Fig . 4 . RNAi - mediated knockdown of the (cid:1) 2 subunit of the AP2 complex and dynamin 2 . dCAD cells were transfected with Stealth RNAi ( Invitro - gen ) using Lipofectamine 2000 ( Invitrogen ) and then incubated at 37\u00b0C in serum - free media for 72 h to allow for protein degradation . The expres - sion of both the (cid:1) 2 subunit of the AP2 complex and the protein dynamin 2 was decreased by (cid:8) 90 % compared with mock siRNA ( nonsense se - quence ) transfection . siRNA oligonucleotides had no effect on the level of control protein (cid:1) - actin ( data not shown ) . To control for protein loading , equal amounts of total protein ( 20 (cid:2) g ) from the cell lysates were added to each well for Western blot analysis as described under Materials and Methods . Both proteins resolved at the expected molecular weights . Data are representative of three separate experiments . 104 McFarland et al . a t A SP ET J ou r n a l s on M a r c h 17 , 2015 m o l ph a r m . a s p e tj ou r n a l s . o r g D o w n l o a d e d fr o m AEA uptake , cells were treated with 25 (cid:2) M SKM 4 - 45 - 1 at 37\u00b0C for 10 min after knockdown of the expression of either the (cid:1) 2 subunit of the AP2 complex or dynamin 2 . Compared with control , RNAi knockdown of the (cid:1) 2 subunit of the AP2 complex caused no observable change in the internalization of SKM 4 - 45 - 1 ( Fig . 7 ) . However , knockdown of dynamin 2 expression abolished the SKM 4 - 45 - 1 fluorescence accumulation in dCAD cells ( Fig . 7 ) . In addition to examining SKM 4 - 45 - 1 accumulation , we investigated the effects of RNAi - mediated knockdown of both the (cid:1) 2 subunit of the AP2 complex and dynamin 2 on the specific uptake of [ 3 H ] AEA . We were surprised to find that neither the (cid:1) 2 subunit of the AP2 complex nor dynamin 2 knockdown displayed any significant effect on the specific uptake of [ 3 H ] AEA ( Fig . 8 ) . To explain the potentially con - flicting result between [ 3 H ] AEA and SKM 4 - 45 - 1 internaliza - tion , we hypothesized that dynamin 2 siRNA - transfected dCAD cells might be accumulating more [ 3 H ] AEA at the plasma membrane instead of internalizing the AEA . If this were true , then FAAH would be unable to metabolize the [ 3 H ] AEA that has been retained at the plasma membrane . Such an effect would be revealed by an increase in intact [ 3 H ] AEA in the dynamin 2 siRNA - transfected cells after a 5 - min uptake assay . dCAD cells transfected with either mock or dynamin 2 siRNA were treated with [ 3 H ] AEA for 5 min in the presence or absence of AM404 . After [ 3 H ] AEA treatment , each treat - ment group of cells was lysed , and the lipids were extracted using chloroform / methanol . TLC was performed on the lipid Fig . 6 . Effect of RNAi knockdown of dynamin 2 and the (cid:1) 2 subunit of the AP2 complex on the internalization of 488 - CT in dCAD cells . Seventy - two hours after transfection with either mock , (cid:1) 2 subunit , or dynamin 2 siRNA , dCAD cells were incubated with 488 - CT for 10 min and imaged by confocal microscopy . Dynamin 2 siRNA - transfected cells showed a reduc - tion of 488 - CT internalization ( indicated by arrows ) compared with mock - transfected dCAD cells . (cid:1) 2 subunit siRNA - transfected cells did not display any difference in 488 - CT internalization compared with mock - transfected dCAD cells , confirming the specificity of the treatments . Images were obtained using confocal microscopy as described under Ma - terials and Methods . The internal fluorescence intensity of the intracel - lular space of all cells in the field of view was quantified using Meta - Morph software ( bottom right ) . Fifteen to 20 cells per experiment were analyzed . Data are representative of three separate experiments . Aster - isks denote a significant difference in intracellular mean fluorescence intensity compared with mock - transfected cells ( analysis of variance and Dunnett\u2019s multiple comparison test ; p (cid:7) 0 . 05 ) . Fig . 5 . Effect of RNAi knockdown of dynamin 2 and the (cid:1) 2 subunit of the AP2 complex on the internalization of 488 - Tf in dCAD cells . Seventy - two hours after transfection with either mock , (cid:1) 2 subunit , or dynamin 2 siRNA , dCAD cells were incubated with 488 - Tf for 10 min and imaged by confocal microscopy . Both (cid:1) 2 subunit - and dynamin 2 siRNA - transfected cells showed a reduction of 488 - Tf internalization ( indicated by arrows ) compared with mock - transfected dCAD cells . Images were obtained using confocal microscopy as described under Materials and Methods . The internal fluorescence intensity of the intracellular space of all cells in the field of view was quantified using MetaMorph software ( Molecular De - vices , Sunnyvale , CA ) ( bottom right ) . Fifteen to 20 cells per experiment were analyzed . Data are representative of three separate experiments . Asterisks denote a significant difference in intracellular mean internal fluorescence intensity compared with mock - transfected cells ( analysis of variance and Dunnett\u2019s multiple comparison test ; p (cid:7) 0 . 01 ) . Fig . 7 . SKM 4 - 45 - 1 internalization by dCAD cells after siRNA transfec - tion . dCAD cells were treated with SKM 4 - 45 - 1 ( 25 (cid:2) M ) for 10 min at 72 h after transfection with mock , (cid:1) 2 subunit , or dynamin 2 siRNA . Knock - down of the (cid:1) 2 subunit of the AP2 complex had no effect on the SKM 4 - 45 - 1 - derived fluorescence signal compared with mock . Knockdown of dynamin 2 in dCAD cells significantly decreased the accumulation of SKM 4 - 45 - 1 - derived fluorescence . Images were obtained using confocal microscopy as described under Materials and Methods . Phase contrast images are shown below each corresponding fluorescence image . Data are representative of three separate experiments . AEA Uptake via Endocytosis in dCAD Cells 105 a t A SP ET J ou r n a l s on M a r c h 17 , 2015 m o l ph a r m . a s p e tj ou r n a l s . o r g D o w n l o a d e d fr o m samples of each treatment group , and the levels of intact [ 3 H ] AEA were quantified by scintillation counting . TLC anal - ysis revealed that the specific levels of intact [ 3 H ] AEA were approximately 2 - fold higher in dCAD cells in which dynamin 2 had been knocked down compared with mock - transfected dCAD cells ( Fig . 9A ) . One explanation for the above - mentioned results could be that the knockdown of dynamin 2 altered FAAH activ - ity . Examination of FAAH protein by Western blot analysis ( Fig . 9B ) and FAAH activity ( Fig . 9C ) after dynamin 2 knockdown indicated that the dynamin 2 knockdown had no detectable effect on FAAH . The lack of effect on FAAH and actin levels also confirms the specificity of our dynamin 2 siRNA ( Fig . 9B ) . Discussion dCAD cells represent a biochemically and morphologically neuronal cell line that rapidly accumulates the endocannabi - noid AEA and displays robust FAAH activity . Thus , dCAD cells may present a novel system useful in the study of endocannabinoid biosynthesis and inactivation . Pharmacological inhibition of endocytosis in RBL - 2H3 cells suggests that an endocytic process is involved in AEA uptake and that this process is clathrin - independent , potentially involving detergent - resistant membrane microdomains such as lipid rafts ( McFarland et al . , 2004 ) . Our present results confirm these findings in the neuronal dCAD cell line as well . Pretreatment with nystatin / progesterone to disrupt detergent - resistant membrane microdomains significantly reduced the specific uptake of [ 3 H ] AEA by dCAD cells by (cid:4) 50 % . An alter - native explanation for our results with nystatin / progesterone could be that the treatment is generally toxic to the cells . Be - cause a major factor determining nonspecific uptake relates to cell number , it is unlikely that the brief treatments caused a decrease in cell viability because nonspecific uptake of AEA was unaltered by the treatment ( data not shown ) . We recognize that there is a lack of specificity associated with the chemical dis - ruption of lipid rafts . Thus , a means of specifically disrupting endocytic processes without causing the potential nonspecific effects associated with the chemical disruption of lipid rafts was needed to assess the role that endocytosis may play in the cellular uptake of AEA . The (cid:1) 2 subunit of the AP2 complex allows for the associa - tion of this complex with clathrin and is necessary for clath - rin - mediated endocytosis to occur ( Huang et al . , 2004 ) . Transferrin is internalized by a clathrin - mediated endocytic process ( Johannes and Lamaze , 2002 ) . As expected , RNAi knockdown of the (cid:1) 2 subunit of the AP2 complex inhibited transferrin internalization in dCAD cells but did not disrupt Fig . 8 . [ 3 H ] AEA uptake by dCAD cells after siRNA transfection . Seventy - two hours after transfection with mock , dynamin 2 , or (cid:1) 2 subunit siRNA , [ 3 H ] AEA uptake assays were performed in 1 (cid:2) KRH at 37\u00b0C as described under Materials and Methods . AM404 ( 100 (cid:2) M ) was used to define nonspecific transport . Data represent mean (cid:5) S . E . M . for three separate experiments performed in triplicate . The cpm values for mock cells were total (cid:6) 2887 (cid:5) 136 and nonspecific (cid:6) 1125 (cid:5) 128 . The cpm values for dynamin 2 knockdown cells were total (cid:6) 2973 (cid:5) 89 and nonspecific (cid:6) 905 (cid:5) 83 . The cpm values for (cid:1) 2 subunit knockdown cells were total (cid:6) 2638 (cid:5) 104 and nonspecific (cid:6) 777 (cid:5) 27 . Fig . 9 . Metabolism of [ 3 H ] AEA in dynamin 2 siRNA - transfected dCAD cells . A , 72 h after transfection with either mock or dynamin 2 siRNA , dCAD cells were treated with [ 3 H ] AEA for 5 min and then lysed . Cellular lipids were extracted from the cell lysates using chloroform / methanol and analyzed by TLC for intact [ 3 H ] AEA . Nonspecific uptake was determined using 100 (cid:2) M AM404 . The graph represents data from three separate experiments . Statistical analysis was performed using a one - sample t test comparing with the value of 100 ( mock control ) . (cid:1) , p (cid:7) 0 . 05 . The cpm values from the TLC analysis for mock cells were total (cid:6) 2986 (cid:5) 795 and nonspecific (cid:6) 2243 (cid:5) 407 . The cpm values from the TLC analysis for dynamin 2 knockdown cells were total (cid:6) 3257 (cid:5) 545 and nonspecific (cid:6) 1793 (cid:5) 198 . B , FAAH expression after RNAi - mediated knockdown of dynamin 2 . Knockdown of dynamin 2 in cCAD cells was performed as described under Materials and Methods except siLentFect Lipid Reagent ( Bio - Rad ) was used as the transfection reagent for 48 h with 10 nM siRNA oligo . Western blots were performed using the polyclonal dynamin 2 antibody , monoclonal FAAH antibody ( 1 : 1000 ; Abnova Corporation , Tai - pei City , Taiwan ) , monoclonal actin antibody ( 1 : 2000 ; Sigma - Aldrich , St . Louis , MO ) . C , FAAH activity after RNAi - mediated knockdown of dy - namin 2 . FAAH activity assays were performed on whole cell lysates as described under Materials and Methods . Data shown represent means (cid:5) S . D . from two experiments performed in quadruplicate . 106 McFarland et al . a t A SP ET J ou r n a l s on M a r c h 17 , 2015 m o l ph a r m . a s p e tj ou r n a l s . o r g D o w n l o a d e d fr o m the cellular accumulation of the cholera toxin B subunit or SKM 4 - 45 - 1 , the fluorescently tagged analog of AEA . The cholera toxin B subunit is known to be internalized via a clathrin - independent ( lipid raft - related ) endocytic process ( Johannes and Lamaze , 2002 ) . Whereas its role in neurons is not completely clear , the small GTPase dynamin 2 is probably involved in both clath - rin - dependent and clathrin - independent endocytic recycling ( Altschuler et al . , 1998 ; Maxfield and McGraw , 2004 ) . Thus , both transferrin and cholera toxin should require dynamin 2 expression for internalization to occur . RNAi knockdown of dynamin 2 significantly reduced both transferrin and cholera toxin accumulation by dCAD cells . Furthermore , dynamin 2 knockdown in dCAD cells inhibited the cellular internaliza - tion of SKM 4 - 45 - 1 . Because the fluorescence associated with SKM 4 - 45 - 1 is only observed after internalization of the compound , it can be used as a marker for actual uptake ( Muthian et al . , 2000 ) . One caveat of using only [ 3 H ] AEA uptake is that conventional uptake assays do not differenti - ate between the tritium bound to the cells or membranes and the AEA that is actually transported into the cells . Taken together , our data suggest that actual internalization of AEA by dCAD cells occurs via a clathrin - independent endocytic process . Furthermore , dynamin 2 seems to play a role in the cellular uptake of AEA by the neuronal dCAD cells . Internal - ization of the B subunit of cholera toxin occurs by a lipid raft - mediated endocytic process ( Johannes and Lamaze , 2002 ) . Dynamin 2 knockdown significantly reduced 488 - CT internalization in dCAD cells consistent with this dynamin 2 - dependent , clathrin - independent endocytic process involv - ing detergent - resistant membrane microdomains . An alter - native approach to the knockdown of dynamin 2 or the (cid:1) subunit of the AP2 complex could be to use siRNA targeted to clathrin itself . We were concerned about potential membrane effects reported with clathrin knockdowns and as such , we opted to focus on proteins downstream of clathrin in the endocytic pathways ( Hinrichsen et al . , 2003 , 2006 ) . Dynamin 2 knockdown displayed a profound inhibitory effect on SKM 4 - 45 - 1 internalization in dCAD cells , and yet it did not have any observable effect on the specific uptake of [ 3 H ] AEA . This discrepancy may be reconciled with the obser - vation that [ 3 H ] AEA uptake assays cannot discriminate be - tween [ 3 H ] AEA that has actually been internalized and [ 3 H ] AEA that is binding to a target on the plasma membrane . Interestingly , a study conducted by Schmid and colleagues showed that expression of dominant - negative mutants of dy - namin proteins inhibited endocytic processes but had no ef - fect on endosomal recycling of membrane components to the cell surface ( Altschuler et al . , 1998 ) . dCAD cells that have lost functioning dynamin 2 protein during the 72 - h incuba - tion after transfection with siRNA may be accumulating the AEA carrier in the detergent - resistant membrane microdo - mains . Precedence for such an effect comes from studies showing that expression of dominant - negative dynamin 2 mutant causes an accumulation of transferrin receptor at the cell surface ( Altschuler et al . , 1998 ) . A similar phenomenon could be occurring with the putative AEA carrier after dy - namin 2 knockdown . The accumulation of detergent - resis - tant membrane components at the cell surface with dynamin 2 knockdown could create a scenario in which dCAD cells , although they are not internalizing [ 3 H ] AEA , are accumulat - ing [ 3 H ] AEA at the membrane . This effect would be due to more [ 3 H ] AEA being able to associate with the plasma mem - brane that is now enriched with the putative carrier protein . This experimental limitation is not an issue in the case of SKM 4 - 45 - 1 internalization assays . Fluorescence from SKM 4 - 45 - 1 is not observed until after the compound is trans - ported into the cell , and the fluorescent component of the molecule is liberated by esterases in the cell ( Muthian et al . , 2000 ) . Thus , excess SKM 4 - 45 - 1 bound only at the membrane is not detectable . We propose that knockdown of dynamin 2 in dCAD cells causes an accumulation of the AEA carrier in detergent - resistant membrane domains at the cell surface . There is no observable difference between the [ 3 H ] AEA uptake in mock siRNA - transfected and dynamin 2 siRNA - transfected dCAD cells because more [ 3 H ] AEA is able to associate with the plasma membrane carrier in the dynamin 2 siRNA - transfected cells . If this is true , then the majority of the radioactivity obtained from the dynamin 2 siRNA - transfected cells should be represented by intact [ 3 H ] AEA . This is because AEA that is only associating with the cell surface and is not internalized will not be available to FAAH for metabolism . Indeed , TLC analysis of lipid extracts from dynamin 2 siRNA - transfected dCAD cells that were treated with [ 3 H ] AEA for 5 min revealed levels of intact [ 3 H ] AEA that were 2 - fold greater than the levels of [ 3 H ] AEA extracted from mock siRNA - transfected dCAD cells treated in the same manner . Furthermore , our results indicate that dy - namin 2 knockdown does not alter FAAH protein or activity . These data support the hypothesis that dynamin 2 siRNA - transfected dCAD cells are able to accumulate more [ 3 H ] AEA on their cell surface than mock - transfected dCAD cells due to an enrichment of the putative AEA carrier in the microdo - mains . If this is the case , then [ 3 H ] AEA uptake assays may not be an ideal method for characterizing AEA internalization after the knockdown of proteins , such as dynamin 2 , that are impor - tant in recycling events that maintain the composition of the plasma membrane . One alternative explanation for our results involves a possible role for the CB1 cannabinoid receptor in the uptake process . CB1 receptors have been shown to signal through lipid raft domains ( Bari et al . , 2005 ) ; thus , binding to CB1 receptors and subsequent internalization would be one possible mechanism for AEA uptake . Unlike RBL - 2H3 cells , we do not expect that dCAD cells contain caveolin - 1 , but we have preliminary evidence that dCAD cells express caveolin - 3 ( data not shown ) . Thus , caveo - lae - related endocytic processes may still be present in neu - rons and are dependent on intact detergent - resistant mem - brane microdomains . In summary , our data , combined with previous experiments in RBL - 2H3 cells ( McFarland et al . , 2004 ) , offer evidence that endocannabinoids are internalized by a lipid raft - or caveolae - related endocytic process . We speculate that the yet unidentified anandamide transporter will in fact be enriched in lipid rafts and functionally partic - ipate in endocytosis . Acknowledgments We thank Jason Parish for helpful discussion and technical assis - tance regarding TLC analysis of intact [ 3 H ] AEA and the Purdue Cytomics Laboratory for technical assistance with microscopic imaging . AEA Uptake via Endocytosis in dCAD Cells 107 a t A SP ET J ou r n a l s on M a r c h 17 , 2015 m o l ph a r m . a s p e tj ou r n a l s . o r g D o w n l o a d e d fr o m References Altschuler Y , Barbas SM , Terlecky LJ , Tang K , Hardy S , Mostov KE , and Schmid SL ( 1998 ) Redundant and distinct functions for dynamin - 1 and dynamin - 2 isoforms . J Cell Biol 143 : 1871 \u2013 1881 . Bari M , Battista N , Fezza F , Finazzi - Agro A , and Maccarrone M ( 2005 ) Lipid rafts control signaling of type - 1 cannabinoid receptors in neuronal cells . J Biol Chem 280 : 12212 \u2013 12220 . Cameron PL , Ruffin JW , Bollag R , Rasmussen H , and Cameron RS ( 1997 ) Identifi - cation of caveolin and caveolin - related proteins in the brain . J Neurosci 17 : 9520 \u2013 9535 . Cravatt BF , Giang DK , Mayfield SP , Boger DL , Lerner RA , and Gilula NB ( 1996 ) Molecular characterization of an enzyme that degrades neuromodulatory fatty - acid amides . Nature 384 : 83 \u2013 87 . Day TA , Rakhshan F , Deutsch DG , and Barker EL ( 2001 ) Role of fatty acid amide hydrolase in the transport of the endogenous cannabinoid anandamide . Mol Phar - macol 59 : 1369 \u2013 1375 . Deutsch DG and Chin SA ( 1993 ) Enzymatic synthesis and degradation of anandam - ide , a cannabinoid receptor agonist . Biochem Pharmacol 46 : 791 \u2013 796 . Di Marzo V , Bisogno T , and De Petrocellis L ( 2001 ) Anandamide : some like it hot . Trends Pharmacol Sci 22 : 346 \u2013 349 . Di Marzo V , Fontana A , Cadas H , Schinelli S , Cimino G , Schwartz JC , and Piomelli D ( 1994 ) Formation and inactivation of endogenous cannabinoid anandamide in central neurons . Nature 372 : 686 \u2013 691 . Glaser ST , Abumrad NA , Fatade F , Kaczocha M , Studholme KM , and Deutsch DG ( 2003 ) Evidence against the presence of an anandamide transporter . Proc Natl Acad Sci U S A 100 : 4269 \u2013 4274 . HashemiSH , LiJY , FaigleR , andDahlstromA ( 2003 ) Adrenergicdifferentiationand SSR2 ( a ) receptor expression in CAD - cells cultured in serum - free medium . Neuro - chem Int 42 : 9 \u2013 17 . Hinrichsen L , Harborth J , Andrees L , Weber K , and Ungewickell EJ ( 2003 ) Effect of clathrin heavy chain - and (cid:3) - adaptin - specific small inhibitory RNAs on endocytic accessory proteins and receptor trafficking in HeLa cells . J Biol Chem 278 : 45160 \u2013 45170 . Hinrichsen L , Meyerholz A , Groos S , and Ungewickell EJ ( 2006 ) Bending a mem - brane : how clathrin affects budding . Proc Natl Acad Sci U S A 103 : 8715 \u2013 8720 . Huang F , Khvorova A , Marshall W , and Sorkin A ( 2004 ) Analysis of clathrin - mediated endocytosis of epidermal growth factor receptor by RNA interference . J Biol Chem 279 : 16657 \u2013 16661 . Johannes L and Lamaze C ( 2002 ) Clathrin - dependent or not : is it still a question ? Traffic 3 : 443 \u2013 451 . Lazaroff M ( 1996 ) A CNS catecholaminergic cell line expresses voltage - gated cur - rents . J Membr Biol 151 : 279 \u2013 291 . Lazaroff M , Qi Y , and Chikaraishi DM ( 1998 ) Differentiation of a catecholaminergic CNS cell line modifies tyrosine hydroxylase transcriptional regulation . J Neuro - chem 71 : 51 \u2013 59 . Matsuda LA , Lolait SJ , Brownstein MJ , Young AC , and Bonner TI ( 1990 ) Structure of a cannabinoid receptor and functional expression of the cloned cDNA . Nature 346 : 561 \u2013 564 . Maxfield FR and McGraw TE ( 2004 ) Endocytic recycling . Nat Rev Mol Cell Biol 5 : 121 \u2013 132 . McFarland MJ , Porter AC , Rakhshan F , Rawat DS , Gibbs RA , and Barker EL ( 2004 ) A role for caveolae / lipid rafts in the uptake and recycling of the endogenous cannabinoid anandamide . J Biol Chem 279 : 41991 \u2013 41997 . Millard EE , Gale SE , Dudley N , Zhang J , Schaffer JE , and Ory DS ( 2005 ) The sterol - sensing domain of the Niemann - Pick C1 ( NPC1 ) protein regulates traffick - ing of low density lipoprotein cholesterol . J Biol Chem 280 : 28581 \u2013 28590 . Munro S , Thomas KL , and Abu - Shaar M ( 1993 ) Molecular characterization of a peripheral receptor for cannabinoids . Nature 365 : 61 \u2013 65 . Muthian S , Nithipatikom K , Campbell WB , and Hillard CJ ( 2000 ) Synthesis and characterization of a fluorescent substrate for the N - arachidonoylethanolamine ( anandamide ) transmembrane carrier . J Pharmacol Exp Ther 293 : 289 \u2013 295 . Qi Y , Wang JK , McMillian M , and Chikaraishi DM ( 1997 ) Characterization of a CNS cell line , CAD , in which morphological differentiation is initiated by serum depri - vation . J Neurosci 17 : 1217 \u2013 1225 . Suri C , Fung BP , Tischler AS , and Chikaraishi DM ( 1993 ) Catecholaminergic cell lines from the brain and adrenal glands of tyrosine hydroxylase - SV40 T antigen transgenic mice . J Neurosci 13 : 1280 \u2013 1291 . Voets T and Nilius B ( 2003 ) TRPs make sense . J Membr Biol 192 : 1 \u2013 8 . Wang H and Oxford GS ( 2000 ) Voltage - dependent ion channels in CAD cells : a catecholaminergic neuronal line that exhibits inducible differentiation . J Neuro - physiol 84 : 2888 \u2013 2895 . Address correspondence to : Dr . Eric L . Barker , Department of Medicinal Chemistry and Molecular Pharmacology , Purdue University , 575 Stadium Mall Dr . , West Lafayette , IN 47907 - 2091 . E - mail : ericb @ pharmacy . purdue . edu 108 McFarland et al . a t A SP ET J ou r n a l s on M a r c h 17 , 2015 m o l ph a r m . a s p e tj ou r n a l s . o r g D o w n l o a d e d fr o m",
    "pandit2020force": "Force and phosphate release from Arp2 / 3 complex promote dissociation of actin filament branches Nandan G . Pandit a , b \ue840 , Wenxiang Cao a , Jeffrey Bibeau a , Eric M . Johnson - Chavarria a , Edwin W . Taylor a , Thomas D . Pollard a , b , c , d \ue840 , and Enrique M . De La Cruz a , b , 1 a Department of Molecular Biophysics and Biochemistry , Yale University , New Haven , CT 06520 ; b Program in Physical and Engineering Biology , Yale University , New Haven , CT 06520 ; c Department of Molecular , Cellular , and Developmental Biology , Yale University , New Haven , CT 06520 ; and d Department of Cell Biology , Yale University , New Haven , CT 06520 Edited by Harry Higgs , Geisel School of Medicine at Dartmouth , Hanover , NH , and accepted by Editorial Board Member Yale E . Goldman April 16 , 2020 ( received for review June 28 , 2019 ) Networks of branched actin filaments formed by Arp2 / 3 complex generate and experience mechanical forces during essential cellular functions , including cell motility and endocytosis . External forces regulate the assembly and architecture of branched actin networks both in vitro and in cells . Considerably less is known about how mechanical forces influence the disassembly of actin filament networks , specifically , the dissociation of branches . We used microfluidics to apply force to branches formed from purified muscle actin and fission yeast Arp2 / 3 complex and observed debranching events in real time with total internal reflection fluorescence microscopy . Low forces in the range of 0 pN to 2 pN on branches accelerated their dissociation from mother filaments more than two orders of magnitude , from hours to < 1 min . Nei - ther force on the mother filament nor thermal fluctuations in mother filament shape influenced debranching . Arp2 / 3 complex at branch junctions adopts two distinct mechanical states with different sensitivities to force , which we name \u201c young / strong \u201d and \u201c old / weak . \u201d The \u201c young / strong \u201d state 1 has adenosine 5 \u2032 - diphosphate ( ADP ) \u2212 P i bound to Arp2 / 3 complex . Phosphate re - lease converts Arp2 / 3 complex into the \u201c old / weak \u201d state 2 with bound ADP , which is 20 times more sensitive to force than state 1 . Branches with ADP \u2212 Arp2 / 3 complex are more sensitive to debranching by fission yeast GMF ( glia maturation factor ) than branches with ADP \u2212 P i \u2212 Arp2 / 3 complex . These findings suggest that aging of branch junctions by phosphate release from Arp2 / 3 complex and mechanical forces contribute to disassembling \u201c old \u201d actin filament branches in cells . actin | Arp2 / 3 complex | branched filament | debranching | force A rp2 / 3 complex forms networks of branched actin filaments that generate and sustain mechanical forces that power cell motility , endocytosis , and vesicle trafficking ( 1 , 2 ) . Membrane - bound proteins , called nucleation - promoting factors , such as WASP activate Arp2 / 3 complex , which then nucleates a branch when it binds to the side of a preexisting \u201c mother \u201d filament ( 3 ) . The new \u201c daughter \u201d filament elongates and pushes against the membrane until it is capped . All of the filaments , including branches formed by Arp2 / 3 complex , must disassemble for recycling to form new filaments and branches . Similar to actin , Arp2 / 3 complex is an ATPase ( 4 \u2013 6 ) . Hydrolysis of bound ATP and subsequent phosphate release have been implicated in controlling branched network dy - namics ( 5 , 7 \u2013 10 ) , but mechanistic details are lacking . The assembly and architecture of branched actin networks are sensitive to force in vitro and in cells ( 11 , 12 ) . Under load , branched actin networks assembled from purified proteins grow more slowly and with a higher branch density ( 13 , 14 ) , but how these mechanical forces directly affect the biochemical interac - tions of branched actin network protein components has not been firmly established . Similar to measurements with purified protein components , branched actin networks in cells respond to external load by increasing density of branched filaments while also reorganizing relative to the membrane ( 15 ) . Networks of branched actin filaments turn over much more rapidly in cells than in vitro when assembled from purified proteins ( 16 \u2013 19 ) . The regulatory proteins cofilin and glia matu - ration factor ( GMF ) accelerate debranching and have been im - plicated in accelerating branched network remodeling and turnover ( 20 \u2013 24 ) . Although not investigated previously , mechanical forces may also affect network disassembly through debranching . We report that mechanical forces promote dissociation of branches formed by Arp2 / 3 complex and that phosphate bound to Arp2 / 3 complex regulates the sensitivity to force . Phosphate release from Arp 2 / 3 complex at branch junctions also regulates debranching by GMF . Thus , phosphate release from the Arp2 / 3 complex could target \u201c older \u201d adenosine 5 \u2032 - diphosphate ( ADP ) \u2212 Arp2 / 3 branches for dissociation while sparing \u201c younger \u201d branches with ADP \u2212 P i \u2212 Arp2 / 3 complex . Results Microfluidics Assay to Measure Dissociation of Branches Formed by Arp2 / 3 Complex under Force . We used fluid flowing through a microfluidics apparatus to apply force to actin filament branches formed by purified fission yeast Arp2 / 3 complex and muscle actin monomers ( Fig . 1 A ) as we observed the dissociation of the branches by fluorescence microscopy . Starting with short fila - ment seeds tethered to the surface of the slide , we assembled branched filaments from purified ATP - actin monomers and Significance Arp2 / 3 complex is an ATPase that binds to the side of a pre - existing actin filament and nucleates an actin filament branch . Growing branched networks experience variable resistance and respond by adapting growth speed , power , and architec - ture . How force influences the dissociation of actin filament branches was not known . We used microfluidics to show that mechanical force promotes the dissociation of actin filament branches and that Arp2 / 3 complex adopts two distinct me - chanical states with different responses to force . Phosphate release from Arp2 / 3 complex increases the sensitivity to both force and the debranching protein GMF . Thus , phosphate re - lease from Arp2 / 3 complex may regulate debranching by force and debranching proteins . Author contributions : N . G . P . and E . M . D . L . C . designed research ; N . G . P . performed re - search ; N . G . P . , W . C . , J . B . , E . M . J . - C . , E . W . T . , T . D . P . , and E . M . D . L . C . contributed new re - agents / analytic tools ; N . G . P . , W . C . , J . B . , E . W . T . , T . D . P . , and E . M . D . L . C . analyzed data ; and N . G . P . , W . C . , J . B . , E . W . T . , T . D . P . , and E . M . D . L . C . wrote the paper . The authors declare no competing interest . This article is a PNAS Direct Submission . H . H . is a guest editor invited by the Editorial Board . Published under the PNAS license . 1 To whom correspondence may be addressed . Email : enrique . delacruz @ yale . edu . Thisarticlecontainssupportinginformationonlineathttps : / / www . pnas . org / lookup / suppl / doi : 10 . 1073 / pnas . 1911183117 / - / DCSupplemental . First published May 27 , 2020 . www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1911183117 PNAS | June 16 , 2020 | vol . 117 | no . 24 | 13519 \u2013 13528 B I O P H Y S I C S A ND C O M P U T A T I O N A L B I O L O G Y Arp2 / 3 complex for 2 min to 4 min . After washing out the soluble proteins , the filaments were allowed to \u201c age \u201d for an additional , variable time with very slow fluid flow . Thereafter , we started the debranching process under constant force by flowing buffer over the surface at a rate of 2 \u03bc L \u00b7 min \u2212 1 to 500 \u03bc L \u00b7 min \u2212 1 until the end of the experiment , while we recorded a series of images . The seeds were immobilized on the slide , but both the mother fila - ments and branches were free to fluctuate ( Fig . 1 A ) . Rapid flow rates flattened branches against the mother fila - ments until they dissociated ( Fig . 1 B and Movies S1 \u2013 S3 ) . Neither the angle of the applied force ( relative to mother filament ori - entation ) , the tension in the mother filament , nor fluctuations in mother filament shape had a strong influence on debranching ( SI Appendix , Fig . S1 and Movies S6 and S7 ) . We added a snap tag to the Arpc5 subunit of Arp2 / 3 complex for labeling with Alexa 488 and simultaneous viewing with Alexa 647 - labeled actin . Force dissociated Arp2 / 3 complex and the daughter filament concurrently , within the 0 . 1 - s time resolution of our imaging ( Fig . 1 D and Movies S4 and S5 ) . Buffer flowed across the sample at 500 \u03bc m \u00b7 s \u2212 1 dissociated both the daughter filament and Arp2 / 3 complex from the field of view by the next frame , making it impossible to determine whether the labeled Arp2 / 3 complex remained bound to the dissociated daughter filament . Piconewton Forces Decreased the Time for Branch Dissociation from Hours to < 1 Min . Branches formed by ATP \u2212 Arp2 / 3 com - plex were stable for many minutes without buffer flow ( Fig . 2 A ) but had a higher probability of dissociating when subjected to the forces produced by the range of buffer flow rates in our exper - iments ( Fig . 2 A ) . The following sections explain how the time course of debranching depends on the applied force and how long the newly formed branches were aged after assembly . At a given flow rate , the force on a branch scales with its length . In some experiments , we show the force on individual branches ( SI Appendix , Fig . S2 ) , but , in most experiments , we report flow rates from which we calculate the average force on branches ( e . g . , Fig . 2 A ) . In samples of branches formed by ATP \u2212 Arp2 / 3 complex and aged for 30 min , the time course of debranching followed a single exponential that depended on the applied force ( Fig . 2 A ) . The observed lifetimes of these branches decreased with force ( Fig . 2 D ) , suggesting a slip bond behavior , so we used the Bell \u2019 s equation ( 25 ) ( Eq . 1 ) to estimate the force sensitivity , \u03c4 obs = \u03c4 0 e \u2212 FdkBT . [ 1 ] Here k B is the Boltzmann constant , T is the absolute tempera - ture , F is force , \u03c4 0 is the branch lifetime in the absence of force , and d is the characteristic distance to the transition state ( 26 ) . The value of d is typically considered a force sensitivity param - eter for bond rupture . An alternative estimate of force depen - dence is the force that reduces the branch lifetime by half ( half - force , F 1 / 2 = 0 . 693 k B T / d ) , which we estimate to be 0 . 054 ( \u00b1 0 . 008 ) pN . Without force , the branch lifetime ( \u03c4 0 ) , estimated from ex - trapolation of the fit of the force dependence ( Fig . 2 D ) , was 106 ( \u00b1 8 ) min [ Table 1 ; k 2 \u2248 1 / \u03c4 0 = 0 . 01 ( \u00b1 0 . 0007 ) min \u2212 1 ] , indicating - 2 0 2 4 time ( s ) - 20 0 20 40 60 80 100 y t i s ne t n I e c ne cs e r ou l F ARP2 / 3 BackgroundDebranching - 2 0 2 4 time ( s ) 0 100 200 300 400 500 y t i s ne t n I e c ne cs e r ou l F Actin Branch Debranching A w o l F 2 \u00b5 L m i n - 1 ( S l o w ) w o l F 500 \u00b5 L m i n - 1 ( F a s t ) 0 s 30 s 300 s B C No Flow Flow D Debranching Glass coated with sparse biotin - BSA ( blue ) to anchor the mother filament via neutravidin ( orange ) Biotinylated Alexa - 568 actin filament seed bound to surface Freely fluctuating Alexa - 647 actin filaments grown from seed Branch formed by Arp2 / 3 complex Pointed End Barbed End Arp2 / 3 complex ointed End Arp2 / 3complexx Fig . 1 . Microfluidics assay to measure Arp2 / 3 complex debranching under force . ( A ) Diagram showing a short segment of an actin filament containing 10 % biotinylated and 15 % Alexa 568 - labeled ( red ) actin subunits immobilized on the neutravidin - coated surface . The surface is passivated with 0 . 2 % tween ( illustrated with gray vertical lines ) . This seed was elongated at its barbed end with 1 . 5 \u03bc M 15 % Alexa 647 - labeled Mg - ATP - actin ( green ) , and Arp2 / 3 complex formed a branch with Alexa 647 - labeled Mg - ATP - actin . The green filaments fluctuate freely and are subject to viscous drag forces applied by fluid flow . ( B ) TIRF microscopy images of representative branched filaments under slow flow ( 2 \u03bc L \u00b7 min \u2212 1 , \u223c 0 . 004 pN of force for a 1 . 5 - \u03bc m branch ; Top ) and fast flow ( 500 \u03bc L \u00b7 min \u2212 1 , \u223c 1 . 02 pN of force for a 1 . 5 - \u03bc m branch ; Bottom ) . Branches are aligned in the direction of flow . ( Scale bar , 1 \u03bc m . ) ( C ) The Arpc5 subunit of the Arp2 / 3 complex was labeled with Alexa 488 via snap tag and tracked during debranching . Time - lapse images with the actin filaments represented in red and the Alexa 488 \u2212 Arp2 / 3 complex located at the junction of the daughter branch and mother filament represented in green . ( Scale bar , 1 \u03bc m . ) ( D ) Top shows the spatially integrated fluorescence intensity of actin at a branch junction as a function of time , used to determine the ob - served debranching event time ( t = 0 ) . Middle shows a kymograph measured across a branched actin filament . Bottom shows the time course of spatially integrated fluorescence intensity of Arpc5 subunit at a branch junction with time aligned to its corresponding actin frame . The fluorescence intensity from Arp2 / 3 complex reproducibly decreased in a single step for all 12 debranching events observed . 13520 | www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1911183117 Pandit et al . that sustained pN forces reduce \u03c4 obs more than two orders of magnitude to < 30 s . The force dependence of individual branch lifetimes ( SI Appendix , Fig . S2 ) yielded a half - force ( F 1 / 2 ) of 0 . 054 ( \u00b1 0 . 008 ) pN and \u03c4 0 = 96 ( \u00b1 6 ) min , comparable to the values estimated from exponential fits of debranching time courses ( Fig . 2 A and D ) . Branches Assembled from ATP \u2212 Arp2 / 3 Complex Dissociate Faster under Force as They Age . As they age , branches assembled from ATP \u2212 Arp2 / 3 complex dissociated progressively faster under force produced by a buffer flow rate of 500 \u03bc L \u00b7 min \u2212 1 ( Fig . 2 B ) . Immediately after the 2 . 6 - min assembly reaction , when the dom - inant nucleotide bound to both Arp2 / 3 complex and the actin filaments is expected to be ADP \u2212 P i , the time course of branch dissociation followed a single exponential with a slow observed lifetime ( \u03c4 s , F , where the subscript F indicates under flow force ) of 3 . 08 ( \u00b1 0 . 05 ) min corresponding to a first - order rate constant k s , F of 0 . 32 ( \u00b1 0 . 005 ) min \u2212 1 . After aging for 30 min , when both Arp2 / 3 complex and the subunits in the actin filaments are expected to have bound ADP , the time course of debranching also followed a single exponential with a 20 - fold shorter lifetime ( \u03c4 f , F ) of 0 . 15 ( \u00b1 0 . 01 ) min ( k f , F = 6 . 67 ( \u00b1 0 . 44 ) min \u2212 1 ) . At intermediate aging times , the time courses followed double exponentials ( Fig . 2 B ) with distinguishable fast and slow phases , indicating that ( at least ) two reactions contributed to debranching . To evaluate how force affects slow and fast debranching , we aged samples for a short time ( \u223c 4 min ) , so the sample would include mixtures of branches with Arp2 / 3 complex with bound A B C D Fig . 2 . Effects of mechanical force and nucleotide bound to Arp2 / 3 complex on the time course of dissociation of actin filament branches . Arp2 / 3 complex branches were assembled in the flow chamber before applying flow as described in Materials and Methods . In A and C , the force on each observed branch was calculated from its length and the flow rate , and then binned at the indicated average forces values ( see Materials and Methods ) . ( A ) The effect of force on the time course of dissociation of actin filament branches formed by ATP - actin monomers and ATP \u2212 Arp2 / 3 complex and aged for 30 min , when most branches had ADP bound to Arp2 / 3 complex . The fraction of branches remaining is plotted . Smooth curves are the best fits of single exponentials to the data . Each trace includes at least 14 branches . ( B ) Dependence of the time course of dissociation of branches formed with ATP - actin monomers and ATP \u2212 Arp2 / 3 complex with different aging times ( 2 . 6 , 4 , 12 , and 30 min ) and the presence of BeF x . For all time courses , 500 \u03bc L \u00b7 min \u2212 1 of buffer flow was applied to the branches for debranching , producing a force of \u223c 1 pN for a branch of 1 . 5 \u03bc m . The force on each branch was not calculated for the data shown . Smooth curves are the best global fits of double exponentials to the data for aging branches and yielded two shared rate constants for debranching : slow k s , F = 0 . 32 ( \u00b1 0 . 005 ) min \u2212 1 and fast k f , F = 6 . 67 ( \u00b1 0 . 44 ) min \u2212 1 . Smooth curves are the best single exponential fits to the data for branches aged for 4 and 30 min with BeF x with lifetimes of 13 . 9 ( \u00b1 0 . 2 ) min \u2212 1 for the sample aged for \u223c 4 min and 14 . 8 ( \u00b1 0 . 09 ) min \u2212 1 for the sample aged for 30 min . Fig . 4 D presents the fractional amplitudes obtained from the double exponential fits . ( C ) The effect of force on the time course of the dissociation of actin filament branches formed by ATP - actin monomers and ATP \u2212 Arp2 / 3 complex and aged for \u223c 4 min using the same data collection and analysis methods as in A . The smooth curves are best fits of the data to single ( F \u2264 0 . 6 pN ) or double ( F > 0 . 6 pN ) exponentials . The fractional amplitudes of the slow phase in the time courses that follow the double exponentials are 0 . 71 \u00b1 0 . 05 ( F = 0 . 98 pN ) , 0 . 39 \u00b1 0 . 06 ( F = 1 . 29 pN ) , 0 . 81 \u00b1 0 . 05 ( F = 1 . 60 pN ) , 0 . 38 \u00b1 0 . 02 ( F = 1 . 93 pN ) , and 0 . 17 \u00b1 0 . 02 ( F = 2 . 86 pN ) . Each trace includes at least 19 branches . ( Inset ) Time courses of branch dissociation under a range of forces for branches formed from ATP \u2212 Arp2 / 3 complex in the presence of 2 mM BeF x and aged for \u223c 4 min . Each trace includes at least 13 branches . The smooth curves are single exponential fits to the time courses . Branches dissociate slowly under 0 . 2 pN of force , so the debranching time course cannot be reliably fitted to obtain the branch lifetime . ( D ) Dependence of branch lifetimes on force for four different conditions : ( filled black circles ) branches formed from ATP \u2212 Arp2 / 3 complex and aged for 30 min to form branches with ADP \u2212 Arp2 / 3 complex ( time courses in A ) ; ( filled red squares ) branches formed from ADP \u2212 BeF x \u2212 Arp2 / 3 complex and aged for \u223c 4 min ( time courses in C , Inset ) ; ( filled pink triangles ) slow debranching phase of branches formed from ATP \u2212 Arp2 / 3 complex and aged for \u223c 4 min ( ADP \u2212 P i branch population ; time courses in C ) ; and ( filled blue triangle ) fast debranching phase of branches formed from ATP \u2212 Arp2 / 3 complex and aged for \u223c 4 min ( ADP branch population ; time courses in C ) . The uncertainty bars for all data represent the SDs from the fits to exponentials shown in A and C . The smooth black curve is the best single exponential fit ( Eq . 1 ) to the ADP \u2212 Arp2 / 3 complex debranching data points , yielding a half - force ( F 1 / 2 ) of 0 . 054 ( \u00b1 0 . 008 ) pN and branch lifetime in the absence of force ( \u03c4 0 ) of 106 ( \u00b1 8 ) min ( observed rate constant = \u03c4 0 \u2212 1 = 0 . 01 ( \u00b1 0 . 007 ) min \u2212 1 ) . Inset shows that 1 ) fast phase lifetimes ( blue triangles ; 0 . 3 min to 0 . 8 min at F > 1 pN ) differ from the slow phase lifetimes ( pink triangles ; 2 min to 4 min at F > 1 pN ; t = 6 . 85 , one - tail t critical = 2 . 13 and P = 0 . 001 by Welch \u2019 s unequal variances t test ) ; 2 ) the slow phase lifetimes ( pink triangles ; 2 min to 4 min at F > 1 pN ) differ from the debranching lifetimes with BeF x ( red squares ; 8 min to 9 min at F > 1 pN ; t = 6 . 65 , one - tail t critical = 2 . 02 and P = 0 . 0006 by Welch \u2019 s unequal variances t test ) ; and 3 ) the fast phase lifetimes ( blue triangles ) do not differ significantly from the debranching with ATP aged for 30 min lifetimes ( black circles ) at F > 1 pN ( t = \u2212 0 . 86 , two - tail t critical = 2 . 45 and P = 0 . 42 by Welch \u2019 s unequal variances t test ) . Pandit et al . PNAS | June 16 , 2020 | vol . 117 | no . 24 | 13521 B I O P H Y S I C S A ND C O M P U T A T I O N A L B I O L O G Y ADP \u2212 P i and ADP and thus exhibit both slow and fast phases of dissociation over a range of forces applied with different flow rates . However , the actual aging times prior to making obser - vations varied . This uncertainty influenced the observed ampli - tudes but not the observed lifetimes , so we only analyzed the lifetimes . The debranching time courses at forces of > 1 pN fol - lowed double exponential decays ( Fig . 2 C ) , yielding the force dependence of the slow and fast phase lifetimes ( Fig . 2 D and Inset ; triangles ) . The lifetimes of the slow phase ( 2 - to 4 - min range ) differed from the fast phase ( 0 . 3 min to 0 . 8 min ) in this force range ( > 1 pN ) , but neither depended strongly on the ap - plied force . At low forces of < 0 . 6 pN , dissociation of branches formed from ATP \u2212 Arp2 / 3 complex followed single exponentials , with lifetimes similar to branches with ADP \u2212 Arp2 / 3 complex ( Fig . 2 D ) . The debranching model presented below accounts for these different time courses . The fast phase lifetime behaved simi - lar to ADP \u2212 Arp2 / 3 complex branches ( i . e . , assembled from ATP \u2212 Arp2 / 3 complex and aged for 30 min ) , consistent with fast debranching population corresponding to ADP \u2212 Arp2 / 3 complex branches . We discuss the force dependence of the slow phase below . The Nucleotide Bound to Arp2 / 3 Complex Influences the Sensitivity of Branches to Force . ATP \u2212 Arp2 / 3 complex hydrolyzes the bound nucleotide upon or soon after branch formation ( 5 , 7 ) followed by dissociation of the \u03b3 - phosphate with an unknown rate con - stant . Therefore , branches formed by ATP \u2212 Arp2 / 3 complex rapidly transition to the ADP \u2212 P i state for an unknown duration before the release of P i . Actin filament branches formed by Arp2 / 3 complex with mutations that slow ATP hydrolysis are more stable than those formed by native Arp2 / 3 complex , so Martin et al . ( 10 ) proposed that the hydrolysis of ATP bound to Arp2 / 3 complex destabilizes branches . We performed a series of experiments to determine whether the nucleotide state of Arp2 / 3 complex could explain the slow and fast debranching states ( Fig . 2 B and C ) . Like Acanthamoeba Arp2 / 3 complex ( 5 ) , Schizosaccharomyces pombe ADP \u2212 Arp2 / 3 complex did not form branches from ATP - actin monomers nor did the ADP \u2212 Arp2 / 3 complex form branches in the presence of 20 mM phosphate ( SI Appendix , Fig . S3 ) . A likely interpretation of this behavior is that the transient intermediate ADP \u2212 P i \u2212 Arp2 / 3 complex is competent to form branches , but ADP \u2212 Arp2 / 3 com - plex binds P i very weakly ( K d > 20 mM ) . On the other hand , S . pombe ADP \u2212 Arp2 / 3 complex formed branches with 2 mM be - ryllium fluoride ( BeF x ) in the buffer ( SI Appendix , Fig . S3 ) , as originally shown for Acanthamoeba Arp2 / 3 complex ( 5 ) . ATP \u2212 Arp2 / 3 complex also forms branches with 2 mM BeF x in the buffer . Branches with ADP \u2212 BeF x \u2212 Arp2 / 3 complex dissociate with indistinguishable time courses whether assembled from ATP \u2212 Arp2 / 3 complex or ADP \u2212 Arp2 / 3 complex in the presence of 2 mM BeF x and aged for \u223c 4 min ( SI Appendix , Fig . S4 ) . We assume that , in both cases , BeF x binds to the ADP \u2212 Arp2 / 3 complex and stabilizes conformations similar to ADP \u2212 P i , as established for actin ( 27 , 28 ) , so we used ADP \u2212 BeF x \u2212 Arp2 / 3 complex branches made with either method . ADP \u2212 BeF x \u2212 Arp2 / 3 complex branches aged for 4 or 30 min in the presence of 2 mM BeF x dissociate with similar time courses and follow single exponen - tials ( Fig . 2 B ) . Thus , branches did not convert from the slowly to the rapidly dissociating state when aged with BeF x . Force is required to dissociate , within our experimental ob - servation period , ADP \u2212 BeF x \u2212 Arp2 / 3 complex branches formed from ATP \u2212 Arp2 / 3 complex in the presence of 2 mM BeF x ( Fig . 2 C , Inset ) . Under low forces ( i . e . , < < 0 . 6 pN ) , < 20 % of branches with ADP \u2212 BeF x \u2212 Arp2 / 3 complex dissociated within 5 h , so branching lifetimes could not be measured reliably . With forces > 0 . 6 pN , branches dissociated with single exponential time courses and observed rate constants ( inverse of lifetimes ) that depended on the applied force ( Fig . 2 C , Inset ) . Therefore , branches with ADP \u2212 BeF x \u2212 Arp2 / 3 complex were more stable under force than branches with ADP \u2212 Arp2 / 3 complex ( formed from ATP \u2212 Arp2 / 3 complex and aged for 30 min ; Fig . 2 D ) . Under a given force , the lifetimes of ADP \u2212 BeF x \u2212 Arp2 / 3 complex branches were longer ( \u223c twofold ) than the slow debranch - ing phase of ATP \u2212 Arp2 / 3 complex branches that had been aged for \u223c 4 min ( Fig . 2 D ) . The observed rate constant of the slow phase debranching reflects the sum of the ADP \u2212 P i state debranching rate constant and the rate constant for conversion of the ADP \u2212 P i \u2212 Arp2 / 3 complex to the ADP state ( see Discussion ) . It is therefore expected to be faster than ADP \u2212 BeF x \u2212 Arp2 / 3 complex debranching , which converts more slowly to ADP \u2212 Arp2 / 3 complex ( Fig . 2 B ) . The Nucleotide State of the Actin Filaments Does Not Influence Debranching . We assembled branches from ATP \u2212 Arp2 / 3 com - plex with ADP or ADP \u2212 P i bound to the subunits in the actin filaments and observed that all had similar time courses of debranching after aging for a given time ( SI Appendix , Fig . S5 ) . Samples with ADP - actin filaments were prepared by assembly from ATP - actin monomers and ATP - actin Arp2 / 3 complex fol - lowed by aging for 30 min . Samples with ADP \u2212 P i actin filaments were prepared by assembly from ATP - actin monomers and ATP \u2212 Arp2 / 3 complex in buffer containing 20 mM phosphate followed by aging for 4 or 30 min ( SI Appendix , Fig . S5 ) . This concentration of phosphate is well above the K d for binding ADP - actin subunits ( 29 ) , so subunits in the filaments likely had ADP \u2212 P i in the active site , assuming that Arp2 / 3 complex binding does not dramatically change the affinity of actin for phosphate . Samples assembled from ATP - actin monomers and ATP \u2212 Arp2 / 3 complex followed by \u223c 4 min of aging had a mixture of ADP and ADP \u2212 P i nucleotide states for both actin subunits and Arp2 / 3 complex . Samples with AMPPNP actin ( a nonhydrolyzable an - alog of ATP ) were assembled from AMPPNP actin monomers and ATP - actin Arp2 / 3 complex in buffer containing 2 mM AMPPNP followed by aging for \u223c 4 min . Since AMPPNP \u2212 Arp2 / 3 complex does not form branches ( ref . 5 and SI Appendix , Fig . S3 ) , all of the branches formed from Arp2 / 3 complex with bound ATP , while the actin filaments had bound AMPPNP . A previous study of bovine Arp2 / 3 complex ( 16 ) reported more branches on mother filament segments with ADP \u2212 P i than segments with ADP . This difference was attributed to slower dissociation of branches aged in buffer with 25 mM phosphate allowing time for them to be stabilized by binding to the slide coated with N - ethylmaleimide ( NEM ) - myosin anchors . However , that study did not measure time course of the dissociation of branches , and we did not compare the rate of branch formation on ADP and ADP \u2212 P i mother filaments , so we do not know whether the sources of Arp2 / 3 complex , the presence of NEM - myosin an - chors on the surface , or other factors account for the apparent difference . Table 1 . Rate constants for branch formation and dissociation in the absence of force and related figures Conversion k conv ( min \u2212 1 ) State 1 debranching k 1 * ( min \u2212 1 ) State 2 debranching k 2 \u2020 ( min \u2212 1 ) Branch formation k form * ( \u03bc M \u2212 1 \u00b7 s \u2212 1 ) 0 . 14 \u00b1 0 . 03 0 . 012 0 . 01 \u00b1 0 . 007 0 . 02 Fig . 4 Fig . 4 Fig . 2 D Fig . 4 * The equations used to determine these parameters are approximations , so uncertainties are not reported . \u2020 From the fit of force - dependent debranching data to Eq . 1 ; k 2 = 1 / \u03c4 0 . The error was calculated from the SD \u03c4 0 in the fit . 13522 | www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1911183117 Pandit et al . Since the debranching kinetics are independent of the nucle - otide ( i . e . , ADP , ADP \u2212 P i , or AMPPPNP ) bound to the mother and daughter filaments , the stabilization of branches by BeF x ( Fig . 2 B \u2013 D ) is likely due to BeF x bound to Arp2 / 3 complex in the branch junction . This suggests that the ADP \u2212 P i \u2212 Arp2 / 3 complex intermediate has different mechanical properties than the ADP \u2212 Arp2 / 3 complex after phosphate dissociation . The lack of an effect of 20 mM phosphate on debranching ( SI Appendix , Fig . S5 ) is consistent with the affinity of S . pombe ADP \u2212 Arp2 / 3 complexes in branch junctions for phosphate being very weak ( K d > 20 mM ) . Force Promotes but BeF x Inhibits Debranching by GMF . GMF ( 20 \u2013 22 ) promotes the dissociation of actin filament branches without applied force , so we measured how the concentration of fission yeast GMF influences the rates of dissociation of actin filament branches with ADP \u2212 BeF x \u2212 Arp2 / 3 complex or ADP \u2212 Arp2 / 3 complex at a low buffer flow rate ( 15 \u03bc L \u00b7 min \u2212 1 ) . This flow rate exerts very little force on branches and had little or no effect on debranching ( Fig . 2 ) . Concentrations of GMF up to 1 \u03bc M did not dissociate branches with ADP \u2212 BeF x \u2212 Arp2 / 3 complex ( Fig . 3 A ) , but nanomolar con - centrations of GMF promoted dissociation of branches with ADP \u2212 Arp2 / 3 complex ( Fig . 3 ) . Time courses of dissociation of branches with ADP \u2212 Arp2 / 3 complex followed single exponentials ( Fig . 3 A ) with lifetimes ( \u03c4 obs ) that depended hyperbolically on the concentration of GMF ( Fig . 3 B ) . Force increased the rate at which GMF dissociated branches with ADP \u2212 Arp2 / 3 complex ( SI Appendix , Fig . S6 ) , but dissoci - ation was slower than predicted if GMF and force increased the rate of dissociation independently ( i . e . , their energetic contri - butions to debranching were additive ) . This raises the possibility that the reactions catalyzed by force and GMF are coupled and / or that debranching follows different pathways in the presence absence of force . Discussion Quantitative Analysis of the Two - State Model for Dissociation of Actin Filament Branches . Our experiments show that the nucleo - tide bound to Arp2 / 3 complex determines the sensitivity of branches to dissociation by physical force . We use a formal de - scription of a simple two - state model ( Fig . 4 A ) to analyze our data and estimate the rate constants for the three reactions and their sensitivities to force . Description of the model . The model ( Fig . 4 A ) assumes that branches form when ATP \u2212 Arp2 / 3 complex binds to the side of a mother filament and nucleates a daughter filament growing at its barbed end . Hydrolysis of ATP in the active sites of Acantha - moeba Arp2 / 3 complex is closely associated with nucleation of the daughter filament ( 5 , 7 ) , so we assume that state 1 branches have Arp2 / 3 complex with bound ADP \u2212 P i . Dissociation of the \u03b3 - phosphate with a rate constant k conv converts branch state 1 to branch state 2 with ADP bound to Arp2 / 3 complex . Branches in either state can dissociate from the mother filament . The model predicts that , overall , net debranching time courses follow dou - ble exponentials with two rate constants . One observed rate constant is the sum of the rate constants for state 1 debranching and conversion ( k obs , 1 ( F ) = k 1 ( F ) + k conv ( F ) ; ( F ) indicates optional force ) . The second observed rate constant corresponds to debranching of state 2 ( k obs , 2 ( F ) = k 2 ( F ) ; SI Appendix , Eqs . S25 and S29 ) . Force accelerates the dissociation of branches with either ADP \u2212 Arp2 / 3 and ADP \u2212 P i \u2212 Arp2 / 3 complex , but branches with ADP \u2212 P i or ADP \u2212 BeF x \u2212 Arp2 / 3 ( mimicking the ADP \u2212 P i state ) complex are much more resistant to force than branches with ADP \u2212 Arp2 / 3 complex . As branches age , phos - phate release converts slowly dissociating , \u201c young and strong \u201d state 1 branches with ADP \u2212 P i \u2212 Arp2 / 3 complex into rapidly dissociating , \u201c old and weak \u201d state 2 branches with ADP \u2212 Arp2 / 3 complex . Formulation of the model . We analyzed the experimental data us - ing a parallel debranching pathway ( Fig . 4 A and SI Appendix , Part 2 and Scheme S1 ) where \u201c young \u201d ( strong , state 1 ) branches convert to \u201c old \u201d ( weak , state 2 ) branches as they age ( Fig . 4 A ) . We fit the aging time dependence of the fractions of slowly and rapidly dissociating branches observed under force as the amplitudes ( A s , F and A f , F , respectively ; Fig . 4 C ) of double A B Fig . 3 . BeF x inhibits debranching by GMF . ( A ) Dependence of the time course of dissociation of branches with ADP \u2212 BeF x \u2212 Arp2 / 3 complex or ADP \u2212 Arp2 / 3 complex on the concentration of GMF . Branches with ADP \u2212 - BeF x \u2212 Arp2 / 3 complex were assembled in buffer containing 0 . 2 mM ATP , 2 mM BeSO 4 , and 10 mM NaF and aged for \u223c 4 min . Branches with ADP \u2212 Arp2 / 3 complex were assembled in buffer with 0 . 2 mM ATP and aged for 30 min to allow for ATP hydrolysis and phosphate dissociation ( Fig . 2 B ) . Debranching was initiated by flowing buffer with GMF at 15 \u03bc L \u00b7 min \u2212 1 and continued throughout the debranching measurements . The smooth curves are the best fits of single exponentials to the data , yielding the ( average ) branch lifetimes ; n = 30 branches for each trace . A concentration of 1 \u03bc M GMF did not dissociate branches with ADP \u2212 BeF x \u2212 Arp2 / 3 complex assembled from ATP - actin and ATP \u2212 Arp2 / 3 complex with BeF x and aged for \u223c 4 min ( open black circles ) . ( B ) Dependence of the lifetimes of branches with ADP \u2212 Arp2 / 3 complex on the concentration of GMF at a buffer flow rate of 15 \u03bc L \u00b7 min \u2212 1 . The line is the best fit of Eq . 3 to the data , yielding a GMF binding affinity ( K d , GMF ) of 40 ( \u00b1 10 ) nM and a maximum debranching rate constant ( k diss , GMF ) of 0 . 31 ( \u00b1 0 . 05 ) min \u2212 1 . At this low flow rate , branches with ADP \u2212 Arp2 / 3 complex ( without GMF ) dissociated with a rate constant ( k diss , 0 = 0 . 014 \u00b1 0 . 0002 min \u2212 1 , indicated by an open circle ) similar to that under zero force ( Fig . 2 B and Table 1 ) . The uncertainty bars are within the data points and represent the SDs of lifetimes in the best single exponential fits of time traces in A . Pandit et al . PNAS | June 16 , 2020 | vol . 117 | no . 24 | 13523 B I O P H Y S I C S A ND C O M P U T A T I O N A L B I O L O G Y exponential fits to the following functions ( SI Appendix , Eqs . S34 and S37 ) : A s , F = be \u2212 ( k conv + k 1 \u2212 k 2 ) ( t age \u2212 t 1 ) 1 + ae \u2212 ( k 1 + k conv \u2212 k 2 ) ( t age \u2212 t 1 ) A f , F = 1 + ( a \u2212 b ) e \u2212 ( k conv + k 1 \u2212 k 2 ) ( t age \u2212 t 1 ) 1 + ae \u2212 ( k 1 + k conv \u2212 k 2 ) ( t age \u2212 t 1 ) = 1 \u2212 A s , F [ 2 ] where t 1 is the 2 . 6 min during which branches formed from ATP \u2212 Arp2 / 3 complex and ATP - actin monomers , t age is aging time , the intrinsic debranching rate constants are k 1 for state 1 and k 2 for state 2 in the absence of force , and k conv is the rate constant for conversion of state 1 to state 2 in the absence of force ( Fig . 4 A and SI Appendix , Part 2 and Scheme S1 ) . The constants a and b are unitless composites of rate constants and initial branch concentrations defined in SI Appendix , Part 2 and Scheme S1 ( SI Appendix , Eqs . S39 and S40 ) . The constant a is force independent and determined only by intrinsic debranching in the absence of an applied force ( SI Appendix , Eq . S39 ) . The constant b is a function of debranching and conversion under force ( SI Appendix , Eq . S40 ) , yielding force - dependent ampli - tudes ( A s , F and A f , F ) when debranching and / or conversion are force dependent . Estimation of the rate constants for branch formation and phosphate dissociation . Global analysis of the data ( Figs . 2 B and 4 C and SI Appendix , Part 2 ) yielded an overall pseudo \u2212 first - order associa - tion rate constant for branch formation of 0 . 12 min \u2212 1 ( SI Ap - pendix , Part 2 ) . Given 100 nM Arp2 / 3 complex used in our experiments , we estimate the second - order association rate constant for Arp2 / 3 complex binding to mother filaments and subsequent activation ( k form ) to be \u223c 0 . 02 \u03bc M \u2212 1 \u00b7 s \u2212 1 ( Table 1 ) , consistent with published reports ( 30 \u2013 32 ) . The best fit of the fractional amplitudes to Eq . 2 ( Fig . 4 C ) yielded a rate constant ( k conv ) of 0 . 14 ( \u00b1 0 . 03 ) min \u2212 1 ( corre - sponding to a lifetime of > 7 min ) for conversion of state 1 branches to state 2 branches . We interpret the conversion re - action as the release of P i from the Arp2 / 3 complex , and thus interpret state 2 branches as ADP \u2212 Arp2 / 3 complex branches . This rate constant inferred for P i release from ADP - P i Arp2 / 3 complex is similar to the rate constant of \u223c 0 . 18 min \u2212 1 for P i release from actin filaments ( 29 ) . We can place a limit on the rate constant for phosphate binding to ADP - Arp2 / 3 complex in a branch junction ( k + P i ) . We interpret conversion to reflect P i release , so k \u2212 P i = k conv = 0 . 14 ( \u00b1 0 . 03 ) min \u2212 1 , and know that P i binds Arp2 / 3 complex in branches with a low affinity ( K d > 20 mM ; SI Appendix , Fig . S5 A ) . Therefore , the second order association rate constant of k + P i is less than 7 M \u2212 1 min \u2212 1 = 0 . 12 M \u2212 1 s \u2212 1 . This value is two orders of magnitude slower than P i binding to ADP - actin subunits in the interior of filaments [ 600 M \u2212 1 min \u2212 1 = 10 M \u2212 1 s \u2212 1 ( 29 ) ] . Estimation of the rate constants for branch dissociation at low force . The rate constant for dissociation of state 2 branches at zero force is k 2 \u2248 0 . 01 min \u2212 1 , based on the intercept of the force - dependence of ADP - Arp2 / 3 complex ( state 2 ) branch lifetimes ( Fig . 2 D ) . The rate constant for debranching from state 1 at zero force ( k 1 ) is \u223c 0 . 012 min \u2212 1 , from the analysis of the aging time - dependence of State 1 Arp2 / 3 - ADP - P i State 2 Arp2 / 3 - ADP Arp2 / 3 - ATP 0 50 100 150 200 250 300 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 state 1 ) de z il a m r on ( s eh c na r b # Aging time ( min ) state 2 net debranching 0 5 10 15 20 25 30 0 . 0 0 . 5 1 . 0 state 1 state 2 f o r m a t i on 0 10 20 30 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 e c r o f r ednu edu t il p m a . c a r F Aging time ( min ) slow / strong fast / weak A B C Fig . 4 . Model and simulations of the pathways of branch formation , aging , and debranching . ( A ) Schematic of our hypothesis . Formation of a branch by ATP \u2212 Arp2 / 3 complex ( red ) is coupled to hydrolysis of ATP bound to the Arps ( 5 , 7 ) with rate constant k form , yielding ADP \u2212 P i \u2212 Arp2 / 3 complex ( orange ) in state 1 . Irreversible phosphate dissociation with rate constant k conv converts state 1 to ADP \u2212 Arp2 / 3 complex ( blue ) in state 2 . Branches dissociate from mother filaments with rate constants k 1 for state 1 and k 2 for state 2 , both sensitive to force . ( B ) Simulated time course of the model showing how the populations of state 1 branches , state 2 branches , and dissociated branches evolve over time in the absence of force . We assumed that Mg - ATP - actin monomers and Mg - ATP \u2212 Arp2 / 3 complex formed branches for 2 . 6 min , when the free proteins were removed , and the reactions continued without ad - ditional branch formation . The red line represents the best single expo - nential fit to the observed debranching ( i . e . , combined from both states ) starting with a normalized value of 1 at the end of branch formation ( 2 . 6 min ) . The experimentally determined or estimated rate constants used in the simulation are k form [ Arp ] = 0 . 12 min \u2212 1 , k conv = 0 . 14 min \u2212 1 , k 1 = 0 . 012 min \u2212 1 , and k 2 = 0 . 01min \u2212 1 . ( C ) Aging time dependence of the fractional the slow and fast phase amplitudes in the debranching time courses under 500 \u03bc L \u00b7 min \u2212 1 of buffer flow , obtained from double exponential fits to the time courses ( Fig . 2 B ) . These data were used for extraction of fundamental rate constants in Table 1 . The best global fits of the two - state model ( Eq . 2 ) to the fractional amplitude data gave rate constants for conversion k conv of 0 . 14 min \u2212 1 , state 1 branch dissociation k 1 of 0 . 012 s \u2212 1 , and branch formation k form , of 0 . 02 \u03bc M \u2212 1 \u00b7 s \u2212 1 without force ( Table 1 ) . The one negative fast phase amplitude at 2 . 6 min results from a net increase in the state 2 branch population during debranching under force after 2 . 6 min aging time . The state 2 branch population is the net sum of depletion from debranching ( negative contri - bution to population , exponential decay ) and gain from conversion of state 1 branches ( positive contribution , exponential rise ) . For short aging times , little or no state 2 branches exist for depletion , and the conversion from state 1 branches , represented by an exponential rise ( negative amplitude ) , dominates the time course ( 54 , 55 ) . The uncertainty bars are SDs of the fractions of branches from the global double exponential fits of debranch - ing time courses with different aging times in Fig . 2 B . 13524 | www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1911183117 Pandit et al . time course of dissociation of branches ( Figs . 2 B and 4 C , Ta - ble 1 , and SI Appendix , Part 2 and Eq . S42 ) . As a result of these two rate constants being similar ( k 1 \u2248 k 2 ) , debranching without force follows a single exponential with a rate constant of \u223c k 2 ( SI Appendix , Part 2 and Eq . S25 and Fig . 4 B ) , even though three reactions ( conversion and debranching from state 1 and state 2 ) occur simultaneously . Time courses also follow single exponentials at low forces ( < 0 . 6 pN ) , independent of the aging time ( Fig . 2 C and D ) . This behavior arises because dissociation is much slower than con - version ( k 1 < < k conv ) , so ADP - P i branches convert to ADP be - fore dissociating . At forces > 1 pN , the observed dissociation time courses follow double exponentials because k conv < k 1 and debranching occurs from both ADP - P i and ADP branches . Estimation of the force sensitivity of the branch dissociation rate constants . Force increases the dissociation rate constants of both branch states , but has a larger effect on state 2 branches with ADP - Arp2 / 3 complexes than state 1 branches with ADP - P i - Arp2 / 3 complexes ( Fig . 2 D ) . The best global fits of the double expo - nential time courses at intermediate aging times yielded a \u223c 20 - fold difference in the lifetimes of the species that dissociated slowly ( k s , F = 0 . 32 ( \u00b1 0 . 005 ) min \u2212 1 ) and rapidly ( k f , F = 6 . 67 ( \u00b1 0 . 44 ) min \u2212 1 ) at a buffer flow rate of 500 \u03bc L min \u2212 1 ( Fig . 2 B ) . Under this force , the observed fast rate constant is the rate constant for the state 2 branch dissociation ( i . e . , k f , F = k 2 , F ) and the observed slow rate constant is the sum of the rate constants for dissociation for state 1 branches and conversion from state 1 to state 2 ( i . e . , k s , F = k 1 , F + k conv , F ; SI Appendix , Part 2 and Eqs . S29 and S41 ) . Force produced by a flow of 500 \u03bc L min \u2212 1 increased the ob - served dissociation of branches \u223c 670 - fold ( from k 2 = 0 . 01 to k 2 , F = 6 . 67 min \u2212 1 ) for those with ADP - Arp2 / 3 complex but only 32 - fold ( from k 1 = 0 . 012 to k 1 , F = k s , F \u2212 k conv , F < k s , F = 0 . 32 min \u2212 1 ; Fig . 2 B and SI Appendix , Fig . S5 ) for those with ADP - P i - Arp2 / 3 complex . This difference in force sensitivity of the two states explains the dramatic acceleration in overall debranching with aging ( Fig . 2 B ) . Furthermore , we estimate that force produced by 500 \u03bc L min \u2212 1 flow increased k conv only two - fold from k conv = 0 . 14 min \u2212 1 in the absence of force to k conv , F = k s , F \u2212 k 1 , F < k s , F = 0 . 32 min \u2212 1 in the presence of force ( SI Appendix , Fig . S5 ) . Implications of the Force Sensitivity of Branches for Their Turnover in Cells . Implications for debranching by myosin motor proteins . A sustained flow force of \u223c 1 pN dissociates ADP - Arp2 / 3 complex branches in < 30 s ( Fig . 2 D ) , raising the possibility that pN contractile forces generated by myosin motors could rapidly debranch and reorganize Arp2 / 3 complex networks . Myosin accelerates net - work disassembly and reorganization in both biomimetic systems ( 33 , 34 ) and in cells ( 35 , 36 ) . We note , however , that the work ( i . e . , energy term ) from the applied force determines the overall debranching acceleration , not the force per se . Implications of Nucleotide - Dependent Force Sensitivity of Arp2 / 3 Complex Debranching . \u201c Young \u201d branches with ADP \u2212 P i \u2212 Arp2 / 3 complex are more resistant to dissociation by force than \u201c old \u201d branches with ADP \u2212 Arp2 / 3 complex . Nucleation - promoting fac - tors associated with membranes ( plasma membrane , vesicles , in - tracellular bacteria ) activate Arp 2 / 3 complex , so \u201c young \u201d branches are located closer to these surfaces than \u201c old \u201d branches . If force is uniformly distributed across the filament network , old branches farthest from the surface with nucleation - promoting factors would be preferentially debranched . This differential response to force may contribute to the ob - served remodeling of cellular branched actin networks under load , including local changes in branch density ( 13 \u2013 15 ) . We apply a simple a minimal kinetic model ( SI Appendix , Part 3 ) to analyze how selective , force - sensitive debranching might influence the distribution of branches in actin networks at the leading edge of cells . The model assumes Arp 2 / 3 complex is activated uniformly at the membrane , such that the branch density distribution is also uniform , on average , along the plane parallel to the cell mem - brane at the leading edge . The model with the experimental pa - rameters determined here predicts that the density of branches decays exponentially ( see also refs . 37 \u2013 39 ) along the axis per - pendicular to the membrane , toward cell interior in both the absence and presence of force , but preferential debranching of old Arp 2 / 3 complexes shortens the branch density decay length and can , under some conditions , increase the local branch density near the membrane ( SI Appendix , Part 3 and Figs . S9 and S10 ) . Thus , external force on the leading edge of cells fa - vors debranching of old branches that have migrated toward the cell interior over young branches still near the membrane and can thus contribute to spatial and vectorial network turnover . External force may also influence binding of cofilin ( 40 ) and GMF ( Fig . 3 ; discussed below ) , both of which display debranching activity ( 20 , 24 ) . GMF Selectively Dissociates ADP \u2212 Arp2 / 3 Complex Branches . Fission yeast GMF increases the rate constant for dissociation of branches with ADP \u2212 Arp 2 / 3 complex up to \u223c 20 - fold , but does not disso - ciate branches with ADP \u2212 BeF x \u2212 Arp 2 / 3 complex ( Fig . 3 ) . The following parallel reactions describe GMF - catalyzed dissociation of branches with ADP \u2212 Arp2 / 3 complex : Arp + GMF \u21cc Arp - GMF \u2193 k diss , 0 \u2193 k diss , GMF , debranched Scheme 1 where k diss , 0 and k diss , GMF are the dissociation rate constants of branches formed by Arp2 / 3 complex alone ( Arp ) and GMF - bound Arp2 / 3 complex ( Arp - GMF ) , and K d , GMF is the affinity of GMF for Arp 2 / 3 complex in a branch junction . This scheme assumes that GMF binding equilibrates rapidly compared to the rate of debranching ( and k + GMF [ GMF ] , k \u2212 GMF > > k diss , 0 , k diss , GMF ) and that [ GMF ] > > [ Arp2 / 3 complex ] . Debranching from the parallel pathway in Scheme 1 follows a single exponential ( 41 ) . The dependence of the observed branch lifetime ( \u03c4 obs ) on the concentration of GMF can be expressed in two ways , first , \u03c4 obs = \u03c4 0 \u2212 ( \u03c4 0 \u2212 \u03c4 GMF ) [ GMF ] K d , GMF ( \u03c4 GMF \u03c4 0 ) + [ GMF ] . [ 3 ] Alternatively , the relationships can also be expressed in terms of observed rate constants ( 41 , 42 ) , k obs = k 0 + ( k GMF \u2212 k 0 ) [ GMF ] K d , GMF + [ GMF ] . [ 4 ] The best fit of Eq . 3 to the dependence of the observed branch lifetime ( \u03c4 obs ) on the concentration of GMF ( Fig . 3 B ) yielded an apparent affinity of GMF for ADP \u2212 Arp2 / 3 complex ( K d , GMF ) of 40 ( \u00b1 10 ) nM and a lifetime of GMF bound to ADP \u2212 Arp2 / 3 complex in branches ( \u03c4 GMF ) of 3 . 3 ( \u00b1 0 . 5 ) min . This corre - sponds to a debranching rate constant ( k GMF = 1 / \u03c4 GMF ) of 0 . 31 ( \u00b1 0 . 05 ) min \u2212 1 . In Eqs . 3 and 4 , k 0 is the debranching rate con - stant and \u03c4 0 is the branch lifetime without GMF under very low flow rate ( 15 \u03bc L \u00b7 min \u2212 1 ) , that is , very low average force . The best fit yielded k 0 = 1 / \u03c4 0 = 0 . 014 ( \u00b1 0 . 0002 ) min \u2212 1 , corresponding to \u03c4 0 = 71 ( \u00b1 1 ) min . These values are slightly faster than the in - trinsic dissociation rate constant for branches with ADP \u2212 Arp2 / 3 complex in the absence of force ( > 100 min ; Fig . 3 ) due to the 15 \u03bc L \u00b7 min \u2212 1 flow rate employed in this experiment . We note , for general readers , that the GMF concentration at half - maximum Pandit et al . PNAS | June 16 , 2020 | vol . 117 | no . 24 | 13525 B I O P H Y S I C S A ND C O M P U T A T I O N A L B I O L O G Y effect is given by K d , GMF ( \u03c4 GMF / \u03c4 0 ) in Eq . 3 and K d , GMF in Eq . 4 , so its value differs depending on how the data are plotted . Reported affinities of GMF for soluble Arp2 / 3 complex ( K d , GMF ) vary widely : > > 10 \u03bc M for ATP \u2212 Arp 2 / 3 complex ( 43 ) ; 0 . 7 \u03bc M for ADP \u2212 Arp2 / 3 complex ( 43 ) ; and 13 nM ( 20 \u2013 22 ) or \u223c 1 \u03bc M ( 44 ) for Arp2 / 3 complex without a specified nucleotide . We found that the measured affinity of GMF for Arp2 / 3 complex in branch junctions bound to both a mother and daughter filament depended on the bound nucleotide . Arp2 / 3 complex used in these studies came from different organisms : Saccharomyces cerevisiae ( 20 \u2013 22 ) , bovine brain ( 43 ) , and S . pombe ( this study and ref . 44 ) . Thus , GMF preferentially dissociates \u201c old \u201d branches with ADP \u2212 Arp2 / 3 complex rather than \u201c young \u201d branches with ADP \u2212 P i \u2212 Arp2 / 3 complex . Given that GMF has a much lower affinity for ATP \u2212 Arp2 / 3 complex in solution than ADP \u2212 Arp2 / 3 complex ( 43 ) , weak binding of GMF to ADP \u2212 P i \u2212 Arp2 / 3 com - plex in branches is the most likely explanation for the resistance of young branches and branches with ADP \u2212 BeF x \u2212 Arp2 / 3 com - plex to dissociation by GMF . The nucleotide bound to Arp2 / 3 complex might also affect the debranching reaction directly . Saturating GMF destabilizes ADP \u2212 Arp2 / 3 complexes , reducing their lifetime from more than 60 min to \u223c 3 min ( Fig . 3 ) . While GMF strongly promotes debranching , these long lifetimes ( 3 min ) at saturating GMF concentrations place limits on the role of GMF in mediating debranching and network turnover in cells . More - rapid debranching may be achieved by combining GMF with force and / or from contributions from other debranching proteins such as cofilin , which dissociates branches on a second time scale at micromolar concentrations without force ( 24 ) . Arp2 / 3 Complex Likely Dissociates with the Daughter Filament . A daughter filament and its associated Arp2 / 3 complex dissociate simultaneously ( within the 100 ms time resolution of our ex - periments ) during debranching events . We cannot eliminate a pathway in which the daughter filament detaches first , followed by rapid release of Arp2 / 3 complex . However , we favor a mechanism where the interface between Arp2 / 3 complex and the mother filament ruptures and simultaneously releases the daugh - ter filament with Arp2 / 3 complex on its pointed end . This mech - anism is consistent with actin filaments rarely fragmenting under debranching conditions , so the rate constant for rupture of the actin \u2212 actin interface is hundreds of times slower than the rate constant for rupturing a branch junction . Given that the interface between Arp2 / 3 complex and the daughter filament is structurally similar on many levels to an actin \u2212 actin interface ( 45 ) , it is likely to rupture slowly like actin filaments . Materials and Methods Protein Purification . Rabbit skeletal muscle actin was purified from back and legmusclesandlabeledonlysineresidueswithNHSestersofAlexa568 , Alexa 647 , or biotin ( 46 , 47 ) or on Cys - 374 with pyrene iodoacetamide ( 48 ) . Actin monomers with bound ATP were passed twice through a desalting column to exchange into G - Buffer ( 20 mM Tris pH 8 , 2 mM CaCl 2 , 10 mM NaN 3 , 0 . 2 mM ATP , and 0 . 5 mM dithiothreitol [ DTT ] ) containing 0 . 2 mM AMPPNP instead of ATP and incubated for 1 h at 25 \u00b0C . Remaining free nucleotides were removed with a desalting column in G - Buffer containing AMPPNP in - stead of ATP prior to polymerization experiments . Arp2 / 3 complex was stored in ATP buffer solution composed of 10 mM Pipes at pH 6 . 8 , 100 mM KCl , 1 mM MgCl 2 , 0 . 25 mM ethylene glycol bis ( \u03b2 - aminoethyl ether ) - N , N , N \u2032 , N \u2032 - tetraacetic acid [ EGTA ] , 0 . 1 mM ATP , and 1 mM DTT . Recombinant fission yeast glutathione S - transferase ( GST ) - VCA was puri - fied from bacteria ( 49 ) . S . pombe Arp2 / 3 complex was purified from the TP150 strain , a protease - deleted S . pombe strain ( 49 ) . Fluorescent Arp2 / 3 complex was generated by conjugating Alexa 488 to a snap tag on the C terminus of the Arpc5 subunit . Bulk polymerization assays with pyrenyl actin and GST - VCA showed that Arp2 / 3 complex with 82 % labeling of the snap tag nucleated 91 % as many actin filaments as unlabeled Arp2 / 3 complex ( 4 . 2 nM vs . 4 . 6 nM of barbed ends at half - maximal polymerization ; SI Ap - pendix , Fig . S6 ) ( refs . 48 and 49 and SI Appendix , Fig . S8 ) . A complementary DNA for S . pombe GMF was cloned and inserted into the pet28a plasmid that includes GST and Tobacco Etch Virus ( TEV ) protease sites . Recombinant GMF was expressed in Rosetta2 BL21 Escherichia coli cells ( Novagen ) and purified using glutathione affinity chromatography . After cleaving off GST with TEV protease , a second round of glutathione chro - matography removed the protease and GST . After fast protein liquid chro - matography on a MonoQ column ( 20 ) , the GMF was greater than 95 % pure when analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis . Microscopy and Microfluidics . A total internal reflection fluorescence ( TIRF ) microscopy system with a Till iMic digital microscope equipped with a \u00d7 100 objective ( Olympus ) and an Andor iXon897 electron multiplying charge - coupled device ( EMCCD ) camera was used in this study . Images were ac - quired at a rate of 5 to 30 frames per second . Coverslips were cleaned with the following solutions , all incubated in a sonicator for 30 min and rinsed in - tensively with Milli - Q water in between steps : 2 % Hellmanex , water , acetone , 1 M HCl , 5 M KOH , and hexane , then silanized with 650 \u03bc L of dichlorodimethyl silane in 500 mL of hexane for 30 min , and subsequently rinsed and sonicated in hexane for 3 \u00d7 1 min . Coverslips were dried using a stream of nitrogen gas and stored in falcon tubes at \u2212 20 \u00b0C for up to 2 mo ( 50 ) . A glass microfluidic chamber was constructed as described in ref . 51 . Briefly , the input and output ports for flow solution were formed in poly - dimethylsiloxane ( PDMS ) using a flat - tip needle . Holes connecting the PDMS to parallel sample chambers were drilled through the glass with a diamond tip bit . Subsequently , a plasma cleaner was employed to bond PDMS to slide glass . Immediately before use , chambers were assembled using parafilm on the glass slide opposite to PDMS ports , and a coverslip was placed over the parafilm and sealed with heat . A fixed rate hydrodynamic flow exerted by an automatic syringe pump applied pullingforcesonactinfilaments . Themagnitudeofpullingforceona branch joint ( F d ; SI Appendix , Fig . S1 A and B ) scales with daughter filament branch length and flow rate . These values span a range up to a few pico - newtons for the daughter filament lengths and flow rates examined here . Pulling force was calculated according to the Batchelor ( 52 ) equation , F d = \u03b7 2 \u03c0 lv ln ( 2 hr ) . [ 5 ] Eq . 5 estimates drag force on a cylindrical filament with correction for the height ( h ) of the filaments from the surface of the flow chamber ( 53 ) . So - lution viscosity ( \u03b7 ) is assumed to equal unity ; l is the ( daughter ) filament length and varies for individual branches . The average actin filament radius is r = 4 nm , and v is the linear fluid velocity in the plane of the filament . We assume an average height ( h ) of 200 nm from the flow chamber surface . The fluid velocity ( v ) is proportional to bulk flow rate and was determined from the movement of 100 nM TetraSpek beads ( Thermo Fisher ) through the sample chamber ( 52 ) . Only beads moving parallel to the surface were employed in our analysis , as the flow velocity profile changes with the height from the chamber surface . Bead flow velocity was measured via ImageJ \u201c Manual Tracking \u201d and used in force calculations ( 52 ) . In the ex - periments presented in Fig . 2 A and C , multiple experiments at different flow rates were performed , and the force on each branch was calculated using the flow rate and length of the branch . Then , debranching data from different flow cells with a range of flow rates were binned according to the force on each branch and plotted as survival time courses , given by the fraction of branches remaining at each time point . For example , the branches in Fig . 2 A were divided into bins of at least 13 branches . The bin size for each data point is approximately the midpoint between the point itself and its neighboring data points ( Fig . 2 D ) . At low forces ( 0 . 002 pN to 0 . 4 pN ) , the bin sizes were small , for example , 0 pN to 0 . 04 pN , 0 . 1 pN to 0 . 2 pN , etc . , and , at higher forces ( > \u223c 1 pN ) , the bin sizes were larger , for example , 0 . 8 pN to 1 . 4 pN , 1 . 4 pN to 2 . 4 pN , etc . Someoftheseparametersareuncertain , sotheestimatedforcevaluesmay be offset systematically for all of our experiments . The least certain pa - rameter is the distance of individual filaments from the surface , which may vary by < 50 % from the average distance of 200 nm assumed here . A 500 \u03bc L \u00b7 min \u2212 1 flow rate produces forces on a 1 . 5 - \u03bc m branch of 0 . 94 pN at a height of 300 nm , 1 . 02 pN at 200 nm , and 1 . 20 pN at 100 nm ( SI appendix in ref . 52 ) . Therefore , the absolute forces and parameters calculated from them may be imprecise , but no more than 20 % , and the relative forces between experi - ments can be compared . Preparation of Branched Actin Filament Networks and Experimental Procedures . KMIE buffer ( 6 , 23 ) ( 10 mM imidazole pH 7 . 0 , 50 mM KCl , 2 mM MgCl 2 , 1 mM EGTA , 0 . 2 mM ATP , 2 mM DTT ) supplemented with 13526 | www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1911183117 Pandit et al . 15 mM glucose , 0 . 02 mg \u00b7 mL \u2212 1 catalase , and 0 . 1 mg \u00b7 mL \u2212 1 glucose oxidase was used for all microscope experiments , unless noted otherwise . Filaments of Alexa 568 \u2212 labeled actin ( Fig . 1 A , red filament segment ; 15 % labeled with Alexa 568 and containing 10 % biotinylated actin subunits ) were polymer - ized in KMIE buffer , sheared by vigorous pipetting , and subsequently mixed with 1 . 5 \u03bc M Alexa 647 - labeled actin monomers ( Fig . 1 A , green segment and branch ) , 0 . 1 \u03bc M Arp2 / 3 complex , and 0 . 5 \u03bc M GST - VCA in KMIE buffer . This solution was incubated for 1 min to 2 min to initiate branch formation . During the branching reaction , green actin monomers elongated from ei - ther the barbed end of short Alexa 568 actin segments or mother filament bound Arp2 / 3 complex VCA , forming mother or daughter filaments , re - spectively . This mixture was pipetted into flow chambers where the short , red actin segments bind the neutravidin - coated coverslip surface , and the branching reaction was allowed to continue for 1 min to 2 min . The total time of branch formation and aging for a given experiment ( e . g . , 2 . 6 min ) is indicated in the text . Branch formation and filament elongation was then terminated by removing untethered proteins ( labeled actin , Arp2 / 3 complex , and GST - VCA ) from the sample chamber with gentle flow of KMIE buffer . Tethered filaments were further aged for different time periods after the removal of unbound protein . Low ( 2 \u03bc L \u00b7 min \u2212 1 ) flow applied during aging ensured irreversible debranching . Filament segments with Alexa 568 actin subunits ( Fig . 1 A , red ) were tethered to the surface , while Alexa 647 actin ( Fig . 1 A , green ) mother and daughter filaments were allowed to freely fluctuate and align with flow . Unlabeled actin monomer ( 0 . 2 \u03bc M ) was in - cluded in flow buffer solution to prevent filament depolymerization . The \u201c aging time \u201d specified in each experiment is the time between the initial mixing of the proteins and application of debranching flow , including the time for branch formation and further aging . The age times were precisely measured in the experiment in Fig . 2 B but are approximate times in all other experiments . Debranching under force was performed by applying flow at a fixed rate ( specified in text ) throughout the experiment . Since unbound proteins were removed after the branch formation period , actin filament elongation was terminated , and individual branch length ( l ) and conse - quently applied force ( F ) on the branch joint were constant ( Eq . 5 ) . Experiments with BeF x were performed two different ways . 1 ) For ex - periments with ATP \u2212 Arp2 / 3 complex ( Fig . 2 C , Inset ) , branches were formed from ATP - actin monomers and ATP \u2212 Arp2 / 3 complex as described above but with 2 mM BeSO 4 and 10 mM NaF included in the buffer at all times , and aged for \u223c 4 min before applying various rates of flow with KMIE buffer supplemented with 2 mM BeSO 4 and 10 mM NaF . For the experiment with BeF x and 30 - min aging ( Fig . 2 B , gray ) , ATP \u2212 Arp2 / 3 complex branches were formed as described above and aged for 30 min with 2 mM BeSO 4 and 10 mM NaF before applying flow with KMIE buffer . 2 ) For experiments with ADP \u2212 Arp2 / 3 complex and BeF x ( Fig . 2 B , brown ) , mother filaments with ATP - actin ( 10 % biotinylated ) monomers were immobilized on the surface as described above , and the KMIE buffer ( which includes ATP ) was washed out and replaced with KMIE ADP buffer supplemented with 2 mM BeSO 4 and 10 mM NaF . Then ADP - actin monomers , ADP \u2212 Arp2 / 3 complex , and GST - VCA in KMIE buffer with 2 mM BeSO 4 and 10 mM NaF and 2 mM ADP instead of 2 mM ATP were gently flowed into the chamber and allowed to form branches . After a \u223c 4 - min branch formation period , all unbound proteins were washed out with KMIE ADP and 2 mM BeSO 4 and 10 mM NaF and maintained under flow as noted . ADP - actin and ADP \u2212 Arp2 / 3 complex were prepared by exchanging nucleotide from ATP - actin and ATP \u2212 Arp2 / 3 com - plex , respectively , using desalting columns . ADP \u2212 Arp2 / 3 complex never formed branches unless 2 mM BeSO 4 and 10 mM NaF was present as reported previously and shown here ( SI Appendix , Fig . S3 ) ( 5 , 6 ) . For experiments with AMPPNP mother filaments , filaments were poly - merized from AMPPNP actin monomers and immobilized on the surface as described above . Branch formation was initiated with ATP \u2212 Arp2 / 3 complex as described above in KMIE buffer with 0 . 2 mM AMPNP instead of 0 . 2 mM ATP . Since AMPPNP \u2212 Arp2 / 3 complex does not form branches ( 5 ) , verified in our experiments ( SI Appendix , Fig . S3 ) , all branches formed from Arp2 / 3 complex with bound ATP . Since the flow buffer included 0 . 2 mM AMPPNP , all actin monomers remained bound to AMPNP . The solution with Arp2 / 3 complex accounted for less than 1 % of the final solution volume , so it in - troduced a negligible amount of ATP . In debranching experiments with GMF with orwithout BeF x , samples were prepared as above except that debranching was initiated by gentle flow of buffer ( 15 \u03bc L \u00b7 min \u2212 1 , very low force ) containing a range of concentrations of GMF , and the flow was maintained throughout the experiment . Data Analysis . Analysis of branch dissociation was done by manual tracking with ImageJ ( https : / / imagej . nih . gov / ) . Branches that were fluctuating at all times and did not stick to the glass surface were selected randomly to minimize bias . These branches were labeled , cataloged , and observed for their entire lifetimes . Origin ( https : / / www . originlab . com / ) was used to fit all data and make plots . For the Arp2 / 3 complex debranching images in Fig . 1 , a Gaussian blur with a sigma radius of 1 . 25 pixels was applied . Then images were background subtracted with a rolling ball radius of 15 pixels and subsequently contrast enhanced . Data Availability . All protein purification protocols are referenced in Mate - rials and Methods . Purification of GMF is described in Materials and Meth - ods . Plasmid and the DNA sequence of GMF are provided in SI Appendix , Part 4 . Details on movies for microscopy experiments are provided in SI Appendix , Part 1 . 1 . T . D . Pollard , J . A . Cooper , Actin , a central player in cell shape and movement . Science 326 , 1208 \u2013 1212 ( 2009 ) . 2 . P . A . Janmey , C . A . McCulloch , Cell mechanics : Integrating cell responses to me - chanical stimuli . Annu . Rev . Biomed . Eng . 9 , 1 \u2013 34 ( 2007 ) . 3 . T . D . Pollard , Regulation of actin filament assembly by Arp2 / 3 complex and formins . Annu . Rev . Biophys . Biomol . Struct . 36 , 451 \u2013 477 ( 2007 ) . 4 . C . Le Clainche , D . Didry , M . F . Carlier , D . Pantaloni , Activation of Arp2 / 3 complex by Wiskott - Aldrich Syndrome protein is linked to enhanced binding of ATP to Arp2 . J . Biol . Chem . 276 , 46689 \u2013 46692 ( 2001 ) . 5 . M . J . Dayel , E . A . Holleran , R . D . Mullins , Arp2 / 3 complex requires hydrolyzable ATP for nucleation of new actin filaments . Proc . Natl . Acad . Sci . U . S . A . 98 , 14871 \u2013 14876 ( 2001 ) . 6 . L . Blanchoin , T . D . Pollard , R . D . Mullins , Interactions of ADF / cofilin , Arp2 / 3 complex , capping protein and profilin in remodeling of branched actin filament networks . Curr . Biol . 10 , 1273 \u2013 1282 ( 2000 ) . 7 . M . J . Dayel , R . D . Mullins , Activation of Arp2 / 3 complex : Addition of the first subunit of the new filament by a WASP protein triggers rapid ATP hydrolysis on Arp2 . PLoS Biol . 2 , E91 ( 2004 ) . 8 . E . Ingerman , J . Y . Hsiao , R . D . Mullins , Arp2 / 3 complex ATP hydrolysis promotes la - mellipodialactinnetworkdisassemblybutisdispensableforassembly . J . CellBiol . 200 , 619 \u2013 633 ( 2013 ) . 9 . C . Le Clainche , D . Pantaloni , M . F . Carlier , ATP hydrolysis on actin - related protein 2 / 3 complexcauses debranchingof dendritic actin arrays . Proc . Natl . Acad . Sci . U . S . A . 100 , 6337 \u2013 6342 ( 2003 ) . 10 . A . C . Martin , M . D . Welch , D . G . Drubin , Arp2 / 3 ATP hydrolysis - catalysed branch dis - sociation is critical for endocytic force generation . Nat . Cell Biol . 8 , 826 \u2013 833 ( 2006 ) . 11 . D . A . Fletcher , R . D . Mullins , Cellmechanicsandthecytoskeleton . Nature 463 , 485 \u2013 492 ( 2010 ) . 12 . L . Blanchoin , R . Boujemaa - Paterski , C . Sykes , J . Plastino , Actin dynamics , architecture , and mechanics in cell motility . Physiol . Rev . 94 , 235 \u2013 263 ( 2014 ) . 13 . P . Bieling et al . , Force feedback controls motor activity and mechanical properties of self - assembling branched actin networks . Cell 164 , 115 \u2013 127 ( 2016 ) . 14 . S . H . Parekh , O . Chaudhuri , J . A . Theriot , D . A . Fletcher , Loading history determines the velocity of actin - network growth . Nat . Cell Biol . 7 , 1219 \u2013 1223 ( 2005 ) . 15 . J . Mueller et al . , Load adaptation of lamellipodial actin networks . Cell 171 , 188 \u2013 200 . e16 ( 2017 ) . 16 . R . E . Mahaffy , T . D . Pollard , Kinetics of the formation and dissociation of actin filament branches mediated by Arp2 / 3 complex . Biophys . J . 91 , 3519 \u2013 3528 ( 2006 ) . 17 . R . E . Mahaffy , T . D . Pollard , Influenceofphalloidinonthe formationof actinfilament branches by Arp2 / 3 complex . Biochemistry 47 , 6460 \u2013 6467 ( 2008 ) . 18 . J . Berro , V . Sirotkin , T . D . Pollard , Mathematical modeling of endocytic actin patch kineticsinfissionyeast : Disassemblyrequiresreleaseofactinfilamentfragments . Mol . Biol . Cell 21 , 2905 \u2013 2915 ( 2010 ) . 19 . V . Sirotkin , J . Berro , K . Macmillan , L . Zhao , T . D . Pollard , Quantitative analysis of the mechanism of endocytic actin patch assembly and disassembly in fission yeast . Mol . Biol . Cell 21 , 2894 \u2013 2904 ( 2010 ) . 20 . M . Gandhi et al . , GMF is a cofilin homolog that binds Arp2 / 3 complex to stimulate filament debranching and inhibit actin nucleation . Curr . Biol . 20 , 861 \u2013 867 ( 2010 ) . 21 . B . L . Goode , M . O . Sweeney , J . A . Eskin , GMF as an actin network remodeling factor . Trends Cell Biol . 28 , 749 \u2013 760 ( 2018 ) . 22 . C . A . Ydenberg et al . , GMF severs actin - Arp2 / 3 complex branch junctions by a cofilin - like mechanism . Curr . Biol . 23 , 1037 \u2013 1045 ( 2013 ) . 23 . H . Kang et al . , Site - specific cation release drives actin filament severing by vertebrate cofilin . Proc . Natl . Acad . Sci . U . S . A . 111 , 17821 \u2013 17826 ( 2014 ) . 24 . C . Chan , C . C . Beltzner , T . D . Pollard , Cofilin dissociates Arp2 / 3 complex and branches from actin filaments . Curr . Biol . 19 , 537 \u2013 545 ( 2009 ) . 25 . G . I . Bell , Modelsforthespecificadhesionofcellstocells . Science 200 , 618 \u2013 627 ( 1978 ) . 26 . S . Rakshit , S . Sivasankar , Biomechanics of cell adhesion : How force regulates the lifetime of adhesive bonds at the single molecule level . Phys . Chem . Chem . Phys . 16 , 2211 \u2013 2223 ( 2014 ) . 27 . A . Muhlrad , P . Cheung , B . C . Phan , C . Miller , E . Reisler , Dynamic properties of actin . Structural changes induced by beryllium fluoride . J . Biol . Chem . 269 , 11852 \u2013 11858 ( 1994 ) . Pandit et al . PNAS | June 16 , 2020 | vol . 117 | no . 24 | 13527 B I O P H Y S I C S A ND C O M P U T A T I O N A L B I O L O G Y 28 . P . Ge , Z . A . O . Durer , D . Kudryashov , Z . H . Zhou , E . Reisler , Cryo - EM reveals different coronin binding modes for ADP - and ADP - BeFx actin filaments . Nat . Struct . Mol . Biol . 21 , 1075 \u2013 1081 ( 2014 ) . 29 . M . F . Carlier , D . Pantaloni , Binding of phosphate to F - ADP - actin and role of F - ADP - - Pi - actin in ATP - actin polymerization . J . Biol . Chem . 263 , 817 \u2013 825 ( 1988 ) . 30 . L . A . Helgeson , B . J . Nolen , Mechanism of synergistic activation of Arp2 / 3 complex by cortactin and N - WASP . eLife 2 , e00884 ( 2013 ) . 31 . B . A . Smith , K . Daugherty - Clarke , B . L . Goode , J . Gelles , Pathway of actin filament branchformationby Arp2 / 3complexrevealedby single - moleculeimaging . Proc . Natl . Acad . Sci . U . S . A . 110 , 1285 \u2013 1290 ( 2013 ) . 32 . B . A . Smith et al . , Three - color single moleculeimagingshows WASPdetachment from Arp2 / 3 complex triggers actin filament branch formation . eLife 2 , e01008 ( 2013 ) . 33 . A . C . Reymann etal . , Actinnetworkarchitecturecandeterminemyosinmotoractivity . Science 336 , 1310 \u2013 1314 ( 2012 ) . 34 . Sonal et al . , Myosin - II activity generates a dynamic steady state with continuous actin turnover in a minimal actin cortex . J . Cell Sci . 132 , jcs219899 ( 2019 ) . 35 . C . A . Wilson et al . , Myosin II contributes to cell - scale actin network treadmilling through network disassembly . Nature 465 , 373 \u2013 377 ( 2010 ) . 36 . N . A . Medeiros , D . T . Burnette , P . Forscher , Myosin II functions in actin - bundle turn - over in neuronal growth cones . Nat . Cell Biol . 8 , 215 \u2013 226 ( 2006 ) . 37 . M . Vinzenz et al . , Actin branching in the initiation and maintenance of lamellipodia . J . Cell Sci . 125 , 2775 \u2013 2785 ( 2012 ) . 38 . L . M . McMillen , D . Vavylonis , Model of turnover kinetics in the lamellipodium : Im - plications of slow - and fast - diffusing capping protein and Arp2 / 3 complex . Phys . Biol . 13 , 66009 ( 2016 ) . 39 . T . Miyoshi et al . , Actinturnover - dependent fast dissociation of capping protein in the dendritic nucleation actin network : Evidence of frequent filament severing . J . Cell Biol . 175 , 947 \u2013 955 ( 2006 ) . 40 . K . Hayakawa , H . Tatsumi , M . Sokabe , Actin filaments function as a tension sensor by tension - dependentbinding of cofilin to the filament . J . CellBiol . 195 , 721 \u2013 727 ( 2011 ) . 41 . D . E . Hannemann , W . Cao , A . O . Olivares , J . P . Robblee , E . M . De La Cruz , Magnesium , ADP , and actin binding linkage of myosin V : Evidence for multiple myosin V - ADP and actomyosin V - ADP states . Biochemistry 44 , 8826 \u2013 8840 ( 2005 ) . 42 . E . V . Wong etal . , Nup159weakensGle1bindingtoDbp5butdoesnotaccelerateADP release . J . Mol . Biol . 430 , 2080 \u2013 2095 ( 2018 ) . 43 . M . Boczkowska , G . Rebowski , R . Dominguez , Glia maturation factor ( GMF ) interacts with Arp2 / 3 complex in a nucleotide state - dependent manner . J . Biol . Chem . 288 , 25683 \u2013 25688 ( 2013 ) . 44 . K . Nakano , I . Mabuchi , Actin - depolymerizing protein Adf1 is required for formation and maintenance of the contractile ring during cytokinesis in fission yeast . Mol . Biol . Cell 17 , 1933 \u2013 1945 ( 2006 ) . 45 . I . Rouiller et al . , The structural basis of actin filament branching by the Arp2 / 3 com - plex . J . Cell Biol . 180 , 887 \u2013 895 ( 2008 ) . 46 . B . R . McCullough , L . Blanchoin , J . L . Martiel , E . M . De la Cruz , Cofilin increases the bending flexibility of actin filaments : Implications for severing and cell mechanics . J . Mol . Biol . 381 , 550 \u2013 558 ( 2008 ) . 47 . W . A . Elam et al . , Phosphomimetic S3D cofilin binds but only weakly severs actin filaments . J . Biol . Chem . 292 , 19565 \u2013 19579 ( 2017 ) . 48 . J . A . Cooper , S . B . Walker , T . D . Pollard , Pyreneactin : Documentationofthevalidityof a sensitive assay for actin polymerization . J . Muscle Res . Cell Motil . 4 , 253 \u2013 262 ( 1983 ) . 49 . B . J . Nolen , T . D . Pollard , Structure and biochemical properties of fission yeast Arp2 / 3 complex lacking the Arp2 subunit . J . Biol . Chem . 283 , 26490 \u2013 26498 ( 2008 ) . 50 . B . Y . Hua et al . , An improved surface passivation method for single - molecule studies . Nat . Methods 11 , 1233 \u2013 1236 ( 2014 ) . 51 . E . M . Johnson - Chavarria , U . Agrawal , M . Tanyeri , T . E . Kuhlman , C . M . Schroeder , Automated single cell microbioreactor for monitoring intracellular dynamics and cell growth in free solution . Lab Chip 14 , 2688 \u2013 2697 ( 2014 ) . 52 . N . Courtemanche , J . Y . Lee , T . D . Pollard , E . C . Greene , Tension modulates actin fil - ament polymerization mediated by formin and profilin . Proc . Natl . Acad . Sci . U . S . A . 110 , 9752 \u2013 9757 ( 2013 ) . 53 . C . Brennen , H . Winet , Fluid - mechanics of propulsion by cilia and flagella . Annu . Rev . Fluid Mech . 9 , 339 \u2013 398 ( 1977 ) . 54 . J . P . Robblee , W . Cao , A . Henn , D . E . Hannemann , E . M . De La Cruz , Thermodynamics of nucleotide binding to actomyosin V and VI : A positive heat capacity change ac - companies strong ADP binding . Biochemistry 44 , 10238 \u2013 10249 ( 2005 ) . 55 . E . W . Taylor , Kinetic studies on the association and dissociation of myosin subfrag - ment 1 and actin . J . Biol . Chem . 266 , 294 \u2013 302 ( 1991 ) . 13528 | www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1911183117 Pandit et al .",
    "akamatsu2020principles": "* Forcorrespondence : padmini . rangamani @ eng . ucsd . edu ( PR ) ; drubin @ berkeley . edu ( DGD ) Competing interests : The authors declare that no competing interests exist . Funding : See page 34 Received : 02 July 2019 Accepted : 16 January 2020 Published : 17 January 2020 Reviewing editor : Patricia Bassereau , Institut Curie , France Copyright Akamatsu et al . This article is distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use and redistribution provided that the original author and source are credited . Principles of self - organization and load adaptation by the actin cytoskeleton during clathrin - mediated endocytosis Matthew Akamatsu 1 , Ritvik Vasan 2 , Daniel Serwas 1 , Michael A Ferrin 1 , Padmini Rangamani 2 * , David G Drubin 1 * 1 Department of Molecular and Cell Biology , University of California , Berkeley , Berkeley , United States ; 2 Department of Mechanical and Aerospace Engineering , University of California , San Diego , La Jolla , United States Abstract Force generation by actin assembly shapes cellular membranes . An experimentally constrained multiscale model shows that a minimal branched actin network is sufficient to internalize endocytic pits against membrane tension . Around 200 activated Arp2 / 3 complexes are required for robust internalization . A newly developed molecule - counting method determined that ~ 200 Arp2 / 3 complexes assemble at sites of clathrin - mediated endocytosis in human cells . Simulations predict that actin self - organizes into a radial branched array with growing ends oriented toward the base of the pit . Long actin filaments bend between attachment sites in the coat and the base of the pit . Elastic energy stored in bent filaments , whose presence was confirmed by cryo - electron tomography , contributes to endocytic internalization . Elevated membrane tension directs more growing filaments toward the base of the pit , increasing actin nucleation and bending for increased force production . Thus , spatially constrained actin filament assembly utilizes an adaptive mechanism enabling endocytosis under varying physical constraints . Introduction Cells polymerize actin filaments to produce force and provide mechanical integrity for a variety of cellular processes , from cytokinesis and cell migration , to membrane reshaping and trafficking ( Pol - lard , 2016 ) . For each cellular process , actin filaments organize into a specific geometry that confers structural integrity and force - generation capacity . Most membrane deformation processes use branched actin networks nucleated by the Arp2 / 3 complex , a branched actin filament network nucle - ator ( Carlsson , 2018 ; Rotty et al . , 2013 ) . On a large ( m m ) length scale , branched actin networks drive the plasma membrane forward during cell migration , such that on the scale of individual actin branches , the membrane shape can be thought of as more or less constant ( Keren et al . , 2008 ; Mueller et al . , 2017 ; Schaus et al . , 2007 ) . However , on a smaller ( sub - micron ) length scale , branched actin networks deform many cellular membranes as part of organelle and vesicle biogene - sis and function ( Rottner et al . , 2017 ) . The relationship between cellular membrane curvature and local actin assembly for each of these \u2018local\u2019 membrane - deformation processes remains relatively unexplored ( Daste et al . , 2017 ) . Clathrin - mediated endocytosis ( CME ) is an especially attractive process for studies of actin\u2019s role in membrane shape changes due to the relatively complete parts list and available quantitative infor - mation about the positions , recruitment timing and biochemical function of many of the participating proteins ( Arasada et al . , 2018 ; Idrissi et al . , 2012 ; Kaksonen et al . , 2005 ; Kaksonen et al . , 2003 ; Mund et al . , 2018 ; Picco et al . , 2015 ; Sochacki et al . , 2017 ; Taylor et al . , 2011 ) . CME is a ubiqui - tous and essential cellular process by which cells take macromolecules from the extracellular space and the plasma membrane into the cell interior ( Kaksonen and Roux , 2018 ) . During CME , the Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 1 of 40 RESEARCH ARTICLE plasma membrane is bent , pinched , and pulled inward in a time frame of ~ 60 s thereby transitioning from a flat sheet into a spherical vesicle ~ 100 nm in diameter . Clathrin and its adaptor proteins establish a coat that generates initial membrane curvature ( Chen et al . , 1998 ; Pearse , 1976 ; Stachowiak et al . , 2012 ) , and BAR ( bin - amphiphysin - rvs ) - domain proteins bind curved membranes and support further membrane curvature ( Buser and Drubin , 2013 ; David et al . , 1996 ; Kishimoto et al . , 2011 ) . During yeast endocytosis , branched actin filaments provide the force required for membrane tubule formation ( Engqvist - Goldstein and Drubin , 2003 ; Idrissi et al . , 2012 ; Kukulski et al . , 2012 ; Picco et al . , 2018 ; Sun et al . , 2006 ; Wang and Carlsson , 2017 ) . In metazoan cells , endocytic pits under high tension stall at a \u2018U\u2019 - shaped intermediate in the absence of functional actin ( Boulant et al . , 2011 ) , implying that actin is required to generate plasma mem - brane shape changes late in CME ( Hassinger et al . , 2017 ; Yarar et al . , 2005 ; Yoshida et al . , 2018 ) . The molecular mechanism by which a network of polarized , branched actin filaments assembles at these sites for productive force generation is poorly understood . Actin network assembly is known to play a key role in membrane shape change in some contexts . For example , mathematical modeling ( Berro et al . , 2010 ; Carlsson and Bayly , 2014 ; Dmitrieff and Ne\u00b4de\u00b4lec , 2015 ; Liu et al . , 2009 ; Mund et al . , 2018 ; Wang et al . , 2016 ) and quantitative fluores - cence imaging in yeast ( Wu and Pollard , 2005 ; Sirotkin et al . , 2010 ; Berro and Pollard , 2014 ; Picco et al . , 2015 ) have established the relationship between actin filament assembly and plasma membrane shape particular to fungi , which have unique mechanical requirements due to very high ( ~ 10 atm ) hydrostatic turgor pressure . However , less is known about actin organization and function in the lower force regime characteristic of metazoan cells . A multiscale modeling effort that accounts for the mechanics of single actin filaments and that is constrained by experimental measurements of actin dynamics , spatial organization of the filaments , and tension in the plasma membrane is required to gain insight into actin organization and force generation capacity . We hypothesize that in localized membrane - reshaping processes such as endocytosis , branched actin networks assemble under specific spatial \u2018boundary conditions , \u2019 which serve as geometrical constraints dictated both by the shape of the membrane and the spatial segregation of membrane - associated proteins that inter - act with actin . These unique spatial boundary conditions on a curved surface , combined with the knowledge of numbers of molecules in cells and known reaction rate constants , provide the eLife digest The outer membrane of a cell is a tight but elastic barrier that controls what enters or leaves the cell . Large molecules typically cannot cross this membrane unaided . Instead , to enter the cell , they must be packaged into a pocket of the membrane that is then pulled inside . This process , called endocytosis , shuttles material into a cell hundreds of times a minute . Endocytosis relies on molecular machines that assemble and disassemble at the membrane as required . One component , a protein called actin , self - assembles near the membrane into long filaments with many repeated subunits . These filaments grow against the membrane , pulling it inwards . But it was not clear how actin filaments organize in such a way that allows them to pull on the membrane with enough force \u2013 and without a template to follow . Akamatsu et al . set about identifying how actin operates during endocytosis by using computer simulations that were informed by measurements made in living cells . The simulations included information about the location of actin and other essential molecules , along with the details of how these molecules work individually and together . Akamatsu et al . also developed a method to count the numbers of molecules of a key protein at individual sites of endocytosis . High - resolution imaging was then used to create 3D pictures of actin and endocytosis in action in human cells grown in the laboratory . The analysis showed the way actin filaments arrange themselves depends on the starting positions of a few key molecules that connect to actin . Imaging confirmed that , like a pole - vaulting pole , the flexible actin filaments bend to store energy and then release it to pull the membrane inwards during endocytosis . Finally , the simulations predicted that the collection of filaments adapts its shape and size in response to the resistance of the elastic membrane . This makes the system opportunistic and adaptable to the unpredictable environment within cells . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 2 of 40 Research article Cell Biology Physics of Living Systems necessary information for multiscale modeling and a mechanistic framework to understand the rela - tionship between plasma membrane mechanics and branched actin assembly and mechanics associ - ated with CME . Using this framework , we sought to answer the following questions : How do branched actin net - works assemble , organize , and produce force around an endocytic pit ? How does the spatial segre - gation of Arp2 / 3 complex activators ( Almeida - Souza et al . , 2018 ; Mund et al . , 2018 ) and actin - binding proteins associated with endocytic coats ( Clarke and Royle , 2018 ; Engqvist - Goldstein et al . , 2001 ; Sochacki et al . , 2017 ) influence this organization ? And finally , how do endo - cytic actin networks adapt to changing loads due to the stochastic environment and changes in membrane tension ? To answer these questions , we combined live - cell molecule counting methods in genome - edited diploid human cells and cryo - electron tomography of intact cells with multiscale modeling of plasma membrane mechanics and actin filament dynamics . Our results show that a mini - mal branched actin network is sufficient to create sustained internalization of an endocytic pit against physiological membrane tension . Actin filament self - organization and bending , which arise from the spatial distribution of actin - coat attachments around the curved endocytic pit , allow the actin net - work to adapt to changing loads . We anticipate that the mechanistic insights gained for actin in mammalian endocytosis will also apply to a variety of local membrane - bending processes carried out by branched actin throughout the cell . Results Multiscale modeling shows that a minimal branched actin network is sufficient to internalize endocytic pits against physiological membrane tension We combined a continuum - mechanics model of the plasma membrane , an agent - based model of actin filament dynamics , quantitative fluorescence microscopy , and electron tomography in cells to determine the molecular mechanism by which branched actin networks produce force during mam - malian clathrin - mediated endocytosis ( Figure 1 , Scheme 1 ) . First , we used a continuum - mechanics model of the plasma membrane ( Alimohamadi et al . , 2018 ; Hassinger et al . , 2017 ; Rangamani et al . , 2014 ) to determine the force - extension relation - ship for clathrin - coated pits stalled at a U - shaped intermediate under high membrane tension ( Figure 1B ) . Here , the extension refers to the extent of pit internalization , which is a displacement in the - Z direction ( Figure 1A \u2013 B ) . Previously , we showed that membrane curvature generation by the endocytic coat during vesicle formation could snap the membrane into a pinched \u2018omega\u2019 shape as a function of membrane tension and the curvature induced by the coat ( Hassinger et al . , 2017 ) , but we did not focus on force produced by the actin cytoskeleton . Here , we modeled the coated mem - brane based on the Helfrich ( 1973 ) energy and applied a linear force to the clathrin - coated pit in increasing value over successive simulations , corresponding to a simplified actin force . Simulations demonstrated that a clathrin - coated pit experiences a nearly linear force - extension relationship until an internalization of ~ 100 nm , at which point the pit can also adopt a pinched ( or \u2018omega\u2019 ) shape , which requires a lower force ( Figure 1C and Figure 1\u2014video 1 ) . We calculated the resistance to internalization as the slope of the force - extension plot for the linear regime and found that it is directly proportional to plasma membrane tension for a wide range of coat rigidities ( Figure 1D ) . Importantly , this direct scaling between resistance to internalization and membrane tension allowed us to treat this step of endocytic pit internalization as a linear spring , with the spring constant cali - brated using measurements of plasma membrane tension in mammalian cells ( Diz - Mun\u02dcoz et al . , 2016 ; Kaplan et al . , in preparation ) . The simple spring - like relationship uncovered above between force and endocytic pit internaliza - tion ( Figure 1D ) allowed us to simplify our mechanical treatment of the plasma membrane while modeling individual actin filaments and actin - binding proteins with realistic kinetics and mechanics ( Figure 1E \u2013 G , Supplementary file 3 ) . We used Cytosim ( Nedelec and Foethke , 2007 ) to construct a filament - based model of the endocytic actin network . This agent - based model allowed us to simu - late the emergent architecture and mechanical functions of branched actin for realistic endocytic geometries . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 3 of 40 Research article Cell Biology Physics of Living Systems 0 - 20 - 40 - 60 - 80 - 100 Internalization ( nm ) 0 5 10 15 20 25 30 35 40 45 F o r c e ( p N ) growing end capped end Hip1R nucleator Arp2 / 3 Actin 3 s 7 s 9 s 11 s 5 s Membrane tension ( pN / nm ) S p r i ng c on s t an t ( p N / n m ) A B E F Y ( n m ) Z ( n m ) X ( n m ) C Actin Arp2 / 3 complex Clathrin Capping protein Hip1R Plasma membrane vesicle A r p2 / 3 c o m p l e x H i p1 R T en s i on c apped end nu c l ea t o r D Tension Number Arp2 / 3 complex G U - shaped pit Transition - shaped pit ? 0 . 15 0 . 2 0 . 25 0 . 3 0 . 35 0 . 4 0 . 45 1 . 6 1 . 4 1 . 2 1 0 . 8 0 . 6 0 . 4 0 . 2 0 Tension ( pN / nm ) N - WASP N - N - N - WA N - WASP Base Coat \u03a9 0 - 50 - 100 0 5 10 15 Time ( s ) I n t e r na li z a t i on ( n m ) 1x2x5x7 . 5x 10x Coat rigidity 0 . 45 0 . 40 0 . 35 0 . 30 0 . 25 0 . 20 0 . 15 0 . 10 0 . 05 S p r i ng c on s t an t Figure 1 . Multiscale modeling shows that a minimal branched actin network is sufficient to internalize endocytic pits against physiological membrane tension . ( A ) Schematic of a section of the cell\u2019s plasma membrane being internalized during mammalian endocytosis depicts plasma membrane deformation against membrane tension ( purple arrows ) countered by the clathrin coat ( yellow ) and the actin cytoskeleton ( red ) . ( B ) Shape of the membrane and pit internalization from continuum mechanics simulations of the endocytic pit experiencing axial ( Z ) forces corresponding to simplified actin forces . To begin with , the plasma membrane ( yellow ) is deformed by a coat with preferred curvature that expands in area until the pit stalls . A net Figure 1 continued on next page Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 4 of 40 Research article Cell Biology Physics of Living Systems We simplified the endocytic pit as a solid , impermeable structure , initially a hemisphere , attached to a flat plasma membrane corresponding to the \u2018U - shaped\u2019 intermediate ( Avinoam et al . , 2015 ; Boulant et al . , 2011 ; Messa et al . , 2014 ; Yarar et al . , 2005 ; Figure 1E ) . The following rules were prescribed for actin filament dynamics . Initially , actin filament - nucleating proteins seed a small num - ber of actin filaments near the endocytic pit . These randomly - oriented \u2018mother filaments\u2019 serve as templates for binding pre - activated Arp2 / 3 complexes , which correspond to the coincidence of Arp2 / 3 complex and its activator N - WASP , arranged in a ring ( Almeida - Souza et al . , 2018 ; Mund et al . , 2018 ) at the base of the endocytic pit ( Idrissi et al . , 2008 ; Kaksonen et al . , 2003 ; Picco et al . , 2015 ; Kaplan et al . , in preparation ) . When an active Arp2 / 3 complex comes in proximity Figure 1 continued force ( red arrows ) is applied downward from the coat and upward into the base of the endocytic pit ( red dotted lines ) . In this simulation , membrane tension was 0 . 2 pN / nm , and the coated area was rigid ( 2400 pN ! nm ) . ( C ) Force versus pit internalization relationships for different values of membrane tension . Internalization is defined as the pit displacement in Z . Shading delineates linear force - internalization regime ( blue ) ; \u2018transition point\u2019 from U to omega shape ( orange ) ; \u2018omega - shaped\u2019 regime where the neck is narrower than the pit diameter and the force required for internalization is lower than at the transition point ( for tensions > 0 . 1 pN / nm ) ( yellow ) . Color matches the three snapshots in B . Parameters are given in Supplementary files 1 and 2 . ( D ) Resistance of pit to internalization versus membrane tension . Resistance ( spring constant ) is defined as absolute value of slope in C for the \u2018U - shaped\u2019 region . Each curve is calculated for a different value of membrane rigidity ( where 1x = 320 pN ! nm , the rigidity of the uncoated plasma membrane ) . ( E ) Computational model of branched actin filament polymerization coupled to endocytic pit internalization . An internalizing endocytic pit is modeled as a sphere with a neck attached to a flat surface by a spring . Active Arp2 / 3 complex ( blue ) is distributed in a ring around the base of the pit . An actin nucleation protein ( pink ) generates an actin filament ( white ) , which polymerizes , stalls under load , and is stochastically capped ( red ) . Arp2 / 3 complexes bind to the sides of actin filaments and nucleate new filaments at a 77 - degree angle , creating new branches . Linker Hip1R ( purple ) is embedded in the pit and binds to actin filaments . Model parameters are given in Supplementary file 3 . ( F ) Graphical output of the simulations from Cytosim ( Nedelec and Foethke , 2007 ) at 2 s intervals . Scale bar : 100 nm . ( G ) Pit internalization over simulated time as a function of the number of available molecules of Arp2 / 3 complex . Average of 16 simulations per condition . Shaded bars are standard deviations . The online version of this article includes the following video and figure supplement ( s ) for figure 1 : Figure supplement 1 . Effect of different actin - and simulation - related parameters on pit internalization dynamics . Figure supplement 2 . Initiation from a pool of diffusing cytoplasmic actin filaments leads to variable timing of internalization . Figure 1\u2014video 1 . Simulations of continuum membrane mechanics model . https : / / elifesciences . org / articles / 49840 # fig1video1 Figure 1\u2014video 2 . Simulation of actin in endocytosis using Cytosim . https : / / elifesciences . org / articles / 49840 # fig1video2 Scheme 1 . Flow chart of multiscale modeling and experimental strategy combining membrane mechanics , actin spatiotemporal dynamics , molecule counting , and cryo - electron tomography . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 5 of 40 Research article Cell Biology Physics of Living Systems with an actin filament , it can bind to the filament and nucleate the growth of a new branched fila - ment at an ~ 77\u02da angle ( Blanchoin et al . , 2000 ) . Growing actin filaments can polymerize , diffuse under thermal fluctuations , and bend under force , and their growing ends are capped stochastically . Filament growth decreases with load according to the Brownian ratchet mechanism ( Mogilner and Oster , 1996 ; Peskin et al . , 1993 ) . Growth of the actin network is coupled to internalization of the endocytic pit by an actin - linking protein ( Hip1 / Hip1R / Epsin , simplified here as Hip1R ) , which is embedded in the coated pit and binds to actin filaments ( Clarke and Royle , 2018 ; Engqvist - Goldstein et al . , 2001 ; Engqvist - Goldstein et al . , 1999 ; Sochacki et al . , 2017 ) . Importantly , most of the parameters in this model have been determined with measurements in vitro or in vivo , includ - ing the dimensions of the endocytic pit , its resistance to internalization ( modeled as a spring , Figure 1D ) , rates of association and dissociation of different proteins , branching angles , capping rates , filament persistence length , and stall force ( Supplementary file 3 and Materials and methods ) . Stochastic simulations of the model showed that this minimal branched actin network internalizes endocytic pits up to ~ 60 nm against physiological membrane tension ( Figure 1F and Figure 1\u2014 video 2 ) . In order to compare different conditions , we used two metrics \u2013 internalization of the pit ( in nm ) over time ( Figure 1G ) and the 95th percentile of internalization ( Figure 1\u2014figure supple - ment 1A ) . Then , we evaluated the robustness of the model to different parameters by conducting a series of parameter sweeps ( Figure 1\u2014figure supplement 1 ) . We found that the extent of internali - zation is robust to a wide range of parameters , including filament stiffness , stall force , and affinity between Hip1R attachments and actin filaments ( Figure 1\u2014figure supplement 1 ) . Initiating the sim - ulations from a cytoplasmic pool of linear actin filaments ( Raz - Ben Aroush et al . , 2017 ) allowed for endocytosis but the timing of the onset of internalization was more variable ( Figure 1\u2014figure sup - plement 2 ) . The extent of internalization was particularly sensitive to the number of available Arp2 / 3 complexes ( Figure 1G ) , indicating a need for precise measurements of this molecule at mammalian endocytic sites . Molecule counting of endogenously GFP - tagged Arp2 / 3 complex in live mammalian cells Motivated by our prediction that internalization rate is sensitive to the number of Arp2 / 3 complexes , we developed a method to count the number of molecules of endogenously GFP - tagged proteins in living mammalian cells ( Figure 2 ) . We adapted the self - assembling GFP - tagged protein nanocages developed by Hsia et al . ( 2016 ) for expression in live cells to create a fluorescence - based calibration curve relating fluorescence intensity of endogenously GFP - tagged proteins to numbers of molecules per endocytic site . Given that the nanocages are derived from bacterial glycolytic enzymes , we made point mutations known to abolish enzymatic activity of the proteins . To slow the diffusion of the intracellular nanocages and facilitate fluorescence measurements , we introduced an inducible dimerization motif to the plasma membrane by fusing the construct to FKBP and coexpressing a pal - mitoylated and myristoylated FRB variant ( Figure 2A ) . The resulting two - component fusion protein transiently associated with the plasma membrane even without rapamycin analog AP21967 , but the extent of association with the plasma membrane increased in a dose - dependent manner with the concentration of AP21967 ( Figure 2\u2014figure supplement 1A \u2013 B and Figure 3\u2014video 1 ) . We mea - sured the fluorescence intensity of four GFP - tagged nanocages of copy number ranging from 12 to 120 copies of GFP per structure using spinning disk confocal microscopy ( Figure 2B ) . After correct - ing for exposure time ( Figure 2\u2014figure supplement 1E \u2013 F ) , uneven illumination intensity , and local background ( Materials and methods ) , the fluorescence intensity per spot was unitary ( Figure 2C ) and directly proportional to the predicted numbers of molecules per structure ( R 2 = 0 . 996 ) ( Figure 2D ) . Using this calibration curve , we measured the numbers of molecules of an E . coli flagel - lar motor protein eGFP - MotB , which resulted in measurements similar to previously published meas - urements ( Figure 2\u2014figure supplement 1G \u2013 I ) . Thus , we established the suitability of this method to relate fluorescence intensity of endogenously GFP - tagged proteins to numbers of molecules inside live mammalian cells . To measure the timing , frequency , and numbers of Arp2 / 3 complexes assembling at sites of cla - thrin - mediated endocytosis , we used CRISPR / Cas9 - mediated genome editing to endogenously tag the C terminus of ArpC3 , a subunit of the Arp2 / 3 complex , with the fluorescent protein tagGFP2 in human induced pluripotent stem cells ( Figure 2E , Figure 2\u2014figure supplement 2A \u2013 B ) . Human Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 6 of 40 Research article Cell Biology Physics of Living Systems 12mer 0 1000 2000 3000 4000 0 20 40 24mer 0 1000 2000 3000 4000 0 5 10 60mer 0 1000 2000 3000 4000 0 20 40 120mer 0 1000 2000 3000 4000 0 20 40 0 50 100 150 Molecules per structure 0 1000 2000 3000 I n t en s i t y pe r s po t ( A . U . ) ! \" # $ % & \u2019 ( ) * ) + , - . / ) * ) 0122 , 220 \u00b1 190 390 \u00b1 130 1000 \u00b1 240 1800 \u00b1 400 12mer 24mer 60mer 120mer O cc u rr en c e s \" 2424me 6060me ! \" # $ % # & % \u2019 % ( ) # * \u2019 FRBFKBPnanocageGFP Intensity ( A . U . ) 120m ArpC3 - GFP ArpC2Arp3Arp2ArpC4 ArpC5 ArpC1 - , 0 60mer - 200 - 175 - 150 - 125 - 100 - 75 - 50 - 25 0 Time ( s ) 0 0 . 2 0 . 4 0 . 6 0 . 8 1 R e l a t i v e I n t e n s i t y ( A . U . ) 3 - 20 0 20 Time ( s ) 0 100 200 300 400 M o l e c u l e s o f A r p C 3 - G F P ! . 4 & 563 # 7 89 . : 9 ! 7 / 6 ; # 7 < 60mer - GFP 120mer - GFP ArpC3 - GFP ArpC3 - GFP AP2 - RFP # = : > . 9 ) / Figure 2 . Molecule counting of endogenously GFP - tagged Arp2 / 3 complex in live human induced pluripotent stem cells . ( A \u2013 D ) Development of a calibration curve relating fluorescence intensity to numbers of molecules in live cells . ( A ) Cartoon of intracellular GFP - tagged 60mer nanocage with inducible plasma membrane tether . Each subunit ( blue ) is tagged with GFP ( green ) and FKBP ( orange ) . FRB ( T2098L ) ( Purple ) is targeted to the plasma membrane by a palmitoylation and myristoylation sequence and dimerizes with FKBP in the presence of rapamycin analog AP21967 . Cartoon showing one of 60 tagged subunits is based on PDB structures 5kp9 , 2y0g , and 4dri . Scale bar 10 nm . ( B ) Inverse contrast fluorescence intensity images of human induced pluripotent stem cells expressing GFP - tagged plasma membrane - bound nanocages . Sum projection of nine 300 nm confocal images . Scale bar : 2 m m . ( C ) Histograms of fluorescence intensity per spot for the four calibration constructs showing mean \u00b1 standard deviation . Images were corrected for uneven illumination and intensity was background - corrected . Data from 305 spots in 15 cells over three experiments . ( D ) Calibration curve relating fluorescence intensity to numbers of molecules in mammalian cells . Line is a linear fit through zero . Error bars are standard deviations . ( E ) Cartoon drawn to scale of Arp2 / 3 complex tagged with GFP at the flexible C - terminus of ArpC3 . Known binding and activation sites are distal to this site . Based on PDB 2p9l . ( F ) Montage of CME event marked by AP2 - tagRFP - T and ArpC3 - tagGFP2 from TIRF imaging . Montage shows 4 s intervals from a movie taken at 2 s intervals . ( G ) Relative fluorescence intensity over time of AP2 - tagRFP - T and ArpC3 - tagGFP2 in endocytic events imaged by TIRF microscopy . Traces were normalized to maximum intensity and averaged . 121 traces from 8 cells in four experiments . Shading is \u00b1 1 s . d . ( H ) Fluorescence micrographs of ( left ) 60mer - tagGFP2 , ( left - center ) 120mer - tagGFP2 , ( right - center ) ArpC3 - tagGFP2 , and ( right ) ArpC3 - tagGFP2 and AP2 - tagRFP - T . White arrows mark spots in which ArpC3 - tagGFP2 and AP2 - tagRFP - T colocalize . Scale bar 2 m m . ( I ) Numbers of molecules of ArpC3 over time . The online version of this article includes the following video and figure supplement ( s ) for figure 2 : Figure 2 continued on next page Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 7 of 40 Research article Cell Biology Physics of Living Systems induced pluripotent stem cells are diploid and thus suitable for molecule - counting measurements when both alleles of the ArpC3 gene are fused to the gene for GFP . C - terminal GFP tags on ArpC3 are more functional than on other subunits of the Arp2 / 3 complex ( Egile et al . , 2005 ; Picco et al . , 2015 ; Sirotkin et al . , 2010 ; Smith et al . , 2013 ) . Cells tagged at both alleles of the ArpC3 gene had twice the fluorescence intensity of cells with a single allele tagged , suggesting direct proportionality between GFP fluorescence intensity and numbers of molecules ( Figure 2\u2014figure supplement 2C \u2013 D ) . These cells also endogenously express a tagRFP - T fusion with the m 2 subunit of the adaptor pro - tein AP2 , allowing us to identify sites of clathrin - mediated endocytosis ( Hong et al . , 2015 ) . We determined the relative timing of AP2 and ArpC3 appearance at endocytic sites using time - lapse TIRF imaging and automated two - color particle tracking ( Dambournet et al . , 2018 ; Hong et al . , 2015 ; Figure 2F ) . The vast majority ( 81 \u00b1 10 % , n = 136 ) of CME events marked by AP2 - RFP culminated in a burst of ArpC3 - GFP fluorescence , prior to internalization of the pit , persisting until the pit internalized ( Figure 2G and Figure 2\u2014video 2 ) . In addition , 24 \u00b1 4 % of ArpC3 - GFP tracks ( n = 145 ) did not colocalize with AP2 . We hypothesize that these are sites of clathrin - indepen - dent endocytosis . Then , using spinning - disk confocal fluorescence microscopy , we compared the fluorescence intensities of ArpC3 - GFP spots and GFP - tagged nanocage proteins in cells to deter - mine the numbers of ArpC3 - GFP molecules at clathrin - mediated endocytosis sites ( Figure 2H ) . Thus , we determined that ~ 200 molecules of Arp2 / 3 complex accumulate at clathrin - mediated endo - cytosis sites over time ( Figure 2I ) . Self - organization of actin filaments into a radial dendritic network drives endocytic internalization Incorporating the Arp2 / 3 molecule number we determined experimentally into our multiscale model , we next conducted simulations of the model to investigate the spatial organization of actin and force generation capacity of the endocytic network ( Figure 3 ) . Strikingly , the actin network self - organized around the endocytic pit . This self - organized network drove the assembly of 150 \u00b1 30 actin filaments ( Figure 3\u2014figure supplement 1A ) containing 5700 \u00b1 1100 monomers ( Figure 3\u2014figure supple - ment 1B ) . Interestingly , only a small number of actin filaments ( < 5 ) grew at any given time because the filaments became capped soon after they were nucleated ( Figure 3\u2014figure supplement 1C ; Berro et al . , 2010 ; Rangamani et al . , 2011 ; Xiong et al . , 2010 ) . Filament lengths were expo - nentially distributed with a final length of 90 \u00b1 80 nm ( Figure 3\u2014figure supplement 1D \u2013 E ) . Actin fil - aments bound to 120 \u00b1 10 Hip1R molecules in the coat ( Figure 3\u2014figure supplement 1F ) . The endocytic pit internalized ~ 60 nm in 10 \u2013 15 s ( Figure 3A and D ) . Based on the initial geometry of the endocytic pit and activated Arp2 / 3 complex , branched actin filaments self - organized into a radial dendritic network : the network attached to the clathrin coat by binding to Hip1R , the pointed ( minus ) ends localized close to the pit and the barbed ( plus ) ends near the base of the pit were ori - ented to grow toward the base of the pit ( Figure 3A \u2013 C and Figure 3\u2014video 1 ) . The axial self - organization of this branched actin network resembles that at the leading edge of cells ( Figure 3\u2014figure supplement 1G \u2013 I ; Maly and Borisy , 2001 ; Mueller et al . , 2017 ; Schaus et al . , 2007 ; Svitkina and Borisy , 1999 ) , with an important difference . Because actin fila - ment attachment sites are located on the coat of the endocytic pit , filaments radiate away from the center of the pit , such that most of the barbed ends orient radially away from the center of the pit rather than toward the coat or neck ( Figure 3E ) . The radial orientation of barbed ends gradually increases from the center of the pit , where there is no preferred orientation , to the periphery , where the barbed end radial orientation is highest ( Figure 3F ) . The extent of radial distribution of the Figure 2 continued Figure supplement 1 . Optimization and validation of fluorescence calibration method . Figure supplement 2 . Generation of genome - edited human induced pluripotent stem cell lines endogenously expressing AP2 - RFP and ArpC3 - GFP . Figure 2\u2014video 1 . Time lapse images of human induced pluripotent stem cells transiently expressing FKBP - 60mer - GFP and treated with 0 . 5 nM AP21967 . https : / / elifesciences . org / articles / 49840 # fig2video1 Figure 2\u2014video 2 . Time - lapse TIRF microscopy image of a human induced pluripotent stem cell endogenously expressing ArpC3 - GFP and AP2 - RFP . https : / / elifesciences . org / articles / 49840 # fig2video2 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 8 of 40 Research article Cell Biology Physics of Living Systems R Z Axial orientation C D XZ B (cid:16)(cid:28)(cid:19)(cid:219) (cid:14)(cid:28)(cid:19)(cid:219) (cid:19)(cid:219) (cid:14)(cid:28)(cid:19)(cid:221) (cid:14)(cid:23)(cid:24)(cid:221) (cid:19)(cid:221) (cid:16)(cid:23)(cid:24)(cid:221) (cid:16)(cid:28)(cid:19)(cid:221) Radial orientation XY BE R Y X 1 - 1 F E H I I G + 1 + 0 . 5 0 - 0 . 5 - 1 random filaments endocytosis Base Neck p i nch i n g pu lli ng A Z ( n m ) X ( nm ) Barbed Barbed Pointed Pointed Base Neck + + 1 s 3 s 5 s 7 s 9 s 11 s 13 s 15 s Axial orientation Radial orientation 0 - 50 - 100 - 150 6420 - 2 - 4 - 6 - 200 0 200 - 200 0 200 - 200 0 200 - 200 0 200 - 200 0 200 - 200 0 200 - 200 0 200 - 200 0 200 R e l a t i v e nu m be r o f end s I n t e r n a li z a t i on ( n m ) Figure 3 . Self - organization of actin filaments into a radial dendritic network drives endocytic internalization . ( A ) ( Left ) Schematic depicting actin barbed ( plus ) or pointed ( minus ) ends . ( Right ) Heat maps of the positions of barbed ends ( red ) or pointed ends ( blue ) relative to the endocytic pit . Color code represents the relative number of ends . Each graph is averaged across 96 simulations and 1 s of simulation time . ( B ) Simulation output of endocytic actin filaments color - coded for axial ( Z ) orientation . Blue filaments orient toward the base of the pit ( + 90\u02da ) and green filaments orient parallel to the base of the pit ( 0\u02da ) . ( C ) Axial orientation of barbed ends . ( Left ) Schematic of axes . R is radial position of barbed end . ( Right ) Heat map of axial orientation of barbed ends as a function of R and Z position . Average of 96 simulations . ( D ) Pit internalization over time ( n = 96 simulations ) . ( E ) Simulation output of endocytic actin filaments color - coded for radial orientation . ( F ) Radially oriented endocytic actin filaments . ( Left ) Schematic of axes . Radial orientation is defined such that + 1 = barbed end oriented away from the center of the pit , and \" 1 = barbed end oriented toward the center of the pit . ( Right ) Heat map of radial orientation of barbed ends as a function of X and Y position ( n = 96 simulations ) . Barbed ends radiate outward . ( G ) Radial orientation of barbed ends over time for 96 simulations . Gray curve is negative control of randomly oriented filaments ( n = 50 filaments in one simulation ) . ( H ) Concentration of barbed ends near the base of the endocytic pit . ( Left ) Schematic of positions of the neck and base of the pit . ( Right ) Number of barbed ends near base ( green ) or neck ( blue ) of pit , defined as within 7 . 5 nm of each surface . ( I ) The majority of forces are directed orthogonal to the base of the pit based on positions of barbed ends in simulations . Shaded bars are standard deviations . The online version of this article includes the following video and figure supplement ( s ) for figure 3 : Figure supplement 1 . Assembly and self - organization of endocytic actin network . Figure 3\u2014video 1 . Simulation of actin in endocytosis with actin filaments color coded for axial orientation . https : / / elifesciences . org / articles / 49840 # fig3video1 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 9 of 40 Research article Cell Biology Physics of Living Systems filaments increases rapidly after time 0 ( Figure 3G ) . An important consequence of this self - organiza - tion is that , based on the position of Hip1R and the Arp2 / 3 complex , more barbed filament ends localize near the base ( 10 \u00b1 4 ends ) than near the neck of the endocytic pit ( 1 \u00b1 1 ends ) ( Figure 3H ) . These data result in an important prediction from our model : an actin network self - organized as described here will produce an axial force during pit internalization ( Figure 3I ) . We predict that the radial dendritic self - organization is a powerful mechanism that makes endocytic actin networks resil - ient to biochemical and mechanical perturbations . Spatial distribution of actin / coat attachments and Arp2 / 3 complex , but not Arp2 / 3 complex density , strongly affects actin self - organization and pit internalization Our finding that self - organized endocytic actin networks grow toward the base of the pit prompted us to explore the molecular mechanism by which actin filaments self - organize . Actin dynamics in association with the endocytic machinery can be thought of as a polymerization engine constrained by two spatial boundary conditions \u2013 active Arp2 / 3 complex at the base of the pit ( Almeida - Souza et al . , 2018 ; Idrissi et al . , 2008 ; Kaksonen et al . , 2003 ; Mund et al . , 2018 ; Picco et al . , 2015 ; Kaplan et al . , in preparation ) and Hip1R / actin attachments on the curved pit surface ( Clarke and Royle , 2018 ; Engqvist - Goldstein et al . , 2001 ; Engqvist - Goldstein et al . , 1999 ; Sochacki et al . , 2017 ; Figure 4A ) . Given that such spatial boundary conditions confer unique mechanical properties and adaptation to loads under flat geometries in vitro ( Bieling et al . , 2016 ) , we aimed to understand how the boundary conditions corresponding to the curved endocytic pit affect endocytic actin organization and internalization . We tested two different scenarios : varying the surface density of Arp2 / 3 complex at the base of the pit and varying Hip1R surface coverage around the pit itself . First , we tested whether the surface density of the Arp2 / 3 complex at the base of the pit affects endocytic internalization because recent studies in vitro and in yeast suggest that the local concen - tration of Arp2 / 3 complex activators is critical for the timing of Arp2 / 3 complex activation and endo - cytic progression ( Case et al . , 2019 ; Sun et al . , 2017 ) . In a series of simulations , we distributed 200 molecules of active Arp2 / 3 complex in a ring of increasing outer diameter to vary the surface den - sity . Surprisingly , we found that varying the surface density of Arp2 / 3 complex along the base of the pit by a factor of 20 had little impact on endocytic outcome ( Figure 4\u2014figure supplement 1 ) . We also explored whether localization of a fraction of Arp2 / 3 complexes at the neck of the pit provided an additional advantage for the endocytic outcome . In this scenario , we distributed 50 of the 200 molecules of the active Arp2 / 3 complex near the neck of the pit . We found that localizing some of the active Arp2 / 3 complex near the neck of the pit did not have an impact on the outcome of simu - lations ( p > 0 . 5 ) ( Figure 4\u2014figure supplement 2D \u2013 E ; Figure 4\u2014figure supplement 2 ) . We next conducted a series of simulations in which we varied the surface distribution of a con - stant number of Hip1R molecules to cover between 1 % ( localized to the tip of the pit ) and 80 % ( up to the neck of the pit ) of the pit ( Figure 4B ) and found that the surface distribution of Hip1R around the endocytic pit strongly impacted endocytic outcome ( Figure 4 ) . Simulations in each of these con - ditions revealed that endocytic internalization depends on the surface distribution of actin - coat attachments around the endocytic site ( Figure 4C and Figure 4\u2014video 1 ) . Both the rate and extent of internalization increased with increasing surface area of Hip1R around the pit ( Figure 4D ) . From a functional standpoint , increased Hip1R surface coverage around the pit drove more barbed ends toward the base of the pit ( Figure 4E ) . This increase in Hip1R surface coverage resulted in an increase in Arp2 / 3 complexes bound in the endocytic actin network ( Figure 4F ) , which in turn nucle - ated more actin filaments ( Figure 4G ) . Simulations showed that a threshold of ~ 100 Hip1R molecules on the pit is necessary for endocytic internalization ( Figure 4\u2014figure supplement 3A ) . The high impact of Hip1R surface distribution on actin filament organization implies that Hip1R molecules dis - tributed broadly around the pit allow for multivalent attachments between the pit and actin fila - ments , resulting in filaments being captured in an orientation conducive to force production . Further examination of the simulations revealed that the Hip1R surface distribution supports a self - organized dendritic actin network via a mechanism of stochastic self - assembly and selection for actin filaments growing toward the base of the pit ( Figure 4\u2014figure supplement 3B ) . Mother fila - ments initially bind and unbind the coat in random orientations ( Figure 4\u2014figure supplement 3B \u2013 C ) . Filaments growing toward the interior of the cell do not template the growth of new branched Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 10 of 40 Research article Cell Biology Physics of Living Systems actin filaments . However , filaments growing toward the base of the pit encounter active Arp2 / 3 com - plex , which catalyzes dendritic nucleation of new actin filaments growing in a similar direction ( Fig - ure 4\u2014figure supplement 3B and D ; Carlsson , 2001 ) . As a result , near the base of the pit , filaments increasingly orient toward the base of the pit over time ( Figure 4\u2014figure supplement 3E \u2013 F ) . Our observations therefore establish an important principle for actin organization in endocytosis : the positions of active Arp2 / 3 complexes are critical for organizing the actin network and determin - ing the direction of force production , while the Hip1R linker distribution is critical for recruiting 40 - 40 - 20 0 20 40 A r p2 / 3 c o m p l e x H i p1 R Boundary 1 : Active Arp2 / 3 complex Boundary 2 : Hip1R attachments B C G A D E C F Time ( s ) N u m be r A r p2 / 3 bound Hip1R coverage ( % ) Hip1R coverage ( % ) Z ( n m ) X ( n m ) Y ( n m ) N u m be r f il a m en t s bound barbed end Hip1RActin 80604020101 11020406080 - 40 - 20 0 20 20 - 40 Z ( nm ) 40 40 0 0 - 40 - 40 40 40 0 0 - 40 - 40 40 40 0 0 - 40 - 40 40 40 0 0 - 40 - 40 40 40 0 0 - 40 - 40 40 40 0 0 - 40 - 40 0 5 10 15 100 50 0 95 % i n t e r na li z a t i on ( n m ) Time ( s ) 0 5 10 15 Time ( s ) Time ( s ) 0 5 10 15 0 5 10 15 B a r bed end s nea r ba s e 200 150 100 50 0 200 150 100 50 0 20 10 0 1 % coverage 10 % coverage 20 % coverage 40 % coverage 60 % coverage 80 % coverage 1 % coverage 10 % coverage 20 % coverage 40 % coverage 60 % coverage 80 % coverage Hip1R coverage Figure 4 . Spatial distribution of actin / Hip1R attachments strongly affects actin self - organization and pit internalization . ( A ) Schematic of spatial boundary conditions from endocytic actin - binding proteins . Positions of active Arp2 / 3 complex ( blue ) and actin / pit attachments via linker proteins such as Hip1R ( purple ) . ( B ) Initial positions of Hip1R around increasingly large pit surface area , from 1 % to 80 % of a sphere . The top ~ 20 % of the sphere is occluded by the neck . ( C ) Snapshots of a series of simulations for different values of Hip1R coverage showing actin distribution at t = 13 s . ( D \u2013 G ) Changes in the endocytic actin network over time as a function of Hip1R coverage ( colors ) . n = 96 simulations . ( D ) Internalization ; ( E ) Number of barbed ends near the base of the pit ( within 7 . 5 nm ) ; ( F ) Number of Arp2 / 3 complexes bound in the endocytic network ; ( G ) Number of actin filaments bound in the endocytic network . Scale bar : 50 nm . The online version of this article includes the following video and figure supplement ( s ) for figure 4 : Figure supplement 1 . Relationship between endocytic outcome and active Arp2 / 3 complex surface density or mother filament nucleating protein surface density at the base of the pit . Figure supplement 2 . A collar of active Arp2 / 3 complex near the neck the pit does not affect endocytic outcome . Figure supplement 3 . Internalization as a function of the number of Hip1R molecules and mechanism of self - organization of endocytic actin filaments . Figure 4\u2014video 1 . Simulations in which the coverage of linker Hip1R around the pit was varied from 1 % to 80 % of a sphere . https : / / elifesciences . org / articles / 49840 # fig4video1 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 11 of 40 Research article Cell Biology Physics of Living Systems mother filaments that activate the Arp2 / 3 complex to direct filament growth to the area surrounding the base of the pit . Bending of endocytic actin filaments contributes to endocytic robustness Given that self - organized actin filaments help to nucleate new actin filaments that grow toward the base of the pit , questions remained about how these filaments contribute to and sustain force pro - duction . Closer examination of the simulations revealed that long actin filaments bend between their attachment sites in the clathrin coat and the base of the pit as a result of assembly confined by the membrane at the base of the pit ( Figure 5A ) . We predict that these bent filaments provide a previ - ously unrecognized means of force production by endocytic actin filaments . To test the prediction that actin filaments bend at sites of mammalian endocytosis , we used cryo - electron tomography on intact vitrified mammalian ( SK - MEL - 2 ) cells . SK - MEL - 2 cells grown on electron - microscopy grids are thin at their periphery ( < 1 m m ) , which makes them suitable for electron tomography studies . Indeed , we found bent actin filaments present at sites of clathrin - mediated endocytosis , between the clathrin coat ( Figure 5\u2014figure supplement 1 ) and the base of the pit ( Figure 5B and Figure 5\u2014video 1 ) , in extended \u2018U\u2019 - shaped clathrin - coated pits similar to the stage modeled in our simulations ( Figure 5C \u2013 D ) . What could be a functional consequence of such bent filaments ? We hypothesized that the bent actin filaments store elastic energy that promotes endocytic internalization . We first quantified the filament bending in simulations and found that many ( 13 \u00b1 3 % ) of the actin filaments bend further than can be accounted for by thermal fluctuations ( Boal and Boal , 2012 ; Mogilner and Oster , 1996 ; Figure 5E and Figure 5\u2014figure supplement 2A ) . Most ( 92 % ) of the bent filaments bent less than the minimum energy expected to sever the filaments ( De La Cruz et al . , 2015b ; Sept and McCammon , 2001 ; Figure 5E ) . Importantly , the bent filaments stored elastic energy by collectively continuing to bend over time , storing up to ~ 750 pN ! nm of elastic energy \u2013 mostly in capped fila - ments ( Figure 5F ) . In the context of pit internalization , the amount of elastic energy stored was larger than the magnitude of work required to internalize endocytic pits ( Figure 5F and Figure 5\u2014 figure supplement 2B ) . The elastic energy stored in bent filaments was ~ 1 % of the total energy required to polymerize the endocytic actin network ( Figure 5\u2014figure supplement 2C \u2013 D ) , with pit internalization constituting ~ 0 . 5 % of the total energy from actin filament assembly ( Figure 5\u2014figure supplement 2E ) . The majority ( 62 \u00b1 20 % ) of filament bending energy came from filaments directly bound to Hip1R , and 78 \u00b1 25 % of the bending energy came from filaments with barbed ends > 5 nm from the coat surface ( Figure 5\u2014figure supplement 2F \u2013 H ) . 17 \u00b1 16 % of bending energy came from filaments with barbed ends near the base of the pit ( Figure 5\u2014figure supplement 2I ) . For fila - ments near the base of the pit , the bending energy was distributed radially such that filaments with barbed ends ~ 130 nm from the center of the pit contribute the most bending energy ( Figure 5\u2014 figure supplement 2J ) . Filament bending serves as an important functional consequence of the self - organization of actin filaments at endocytic sites ( Figure 4 ) . With high Hip1R surface coverage around the pit , filaments directed to grow toward the base of the pit bend , storing elastic energy ( Figure 5\u2014figure supple - ment 2K \u2013 L ) . This elastic energy can be harnessed gradually under thermal fluctuations to drive endocytic internalization through a ratchet mechanism ( Mogilner and Oster , 1996 ) . To test the hypothesis that energy stored in bent actin filaments can promote endocytic internali - zation , we conducted simulations in which the resistance from membrane tension was released at a late time point ( t = 10 s , internalization ~ 50 nm ) along with capping filament growth ( Figure 5G ) . This scenario allowed us to test how the stored energy in the bent filaments ( rather than force gener - ated by growing filaments ) can promote internalization in response to an abrupt decrease in tension . We found that once membrane tension decreases , pit internalization sharply increases ( Figure 5H and Figure 5\u2014video 2 ) and filament bending near the base of the pit gradually decreases by 50 % with wide variance ( Figure 5I ) . Thus , we found that in addition to generating force actively by fila - ment growth ( Figure 3 ) , the endocytic actin network stores potential energy in the form of bent fila - ments that can promote internalization even after filaments have stopped growing . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 12 of 40 Research article Cell Biology Physics of Living Systems ! \" # $ \" % & \u2019 ( # $ ) & \u2019 ( # $ * + , ( - . + / 0 , + 1 , + / - 2 + - ( - . + / 0 , + 1 , + / - 2 + - ( - . + / 0 , + + , ( - . + / 0 , + 3 4 5 6 f il a m en t s 21 - 07 . ( 8 . + 90 + : ; 2 / / 07 . 8 . + 90 + : / . 7 . < 0 + : Z = 0 nm Z = 37 nm = > ? @ A . B8 < 2 + . ( B . 1C2 + 01 / ( / 0BDE2 - 0 , + = 1 - 0 + ( B . 1C2 + 01 / ( / 0BDE2 - 0 , + ? < F , % . E . 1 - < , + ( - , B , : < 2B ? D - ( / ; < 0 + : 4 + - . < + 2E0G . / HD < - C . < I ? 2 ; H0E2B . + - / ? D - ( 12 ; ; . 9 0 + - . < + 2E0G2 - 0 , + : < , J0 + : 1500 1000 500 0 200 100 0 90 75 60 45 30 15 0 0 5 10 15 0 5 10 15 0 - 50 - 100 0 5 10 15 0 100 200 300 400 500 Time ( s ) Time ( s ) Time ( s ) Free filament length ( nm ) B end i ng ene r g y ( p N (cid:278) n m ) B end i ng ene r g y ( p N (cid:278) n m ) I n t e r n a li z a t i on ( n m ) B end i ng ang l e ( deg r ee s ) Figure 5 . Bending of endocytic actin filaments stores elastic energy for pit internalization . ( A ) Snapshot of simulation showing filaments bent between the attachment site in the coat and the base of the pit . Also see Figure 1F . Yellow arrowheads point to a bent actin filament . ( B ) Tomographic slice of cryo - electron tomogram of an SK - MEL - 2 cell . Long actin filaments ( yellow arrowheads ) bend along the clathrin - coated pit between the coat and the base of the pit . ( C ) Snapshot of membrane mechanics simulation under an internalization force with 60 nm internalization . ( D ) Slice of the same tomogram as shown in B at a different Z - level ( + 37 nm ) in which the coated plasma membrane ( white arrowheads ) is visible . Scale bar for A - D : 50 nm . ( E ) Heat map of the bending angle and free filament length of endocytic actin filaments in simulations . Color code is number of filaments ( summed for all time points , average of 24 simulations ) . Lines demarcate the magnitude of energy stored in these filaments , based on the theory of elastic beam rigidity for filaments of persistence length 10 m m ( Materials and methods ) , in units of k B T ( 4 . 1 pN ! nm ) . Purple lines : filament conformations expected from thermal fluctuations ( passive bending ) . White lines : filament bending greater than from thermal fluctuations ( active bending ) . Magenta lines : lower Figure 5 continued on next page Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 13 of 40 Research article Cell Biology Physics of Living Systems Inhibiting Arp2 / 3 complex activity stalls endocytosis We next investigated how inhibiting the activity of Arp2 / 3 complex would affect endocytosis ( Fig - ure 6 ) . Our simulations , conducted by varying the nucleation rate of Arp2 / 3 complex , predicted that inhibiting Arp2 / 3 complex activity stalls endocytosis ( Figure 6A ) . Endocytosis was inhibited when Arp2 / 3 complex nucleation rates fell below the basal value of 1 filament per second ( Beltzner and Pollard , 2008 ) , and was insensitive to increased rates of nucleation ( Figure 6B ) . We validated this relationship with experiments modulating Arp2 / 3 complex activity in cells . The small molecule inhibi - tor CK - 666 prevents the Arp2 / 3 complex from nucleating actin filaments ( Hetrick et al . , 2013 ; Figure 5 continued limit for bending energy expected to sever filaments ( De La Cruz et al . , 2015b ) . ( F ) Total elastic energy stored in bent capped ( red ) or growing ( green ) endocytic actin filaments during simulation over time compared to mean energy necessary for internalization ( black ) ( n = 96 simulations ) . ( G ) Schematic of an in silico experiment to test the mechanical function of bent endocytic actin filaments . At t = 10 s , the membrane tension was reduced to zero , and the filaments were capped . ( H ) Internalization ( green ) after spring cut and filament capping , compared to simulation with no change in tension ( black , same data as Figure 3D ) . n = 48 simulations . ( I ) Bending energy of endocytic actin filaments with barbed ends near base of pit over time . Release of tension and filament capping at t = 10 s ( green ) compared to no change in tension ( black ) . The online version of this article includes the following video and figure supplement ( s ) for figure 5 : Figure supplement 1 . Hexagonal and pentagonal lattices in tomogram of clathrin - coated pit . Figure supplement 2 . Energetics of endocytic actin network . Figure 5\u2014video 1 . Cryo - electron tomogram of SK - MEL - 2 cell grown on holey carbon grid and vitrified , related to Figure 5 . https : / / elifesciences . org / articles / 49840 # fig5video1 Figure 5\u2014video 2 . Simulation of actin in endocytosis in which , at t = 10 s , filaments were all capped and the membrane tension was reduced to 0 pN / nm . https : / / elifesciences . org / articles / 49840 # fig5video2 ! \" # Arp2 / 3 nucleation rate ( filaments per second ) A $ % & \u2019 ( ) * + , - & . / 0 12 + . / 345 , 6 # 7888 100 75 50 25 0 0 0 . 2 0 . 4 0 0 . 2 0 . 4 0 0 . 2 0 . 4 0 50 100 150 200 250 0 0 . 2 0 . 4 Lifetime ( s ) Time ( s ) F r equen cy 9 * : ; * # 7888 < * : ; * # 7888 \u2019 < * : ; * # 7888 < 9 * : ; * # 7888 # . 34 = % 56 > ? @ A * ! B63 - 56\u2019 > C @ A 95 % i n t e r na li z a t i on ( n m ) 0 5 10 ref - track lifetimes : 10 . 00 - 87 . 34 ( n = 182 ) , 87 . 34 - 164 . 67 ( n = 14 ) , 164 . 67 - 242 . 01 ( n = 18 ) , - 200 - 150 - 100 - 50 0 0 0 . 2 0 . 4 0 . 6 0 . 8 A v e r age I n t en s i t y ref - track lifetimes : 14 . 00 - 90 . 00 ( n = 19 ) , 90 . 00 - 166 . 01 ( n = 3 ) , 166 . 01 - 242 . 01 ( n = 3 ) , - 200 - 150 - 100 - 50 0 0 0 . 2 0 . 4 0 . 6 0 . 8 A v e r age I n t en s i t y ref - track lifetimes : 10 . 00 - 87 . 34 ( n = 71 ) , 87 . 34 - 164 . 67 ( n = 10 ) , 164 . 67 - 242 . 01 ( n = 13 ) , - 200 - 150 - 100 - 50 0 0 0 . 2 0 . 4 0 . 6 0 . 8 A v e r age I n t en s i t y cor - track lifetimes : 10 . 00 - 87 . 34 ( n = 21 ) , 87 . 34 - 164 . 67 ( n = 8 ) , 164 . 67 - 242 . 01 ( n = 6 ) , - 200 - 150 - 100 - 50 0 0 0 . 2 0 . 4 0 . 6 0 . 8 A v e r age I n t en s i t y # . 34 = % 56 > ? @ A * ! B63 - 56\u2019 > C @ A N o r m a li z ed i n t en s i t y ( A . U . ) 9 * : ; * # 7888 < * : ; * # 7888 \u2019 < * : ; * # 7888 < 9 * : ; * # 7888 Figure 6 . Inhibiting Arp2 / 3 complex nucleation activity stalls endocytosis . ( A ) Schematic of the model parameter corresponding to Arp2 / 3 nucleation activity , and the step inhibited by the small molecule CK - 666 . ( B ) Internalization as a function of Arp2 / 3 complex nucleation rate . Orange region highlights parameter sensitivity , and green region highlights parameter insensitivity . n = 96 simulations . Reducing Arp2 / 3 nucleation rate reduces internalization as seen in the orange region . ( C ) Histograms of endocytic lifetime in SK - MEL - 2 cells endogenously expressing clathrin light chain CLTA - tagRFP - T and dynamin2 - eGFP and treated with CK - 666 . n = 368 tracks from 10 cells . ( D ) Fluorescence intensity over time for endocytic tracks marked by clathrin - RFP and dynamin2 - GFP in SK - MEL - 2 cells treated with 0 . 1 % DMSO ( 0 m M ) or the indicated concentration of CK - 666 for 45 min . Fluorescence events were tracked automatically ( Materials and methods ) . Tracks in which GFP and RFP colocalized are shown . Each track was normalized to its maximum intensity and then all tracks were averaged and aligned to the time of the disappearance of the clathrin - RFP signal . The lifetimes of these events are plotted in D . Shaded bars are standard deviations . The online version of this article includes the following figure supplement ( s ) for figure 6 : Figure supplement 1 . Effect of Arp2 / 3 complex inhibitor CK - 666 on lifetimes of endogenously tagged markers of endocytosis . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 14 of 40 Research article Cell Biology Physics of Living Systems Nolen et al . , 2009 ) . Treatment of SK - MEL - 2 cells with CK - 666 inhibited endocytic progression , as marked by the lifetimes of endogenously tagged AP2 - RFP or dynamin2 - GFP at endocytic sites , in a dose - dependent and time - dependent manner ( Figure 6C \u2013 D and Figure 6\u2014figure supplement 1A ) . Adaptation of the endocytic actin network to changes in membrane tension Because we and others previously modeled that membrane tension plays an important role in mem - brane bending during the formation of an endocytic pit ( Hassinger et al . , 2017 ; Rangamani et al . , 2014 ; Walani et al . , 2014 ) we next varied the value of membrane tension in simulations to under - stand the relationship between tension , actin filament bending , and actin assembly ( Figure 7A ) . In our simulations , endocytic progression attenuated in a tension - dependent manner ( Figure 7A \u2013 E ) , consistent with previous modeling ( Hassinger et al . , 2017 ) and experimental observations ( Boulant et al . , 2011 ; Ferguson et al . , 2017 ; Ferguson et al . , 2016 ; Wu et al . , 2017 ) . However , at higher membrane tensions , endocytosis persisted better than expected for a non - adapting network , suggesting the existence of an adaptive mechanism ( Figure 7E ) . Therefore , we sought to under - stand how endocytic actin networks adapt to increases in load . We found that under low tension ( 0 . 015 pN / nm ) , endocytic pits internalize strongly ( Figure 7B and E ) and few barbed ends encounter the base of the pit ( Figure 7F ) , with fewer Arp2 / 3 complexes recruited to the network ( Figure 7\u2014figure supplement 1A ) and a correspondingly low filament bending energy ( Figure 7H ) . Under > 50 x higher membrane tension ( 1 pN / nm ) , endocytic ! \" # $ % & \u2019 ( ) * + , - & + . * / - 01 ( ) * + , - & + 2 - 34 ( ) * + , - & + 5 6 7 2 8 * + , - & + ! 9 ) - + : / : ; ) : ) - & + < 6 - = : 1 * + ) > * + / - + 3 B a r bed end s nea r ba s e A c t i n f il a m en t s 95 % i n t e r na li z a t i on ( n m ) Tension ( pN / nm ) Tension ( pN / nm ) Tension ( pN / nm ) Tension ( pN / nm ) 100 50 0 0 . 01 0 . 1 1 10 0 . 01 0 . 1 1 10 0 . 01 0 . 1 1 10 0 . 01 0 . 1 1 10 B end i ng ene r g y ( p N (cid:278) n m ) Figure 7 . Adaptation of endocytic actin network to changes in membrane tension . ( A ) Schematic depicting possible adaptation of the actin network to membrane tension via self - organization and bending . ( B \u2013 D ) Snapshots of simulations from the same time point ( 14 s ) for ( B ) low membrane tension ( 0 . 015 pN / nm ) ; ( C ) medium membrane tension ( 0 . 15 pN / nm ) ; ( D ) high membrane tension ( 1 pN / nm ) . Scale bar is 50 nm . ( E \u2013 H ) Changes in the endocytic actin network as a function of membrane tension . n = 144 simulations . ( E ) Internalization ; ( F ) Number of barbed ends near base of pit ; ( G ) Number of actin filaments in Hip1R - bound network ; ( H ) Bending energy for filaments with barbed ends near base of pit . Mean \u00b1 standard deviation of time points in the last 5 s of simulations . Dashed line in ( E ) is expected internalization based on constant energy usage with 0 . 01 pN / nm condition as reference ( see Methods ) . The online version of this article includes the following video and figure supplement ( s ) for figure 7 : Figure supplement 1 . Membrane tension - dependent adaptation by the actin network . Figure 7\u2014video 1 . Simulations of actin in endocytosis with different values of membrane tension ( Low , 0 . 01 pN / nm ; Medium , 0 . 15 pN / nm ; High , 1 pN / nm ) . https : / / elifesciences . org / articles / 49840 # fig7video1 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 15 of 40 Research article Cell Biology Physics of Living Systems internalization slowed but was not abolished ( Figure 7E ) . For these pits , more barbed ends encoun - tered the base of the pit ( Figure 7F ) , binding more Arp2 / 3 complexes ( Figure 7\u2014figure supple - ment 1A ) to nucleate more actin filaments ( Figure 7G ) and increasing the total actin filament bending energy near the base of the pit ( Figure 7H ) . As a result , upon increasing membrane ten - sion , the overall endocytic energy efficiency increased ( Figure 7\u2014figure supplement 1B ) . Thus , the self - organization of the endocytic actin network allows it to adapt to elevated membrane tension by nucleating more filaments at the base of the pit . Arp2 / 3 complex activity and Hip1R / actin attachments are critical for allowing actin filaments to drive endocytic pit internalization and adapt to changing tension Having established that endocytic internalization depends on two spatially confined boundary condi - tions \u2013 Hip1R / actin attachments at the curved pit ( Figure 4 ) and active Arp2 / 3 complex activity at the base of the pit ( Figure 6 ) \u2013 we next investigated how these boundary conditions alter the endo - cytic response to membrane tension ( Figure 8A ) . We systematically varied membrane tension and Arp2 / 3 complex activity in our model to gener - ate a phase diagram of endocytic internalization as a function of membrane tension and Arp2 / 3 com - plex activity ( Figure 8B ) . This phase diagram shows that cells with high membrane tension are especially sensitive to changes in Arp2 / 3 complex nucleation rate ( Kaplan et al . , in preparation ) , whereas cells with low membrane tension carry out endocytosis even with low Arp2 / 3 complex activ - ity , consistent with experimental observations ( Boulant et al . , 2011 ) . We hypothesized that actin network self - organization arising from the broad Hip1R distribution around the pit ( Figure 4 ) and filament bending ( Figure 5 ) might allow for the endocytic actin net - work to change its organization and force - producing capacity under elevated loads ( Figure 7 ) . To test this hypothesis , we conducted simulations in which Hip1R coverage was varied for different val - ues of plasma membrane tension ( Figure 8A and Figure 7\u2014video 1 ) . We found that the endocytic actin network\u2019s ability to adapt to load ( Figure 7 ) depends on Hip1R coverage around the pit ( Figure 8D \u2013 F ) . As the coverage of Hip1R around the pit increased , actin\u2019s ability to adapt to changes in membrane tension also increased , as measured by the number of barbed ends near the base of the pit ( Figure 8D ) , the binding of active Arp2 / 3 complex at the base of the pit ( Figure 8E ) , subsequent nucleation of additional actin filaments ( Figure 8F ) , and bending of actin filaments near the base of the pit ( Figure 8G ) . We conclude that sufficient Hip1R coverage around the pit ( Clarke and Royle , 2018 ; Sochacki et al . , 2017 ) allows endocytic actin filaments to orient in such a way that they can encounter more Arp2 / 3 complexes at the base of the pit to nucleate more actin fil - aments . This spatial organization allows the actin network to adapt to sustain force production under a range of opposing loads ( Figure 8H ) . Discussion Understanding the relationship between actin filament assembly , actin network organization , and force generation on the plasma membrane requires iterative feedback between experimental meas - urements and computational modeling . An ultimate goal of this study was to relate single actin fila - ment mechanics to force generation by the collective actin filament network in CME ( Lacayo et al . , 2007 ) . We integrated modeling and quantitative cellular measurements to show that a minimal actin network composed of actin , the Arp2 / 3 complex and capping protein , with linker attachments in the clathrin coat and rates constrained by cellular and biochemical measurements , is able to generate sufficient force to internalize endocytic pits against mammalian plasma membrane tension . Approxi - mately 200 Arp2 / 3 complexes constitutively assemble at sites of endocytosis in human induced plu - ripotent stem cells . Endocytic actin filaments self - organize into a radial dendritic array , in which filaments grow toward the base of the pit . These filaments bend and store elastic energy , which sup - ports internalization . The endocytic actin network adapts to changes in membrane tension by driving more filaments to the base of the pit and increasing filament bending , which supports a higher load and nucleates more actin filaments . Four lines of experimental evidence support our model ( Figure 8\u2014figure supplement 1 ) . Two pieces of evidence serve as model validation based on published data and two more are based on experiments conducted in this study . Previous experiments from our lab showed that knocking down Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 16 of 40 Research article Cell Biology Physics of Living Systems Hip1R in HeLa cells inhibited endocytosis ( Engqvist - Goldstein et al . , 2004 ) . Consistent with these observations , our simulations showed that a threshold number of actin linkers such as Hip1R is nec - essary for endocytic internalization ( Figure 4\u2014figure supplement 3A and Figure 8\u2014figure supple - ment 1A \u2013 B ) . This threshold appears necessary to allow efficient filament capture by the coat and force transmission from the actin network to the coat . Second , experiments showed that capping actin filament elongation with the small molecule compound Cytochalasin inhibits CME , reflected by an increase in stalled endocytic sites marked by clathrin - RFP ( Brady et al . , 2010 ) or slower A r p2 / 3 c o m p l e x D Low Medium High E F Hip1R coverage ( % ) Hip1Rcoverage Tension Tension 0 20 40 60 80 100 50 0 95 % i n t e r na li z a t i on ( n m ) 3 . Nucleates new filaments Actin Arp 2 / 3 complex Clathrin N - WASP Hip1R 1 . Broad Hip1R coverage in clathrin coat 2 . Drives barbed ends to base 4 . Filaments bend to store elastic energy T en s i on ( p N / n m ) 0 . 1 0 . 01 0 . 001 0 . 0001 Arp2 / 3 nucleation rate ( filaments per second ) 0 1 10 H A B C 95 % internalization ( nm ) Hip1R coverage ( % ) Hip1R coverage ( % ) Hip1R coverage ( % ) Hip1R coverage ( % ) G B a r bed e n d s nea r ba s e A c t i n f il a m en t s B ound A r p2 / 3 c o m p l e x B end i ng e n e r g y ( p N (cid:278) n m ) Figure 8 . Arp2 / 3 complex activity and Hip1R / actin attachments are critical for allowing actin filaments to drive endocytic pit internalization and adapt to changing tension . ( A ) Schematic of Arp2 / 3 complex activity and Hip1R coverage along with membrane tension . ( B ) Phase diagram of endocytic internalization as a function of membrane tension and Arp2 / 3 complex nucleation rate shown on a log - log plot . Dotted lines are values taken from the literature ( Beltzner and Pollard , 2008 ; Diz - Mun\u02dcoz et al . , 2016 ) . ( C \u2013 G ) Changes in the endocytic actin network as a function of Hip1R coverage for different values of membrane tension . Low tension = 0 . 015 pN / nm ; medium tension = 0 . 15 pN / nm ; high tension = 1 pN / nm . n = 288 simulations . ( C ) Internalization ; ( D ) Number of barbed ends near base of pit ; ( E ) Number of Arp2 / 3 complexes bound in network ; ( F ) Number of actin filaments bound in network ; ( G ) Bending energy of filaments with barbed ends near the base of the pit . Mean \u00b1 standard deviation of time points in the last 5 s of simulations . ( H ) Summary of load - dependent adaptation of self - organizing endocytic actin network due to spatial segregation of active Arp2 / 3 complex at the base and Hip1R in a broad distribution within the clathrin coat . The online version of this article includes the following figure supplement ( s ) for figure 8 : Figure supplement 1 . Summary of predictions of the model supported by experimental data in the current manuscript and in the literature . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 17 of 40 Research article Cell Biology Physics of Living Systems accumulation of dynamin2 - GFP at endocytic sites ( Grassart et al . , 2014 ) . We also showed in our model that capping rate is an important parameter for progression of CME ; our simulations show that increasing the capping rate of actin filaments inhibits CME , presumably because increasing cap - ping decreases the total amount of actin ( Figure 1\u2014figure supplement 1C and Figure 8\u2014figure supplement 1C \u2013 D ) . In this study , our simulations predicted that actin filaments bend around endo - cytic pits . These bent filaments store elastic energy for subsequent force production much as a pole vaulter\u2019s pole bends and stores energy for delayed force production . Using cryo - electron tomogra - phy of intact cells , we observed bent actin filaments at sites of endocytosis in mammalian cells ( Fig - ure 5 and Figure 8\u2014figure supplement 1E \u2013 F ) . Finally , we also predicted that inhibiting Arp2 / 3 complex activity below its basal nucleation rate of 1 filament per second inhibits endocytosis in silico and this prediction was validated in cells using pharmacological agents ( Figure 6C , D and Figure 8\u2014 figure supplement 1G \u2013 H ) . Without sufficient Arp2 / 3 complex , CME fails due to insufficient force production . Three main conclusions resulted from our study . First , we found that the spatial segregation of Arp2 / 3 complex activation and Hip1R linker proteins on the clathrin coat are important factors for effective force generation . Unlike actin organization at the leading edge of a migrating cell wherein only one boundary condition at the plasma membrane is sufficient to enable force - generation capac - ity to be inferred ( Abercrombie , 1980 ; Bieling et al . , 2016 ; Mogilner and Edelstein - Keshet , 2002 ) , in CME two boundary conditions are required \u2013 one at the base of the pit for actin polymeri - zation against the plasma membrane and the second on the coat of the pit for attachment of the growing actin filaments so forces are transmitted to the pit to drive internalization . In our model , we used Hip1R as a surrogate for Hip1 , Hip1R , and Epsin1 / 2 / 3 , which cooperatively link actin filaments to the coat ( Brett et al . , 2006 ; Chen et al . , 1998 ; Messa et al . , 2014 ; Senetar et al . , 2004 ; Skruzny et al . , 2012 ) . We conclude that based on the relative positions of Arp2 / 3 complex activa - tors and actin filament linkers , the resultant self - organized actin network orients to produce force orthogonal to the base of the pit rather than producing a pinching force on the neck ( Collins et al . , 2011 ; Hassinger et al . , 2017 ) . Pinching forces are achieved by the spontaneous curvature of the rigid coat ( Alimohamadi et al . , 2018 ; Foret , 2014 ; Hassinger et al . , 2017 ) . Any constriction forces generated by actin polymerization at the neck would likely occur at a later stage of endocytosis than is the focus of our model , and the filaments would need to be nucleated by a spatially distinct set of Arp2 / 3 activating proteins around the neck , or by an interaction between other actin filaments and dynamin , but the mechanism for arranging and anchoring such a network has not been elucidated ( Ma and Berro , 2018 ) . Second , the effective anchoring of actin filaments to the surface of the pit depends on the distri - bution of linker proteins on the pit surface . Since these linker proteins are embedded within the cla - thrin coat ( Clarke and Royle , 2018 ; Engqvist - Goldstein et al . , 2001 ; Sochacki et al . , 2017 ) , their surface coverage is directly proportional to the coat coverage on the endocytic pit . This observation suggests that one possible function of a large coat is for the actin - linking proteins Hip1 , Hip1R and Epsin to cover enough surface area to provide leverage for internalization . The role of linker protein coverage in force generation also has implications for the flat - to - curved transition earlier in endocy - tosis , when the membrane either begins to bend from a flat pre - assembled coat or continually deforms while the coat assembles ( Avinoam et al . , 2015 ; Bucher et al . , 2018 ; Scott et al . , 2018 ; Sochacki and Taraska , 2019 ) . In cases when the clathrin coat gradually increases in area during membrane deformation , our findings imply that actin polymerization may be ineffective until the coat reaches a threshold size ( Avinoam et al . , 2015 ; Sun et al . , 2017 ) , with membrane tension con - trolling a switch for the flat - to - curved transition ( Bucher et al . , 2018 ; Scott et al . , 2018 ) . Future work will investigate the relationship between coat topology and actin forces during the initiation of endocytosis . Third , we showed a significant fraction of endocytic actin filaments bend under force . We predict that the bent filaments , whose existence we confirmed by cryo - electron tomography of intact cells , contribute to successful endocytic internalization in at least two ways . First , they might contribute to the resilience of endocytosis by preventing backsliding of the pit . Second , we expect that they con - tribute to internalization by releasing stored elastic energy when they straighten out under thermal fluctuations , consistent with the elastic Brownian ratchet mechanism for actin - mediated force pro - duction ( De La Cruz and Gardel , 2015a ; Mogilner and Oster , 1996 ) . Here , filament bending occurs to a greater extent and for a longer time than previously described for coherent flat surfaces like the Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 18 of 40 Research article Cell Biology Physics of Living Systems leading edge , possibly due to the curved geometry of endocytic pits . Fixing the filament orientation at one end increases bending energy ( De La Cruz et al . , 2015b ; Ferna\u00b4ndez et al . , 2006 ) , which is accomplished here by multivalent attachments from Hip1R . Previous studies overlooked the role of actin filament bending at endocytic sites because of the predicted short length of filaments based on population averages ( Berro et al . , 2010 ) and the possible loss of less densely branched filaments during the preparation process for platinum - replica electron microscopy ( detergent extraction or sonication - based unroofing ) ( Collins et al . , 2011 ) . The load response of branched actin networks in vitro can be reversible due to filament bending ( Chaudhuri et al . , 2007 ) , or permanent from a change in filament architecture ( Bieling et al . , 2016 ; Parekh et al . , 2005 ) . In our simulations , some of the elastic energy from bent filaments is released as internalization increases , suggesting a revers - ible compression of the network to store elastic energy ( Figure 5 ) . However , a significant fraction of filament bending is retained after the spring is released , which suggests that load also changes the intrinsic structure of the network ( Bieling et al . , 2016 ) . Importantly , the results presented here demonstrate a mechanism of active adaptation by the endocytic actin network to changes in load ( Figure 7 ) . Different cell types , different locations in the same cell , and different stages of endocytosis at the same location can have different membrane tension values at different times ( Shi et al . , 2018 ) . Under flat geometries , branched actin networks adapt to load ( Bieling et al . , 2016 ; Mueller et al . , 2017 ) . Here , the distribution of Hip1R linkers around the pit directs more filaments to grow toward the base of the pit ( Figure 4 ) , which nucleates more filaments autocatalytically and increases filament bending ( Figure 5 ) , thereby supporting greater internalization ( Figure 7 ) . It is now important to determine whether the principles of actin filament self - organization and load adaptation identified here also apply to endocytic actin in the higher force regime characteristic of fungi . An agent - based model of endocytic actin networks in yeast predicted that barbed filament ends radiate away from the center of pit in the XY plane ( Mund et al . , 2018 ) . However , the > 200 x larger force requirements in this organism lead to a different axial organization of the filaments , less filament bending , and a distinct mechanism of force production sufficient to counteract high turgor pressure . Understanding the mechanical function and load adaptation in the \u2018soft\u2019 regime studied here is likely to apply to a variety of cellular membrane bending processes employing branched actin networks , including the assembly and maturation of endosomes , lysosomes , and autophagosomes ( Rottner et al . , 2017 ) . Finally , we acknowledge that our model represents a minimal core actin machinery necessary for endocytic internalization in mammalian cells . This feature of our study was necessary so the number of free parameters could be limited . Future models will add complexity to test roles for filament crosslinking , filament severing , and myosin - I motor activity , among other factors . It is anticipated that these additional features will increase our understanding of the force - generation capability and overall efficiency of the endocytic actin network . Materials and methods Key resources table Reagent type ( species ) or resource Designation Source or reference Identifiers Additionalinformation Gene ( H . sapiens ) ArpC3 HGNC : 706 ARPC3 Cell line ( human ) WTC - 10 hiPSC AP2 - tagRFP - t ArpC3 - tagGFP2 This study Cell line maintained in D . Drubin lab Transfected construct ( H . sapiens ) 12mer - tagGFP2 - FKBP This study Plasmid to transiently express calibration construct Transfected construct ( H . sapiens ) 24mer - tagGFP2 - FKBP This study Plasmid to transiently express calibration construct Continued on next page Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 19 of 40 Research article Cell Biology Physics of Living Systems Continued Reagent type ( species ) or resource Designation Source or reference Identifiers Additionalinformation Transfected construct ( H . sapiens ) 60mer - tagGFP2 - FKBP This study Plasmid to transiently expresscalibration construct Transfected construct ( H . sapiens ) 120mer - tagGFP2 - FKBP This study Plasmid to transiently expresscalibration construct Antibody mouse monoclonal anti - GAPDH ProteinTech 10494 \u2013 1 - AP ( RRID : AB _ 2263076 ) ( 1 : 5000 dilution ) Antibody tag ( C , G , Y ) FP Evrogen 12101231265 ( 1 : 2500 dilution ) Sequence - based reagent ArpC3 crRNA This paper crRNA CCGGGCUCCCUUCACUGUCC Sequence - based reagent ArpC3 _ sequencing primer This paper PCR primers ACTTATTCTTATTAAGCGCCAGC Sequence - based reagent ArpC3 _ sequencing primer This paper PCR primers CAGGGCTCTGGAGACGGT Commercialassayorkit Lipofectamine Stem Thermo Fisher STEM00003 Chemical compound , drug AP21967 Clontech 635056 Chemical compound , drug CK - 666 Sigma SML0006 Software , algorithm Cytosim Nedelec and Foethke , 2007 https : / / github . com / DrubinBarnes / Akamatsu _ CME _ manuscript Software , algorithm MATLAB Mathworks , Inc R2017b Software , algorithm Python Python . org 3 . 7 Software , algorithm m - Track Jaqaman et al . , 2008 https : / / github . com / DrubinBarnes / Akamatsu _ CME _ manuscript Mathematical modeling We combined a continuum membrane mechanics model with filament - based simulations of actin polymerization coupled to endocytic pit internalization to develop a multiscale model of actin in mammalian endocytosis . In the continuum model , the bending of the membrane was explicitly mod - eled and the contributions of the actin cytoskeleton were simplified to an applied localized force , while the agent - based model simplified the membrane and explicitly modeled actin and associated binding proteins ( Scheme 1 ) . We iteratively ran simulations in each module . The results from each module improved the assumptions of the other , over successive rounds of simulation and comparison to experimental measurements . In particular , experiments focused our attention on the internalization of the U - shaped pit and its transition to omega shape for both the membrane and actin modules . Experi - ments and the actin module informed the location of actin forces for the membrane mechanics model . The membrane mechanics simulations in turn informed the initial pit shape and force / exten - sion relationship for the actin module . Thus , these two modules synergistically provided collective information about how actin organization and dynamics couple to the bending and internalization of the clathrin - coated pit . Membrane mechanics module Continuum mechanics modeling of the plasma membrane allows a quantitative understanding of the relationship between applied forces and the shape of the membrane ( Dere\u00b4nyi et al . , 2002 ; Rangamani et al . , 2013 ) . Bending the membrane requires energy , such that pulling a tether from a flat membrane requires increasing force until the membrane adopts a tubule shape ( Dere \u00b4 nyi et al . , 2002 ; Alimohamadi et al . , 2018 ) . Adding a region with spontaneous curvature ( corresponding to the endocytic coat ) can help lower this energy barrier against moderate membrane tension ( Rangamani et al . , 2013 ; Hassinger et al . , 2017 ) . Forces due to actin polymerization can also help overcome the energy barrier ( Hassinger et al . , 2017 ) , but the relationship between applied actin Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 20 of 40 Research article Cell Biology Physics of Living Systems forces and coated membrane shape has not been explored quantitatively . The following assump - tions guide our model of the coated plasma membrane : . Membrane curvature generated due to forces or protein - induced spontaneous curvature is much larger than the thickness of the bilayer . Based on this assumption , we model the lipid bilayer as a thin elastic shell with a bending energy given by the Helfrich - Canham energy , which is valid for radii of curvatures much larger than the thickness of the bilayer ( Helfrich , 1973 ) . . We neglect the surrounding fluid flow or inertial dynamics and assume that the membrane is at mechanical equilibrium at all times ( Naghdi , 1957 ; Steigmann et al . , 2003 ) . This assump - tion is commonly used in the modeling of membrane curvature to keep the mathematics trac - table ( Alimohamadi et al . , 2018 ; Hassinger et al . , 2017 ; Rangamani et al . , 2014 ; Steigmann et al . , 2003 ; Vasan et al . , 2018 ) . . The membrane is incompressible because the energetic cost of stretching the membrane is high ( Steigmann et al . , 2003 ; Steigmann , 1999 ) . This constraint is implemented using a Lagrange multiplier ( Alimohamadi et al . , 2018 ; Rangamani et al . , 2014 ; Rangamani et al . , 2013 ) . . Finally , for simplicity in the numerical simulations , we assume that the membrane in the region of interest is rotationally symmetric . The following derivation can also be found in Hassinger et al . ( 2017 ) . We use a modified form of the Helfrich energy defined as W \u00f0 H ; K ; ! a \u00de \u00bc k \u00f0 H \" C \u00f0 ! a \u00de\u00de 2 \u00fe ! k K A general force balance on the membrane can be written as ! ! s \u00fe p n \u00bc f ; where is ! ! surface divergence , s is the stress vector , p is the pressure difference between the inside and outside of the volume bounded by the membrane , and f is any externally applied force per unit area on the membrane . The stress vector can be split into normal and tangential components as s a \u00bc T a \u00fe S a n ; where T a \u00bc T ab a b ; T ab \u00bc s ab \u00fe b b \" M \" a ; S a \u00bc \" M ab ; b : s ab and M ab can also be written as s ab \u00bc # q F \u00f0 # ; H ; K ; x a \u00de q a ab \u00fe q F \u00f0 # ; H ; K ; x a \u00de q a ba ! \" ; M ab \u00bc # 2 q F \u00f0 # ; H ; K ; x a \u00de q b ab \u00fe q F \u00f0 # ; H ; K ; x a \u00de q b ba ! \" ; Here \u00f0 a ab \u00de \u00bc \u00f0 a ab \u00de is the dual metric or first fundamental form , b ab is the second fundamental form and # is the surface mass density . Using the first and second fundamental form , we can define H ( mean curvature ) and K ( Gaussian curvature ) as H \u00bc 1 2 a ab b ab ; K \u00bc 1 2 \" ab \" l \" b al b b \" : where \" ab is the permutation tensor defined by \" 12 \u00bc \" \" 21 \u00bc 1 \ufb03\ufb03 a p ; \" 11 \u00bc \" 22 \u00bc 0 . We then define an area incompressibility constraint by rewriting the free energy density as F \u00f0 # ; H ; K ; x a \u00de \u00bc ~ F \u00f0 H ; K ; x a \u00de \" g \u00f0 x a ; t \u00de # : where g \u00f0 x a ; t \u00de is a Lagrange multiplier field required to impose invariance of # on the whole of the surface . This free energy density relates to the Helfrich energy density as Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 21 of 40 Research article Cell Biology Physics of Living Systems W \u00bc # ~ F Combining these equations , we can get the stress equations s ab \u00bc \u00f0 l \u00fe W \u00de a ab \" \u00f0 2 HW H \u00fe 2 k W K \u00de a ab \u00fe W H ~ b ab ; and M ab \u00bc 1 2 W H a ab \u00fe W k ~ b ab ; where l \u00bc \" \u00f0 g \u00fe W \u00de Simplifying this further , we can get the shape equation ( normal balance ) p \u00fe f ! n \u00bc D 1 2 W H \u00fe\u00f0 W K \u00de ; ab ~ b ab \u00fe W H \u00f0 2 H 2 \" K \u00de\u00fe 2 H \u00f0 KW K \" W \u00de \" 2 l H ; and the tangential balance q W q x a j exp \u00fe l ; a ! a ba \u00bc f : a s : where \u00f0\u00de j exp denotes the explicit derivative respect to coordinate ! a . To further simplify the equa - tions , we define the coordinate system as axisymmetric using r \u00f0 s ; ! \u00de \u00bc r \u00f0 s \u00de e r \u00f0 ! \u00de\u00fe z \u00f0 s \u00de k : We define a such that r 0 \u00f0 s \u00de \u00bc cos \u00f0 \u00de , z 0 \u00f0 s \u00de \u00bc sin \u00f0 \u00de and n \u00bc \" sin e r \u00f0 ! \u00de\u00fe cos k Using this , we can write the mean curvature ( H ) and Gaussian curvature ( K ) as H \u00bc 1 2 \u00f0 k v \u00fe k t \u00de \u00bc 1 2 \u00f0 0 \u00fe r \" 1 sin \u00de K \u00bc k t k v \u00bc 0 sin r : We also introduce L \u00bc 1 2 k r \u00f0 W H \u00de 0 allowing us to formulate a system of ordinary differential equations ( ODE\u2019s ) as function of arc length s r 0 \u00bc cos ; z 0 \u00bc sin ; r 0 \u00bc 2 rH \" r 0 \u00bc \" sin ; rH 0 \u00bc L \u00fe rC 0 ; L 0 r \u00bc pk \u00fe f ! n k \u00fe 2 H \u00f0 H \" C \u00de 2 \u00fe lk \" 2 \u00f0 H \" C \u00de h i H 2 \u00fe\u00f0 H \" r \" 1 sin \u00de 2 h i l 0 \u00bc 2 k \u00f0 H \" C \u00de C 0 \" f ! a s : This can also be written as a function of membrane area using a \u00f0 s \u00de \u00bc 2 p Z s 0 r \u00f0 $ \u00de d $ ! da ds \u00bc 2 p r : Here , we choose to non - dimensionalize the system using a \u00bc a 2 p R 2 0 ; x \u00bc r R 0 ; y \u00bc y R 0 ; h \u00bc HR 0 ; c \u00bc CR 0 ; l \u00bc LR 0 l \u2019 \u00bc l R 20 k 0 ; p \u2019 \u00bc pR 30 k 0 f \u2019 \u00bc fR 30 k 0 ; k \u2019 \u00bc k k 0 giving us the system of equations Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 22 of 40 Research article Cell Biology Physics of Living Systems x _ x \u00bc cos ; x _ y \u00bc sin ; x 2 _ \u00bc 2 xh \" sin ; x 2 _ h \u00bc l \u00fe x 2 _ c ; _ l \u00bc p \u2019 k \u2019 \u00fe f \u2019 ! n k \u2019 \u00fe 2 h \u00f0 h \" c \u00de 2 \u00fe l \u2019 k \u2019 h i \" 2 \u00f0 h \" c \u00de h 2 \u00fe\u00f0 h \" x \" 1 sin \u00de 2 h i ; _ l \u2019 \u00bc 2 k \u2019 \u00f0 h \" c \u00de _ c \" f \u2019 ! a s x : We define a spatially varying spontaneous curvature as c \u00bc c 0 \u2019 0 : 5 \u00f0 1 \" tanh \u00f0 g \u2019\u00f0 a \" a 0 \u00de\u00de\u00de where a is the non - dimensional membrane area , a 0 is the non - dimensional membrane area of the protein coat , g is a constant and c 0 is the coat spontaneous curvature . The parameters used for the spontaneous curvature simulations are specified in Supplementary file 1 : To perform the coat pulling simulations , we applied an axial force acting downward along the protein coat and upward along the base of the pit such that the net force integrates to 0 ( we do this by scaling the applied force by the area over which it is applied ) . This force function was defined as f \u00bc f 0 \u2019\u00f0 0 : 5 \u00f0\u00f0 1 \" tanh \u00f0 g \u2019\u00f0 a \" a 0 \u00de\u00de\u00de = a 0 \" \u00f0 tanh \u00f0 g \u2019\u00f0 a \" a in \u00de\u00de \" tanh \u00f0 g \u2019\u00f0 a \" a out \u00de\u00de\u00de = \u00f0 a out \" a in \u00de\u00de\u00de where a 0 is the non - dimensional coat area , a is the non \" dimensional membrane area , a out \" a in is the area of force applied at the base of the pit . a in corresponds to an inner radius r in and a out corre - sponds to an outer radius r out within which the upward force is applied . These parameters are specified in Supplementary file 2 . To simulate the pinched ( \u2018omega - shaped\u2019\u201d ) curves at high membrane tension , we provided an ini - tial guess of an \u2018omega - shaped\u2019 membrane from a lower membrane tension . We did this because the simulations stalled at U shapes at an internalization of about 100 nm . Providing this initial guess led to solutions for membrane shape and force beyond 100 nm , as seen in Figure 1C . Further , to fully explore the space of solutions , we ran the simulations backward by starting from an \u2018omega shaped\u2019 pit at a large internalization and then decreasing the internalization . In Figure 1C , we plot - ted the curves for negative internalization > = the farthest internalization for the U shaped pit ( generally ~ 100 nm ) . Values of membrane tension in Figure 1C are [ 0 : 051 : 0 . 05 : 0 . 451 ] pN / nm , and rounded to two significant digits in the figure for clarity . Actin module We used Cytosim ( Nedelec and Foethke , 2007 ) to model the polymerization of a branched actin network coupled to the internalization of a clathrin - coated pit . This approach simplified the pit as a bead attached to a flat boundary ( the plasma membrane ) by a spring . This assumption of a linear force - extension relationship was validated in Figure 1 . Actin filaments and actin - binding proteins ( Arp2 / 3 complex , Hip1R ) were explicitly simulated as individual objects ( agents ) . Cytosim calculates the forces on each segment of actin from rules such as diffusion , confinement , growth , and binding based on Brownian dynamics . Assumptions in Cytosim Cytosim simulates the movement of actin model points within a boundary according to constrained Langevin dynamics ( Nedelec and Foethke , 2007 ) , which accounts for the diffusion , bending , and forces of actin filaments , as well as the diffusion and binding of actin - binding proteins detailed below . 1 . Force balance equations ( Nedelec and Foethke , 2007 ) section 7 . 1 : a . All points in the simulation follow constrained Langevin dynamics : b . dx \u00bc \" F \u00f0 x ; t \u00de dt \u00fe dB \u00f0 t \u00de i . where m is defined as an effective mobility , which takes on a different value for each type of object . 2 . Mobilities of diffusing objects : a . The simulated Brownian motion dB \u00f0 t \u00de of objects of radius r is governed by a uniformly dis - tributed random \u00bd 0 ; 1 ) \u2019 \ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03 k B T = 3 ph p number at each time point , where h is the viscosity of the cytoplasm . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 23 of 40 Research article Cell Biology Physics of Living Systems b . Their movement is governed by a mobility m i . For model points of an actin filament , \" \u00bc log \u00f0 L = d \u00de $ \u00f0 3 ph L \u00de for a rod of diameter d , length L and cytoplasmic viscosity h . This mobility term ignores the bending of the filaments . c . The endocytic pit is modeled as a solid , with bulk fluid viscosity associated with pit trans - lational movement and a viscoelastic confinement to the cell surface : dx \u00bc \" F \u00f0 x ; t \u00de dt \u00fe dB \u00f0 t \u00de , where \" \u00bc 6 ph r . 3 . Confinement of objects : a . Objects are confined within a boundary ( cell surface ) according to a harmonic spring potential F \u00bc kx . i . The endocytic pit a distance z from the cell surface experiences a force F \u00bc kz . ii . Actin filaments are confined inside the cell wherein each model point at distance z outside the cell experiences a force F \u00bc kz . 4 . Bending elasticity of filament model points : a . Filament model points are connected via linear elasticity according to a flexural rigidity k , which is the persistence length L p multiplied by k B T . b . The bending elasticity is treated as linear ( see Limitations ) such that for three connected actin model points m 0 ; m 1 ; m 2 the force for those points is F \u00bc k \u00f0 p = L \u00de 3 \u00f0 m 0 \" 2 m 1 \u00fe m 2 \u00de , where k is the flexural rigidity , p is the number of model points , and L is the length of the filament . 5 . Actin - binding proteins a . Hip1R binds actin filaments according to a binding rate and binding radius ( probability of binding when a filament is within the binding radius ) . This general actin - binding protein is a simplification of the multiple interacting proteins that link actin to the coat , including Hip1 and Epsin1 / 2 / 3 ( Brett et al . , 2006 ; Chen et al . , 1998 ; Messa et al . , 2014 ; Senetar et al . , 2004 ; Skruzny et al . , 2012 ) . b . Arp2 / 3 complex was developed as a special - case \u2018fork\u2019 class with two coupled ends . One end binds actin filaments , and the other nucleates a new actin filament , provided the first end bound an actin filament ( this is defined as trans - activation in Cytosim ) . In the \u2018fork\u2019 class , the two ends are constrained geometrically at a resting angle with a given resis - tance to torque ( angular stiffness ) similar to 4b above ( Mund et al . , 2018 ) . Assumptions for modeling mammalian clathrin - mediated endocytosis in cytosim Geometry Endocytic pit : We used our membrane mechanics simulations ( Figure 1 ; Hassinger et al . , 2017 ) to estimate the dimensions of the endocytic pit for physiological values of membrane tension and rigid - ity of the membrane and clathrin coat . Under these conditions the clathrin coat bends the plasma membrane into a U - shaped hemisphere ( Figure 1 ; Boulant et al . , 2011 ; Messa et al . , 2014 ; Yarar et al . , 2005 ) . We initialized the pit as a hemisphere 90 nm in diameter ( Avinoam et al . , 2015 ; Collins et al . , 2011 ) . As the pit internalizes , a smaller neck is exposed ( Figure 1 ) , which is modeled as a sphere with a cylindrical neck of diameter 60 nm . Internalization is defined as a displacement in the - Z direction ( Figure 1A ) . Active Arp2 / 3 complex : We collapsed the activation steps of Arp2 / 3 complex into a single spe - cies , active Arp2 / 3 complex , that resides on the plasma membrane from the beginning of the simula - tion . This models the cellular process , in which soluble Arp2 / 3 complex is inactive until it encounters its activator N - WASP at the plasma membrane . N - WASP binds the plasma membrane via a PI ( 4 , 5 ) P2 - binding site ( which relieves its own autoinhi - bition ) ( Rohatgi et al . , 2000 ) . Additional proteins can bind different regions of N - WASP to increase its level of activation , including the GTPase Cdc42 , actin nucleator cortactin , and BAR protein SNX9 . Because the activation rate and concentrations of these proteins are not yet known , we considered fully active N - WASP ( similar to the VCA region alone ) rather than modeling the individual activation steps . Furthermore , rather than explicitly modeling N - WASP , we used pre - activated Arp2 / 3 com - plex , which models the coincidence of active N - WASP with soluble Arp2 / 3 complex on the plasma membrane . This active Arp2 / 3 complex can template new branched actin filaments when in proxim - ity of an existing \u2018mother\u2019 actin filament . Thus , this model aims to functionally capture Arp2 / 3 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 24 of 40 Research article Cell Biology Physics of Living Systems complex activation and the geometry of branched actin filament nucleation , rather than explicitly modeling each molecule involved in the process of Arp2 / 3 complex activation . N - WASP ( or its homologues Las17 / WASP in yeast ) accumulates earlier in endocytosis ( Taylor et al . , 2011 ) until a \u2018threshold\u2019 concentration triggers actin assembly ( Sun et al . , 2017 ) ; here we initialize all the active Arp2 / 3 complex on the plasma membrane at the beginning of the simula - tion , and it is used over the course of the simulation . Therefore we model the phase in which a threshold value of Arp2 / 3 complex activators has accumulated at the endocytic site and is ready to trigger actin polymerization . We assumed that activated Arp2 / 3 complex resides in a ring around the base of the endocytic pit . This feature has been shown for the budding yeast homologue of N - WASP , Las17 ( Picco et al . , 2015 ) . Endocytic actin polymerizes from the base of the pit in budding yeast ( Idrissi et al . , 2008 ; Kaksonen et al . , 2003 ; Mund et al . , 2018 ; Picco et al . , 2015 ) and in mammalian cells ( Kaplan et al . , in preparation ; Almeida - Souza et al . , 2018 ) , consistent with active Arp2 / 3 complex residing in a ring at the base of the pit . In our fluorescence micrographs , Arp2 / 3 complex is diffraction - limited , so the outer diameter of this ring is * 250 nm . The inner diameter of the ring of Arp2 / 3 complex corre - sponds to the width of the neck of the pit , 60 nm . In budding yeast the Las17 ring outer diameter is ~ 140 nm ( Mund et al . , 2018 ) , which corresponds to a surface density of ~ 3000 molecules / m m 2 . We conservatively set the outer radius of the ring to be 240 nm , which also corresponds to a surface density of ~ 3000 molecules / m m 2 . Estimates of the surface density of in vitro and in vivo patterned activators of Arp2 / 3 complex ( also called nucleation - promoting factors ) range from ~ 3000 \u2013 19000 molecules / m m 2 ( Bieling et al . , 2018 ; Case et al . , 2019 ; Ditlev et al . , 2012 ) . Filament attachments to endocytic pit In mammalian cells , Hip1R and Epsin connect actin filaments to the endocytic pit ( Brett et al . , 2006 ; Chen et al . , 1998 ; Engqvist - Goldstein et al . , 2001 ; Messa et al . , 2014 ; Senetar et al . , 2004 ; Skruzny et al . , 2012 ) . Both are present throughout the clathrin coat ( Clarke and Royle , 2018 ; Sochacki et al . , 2017 ) . We wrote a script in Matlab to uniformly distribute Hip1R molecules around a desired coverage of a sphere , from 1 % to 80 % of a sphere ( Figure 4 ) . In most simulations Hip1R was distributed in 60 % of a sphere . Modeling actin filament dynamics Stall force : Filament polymerization slows under applied load according to the Brownian Ratchet the - ory ( Peskin et al . , 1993 ) . This is treated in Cytosim as growth velocity that slows exponentially under applied load , which is reasonable within the time scales of endocytosis . Modeling filament capping Previous filament - based models of actin in endocytosis modeled actin filaments with uniform lengths or that grow until a maximum length ( Mund et al . , 2018 ) , while others took into account stochastic capping without diffusion or bending ( Wang et al . , 2016 ) . We adapted an existing property in Cyto - sim to model stochastic filament capping , such that the capping events were exponentially distrib - uted . We modeled actin filaments using Cytosim\u2019s \u2018classic\u2019 fiber class , which treats the filament as growing from its plus end , with a stochastic probability of depolymerizing ( corresponding to catas - trophe for microtubules ) . We set the depolymerization rate of shrinking filaments to be 0 with no recovery rate . Thus these filaments become irreversibly capped after an exponential wait time char - acterized by the parameter catastrophe _ rate ( which we define as the capping rate ) . Because the probability that capping protein binds to the barbed end of the filament is exponentially distributed , filament lengths are exponentially distributed ( Figure 3\u2014figure supplement 1F ) . We set the rate of capping to achieve a desired mean filament length based on the expected distributions of actin fila - ment lengths for a given capping rate . In Cytosim , the catastrophe rate can be set to depend on a combination of applied load and growth velocity , which we did not include in our model of actin fila - ment capping . Source of actin mother filaments Active Arp2 / 3 complex requires a mother filament from which to nucleate a new actin filament at a defined angle ( Amann and Pollard , 2001 ; Mullins et al . , 1997 ) . Therefore , the polarity of the Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 25 of 40 Research article Cell Biology Physics of Living Systems mother filament defines the polarity of the resultant branched actin network . Our study uses diffus - ing linear actin filament nucleating proteins ( Balzer et al . , 2018 ; Basu and Chang , 2011 ; Wagner et al . , 2013 ) to seed a defined number of randomly oriented mother filaments near the endocytic site . Alternatively , simulations using a pool of cytoplasmic linear actin filaments ( Raz - Ben Aroush et al . , 2017 ) allowed for similar internalization , but with less reliable timing of initiation . More detailed studies of the mechanism of actin nucleation and mother filament generation are necessary . Limitations Endocytic pit internalization is simplified as an elastic spring , with a linear relationship between force and extension . We show in Figure 1 that this linear relationship is characteristic of coated plasma membranes under force up until a threshold internalization of ~ 100 nm . Future studies will treat the coated membrane as a 3D , force - dependent curving surface ; such an approach is outside the scope of the present work . We focus our model on the minimal actin machinery required to produce force . We have not included crosslinkers ( Ma and Berro , 2018 ) or myosin I , both of which are expected to increase the network\u2019s ability to produce force . The effects of these two proteins on mammalian endocytosis will be treated in a future study . The treatment of filament bending elasticity as linear is valid for small deflections of individual actin model points . Importantly , the outcomes of our simulations did not depend on the frequency of segmentation of actin model points ( which change the magnitude of deflection between individ - ual actin model points ) . Filament twist or twist - bend coupling , which increases the total energy stored in bent actin filaments ( De La Cruz et al . , 2015b ) , is not considered in Cytosim , and requires a more detailed modeling approach considering each subunit . Cytosim does not implement hydro - dynamics for curved filaments , so the diffusion of these filaments is approximated as the motion for a linear filament . Arp2 / 3 complex preferentially binds to the curved sides of actin filaments ( Risca et al . , 2012 ) . We do not include this assumption in our model . We expect that the self - organization and robustness exhibited by our minimal actin network would be enhanced by this assumption , given that Arp2 / 3 complex at the base of the pit encounters many curved sides of actin filaments . Parameter values We derived most parameters from experimental data in the literature , and made measurements for some measurements not available ( Supplementary file 3 ) . We varied the remaining parameters to show their effect on the outcome of the simulations . Discussion of each parameter follows below . Membrane tension We used the relationship between internalization resistance and membrane tension ( Figure 1 ) to cal - ibrate the spring stiffness in our agent - based simulations . We relied on values of membrane tension measured in human skin melanoma SK - MEL - 2 cells , based on atomic force microscope membrane tether rupture forces and the assumption that the rigidity of the plasma membrane is 320 pN ! nm ( Dimova , 2014 ; Diz - Mun\u02dcoz et al . , 2016 ; Kaplan et al . , in preparation ) . Association rate constants The biochemical association rate constant , k on , is given in units of m M \" 1 s \" 1 . In Cytosim , we input the association probability between actin and a binding protein as a binding _ rate , in units of s \" 1 . These two rates can be related by the following relationship : k on \u00bc binding rate \u2019 capture volume ; which is defined as p \u2019 capture radius 2 \u2019 filament length ( Francois Nedelec , personal communication ) . This gives an order - of - magnitude scaling relationship to convert between k on and binding rate , considering that cytosim does not treat explicit binding sites on the filament ( Francois Nedelec , personal communication ) . Arp2 / 3 complex Branching angle : Based on Blanchoin et al . ( 2000 ) we set the branching angle of Arp2 / 3 complex to be 77 \u00b1 13\u02da , as measured for bovine Arp2 / 3 complex . Acanthomeba Arp2 / 3 complex adopts closer Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 26 of 40 Research article Cell Biology Physics of Living Systems to a 70\u02da branching angle . In NIH3T3 cells preserved by cryo - fixation , branch angles in the lamellipo - dium are 77 \u00b1 8\u02da ( Vinzenz et al . , 2012 ) . Importantly , Blanchoin et al . ( 2000 ) was the only in vitro study we are aware of that measured the variance of branching angles , which was converted to an angular stiffness ( 0 . 076 pN / rad 2 ) . Therefore we set the resting angle of Arp2 / 3 branches as 1 . 344 rad with an angular stiffness of 0 . 076 pN / rad 2 . Unbinding rate We used the k off measured by Beltzner and Pollard ( 2008 ) : 0 . 003s - 1 . Nucleation rate Based on the association rate constant between activated Arp2 / 3 complex and actin filaments of 0 . 00015 m M \" 1 s \" 1 ( Beltzner and Pollard , 2008 ) , we set the binding _ rate of Arp2 / 3 to actin to be 7 s \" 1 , given a capture radius of 10 nm and filament length of 100 nm ( see relationship between these parameters above ) . We note that this calculation of binding rate depends inversely with the filament length ( and number of filaments ) which are not directly comparable in our simulations given that the filaments and Arp2 / 3 are generally not freely diffusing . Still , it was remarkable that our best estimate for binding rate gave reasonable nucleation kinetics , and served as a threshold for timely internaliza - tion of the endocytic pit , whereas previous deterministic models needed to increase the association rate constant by 300 - 600x for efficient nucleation ( Beltzner and Pollard , 2008 ; Berro et al . , 2010 ) . These ODE models did not have spatial considerations , so this suggests that the spatial and tempo - ral confinement of actin filaments and the high local concentration of active Arp2 / 3 complex in our simulations accounted for most of this difference . Thus the local geometry has a significant ( > 2 orders of magnitude ) effect on the effective nucleation rate . Actin Growth rate : In cells the cytoplasmic concentration of actin is 60 \u2013 100 m M ( Haugwitz et al . , 1994 ; Wu and Pollard , 2005 ) . In mammalian cells , a subset of this actin is available as polymerizable actin , both due to monomer - sequestering proteins ( thymosin B4 ) and due to only a subset of monomers being ATP - bound . We conservatively set the concentration of available polymerizable actin to be 20 m M , which given the association rate constant of ATP - actin of 11 . 6 subunits / m M / s ( Pollard , 1986 ) corresponds to a polymerization rate of 500 nm ( 182 subunits ) per second . Capping rate : The mean length of filaments in mammalian endocytosis has not been measured . We relied on the estimates from Berro et al . ( 2010 ) ; Sirotkin et al . ( 2010 ) which showed that for fis - sion yeast filaments were an average of 150 nm in length . We set the capping rate to be 6 . 3 / s , which set the mean filament length at 150 nm . We varied the rate of capping in our simulations . Less actin capping resulted in greater internalization , due to more actin ( Figure 1\u2014figure supplement 1 ) . However , the resultant amount of actin is larger than the amount of actin measured in CME in other organisms ( Picco et al . , 2015 ; Sirotkin et al . , 2010 ) . Stall force : The stall force scales with the load applied and the concentration of actin monomers available ( Peskin et al . , 1993 ) . At 4 m M actin the filaments\u2019 stall force was measured to be 1 \u2013 2 pN ( Footer et al . , 2007 ) . With 40 m M actin the filaments could theoretically stall at up to 9 pN force per filament ( Dmitrieff and Ne\u00b4de\u00b4lec , 2016 ) . For the ~ 20 m M actin that we assumed was available for polymerization , the stall force was ~ 5 pN . Surprisingly , the extent of internalization varied weakly with stall force ( Figure 1\u2014figure supplement 1 ) , suggesting that actin used another mode of force generation than elongation directly against the membrane ( Figure 5 ) . Persistence length : We set the persistence length of actin filaments to be 10 m m , which corre - sponds to a flexural rigidity of 0 . 041 pN ! m m 2 ( McCullough et al . , 2008 ) . Previous modeling studies used a value of 20 m m , based on measurements of actin filaments labeled with phalloidin ( Gittes et al . , 1993 ) , which stiffens actin filaments ( Isambert et al . , 1995 ; Pfaendtner et al . , 2010 ) . Changing the persistence length of actin between 1 and 20 m m had a minor effect on pit internaliza - tion ( Figure 1\u2014figure supplement 1 ) . Hip1R In mammalian endocytosis , several proteins link actin filaments to the clathrin coat via phosphoinosi - tide - and clathrin - binding domains and actin - binding domains , including Hip1 , Hip1R , and Epsin Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 27 of 40 Research article Cell Biology Physics of Living Systems ( Brett et al . , 2006 ; Chen et al . , 1998 ; Messa et al . , 2014 ; Senetar et al . , 2004 ; Skruzny et al . , 2012 ) . Our general linker protein , named in the text as Hip1R , is a surrogate for all proteins that link actin to the coat . Endocytic actin - binding proteins Hip1 and Hip1R use a conserved domain to bind actin with sur - prisingly weak affinity . This domain , which is alternately named the THATCH ( Talin - Hip1 / R / Sla2p Actin - Tethering C - terminal Homology ) , Talin - family , or I / LWEQ domain , has been isolated and stud - ied by several groups . We fit the binding results of previous Hip1R binding experiments in our lab ( Engqvist - Goldstein et al . , 1999 ) to estimate a binding affinity between Hip1R and actin as ~ 400 nM , and the affinity between clathrin cages and Hip1R to be ~ 1 nM . Both sets of data were fit well by the quadratic binding curve ( Pollard , 2010 ) , \u00bd LR ) = \u00bd L ) \u00bc \u00f0\u00f0\u00bd R ) \u00fe\u00bd L ) \u00fe K d \u00de \" \u00f0\u00f0\u00bd R ) \u00fe\u00bd L ) \u00fe K d \u00de 2 \" 4 \u2019\u00bd R ) \u2019\u00bd L ) \u00de 1 = 2 = \u00f0 2 \u2019\u00bd L ) \u00de where \u00bd L ) is the concentration of actin or clathrin and \u00bd R ) is the concentration of Hip1R , with r 2 = 0 . 94 and 0 . 99 , respectively ( data not shown ) . Other studies measured a weaker affinity between Hip1R and actin : K d = 3 . 4 m M , or 2 . 5 m M for Hip1 ( Senetar et al . , 2004 ) . In the presence of the first helix of the five - helix bundle comprising the THATCH domain , actin affinity further decreases ( Senetar et al . , 2004 ) . Epsin has two actin - binding domains with very weak ( K d > 10 m M ) or unknown affinity to actin ( Messa et al . , 2014 ; Skruzny et al . , 2012 ) . For our linker protein we used a combina - tion of rate constants such that k off / k on ~ K d of 400 nM . Compared with dilute reactions , in an endo - cytic geometry actin filaments grow near the coat , so the actin filaments bind Hip1R more frequently . As a result we found that a relatively low binding rate was sufficient for binding between actin and Hip1R in our simulations . We varied the off rate of Hip1R and found that that , surprisingly , the internalization was robust to Hip1R off rate between 0 . 001 and 10 s \" 1 ( Figure 1\u2014figure supple - ment 1 ) . With such weak affinity , + 100 linking molecules are required for robust endocytosis ( Figure 4 ) . The following considerations support the likelihood that a sufficient number of actin - linking proteins reside in the clathrin - coated pit . In yeast , endocytic sites accumulate ~ 100 \u2013 150 molecules of the Hip1R homologue End4p and 230 molecules of the actin - binding protein Pan1 ( with K d to actin = 2 . 9 m M ) . Estimation from platinum replica images of clathrin - coated pits in SK - MEL - 2 cells ( Sochacki et al . , 2017 ) suggest that clathrin cages have approximately 55 \u00b1 12 \u2018faces\u2019 ( pentagons and hexagons ) , or up to 90 faces in HeLa cells . If the cage accumulates one Hip1R dimer per face , this would lead to 110 \u2013 180 molecules of Hip1R , plus molecules of Hip1 , Epsin1 , Epsin2 , and Epsin3 . From a similar analysis , SK - MEL - 2 cells have ~ 66 \u00b1 12 vertices ( triskelia ) , which corresponds to ~ 200 \u00b1 40 clathrin heavy chains and ~ 200 \u00b1 40 light chains . Since Hip1R binds clathrin light chain , a 1 : 1 ratio of these proteins would again suggest ~ 200 Hip1R molecules in the clathrin coat . Addi - tionally , actin binding by these linker proteins is likely highly multivalent . Mammalian Epsin proteins hexamerize in vitro via their membrane - binding ENTH domains , and Hip1 and Hip1R can dimerize with each Epsin through its ANTH domain by sharing a Pi ( 4 , 5 ) P2 molecule ( Garcia - Alai et al . , 2018 ; Skruzny et al . , 2015 ) . Adding an additional layer of multivalency , Hip1R and Hip1 hetero - and homodimerize via coiled - coil domains and a C - terminal dimerization motif ( Brett et al . , 2006 ; Chen and Brodsky , 2005 ; Engqvist - Goldstein et al . , 2001 ; Niu and Ybe , 2008 ; Senetar et al . , 2004 ) . Therefore , it is quite likely that a sufficient number of actin - linking proteins cover the coat . Simulation environment parameters The internalization of endocytic pits was not sensitive to other simulation parameters including the segmentation of actin ( 1 and 100 nm per model point ; 1 m m model points introduced additional vari - ability ) , confinement force for actin within the cell , and time step of the simulation ( Figure 1\u2014figure supplement 1 ) . The viscosity of the cytoplasm and the endocytic pit weakly affected extent of inter - nalization ( Figure 1\u2014figure supplement 1 ) . Modifications to source code We added a method \u2018confine _ first _ surface\u2019 in which only the first segment of the multi - point bead is under confinement , and that the bead does not undergo angular displacements . We added data reporting methods , including ( 1 ) reporting the Arp2 / 3 complex branch angle ( Francois Nedelec ) ; ( 2 ) reporting the Hip1R - attached filament ID numbers ; ( 3 ) visualizing the axial Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 28 of 40 Research article Cell Biology Physics of Living Systems orientation of actin segments with respect to the plasma membrane ( implemented as 75 % of the hsv colormap ) . Comparison to theory Our calculation of the elastic energy stored in bent actin filaments is derived from the theory of deforming elastic beams ( Boal ) . Specifically , the bending energy E is determined by \u00f0 k B TL p ! 2 \u00de = \u00f0 2 l \u00de , where k B is the Boltzmann constant , T is temperature , L p is the filament persistence length ( 10 m m for actin ) , ! is the bending angle , and l the free filament length ( contour length of the filament between the attachment site and barbed end of the filament ) ( Boal , equations 3 . 15 and 3 . 21 ) . In Figure 5 we reference ( Sept and McCammon , 2001 ) and ( De La Cruz et al . , 2015b ) to esti - mate the bending energy associated with fragmentation of bare actin filaments . Based on the energy associated with removal of two longitudinal contacts and one lateral contact between mono - mers in an actin filament ( Sept and McCammon , 2001 ) , the elastic energy associated with fragment - ing an actin filament is 26 \u2013 28 k B T ( De La Cruz et al . , 2015b ) . The rate constants in De La Cruz and Gardel ( 2015a ) associated with the probability of severing are not considered in this study . Running simulations We wrote custom scripts in bash to run parallel simulations on a high - performance computing server . Analysis of simulations We wrote custom code in Python ( 3 . 7 ) with Jupyter Notebook ( Project Jupyter ) to read , analyze , and plot the simulations obtained from Cytosim . X , Y = 0 is defined as the center of the pit . Energy associated with polymerization was defined as 5 pN * 2 . 75 nm = 13 . 5 pN ! nm per binding event . This code is available at the following website : https : / / github . com / DrubinBarnes / Akamatsu _ CME _ manuscript ( Akamatsu , 2019 ; copy archived at https : / / github . com / elifesciences - publications / Aka - matsu _ CME _ manuscript ) . Internalization energy We used the relationship E \u00bc 1 = 2 kx 2 to estimate the expected internalization for a non - adapting machine . From the internalization for k = 0 . 01 pN / nm ( low tension ) , we calculated the expected internalization using the same energy for different values of k . Specifically , for a single value of load , we calculated the work output based on energy ( work output ) E \u00bc 1 = 2 kx 2 , where k is the load ( in pN / nm ) and x is the internalization ( in nm ) . For an adapting network , the work output changes with load . For a non - adapting network , it does not , so as load k increases , internalization x decreases . This relationship of x as a function of k for constant E is plotted as the dotted line in Figure 7E . Energy efficiency is defined as the energy ( of internalization or stored in bent filaments ) divided by the total polymerization energy . Radial orientation metric : We defined the radial orientation as the sum of the dot products of the direction of the barbed end with the unit vector in X and Y , such that \" 1 = barbed ends oriented toward the origin ( the center of the pit ) , 0 = oriented tangent to the pit , and + 1 = oriented away from the center of the pit . 95 % internalization is defined as the 95th percentile of internalization . We use the first 12 s of sim - ulations to allow for comparison with simulations that terminated after t = 12 s . For Figure 4\u2014figure supplement 2 , we compared 95 % internalization in two conditions using the Welch\u2019s t - test in Python ( 3 . 7 ) with scipy . stats . ttest _ ind ( equal _ var = False ) . Barbed ends near base / neck : We chose a distance of 7 . 5 nm as a metric for proximity to the membrane ( base or neck ) as this was the width of 1 \u2013 2 actin monomers and less than the radial dis - tance between the neck surface and the pit diameter . The absolute value of this metric did not change the results appreciably . Experimental method details Cell line maintenance WTC - 10 human induced pluripotent stem cells ( hiPSCs ) were obtained from the lab of Bruce Conklin and genome edited using TALENs to endogenously express AP2 - tagRFP - T at one allele at an Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 29 of 40 Research article Cell Biology Physics of Living Systems internal loop of the m 2 subunit ( Hong et al . , 2015 ) . We grew these cells on matrigel ( hESC - Qualified Matrix , Corning ) ( 80 m g / mL , 1 mL / well ) in StemFlex ( Thermo Fisher ) with Penicillin / Streptomycin ( Thermo Fisher ) , and passaged with Gentle Cell Dissociation reagent ( EDTA - based clump passaging ; StemCell Technologies ) . Parental and genome - edited cells were tested for mycoplasma and authen - ticated by STR profiling . For single - cell applications ( genome editing , flow cytometry , transfections ) we trypsinized the cells with the recombinant trypsin TrypLE Select ( Thermo Fisher ) and grew the cells in StemFlex supplemented with the specific rho kinase inhibitor RevitaCell ( Thermo Fisher ) . SK - MEL - 2 cell culture SK - MEL - 2 cells endogenously expressing clathrin light chain CLTA - RFP and dynamin2 - eGFP ( Doyon et al . , 2011 ) were cultured in DMEM / F12 ( Thermo Fisher ) supplemented with 10 % FBS ( HyClone ) and Penicillin / Streptomycin ( Thermo Fisher ) . A day before imaging , the cells were seeded on 25 mm diameter glass coverslips ( Fisher Scientific ) . Cell line construction We followed ( Dambournet et al . , 2014 ) to create lines human induced pluripotent ( hiPS ) cells with genetically encoded fluorescent fusion proteins . To the AP2 - RFP cells described above , we used the following Cas9 ribonuclear protein electroporation method for genome editing . Gibson assembly of DNA G - blocks ( IDT ) was used to construct a donor plasmid containing the tagGFP2 gene , codon optimized for mammalian cell expression , between 500 base pair homology arms flanking the 3\u2019 ter - minus of the ArpC3 gene . S . pyogenes NLS - Cas9 was purified in the University of California Berkeley QB3 MacroLab and frozen at \" 80 \u02daC until use . TracrRNA and crRNAs were purchased from IDT . Cells were trypsinized in TrypLE select ( Thermo Fisher ) , mixed with donor plasmid ( final concentra - tion 3 . 8 m M ) and 240 pmol Cas9 ( final concentration 1 . 75 m M ) complexed with 100 m M crRNA and tracrRNA ( final concentration 3 . 7 m M ) , and electroporated with an Amaxa nucleofector in stem cell nucleofector reagent ( Lonza ) . The sequence of the crRNA to target the C terminus of the ArpC3 gene was CCGGGCUCCCUUCACUGUCC . Cells were seeded on matrigel - coated 6 - well plates in StemFlex supplemented with the rho - kinase inhibitor RevitaCell and media was changed 24 hr later . Three days after electroporation the cells were bulk sorted for GFP fluorescence with a BD Biosci - ence Influx sorter ( BD Bioscience ) . Fluorescence intensities of cell populations were analyzed using the flow cytometry software FlowJo ( FlowJo , LLC ) . Around one week later the cells were re - sorted into matrigel - coated 96 - well plates at densities of 5 , 10 , or 20 cells per well . Positive clones were confirmed by PCR and sequencing of the genomic DNA locus . From genomic DNA and fluorescent cell sorting analysis we determined that both alleles of ArpC3 were tagged with tagGFP2 . We sequenced genomic DNA extracts from the cell lines at the insertion sites to confirm that there were no insertions or deletions at the site of GFP insertion . We isolated genomic DNA using the DNEasy blood and tissue DNA isolation kit ( Qiagen ) . The primers used for PCR amplification of genomic DNA were TCAGGGTGGCTTTCTCTCCT and CCAGAGCTGCAACCAGTACA . The primers used for sequencing the ArpC3 allele were ACTTATTCTTATTAAGCGCCAGC and CAGGGCTCTGGA - GACGGT . Western blotting We pelleted 1 \u2013 2 wells of cells from a 6 - well plate ( ~ 10 6 cells ) at 4 \u02daC and lysed the cells in 50 mM Hepes pH 7 . 4 , 150 mM NaCl , 1 mM MgCl 2 , 0 . 1 % Triton X - 100 , and cOmplete Mini EDTA - free prote - ase inhibitor ( Sigma Aldrich ) . Extracts were separated electrophoretically in sample buffer with 80 mM DTT on 8 % polyacrylamide gels and transferred to nitrocellulose membranes . Membranes were blocked with 5 % nonfat milk in PBS , probed with mouse monoclonal anti - GAPDH ( ProteinTech 10494 \u2013 1 - AP ) at 1 : 5000 dilution and rabbit polyclonal antibody against tag ( C , G , Y ) FP ( Evrogen AB121 Lot 12101231265 ) at 1 : 2500 dilution in PBS with 0 . 1 % TWEEN - 20 and 1 : 100 PBS / milk , followed by incubation in donkey anti - mouse CF680 and donkey anti - rabbit CF800 ( Li - Cor Bioscences ) . Washed membranes were imaged on a Li - Cor Odyssey Clx infrared fluorescence imager ( Li - Cor Biosciences ) . Choice of subunit to tag : We chose the location of the GFP tag based on available structural , biochemical , cell biological , and genetic data on the functionality of fluorescent fusion proteins of subunits of the Arp2 / 3 complex . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 30 of 40 Research article Cell Biology Physics of Living Systems ArpC3 ( p21 ; Arc18 in budding yeast ) tags are more functional than tags on Arp2 , Arp3 , or ArpC5 ( Arc15 in budding yeast ) ( Egile et al . , 2005 ; Sirotkin et al . , 2010 ; Smith et al . , 2013 ; Picco et al . , 2015 ) . Single - particle electron microscopy reconstructions and the crystal structure ( PDB : 2p9l ) of Arp2 / 3 complex show that the C terminus of ArpC3 is flexible and does not sterically interfere with Arp2 / 3 complex\u2019s binding site to actin filaments or its activators ( VCA ) . Constructs for intracellular fluorescence - based standard curve We adapted self - assembling protein nanocages of defined copy number ( Hsia et al . , 2016 ) to con - struct a fluorescence calibration standard in live mammalian cells . These trimeric proteins were engi - neered by Hsia et al . ( 2016 ) at protein - protein interfaces to self - assemble into a 60mer ( KPDG aldolase ) that can be tagged at the N or and C termini with GFP to yield an average of 60 or 120 copies of GFP . Alternatively a two - component 24mer ( alamin adenosyl transferase and 5 - carboxy - methyl - 2 - hydroxymuconate isomerase [ King et al . , 2014 ] ) can be tagged one or both components with GFP to yield an average 12 or 24 copies of GFP per structure ( Hsia et al . , 2016 ) . DNA con - structs codon - optimized for mammalian expression were synthesized ( IDT ) with alanine mutations at K129 ( KPDG aldolase ) and R72 ( 5 - carboxymethyl - 2 - hydroxymuconate isomerase ) to abolish enzy - matic activity . An E126L mutation in the transferase ( King et al . , 2014 ) is predicted to abolish its enzymatic activity . The synthetic construct included GS repeat linkers and tagGFP2 codon optimized for mammalian cell expression . We inducibly tethered the nanocages to the plasma membrane using an N - terminal myristolation and palmitoylation motif and the FKBP / FRB * dimerization system , where FRB * is a T2098L variant of FRB that binds a rapamcyin analog , AP21967 ( Clontech ) , and does not bind endogenous mTOR . These constructs bound weakly to the plasma membrane even in the absence of AP21967 , presumably due to multivalent weak affinity binding . We transiently expressed these plasmid constructs into hiPS ( or SK - MEL - 2 ) cells with Lipofectamine Stem ( Thermo Fisher ) . After 2 days we imaged the cells along with the genome edited cells using similar imaging settings . TIRF microscopy Cells were imaged on either an Olympus IX - 81 or Nikon Ti - 2 inverted microscope fitted with TIRF optics . The IX - 81 microscope used a 60 , 1 . 49 NA objective ( Olympus ) and an Orca Flash 4 . 0 sCMOS camera ( Hamamatsu ) . Cells were illuminated with solid - state lasers ( Melles Griot ) with simul - taneous acquisition by a DV - 2 image splitter ( MAG Biosystems ) . The microscope was maintained at 37 \u02daC with a WeatherStation chamber and temperature controller ( Precision Control ) and images were acquired using Metamorph software . The Nikon Ti2 microscope was equipped with a motor - ized stage ( Nikon ) , automated Z focus control , LU - N4 integrated four - wavelength solid state laser setup , TIRF illuminator ( Nikon ) , quad wavelength filter cube , NI - DAQ triggering acquisition ( National Instruments ) , an Orca Flash 4 . 0 sCMOS camera ( Hamamatsu ) , and triggerable filter wheel ( Finger Lakes Intstrumentation ) with 525 / 50 and 600 / 50 wavelength emission filters . Cells were seeded on autoclaved 25 mm # 1 . 5 round coverslips coated with 1 mL matrigel ( 80 m g / mL ) or recombinant Vitronectin - N diluted in PBS ( Thermo Fisher ) . Cells were maintained at 37 \u02daC with a stage top incuba - tor ( OKO Lab ) and images were acquired with Nikon Elements . CK - 666 experiments CK - 666 ( Sigma ) was reconstituted in DMSO and diluted in imaging media prior to treatment . Cells were treated prior to or during imaging . Cells were treated for 45 min prior to imaging unless other - wise indicated in the text . \u20180 m M\u2019 treatment corresponds to 0 . 1 % DMSO treatment . Confocal microscopy We imaged cells on a Nikon Eclipse Ti inverted microscope ( Nikon Instruments ) fitted with a CSU - X spinning disk confocal head ( Yokogawa ) , four solid - state lasers ( Nikon ) , IXon X3 EMCCD camera ( Andor ) , and emission filter wheel ( Sutter Instruments ) . The imaging area was kept at 37 \u02daC with 5 % CO2 ( In Vivo Scientific , LCI ) . We used a 100 , 1 . 45 NA Plan Apo oil immersion objective ( Nikon ) . Images were acquired using Nikon Elements . We generally imaged 3 \u2013 7 z slices with 300 nm z spac - ing at 3 s time intervals . Cells were seeded on sterile 4 - chambered or 8 - chambered # 1 . 5H ( 0 . 170 \u00b1 0 . 005 mm ) cover glasses ( CellVis ) . We imaged the cells in media supplemented with HEPES and the antioxidant oxyrase ( OxyFluor ) with substrate lactate . For quantitative fluorescence Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 31 of 40 Research article Cell Biology Physics of Living Systems experiments we limited the cells\u2019 exposure to ambient light , brightfield light and 488 nm laser light prior to acquisition , and used the same laser power during acquisition to compare experiments ( gen - erally 10 % acousto - optic tunable filter ( AOTF ) power ) . For most experiments , we used DMEM / F12 without phenol red ( Thermo Fisher ) supplemented with the protein supplement used in StemFlex media ( Thermo Fisher ) , which gave similar fluorescence intensity results as cells imaged in StemFlex ( which has phenol red ) . Time - lapse imaging of nanocages We treated cells with a range of concentrations of AP21967 for > 45 min and then imaged them on a spinning disk confocal microscope at 0 . 3 s intervals for 1 min . Image correction We took images of dilute GFP or autofluorescent Luria Broth ( LB ) to correct for uneven illumination in the imaging plane . The exposure time and laser power both scaled linearly on our instrument ( Fig - ure 2\u2014figure supplement 1 ) which allowed us to adjust for dimmer or brighter signals by changing the exposure time . Cryo - electron tomography sample preparation Holey carbon grids ( Quantifoil R2 / 1 , 200 mesh , gold ) were glow discharged using a Pelco SC - 6 sput - ter coater and sterilized in 70 % ethanol in H 2 O . SK - MEL - 2 cells were plated in DMEM / F12 ( Gibco ) supplemented with 10 % fetal bovine serum ( premium grade , VWR Seradigm Life Science ) and 1 % Penicillin / Streptomycin ( Gibco ) onto grids and incubated overnight at 37\u02daC and 5 % CO 2 in a cell cul - ture incubator . Samples were blotted and plunge frozen using a Vitrobot Mark IV ( FEI ) after addition of 10 nm BSA Gold Tracer ( Electron Microscopy Sciences ) . Cryo - electron tomography data recording and processing Tilt - series were recorded on a Titan Krios operated at 300 kV ( FEI ) equipped with a Quantum energy filter ( Gatan ) and a K2 direct electron detecting device ( Gatan ) at 2 . 97 A\u02da pixel size and a target defo - cus of \" 2 m m . SerialEM ( Mastronarde , 2005 ) in low - dose mode was used for automated tilt series acquisition using a bidirectional tilt scheme covering a whole tilt range from + 60\u02da to \" 60\u02da with a base increment of 2\u02da and a total electron dose of 100e - / A\u02da 2 . Tomograms were generated in IMOD ( Kremer et al . , 1996 ) using the gold beads as fiducials for tilt - series alignment . For Figure 5 , tomo - grams were reconstructed using the simultaneous iterative reconstruction technique ( SIRT ) recon - struction algorithm , then filtered using the Nonlinear Anisotropic Diffusion ( NAD ) filter in IMOD and binned by a factor of 2 . For Figure 5\u2014video 1 , we used the backprojection algorithm for tomogram reconstruction followed by a smoothing filter ( Clip smooth function in IMOD ) . The U - shaped pit shown is one of six sites of CME we have identified in our tomograms , all of which have bent actin filaments around the pit to varying degrees . The bent filaments are especially prominent in the tomogram shown because of the orientation of the pit with respect to the missing wedge effect . A detailed treatment of these tomograms will be the subject of a subsequent study . Data analysis Calibration curve : We used a combination of custom - written and publicly available image analysis software in Fiji ( 1 . 52i ) and Matlab ( r2017b ) to analyze the traces of fluorescence intensity per spot , for multiple Z slices and time points . To measure fluorescence intensity per spot of GFP - tagged nanocages , we wrote a toolset in Fiji to select and circular regions of interest eight pixels ( 1 . 1 m m ) in diameter , and used cross - correlation to center the regions around the intensity - based center of mass . We selected as background regions of concentric circles one pixel larger than the original region of interest . The toolset measured the fluorescence intensity per spot , which we subtracted by the area - corrected background intensity to yield fluorescence intensity per spot for the four con - structs . Some analysis functions were adapted from published software ( Akamatsu et al . , 2017 ; Epstein et al . , 2018 ; McCormick et al . , 2013 ) . We measured only spots that were contained within the slices imaged and were single stationary spots . For comparison of fluorescence to eGFP - MotB , we used smaller ( 6 - pixel ) ROIs . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 32 of 40 Research article Cell Biology Physics of Living Systems We plotted background - subtracted fluorescence intensity per spot as a function of predicted copy number per structure to obtain a calibration curve relating fluorescence intensity per spot to numbers of molecules per structure . For the curve in Figure 3D , we combined data from three experiments with different imaging conditions by defining the average 60mer - GFP intensity per experiment as 1000 arbitrary units . Lines are linear fits through zero with r 2 calculated by linear least - squares fitting . Time - lapse fluorescence quantification : We made modifications to automated MATLAB - based tracking software ( Aguet et al . , 2013 ; Hong et al . , 2015 ) to track and analyze fluorescence - intensity time lapse data of genome - edited cells . The core tracking program ( based on the software package m - track ) automatically identifies fluorescent spots and connects them as tracks by minimizing the lin - ear assignment problem ( Jaqaman et al . , 2008 ) . We used stringent tracking parameters with gap size 0 and search radius 0 \u2013 2 . 3 pixels ( 248 nm ) . GFP and RFP tracks with high variability in the inten - sity / time profile were automatically rejected ( Ferguson et al . , 2017 ) as well as tracks * 3 s in dura - tion ( Dambournet et al . , 2018 ) and the remaining tracks were associated spatiotemporally according to a cost matrix ( Hong et al . , 2015 ) . We used two track rejection schemes . In the first , users were presented with fluorescence montages and XY coordinates of the tracks to assess the fidelity of tracking for each event ( Hong et al . , 2015 ) . In the second , tracks were automatically rejected based signal - to - noise ratio ( > 1 . 05 ) and proximity to neighboring tracks ( > 525 nm ) ( Hong et al . , 2015 ) . We checked that the manual and automatic track rejection schemes yielded similar results ( lifetime distributions and intensity versus time plots ) as well as to manual , kymograph - based quantification of lifetimes ( below ) . From the above workflow ( Dambournet et al . , 2018 ) we increased throughput by connecting all steps into an automated tracking pipeline requiring minimal user input . For SK - MEL - 2 cells expressing CLTA - RFP and DNM2 - GFP , we tracked regions that were not near the nucleus ( which has a concentration of Golgi - derived clathrin budding ) and that did not have large , bright , persistent structures containing invariant RFP and GFP signals ( \u2018plaques , \u2019 which are likely sites of adhesion ) . Alignment method : For Figure 2G , we aligned tracks based on the time point after the peak intensity in which 50 % of the fluorescence remained in the GFP channel . We normalized the fluores - cence intensity to compare movies from different imaging conditions . For Figure 2I , we aligned the tracks based on the maximum intensity . For Figure 6D , we aligned the tracks based on the disap - pearance of the RFP signal . This code is available at the following website : https : / / github . com / Dru - binBarnes / Akamatsu _ CME _ manuscript . Manual track analysis : We wrote a Fiji toolset that generated two - color kymographs from user - defined regions of interest , and then quantified the lifetime based on the lengths of the kymographs ( from user - defined regions on the kymographs ) . This manual analysis was used in Figure 6\u2014figure supplement 1A and for verification of the automated tracking scheme . This code is available at the following website : https : / / github . com / DrubinBarnes / Akamatsu _ CME _ manuscript . Nanocage particle tracking : To track the membrane - tethered nanocages in 2D in cells we used TrackMate , a plugin available in Fiji that optimizes the Linear Assignment Problem ( LAP ) ( Jaqaman et al . , 2008 ) . We detected spots using an estimate of 0 . 5 m m and threshold of 15 , using a median filter and sub - pixel localization . We used the simple LAP tracker with a maximum linking and gap closing distance of 1 m m and two frames and a minimum track length of 4 frames . Calculating numbers of molecules per endocytic site : We calculated the fluorescence intensity of background - subtracted ArpC3 - GFP spots colocalized with AP2 - RFP spots from single time - point images using the same background correction approach described for the calibration curve . We used the slope of the calibration curve to convert fluorescence intensity to numbers of molecules of ArpC3 - GFP . Because the standard is inside cells , this standard controls for fluorescence environment and fluorescent protein folding and maturation . We used the resultant histogram of numbers of ArpC3 - GFP per spot and the time - lapse fluorescence intensity data in Figure 2G to create the graph in Figure 2I of numbers of molecules of ArpC3 - GFP over time . Acknowledgements We would like to thank Dan Fletcher , Johannes Schoeneberg , and Jasmine Nirody for insightful com - ments on the manuscript ; Julian Hassinger for advice on the continuum mechanics model and gener - ating movies of the membrane simulations ; Francois Nedelec for training and advice in Cytosim , Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 33 of 40 Research article Cell Biology Physics of Living Systems discussion of the relationship between association rate constants and binding rates , and sharing the \u2018Fork . cc\u2019 class for modeling Arp2 / 3 complex ; Johannes Schoeneberg and Logan Akamatsu for help generating parallel simulation shell scripts ; Sun Hong and Meiyan Jin for generating and validating the AP2 - tagRFP - T human induced pluripotent cell line ; and Elizabeth Li for contributions to automat - ing the particle tracking pipeline . In addition , we would like to thank the UC Berkeley High Perfor - mance Computing cluster for training and server space for parallel simulations ; the UC Berkeley QB3 MacroLab for purified S . pyogenes NLS - Cas9 ; the UC Berkeley Cancer Research Laboratory Molecu - lar Imaging Center with support from the Gordon and Betty Moore Foundation ; Karen M Davies and Jonathan Remis , Lawrence Berkeley National Labs Donner Building Cryo EM facility ; Daniel B Toso and Paul Tobias , UC Berkeley Berkeley Bay Area Cryo - EM facility ; NIH MIRA R35GM118149 to DGD ; a postdoctoral fellowship from the Arnold and Mabel Beckman Foundation to MA ; postdoc - toral fellowship LT000234 / 2018L from the Human Frontier Science Program to DS ; ARO W911NF1610411 and Office of naval research N00014 - 17 - 1 - 2628 to PR . Additional information Funding Funder Grant reference number Author National Institutes of Health R35GM118149 David G Drubin Arnold and Mabel Beckman Foundation Matthew Akamatsu Human Frontier Science Pro - gram LT000234 / 2018 - L Daniel Serwas Army Research Office W911NF1610411 Padmini Rangamani Office of Naval Research N00014 - 17 - 1 - 2628 Padmini Rangamani The funders had no role in study design , data collection and interpretation , or the decision to submit the work for publication . Author contributions Matthew Akamatsu , Conceptualization , Software , Formal analysis , Funding acquisition , Investigation , Visualization , Methodology , Project administration ; Ritvik Vasan , Michael A Ferrin , Software , Formal analysis , Investigation , Methodology ; Daniel Serwas , Investigation , Methodology ; Padmini Ranga - mani , Conceptualization , Supervision , Funding acquisition , Visualization , Project administration ; David G Drubin , Conceptualization , Resources , Supervision , Funding acquisition , Visualization , Proj - ect administration Author ORCIDs Matthew Akamatsu https : / / orcid . org / 0000 - 0002 - 0286 - 5310 Daniel Serwas http : / / orcid . org / 0000 - 0001 - 9010 - 7298 Michael A Ferrin https : / / orcid . org / 0000 - 0002 - 9899 - 1169 Padmini Rangamani https : / / orcid . org / 0000 - 0001 - 5953 - 4347 David G Drubin https : / / orcid . org / 0000 - 0003 - 3002 - 6271 Decision letter and Author response Decision letter https : / / doi . org / 10 . 7554 / eLife . 49840 . sa1 Author response https : / / doi . org / 10 . 7554 / eLife . 49840 . sa2 Additional files Supplementary files . Supplementary file 1 . Model parameters for continuum membrane mechanics model . . Supplementary file 2 . Parameters for coat pulling continuum mechanics simulations . Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 34 of 40 Research article Cell Biology Physics of Living Systems . Supplementary file 3 . Parameters in the model . . Transparent reporting form Data availability All code associated with simulation and analysis is available at https : / / github . com / DrubinBarnes / Akamatsu _ CME _ manuscript ( copy archived at https : / / github . com / elifesciences - publications / Aka - matsu _ CME _ manuscript ) . References Abercrombie M . 1980 . The crawling movement of metazoan cells . Proceedings of the Royal Society of London Series B Biological Sciences 207 : 129 \u2013 147 . DOI : https : / / doi . org / 10 . 1098 / rspb . 1980 . 0017 Aguet F , Antonescu CN , Mettlen M , Schmid SL , Danuser G . 2013 . Advances in analysis of low signal - to - noise images link dynamin and AP2 to the functions of an endocytic checkpoint . Developmental Cell 26 : 279 \u2013 291 . DOI : https : / / doi . org / 10 . 1016 / j . devcel . 2013 . 06 . 019 , PMID : 23891661 Akamatsu M , Lin Y , Bewersdorf J , Pollard TD . 2017 . Analysis of interphase node proteins in fission yeast by quantitative and superresolution fluorescence microscopy . Molecular Biology of the Cell 28 : 3203 \u2013 3214 . DOI : https : / / doi . org / 10 . 1091 / mbc . e16 - 07 - 0522 , PMID : 28539404 Akamatsu M . 2019 . Code associated with Akamatsu et al . manuscript . e0d5426 . GitHub . https : / / github . com / DrubinBarnes / Akamatsu _ CME _ manuscript Alimohamadi H , Vasan R , Hassinger JE , Stachowiak JC , Rangamani P . 2018 . The role of traction in membrane curvature generation . Molecular Biology of the Cell 29 : 2024 \u2013 2035 . DOI : https : / / doi . org / 10 . 1091 / mbc . E18 - 02 - 0087 , PMID : 30044708 Almeida - Souza L , Frank RAW , Garc\u0131\u00b4a - Nafr\u0131\u00b4a J , Colussi A , Gunawardana N , Johnson CM , Yu M , Howard G , Andrews B , Vallis Y , McMahon HT . 2018 . A flat BAR protein promotes actin polymerization at the base of Clathrin - Coated pits . Cell 174 : 325 \u2013 337 . DOI : https : / / doi . org / 10 . 1016 / j . cell . 2018 . 05 . 020 , PMID : 29887380 Amann KJ , Pollard TD . 2001 . The Arp2 / 3 complex nucleates actin filament branches from the sides of pre - existing filaments . Nature Cell Biology 3 : 306 \u2013 310 . DOI : https : / / doi . org / 10 . 1038 / 35060104 , PMID : 11231582 Arasada R , Sayyad WA , Berro J , Pollard TD . 2018 . High - speed superresolution imaging of the proteins in fission yeast clathrin - mediated endocytic actin patches . Molecular Biology of the Cell 29 : 295 \u2013 303 . DOI : https : / / doi . org / 10 . 1091 / mbc . E17 - 06 - 0415 , PMID : 29212877 Avinoam O , Schorb M , Beese CJ , Briggs JA , Kaksonen M . 2015 . Endocytic sites mature by continuous bending and remodeling of the clathrin coat . Science 348 : 1369 \u2013 1372 . DOI : https : / / doi . org / 10 . 1126 / science . aaa9555 , PMID : 26089517 Balzer CJ , Wagner AR , Helgeson LA , Nolen BJ . 2018 . Dip1 Co - opts features of branching nucleation to create linear actin filaments that activate WASP - Bound Arp2 / 3 complex . Current Biology 28 : 3886 \u2013 3891 . DOI : https : / / doi . org / 10 . 1016 / j . cub . 2018 . 10 . 045 , PMID : 30471998 Basu R , Chang F . 2011 . Characterization of dip1p reveals a switch in Arp2 / 3 - dependent actin assembly for fission yeast endocytosis . Current Biology 21 : 905 \u2013 916 . DOI : https : / / doi . org / 10 . 1016 / j . cub . 2011 . 04 . 047 , PMID : 21620704 Beltzner CC , Pollard TD . 2008 . Pathway of actin filament branch formation by Arp2 / 3 complex . Journal of Biological Chemistry 283 : 7135 \u2013 7144 . DOI : https : / / doi . org / 10 . 1074 / jbc . M705894200 , PMID : 18165685 Berro J , Sirotkin V , Pollard TD . 2010 . Mathematical modeling of endocytic actin patch kinetics in fission yeast : disassembly requires release of actin filament fragments . Molecular Biology of the Cell 21 : 2905 \u2013 2915 . DOI : https : / / doi . org / 10 . 1091 / mbc . e10 - 06 - 0494 Berro J , Pollard TD . 2014 . Synergies between Aip1p and capping protein subunits ( Acp1p and Acp2p ) in clathrin - mediated endocytosis and cell polarization in fission yeast . Molecular Biology of the Cell 25 : 3515 \u2013 3527 . DOI : https : / / doi . org / 10 . 1091 / mbc . e13 - 01 - 0005 , PMID : 25143407 Bieling P , Li TD , Weichsel J , McGorty R , Jreij P , Huang B , Fletcher DA , Mullins RD . 2016 . Force feedback controls motor activity and mechanical properties of Self - Assembling branched actin networks . Cell 164 : 115 \u2013 127 . DOI : https : / / doi . org / 10 . 1016 / j . cell . 2015 . 11 . 057 , PMID : 26771487 Bieling P , Hansen SD , Akin O , Li Tai - De , Hayden CC , Fletcher DA , Mullins RD . 2018 . WH2 and proline - rich domains of wasp - family proteins collaborate to accelerate actin filament elongation . The EMBO Journal 37 : 102 \u2013 121 . DOI : https : / / doi . org / 10 . 15252 / embj . 201797039 Blanchoin L , Amann KJ , Higgs HN , Marchand JB , Kaiser DA , Pollard TD . 2000 . Direct observation of dendritic actin filament networks nucleated by Arp2 / 3 complex and WASP / Scar proteins . Nature 404 : 1007 \u2013 1011 . DOI : https : / / doi . org / 10 . 1038 / 35010008 , PMID : 10801131 Boal D , Boal DH . 2012 . Mechanics of the Cell . Cambridge University Press . DOI : https : / / doi . org / 10 . 1017 / CBO9780511810954 Boulant S , Kural C , Zeeh J - C , Ubelmann F , Kirchhausen T . 2011 . Actin dynamics counteract membrane tension during clathrin - mediated endocytosis . Nature Cell Biology 13 : 1124 \u2013 1131 . DOI : https : / / doi . org / 10 . 1038 / ncb2307 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 35 of 40 Research article Cell Biology Physics of Living Systems Brady RJ , Damer CK , Heuser JE , O\u2019Halloran TJ . 2010 . Regulation of Hip1r by epsin controls the temporal and spatial coupling of actin filaments to clathrin - coated pits . Journal of Cell Science 123 : 3652 \u2013 3661 . DOI : https : / / doi . org / 10 . 1242 / jcs . 066852 , PMID : 20923836 Brett TJ , Legendre - Guillemin V , McPherson PS , Fremont DH . 2006 . Structural definition of the F - actin - binding THATCH domain from HIP1R . Nature Structural & Molecular Biology 13 : 121 \u2013 130 . DOI : https : / / doi . org / 10 . 1038 / nsmb1043 , PMID : 16415883 Bucher D , Frey F , Sochacki KA , Kummer S , Bergeest JP , Godinez WJ , Kra\u00a8usslich HG , Rohr K , Taraska JW , Schwarz US , Boulant S . 2018 . Clathrin - adaptor ratio and membrane tension regulate the flat - to - curved transition of the clathrin coat during endocytosis . Nature Communications 9 : 1 \u2013 13 . DOI : https : / / doi . org / 10 . 1038 / s41467 - 018 - 03533 - 0 , PMID : 29549258 Buser C , Drubin DG . 2013 . Ultrastructural imaging of endocytic sites in Saccharomyces cerevisiae by transmission electron microscopy and immunolabeling . Microscopy and Microanalysis 19 : 381 \u2013 392 . DOI : https : / / doi . org / 10 . 1017 / S1431927612014304 , PMID : 23458500 Carlsson AE . 2001 . Growth of branched actin networks against obstacles . Biophysical Journal 81 : 1907 \u2013 1923 . DOI : https : / / doi . org / 10 . 1016 / S0006 - 3495 ( 01 ) 75842 - 0 , PMID : 11566765 Carlsson AE . 2018 . Membrane bending by actin polymerization . Current Opinion in Cell Biology 50 : 1 \u2013 7 . DOI : https : / / doi . org / 10 . 1016 / j . ceb . 2017 . 11 . 007 , PMID : 29207306 Carlsson AE , Bayly PV . 2014 . Force generation by endocytic actin patches in budding yeast . Biophysical Journal 106 : 1596 \u2013 1606 . DOI : https : / / doi . org / 10 . 1016 / j . bpj . 2014 . 02 . 035 , PMID : 24739159 Case LB , Zhang X , Ditlev JA , Rosen MK . 2019 . Stoichiometry controls activity of phase - separated clusters of actin signaling proteins . Science 363 : 1093 \u2013 1097 . DOI : https : / / doi . org / 10 . 1126 / science . aau6313 , PMID : 30846599 Chaudhuri O , Parekh SH , Fletcher DA . 2007 . Reversible stress softening of actin networks . Nature 445 : 295 \u2013 298 . DOI : https : / / doi . org / 10 . 1038 / nature05459 Chen H , Fre S , Slepnev VI , Capua MR , Takei K , Butler MH , Di Fiore PP , De Camilli P . 1998 . Epsin is an EH - domain - binding protein implicated in clathrin - mediated endocytosis . Nature 394 : 793 \u2013 797 . DOI : https : / / doi . org / 10 . 1038 / 29555 , PMID : 9723620 Chen CY , Brodsky FM . 2005 . Huntingtin - interacting prote in 1 ( Hip1 ) and Hip1 - related protein ( Hip1R ) bind the conserved sequence of clathr in light chains and thereby influence clathr in assembly in vitro and act in distribution in vivo . Journal of Biological Chemistry 280 : 6109 \u2013 6117 . DOI : https : / / doi . org / 10 . 1074 / jbc . M408454200 , PMID : 15533940 Clarke NI , Royle SJ . 2018 . FerriTag is a new genetically - encoded inducible tag for correlative light - electron microscopy . Nature Communications 9 : 1 \u2013 10 . DOI : https : / / doi . org / 10 . 1038 / s41467 - 018 - 04993 - 0 Collins A , Warrington A , Taylor KA , Svitkina T . 2011 . Structural organization of the actin cytoskeleton at sites of clathrin - mediated endocytosis . Current Biology 21 : 1167 \u2013 1175 . DOI : https : / / doi . org / 10 . 1016 / j . cub . 2011 . 05 . 048 , PMID : 21723126 Dambournet D , Hong SH , Grassart A , Drubin DG . 2014 . Tagging endogenous loci for live - cell fluorescence imaging and molecule counting using ZFNs , TALENs , and Cas9 . Methods in Enzymology 546 : 139 \u2013 160 . DOI : https : / / doi . org / 10 . 1016 / B978 - 0 - 12 - 801185 - 0 . 00007 - 6 , PMID : 25398339 Dambournet D , Sochacki KA , Cheng AT , Akamatsu M , Taraska JW , Hockemeyer D , Drubin DG . 2018 . Genome - edited human stem cells expressing fluorescently labeled endocytic markers allow quantitative analysis of clathrin - mediated endocytosis during differentiation . The Journal of Cell Biology 217 : 3301 \u2013 3311 . DOI : https : / / doi . org / 10 . 1083 / jcb . 201710084 Daste F , Walrant A , Holst MR , Gadsby JR , Mason J , Lee JE , Brook D , Mettlen M , Larsson E , Lee SF , Lundmark R , Gallop JL . 2017 . Control of actin polymerization via the coincidence of phosphoinositides and high membrane curvature . The Journal of Cell Biology 216 : 3745 \u2013 3765 . DOI : https : / / doi . org / 10 . 1083 / jcb . 201704061 , PMID : 2 8923975 David C , McPherson PS , Mundigl O , de Camilli P . 1996 . A role of amphiphysin in synaptic vesicle endocytosis suggested by its binding to dynamin in nerve terminals . PNAS 93 : 331 \u2013 335 . DOI : https : / / doi . org / 10 . 1073 / pnas . 93 . 1 . 331 , PMID : 8552632 De La Cruz EM , Martiel JL , Blanchoin L . 2015b . Mechanical heterogeneity favors fragmentation of strained actin filaments . Biophysical Journal 108 : 2270 \u2013 2281 . DOI : https : / / doi . org / 10 . 1016 / j . bpj . 2015 . 03 . 058 , PMID : 25954 884 De La Cruz EM , Gardel ML . 2015a . Actin mechanics and fragmentation . Journal of Biological Chemistry 290 : 17137 \u2013 17144 . DOI : https : / / doi . org / 10 . 1074 / jbc . R115 . 636472 , PMID : 25957404 Dere\u00b4nyi I , Ju\u00a8licher F , Prost J . 2002 . Formation and interaction of membrane tubes . Physical Review Letters 88 : 37 . DOI : https : / / doi . org / 10 . 1103 / PhysRevLett . 88 . 238101 Dimova R . 2014 . Recent developments in the field of bending rigidity measurements on membranes . Advances in Colloid and Interface Science 208 : 225 \u2013 234 . DOI : https : / / doi . org / 10 . 1016 / j . cis . 2014 . 03 . 003 , PMID : 24666592 Ditlev JA , Michalski PJ , Huber G , Rivera GM , Mohler WA , Loew LM , Mayer BJ . 2012 . Stoichiometry of Nck - dependent actin polymerization in living cells . The Journal of Cell Biology 197 : 643 \u2013 658 . DOI : https : / / doi . org / 10 . 1083 / jcb . 201111113 , PMID : 22613834 Diz - Mun\u02dcoz A , Thurley K , Chintamen S , Altschuler SJ , Wu LF , Fletcher DA , Weiner OD . 2016 . Membrane tension acts through PLD2 and mTORC2 to limit actin network assembly during neutrophil migration . PLOS Biology 14 : e1002474 . DOI : https : / / doi . org / 10 . 1371 / journal . pbio . 1002474 , PMID : 27280401 Dmitrieff S , Ne\u00b4de\u00b4lec F . 2015 . Membrane mechanics of endocytosis in cells with turgor . PLOS Computational Biology 11 : e1004538 . DOI : https : / / doi . org / 10 . 1371 / journal . pcbi . 1004538 , PMID : 26517669 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 36 of 40 Research article Cell Biology Physics of Living Systems Dmitrieff S , Ne\u00b4de\u00b4lec F . 2016 . Amplification of actin polymerization forces . The Journal of Cell Biology 212 : 763 \u2013 766 . DOI : https : / / doi . org / 10 . 1083 / jcb . 201512019 , PMID : 27002174 Doyon JB , Zeitler B , Cheng J , Cheng AT , Cherone JM , Santiago Y , Lee AH , Vo TD , Doyon Y , Miller JC , Paschon DE , Zhang L , Rebar EJ , Gregory PD , Urnov FD , Drubin DG . 2011 . Rapid and efficient clathrin - mediated endocytosis revealed in genome - edited mammalian cells . Nature Cell Biology 13 : 331 \u2013 337 . DOI : https : / / doi . org / 10 . 1038 / ncb2175 , PMID : 21297641 Egile C , Rouiller I , Xu XP , Volkmann N , Li R , Hanein D . 2005 . Mechanism of filament nucleation and branch stability revealed by the structure of the Arp2 / 3 complex at actin branch junctions . PLOS Biology 3 : e383 . DOI : https : / / doi . org / 10 . 1371 / journal . pbio . 0030383 , PMID : 16262445 Engqvist - Goldstein AE , Kessels MM , Chopra VS , Hayden MR , Drubin DG . 1999 . An actin - binding protein of the Sla2 / Huntingtin interacting protein 1 family is a novel component of clathrin - coated pits and vesicles . The Journal of Cell Biology 147 : 1503 \u2013 1518 . DOI : https : / / doi . org / 10 . 1083 / jcb . 147 . 7 . 1503 , PMID : 10613908 Engqvist - Goldstein AE , Warren RA , Kessels MM , Keen JH , Heuser J , Drubin DG . 2001 . The actin - binding protein Hip1R associates with clathrin during early stages of endocytosis and promotes clathrin assembly in vitro . The Journal of Cell Biology 154 : 1209 \u2013 1224 . DOI : https : / / doi . org / 10 . 1083 / jcb . 200106089 , PMID : 11564758 Engqvist - Goldstein AE , Zhang CX , Carreno S , Barroso C , Heuser JE , Drubin DG . 2004 . RNAi - mediated Hip1R silencing results in stable association between the endocytic machinery and the actin assembly machinery . Molecular Biology of the Cell 15 : 1666 \u2013 1679 . DOI : https : / / doi . org / 10 . 1091 / mbc . e03 - 09 - 0639 , PMID : 14742709 Engqvist - Goldstein AE , Drubin DG . 2003 . Actin assembly and endocytosis : from yeast to mammals . Annual Review of Cell and Developmental Biology 19 : 287 \u2013 332 . DOI : https : / / doi . org / 10 . 1146 / annurev . cellbio . 19 . 111401 . 093127 , PMID : 14570572 Epstein AE , Espinoza - Sanchez S , Pollard TD . 2018 . Phosphorylation of Arp2 is not essential for Arp2 / 3 complex activity in fission yeast . Life Science Alliance 1 : e201800202 . DOI : https : / / doi . org / 10 . 26508 / lsa . 201800202 , PMID : 30456391 Ferguson JP , Willy NM , Heidotting SP , Huber SD , Webber MJ , Kural C . 2016 . Deciphering dynamics of clathrin - mediated endocytosis in a living organism . The Journal of Cell Biology 214 : 347 \u2013 358 . DOI : https : / / doi . org / 10 . 1083 / jcb . 201604128 , PMID : 27458134 Ferguson JP , Huber SD , Willy NM , Aygu\u00a8n E , Goker S , Atabey T , Kural C . 2017 . Mechanoregulation of clathrin - mediated endocytosis . Journal of Cell Science 130 : 3631 \u2013 3636 . DOI : https : / / doi . org / 10 . 1242 / jcs . 205930 , PMID : 28923837 Ferna\u00b4ndez P , Pullarkat PA , Ott A . 2006 . A master relation defines the nonlinear viscoelasticity of single fibroblasts . Biophysical Journal 90 : 3796 \u2013 3805 . DOI : https : / / doi . org / 10 . 1529 / biophysj . 105 . 072215 , PMID : 16461394 Footer MJ , Kerssemakers JW , Theriot JA , Dogterom M . 2007 . Direct measurement of force generation by actin filament polymerization using an optical trap . PNAS 104 : 2181 \u2013 2186 . DOI : https : / / doi . org / 10 . 1073 / pnas . 0607052104 , PMID : 17277076 Foret L . 2014 . Shape and energy of a membrane bud induced by protein coats or viral protein assembly . The European Physical Journal E 37 : 42 . DOI : https : / / doi . org / 10 . 1140 / epje / i2014 - 14042 - 1 Garcia - Alai MM , Heidemann J , Skruzny M , Gieras A , Mertens HDT , Svergun DI , Kaksonen M , Uetrecht C , Meijers R . 2018 . Epsin and Sla2 form assemblies through phospholipid interfaces . Nature Communications 9 : 1 \u2013 13 . DOI : https : / / doi . org / 10 . 1038 / s41467 - 017 - 02443 - x , PMID : 29362354 Gittes F , Mickey B , Nettleton J , Howard J . 1993 . Flexural rigidity of microtubules and actin filaments measured from thermal fluctuations in shape . The Journal of Cell Biology 120 : 923 \u2013 934 . DOI : https : / / doi . org / 10 . 1083 / jcb . 120 . 4 . 923 , PMID : 8432732 Grassart A , Cheng AT , Hong SH , Zhang F , Zenzer N , Feng Y , Briner DM , Davis GD , Malkov D , Drubin DG . 2014 . Actin and dynamin2 dynamics and interplay during clathrin - mediated endocytosis . The Journal of Cell Biology 205 : 721 \u2013 735 . DOI : https : / / doi . org / 10 . 1083 / jcb . 201403041 , PMID : 24891602 Hassinger JE , Oster G , Drubin DG , Rangamani P . 2017 . Design principles for robust vesiculation in clathrin - mediated endocytosis . PNAS 114 : E1118 \u2013 E1127 . DOI : https : / / doi . org / 10 . 1073 / pnas . 1617705114 , PMID : 2 8126722 Haugwitz M , Noegel AA , Karakesisoglou J , Schleicher M . 1994 . Dictyostelium amoebae that lack G - actin - sequestering profilins show defects in F - actin content , Cytokinesis , and development . Cell 79 : 303 \u2013 314 . DOI : https : / / doi . org / 10 . 1016 / 0092 - 8674 ( 94 ) 90199 - 6 , PMID : 7954798 Helfrich W . 1973 . Elastic properties of lipid bilayers : theory and possible experiments . Zeitschrift Fu\u00a8r Naturforschung C 28 : 693 \u2013 703 . DOI : https : / / doi . org / 10 . 1515 / znc - 1973 - 11 - 1209 Hetrick B , Han MS , Helgeson LA , Nolen BJ . 2013 . Small molecules CK - 666 and CK - 869 inhibit actin - related protein 2 / 3 complex by blocking an activating conformational change . Chemistry & Biology 20 : 701 \u2013 712 . DOI : https : / / doi . org / 10 . 1016 / j . chembiol . 2013 . 03 . 019 , PMID : 23623350 Hong SH , Cortesio CL , Drubin DG . 2015 . Machine - Learning - Based analysis in Genome - Edited cells reveals the efficiency of Clathrin - Mediated endocytosis . Cell Reports 12 : 2121 \u2013 2130 . DOI : https : / / doi . org / 10 . 1016 / j . celrep . 2015 . 08 . 048 , PMID : 26387943 Hsia Y , Bale JB , Gonen S , Shi D , Sheffler W , Fong KK , Nattermann U , Xu C , Huang P - S , Ravichandran R , Yi S , Davis TN , Gonen T , King NP , Baker D . 2016 . Design of a hyperstable 60 - subunit protein icosahedron . Nature 535 : 136 \u2013 139 . DOI : https : / / doi . org / 10 . 1038 / nature18010 Idrissi FZ , Gro\u00a8tsch H , Ferna\u00b4ndez - Golbano IM , Presciatto - Baschong C , Riezman H , Geli MI . 2008 . Distinct acto / myosin - I structures associate with endocytic profiles at the plasma membrane . The Journal of Cell Biology 180 : 1219 \u2013 1232 . DOI : https : / / doi . org / 10 . 1083 / jcb . 200708060 , PMID : 18347067 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 37 of 40 Research article Cell Biology Physics of Living Systems Idrissi FZ , Blasco A , Espinal A , Geli MI . 2012 . Ultrastructural dynamics of proteins involved in Endocytic budding . PNAS 109 : E2587 \u2013 E2594 . DOI : https : / / doi . org / 10 . 1073 / pnas . 1202789109 , PMID : 22949647 Isambert H , Venier P , Maggs AC , Fattoum A , Kassab R , Pantaloni D , Carlier MF . 1995 . Flexibility of actin filaments derived from thermal fluctuations . Effect of bound nucleotide , phalloidin , and muscle regulatory proteins . Journal of Biological Chemistry 270 : 11437 \u2013 11444 . DOI : https : / / doi . org / 10 . 1074 / jbc . 270 . 19 . 11437 , PMID : 7744781 Jaqaman K , Loerke D , Mettlen M , Kuwata H , Grinstein S , Schmid SL , Danuser G . 2008 . Robust single - particle tracking in live - cell time - lapse sequences . Nature Methods 5 : 695 \u2013 702 . DOI : https : / / doi . org / 10 . 1038 / nmeth . 1237 , PMID : 18641657 Kaksonen M , Sun Y , Drubin DG . 2003 . A pathway for association of receptors , adaptors , and actin during endocytic internalization . Cell 115 : 475 \u2013 487 . DOI : https : / / doi . org / 10 . 1016 / S0092 - 8674 ( 03 ) 00883 - 3 , PMID : 14622601 Kaksonen M , Toret CP , Drubin DG . 2005 . A modular design for the clathrin - and actin - mediated endocytosis machinery . Cell 123 : 305 \u2013 320 . DOI : https : / / doi . org / 10 . 1016 / j . cell . 2005 . 09 . 024 , PMID : 16239147 Kaksonen M , Roux A . 2018 . Mechanisms of clathrin - mediated endocytosis . Nature Reviews Molecular Cell Biology 3 : 1 \u2013 14 . DOI : https : / / doi . org / 10 . 1038 / nrm . 2017 . 132 Keren K , Pincus Z , Allen GM , Barnhart EL , Marriott G , Mogilner A , Theriot JA . 2008 . Mechanism of shape determination in motile cells . Nature 453 : 475 \u2013 480 . DOI : https : / / doi . org / 10 . 1038 / nature06952 , PMID : 18497 816 King NP , Bale JB , Sheffler W , McNamara DE , Gonen S , Gonen T , Yeates TO , Baker D . 2014 . Accurate design of co - assembling multi - component protein nanomaterials . Nature 510 : 103 \u2013 108 . DOI : https : / / doi . org / 10 . 1038 / nature13404 Kishimoto T , Sun Y , Buser C , Liu J , Michelot A , Drubin DG . 2011 . Determinants of endocytic membrane geometry , stability , and scission . PNAS 108 : E979 \u2013 E988 . DOI : https : / / doi . org / 10 . 1073 / pnas . 1113413108 , PMID : 22006337 Kremer JR , Mastronarde DN , McIntosh JR . 1996 . Computer visualization of three - dimensional image data using IMOD . Journal of Structural Biology 116 : 71 \u2013 76 . DOI : https : / / doi . org / 10 . 1006 / jsbi . 1996 . 0013 , PMID : 8742726 Kukulski W , Schorb M , Kaksonen M , Briggs JA . 2012 . Plasma membrane reshaping during endocytosis is revealed by time - resolved electron tomography . Cell 150 : 508 \u2013 520 . DOI : https : / / doi . org / 10 . 1016 / j . cell . 2012 . 05 . 046 , PMID : 22863005 Lacayo CI , Pincus Z , VanDuijn MM , Wilson CA , Fletcher DA , Gertler FB , Mogilner A , Theriot JA . 2007 . Emergence of large - scale cell morphology and movement from local actin filament growth dynamics . PLOS Biology 5 : e233 . DOI : https : / / doi . org / 10 . 1371 / journal . pbio . 0050233 , PMID : 17760506 Leake MC , Chandler JH , Wadhams GH , Bai F , Berry RM , Armitage JP . 2006 . Stoichiometry and turnover in single , functioning membrane protein complexes . Nature 443 : 355 \u2013 358 . DOI : https : / / doi . org / 10 . 1038 / nature05135 , PMID : 16971952 Liu J , Sun Y , Drubin DG , Oster GF . 2009 . The mechanochemistry of endocytosis . PLOS Biology 7 : e1000204 . DOI : https : / / doi . org / 10 . 1371 / journal . pbio . 1000204 , PMID : 19787029 Ma R , Berro J . 2018 . Structural organization and energy storage in crosslinked actin assemblies . PLOS Computational Biology 14 : e1006150 . DOI : https : / / doi . org / 10 . 1371 / journal . pcbi . 1006150 , PMID : 29813051 Maly IV , Borisy GG . 2001 . Self - organization of a propulsive actin network as an evolutionary process . PNAS 98 : 11324 \u2013 11329 . DOI : https : / / doi . org / 10 . 1073 / pnas . 181338798 , PMID : 11572984 Mastronarde DN . 2005 . Automated electron microscope tomography using robust prediction of specimen movements . Journal of Structural Biology 152 : 36 \u2013 51 . DOI : https : / / doi . org / 10 . 1016 / j . jsb . 2005 . 07 . 007 , PMID : 16182563 McCormick CD , Akamatsu MS , Ti SC , Pollard TD . 2013 . Measuring affinities of fission yeast spindle pole body proteins in live cells across the cell cycle . Biophysical Journal 105 : 1324 \u2013 1335 . DOI : https : / / doi . org / 10 . 1016 / j . bpj . 2013 . 08 . 017 , PMID : 24047983 McCullough BR , Blanchoin L , Martiel JL , De la Cruz EM . 2008 . Cofilin increases the bending flexibility of actin filaments : implications for severing and cell mechanics . Journal of Molecular Biology 381 : 550 \u2013 558 . DOI : https : / / doi . org / 10 . 1016 / j . jmb . 2008 . 05 . 055 , PMID : 18617188 Messa M , Ferna\u00b4ndez - Busnadiego R , Sun EW , Chen H , Czapla H , Wrasman K , Wu Y , Ko G , Ross T , Wendland B , De Camilli P . 2014 . Epsin deficiency impairs endocytosis by stalling the actin - dependent invagination of endocytic clathrin - coated pits . eLife 3 : e03311 . DOI : https : / / doi . org / 10 . 7554 / eLife . 03311 , PMID : 25122462 Mogilner A , Edelstein - Keshet L . 2002 . Regulation of actin dynamics in rapidly moving cells : a quantitative analysis . Biophysical Journal 83 : 1237 \u2013 1258 . DOI : https : / / doi . org / 10 . 1016 / S0006 - 3495 ( 02 ) 73897 - 6 , PMID : 12202352 Mogilner A , Oster G . 1996 . Cell motility driven by actin polymerization . Biophysical Journal 71 : 3030 \u2013 3045 . DOI : https : / / doi . org / 10 . 1016 / S0006 - 3495 ( 96 ) 79496 - 1 , PMID : 8968574 Mueller J , Szep G , Nemethova M , de Vries I , Lieber AD , Winkler C , Kruse K , Small JV , Schmeiser C , Keren K , Hauschild R , Sixt M . 2017 . Load adaptation of lamellipodial actin networks . Cell 171 : 188 \u2013 200 . DOI : https : / / doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 , PMID : 28867286 Mullins RD , Stafford WF , Pollard TD . 1997 . Structure , subunit topology , and actin - binding activity of the Arp2 / 3 complex from Acanthamoeba . The Journal of Cell Biology 136 : 331 \u2013 343 . DOI : https : / / doi . org / 10 . 1083 / jcb . 136 . 2 . 331 , PMID : 9015304 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 38 of 40 Research article Cell Biology Physics of Living Systems Mund M , van der Beek JA , Deschamps J , Dmitrieff S , Hoess P , Monster JL , Picco A , Ne\u00b4de\u00b4lec F , Kaksonen M , Ries J . 2018 . Systematic nanoscale analysis of endocytosis links efficient vesicle formation to patterned actin nucleation . Cell 174 : 884 \u2013 896 . DOI : https : / / doi . org / 10 . 1016 / j . cell . 2018 . 06 . 032 , PMID : 30057119 Naghdi PM . 1957 . On the theory of thin elastic shells . Quarterly of Applied Mathematics 14 : 369 \u2013 380 . DOI : https : / / doi . org / 10 . 1090 / qam / 84284 Nedelec F , Foethke D . 2007 . Collective Langevin dynamics of flexible cytoskeletal fibers . New Journal of Physics 9 : 427 . DOI : https : / / doi . org / 10 . 1088 / 1367 - 2630 / 9 / 11 / 427 Niu Q , Ybe JA . 2008 . Crystal structure at 2 . 8 A of Huntingtin - interacting protein 1 ( HIP1 ) coiled - coil domain reveals a charged surface suitable for HIP1 protein interactor ( HIPPI ) . Journal of Molecular Biology 375 : 1197 \u2013 1205 . DOI : https : / / doi . org / 10 . 1016 / j . jmb . 2007 . 11 . 036 , PMID : 18155047 Nolen BJ , Tomasevic N , Russell A , Pierce DW , Jia Z , McCormick CD , Hartman J , Sakowicz R , Pollard TD . 2009 . Characterization of two classes of small molecule inhibitors of Arp2 / 3 complex . Nature 460 : 1031 \u2013 1034 . DOI : https : / / doi . org / 10 . 1038 / nature08231 , PMID : 19648907 Parekh SH , Chaudhuri O , Theriot JA , Fletcher DA . 2005 . Loading history determines the velocity of actin - network growth . Nature Cell Biology 7 : 1219 \u2013 1223 . DOI : https : / / doi . org / 10 . 1038 / ncb1336 , PMID : 16299496 Pearse BM . 1976 . Clathrin : a unique protein associated with intracellular transfer of membrane by coated vesicles . PNAS 73 : 1255 \u2013 1259 . DOI : https : / / doi . org / 10 . 1073 / pnas . 73 . 4 . 1255 , PMID : 1063406 Peskin CS , Odell GM , Oster GF . 1993 . Cellular motions and thermal fluctuations : the brownian ratchet . Biophysical Journal 65 : 316 \u2013 324 . DOI : https : / / doi . org / 10 . 1016 / S0006 - 3495 ( 93 ) 81035 - X , PMID : 8369439 Pfaendtner J , Lyman E , Pollard TD , Voth GA . 2010 . Structure and dynamics of the actin filament . Journal of Molecular Biology 396 : 252 \u2013 263 . DOI : https : / / doi . org / 10 . 1016 / j . jmb . 2009 . 11 . 034 , PMID : 19931282 Picco A , Mund M , Ries J , Ne\u00b4de\u00b4lec F , Kaksonen M . 2015 . Visualizing the functional architecture of the endocytic machinery . eLife 4 : e04535 . DOI : https : / / doi . org / 10 . 7554 / eLife . 04535 Picco A , Kukulski W , Manenschijn HE , Specht T , Briggs JAG , Kaksonen M . 2018 . The contributions of the actin machinery to endocytic membrane bending and vesicle formation . Molecular Biology of the Cell 29 : 1346 \u2013 1358 . DOI : https : / / doi . org / 10 . 1091 / mbc . E17 - 11 - 0688 , PMID : 29851558 Pollard TD . 1986 . Rate constants for the reactions of ATP - and ADP - actin with the ends of actin filaments . The Journal of Cell Biology 103 : 2747 \u2013 2754 . DOI : https : / / doi . org / 10 . 1083 / jcb . 103 . 6 . 2747 , PMID : 3793756 Pollard TD . 2010 . A guide to simple and informative binding assays . Molecular Biology of the Cell 21 : 4061 \u2013 4067 . DOI : https : / / doi . org / 10 . 1091 / mbc . e10 - 08 - 0683 , PMID : 21115850 Pollard TD . 2016 . Actin and Actin - Binding proteins . Cold Spring Harbor Perspectives in Biology 8 : a018226 . DOI : https : / / doi . org / 10 . 1101 / cshperspect . a018226 , PMID : 26988969 Rangamani P , Fardin M - A , Xiong Y , Lipshtat A , Rossier O , Sheetz MP , Iyengar R . 2011 . Signaling network triggers and membrane physical properties control the actin Cytoskeleton - Driven isotropic phase of cell spreading . Biophysical Journal 100 : 845 \u2013 857 . DOI : https : / / doi . org / 10 . 1016 / j . bpj . 2010 . 12 . 3732 Rangamani P , Agrawal A , Mandadapu KK , Oster G , Steigmann DJ . 2013 . Interaction between surface shape and intra - surface viscous flow on lipid membranes . Biomechanics and Modeling in Mechanobiology 12 : 833 \u2013 845 . DOI : https : / / doi . org / 10 . 1007 / s10237 - 012 - 0447 - y Rangamani P , Mandadap KK , Oster G . 2014 . Protein - Induced membrane curvature alters local membrane tension . Biophysical Journal 107 : 751 \u2013 762 . DOI : https : / / doi . org / 10 . 1016 / j . bpj . 2014 . 06 . 010 Raz - Ben Aroush D , Ofer N , Abu - Shah E , Allard J , Krichevsky O , Mogilner A , Keren K . 2017 . Actin turnover in lamellipodial fragments . Current Biology 27 : 2963 \u2013 2973 . DOI : https : / / doi . org / 10 . 1016 / j . cub . 2017 . 08 . 066 , PMID : 28966086 Risca VI , Wang EB , Chaudhuri O , Chia JJ , Geissler PL , Fletcher DA . 2012 . Actin filament curvature biases branching direction . PNAS 109 : 2913 \u2013 2918 . DOI : https : / / doi . org / 10 . 1073 / pnas . 1114292109 , PMID : 22308368 Rohatgi R , Ho HY , Kirschner MW . 2000 . Mechanism of N - WASP activation by CDC42 and phosphatidylinositol 4 , 5 - bisphosphate . The Journal of Cell Biology 150 : 1299 \u2013 1310 . DOI : https : / / doi . org / 10 . 1083 / jcb . 150 . 6 . 1299 , PMID : 10995436 Rottner K , Faix J , Bogdan S , Linder S , Kerkhoff E . 2017 . Actin assembly mechanisms at a glance . Journal of Cell Science 130 : 3427 \u2013 3435 . DOI : https : / / doi . org / 10 . 1242 / jcs . 206433 , PMID : 29032357 Rotty JD , Wu C , Bear JE . 2013 . New insights into the regulation and cellular functions of the ARP2 / 3 complex . Nature Reviews Molecular Cell Biology 14 : 7 \u2013 12 . DOI : https : / / doi . org / 10 . 1038 / nrm3492 , PMID : 23212475 Schaus TE , Taylor EW , Borisy GG . 2007 . Self - organization of actin filament orientation in the dendritic - nucleation / array - treadmilling model . PNAS 104 : 7086 \u2013 7091 . DOI : https : / / doi . org / 10 . 1073 / pnas . 0701943104 , PMID : 17440042 Scott BL , Sochacki KA , Low - Nam ST , Bailey EM , Luu Q , Hor A , Dickey AM , Smith S , Kerkvliet JG , Taraska JW , Hoppe AD . 2018 . Membrane bending occurs at all stages of clathrin - coat assembly and defines endocytic dynamics . Nature Communications 9 : 419 . DOI : https : / / doi . org / 10 . 1038 / s41467 - 018 - 02818 - 8 , PMID : 29379015 Senetar MA , Foster SJ , McCann RO . 2004 . Intrasteric inhibition mediates the interaction of the I / LWEQ module proteins Talin1 , Talin2 , Hip1 , and Hip12 with actin . Biochemistry 43 : 15418 \u2013 15428 . DOI : https : / / doi . org / 10 . 1021 / bi0487239 , PMID : 15581353 Sept D , McCammon JA . 2001 . Thermodynamics and kinetics of actin filament nucleation . Biophysical Journal 81 : 667 \u2013 674 . DOI : https : / / doi . org / 10 . 1016 / S0006 - 3495 ( 01 ) 75731 - 1 , PMID : 11463615 Shi Z , Graber ZT , Baumgart T , Stone HA , Cohen AE . 2018 . Cell membranes resist flow . Cell 175 : 1769 \u2013 1779 . DOI : https : / / doi . org / 10 . 1016 / j . cell . 2018 . 09 . 054 , PMID : 30392960 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 39 of 40 Research article Cell Biology Physics of Living Systems Sirotkin V , Berro J , Macmillan K , Zhao L , Pollard TD . 2010 . Quantitative analysis of the mechanism of endocytic actin patch assembly and disassembly in fission yeast . Molecular Biology of the Cell 21 : 2894 \u2013 2904 . DOI : https : / / doi . org / 10 . 1091 / mbc . e10 - 02 - 0157 , PMID : 20587778 Skruzny M , Brach T , Ciuffa R , Rybina S , Wachsmuth M , Kaksonen M . 2012 . Molecular basis for coupling the plasma membrane to the actin cytoskeleton during clathrin - mediated endocytosis . PNAS 109 : E2533 \u2013 E2542 . DOI : https : / / doi . org / 10 . 1073 / pnas . 1207011109 , PMID : 22927393 Skruzny M , Desfosses A , Prinz S , Dodonova SO , Gieras A , Uetrecht C , Jakobi AJ , Abella M , Hagen WJ , Schulz J , Meijers R , Rybin V , Briggs JA , Sachse C , Kaksonen M . 2015 . An organized co - assembly of clathrin adaptors is essential for endocytosis . Developmental Cell 33 : 150 \u2013 162 . DOI : https : / / doi . org / 10 . 1016 / j . devcel . 2015 . 02 . 023 , PMID : 25898165 Smith BA , Daugherty - Clarke K , Goode BL , Gelles J . 2013 . Pathway of actin filament branch formation by Arp2 / 3 complex revealed by single - molecule imaging . PNAS 110 : 1285 \u2013 1290 . DOI : https : / / doi . org / 10 . 1073 / pnas . 1211164110 Sochacki KA , Dickey AM , Strub M - P , Taraska JW . 2017 . Endocytic proteins are partitioned at the edge of the clathrin lattice in mammalian cells . Nature Cell Biology 19 : 352 \u2013 361 . DOI : https : / / doi . org / 10 . 1038 / ncb3498 Sochacki KA , Taraska JW . 2019 . From flat to curved clathrin : controlling a plastic ratchet . Trends in Cell Biology 29 : 241 \u2013 256 . DOI : https : / / doi . org / 10 . 1016 / j . tcb . 2018 . 12 . 002 , PMID : 30598298 Stachowiak JC , Schmid EM , Ryan CJ , Ann HS , Sasaki DY , Sherman MB , Geissler PL , Fletcher DA , Hayden CC . 2012 . Membrane bending by protein - protein crowding . Nature Cell Biology 14 : 944 \u2013 949 . DOI : https : / / doi . org / 10 . 1038 / ncb2561 , PMID : 22902598 Steigmann DJ . 1999 . Fluid films with curvature elasticity . Archive for Rational Mechanics and Analysis 150 : 127 \u2013 152 . DOI : https : / / doi . org / 10 . 1007 / s002050050183 Steigmann D , Baesu E , Rudd R , Belak J , McElfresh M . 2003 . On the variational theory of cell - membrane equilibria . Interfaces and Free Boundaries 5 : 357 \u2013 366 . DOI : https : / / doi . org / 10 . 4171 / IFB / 83 Sun Y , Martin AC , Drubin DG . 2006 . Endocytic internalization in budding yeast requires coordinated actin nucleation and myosin motor activity . Developmental Cell 11 : 33 \u2013 46 . DOI : https : / / doi . org / 10 . 1016 / j . devcel . 2006 . 05 . 008 , PMID : 16824951 Sun Y , Leong NT , Jiang T , Tangara A , Darzacq X , Drubin DG . 2017 . Switch - like Arp2 / 3 activation upon WASP and WIP recruitment to an apparent threshold level by multivalent linker proteins in vivo . eLife 6 : e29140 . DOI : https : / / doi . org / 10 . 7554 / eLife . 29140 , PMID : 28813247 Svitkina TM , Borisy GG . 1999 . Arp2 / 3 complex and actin depolymerizing factor / cofilin in dendritic organization and treadmilling of actin filament array in Lamellipodia . The Journal of Cell Biology 145 : 1009 \u2013 1026 . DOI : https : / / doi . org / 10 . 1083 / jcb . 145 . 5 . 1009 , PMID : 10352018 Taylor MJ , Perrais D , Merrifield CJ . 2011 . A high precision survey of the molecular dynamics of mammalian clathrin - mediated endocytosis . PLOS Biology 9 : e1000604 . DOI : https : / / doi . org / 10 . 1371 / journal . pbio . 1000604 , PMID : 21445324 Vasan R , Akamatsu M , Scho\u00a8neberg J , Rangamani P . 2018 . Intracellular Membrane Trafficking : Modeling Local Movements in Cells . In : Stolarska M , Tarfulea N ( Eds ) . Cell Movement : Modeling and Applications . Springer International Publishing . p . 259 \u2013 301 . DOI : https : / / doi . org / 10 . 1007 / 978 - 3 - 319 - 96842 - 1 _ 9 Vinzenz M , Nemethova M , Schur F , Mueller J , Narita A , Urban E , Winkler C , Schmeiser C , Koestler SA , Rottner K , Resch GP , Maeda Y , Small JV . 2012 . Actin branching in the initiation and maintenance of lamellipodia . Journal of Cell Science 125 : 2775 \u2013 2785 . DOI : https : / / doi . org / 10 . 1242 / jcs . 107623 , PMID : 22431015 Wagner AR , Luan Q , Liu SL , Nolen BJ . 2013 . Dip1 defines a class of Arp2 / 3 complex activators that function without preformed actin filaments . Current Biology 23 : 1990 \u2013 1998 . DOI : https : / / doi . org / 10 . 1016 / j . cub . 2013 . 08 . 029 , PMID : 24120641 Walani N , Torres J , Agrawal A . 2014 . Anisotropic spontaneous curvatures in lipid membranes . Physical Review E 89 : 693 \u2013 698 . DOI : https : / / doi . org / 10 . 1103 / PhysRevE . 89 . 062715 Wang X , Galletta BJ , Cooper JA , Carlsson AE . 2016 . Actin - Regulator feedback interactions during endocytosis . Biophysical Journal 110 : 1430 \u2013 1443 . DOI : https : / / doi . org / 10 . 1016 / j . bpj . 2016 . 02 . 018 , PMID : 27028652 Wang X , Carlsson AE . 2017 . A master equation approach to actin polymerization applied to endocytosis in yeast . PLOS Computational Biology 13 : e1005901 . DOI : https : / / doi . org / 10 . 1371 / journal . pcbi . 1005901 , PMID : 2 9240771 Wu X - s , Elias S , Liu H , Heureaux J , Wen PJ , Liu AP , Kozlov MM , L - g W . 2017 . Membrane tension inhibits rapid and slow endocytosis in secretory cells . Biophysj 113 : 2406 \u2013 2414 . DOI : https : / / doi . org / 10 . 1016 / j . bpj . 2017 . 09 . 035 Wu JQ , Pollard TD . 2005 . Counting cytokinesis proteins globally and locally in fission yeast . Science 310 : 306 \u2013 310 . DOI : https : / / doi . org / 10 . 1126 / science . 1113230 Xiong Y , Rangamani P , Fardin MA , Lipshtat A , Dubin - Thaler B , Rossier O , Sheetz MP , Iyengar R . 2010 . Mechanisms controlling cell size and shape during isotropic cell spreading . Biophysical Journal 98 : 2136 \u2013 2146 . DOI : https : / / doi . org / 10 . 1016 / j . bpj . 2010 . 01 . 059 , PMID : 20483321 Yarar D , Waterman - Storer CM , Schmid SL . 2005 . A dynamic actin cytoskeleton functions at multiple stages of clathrin - mediated endocytosis . Molecular Biology of the Cell 16 : 964 \u2013 975 . DOI : https : / / doi . org / 10 . 1091 / mbc . e04 - 09 - 0774 , PMID : 15601897 Yoshida A , Sakai N , Uekusa Y , Imaoka Y , Itagaki Y , Suzuki Y , Yoshimura SH . 2018 . Morphological changes of plasma membrane and protein assembly during clathrin - mediated endocytosis . PLOS Biology 16 : e2004786 . DOI : https : / / doi . org / 10 . 1371 / journal . pbio . 2004786 , PMID : 29723197 Akamatsu et al . eLife 2020 ; 9 : e49840 . DOI : https : / / doi . org / 10 . 7554 / eLife . 49840 40 of 40 Research article Cell Biology Physics of Living Systems",
    "okrut2015allosteric": "Allosteric N - WASP activation by an inter - SH3 domain linker in Nck Julia Okrut a , b , Sumit Prakash a , Qiong Wu c , Mark J . S . Kelly d , and Jack Taunton a , b , 1 a Department of Cellular and Molecular Pharmacology , University of California , San Francisco , CA 94158 ; b The Howard Hughes Medical Institute Summer Institute , Marine Biological Laboratory , Woods Hole , MA 02543 ; c Department of Biophysics , University of Texas Southwestern Medical Center , Dallas , TX 75390 ; and d Department of Pharmaceutical Chemistry , University of California , San Francisco , CA 94158 Edited by Brenda A . Schulman , St . Jude Children \u2019 s Research Hospital , Memphis , TN , and approved October 20 , 2015 ( received for review June 3 , 2015 ) Actin filament networks assemble on cellular membranes in response to signals that locally activate neural Wiskott \u2013 Aldrich - syndrome pro - tein ( N - WASP ) and the Arp2 / 3 complex . An inactive conformation of N - WASP is stabilized by intramolecular contacts between the GTPase binding domain ( GBD ) and the C helix of the verprolin - homology , connector - helix , acidic motif ( VCA ) segment . Multiple SH3 domain - containing adapter proteins can bind and possibly activate N - WASP , but it remains unclear how such binding events relieve autoinhibition to unmask the VCA segment and activate the Arp2 / 3 complex . Here , we have used purified components to reconstitute a signaling cascade driven by membrane - localized Src homology 3 ( SH3 ) adapters and N - WASP , resulting in the assembly of dynamic actin networks . Among six SH3 adapters tested , Nck was the most potent activator of N - WASP \u2013 driven actin assembly . We identify within Nck a previously unrecognized activation motif in a linker between the first two SH3 domains . This linker sequence , reminiscent of bacterial virulence fac - tors , directly engages the N - WASP GBD and competes with VCA bind - ing . Our results suggest that animals , like pathogenic bacteria , have evolved peptide motifs that allosterically activate N - WASP , leading to localized actin nucleation on cellular membranes . signal transduction | actin cytoskeleton | SH3 adapter | Nck | N - WASP A ctin polymerization provides the force that drives membrane protrusion during cell motility as well as the propulsion of endocytic vesicles and intracellular pathogens . Branched actin networks are assembled on the surface of cellular membranes , where actin monomers are incorporated into the membrane - apposed ends of growing actin filaments ( 1 \u2013 3 ) . The formation of new branches is initiated by the Arp2 / 3 complex , which requires allosteric activation by membrane - associated nucleation - promoting factors . Neural Wiskott \u2013 Aldrich - syndrome protein ( N - WASP ) is an essential nucleation - promoting factor that integrates and trans - duces membrane - localized signals to the Arp2 / 3 complex . N - WASP constitutes a regulatory hub whose localization and activation state govern the spatiotemporal dynamics of actin network formation . Under resting conditions , N - WASP exists in an autoinhibited conformation in the cytoplasm . Signaling from tyrosine kinases , GTPases , and acidic phospholipids cooperatively activates N - WASP on the membrane ( 4 \u2013 7 ) . Two principal themes have emerged to describe N - WASP regulatory mechanisms : allo - steric activation and oligomerization ( 8 ) . Allosteric activation dis - rupts intramolecular autoinhibitory contacts between the C helix and the GTPase binding domain ( GBD ) . These interactions maintain N - WASP in a closed conformation that sterically occludes its carboxyl - terminal verprolin - homology , connector - helix , acidic motif ( VCA ) segment ( 9 , 10 ) ( Fig . 3 A ) . The small GTPase Cdc42 is the archetypal allosteric N - WASP activator . Cdc42 binds directly to the GBD and releases the VCA segment ( 11 ) , which subsequently binds the Arp2 / 3 complex and promotes actin filament nucleation from the side of a preexisting actin filament . N - WASP oligomeri - zation or clustering , mediated by signaling adapter proteins and acidic phospholipids , facilitates simultaneous interaction of two N - WASP molecules with one Arp2 / 3 complex . Simulta - neous engagement of the constitutively inactive Arp2 / 3 complex by two VCA - type ligands is required for Arp2 / 3 - mediated actin nucleation ( 12 , 13 ) . Src homology 3 ( SH3 ) domain - containing adapter proteins have also been shown to activate N - WASP . Such signaling adapters often harbor multiple SH3 domains , each capable of binding a canonical polyproline motif ( 14 ) . Genetic , cell biological , and biochemical evidence supports a role for the SH2 / SH3 adapter protein Nck in the activation of N - WASP ( 15 \u2013 22 ) . Phosphorylated tyrosine residues on membrane receptors , such as the podocyte adhesion receptor nephrin and the vaccinia virus membrane protein A36R , localize Nck to the plasma membrane through its SH2 domain . Nck then directly binds N - WASP and the N - WASP \u2013 associated protein WIP , leading to localized actin polymerization ( 15 , 20 , 21 , 23 ) . Similar to Nck , Grb2 is an SH2 / SH3 adapter that binds and activates N - WASP ( 24 , 25 ) , acting in concert with Nck to promote actin - dependent vaccinia virus motility ( 26 ) . Other SH3 adapter proteins implicated in N - WASP activation include Crk - II ( 27 ) , cortactin ( 28 ) , Toca / CIP4 ( 29 , 30 ) , and Tks4 / 5 ( 31 ) ( Fig . 1 A ) . The mechanism by which SH3 adapters activate N - WASP is only partially understood . Although SH3 - mediated oligomerization has been shown to increase N - WASP activity , these experiments were performed with a constitutively active N - WASP mutant lacking the GBD \u2013 VCA autoinhibitory in - teraction ( 32 ) . Key unresolved questions are whether and how SH3 adapters counteract the autoinhibitory interactions that restrain access to N - WASP \u2019 s VCA segment . Here , we systematically compare the ability of distinct SH3 adapters to assemble actin and Arp2 / 3 networks while localized to a membrane surface . To this end , we have biochemically Significance Actin is a monomeric protein that can polymerize into branched networks . Actin polymerization acts like an engine to drive cell movement and is regulated by multiple interacting proteins on the cell membrane . To understand the molecular details of how cells transmit signals from the membrane to the actin polymer - ization engine , we reconstituted this process in a test tube using seven purified proteins and membrane - coated glass beads . Us - ing this \u201c biomimetic \u201d system , we discovered a sequence motif in the human protein Nck that activates a core component of the actin polymerization engine . This motif shares similarity with certain bacterial virulence factors that stimulate actin polymeri - zation in infected human cells , suggesting that similar activation mechanisms have evolved in humans and bacterial pathogens . Author contributions : J . O . and J . T . designed research ; J . O . , Q . W . , and M . J . S . K . performed research ; J . O . and S . P . contributed new reagents / analytic tools ; J . O . , S . P . , Q . W . , and M . J . S . K . analyzed data ; and J . O . and J . T . wrote the paper . The authors declare no conflict of interest . This article is a PNAS Direct Submission . Datadeposition : TheNMRchemicalshiftshavebeendepositedin theBiologicalMagnetic Resonance Data Bank ( accession no . 26643 ) . 1 To whom correspondence should be addressed . Email : jack . taunton @ ucsf . edu . Thisarticlecontainssupportinginformationonlineatwww . pnas . org / lookup / suppl / doi : 10 . 1073 / pnas . 1510876112 / - / DCSupplemental . E6436 \u2013 E6445 | PNAS | Published online November 9 , 2015 www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1510876112 D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 . reconstituted the N - WASP / actin signaling cascade from pure components using a membrane bilayer supported on silica micro - spheres . We find that membrane - localized actin assembly varies dramatically depending on the SH3 adapter , with only Nck showing robust actin assembly . Structure \u2013 function analysis revealed a pre - viously unidentified N - WASP activation motif embedded within a 45 - aa linker that connects the first two SH3 domains of Nck . This conserved inter - SH3 domain linker binds directly to the N - WASP GBD and in concert with the SH3 domains , potently stimulates actin network assembly on membranes . Nck thus uses both allo - steric and oligomerization - based mechanisms to activate N - WASP . Results Polarized Actin Network Assembly Promoted by SH3 Adapters and N - WASP . We initially compared six full - length SH3 adapter pro - teins ( Fig . 1 A ) using a well - established reconstituted motility system that contains purified N - WASP ( amino acids 151 \u2013 501 ; lacking the N - terminal EVH1 domain ) , Arp2 / 3 complex , actin , and the actin Nck1 1 % NTAD 2 . 5 % NTAD 5 % NTAD Grb2 Crk - II cortactin Phase Alexa - actin Phase Alexa - actin Phase Alexa - actin T ks4 CIP4 CIP4 BAR SH3 SH3A SH3B SH3D SH3C PH cortactin Arp2 / 3 Actin Binding Repeats SH3 Tks4 SH3A SH3B SH2 Grb2 SH3A SH3B SH3C SH2 Nck1 SH3A SH3B SH2 Crk - II N ck 1 G r b2 C r k - II c o r t a c t i n C I P 4 T ks 4 0 . 0 5 . 0 10 . 0 15 . 0 20 . 0 25 . 0 T a il L e ng t h [ \u00b5 m ] * * NTAD 1 % 2 . 5 % 5 % A C B D E F NTAD + actin , Arp2 / 3 , cofilin , profilin , capping protein lipid - coated glass bead S H 3 a d a p t e r N - W A SP S H 3 a d a p t e r s N ck 1 G r b2 C r k - II c o r t a c t i n C I P 4 T ks 4 0 10000 20000 30000 40000 50000 60000 70000 A l exa568 - S H 3 a d a p t e r [ AU ] NTAD 1 % 2 . 5 % 5 % NTAD : 1 % 2 . 5 % 5 % N ck 1 G r b2 C r k - II c o r t a c t i n C I P 4 T ks 4 0 500 1000 1500 2000 2500 A l exa488 - ac t i n [ AU ] * * * Fig . 1 . Membrane - associated actin networks assembled by SH3 adapters and N - WASP . ( A ) Domain organization of SH3 adapter proteins used in this study . ( B ) Experimental strategy for localization of His - tagged SH3 adapters using NTAD - doped membranes supported on silica microspheres . ( C ) Phase contrast and fluorescence images of lipid - coated beads ( NTAD density : 1 % , 2 . 5 % , or 5 % ) incubated with the indicated SH3 adapter ( 250 nM ) , N - WASP ( 50 nM ) , Alexa488 actin , and actin regulatory components . After 15 min , reactions were fixed with glutaraldehyde . ( Scale bar : 5 \u03bc m . ) ( D ) Integrated fluorescence intensity of Alexa488 actin tails . ( E ) Actin tail length . ( F ) Integrated fluorescence intensity of bead - localized Alexa568 SH3 adapters ( 250 nM ) . ( Mean \u00b1 SD ; n > 20 . ) Asterisks indicate a significant difference between Nck1 and Grb2 at 1 % and 5 % NTAD : * * P < 0 . 01 ; * * * P < 0 . 001 , respectively . Okrut et al . PNAS | Published online November 9 , 2015 | E6437 B I O C H E M I S T RY P N A S P L U S D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 . regulatory proteins profilin , cofilin , and capping protein ( 33 , 34 ) . As described in the Introduction , all six SH3 adapters have been implicated in cellular N - WASP activation and have been shown to bind the N - WASP proline - rich region . Because N - WASP \u2013 medi - ated actin assembly occurs on the cytosolic surface of cellular membranes , we used lipid bilayers supported on silica microspheres ( 7 , 35 , 36 ) . Supported lipid bilayers were supplemented with a nickel - chelating lipid [ 1 , 2 - dioleoyl - sn - glycero - 3 - [ ( N - ( 5 - amino - 1 - carboxypentyl ) iminodiacetic acid ) succinyl ] ( NTAD ) ] to localize His 6 - tagged versions of the SH3 adapter proteins ( Fig . 1 B ) . To test a range of SH3 adapter densities , we varied the molar fraction of NTAD in the lipid bilayers ( 1 % , 2 . 5 % , and 5 % NTAD , mol % ) . Lipid - coated beads were incubated in a solution containing the purified full - length SH3 adapter ( 250 nM ) , Alexa488 - labeled actin , N - WASP , and the actin regulatory components used at concentrations that were previously shown to support bead motility ( 7 , 33 ) . After 15 min , reactions were fixed with glutar - aldehyde , and the bead - associated actin structures were analyzed by wide - field epifluorescence imaging . Two measurements were used to quantify bead - associated actin : integrated fluorescence intensity of Alexa488 actin and actin tail length . Nck was the only SH3 adapter that induced polarized actin tails on > 95 % of the beads at the lowest NTAD density tested ( Fig . 1 C and Fig . S1 ) . Image analysis revealed that , in reactions driven by Nck , bead - associated actin intensity and tail length were already maximal at the lowest NTAD density ( Fig . 1 D and E ) . Moreover , Nck - coated beads displayed robust actin - dependent motility when imaged live in a flow chamber containing N - WASP , Arp2 / 3 complex , and actin regulatory components ( Movie S1 ) . By contrast , Grb2 showed substantially weaker ac - tivity . Grb2 - induced tails were not detected at 1 % NTAD , and only diffuse actin \u201c clouds \u201d or short tails were observed at 2 . 5 % and 5 % NTAD , respectively ( Fig . 1 C , row 2 ) . The Grb2 - induced tails at 5 % NTAD density contained one - half as much actin as the Nck - induced tails obtained at 1 % NTAD and were signifi - cantly shorter ( Fig . 1 D and E ) . A thin shell of bead - associated actin was visible in reactions containing cortactin ( Fig . 1 C and D ) , consistent with its ability to directly activate Arp2 / 3 and bind actin filaments ( 37 ) . However , cortactin did not promote the assembly of polarized actin tails . Reactions containing Crk - II , CIP4 , or Tks4 were devoid of visible actin structures . To ensure that the observed effects in the actin tail assembly assay were not caused simply by differential recruitment of the SH3 adapters , we used live imaging to quantify the fluorescence intensity of Alexa568 - labeled adapters on the bead surface . Despite the common recruitment mechanism through hexahistidine tag and NTAD lipids , the adapters localized to different extents . At an NTAD density of 1 % , the fluorescence intensities measured for CIP4 and Tks4 were significantly higher than for Nck , Grb2 , Crk - II , and cortactin , all of which were used at a concentration of 250 nM ( Fig . 1 F , blue bars ) . We speculate that these differences arise from intrinsic lipid binding and / or a propensity to oligomerize on membrane surfaces . Most importantly , the Nck density at 1 % NTAD was significantly lower than the density of the other five adapters at 5 % NTAD . Despite achieving higher membrane den - sities than Nck , the other SH3 adapters were significantly less active or inactive . These results show that , of six SH3 adapters tested , Nck has by far the strongest ability to assemble polarized actin networks on the surface of supported membranes . A potential explanation for Nck \u2019 s superior activity is that it re - cruits more N - WASP to the membrane compared with the other SH3 adapters . To test this , we measured the density of Alexa568 - labeled N - WASP recruited to lipid beads in the presence of the His - tagged adapters . In initial experiments , Nck and Grb2 recruited similar levels of N - WASP , but the other SH3 adapters recruited substantially lower amounts . To compensate for these differences in apparent affinity , we increased the N - WASP concentration to 200 nM . Additionally , we increased the NTAD density to 5 % for Crk - II , cortactin , CIP4 , and Tks4 , while maintaining 1 % NTAD for Nck and Grb2 . Under these conditions , N - WASP levels recruited by all of the SH3 adapters were within approximately twofold of each other ( Fig . 2 A ) . Despite the similar N - WASP density observed with all six SH3 adapters , Nck showed the strongest actin assembly activity ( Fig . 2 B ) . Tails assembled by Grb2 were approximately threefold shorter and incorporated approximately twofold less actin compared with Nck ( Fig . 2 C ) . Similar to our previous results ( Fig . 1 C and D ) , Crk - II , cortactin , CIP4 , and Tks4 did not assemble any detectable actin structures . These data indicate that the increased ability of Nck to assemble membrane - associated actin networks cannot be explained solely by enhanced N - WASP recruitment . We therefore considered the possibility that , among the SH3 adapters tested , Nck might deploy a unique N - WASP activation mechanism . Discovery of an Inter - SH3 Domain N - WASP Activation Motif . Nck is recruited to cellular membranes through its SH2 domain , which binds phosphorylated tyrosine residues on integral membrane receptors or membrane - associated adapter proteins . Nephrin is an Nck receptor on the plasma membrane of podocytes . Three phosphorylated tyrosine residues on nephrin bind the SH2 domain of Nck , and this interaction is essential for regulating podocyte actin dynamics in an N - WASP \u2013 dependent manner ( 20 , 21 ) . To explore the mechanism of Nck - mediated N - WASP activation on membranes , we used a His - tagged , triphosphorylated nephrin peptide ( His 8 - pY 3 - nephrin ) , which recruits untagged Nck to the surface of NTAD - containing membranes ( 38 ) . Lipid - coated beads with 1 % NTAD were preincubated with His 8 - pY 3 - nephrin and subsequently added to a solution containing untagged full - length Nck , N - WASP , and the actin regulatory components described above ( Fig . 3 A ) . Similar to His - tagged Nck recruited through NTAD lipids ( Fig . 1 C ) , untagged Nck recruited through pY 3 - nephrin promoted the assembly of actin tails on > 95 % of the lipid - coated beads ( Fig . 3 C and Fig . S2 ) . Actin assembly promoted by untagged Nck was strictly dependent on the addition of His 8 - pY 3 - nephrin , and Nck mutants lacking the SH2 domain were inactive in this system . Using the pY 3 - nephrin system , we performed a deletion analysis of Nck ( Fig . 3 B ) . Whereas deletion of the amino - terminal SH3A domain had no effect , additional deletion of the inter - SH3AB linker led to a drastic reduction in actin tails ( Fig . 3 C ) . The few residual actin structures formed by this deletion construct con - tained threefold lower actin fluorescence , and the average tail length was reduced by 85 % ( Fig . 3 D and E ) . To exclude the possibility that decreased actin assembly was a trivial result of re - duced membrane recruitment , we measured surface densities of the three Nck constructs in a separate experiment . Both deletion mutants were recruited similarly to one another and to a slightly higher extent than full - length Nck ( Fig . 3 F ) . Hence , the Nck mutant lacking the inter - SH3AB linker is recruited to pY 3 - nephrin but is specifically defective in promoting actin assembly ( Fig . 3 F ) . These experiments identify for the first time , to our knowledge , a 45 - aa linker motif between the first two SH3 domains of Nck that strongly potentiates N - WASP \u2013 dependent actin assembly on membranes . Direct Binding of the Inter - SH3 Linker to the N - WASP GBD . We hy - pothesized that the inter - SH3AB linker ( Nck1 amino acids 61 \u2013 105 ) might interact directly with N - WASP . Within this linker is a 9 - aa motif ( amino acids 67 \u2013 75 ) that is highly conserved across species ( Fig . 4 A and B ) . Moreover , the spacing of the hydrophobic resi - dues suggested the possibility of forming an amphipathic helix ( Fig . 4 B and Fig . S3 ) . Several amphipathic helical peptides have been shown to bind N - WASP through its GBD . In the autoinhibited state , the N - WASP GBD binds intramolecularly to the amphipathic C helix in the carboxyl - terminal VCA region ( 10 ) ( Fig . 3 A ) . Am - phipathic helices can also activate N - WASP : helical motifs within the Escherichia coli virulence factors , EspF U and EspF , bind the GBD in competition with the inhibitory C helix ( 39 \u2013 41 ) . To our E6438 | www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1510876112 Okrut et al . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 . knowledge , such an activation mechanism has only been docu - mented with bacterial pathogens and has not been found with endogenous eukaryotic activators . To test whether the inter - SH3AB linker can directly interact with the N - WASP GBD , we first performed GST pull - down ex - periments with purified proteins . Full - length Nck bound to gluta - thione beads loaded with N - WASP GST - GBD but not to beads loaded with a structurally and functionally related GST - GBD de - rived from the PAK1 kinase ( Fig . 4 C ) . These results indicate a specific , direct , and previously undocumented interaction between Nck and the N - WASP GBD . Deletion of the SH2 and the SH3A domains from Nck did not affect the interaction ( Fig . 4 D ) . How - ever , additional deletion of only 18 amino acids from the inter - SH3AB linker , including the putative amphipathic helix , com - pletely abrogated binding to the GBD along with deletion of the entire linker ( Fig . 4 D ) . To further assess the importance of the inter - SH3AB linker , we used a construct lacking the SH2 domain ( Nck amino acids 1 \u2013 270 ) . Replacing either the entire linker ( amino acids 67 \u2013 105 ) or only the hydrophobic motif ( amino acids 67 \u2013 77 ) with a Gly - Ser linker of equal length abolished the interaction ( Fig . 4 E ) . Finally , GBD binding was completely prevented by four ala - nine mutations at hydrophobic positions within the putative am - phipathic helix [ Ile67 / Val68 / Leu71 / Leu75 ( 4A ) ] . These results strongly suggest that the inter - SH3AB linker , implicated above in promoting N - WASP \u2013 dependent actin assembly ( Fig . 3 C \u2013 E ) , mediates direct interactions between Nck and the N - WASP GBD . We next evaluated actin comet tail assembly by full - length Nck containing the four Ala mutations in the inter - SH3AB linker . In the presence of lipid - coated beads with a high NTAD density ( 1 % ) , 4A Nck formed tails similar to the WT ( Fig . S4 ) . However , with a lower density of 0 . 25 % NTAD , 4A Nck was defective , producing two - to threefold less actin than WT Nck ( P < 0 . 001 ) ( Fig . S4 A \u2013 C ) . We hypothesized that , at higher NTAD densities , polyvalent SH3 - mediated interactions could potentially compen - sate for the defective linker in 4A Nck . To test this , we introduced the four Ala mutations in a construct lacking the SH3A domain B CIP4 cortactin Tks4 Nck1 Crk - II Phase Alexa - actin A C Grb2 0 . 0 10 . 0 20 . 0 30 . 0 N ck 1 , 1 % N T A D G r b2 , 1 % N T A D C r k - II , 5 % N T A D c o r t a c t i n , 5 % N T A D C I P 4 , 5 % N T A D T ks 4 , 5 % N T A D T a il L e ng t h [ \u03bc m ] * * * 0 400 800 1200 1600 2000 2400 A l exa488 - ac t i n [ AU ] * * * NTAD 1 % 5 % N ck 1 , 1 % N T A D G r b2 , 1 % N T A D C r k - II , 5 % N T A D c o r t a c t i n , 5 % N T A D C I P 4 , 5 % N T A D T ks 4 , 5 % N T A D 0 1000 2000 3000 4000 5000 6000 7000 A l exa568 - N - W A SP [ AU ] Fig . 2 . Nck is the most effective SH3 adapter in promoting N - WASP \u2013 dependent actin assembly . ( A ) Integrated fluorescence intensity and representative images of bead - localized Alexa568 N - WASP ( 200 nM ) incubated with NTAD - doped lipid - coated beads and the indicated His - SH3 adapters ( 250 nM ) . To compensate for differences in N - WASP recruitment efficiency , the NTAD density was increased to 5 % for Crk - II , cortactin , CIP4 , and Tks4 , whereas 1 % NTAD was used for Nck1 and Grb2 . ( B ) Representative phase contrast and fluorescence images of actin assembly reactions under conditions reported in A . ( Scale bars : 5 \u03bc m . ) ( C ) Integrated fluorescence intensity of Alexa488 actin tails and tail lengths ( mean \u00b1 SD ; n > 20 ) . Asterisks indicate significant differences between Nck and Grb2 : * * * P < 0 . 001 . Okrut et al . PNAS | Published online November 9 , 2015 | E6439 B I O C H E M I S T RY P N A S P L U S D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 . ( Nck amino acids 61 \u2013 377 ) . Here , the 4A mutant was strongly compromised in its ability to assemble actin tails , displaying a threefold reduction in polymerized actin and a fivefold decrease in tail length , even at 1 % NTAD ( P < 0 . 001 ) ( Fig . 4 F and Fig . S5 A and B ) . Collectively , these data establish a role for the amphi - pathic motif in actin comet tail formation , which becomes essen - tial under conditions of low pY 3 - nephrin density or when the first SH3 domain is removed . Structural studies of the WASP GBD ( 68 % sequence identity to N - WASP GBD ) have revealed a high degree of conforma - tional plasticity . In the absence of a binding partner , the WASP GBD is partially folded or unstructured . Binding of the auto - inhibitory C helix ( 10 ) or an activating EspF U - derived peptide ( 40 ) causes the GBD to adopt a mostly helical globular fold . To test whether the Nck linker behaves similarly , we recorded heter - onuclear single - quantum coherence ( HSQC ) spectra of 15 N - labeled N - WASP GBD in the presence and absence of the unlabeled Nck linker peptide ( amino acids 61 \u2013 106 ) . The GBD spectrum in the presence of the Nck linker peptide revealed greater peak disper - sion and more uniform peak intensities and line widths compared with the GBD alone ( Fig . 5 A ) , consistent with a more compact , folded domain . Based on chemical shift differences from NMR spectra recorded with increasing concentrations of the Nck pep - tide , we determined the equilibrium K d to be 33 \u00b1 8 \u03bc M ( Fig . 5 B and Fig . S6 ) . We also recorded the N - WASP GBD spectrum with the VCA peptide and observed similar chemical shifts to those recorded in the presence of the Nck linker peptide : 80 % of the peaks show a weighted chemical shift difference of less than 0 . 1 ppm ( Fig . S7 ) . To characterize the GBD \u2013 Nck linker complex in greater detail , we used standard triple - resonance experiments to assign the back - bone ( 86 % completion ) of a construct comprising the N - WASP GBD covalently linked to the inter - SH3AB segment from Nck ( Fig . S8 ) ( Biological Magnetic Resonance Data Bank accession no . 26643 ) . Analysis of the HN , N , CO , C \u03b1 , and C \u03b2 chemical shifts using Talos + ( 42 ) indicates secondary structure similar to the WASP GBD in complex with the autoinhibitory C helix ( Fig . 5 C and Fig . S9 A ) ( 10 ) . Importantly , the Talos + analysis predicts a helical conformation for the Nck hydrophobic motif ( amino acids 66 \u2013 75 ) . Moreover , the difference between the secondary C \u03b1 and C \u03b2 chemical shifts for residues within the hydrophobic motif is positive ( \u0394\u03b4 [ C \u03b1 \u2212 C \u03b2 ] > 0 ) ( Fig . 5 D and Fig . S9 B ) , consistent with a helical conformation ( 43 ) . Collectively , these data suggest that the structure of the GBD \u2013 Nck linker complex is similar to that of the GBD \u2013 VCA complex . A prediction of this model is that binding of the C helix and the Nck linker to the GBD are mutually exclusive . Consistent with competitive binding , addition of in - creasing concentrations of VCA abrogated binding of full - length Nck to the GBD ( Fig . 5 E and F ) . Based on these results , we conclude that Nck binds N - WASP through not only its SH3 do - mains but also , its inter - SH3AB linker . These interactions may act in a cooperative manner to destabilize autoinhibitory contacts between the GBD and VCA regions , resulting in allosteric acti - vation of N - WASP ( Fig . 6 ) . Phase actin SH3A SH3B SH3C SH2 SH3B SH3C SH2 SH3B SH3C SH2 Nck1 full - length Nck1 aa 61 - 377 Nck1 aa 106 - 377 A C B Alexa488 D Nck fl . Nck aa 61 - 377 Nck aa 106 - 377 G B D V C A P RR N - W A SP + actin , Arp2 / 3 , cofilin , profilin , capping protein lipid - coated glass bead p Y 3 - n e ph r i n pY N ck NTAD F E SH2 SH3 SH3 SH3 0 2000 4000 6000 8000 N ck f l . N ck aa 61 - 377 N ck aa 106 - 377 A l e x a568 - N ck [ A U ] 0 . 0 4 . 0 8 . 0 12 . 0 16 . 0 20 . 0 24 . 0 28 . 0 N ck f l . N c k aa 61 - 377 N ck aa 106 - 377 T a il Leng t h [ m ] * * * 0 400 800 1200 1600 2000 2400 N ck f l . N c k aa 61 - 377 N c k aa 106 - 377 A l e x a488 - a c t i n [ A U ] * * * Fig . 3 . An inter - SH3 linker in Nck promotes actin network assembly by N - WASP . ( A ) Experimental strategy for localization of Nck to membranes con - taining pY 3 - nephrin . ( B ) Nck constructs used in the deletion analysis . ( C ) Representative images of actin tails formed by Nck deletion constructs . Motility reactions contained 200 nM N - WASP and 100 nM Nck . ( Scale bar : 5 \u03bc m . ) ( D ) Integrated fluorescence intensity of Alexa488 actin tails . ( E ) Actin tail length . ( F ) Integrated fluorescence intensity of bead - localized Alexa568 Nck constructs . ( Mean \u00b1 SD ; n > 20 ; * * * P < 0 . 001 . ) fl . , Full length ; PRR , proline - rich region . E6440 | www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1510876112 Okrut et al . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 . Discussion In response to membrane - localized signals , N - WASP binds and activates the Arp2 / 3 complex , promoting the assembly of branched actin networks and the physical movement and deformation of cellular membranes . In addition to the canonical activators , Cdc42 and PIP2 , membrane - localized SH3 adapter proteins have been implicated in N - WASP activation . It has remained unclear how SH3 adapters disrupt the intramolecular interactions that main - tain N - WASP in an autoinhibited state . Using purified components , we reconstituted N - WASP \u2013 dependent actin assembly on a lipid bilayer supported on silica microspheres , allowing side by side comparison of multiple SH3 adapters . We find that the SH2 / SH3 adapter Nck activates N - WASP more po - tently than all of the other SH3 adapters , including Grb2 , Crk - II , cortactin , CIP4 , and Tks4 . Although previous studies have focused on SH3 / polyproline interactions to explain N - WASP activation by Nck , we identify here an activation motif in the linker between the first two SH3 domains ( amino acids 67 \u2013 75 ) . This motif , which had no assigned function before this work , binds directly to the N - WASP GBD and competes with the autoinhibitory C helix in the VCA region , leading to allosteric activation of N - WASP . The inter - SH3AB activation motif is conserved among Nck orthologs , suggesting that the N - WASP GBD interaction and activation mechanism is also evolutionarily conserved . N - WASP activation by the inter - SH3 linker is reminiscent of the mechanism used by the bacterial virulence factors , EspF U and EspF , to nu - cleate actin assembly in host cells . The activation motif in EspF U and EspF has the same pattern of hydrophobic residues as the Nck motif ( Fig . 4 B ) ( \u03a6 - \u03a6 - x - x - \u03a6 - x - x - x - \u03a6 ; \u03a6 is any hydrophobic residue and x is any residue ) ( 39 \u2013 41 ) . Structural studies by mul - tidimensional NMR have shown that the EspF U peptide adopts a helical conformation when bound to the WASP GBD ( 40 ) . Likewise , our analysis of secondary chemical shifts strongly sug - gests that the Nck hydrophobic motif also adopts a helical con - formation with the side chains of Ile67 , Val68 , Leu71 , and Leu75 lying on one face of the helix ( Fig . S3 ) . Similar to linker segments in many other proteins ( 44 ) , the Nck inter - SH3AB linker is multifunctional . In addition to acti - vating N - WASP through GBD binding , the inter - SH3AB linker has been shown to bind intramolecularly to the SH3B domain ( 45 ) and mediate Nck homo - oligomerization and phase separa - tion ( 46 ) . The N - WASP C - helix motif is similarly multifunc - tional , binding in a mutually exclusive manner to G actin and the Arp2 / 3 complex ( 47 ) in addition to mediating autoinhibition by intramolecular GBD interactions . Our results , combined with previous studies , suggest that Nck can activate N - WASP by two complementary mechanisms . First , the Nck inter - SH3 linker can bind directly to the N - WASP GBD , NCK1 , HUMAN RKASIVKNLKDTLGIGK NCK1 , FISH RKASIVKNLKDTLGIGK NCK2 , HUMAN RKGSLVKNLKDTLGLGK NCK , CHICKEN KKGSLVKNLKDTLGLGK NCK , FLY - KPSLFDSIKKKVKKGS NCK , WORM - KESIVDKAKGTIKGLA N - WASP , C - helix - TSGIVGALMEVMQ - - - EspFu , helix - LPDVAQRLMQMLA - - - EspF , helix - LPPIAQALKDMLA - - - consensus \u03a6\u03a6xx\u03a6xxx\u03a6 SH3A SH3B SH3C SH2 A aa 2 - 60 115 - 165 190 - 252 282 - 377 61 71 81 91 101 SARKASIVKN LKDTLGIGKV KRKPSVPDSA SPADDSFVDP GERLY N - WASP activation motif Nck1 PAK1 GST - GBD N - WASP GST - GBD B S B S B 10075 50 37 25 Nck GST - GBD GST - GBD w t G S aa 67 - 77 4 A G S aa 67 - 105 Nck aa 1 - 270 D aa 61 - 270 aa 79 - 270 aa 106 - 270 C B S B S B S N - WASP GST - GBD 37 25 E B S B S B S N - WASP GST - GBD B S Nckaa 61 - 377 wt Nckaa 61 - 377 4A Phase Alexa - actin F Fig . 4 . The inter - SH3AB linker contains a conserved motif and binds directly to the N - WASP GBD . ( A ) Nck domain structure and sequence of the inter - SH3AB linker . ( B ) Sequence alignment comparing the Nck inter - SH3 linker from various organisms and previously known N - WASP GBD ligands . ( C ) Coomassie - stained gel of GST pull - down samples . Immobilized GST - GBD from N - WASP or PAK1 was used to pull down full - length Nck . ( D ) N - WASP GST - GBD was used to pull down the indicated Nck deletion constructs . ( E ) N - WASP GST - GBD was used to pull down Nck amino acids 1 \u2013 270 WT and mutants . Samples comprising 1 . 25 % of the unbound supernatant ( S ) and 12 . 5 % of the bound fraction ( B ) were separated by SDS / PAGE and stained with Coomassie . GS aa 67 \u2013 105 , substitution of the entire SH3AB linker with a Gly - Ser linker ; GS aa 67 \u2013 77 , substitution of the hydrophobic motif with a Gly - Ser linker . ( F ) Representative images of actin tails formed by Nck amino acids 61 \u2013 377 WT and 4A mutant . Motility reactions contained 200 nM N - WASP and 100 nM Nck . ( Scale bar : 5 \u03bc m . ) Okrut et al . PNAS | Published online November 9 , 2015 | E6441 B I O C H E M I S T RY P N A S P L U S D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 . thereby releasing the critical VCA segment . We note that the affinity of this interaction ( K d = 33 \u00b1 8 \u03bc M ) is probably insufficient to activate autoinhibited N - WASP on its own ; cooperative binding of the linker , along with one or more SH3 domains , is likely re - quired to disrupt N - WASP \u2019 s autoinhibitory interactions . Second , Nck can promote N - WASP oligomerization . When N - WASP binds to a poly - SH3 adapter that is itself oligomerized by a third multivalent adapter , phase separation can occur , enforcing a high density of N - WASP molecules ( 38 , 48 ) . On recruitment to the membrane by pY 3 - nephrin / Nck complexes , N - WASP molecules can presumably equilibrate between auto - inhibited and activated states . At high local densities of pY 3 - nephrin and Nck , the density of activated N - WASP molecules bound to the Arp2 / 3 complex may be sufficient to nucleate actin assembly without GBD engagement by the Nck hydrophobic motif . This model may explain our observation that , at high NTAD densities , the Nck 4A mutant ( with a defective hydrophobic motif ) assembles actin comet tails similar to the WT ( Fig . S4 ) . By contrast , a functional hydro - phobic motif is essential at low NTAD densities or when the SH3A domain is deleted ( Fig . 4 F and Figs . S4 and S5 ) . In the cellular context , a similar dependency on the Nck hydrophobic motif may occur under conditions where the local density of Nck is below the critical concentration required for phase separation ( for example , when the balance of kinase and phosphatase activity is shifted to produce low phosphotyrosine levels on membrane - associated receptors that recruit Nck ) . Consistent with previous observations ( 48 ) , Nck and N - WASP phase separate in our hands when mixed together at micromolar concentrations . However , under the conditions of our bead motility assays ( 100 \u2013 250 nM Nck and 50 \u2013 200 nM N - WASP ) , we do not observe phase separation in solution . We also have not detected macroscopic phase separation on the lipid - coated beads in contrast to a recent study , in which macroscopic , phase - sep - arated domains containing pY 3 - nephrin , Nck , and N - WASP were observed on planar lipid bilayers ( 38 ) . Nevertheless , it is likely that membrane - associated Nck and N - WASP can form higher - order oligomers in our system , especially at high densities of NTAD ( and pY 3 - nephrin ) ; such oligomeric assemblies may be important for promoting localized actin nucleation ( Fig . 6 ) . The quantitative relationships between Nck / N - WASP signal - ing inputs on the one hand \u2014 tyrosine kinase / phosphatase activ - ities , phosphotyrosine valency , and phase separation \u2014 and actin 0 0 . 05 0 . 1 0 . 15 0 . 2 0 100 200 300 400 500 600 \u2206 \u03b4 ( HN , N ) Nck linker [ \u03bc M ] E Nck fl VCA 0 2 . 5 5 10 25 50 GST - GBD 50 37 25 20 VCA [ \u03bc M ] F C D - 2 - 1 0 1 2 3 4 5 6 7 8 A R K A S I V K N L K D T L G I G K V \u2206 \u03b4 ( C \u03b1 - C \u03b2 ) N ck aa 62 - 80 I V L L WASP - C - helix SGFKHVSHVGWDPQNGFDVNNLDPDLRSLFSRAGI N - WASP - Nck SNFQHIGHVGWDPNTGFDLNNLDPELKNLFDMCGI Talos bb hhhhhhhhhh WASP - C - helix SEAQLTDAETSKLIYDFIEDQGGLEAVRQEMRRQ N - WASP - Nck SEAQLKDRETSKVIYDFIEKTGGVEAVKNELRRQ Talos hhhh hhhhhhhhhhhhh hhhhhhhhhh WASP - C - helix GGSGGSQSSEGLVGALMHVMQKRSRAIH N - WASP - Nck GGSGGSARKASIVKNLKDTLGIGKVKRK Talos hh hhhhhhhhhh 0 0 . 2 0 . 4 0 . 6 0 . 8 1 1 . 2 0 10 20 30 40 50 60 R e l a t i ve B a nd I n t e n s i t y VCA concentration [ \u03bc M ] Nck VCA 7 . 9 8 . 0 8 . 1 114 . . 0 114 . . 5 Nck linkerequiv 00 . 2 0 . 4 0 . 7 11 . 5 23458 \u03b4 1 H [ ppm ] \u03b4 15 N [ pp m ] \u03b4 15 N [ pp m ] GBD alone GBD + Nck aa 61 - 106 \u03b4 1 H [ ppm ] 6 . 5 7 . 0 7 . 5 8 . 0 8 . 5 9 . 0 9 . 5 110 115 120 125 130 A B Fig . 5 . NMR analysis reveals direct binding of the Nck inter - SH3 linker and the N - WASP GBD . ( A ) [ 15 N , 1 H ] - heteronuclear single - quantum coherence ( HSQC ) NMR spectra of 15 N - labeled N - WASP GBD acquired in the presence and absence of unlabeled Nck linker peptide ( amino acids 61 \u2013 106 ) . ( B ) NMR titration of 15 N - labeled GBD ( 65 \u03bc M ) with the Nck linker peptide . Chemical shift changes , along with the corresponding curve fit , are shown for a representative peak from the [ 15 N , 1 H ] - HSQC spectrum . A binding K d of 33 \u00b1 8 \u03bc M was determined by fitting titration curves to a total of 17 peaks . ( C ) Sequence alignment of a WASP - GBD C - helix construct ( 10 ) and the N - WASP GBD Nck linker fusion protein . The secondary structure of the WASP \u2013 C helix complex ( 1EJ5 . pdb ) is in - dicated . ( D ) Secondary structure prediction of the N - WASP Nck fusion protein by Talos + is shown . ( E ) Coomassie - stained gel from a competition pull - down experiment ( bound fraction ) . N - WASP GST - GBD was used to pull down Nck in the presence of increasing N - WASP VCA . ( F ) Integrated band intensities from the Coomassie - stained gel showing Nck and VCA bound to GST - GBD . E6442 | www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1510876112 Okrut et al . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 . network assembly and force production on the other hand re - main to be elucidated . We anticipate that reconstituted signaling and actin assembly systems of even higher complexity will be essential for obtaining a complete mechanistic description of emergent phenomena at the intersection of signaling and cyto - skeletal dynamics . Materials and Methods ProteinExpressionandPurification . Expression plasmids , constructedfrompET vectors , were transformed into E . coli BL21 ( DE3 ) CodonPlus RIL ( construct details are in Table S1 ) . Cultures ( 2 \u2013 4 L ) were grown in LB at 18 \u00b0C overnight after induction with 0 . 5 mM isopropyl \u03b2 - D - 1 - thiogalactopyranoside . For isotope labeling , M9 minimal medium was supplemented with 13 C - glucose and / or 15 NH 4 Cl , and bacteria were induced for 6 h at 30 \u00b0C . Bacterial cell pellets were lysed using a microfluidizer ( Microfluidics ) in 20 mM Hepes , pH 7 . 5 , 500 mM NaCl , 5 \u2013 10 % glycerol , and 4 mM \u03b2 - mercapto - ethanol with protease inhibitor ( 4 mM Benzamidine or 1 mM di - isopropyl - fluorophosphate ) . For Nck and Grb2 , 30 mM arginine was added to the lysis buffer . Lysates were cleared ( 24 , 000 \u00d7 g for 40 min at 4 \u00b0C ) and incubated in batch with Ni - nitriloacetic acid - agarose ( Qiagen ) . The beads were washed with ATP buffer ( lysis buffer supplemented with 10 mM MgCl 2 , 2 mM ATP , 30 mM KCl ) and a high - salt buffer ( lysis buffer with 1 M NaCl ) , and the protein was eluted with 400 mM imidazole . Poly - His tags were cleaved from Nck , N - WASP ( amino acids 151 \u2013 501 ) , cofilin , and profilin by incubation with tobacco etch virus ( TEV ) protease overnight . Tks4 and N - WASP were further purified with a heparin column ( HiTrap ; GE Healthcare ) , eluting with 50 \u2013 1500 mM NaCl in storage buffer . N - WASP was stored in 20 mM Hepes , pH 7 . 5 , 300 mM NaCl , and 1 mM tris ( 2 - carboxyethyl ) phosphine ( TCEP ) . Tks4 was stored in 20 mM MES , pH 6 . 5 , 250 mMNaCl , and 5 % glycerol . Allother proteinswere furtherpurified by size exclusion chromatography on S75 or S200 columns ( GE Healthcare ) equilibrated with storage buffer . The following storage buffers were used : Grb2 and Nck constructs ( 20 mM Hepes , pH 7 . 5 , 500 mM NaCl , 0 . 5 mM TCEP , 10 % glycerol ) , CIP4 ( 20 mM Hepes , pH 7 . 5 , 150 mM NaCl ) , cortactin ( 50 mM Tris , pH 7 . 4 , 100 mM NaCl , 10 % glycerol ) , cofilin ( 20 mM Tris , pH 8 , 50 mM NaCl , 1 mM DTT ) , and profilin ( 20 mM Tris , pH 8 , 1 mM DTT , 1 mM EDTA ) . GST - tagged N - WASP GBD and covalently linked GBD \u2013 Nck linker complex were purified on glutathione - Sepharose 4B ( GE Healthcare ) . GST fusion protein was eluted with 30 mM glutathione ( GSH ) followed by dialysis in 20 mM Hepes , pH 7 . 5 , 100 mM NaCl , and 1 mM DTT . For NMR experiments , the GST tag was cleaved by incubation with TEV protease followed by size exclusion chromatography . To purify N - WASP VCA and Nck inter - SH3AB linker , cell pellets were lysed by incubation with 3 M guanidinium HCl overnight at room temperature ( RT ) . The cleared lysate was applied to a Ni - chelating column ( GE Healthcare ) and washed with 1 M NaCl . Proteins were eluted with 0 \u2013 400 mM imidazole followed by dialysis in storage buffer ( Nck linker : 20 mM Hepes , pH 7 . 5 , 200 mM NaCl ; VCA : 20 mM Hepes , pH 7 . 5 , 200 mM NaCl , 1 mM TCEP ) . The following proteins were purified as previously described : actin ( 49 ) , capping protein ( 50 ) , and Arp2 / 3 complex ( 51 ) , except that Arp2 / 3 complex was purified from frozen bovine thymus ( Pel - Freez Biologicals ) . Fluorescent Labeling . Unless stated otherwise , proteins ( 5 \u2013 50 \u03bc M ) were la - beled with 1 M eq Alexa568 maleimide ( Invitrogen ) on ice for 1 h . Crk - II was labeled overnight at 4 \u00b0C with 10 eq Alexa568 NHS ( Invitrogen ) . Unreacted dye was removed by gel filtration . F - actin was labeled for 6 h on ice with 2 eq Alexa488 - TFP ( Invitrogen ) ; unreacted dye was removed by pelleting the actin filaments followed by actin depolymerization . The extent of labeling ( percentage of dye - labeled protein ) was calculated by dividing the concen - tration of the protein - bound dye ( determined by measuring the fluores - cence emission intensity relative to a standard dilution series of the free dye ) by the protein concentration . Protein concentrations were calculated from the absorbance at 280 nm after subtracting the dye contribution . Preparation of Lipid - Coated Beads . Chloroform solutions of 1 , 2 - dioleoyl - sn - glycero - 3 - phosphocholine , NTAD ( nickel salt ; Avanti Polar Lipids ) , and Atto390 - 1 , 2 - dioleoyl - sn - glycero - 3 - phosphoethanolamine ( Atto390 - DOPE ; ATTO - TEC ) were mixed in molar ratios of 94 : 5 : 1 , 96 . 5 : 2 . 5 : 1 , and 98 : 1 : 1 of 1 , 2 - dioleoyl - sn - glycero - 3 - phosphocholine : NTAD : Atto390 - DOPE , evaporated in a stream of argon , and dried under high vacuum for at least 2 h . Lipids were resus - pended in vesiclebuffer ( 20 mMHepes , pH7 . 5 , 100 mMNaCl , 330 mMsucrose ) to a final concentration of 5 mM , sonicated for 20 s , subjected to five freeze \u2013 thaw cycles , and stored at \u2212 80 \u00b0C . A suspension of glass beads ( 3 . 5 \u03bc L ; 2 . 3 - \u03bc m diameter ; Total 2 . 8 \u00d7 10 7 ; Bangs Laboratories ) was mixed with 5 \u03bc L lipid suspension and diluted to a final volume of 45 \u03bc L in vesicle buffer ( 6 . 2 \u00d7 10 8 beads mL \u2212 1 ) . The suspension was sonicated ( 3 \u00d7 10 s ) in a bath sonicator ( Branson 2510 ) and incubated while rotating for 30 min at RT . The beads were pelleted by pulse centrifugation , washed two times with 200 \u03bc L vesicle buffer , resuspended in 45 \u03bc L vesicle buffer , and used within 8 h . Motility Assay with Different SH3 Adapters . Lipid - coated beads ( 6 . 1 \u00d7 10 6 beads mL \u2212 1 ) containing 1 % , 2 . 5 % , and 5 % NTAD were incubated with 0 . 25 \u03bc M His 6 - tagged SH3 adapter ( 20 % Alexa568 labeled ) , 0 . 05 \u03bc M N - WASP , and motility components [ 9 \u03bc M actin ( 5 % Alexa488 labeled ) , 0 . 075 \u03bc M Arp2 / 3 , 0 . 05 \u03bc M capping protein , 2 . 6 \u03bc M profilin , 3 . 5 \u03bc M cofilin ] in motility buffer ( 10 mM Hepes , pH 7 . 5 , 2 mM MgCl 2 , 50 mM KCl , 50 mM NaCl , 1 mg / mL BSA , 2 . 5 mM ATP , 5 mM TCEP , 1 . 3 mM Dabco , 20 mM ascorbic acid , 0 . 2 % methylcellulose ) . Samples were incubated for 15 min at RT while rotating . For quantitative image analysis , the reaction was diluted 1 : 1 in 3 % glutaraldehyde ( Ricca Chemical Company ) , spotted on a microscopy slide , covered with a coverslip , and imaged within 1 h . Motility Reactions Using pY 3 - nephrin and Nck . For each assay , lipid - coated beads ( 7 . 6 \u00d7 10 7 mL \u2212 1 ) were preincubated ( 15 min at RT ) with 0 . 5 \u03bc M His 8 - pY 3 - nephrin ( 38 ) in buffer containing 10 mM Hepes , pH 7 . 5 , 1 mg / mL BSA , and 50 mM KCl . Phosphonephrin - loaded beads were diluted eightfold in motility buffer containing 0 . 1 \u03bc M Nck ( 20 % Alexa568 labeled ) , 0 . 2 \u03bc M N - WASP , and actin motility components as described above . N - WASP and SH3 Adapter Density Measurements . To compare the surface densities of SH3 adapters or N - WASP , we quantified the Alexa568 signal on the lipid - coated beads . Indicated amounts of Alexa568 - labeled SH3 adapters or unlabeled SH3 adapter and Alexa568 N - WASP were incubated with lipid - coated beads in motility buffer for 15 min while rotating . The suspension ( \u223c 6 \u03bc L ) was transferred to a flow chamber that had been passivated with \u2013 poly - ( L - lysine ) - PEG ( 3 mg / mL ) ( 52 ) and imaged live as described below . Microscopy and Image Analysis . Epifluorescence and phase contrast images of fixed motility reactions were acquired with a 60 \u00d7 objective ( N . A . 1 . 4 ) on an Olympus IX70 Inverted Microscope equipped with a Coolsnap HQ CCD Camera ( Photometrics ) . Flat field - corrected images were analyzed using a CellProfiler Pipeline ( 53 ) to automatically locate beads , threshold images ( Background Global Method ) , and determine the integrated fluorescence intensity of Alexa488 actin associated with each bead . Live confocal images were acquired to quantify the density of N - WASP or SH3 adapters on lipid - coated beads . Images were obtained using a 100 \u00d7 objective ( N . A . 1 . 49 ) on a Nikon ECLIPSE Ti Inverted Microscope equipped with a Yokoyama CSU - X1 Spinning Disk , a Hamamatsu EM CCD 9100 \u2013 13 Camera , and 50 - mW lasers [ solid state , 405 nm ; diode - pumped solid - state ( DPSS ) , 561 nm ] . For density measurements , flat field - corrected images were analyzed for bead fluorescence using a Mathematica script . The script inactive G BD V P RR C A active membrane G BD V C A SH3A SH2 SH3B SH3C PxxP PxxP Nck pY N - WASP Fig . 6 . Proposed mechanism of N - WASP activation by Nck . In the auto - inhibited conformation , the N - WASP C helix binds intramolecularly to the GBD . Nck binds to tyrosine - phosphorylated receptors on the membrane and activates N - WASP by competitively displacing the C helix from the GBD with its inter - SH3AB linker . The released VCA segment is available to activate the Arp2 / 3 complex . Polyvalent interactions between the SH3 domains of Nck and the proline - rich region ( PRR ) of N - WASP can lead to higher - order olig - omer formation , increasing the local density of activated N - WASP molecules . Okrut et al . PNAS | Published online November 9 , 2015 | E6443 B I O C H E M I S T RY P N A S P L U S D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 . identifies bead positions based on the Atto390 - DOPE lipid fluorescence , quantifies bead fluorescence in the Alexa568 channel ( SH3 adapter or N - WASP ) , and subtracts the median background value of a circular region outside of the bead . P values were calculated using an unpaired , two - tailed Student \u2019 s t test . Pull - Down Assays . GST - tagged N - WASP GBD ( 1 . 5 nmol ; amino acids 196 \u2013 274 ) or PAK1 - GBD ( 1 . 5 nmol ; amino acids 67 \u2013 150 ; Cytoskeleton Inc . ) was immo - bilized on GSH Sepharose 4B ( 20 \u03bc L per sample ; GE Healthcare ) for 15 min at 4 \u00b0C . The beads were washed two times with Nck storage buffer . Nck con - structs ( 1 . 5 nmol total ) were added to the beads and incubated for 2 h at a final concentration of 5 \u03bc M . The beads were washed two times , and the supernatant as well as the bead samples were subjected to SDS / PAGE . The Coomassie - stained gels were analyzed using a Licor Gel Imaging System . For competitive pull - down assays , N - WASP VCA ( 0 \u2013 50 \u03bc M ) was added to the binding reaction . NMR Spectroscopy . All NMR chemical shift assignment experiments were performed at 298 K onan Agilent DD2 600 - MHz spectrometer equipped with a cold probe using NMRPipe for data processing ( 54 ) and NMRView for analysis ( 55 ) . Backbone chemical shift assignments were obtained using standard 3D 15 N - edited triple - resonance experiments including HNCACB , CBCA ( CO ) NH , and HNCO spectra ( 56 ) . 13 C - , 15 N - labeled N - WASP Nck fusion protein in 20 mM Hepes , pH 7 . 5 , 100 mM NaCl , 1 mM DTT , 1 mM EDTA , 0 . 01 % NaN 3 , and 5 % D 2 O was used at a concentration of 550 \u03bc M . NMR titration data were acquired at 300 K on a Bruker DRX500 spectrometer ( Topspin , version 1 . 3 ) equipped with a cryogenic probe . 15 N - labeled N - WASP GBD ( 65 \u03bc M ; in 20 mM Hepes , pH 7 . 5 , 200 mM NaCl , 1 mM DTT , 0 . 01 % NaN 3 , 5 % D 2 O ) was titrated with the Nck linker peptide ( 0 \u2013 520 \u03bc M ) . Titration curves from 17 peaks were fit using CCPN Analysis ( 57 ) . Weighted chemical shift differences were calculated using the following equation : \u0394\u03b4 = ( ( \u0394\u03b4 N \u00d7 0 . 2 ) 2 + \u0394\u03b4 H2 ) 1 / 2 . ACKNOWLEDGMENTS . We thank Justin Farlow for help with image analysis , Sudeep Banjade and Michael Rosen for providing triphosphorylated nephrin peptide , and Sebastian Rumpf for critical reading of the manuscript . This research was supported by funds from Tobacco - Related Disease Research Program of the University of California Grant 19FT - 0090 ( to J . O . ) and the Howard Hughes Medical Institute . NMR spectroscopy at University of Texas Southwestern Medical Center is supported by NIH instrumentation Grant 1S10OD018027 - 01 . 1 . Le Clainche C , Carlier M - F ( 2008 ) Regulation of actin assembly associated with pro - trusion and adhesion in cell migration . Physiol Rev 88 ( 2 ) : 489 \u2013 513 . 2 . Pollard TD , Cooper JA ( 2009 ) Actin , a central player in cell shape and movement . Science 326 ( 5957 ) : 1208 \u2013 1212 . 3 . Fletcher DA , Mullins RD ( 2010 ) Cell mechanics and the cytoskeleton . Nature 463 ( 7280 ) : 485 \u2013 492 . 4 . PrehodaKE , ScottJA , MullinsRD , LimWA ( 2000 ) Integrationofmultiplesignalsthrough cooperative regulation of the N - WASP - Arp2 / 3 complex . Science 290 ( 5492 ) : 801 \u2013 806 . 5 . Papayannopoulos V , et al . ( 2005 ) A polybasic motif allows N - WASP to act as a sensor of PIP ( 2 ) density . Mol Cell 17 ( 2 ) : 181 \u2013 191 . 6 . Torres E , Rosen MK ( 2006 ) Protein - tyrosine kinase and GTPase signals cooperate to phosphorylate and activate Wiskott - Aldrich syndrome protein ( WASP ) / neuronal WASP . J Biol Chem 281 ( 6 ) : 3513 \u2013 3520 . 7 . Co C , Wong DT , Gierke S , Chang V , Taunton J ( 2007 ) Mechanism of actin network attachment to moving membranes : Barbed end capture by N - WASP WH2 domains . Cell 128 ( 5 ) : 901 \u2013 913 . 8 . Padrick SB , Rosen MK ( 2010 ) Physical mechanisms of signal integration by WASP family proteins . Annu Rev Biochem 79 : 707 \u2013 735 . 9 . Miki H , Sasaki T , Takai Y , Takenawa T ( 1998 ) Induction of filopodium formation by a WASP - related actin - depolymerizing protein N - WASP . Nature 391 ( 6662 ) : 93 \u2013 96 . 10 . Kim AS , Kakalis LT , Abdul - Manan N , Liu GA , Rosen MK ( 2000 ) Autoinhibition and acti - vation mechanisms of the Wiskott - Aldrich syndrome protein . Nature 404 ( 6774 ) : 151 \u2013 158 . 11 . Abdul - Manan N , et al . ( 1999 ) Structure of Cdc42 in complex with the GTPase - binding domain of the \u2018 Wiskott - Aldrich syndrome \u2019 protein . Nature 399 ( 6734 ) : 379 \u2013 383 . 12 . Padrick SB , Doolittle LK , Brautigam CA , King DS , Rosen MK ( 2011 ) Arp2 / 3 complex is bound and activated by two WASP proteins . Proc NatlAcadSci USA 108 ( 33 ) : E472 \u2013 E479 . 13 . Ti S - C , Jurgenson CT , Nolen BJ , Pollard TD ( 2011 ) Structural and biochemical charac - terization of two binding sites for nucleation - promoting factor WASp - VCA on Arp2 / 3 complex . Proc Natl Acad Sci USA 108 ( 33 ) : E463 \u2013 E471 . 14 . LiSS - C ( 2005 ) SpecificityandversatilityofSH3andotherproline - recognitiondomains : Structural basis and implications for cellular signal transduction . Biochem J 390 ( Pt 3 ) : 641 \u2013 653 . 15 . Frischknecht F , et al . ( 1999 ) Actin - based motility of vaccinia virus mimics receptor tyrosine kinase signalling . Nature 401 ( 6756 ) : 926 \u2013 929 . 16 . Gruenheid S , et al . ( 2001 ) Enteropathogenic E . coli Tir binds Nck to initiate actin pedestal formation in host cells . Nat Cell Biol 3 ( 9 ) : 856 \u2013 859 . 17 . Rohatgi R , Nollau P , Ho HY , Kirschner MW , Mayer BJ ( 2001 ) Nck and phosphatidyli - nositol 4 , 5 - bisphosphate synergistically activate actin polymerization through the N - WASP - Arp2 / 3 pathway . J Biol Chem 276 ( 28 ) : 26448 \u2013 26452 . 18 . Campellone KG , et al . ( 2004 ) Clustering of Nck by a 12 - residue Tir phosphopeptide is sufficient to trigger localized actin assembly . J Cell Biol 164 ( 3 ) : 407 \u2013 416 . 19 . RiveraGM , Brice\u00f1o CA , TakeshimaF , SnapperSB , MayerBJ ( 2004 ) Inducibleclustering of membrane - targeted SH3 domains of the adaptor protein Nck triggers localized actin polymerization . Curr Biol 14 ( 1 ) : 11 \u2013 22 . 20 . Jones N , et al . ( 2006 ) Nck adaptor proteins link nephrin to the actin cytoskeleton of kidney podocytes . Nature 440 ( 7085 ) : 818 \u2013 823 . 21 . Verma R , et al . ( 2006 ) Nephrin ectodomain engagement results in Src kinase activa - tion , nephrin phosphorylation , Nck recruitment , and actin polymerization . J Clin Invest 116 ( 5 ) : 1346 \u2013 1359 . 22 . Mohamed AM , Boudreau JR , Yu FPS , Liu J , Chin - Sang ID ( 2012 ) The Caenorhabditis elegans Eph receptor activates NCK and N - WASP , and inhibits Ena / VASP to regulate growth cone dynamics during axon guidance . PLoS Genet 8 ( 2 ) : e1002513 . 23 . Donnelly SK , Weisswange I , Zettl M , Way M ( 2013 ) WIP provides an essential link between Nck and N - WASP during Arp2 / 3 - dependent actin polymerization . Curr Biol 23 ( 11 ) : 999 \u2013 1006 . 24 . She HY , et al . ( 1997 ) Wiskott - Aldrich syndrome protein is associated with the adapter protein Grb2 and the epidermal growth factor receptor in living cells . Mol Biol Cell 8 ( 9 ) : 1709 \u2013 1721 . 25 . Carlier MF , et al . ( 2000 ) GRB2 links signaling to actin assembly by enhancing in - teraction of neural Wiskott - Aldrich syndrome protein ( N - WASp ) with actin - related protein ( ARP2 / 3 ) complex . J Biol Chem 275 ( 29 ) : 21946 \u2013 21952 . 26 . Scaplehorn N , et al . ( 2002 ) Grb2 and Nck act cooperatively to promote actin - based motility of vaccinia virus . Curr Biol 12 ( 9 ) : 740 \u2013 745 . 27 . Tang DD , Zhang W , Gunst SJ ( 2005 ) The adapter protein CrkII regulates neuronal Wiskott - Aldrich syndrome protein , actin polymerization , and tension development during contractile stimulation of smooth muscle . J Biol Chem 280 ( 24 ) : 23380 \u2013 23389 . 28 . Kowalski JR , et al . ( 2005 ) Cortactin regulates cell migration through activation of N - WASP . J Cell Sci 118 ( Pt 1 ) : 79 \u2013 87 . 29 . Takano K , Toyooka K , Suetsugu S ( 2008 ) EFC / F - BAR proteins and the N - WASP - WIP complex induce membrane curvature - dependent actin polymerization . EMBO J 27 ( 21 ) : 2817 \u2013 2828 . 30 . Pichot CS , et al . ( 2010 ) Cdc42 - interacting protein 4 promotes breast cancer cell in - vasion and formation of invadopodia through activation of N - WASp . Cancer Res 70 ( 21 ) : 8347 \u2013 8356 . 31 . Oikawa T , Itoh T , Takenawa T ( 2008 ) Sequential signals toward podosome formation in NIH - src cells . J Cell Biol 182 ( 1 ) : 157 \u2013 169 . 32 . Padrick SB , et al . ( 2008 ) Hierarchical regulation of WASP / WAVE proteins . Mol Cell 32 ( 3 ) : 426 \u2013 438 . 33 . Loisel TP , Boujemaa R , Pantaloni D , Carlier MF ( 1999 ) Reconstitution of actin - based motility of Listeria and Shigella using pure proteins . Nature 401 ( 6753 ) : 613 \u2013 616 . 34 . MullinsRD , HansenSD ( 2013 ) In vitrostudiesof actinfilamentandnetworkdynamics . Curr Opin Cell Biol 25 ( 1 ) : 6 \u2013 13 . 35 . Lacayo CI , et al . ( 2012 ) Choosing orientation : Influence of cargo geometry and ActA polarization on actin comet tails . Mol Biol Cell 23 ( 4 ) : 614 \u2013 629 . 36 . Koronakis V , et al . ( 2011 ) WAVE regulatory complex activation by cooperating GTPases Arf and Rac1 . Proc Natl Acad Sci USA 108 ( 35 ) : 14449 \u2013 14454 . 37 . Uruno T , et al . ( 2001 ) Activation of Arp2 / 3 complex - mediatedactin polymerizationby cortactin . Nat Cell Biol 3 ( 3 ) : 259 \u2013 266 . 38 . Banjade S , Rosen MK ( 2014 ) Phase transitions of multivalent proteins can promote clustering of membrane receptors . eLife 3 : e04123 . 39 . Alto NM , et al . ( 2007 ) The type III effector EspF coordinates membrane trafficking by the spatiotemporalactivationoftwoeukaryoticsignalingpathways . JCellBiol 178 ( 7 ) : 1265 \u2013 1278 . 40 . Cheng H - C , Skehan BM , Campellone KG , Leong JM , Rosen MK ( 2008 ) Structural mechanism of WASP activation by the enterohaemorrhagic E . coli effector EspF ( U ) . Nature 454 ( 7207 ) : 1009 \u2013 1013 . 41 . Sallee NA , et al . ( 2008 ) The pathogen protein EspF ( U ) hijacks actin polymerization using mimicry and multivalency . Nature 454 ( 7207 ) : 1005 \u2013 1008 . 42 . Shen Y , Delaglio F , Cornilescu G , Bax A ( 2009 ) TALOS + : A hybrid method for predicting proteinbackbonetorsionanglesfromNMRchemicalshifts . JBiomolNMR 44 ( 4 ) : 213 \u2013 223 . 43 . Spera S , Bax A ( 1991 ) Empirical correlation between protein backbone conformation and C . alpha . and C . beta . 13C nuclear magnetic resonance chemical shifts . J Am Chem Soc 113 ( 14 ) : 5490 \u2013 5492 . 44 . Hsu WL , et al . ( 2013 ) Exploring the binding diversity of intrinsically disordered pro - teins involved in one - to - many binding . Protein Sci 22 ( 3 ) : 258 \u2013 273 . 45 . Takeuchi K , Sun Z - YJ , Park S , Wagner G ( 2010 ) Autoinhibitory interaction in the multidomain adaptor protein Nck : Possible roles in improving specificity and func - tional diversity . Biochemistry 49 ( 27 ) : 5634 \u2013 5641 . 46 . BanjadeS , et al . ( 2015 ) Conserved interdomain linker promotes phase separation ofthe multivalent adaptor protein Nck . Proc Natl Acad Sci USA 112 : E6426 \u2013 E6435 . 47 . Kelly AE , Kranitz H , D\u00f6tsch V , Mullins RD ( 2006 ) Actin binding to the central domain of WASP / Scar proteins plays a critical role in the activation of the Arp2 / 3 complex . J Biol Chem 281 ( 15 ) : 10589 \u2013 10597 . 48 . Li P , et al . ( 2012 ) Phase transitions in the assembly of multivalent signalling proteins . Nature 483 ( 7389 ) : 336 \u2013 340 . 49 . PardeeJD , SpudichJA ( 1982 ) Purificationofmuscleactin . MethodsEnzymol 85 ( PtB ) : 164 \u2013 181 . 50 . Palmgren S , Ojala PJ , Wear MA , Cooper JA , Lappalainen P ( 2001 ) Interactions with PIP2 , ADP - actinmonomers , and capping protein regulatethe activity and localization of yeast twinfilin . J Cell Biol 155 ( 2 ) : 251 \u2013 260 . E6444 | www . pnas . org / cgi / doi / 10 . 1073 / pnas . 1510876112 Okrut et al . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 . 51 . Egile C , et al . ( 1999 ) Activationof the CDC42 effectorN - WASP by the Shigella flexneri IcsA protein promotes actin nucleation by Arp2 / 3 complex and bacterial actin - based motility . J Cell Biol 146 ( 6 ) : 1319 \u2013 1332 . 52 . Bieling P , Telley IA , Hentrich C , Piehler J , Surrey T ( 2010 ) Fluorescence microscopy assays on chemically functionalized surfaces for quantitative imaging of microtubule , motor , and + TIP dynamics . Methods Cell Biol 95 ( 2010 ) : 555 \u2013 580 . 53 . LamprechtMR , SabatiniDM , CarpenterAE ( 2007 ) CellProfiler : Free , versatilesoftware for automated biological image analysis . Biotechniques 42 ( 1 ) : 71 \u2013 75 . 54 . Delaglio F , et al . ( 1995 ) NMRPipe : A multidimensional spectral processing system based on UNIX pipes . J Biomol NMR 6 ( 3 ) : 277 \u2013 293 . 55 . Johnson BA ( 2004 ) Using NMRView to visualize and analyze the NMR spectra of macromolecules . Methods Mol Biol 278 : 313 \u2013 352 . 56 . Muhandiram DR , Kay LE ( 1994 ) Gradient - enhanced triple - resonance three - dimensional NMR experiments with improved sensitivity . J Magn Reson B 103 ( 3 ) : 203 \u2013 216 . 57 . Vranken WF , et al . ( 2005 ) The CCPN data model for NMR spectroscopy : Development of a software pipeline . Proteins 59 ( 4 ) : 687 \u2013 696 . Okrut et al . PNAS | Published online November 9 , 2015 | E6445 B I O C H E M I S T RY P N A S P L U S D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S A RC S - S E R I A L S on S e p t e m b e r 20 , 2023 fr o m I P a dd r e ss 205 . 175 . 106 . 245 .",
    "sathe2018small_supplement": "Small GTPases and BAR domain proteins regulate branched actin 1 polymerization for clathrin and dynamin - independent endocytosis 2 Sathe and Muthukrishnan et al 3 4 Supplementary Figures 5 6 Supplementary Figure 1 : Characterization of endocytosis in AGS cells . ( a - b ) Endocytic 7 route of SecGFP - GPI in AGS cells . Single confocal plane shows a 3 minute pulse of \u03b1 - GFP 8 Fab [ a - b : magenta in Merge ] at 37\u00b0C ( upper panel ) and 30\u00b0C ( lower panel ) along with TMR - 9 dextran ( a : green in Merge ) or A647 - Tf ( b : green in Merge ) in AGS cells transiently 10 transfected with SecGFP - GPI . Insets show a magnified view of the marked areas . ( c ) Plot 11 ( top ) showing quantification of the fraction of GFP - GPI endocytic vesicles containing fluid 12 or Tf . The number of cells is shown below the graph . ( d ) Schematic of pH pulsing analysis , 13 steps ( 1 - 3 ) used for identifying and quantifying the fluorescence spots associated with newly 14 formed endocytic vesicles in the sequential frames of the pH 5 montage . Step 1 - New spot 15 identification : Each spot ( green ) in i th frame is compared with the previous frame and is 16 considered new if no nearest neighbour is found by euclidean distance search within 5 pixels 17 ( 1 pixel = 84nm ) . Random ( red ) are generated randomly within the cell mask . Step 2 - Spot 18 mask ( green filled ) and background mask ( green dashed ) considered around the centroid the 19 new spot identified in the step - 1 [ also see pH 5 frame in the middle panel of ( Figure 1a ) ] . 20 The process is repeated for Random spot mask ( red filled ) and background mask ( red 21 dashed ) . Step 3 - schematic of the temporal profile of a new spot ( black ) and a random spot . 22 Multiple traces pooled from different spots and cells over different days was averaged and 23 shown in rest of the figures . The x - axis represents time and y - axis represents fold change in 24 intensity in the spot over the background . See S . I . for a detailed description . ( e ) AGS cells 25 were pre - treated with either pH 7 . 3 or pH 5 . 5 buffer for 5 minutes followed by 5 - minute fluid 26 uptake in pH 7 . 3 buffer . Data was pooled from 2 independent experiments and the number of 27 cells indicated in the graph . ( f ) Untreated AGS ( Control , ( i ) ) or LG186 - treated AGS ( ii ) were 28 incubated for 2 minutes at 37\u00b0C with 10mg / ml HRP as a fluid phase marker before 29 processing for electron microscopy . Endocytic structures close to the plasma membrane ( PM ) 30 are filled with the electron dense peroxidase precipitate . Control cells show a range of 31 endocytic structures including vesicular structures ( CCP / Cav & EE ) ( a pair of small 32 arrowheads ) and tubular / ring - shaped putative CLIC / GEECs ( large arrowhead ) but the drug - 33 treated cells show predominant labelling of vesicular profiles . Histogram shows mean 34 endocytic structures quantified per cell ( n = 5 ) . CCP / Cav represent vesicles derived from 35 clathrin mediate or caveolar endocytosis and EE represent early endosomes ( See S . I . ) . ( g ) 36 The histogram shows quantification of 5 - minute fluid - uptake in AGS cells when treated with 37 indicated concentrations of LG186 or DMSO . Data was pooled from 2 independent 38 experiments and the number of cells indicated in the graph . Scale bar , 5\u00b5m ( a - b ) , 4\u00b5m ( a - b , 39 inset ) , 1 \u00b5m ( f ) & 20\u00b5m ( e and g ) respectively . Error bars ( c ) represent s . e . m . and ( e - g ) s . d . 40 respectively . p - value < 0 . 001 ( * * ) 2 - sample student\u2019s T - test ( f ) and Mann - Whitney U test 41 ( e ) . 42 43 Supplementary Figure 2 : Characterization of the pH pulsing assay for visualizing 44 SecGFP - GPI endocytic vesicle formation . ( a - b ) Graphs show the average normalized 45 fluorescence intensity versus time traces for the recruitment of TagRFPt - CDC42 ( CDC42 46 All ) to all forming SecGFP - GPI endocytic sites ( a ) or endocytic sites that do not co - detect 47 TagRFPt - CDC42 ( CDC42 NoColoc ; 44 % ) and its corresponding random intensity trace ( n , 48 Table 1 ) . ( c - g ) Graphs show the average normalized fluorescence intensity versus time trace 49 for the recruitment of mCherry - dynamin to forming SecTfR endocytic sites [ c ; ( n = 21 50 SecTfR and 1448 random spots from 6 cells , 2 experiments ) , mCherry - clathrin to forming 51 SecGFP - GPI endocytic sites [ d ; n , Table 1 ] , mCherry - dynamin [ e ; n , Table 1 ] to forming 52 SecGFP - GPI endocytic sites , mCherry - ARF1 to forming SecGFP - GPI endocytic sites [ f ; ( n = 53 228 , as in Figure 1e ) ; Note the extended time axis ] , mCherryGBF1 to forming SecGFP - GPI 54 endocytic sites [ g ; ( n = 75 , as in Figure 1e ) ; Note the extended time axis ] , and their 55 corresponding random intensity trace . A representative montage depicted below ( c - e ) . 56 Arrowheads indicate the spot . The random traces were derived from randomly assigned spots 57 of the same radius as the endocytic regions , as detailed in S . I . Endocytic distribution at each 58 time point was compared to the random distribution by Mann - Whitney U test and the log 10 59 ( p ) [ log 10 ( 0 . 05 ) is - 1 . 3 and log 10 ( 0 . 001 ) is - 2 . 5 ] is plotted below each trace ( a , & c - g ) . Error 60 bars represent s . e . m . for ( a , & c - g ) . Scale bar , 1 . 5\u00b5m ( c - e ) . 61 62 Supplementary Figure 3 : RNAi screen reveals BAR domain proteins involved in CG 63 endocytosis . Representative images for data shown in Figure 2b . Scale bar is 20\u00b5m . 64 Supplementary Figure 4 : IRSp53 is involved in CG endocytosis . ( a ) Histogram ( top ) 65 showing quantification of mean 5 - minute normalized TfR uptake in IRSp53 WT cells when 66 treated with 10\u00b5M LG186 for 45 minutes along with the representative images ( bottom ) . 67 Data was pooled from 2 independent experiments and the number of cells is indicated in the 68 figure . ( b ) Histogram ( top ) shows mean number of endocytic structures per cell from the 69 electron microscope images ( below ) . Data pooled from 3 independent blocks with 5 cells 70 each . Untreated WT MEFs ( WT , i ) , IRSp53 null MEFs ( IRSp53 - / - , ii ) or LG186 - treated WT 71 MEFs ( LG186 , iii ) were incubated for 2 minutes at 37\u00b0C with 10mg / ml HRP as a fluid phase 72 marker before processing for electron microscopy . Endocytic structures close to the plasma 73 membrane ( PM ) are filled with the electron dense peroxidase precipitate . WT cells ( left ) 74 show a range of endocytic structures including vesicular structures ( double arrowheads ) and 75 tubular / ring - shaped putative CLIC / GEECs ( large arrowheads ) but the KO cells ( middle ) and 76 LG186 - treated ( right ) cells show predominant labelling of vesicular profiles . ( c ) Electron 77 micrographs of AGS cells co - transfected GFP - IRSp53 and GBP - Apex . ( i - ii ) The reaction 78 product is highly patched ( arrowheads ) on the plasma membrane ( PM ) along with zoomed 79 regions on the right . ( iii ) Double arrowheads indicate specific labelling within defined 80 microdomains of filopodia . ( d ) Plot ( top ) showing quantification of co - localizing IRSp53 81 with p21 ( ARP2 / 3 complex subunit ) using ImageJ plugin ( Van Steensel\u2019s CCF , See 82 Methods . ) when compared with its random . Representative ( bottom ) images of AGS cells co - 83 expressing mEmerald - p21 subunit with mCherry - IRSp53 , which were fixed and imaged with 84 TIRFM , with zoomed inset at top left corner . Data was pooled from 2 independent 85 experiments and the cell number is indicated below the graph . ( e ) ( 1a - c ) Wide - field images of 86 cells at 1x ( 1a ) , 4x ( 1b ) along with a representative confocal slice of a cell ( 4x , 1c ) . ( 2a - c ) 87 Inset from the 4x cell ( Supplementary Movie 5 ) depicting enrichment of IRSp53 ( magenta , 88 2b ) within filopodia ( CD44 , green , 2a ) along with the merge ( 2c ) . ( 3 - 6 ) Insets of CD44 89 ( green ) labelled invaginations showing recruitment of IRSp53 ( magenta ) at various stages . 90 Each example depicts from left to right CD44 , IRSp53 and Merge respectively . Scale bar , 91 20\u00b5m ( a ) , 1\u00b5m ( b - c ) , 5\u00b5m ( d ) , 20\u00b5m ( e , 1a - b ) , 5\u00b5m ( e , 1c ) , 1\u00b5m ( e , 2a - c & 3 - 6 ) 92 respectively . p - value < 0 . 01 ( * ) , and 0 . 001 ( * * ) by Mann - Whitney U Test ( d ) and 2 - sample 93 student\u2019s T - test ( b ) . 94 95 Supplementary Figure 5 : ARP2 / 3 is negatively regulated by PICK1 for CG endocytosis 96 independent of N - WASP . ( a ) Histogram ( left ) showing quantification of 5 - minute pulse 97 fluid - phase in AGS cells treated with either DMSO or 10\u00b5M SMIFH2 along with 98 representative images ( right ) . Data was pooled from 2 independent experiments and the cell 99 number is indicated in the figure . ( b ) Plot ( left ) showing quantification of co - localizing 100 PICK1 with ARP3 or ARF1 using ImageJ plugin ( Van Steensel\u2019s CCF , See Methods ) when 101 compared with its random . Data was pooled from 2 independent experiments and the number 102 the cell number is indicated below the graph . Representative ( right ) images of AGS cells co - 103 expressing GFP - ARF1 with TagRFP - PICK1 or GFP - PICK1 and mCherry - ARP3 which were 104 fixed and imaged with TIRFM , with zoomed inset at bottom right corner . ( c ) Histogram ( left ) 105 showing quantification of mean 5 - minute pulse fluid - phase in AGS cells overexpressing 106 either pIRES - PICK1 WT or PICK1 KK - EE mutant . Data was pooled from 2 independent 107 experiments and the number of cells is indicated in the figure . Error bar represents s . d . ( a , & 108 c ) Scale bar , 20\u00b5m ( a & c ) , 5\u00b5m ( b ) . p - value < 0 . 01 ( * ) , and 0 . 001 ( * * ) by Mann - Whitney U 109 Test ( a - c ) . 110 Supplementary Figure 6 : Fold change over time profiles of SecGFP - GPI spots that did 111 not show a co - detected X - FP spot . ( a - h ) Graphs show the average normalized fluorescence 112 intensity versus time traces for the recruitment of mCherry - ARF1 NoColoc ( a ) , mCherry - 113 GBF1 NoColoc ( b ) , mCherry - IRSp53 NoColoc ( c ) , mCherry - ARP3 NoColoc ( d ) , pRuby - 114 Lifeact NoColoc ( e ) , TagRFPt - PICK1 NoColoc ( f ) , mCherry - NWASP NoColoc ( g ) 115 compared to their respective random trace . The random traces were derived from randomly 116 assigned spots of the same radius as the endocytic regions , as detailed in S . I . They represent 117 the fraction of SecGFP - GPI that failed to show a co - detected the X - FP spots . For n see Table 118 1 and S . I . for further information . 119 Supplementary Methods and Materials 120 Chemicals and reagents : All reagents were purchased from Sigma unless otherwise 121 mentioned . TMR - Dextran ( 10kDa ) and Alexa TM dyes were purchased from Molecular Probes 122 ( Eugene , OR ) . Transfection was carried out using FuGENE6 reagent ( Promega , USA ) as per 123 the manufacturer\u2019s instructions . LG186 was synthesized in the laboratory of Dr . Ram A . 124 Vishwakarma ( Indian Institute of Integrative Medicine ( CSIR ) , Canal Road , Jammu 180001 , 125 India ) . Heat inactivated fetal bovine serum ( 16000044 ) . Hygromycin B ( 10687 - 010 ) was 126 obtained from Invitrogen . CK - 666 ( SML0006 - 5MG ) , SMIFH2 ( S4826 - 5MG ) , FSC231 127 ( 529531 - 10MG , Millipore ) , Puromycin ( P2233 ) . PI - PLC ( purified in - house , used at 128 50\u00b5g / ml ) . hTf was Iron loaded and purified in gel filtration 1 and was fluorescently labelled 129 with ( A 10235 ) , Alexa - 568 ( A 10238 ) or Alexa - 647 ( A 20173 ) as per manufacturer\u2019s 130 instructions . Acryloyl X \u2013 SE ( Life technologies : A20770 ) . Sodium acrylate ( 408220 , 131 Sigma ) . 132 Cell culture : FR - AGS ( adenogastric carcinoma cells stably transfected with FR - GPI ) 2 were 133 maintained in 10 % FBS HF12 ( HiMedia , India ) . S2R + TfR cells 2 , 3 were grown in 10 % FBS 134 Schneider\u2019s Drosophila Media ( 21720 - 024 , Gibco TM ) . IRSp53 - / - ( pBABE - puro empty ) and 135 IRSp53 - / - pBABE - puro - IRSp53WT mouse embryonic fibroblasts 4 were grown in 20 % FBS 136 DMEM ( 31053 - 028 , Gibco TM ) supplemented by with 1 \u00b5g / ml Puromycin . All media were 137 supplemented with L - Glutamine Penicillin Streptomycin Solution ( Sigma , G1146 - 100ml ) . 138 CK - 666 treatment was done for 90 minutes . SMIFH2 treatment was done for 60 - 90 minutes 139 until lamellipodia formation was observed . FSC231 treatment was done for 60 minutes . 140 LG186 treatment was done in serum free media for 45 minutes ( MEFs ) and 30 minutes ( FR - 141 AGS ) . 142 Antibodies : \u03b1 - hTfR monoclonal antibody was purified from mouse hybridoma , OKT9 5 143 ( National Centre for Cell Science , India , used at 1 : 100 ) , \u03b1 - PICK1 antibody ( PA1 - 073 , 144 Thermo Pierce , used at 1 : 50 ) . \u03b1 - mTfR antibody ( 553264 , Becton Dickinson , used at 1 : 50 ) 145 and Cy3 - Mov18 ( 5\u00b5g / mL ) 5 , 6 . Fab fragments of \u03b1 - GFP monoclonal antibodies generated 146 using papain digestion subsequently fluorescently labelled with Alexa - 647 ( used at 1 : 100 ) . \u03b1 - 147 GFP antibody ( ab290 , abcam , 1 : 200 ) . \u03b1 - CD44 antibody ( IM7 clone , 14 - 0441 - 82 , 148 eBiosciences , 1 : 150 ) . \u03b1 - IRSp53 ( HPA023310 , Sigma , 1 : 50 ) . 149 Plasmids : SecGFP - GPI was made by site - directed mutagenesis , F64L and S65T in ecliptic - 150 GPI 7 was a gift from Gero Miesenb\u04e7ck ( University of Oxford ) . pIRES - EGFP - PICK KK - EE 151 and pIRES - EGFP - N - WASP - CA domain were a gift from J . Hanley ( University of Bristol , 152 UK ) 8 . GFP - N - WASP - VCA and GFP - N - WASP\u0394VCA were a gift from Mike Way ( Francis 153 Crick Institute ) . ARF1 - mCherry was a gift from Paul Melan\u00e7on 9 ( University of Alberta ) . 154 GBF1 - mCherry was a gift from Catherine Jackson ( Institut Jacques Monod ) 10 . pRuby - 155 Lifeact was a gift from Roland Wedlich - Soeldner ( Max Planck Institute of Biochemistry ) 11 . 156 IRSp53 - mCherry , SecGFP - TfR 12 , psPAX2 and pMD2 . G were gifts from Dr Marcus J . 157 Taylor ( UCSF / NCBS ) . GFP - PICK1 and TagRFPt - PICK1 was a gift from Harvey McMahon 158 ( Medical Research Council , UK ) . Tag - RFPt - CDC42 was made in the laboratory by sub - 159 cloning GFP - CDC42 into a Tag - RFPt vector . Tag - RFPt vector was purchased from Evrogen 160 ( Catalogue # FP141 ) . GFP - IRSp53 and mutants were generated in the laboratory of Giorgio 161 Scita ( IFOM ) . Dyn2 - pmCherryN1 ( Addgene plasmid # 27689 ) , CLC - pmCherryC1 ( Addgene 162 plasmid # 27680 ) and Arp3 - pmCherryC1 ( Addgene plasmid # 27682 ) , mCherry - WASP - N - 18 163 ( Addgene plasmid # p55164 ) . GFP - binding peptide conjugated to APEX2 ( Addgene plasmid 164 # 67651 ) . shRNA against hPICK1 ( V3LHS _ 347038 , RHS4531 - EG9463 , 165 5\u2019TTCTTCAACACAATGTCCA3\u2019 ) and GIPZ Non - silencing Lentiviral shRNA Control 166 ( RHS4346 ) were purchased from Open Biosystems . 167 Generation of lentiviral stable lines : To generate GFP - IRSp53 and its different mutant 168 addback stable lines in IRSp53 - / - mouse embryonic fibroblasts and shRNA stable line for 169 FR - AGS , second generation lentiviral packaging systems was used following the protocol 170 from Trono lab 13 . HEK293T cells was used for virus generation and PEI ( polyethylenimine ) 171 was used for transfection . The ratio of DNA to PEI used for transfection is gene of interest : 172 packaging ( psPAX2 ) : envelope ( pMD2 . G ) in 1200ng : 1200ng : 1200ng for 3 wells of a 6 - well 173 plate . This DNA mix is added to serum free DMEM ( 327 \u00b5L / well ) and PEI ( 18 . 8 \u00b5L / well ) 174 and the transfection mix is incubated at room temperature for 20 minutes and added to the 175 cells . PEI is used at 1mg / mL and is heated to 70\u2103 for 10 minutes before transfection . 176 Viruses were collected after 2 - 3 days and syringe filtered using 200 \u00b5m filter and added to 177 60 % confluent culture . Typically after 3 - 4 days of infection gene expression was observed . 178 A similar procedure was followed to generate stable shRNA lines by in FR - AGS . 179 Immunostaining : Antibodies were tested for their specificity by western blotting if it was 180 suitable . Additionally expression was verified with published reports and was compared with 181 the expression of the fluorescently tagged version of the gene . Cells were fixed using 4 % 182 paraformaldehyde for 20 minutes at room temperature . Cells were permeablised with 0 . 1 % 183 TritonX at for 10 minutes followed by blocking for 1 hour using 2mg / mL BSA in PBS . \u03b1 - 184 PICK1 was diluted in blocking buffer ( 1 : 50 ) and added to the cells overnight at 4\u00b0C , 185 followed by A647 - Goat - \u03b1 - Rabbit for 1 hour at room temperature . 186 Image Analysis 187 pH pulsing assay : A semi - automated analysis was developed on MATLAB ( M . S . ) to identify 188 newly formed endosomes in the pH pulsing assay and trace their intensity over time in pH 7 , 189 pH 5 and RFP channels . The following steps ( a - g ) describe the procedure used : 190 a ) Identification of new endocytic structures : Two consecutive pH 5 frames were segmented 191 using MATLAB based tracking software which segments using sub - pixel localisation of 192 particles using radial symmetry centre 14 . The x - y coordinate of each particle in the i th 193 frame was compared with the rest of the particles in the i - 1 frame and the nearest 194 neighbour was located . If the nearest neighbour for a particle in the i th frame was found in 195 the i - 1 frame within 5 - pixel radius of its current co - ordinate , then that particle was 196 rejected . If the nearest neighbour was not found within a 5 - pixel radius of its current co - 197 ordinate and it stayed on for the next frame , then that particle was considered \u2018new\u2019 . 198 b ) Intensity calculation and bleach correction : A mask of radius 3 pixels ( 252 nm ) grown 199 from the centroid of the new spot was used for all the calculations . For the local 200 background , an annulus of radius 6 - 8 pixels ( 8 - 10 pixels for PICK1 ) was taken from the 201 centroid of the new spot . The average intensity of a spot is normalised to the average 202 intensity of the local annulus . ( mean intensity of spot - 20th percentile of local 203 background ) / ( mean intensity of local background - 20 percentile of local background ) 204 ( \u2211\ud835\udc3c \ud835\udc60 \ud835\udc5b \ud835\udc60 \u2212 20\ud835\udc61\u210e \ud835\udc5d\ud835\udc52\ud835\udc5f\ud835\udc50\ud835\udc52\ud835\udc5b\ud835\udc61\ud835\udc56\ud835\udc59\ud835\udc52 \ud835\udc5c\ud835\udc53 \ud835\udc3c \ud835\udc4f ) ( \u2211\ud835\udc3c \ud835\udc4f \ud835\udc5b \ud835\udc4f \u2212 20\ud835\udc61\u210e \ud835\udc5d\ud835\udc52\ud835\udc5f\ud835\udc50\ud835\udc52\ud835\udc5b\ud835\udc61\ud835\udc56\ud835\udc59\ud835\udc52 \ud835\udc5c\ud835\udc53 \ud835\udc3c \ud835\udc4f ) . \u2044 Each spot was 205 considered if its average intensity was at least 20 % higher than the local background to 206 throw out low signal to noise spots . Additionally , we demanded that in the subsequent 207 frame the average intensity of the spot should be within 10 % so rule out a spot coming 208 into the TIRF plane , instead of it being recently pinched endocytic vesicle . A spot\u2019s 209 intensity will increase as it approaches the plasma membrane , while a recently pinched 210 endocytic vesicle will not show a rise in the intensity . These newly identified structures in 211 the TIRF plane were used as fiducial markers of the event . 212 c ) Random spots generation : Arbitrary regions using the masks defined in step ( b ) within 213 the cell boundary was chosen . The intensity of such spots is henceforth called \u201cRandom\u201d 214 was calculated as discussed in Step ( b ) and was used to compare the behaviour of X - FP at 215 the new SecGFP - GPI spots vs . random locations in the cell . 216 d ) False positive removal : In addition to the criteria mentioned in Step ( b ) , separately , 217 montages of each event depicting frames from - 18s to + 18s were generated and used to 218 manually verify the appearance of a new event . In addition to it , a spot can be mistakenly 219 identified as new due to the sudden lateral movement into the frame or growth of a sub - 220 threshold spot , appearing in the TIRF plane from the inside of the cell . These events are 221 screened out by manually verifying the absence of such occurrences in the frames prior to 222 0s in the montages . 223 e ) Manual classification of the X - FP spots : In addition to removing the false positives , the 224 montages generated at the Step ( d ) underwent manual check to classify the new SecGFP - 225 GPI spots into two groups based on whether X - FP co - detection was observed or not . One 226 population exhibited co - localisation of X - FP with SecGFP - GPI during the time window 227 of - 18 to + 18s for at least 1 frame , while the second failed to register any co - localization . 228 We called them X - FP Coloc ( Co - detection of X - FP and SecGFP - GPI ) and X - FP NoColc 229 ( the remainder ) . The X - FP NoColoc profile was comparable to Random ( Supplementary 230 Fig . 2b and 6 ) indicative of lack of recruitment in that fraction . This was supported by X - 231 FP All and X - FP Coloc having similar profiles for most CG pathway molecules 232 ( Compare Supplementary Fig . 2a and Fig . 1d ) . Although the fraction of X - FP NoColc was 233 variable , we observed that for molecules that were involved in CG endocytosis , it was 234 typically between 20 - 40 % . On the other hand , for the non - CG pathway molecules like 235 Clathrin , Dynamin and N - WASP it was around 60 - 70 % . Thus , the endocytic sites 236 detected by our assay consisted of two populations wherein one fraction exhibited an 237 accumulation of X - FP while the second fraction failed to show a discernable 238 accumulation . As the removal of the events that did not coincide with the presence of X - 239 FP did not alter their recruitment profile , they were discarded from further analysis . 240 Despite removing , the false positive via both automatic and manual methods there exists a 241 fraction of SecGFP - GPI endocytic sites that fails to associate with X - FP . The reasons for 242 not detecting X - FP at every endocytic event is both a function of both the signal and noise 243 in the data , and it reflects a genuine lack of recruitment at some endocytic events . 244 The number of SecGFP - GPI spots obtained for every molecule , the fraction of X - FP that 245 were associated with the SecGFP - GPI spots , the number of random spots along with the 246 number of cells and experiments is detailed in Table 1 . 247 f ) Plotting : Intensity , corrected for bleaching and normalised , was plotted against time for 248 each event and then averaged across all spots . The average data was compared to average 249 intensity obtained from multiple randomly picked regions from inside the mask of the 250 cell . The random data points ranged from 1800 to 5000 . Y - axis represents fold change 251 over the local background . All the pH pulsing trace used circular mask of radius 3 pixels 252 ( 250 nm ) whose mean intensity was normalized to the local background mask donut of 253 size 6 - 8 pixels ( 470 - 672 nm ) except for IRSp53 ( additional masks of circular radius 2 254 pixels , ( 168 nm ) , donut 3 - 5 pixels ( 250 - 420 nm normalized to background donut 6 - 8 255 pixels ( 470 - 672 nm ) ) and PICK1 which was normalized to background donut 8 - 10 pixels 256 ( 672 - 840 nm ) . 257 g ) Correlation analysis : The average intensity trace over time between two molecules was 258 compared by using MATLAB function called \u2018corrcoef\u2019 that calculates Pearson 259 correlation coefficient wherein the p - value is calculated by t - statistic . 260 Microscope : For population - based endocytic assays quantification of uptake of endocytic 261 tracers , images were obtained using either low magnification objective . For pH pulsing assay 262 the imaging was performed in a TIRF setup with custom designed chamber to maintain 263 constant temperature . The imaging was done at 30\u00b0C . For residence time imaging the 264 imaging was done on a TIRF setup at 37\u00b0C . Confocal imaging was done with spinning disk 265 setup . 266 A detailed list of microscopes used for this manuscript is below . 267 1 . Nikon TE300 - Objectives used in this setup were 20x 0 . 75 NA , 40x 0 . 65 NA . The 268 images collected using an EMCCD camera Ixon ( Andor Technologies ) using \u00b5Manager 269 15 . This microscope was utilized for population level endocytic assay imaging . 270 2 . Nikon TE 2000 equipped with TIRF - Objective used in this setup were 20x 0 . 75 NA , 40x 271 0 . 65 NA and 100x 1 . 49 NA oil . The camera in this setup is CCD cascade camera 272 ( Photometrics Inc . , USA ) . Images were acquired using Metamorph TM and \u00b5Manager . This 273 microscope was utilized for pH pulsing assays , residence time , TIRF based co - 274 localization and population level endocytic assay imaging . 275 3 . Nikon Eclipse Ti equipped with TIRF setup fed by Agilent laser combiner MCL400 276 ( Agilent technologies ) . Laser lines used were 488 and 561 along with epi - fluorescence 277 lamp . Objectives used in this setup were 20x 0 . 75 NA , 40x NA and 100x 1 . 49 NA 278 objective with EMCCD camera ( Photometrics Inc . , USA ) . Image acquisition was done 279 with \u00b5Manager . This microscope was utilized for residence time , TIRF based co - 280 localization and population level endocytic assay imaging . 281 4 . Nikon Eclipse Ti equipped with confocal spinning disk unit ( Yokogawa CSU - 22 scan 282 head ) . Images were collected with 100x 1 . 4 NA oil Nikon objective with an EMCCD 283 camera ( Andor ixon + 897 ) . Images were acquired using Andor iQ2 with Python scripting . 284 This microscope was utilized for confocal imaging . 285 5 . Perkin Elmer equipped with confocal spinning disk unit ( Yokogawa CSU - X1 scan head ) . 286 Images were collected with 100x 1 . 4 NA oil Olympus objective with Hamamatsu 287 IMAGE EM X2 EMCCD camera . Images were acquired using Volocity software . This 288 microscope was utilized for confocal imaging . 289 Supplementary References 290 1 . Mayor , S . , Presley , J . F . & Maxfield , F . R . Sorting of Membrane - Components From 291 Endosomes and Subsequent Recycling To the Cell - Surface Occurs By a Bulk Flow 292 Process . J . Cell Biol . 121 , 1257 \u2013 1269 ( 1993 ) . 293 2 . Gupta , G . D . et al . Population distribution analyses reveal a hierarchy of molecular 294 players underlying parallel endocytic pathways . PLoS One 9 , ( 2014 ) . 295 3 . Gupta , G . D . et al . Analysis of endocytic pathways in Drosophila cells reveals a 296 conserved role for GBF1 in internalization via GEECs . PLoS One 4 , ( 2009 ) . 297 4 . Disanza , A . et al . CDC42 switches IRSp53 from inhibition of actin growth to 298 elongation by clustering of VASP . 2735 \u2013 2750 ( 2013 ) . doi : 10 . 1038 / emboj . 2013 . 208 299 5 . Kumari , S . & Mayor , S . ARF1 is directly involved in dynamin - independent 300 endocytosis . Nat . Cell Biol . 10 , 30 \u2013 41 ( 2008 ) . 301 6 . Coney , L . R . et al . Cloning of a tumor - associated antigen : MOv18 and MOv19 302 antibodies recognize a folate - binding protein . Cancer Res . 51 , 6125 \u2013 6132 ( 1991 ) . 303 7 . Miesenb\u00f6ck , G . , De Angelis , D . A . & Rothman , J . E . Visualizing secretion and 304 synaptic transmission with pH - sensitive green fluorescent proteins . Nature 394 , 192 \u2013 5 305 ( 1998 ) . 306 8 . Rocca , D . L . , Martin , S . , Jenkins , E . L . & Hanley , J . G . Inhibition of Arp2 / 3 - mediated 307 actin polymerization by PICK1 regulates neuronal morphology and AMPA receptor 308 endocytosis . Nat . Cell Biol . 10 , 259 \u2013 71 ( 2008 ) . 309 9 . Chun , J . , Shapovalova , Z . , Dejgaard , S . Y . , Presley , J . F . & Melan\u00e7on , P . 310 Characterization of class I and II ADP - ribosylation factors ( Arfs ) in live cells : GDP - 311 bound class II Arfs associate with the ER - Golgi intermediate compartment 312 independently of GBF1 . Mol . Biol . Cell 19 , 3488 \u2013 500 ( 2008 ) . 313 10 . Bouvet , S . , Golinelli - Cohen , M . - P . , Contremoulins , V . & Jackson , C . L . Targeting of 314 the Arf - GEF GBF1 to lipid droplets and Golgi membranes . J . Cell Sci . 126 , 4794 \u2013 805 315 ( 2013 ) . 316 11 . Riedl , J . et al . Lifeact : a versatile marker to visualize F - actin . Nat . Methods 5 , 605 \u2013 7 317 ( 2008 ) . 318 12 . Taylor , M . J . , Perrais , D . & Merrifield , C . J . A high precision survey of the molecular 319 dynamics of mammalian clathrin - mediated endocytosis . PLoS Biol . 9 , e1000604 320 ( 2011 ) . 321 13 . Barde , I . , Salmon , P . & Trono , D . Production and titration of lentiviral vectors . Curr . 322 Protoc . Neurosci . 1 \u2013 23 ( 2010 ) . doi : 10 . 1002 / 0471142301 . ns0100s37 323 14 . Parthasarathy , R . Rapid , accurate particle tracking by calculation of radial symmetry 324 centers . Nat . Methods 9 , 724 \u2013 726 ( 2012 ) . 325 15 . Stuurman , N . , Edelstein , A . D . , Amodaj , N . , Hoover , K . H . & Ronald , D . Computer 326 Control of Microscopes using \u03bc Manager . 1 \u2013 22 ( 2011 ) . 327 doi : 10 . 1002 / 0471142727 . mb1420s92 . Computer 328",
    "messa2014epsin": "elifesciences . org Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 1 of 25 Epsin deficiency impairs endocytosis by stalling the actin - dependent invagination of endocytic clathrin - coated pits Mirko Messa 1 , Rub\u00e9n Fern\u00e1ndez - Busnadiego 1\u2020 , Elizabeth Wen Sun 1 , Hong Chen 1\u2021 , Heather Czapla 1 , Kristie Wrasman 2 , Yumei Wu 1 , Genevieve Ko 1 , Theodora Ross 3 , Beverly Wendland 2 , Pietro De Camilli 1 * 1 Program in Cellular Neuroscience , Neurodegeneration and Repair , Department of Cell Biology , Howard Hughes Medical Institute , Yale University School of Medicine , New Haven , United States ; 2 Department of Biology , Johns Hopkins University , Baltimore , United States ; 3 Department of Internal Medicine , UT Southwestern Medical Center , Dallas , United States Abstract Epsin is an evolutionarily conserved endocytic clathrin adaptor whose most critical function ( s ) in clathrin coat dynamics remain ( s ) elusive . To elucidate such function ( s ) , we generated embryonic fibroblasts from conditional epsin triple KO mice . Triple KO cells displayed a dramatic cell division defect . Additionally , a robust impairment in clathrin - mediated endocytosis was observed , with an accumulation of early and U - shaped pits . This defect correlated with a perturbation of the coupling between the clathrin coat and the actin cytoskeleton , which we confirmed in a cell - free assay of endocytosis . Our results indicate that a key evolutionary conserved function of epsin , in addition to other roles that include , as we show here , a low affinity interaction with SNAREs , is to help generate the force that leads to invagination and then fission of clathrin - coated pits . DOI : 10 . 7554 / eLife . 03311 . 001 Introduction Clathrin - mediated endocytosis involves a complex set of factors besides clathrin itself and the classical clathrin adaptors . These factors help coordinate nucleation of the clathrin coat with cargo selection , membrane invagination and fission ( Schmid , 1997 ; Slepnev and De Camilli , 2000 ; Kaksonen et al . , 2003 ; Merrifield et al . , 2005 ; Ferguson and De Camilli , 2012 ) . An important player in these pro - cesses is epsin , the collective name for a family of evolutionarily conserved clathrin - associated pro - teins ( Chen et al . , 1998 ; Wendland , 1999 , 2002 ; Ford et al . , 2002 ; Chen and De Camilli , 2005 ) , which in mammals is represented by three isoforms : epsin 1 , 2 , and 3 ( encoded by the Epn1 , Epn2 , and Epn3 genes , respectively [ Ko et al . , 2010 ] ) . Epsin was identified as a major interactor of Eps15 ( Chen et al . , 1998 ) , another clathrin coat associated protein . It comprises a membrane binding N - terminal ENTH ( Epsin N - Terminal Homology ) domain , which is followed by ubiquitin - interacting motifs ( UIMs [ Polo et al . , 2002 ] ) and a long sequence ( tail ) predicted to be primarily unfolded and flexible ( Wendland , 2002 ) . The core of the ENTH domain is preceded by a short sequence that is unfolded in solution but folds into an amphipathic \u03b1 - helix upon binding to PI ( 4 , 5 ) P 2 . The hydrophobic portion of the helix partially penetrates the bilayer , thus conferring membrane curvature generation and sensing properties to the protein ( Itoh et al . , 2001 ; Ford et al . , 2002 ) . Epsin ' s disordered tail binds compo - nents of the clathrin coat via multiple short amino acid motifs : \u2018clathrin boxes\u2019 bind clathrin , DPW / F motifs bind the appendage domain of AP - 2 , and NPF motifs bind the EH domains of Eps15 and intersectin ( Chen et al . , 1998 ; Rosenthal et al . , 1999 ; Drake , 2000 ; Shih et al . , 2002 ; Overstreet et al . , 2003 ) . * For correspondence : pietro . decamilli @ yale . edu Present address : \u2020 Max - Planck - Institut f\u00fcr Biochemie , Martinsried , Germany ; \u2021 Cardiovascular Biology Research Program , Oklahoma Medical Research Foundation , Oklahoma City , United States Competing interests : The authors declare that no competing interests exist . Funding : See page 21 Received : 07 May 2014 Accepted : 12 August 2014 Published : 13 August 2014 Reviewing editor : Suzanne R Pfeffer , Stanford University , United States Copyright Messa et al . This article is distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use and redistribution provided that the original author and source are credited . RESEARCH ARTICLE Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 2 of 25 Research article As epsin binds ubiquitin and genetically interacts with enzymes of ubiquitin metabolism ( Cadavid et al . , 2000 ; Chen et al . , 2002 ; Polo et al . , 2002 ; Shih et al . , 2002 ; Chen et al . , 2003 ; Sigismund et al . , 2005 ) , it was proposed to function as a clathrin adaptor for ubiquitinated cargo . Strong evidence for such a role came from the demonstration of Notch signaling defects in epsin ( liquid facets ) mutant flies , as Notch signaling is critically dependent upon ubiquitin - dependent endocytosis of Notch ligands ( Overstreet et al . , 2003 ; Xie et al . , 2012 ) . However , other findings pointed to a general house - keeping role of epsin in clathrin - mediated endocytosis . Absence of the two epsins ( Ent1 and Ent2 ) in yeast is lethal , while hypomorphic Ent1 or Ent2 mutations result in defects in endocytosis and actin dynamics ( Wendland , 1999 ; Aguilar et al . , 2003 ; Skruzny et al . , 2012 ) . Impairments in clathrin and actin function were also observed in epsin null Dictyostelium mutants ( Brady et al . , 2008 ; 2010 ) . In both these unicellular organisms , epsin functions in close cooperation with Sla2 / Hip1R , another evolu - tionarily conserved clathrin accessory factor ( Brady et al . , 2008 ; 2010 ; Skruzny et al . , 2012 ) . However , a link between epsin and Hip1R in metazoan cells has not been reported . Hip1 family members ( Hip1 and Hip1R in mammals ) comprise an N - terminal ANTH domain fol - lowed by unfolded regions that bracket a coiled - coil region and a C - terminal THATCH ( talin - HIP1 / R / Sla2p actin - tethering C - terminal homology ) domain ( Engqvist - Goldstein et al . , 1999 ; Wilbur et al . , 2008 ; Skruzny et al . , 2012 ) . The coiled - coil region can homo - heterodimerize and also binds clathrin light chain ( Engqvist - Goldstein et al . , 2001 ; Metzler et al . , 2001 ; Legendre - Guillemin et al . , 2002 ; Gottfried et al . , 2010 ) . The THATCH domain is an actin - binding module ( Yang et al . , 1999 ; Engqvist - Goldstein et al . , 2001 ; Brett et al . , 2006 ; Wilbur et al . , 2008 ) . Accordingly , Sla2 / Hip1R binds actin and is thought to function as a major link between the clathrin coat and actin . Studies in yeast have additionally shown that the ENTH domain of epsin and the ANTH domain of Sla2 interact with each other , and the two proteins function together in providing a link between the endocytic coat and the actin cytoskeleton ( Skruzny et al . , 2012 ) . eLife digest Clathrin - dependent endocytosis is one of the mechanisms used by cells to internalize specific proteins ( cargo ) from their surface . First , the cargo interacts with adaptor proteins that help cluster them in the cell ' s outer membrane , called the plasma membrane . This causes the protein clathrin to assemble into a lattice at the cytosolic side of the plasma membrane and deform the membrane into a pit . The pit grows deeper over time as more clathrin molecules assemble , eventually resulting in a deeply invaginated clathrin - coated pit that encloses the cargo to be taken up by the cell . The clathrin - coated pit then pinches off inside the cell in a process called fission to form a bubble - like structure called a vesicle , which transports the molecule to its destination . The deep invagination of clathrin - coated pits that leads to fission is assisted by actin , a protein that assembles into filaments that are suggested to generate the forces needed for this process . Many other factors are also involved . One of them is epsin , the collective name for a family of three very similar proteins in mammalian cells . Epsin binds to several other proteins implicated in clathrin - dependent endocytosis , including clathrin itself , and to plasma membrane proteins specifically \u2018tagged\u2019 for internalization . In addition , a portion of the epsin molecule can insert into the plasma membrane and help it to curve , which is important for forming the invaginated pit . However , due to the number of possible functions epsin could perform , its main role has remained elusive . Messa et al . created mouse cells that lack all three epsin proteins . Although these cells can form clathrin - coated pits , they struggle to develop into vesicles . The normal linking of the actin filaments to the clathrin coat does not occur , and another protein called Hip1R that also participates in clathrin - mediated endocytosis and links clathrin to actin , no longer accumulates at the clathrin - coated pits . Messa et al . also find that epsins can bind directly to actin . Overall , these results suggest that a main role of epsin is to help actin interact with the clathrin - coated pits and generate the force required for a pit to develop into a vesicle . However , epsin also performs many other roles , including recruiting a membrane protein ( a so - called SNARE ) that directs the fate of the vesicle to the clathrin - coated pit . Additionally , Messa et al . find that cells lacking all three epsins have problems dividing correctly . More research is required to establish whether this effect is also due to epsin ' s interaction with the cell ' s actin cytoskeleton . DOI : 10 . 7554 / eLife . 03311 . 002 Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 3 of 25 Research article In addition to roles of epsin mediated by protein \u2013 protein interactions , membrane remodeling prop - erties resulting from the amphipathic helix at the N - terminus of its ENTH domain have been implicated in the clathrin - dependent endocytic reaction . In vitro studies showed that this helix confers , upon the ENTH domain , the property to induce bilayer curvature and even to fragment bilayer tubules into vesicles , thus pointing to a potential role of the epsin in fission ( Itoh et al . , 2001 ; Ford et al . , 2002 ; Boucrot et al . , 2012 ) . Surprisingly , in view of this evidence for an important housekeeping role of epsin in endocytosis , the germline knockout ( KO ) of the mouse Epn1 and Epn2 genes that encode the two major ubiqui - tously expressed mammalian epsins , epsin 1 and 2 , did not block the early embryonic development ( Chen et al . , 2009 ) . Arrest of embryonic development occurred only at E9 . 5 \u2013 E10 , with a pattern sug - gestive of impaired Notch signaling , while no obvious defects in clathrin - mediated endocytosis were observed in fibroblasts derived from these embryos ( Chen et al . , 2009 ) . Moreover , studies of epsin 1 and 2 conditional double KO endothelial cells revealed a selective defect in the internalization of ubiq - uitinated VEGF receptor ( Pasula et al . , 2012 ) . However , a recent study based on RNAi - mediated knock - down ( KD ) in fibroblastic cells reported that the KD of all the three epsins produces a global impairment of clathrin - mediated endocytosis , which was attributed to a defect of the fission reaction ( Boucrot et al . , 2012 ) . The goal of the present study was to provide conclusive evidence about the function ( s ) and sites of action of epsin in endocytosis using a gene KO strategy to completely eliminate all epsins . Our results , which capitalize on triple KO ( TKO ) cells generated from conditional epsin TKO mice , show that epsin provides a link between the clathrin coat and actin and is needed for the transition of pits from a shallow to a deeply invaginated state . As in unicellular organisms , epsin acts in concert with Hip1R but with differences from yeast to mammals . An additional function of epsin is a low affinity interaction of its ENTH domain with synaptobrevin 2 / VAMP2 ( Syb2 ) that may help ensure the presence of a vesic - ular SNARE in the budding vesicle . Results Generation of conditional epsin triple knockout cells As the germline deletion of even only two Epn genes results in embryonic lethality , a conditional approach was used to generate Epn1 , Epn2 , and Epn3 triple KO cells . Towards this goal , Epn1 loxP / loxP ( Figure 1\u2014figure supplement 1A , see also [ Pasula et al . , 2012 ] ) mice were crossed with Epn2 ( Chen et al . , 2009 ) and Epn3 KO ( Ko et al . , 2010 ) mice to generate Epn1 loxP / loxP ; Epn2 \u2212 / \u2212 ; Epn3 \u2212 / \u2212 animals , which were viable and fertile with no obvious pathological phenotypes . These mice were subsequently interbred with mice transgenic for 4 - hydroxy - tamoxifen ( OHT ) - inducible Cre recombinase [ Cre - ER , Badea et al . , 2003 ] to obtain Epn1 loxP / loxP ; Epn2 \u2212 / \u2212 ; Epn3 \u2212 / \u2212 ; Cre - ER + / 0 animals . These mice did not exhibit obvious defects either . Conditional epsin TKO mouse embryonic fibroblasts ( MEFs ) were derived from these animals via treatment with OHT , whose action was confirmed by the translocation of Cre into the nucleus ( Figure 1\u2014figure supplement 1B ) , and typically examined after 7 days of OHT treatment ( after 9 days of treatment cell death started to occur ) . Treated MEFs derived from the same litters but harboring a wild - type ( WT ) epsin 1 allele did not exhibit any obvious differences from OHT - treated WT fibroblasts in terms of proliferation , endocytosis and actin organization . In the experiments described below OHT - treated WT MEFs were used as controls . Immunoblotting with epsin isoform - specific antibodies demonstrated the near complete disap - pearance of epsin 1 from conditional TKO cell extracts in response to OHT ( Figure 1A ) and confirmed the absence of epsin 2 and 3 ( Figure 1B ) . The extremely small amount of residual epsin 1 was likely explained by delayed gene recombination in a few cells , where residual epsin 1 immunoreactivity , much lower than in controls , was observed by immunofluorescence . A C - terminal epsin 2 fragment ( MW about 40 kDa ) was detected in cells containing the epsin 2 KO allele , possibly reflecting an alter - native start site ( Figure 1\u2014figure supplement 1C ) . However , the KD of this fragment by RNAi in TKO cells did not produce phenotypic changes in addition to the ones described below , indicating that it does not play a relevant role . OHT - dependent loss of epsin 1 expression was also validated by the anti - epsin 1 immunofluorescence , as the typical punctate epsin 1 signal ( reflecting clathrin - coated pits [ Chen et al . , 1998 ] ) was completely absent ( Figure 1C ) . We also mated mice with mutations in all the three Epn genes with nestin - Cre transgenic mice ( Tronche et al . , 1999 ) to obtain brain - specific TKO mice ( genotype : Epn1 loxP / loxP ; Epn2 \u2212 / \u2212 ; Epn3 \u2212 / \u2212 ; Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 4 of 25 Research article Figure 1 . Mitotic defects in epsin TKO fibroblasts . ( A ) Anti - epsin 1 immunoblot shows the disappearance of epsin 1 from Epn 1 loxP / loxP ; Epn 2 \u2212 / \u2212 ; Epn 3 \u2212 / \u2212 ; Cre - ER + / 0 cells after 7 - day treatment with 4 - hydroxy - tamoxifen ( OHT ) . Tubulin was used as a loading control ( top ) . Densitometric analysis of the epsin 1 band during OHT - treatment ( bottom ) . ( B ) Anti - epsin immunoblots with isoform - specific antibodies showing OHT - treatment for 7 days results in the loss of epsin 1 in epsin 2 / 3 double knock - out ( DKO ) thus generating triple KO ( TKO ) cells . Wild type ( WT ) cell lysate was Figure 1 . Continued on next page Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 5 of 25 Research article Nestin - Cre + / 0 ) and their controls with a WT epsin 1 allele ( Epn1 loxP / + ; Epn2 \u2212 / \u2212 ; Epn3 \u2212 / \u2212 ; Nestin - CRE + / 0 ) . Nestin - Cre epsin TKO animals were born , but at a lower than expected Mendelian ratio . A few days after birth they began to develop locomotor dysfunction , failed to gain weight ( Figure 1\u2014figure supplement 1D ) , and rapidly deteriorated , typically dying before the end of the fourth week . Western blot analysis of extracts of their brains at the beginning of the fourth week , confirmed the absence of all the three epsin genes ( Figure 1\u2014figure supplement 1E ) . In this study , we focused on the impact of the lack of epsin on basic cellular properties using MEFs and cell - free assays . Brain specific epsin TKO mice will be characterized in a future study , but we made use of their brains as the source of epsin TKO brain cytosol in the cell - free assay described below . Epsin TKO cells have defects in cytokinesis Microscopy of the TKO MEFs revealed them to be abnormally large compared to WT cells ( contour in Figure 1C ) . Furthermore , these cells contained multiple clustered nuclei or very large and abnormally shaped nuclei , as shown by both DAPI staining and Cre immunofluorescence ( Figure 1D , Figure 1\u2014 figure supplement 1B , F ) , in contrast to the single nuclei present in the smaller WT cells . This finding could be explained by the reported role of epsin 1 and 2 in mitotic spindle organization ( Liu and Zheng , 2009 ) , a defect that can result in abnormal chromosome segregation and impaired cytokinesis . Accordingly , analysis of WT and TKO cells during 48 - hours of OHT treatment revealed that the increase in either cell size or size / number of nuclei of TKO cells was accompanied by a strong cell number reduction ( Figure 1E ) . Additionally , immunostaining of cells synchronized by double - thymidine treat - ment ( Banfalvi , 2011 ) for anti - Aurora B kinase , a marker of midbodies of telophase ( [ Banerjee et al . , 2014 ] inset of Figure 1D ) , showed a striking decrease of cytokinesis profiles in TKO ( Figure 1F ) . As the focus of our study is the role of epsin in endocytosis , we did not explore mechanistic aspects of this cytokinesis defect further . We note , however , that other endocytic proteins , including clathrin itself , have been implicated in the organization of mitotic scaffolds ( Royle , 2013 ; Kaur et al . , 2014 ) . Endocytic delay at early / mid - stage clathrin - coated pits We next analyzed the impact of the complete absence of the three epsins on clathrin - mediated endo - cytosis . Immunostaining of TKO cells for \u03b1 - adaptin ( a subunit of the clathrin adaptor complex AP - 2 ) and for clathrin - light chain ( CLC ) showed an increase in the density of clathrin - coated pits relative to controls ( Figure 2A \u2013 D ) . Additionally , most pits occurred in small clusters rather than as single - puncta ( insets in Figure 2A \u2013 D ) . Total levels of clathrin and \u03b1 - adaptin , however , were not changed ( Figure 2\u2014 figure supplement 1A ) , suggesting that the increase in pits reflects an increase in the pool of assem - bled clathrin coats ( Figure 2E ) . Clathrin - coated pit dynamic was also altered , as live imaging with spinning disk confocal microscopy of cells transfected with \u03bc 2 - adaptin - GFP ( another subunit of the AP - 2 complex ) revealed that the turnover rate of pits in TKO cells was much lower than in WT cells ( Figure 2F , G ) . In agreement with these morphological changes , an impairment of the internalization of fluorescent transferrin , a cargo of endocytic clathrin - coated pits , was also observed in TKO cells . Following preincu - bation with Alexa594 - transferrin on ice and subsequent incubation at 37\u00b0C for 15 min , the bulk of transferrin remained at the cell surface in TKO cells , while remaining transferrin was intracellular in used as a control . ( C ) Anti - epsin 1 immunofluorescence shows that the typical punctate epsin 1 signal of WT cells ( top ) was completely absent in TKO cells ( bottom ) . The perimeter of the TKO cell is indicated by a dotted white line and demonstrates a very large size relative to WT cells . The insets show higher magnification of the boxed regions . ( D ) DAPI staining showing single nuclei in WT and multiple nuclei in a TKO cell . The inset of the WT field shows the accumulation of AuroraB kinase immunoreactivity at the mid - body during cytokinesis . ( E ) The increase in cell number during a 3 - day incubation is lower in TKO cells . Cells were counted at day 7 and 9 after addition of OHT ( * * p < 0 . 01 , Student ' s t test , n = 3 experiments ) . ( F ) As shown by a morphometric analysis , cytokinesis events are only rarely observed in TKO cells . Scale bar represents 10 \u03bc m . Data are represented as mean \u00b1 SEM . See also Figure 1\u2014figure supplement 1 . DOI : 10 . 7554 / eLife . 03311 . 003 The following figure supplement is available for figure 1 : Figure supplement 1 . The generation of triple KO cells , nestin Cre brain specific triple KO mice , and the observation of nuclear defect in TKO cells . DOI : 10 . 7554 / eLife . 03311 . 004 Figure 1 . Continued Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 6 of 25 Research article Figure 2 . Absence of epsin stalls endocytic clathrin - coated pit maturation at an early stage . Confocal microscopy . ( A \u2013 D ) Immunofluorescence staining for \u03b1 - adaptin ( A and B ) and clathrin light chain ( C and D ) indicates an increase of clathrin - coated pits ( CCPs ) number in cells that lack all three epsins . In Figure 2 . Continued on next page Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 7 of 25 Research article control cells ( Figure 2H , I ) . This change correlated with a major shift of overexpressed and endogenous transferrin receptor from a punctate localization in the cytoplasm ( WT cells ) to a plasma membrane localization ( TKO cells ) , as demonstrated by both immunofluorescence ( Figure 2J , K ) and a cell surface biotinylation assay ( Figure 2L ) . Transferrin internalization impairment could be rescued by electropo - ration of epsin1 - GFP cDNA ( Figure 2\u2014figure supplement 1B \u2013 D ) . Likewise , a major shift of the local - ization of synaptobrevin 2 / VAMP2 ( Syb2 ) , a vesicular SNARE that is internalized by clathrin - mediated endocytosis , was observed in TKO MEFs stably expressing Syb2 - HA , ( this construct harbors the HA epitope at the very short C - terminal non - cytosolic portion of Syb2 ) . Total Syb2 immunoreactivity had a predominant punctate intracellular distribution in control MEFs , and an additional strong plasma membrane localization in TKO ( Figure 2M , O ) . The presence of an abundant surface exposed pool of Syb2 selectively in TKO MEFs was confirmed by the surface immunofluorescence for the HA epitope ( Figure 2N , P ) and by a cell surface biotinylation assay ( Figure 2Q , R ) . When cells were examined by electron microscopy to assess the stage at which clathrin - mediated endocytosis is impaired in TKO cells , the major increase was observed for shallow and U - shaped pits ( Figure 2S , T ) . This observation may seem in contrast with the reported accumulation of the so - called multiheaded coated - structures at the plasma membrane of cells subjected to RNAi - dependent KD of all the three epsins ( Boucrot et al . , 2012 ) . Such a phenotype was interpreted as a defect in the fission reaction . However , multiheaded - coated structures , which were not clearly detected in our cells ( Figure 2S , T ) , could instead reflect clustering of shallow / intermediate stage pits at the pre - fission stage . In agreement with an arrest at early stage , dynamin 2 , endophillin 2 , and myosin 1E ( a myosin implicated in clathrin - mediated endocytosis [ Cheng et al . , 2012 ] ) , three proteins whose localization at pits peaks at the time of fission ( Taylor et al . , 2011 ) were present at the subset of very late pits in WT cells , while in TKO cells they had a diffuse cytosolic distribution ( Figure 2U ) . Enhanced actin accumulation at epsin - deficient clathrin - coated pits A stalling of endocytic clathrin - coated pits prior to the deep invagination stage has been observed upon manipulations affecting actin nucleation and dynamics ( Shupliakov et al . , 2002 ; Ferguson et al . , 2009 ) , leading us to explore actin localization in TKO cells . Phalloidin staining for F - actin revealed only few of the actin stress fibers typically observed in control cells ( Figure 3A , B ) and a strong accumulation of F - actin foci in TKO cells ( Figure 3B , D ) , sometimes in the form of elongated curved structure at the cell cortex ( inset in Figure 3B ) . Similar results were observed upon immu - nostaining for Arp 2 / 3 ( an actin nucleating complex that functions with N - WASP [ Figure 3E ] ) . A differ - ence in actin organization relative to WT cells was also observed by live total internal reflection TKO cells , clathrin - coated pits generally occur in small clusters . Insets show the boxed regions at high magnification . Note the large size of TKO cells relative to control . ( E ) CCP number in WT and TKO cells as assessed by \u03b1 - adaptin immunofluorescence ( * * * p < 0 . 001 , Student ' s t test , n = 10 cells / genotype ) . ( F ) Kymographs from a time series of WT and TKO cell expressing \u03bc 2 - adaptin - GFP . Each line represents a single \u03bc 2 - GFP spot . Note the short length of the lines for WT , reflecting the turnover of the pits and the continuous lines in TKO cells , reflecting an arrest of the pit maturation . ( G ) Clathrin - coated pit turnover ( appearance and disappearance events ) as analyzed by spinning - disk confocal imaging of \u03bc 2 - adaptin - GFP fluorescence in WT and TKO cells ( n = 5 cells / genotype ) . ( H and I ) Impaired uptake of pre - bound Alexa594 - transferrin ( Tf ) in TKO cells during a 15 - min incubation . In WT , the bulk of Tf was internalized , while in TKO Tf remained at the cell surface . ( J and K ) Transferrin receptor ( TfR ) - GFP predominantly localizes in intracellular vesicles in WT but at the cell surface in TKO cells . ( L ) A surface biotinylation assay reveals elevated amounts of endogenously expressed TfR at the plasma membrane of TKO cells relative to WT , as assessed by anti - TfR immunoblotting of streptavidin affinity - purified material . ( M \u2013 P ) Increased surface localization of stably expressed Syb2 - HA in TKO fibroblasts as shown by total ( M \u2013 O ) and surface - only ( N \u2013 P ) immunofluorescence . ( Q and R ) A surface biotinylation performed as in ( L ) demonstrating an increased fraction of cell surface exposed Syb2 - HA in TKO cells relative to WT ( * * p < 0 . 01 , Student ' s t test , n = 4 experiments , Surf : surface , Int : internal ) . ( S and T ) Representative electron microscopy images of different stage endocytic clathrin - coated intermediates in TKO cells ( S ) and quantification of the corresponding stages ( T , * * p < 0 . 01 , * * * p < 0 . 001 , n = 33 cells / genotype , one - way ANOVA ) . ( U ) Comparative analysis of the localization of clathrin immunoreactivity ( CLC ) with the localization of dynamin 2 , endophillin 2 , and myosin 1E immunoreactivities . In WT cells , these three proteins co - localize with a subset of clathrin - coated pits ( examples are indicated by small white arrows ) , which represent late - stage pits . In TKO cells , where more numerous clathrin - coated pits are observed , the punctate localization of dynamin 2 , endophillin 2 , and myosin 1E is completely lost . Scale bars : 10 \u03bc m for ( A \u2013 D , H \u2013 K ) , 20 \u03bc m for ( M \u2013 P ) , 5 \u03bc m for ( U ) , and 200 nm for ( S ) . In E , G , R , and T black bars indicate WT and red bars epsin TKO . See also Figure 2\u2014figure supplement 1 . DOI : 10 . 7554 / eLife . 03311 . 005 The following figure supplement is available for figure 2 : Figure supplement 1 . Endocytic defect in epsin TKO cells is not due to alteration in endocytic protein levels and it is rescued by epsin1 \u2013 GFP . DOI : 10 . 7554 / eLife . 03311 . 006 Figure 2 . Continued Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 8 of 25 Research article ( TIRF ) microscopy of TKO cells , co - expressing the calponin homology domain of utrophin fused to mCherry ( CH Utr \u2013 mCh ) and CLC fused to GFP ( CLC \u2013 GFP ) , as there was a general increase in the pool of actin associated with clathrin spots ( Figure 3F ) . The expression of full - length epsin1 \u2013 GFP ( Figure 2\u2014figure supplement 1B \u2013 D ) rescued actin cytoskeleton changes ( Figure 3C , D ) . Epsin is required for the recruitment of Hip1R at endocytic pits Interestingly , while no significant alterations of the levels of several major components of the clathrin - dependent endocytic machinery were observed in TKO cells ( Figure 2\u2014figure supplement 1A ) , the level of Hip1R was increased ( Figure 4A ) . This observation strengthened the idea that epsin and Hip1R are functionally interconnected , as suggested by studies in yeast and Dictyostelium ( Baggett et al . , 2003 ; Brady et al . , 2010 ; Skruzny et al . , 2012 ) , and prompted us to explore the interplay of epsin and Hip1R . Pull - down experiments from rat brain homogenate , using ENTH domain of epsin 1 as bait revealed an enrichment of Hip1R in the affinity - purified material in the sample also containing PI ( 4 , 5 ) P 2 ( Figure 4B ) . Additionally , the typical clathrin - coated pit - like punctate localization ( Engqvist - Goldstein et al . , 1999 ) of endogenous ( immunofluorescence , Figure 4C ) and exogenous ( TIRF microscopy of Hip1R - GFP , Figure 4F ) Hip1R was replaced in TKO cells by a diffuse signal ( Figure 4D , F ) . Expression of epsin1 - GFP in TKO cells ( Figure 2\u2014figure supplement 1B \u2013 D ) rescued Hip1R localization ( Figure 4E ) . We conclude that the localization of HipR at endocytic clathrin - coated pits is heavily dependent on epsin . Figure 3 . Abnormal actin distribution in epsin deficient cells . ( A \u2013 C ) Confocal microscopy images . Phalloidin staining of WT and epsin TKO cells shows a major loss of stress fibers with a corresponding accumulation of elongated F - actin foci ( see inset ) in TKO cells . These changes were rescued by the expression of epsin1 \u2013 GFP ( C ) . ( D ) Quantification of the actin foci shown in ( A \u2013 C , * * * p < 0 . 001 , n = 8 cells / conditions , one - way ANOVA ) . ( E ) Phalloidin staining and Arp 2 / 3 immunoreactivity in a WT and a TKO cell showing the co - localization of Arp 2 / 3 with the actin foci , as visualized by confocal microscopy . ( F ) TIRF microscopy of WT and TKO cells expressing clathrin light chain \u2013 GFP ( CLC \u2013 GFP ) and the F - actin binding protein utrophin \u2013 mCherry ( CH Utr \u2013 mCh ) . Note the increase in F - actin ( CH Utr \u2013 mCh signal ) typically surrounding the clathrin - coated pits , in the cortical region ( TIRF plane ) of the TKO cell . Virtually all pits are positive for CH Utr \u2013 mCh . Scale bars : 10 \u03bc m for ( A \u2013 C ) , 5 \u03bc m for ( E and F ) . Data are represented as mean \u00b1 SEM . DOI : 10 . 7554 / eLife . 03311 . 007 Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 9 of 25 Research article Figure 4 . Epsin is required for the recruitment of Hip1R at the endocytic clathrin - coated pits . ( A ) Immunoblot analysis of Hip1R ( left ) shows an increase ( as quantified in the right ) of its levels in epsin TKO cells . ( B ) Pull down experiments from rat brain homogenate , using ENTH \u2013 GST or GST alone as bait , shows the affinity - purification of Hip1R in the presence of diC8 - PI ( 4 , 5 ) P 2 [ PI ( 4 , 5 ) P 2 , top ] . Coomassie blue stained gel of the baits ( bottom ) . ( C \u2013 E ) Anti - Hip1R immunofluorescence of representative WT and TKO cells . The clathrin - coated pit pattern of Hip1R in WT ( see also high magnification insets ) was replaced by a diffuse localization in the TKO cell , but was rescued by expression of epsin1 \u2013 GFP . ( F ) WT and TKO cells were transfected with Hip1R \u2013 GFP and imaged by live TIRF microscopy ( fluorescence is shown in black ) . The punctate accumulation ( clathrin - coated pits ) of Hip1R at the cortex of WT was lost in the TKO cell . ( G and H ) siRNA - mediated knockdown of Hip1R ( G , Ctrl : non - transfected control ; NC : scramble control ; O1 : Hip1R specific double - stranded siRNA ) does not affect epsin localization in HeLa cells as shown by epsin immunofluorescence ( H ) . Scale bars represent 10 \u03bc m . Data are represented as mean \u00b1 SEM . DOI : 10 . 7554 / eLife . 03311 . 008 Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 10 of 25 Research article In contrast , the localization of epsin at pits was independent of Hip1R , as Hip1R KD in HeLa cells by RNAi ( Figure 4G ) did not affect epsin localization ( Figure 4H ) . Absence of epsin impairs the coupling of clathrin - dependent budding to actin growth in a cell - free assay To address the role of epsin at the interface of endocytic clathrin coats and actin dynamics , we turned to a cell - free system developed in our laboratory ( Wu et al . , 2010 ; Wu and De Camilli , 2012 ) . The approach consists of incubating glass - attached plasma membrane sheets , produced by cell sonication , with purified brain cytosol in the presence of ATP and the non - hydrolyzable GTP analogue GTP \u03b3 S . Under these conditions , where endocytic invagination is not compensated by internal pressure , the incubation triggers the formation of endocytic tubules that are capped by clathrin - coated pits and are surrounded by F - BAR proteins of the FBP17 family . Consistent with the role of these proteins in actin nucleation , elongation of the tubules is actin - dependent ( Wu et al . , 2010 ; Wu and De Camilli , 2012 ) . Plasma membranes obtained by sonication of PTK2 cells expressing PM - anchored GFP exhibit a homogeneous fluorescence when kept in cytosolic buffer ( Wu and De Camilli , 2012 ) . Upon incuba - tion with WT mouse brain cytosol and nucleotides , growth of the tubular invaginations was revealed in en face views by the resulting GFP puncta ( Figure 5\u2014figure supplement 1A ) , and \u2018in side\u2019 views by the growth of short columns of GFP fluorescence perpendicular to the plane of the coverslip ( Figure 5A , left PM - GFP panel ) . The presence of clathrin at the tips of the tubules was confirmed by immunofluo - rescence ( Figure 5A , left clathrin panel ) . Clathrin signal had an elongated appearance \u2018in side\u2019 views , given the low resolution of confocal microscopy in the Z dimension . Epsin immunoreactivity precisely overlapped with clathrin ( Figure 5A , left epsin panel ) indicating epsin localization throughout the clathrin coat . Clathrin - capped invaginations also formed on plasma membrane sheets incubated with epsin TKO brain cytosol ( Figure 5A , right panels and Figure 5\u2014figure supplement 1B ) . However , important dif - ferences relative to incubations with control cytosol were observed . In membrane sheets incubated with control brain cytosol , F - actin polymerized in alignment with the invaginations thus generating a well - organized scaffold that contributes to maintain their parallel orientation perpendicular to the plasma membrane ( Figure 5B , left panels ) . On the contrary , in membranes incubated with TKO cytosol , the F - actin network around the invaginations was exaggerated and disorganized ( Figure 5B , right panels ) , leading to clumping of the tips of the invaginations in a meshwork of actin . As a result of this clumping , invaginations often had an oblique orientation ( Figure 5A , B , right panels and Figure 5\u2014figure supplement 1B ) . Such disorganization was confirmed by electron microscopic observations of sections cut parallel to the substrate . Regularly spaced cross - sectioned tubules interspersed within an actin meshwork were observed with WT cytosol , but numerous obliquely cut tubules embedded in a dense actin matrix were visible with TKO cytosol ( Figure 5\u2014figure supplement 1C , D ) . Another striking modification related to an abnormal actin cytoskeleton was the mislocalization of myosin 1E . As shown by immunofluorescence , this protein was localized at the bottom of the tubular invaginations in sheets incubated with WT cytosol , but at the tips in sheets incubated with epsin TKO brain cytosol ( Figure 5C and higher magnifications ) . Epsin is required for the recruitment of Hip1R to endocytic invaginations in the cell - free assay The differences in the cell - free assay using WT vs TKO cytosol were consistent with abnormal actin nucleation observed at clathrin - coated pits of intact TKO cells . Furthermore , in agreement with these observations , Hip1R immunoreactivity was present at the tips of the invaginations in sheet prepara - tions incubated with WT cytosol , but was largely absent from the tubules incubated with TKO cytosol ( Figure 5D ) . This was in spite of the higher levels of Hip1R in the TKO brain cytosol ( Figure 5\u2014figure supplement 1E ) in agreement with the higher levels of Hip1R in TKO fibroblasts ( Figure 4A ) . Thus , also in this system , Hip1R recruitment requires epsin . We also prepared cytosol from brains of mice lacking Hip1R and its close homologue Hip1 ( Hip1 / Hip1R double KO ( DKO ) mice [ Bradley et al . , 2007 ] ) . These mice are typically dwarfed , afflicted with severe spinal defects , and die in early adulthood ( Bradley et al . , 2007 ) . When membrane sheets were incubated with Hip1 / Hip1R DKO brain cytosol , differences from WT were observed that were qualita - tively similar to those obtained with the epsin TKO cytosol , but quantitatively milder ( Figure 5E , F ) . Tubular invaginations were longer and some exaggerated and disorganized actin was also observed Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 11 of 25 Research article Figure 5 . Coupling of clathrin - dependent budding to actin dynamics in a cell - free assay is perturbed by the absence of epsin . Plasma membrane sheets of PTK2 cells expressing PM - anchored GFP ( PM \u2013 GFP ) were incubated in the presence of WT , epsin TKO or Hip1 / Hip1R double KO ( DKO ) brain cytosol , and nucleotides as described in Wu et al . ( 2010 ) , fixed , immunostained , and observed by confocal microscopy . Views orthogonal to the substrate are shown . ( A ) Immunofluorescence staining for clathrin light chain and epsin 1 . The GFP - positive columns represent narrow tubular invaginations of the plasma membrane capped by clathrin - coated pits , as revealed by the presence of clathrin and epsin immunoreactivity ( Wu et al . , 2010 ) . In the control preparation ( left ) tubules are straight and perpendicular to the substrate , while in the preparation incubated with TKO cytosol ( right , note the absence of epsin immunoreactivity ) they have a more disordered orientation . ( B ) Phalloidin staining of the membrane sheets revealing a well - organized actin scaffold around the tubules after incubation with WT cytosol ( left ) and an exaggerated and disorganized F - actin network in sheets incubated with TKO cytosol ( right ) . ( C ) Myosin 1E immunoreactivity localizes at the bottom of tubular invaginations in sheets incubated with WT cytosol ( left ) , and at the tip of the invaginations after incubation with TKO cytosol ( right ) . Details of the sheets are shown at high magnification below the main panels . ( D ) Hip1R immunoreactivity is present at the tip of tubular invaginations in the control samples ( left ) , but is absent in sheets incubated with epsin TKO cytosol ( right ) . ( E ) Quantification of F - actin polymerization upon incubation with the cytosol of the three tested genotypes ( WT , epsin TKO , and Hip1 / Hip1R DKO ) . The average phalloidin fluorescence per unit area of membrane sheets was calculated ( n = 10 sheets / conditions , * * * * p < 0 . 0001 , one - way ANOVA ) . ( F ) Phalloidin staining of sheets incubated with Hip1 / Hip1R double KO ( right ) cytosol reveals an exaggerated and disorganized F - actin network relative to WT ( left ) , but not as prominent as that observed in preparations incubated with epsin TKO cytosol . ( G ) Immunofluorescence staining for epsin 1 and Hip1R shows that the presence of epsin 1 at the tips of the invagination is not strongly modified by the absence of Hip1 and Hip1R . ( H ) The disordered C - terminal tail of epsin binds F - actin . GST fused epsin 1 fragments ( ENTH domain , DPW , NPF , and DPW - NPF containing regions ) were incubated ( 5 \u03bc M final concentration ) with previously polymerized F - actin ( 15 \u03bc M ) and then subjected to ultracentrifugation followed by SDS - PAGE and anti - GST immunoblotting of the supernatant ( S ) and pellet ( P ) materials . Scale bars : 5 \u03bc m . Data are represented as mean \u00b1 SEM . See also Figure 5\u2014figure supplement 1 . DOI : 10 . 7554 / eLife . 03311 . 009 The following figure supplement is available for figure 5 : Figure supplement 1 . Nestin Cre epsin TKO brain cytosol has increased Hip1R level and disrupted membrane tubulation in a cell - free assay . DOI : 10 . 7554 / eLife . 03311 . 010 Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 12 of 25 Research article ( Figure 5E , F ) . Importantly , localization of epsin 1 at the tips of the invaginations was not affected by the absence of Hip1 / Hip1R in the cytosol ( Figure 5G ) , once again as in the case of intact cells . Direct binding of epsin to actin Altogether , the results discussed above support a role for epsin as a critical factor required for the temporal and spatial coordination between clathrin - mediated endocytosis and actin dynamics . We thus explored whether epsin interacts directly with actin by incubating recombinant epsin 1 fragments ( Figure 5\u2014figure supplement 1F ) with purified F - actin followed by co - sedimentation . This analysis revealed that both the DPW and NPF motif containing regions of the protein ( Chen et al . , 1998 ) , but not the ENTH domain , bind actin ( Figure 5H ) . Thus , there appear to be at least two actin - binding sites in epsin . The site in the NPF motif containing region likely corresponds to the actin cytoskeleton - binding ( ACB ) site , previously identified in yeast epsin ( Ent1 , Skruzny et al . , 2012 ) . A low affinity binding between epsin and synaptobrevin 2 / VAMP2 Collectively , previous findings from the literature and the findings described above are consistent with roles for epsin from early to late stage clathrin - coated pits , and more generally for an essential role of this protein in clathrin - mediated endocytosis in mammalian fibroblasts . It was therefore of interest to determine whether epsin , like several clathrin adaptors with ENTH / ANTH domains ( Hirst et al . , 2004 ; Miller et al . , 2007 , 2011 ; Koo et al . , 2011 ) , binds SNAREs . SNARE binding by ENTH / ANTH family proteins helps coordinate bud formation with the incorporation of an appropriate SNARE in the nas - cent vesicle to direct its fate after fission . Such binding in the case of epsin would likely be of very low affinity , as it was not detected in previous investigations . However , even a low affinity binding could be of physiological significance in the context of other synergistic interaction between epsin , the mem - brane bilayer and other coat proteins . Pull - downs from mouse brain lysate using GST \u2013 ENTH domain of epsin 1 as bait , followed by analysis of the affinity purified material by mass spectroscopy , identified Syb2 as one of the top hits ( Figure 6\u2014 figure supplement 1A ) . A direct interaction was next explored by assessing binding to the ENTH domain of epsin 1 ( with an His tag at the C - terminus ) to increasing concentrations of the cytosolic portion of Syb2 fused to GST ( Figure 6A ) . The very low affinity of this interaction ( K D 470 \u00b1 30 \u03bc M ) was enhanced by the presence in the incubation medium of IP 6 ( K D 80 \u00b1 7 . 9 \u03bc M , Figure 6B , D and Figure 6\u2014figure supplement 1B ) , which induces the folding of helix zero of the ENTH domain ( Itoh et al . , 2001 ; Ford et al . , 2002 ) . In contrast , the deletion of helix zero from the ENTH construct abolished Syb2 binding in either the presence or the absence of IP 6 ( Figure 6C ) . Further analysis of the interaction of Syb2 with the ENTH domain using Syb2 fragments ( Figure 6E ) identified the SNARE motif of Syb2 ( Figure 6F , G ) , and more specifically the N - terminal portion of the motif ( residues 29 \u2013 60 , Figure 6H , I ) , as the region respon - sible for binding . Interestingly , previously reported interactions of vesicular SNAREs with ENTH / ANTH proteins also involve their SNARE motifs ( Koo et al . , 2011 ; Miller et al . , 2011 ) , pointing to an interaction mutually exclusive with SNARE complex formation , as expected for an \u2018endocytic\u2019 interaction . The interaction of epsin with Syb2 also occurs in intact cells , as shown by anti - HA co - immunoprecipitation experiments from HeLa cells expressing FLAG - tagged epsin 1 and either HA - tagged full - length Syb2 ( Figure 6\u2014figure supplement 1C ) , or deletion constructs of Syb2 ( Figure 6\u2014figure supplement 1D ) . Only constructs including the N - terminal portion of the SNARE motif of Syb2 co - precipitate epsin 1 ( Figure 6\u2014figure supplement 1E ) . The low affinity interaction of epsin with an exocytic SNARE is evolutionarily conserved as yeast Snc1 ( a Syb2 homologue ) bound the ENTH domain of Ent1 in a recombinant pull - down assay ( Figure 6J ) . Discussion Previous reports have addressed the role of epsin in endocytosis . However , studies in different organ - isms and in cell - free systems had emphasized the different aspects of its function , so that a complete picture of its physiological role ( s ) has remained elusive . Elucidation of epsin function in mammalian cells had been complicated by the existence of the three Epn genes with overlapping functions . To overcome this problem , we have generated cells that lack all the three epsins by a conditional KO approach involving the OHT - dependent disruption of the Epn1 gene in epsin 2 and 3 double KO mice . A main conclusion of our study is the occurrence of a close coupling between the function of epsin and the dynamics of the actin cytoskeleton . While such coupling had been suggested by studies in unicellular organisms , studies of epsin in cells of metazoan had emphasized its role as a bilayer Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 13 of 25 Research article Figure 6 . Epsin directly binds synaptobrevin 2 / VAMP2 . ( A ) Interaction of purified epsin ENTH \u2013 His6 ( 800 ng ) with increasing amounts of GST - cytosolic portion of synaptobrevin2 ( Syb2 ) as revealed by anti - His immunoblotting of bound material in a GST pull - down ( top ) . Coomassie blue stained gel of the baits ( bottom ) . ( B ) The interaction of epsin ENTH \u2013 His6 with equal amounts of GST \u2013 Syb2 ( Syb2 ) is enhanced by the presence of soluble diC8 - PI ( 4 , 5 ) P 2 ( PI ( 4 , 5 ) P 2 ) , IP 3 or IP 6 ( final concentration 50 \u03bc M ) . Top : anti - His immunoblotting . Bottom : Coomassie blue stained Figure 6 . Continued on next page Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 14 of 25 Research article deforming protein ( Ford et al . , 2002 ; Boucrot et al . , 2012 ) and as a cargo - specific adaptor ( Polo et al . , 2002 ; Overstreet et al . , 2003 ; Chen and De Camilli , 2005 ; Meloty - Kapella et al . , 2012 ) . A link of epsin to actin function at endocytic clathrin - coated pits , and in particular a role as a factor that ensures a \u2018controlled\u2019 actin growth at sites of endocytosis , is supported by several observations . First , Hip1R , which had so far been considered the major link of endocytic clathrin coats to actin in mamma - lian cells ( Engqvist - Goldstein et al . , 2001 ; Le Clainche et al . , 2007 ) , is not recruited to endocytic clathrin - coated pits in the absence of epsin . This is consistent with the interaction and functional part - nership of epsin with Sla2 / Hip1R , previously demonstrated in yeast and Dictyostelium , although in yeast it is Sla2 that recruits epsin to clathrin - coated pits ( Skruzny et al . , 2012 ) . Second , the lack of epsin results in a disruption of the normal actin cytoskeleton with a major loss of stress fibers and an accumulation of actin foci at the cell surface , typically in proximity of arrested endocytic clathrin - coated pits . Third , epsin binds actin directly via its disordered tail region . Fourth , a cell - free assay of endocytosis involving plasma membrane sheets and brain cytosol revealed that relative to WT , epsin - deficient cytosol results in an exaggerated and abnormal growth of the actin cytoskeleton . These find - ings suggest a role for epsin in orchestrating a coupling between actin and endocytic clathrin coats that ( 1 ) limits excessive actin growth and ( 2 ) helps mediate an effect of actin polymerization on bud invagination . A functional partnership between epsin and Hip1R explains why the KD of Hip1R results in a similar enhanced and abnormal actin nucleation at endocytic sites , with the formation of large actin bundles projecting away from these sites ( Engqvist - Goldstein et al . , 2004 ) . The excess of actin nucleation in the absence of epsin is of special interest as in yeast the ENTH domains of Ent1 and 2 bind two Cdc42 \u2013 GAPs ( GTPase activating proteins ) , Rga1 and Rga2 , and this interaction is essential for function ( Aguilar et al . , 2006 ) . Interestingly RLIP76 / RalBP1 , a protein with Cdc42 GAP activity , interacts with the ENTH domain of epsin ( Ross\u00e9 et al . , 2003 ) , although there is no clear sequence homology between RLIP76 / RalBP1 and Rga proteins . In view of the modulatory role of Cdc42 in the nucleation of N - WASP - Arp 2 / 3 - dependent actin nucleation at endocytic clathrin - coated pits , the recruitment of a Cdc42 \u2013 GAP by epsin could help explain why actin nucleation is enhanced in epsin null cells . A very strong link between epsin and actin is consistent with evidence for a general and evolution - arily conserved role of actin in clathrin - mediated endocytosis , which is more prominent under condi - tions of high plasma membrane tensions ( Merrifield et al . , 2005 ; Kaksonen et al . , 2006 ; Ferguson et al . , 2009 ; Boulant et al . , 2011 ; Taylor et al . , 2012 ) . The force of actin polymerization may help drive the deep invagination of the pits and also assist in dynamin - dependent fission ( Itoh et al . , 2005 ; Roux et al . , 2006 ) by developing a force that drives the deeply invaginated bud away from the plasma membrane . The occurrence of this force is well demonstrated by the dramatic actin - dependent elon - gation of the clathrin - coated pit necks in living cells and in cell - free systems ( Ferguson et al . , 2009 ; Wu et al . , 2010 ) , when the fission reaction is impaired by the lack of dynamin ( living cells ) or block of its GTPase activity ( cell - free system ) , respectively . Clearly , however , epsin does not simply function as gel of the baits . EB : empty beads . ( C ) The interaction of epsin ENTH \u2013 His6 with equal amounts of GST \u2013 Syb2 is lost in a construct missing the N - terminal helix zero ( \u0394 Helix0 ) irrespective of the presence of IP 6 ( final concentration 50 \u03bc M ) . EB : empty beads . ( D ) Quantitative analysis of the binding of ENTH \u2013 His6 to increasing amounts of GST \u2013 Syb2 in the presence or absence of IP 6 , as revealed by densitometry of anti - His immunoreactivity in western blots of bound material . K D s are also indicated . ( E ) Schematic representation of constructs used for ( F \u2013 I ) . SNARE motif and transmembrane region ( TM ) are denoted by dotted lines . ( F and G ) Pull - down of ENTH \u2013 His6 by GST fusions of different cytosolic fragments of Syb2 . Anti - His immunoblotting ( F ) shows binding only to SNARE motif containing fragments ( top ) . Coomassie blue stained gel of the baits ( Bottom ) . Quantitative analysis of the results is shown in ( G ) . ( H and I ) ENTH \u2013 His6 domain binding to SNARE motif fragments ( H ) and corresponding quantification ( I ) . Anti - His immunoblotting detects ENTH interaction with the N - terminal portion of the SNARE motif . ( J ) Purified yeast His - Snc1 binds a GST fusion of the ENTH domain of Ent1 ( yeast epsin 1 ) in a pull down assay ( top ) . Coomassie blue stained gel of the baits ( Bottom ) . Corresponding quantification is shown . Data are represented as mean \u00b1 SEM . See also Figure 6\u2014figure supplement 1 . DOI : 10 . 7554 / eLife . 03311 . 011 The following figure supplement is available for figure 6 : Figure supplement 1 . Epsin binding to Syb2 : mass spectroscopy analysis and interaction in intact cells . DOI : 10 . 7554 / eLife . 03311 . 012 Figure 6 . Continued Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 15 of 25 Research article a negative regulator of actin polymerization , but it is required for proper spatial organization of the actin - based cytoskeleton , as exemplified by the abnormal localization of myosin 1E at sites of endo - cytic invaginations on plasma membrane sheets incubated with epsin TKO cytosol . We note that partial deficiency of epsin function has a major impact on intercellular signaling medi - ated by Notch ( Overstreet et al . , 2003 ; Chen et al . , 2009 ) . It was proposed that Notch activation requires a pulling force applied to Notch by the endocytosis of its ligand in the neighboring cell and that epsin may be essential for the generation of such force ( Meloty - Kapella et al . , 2012 ; Musse et al . , 2012 ; Shergill et al . , 2012 ) . Based on our findings , such a requirement for epsin may be linked to its role in the control of actin dynamics at clathrin - coated pits . Reported roles of epsin in clathrin - independent endocytosis may also be linked to its function in actin regulation ( Sigismund et al . , 2005 ) . Another important finding of our study is a weak interaction of epsin ' s ENTH domain with the SNARE synaptobrevin 2 / VAMP2 . As intracellular vesicles need SNAREs to fuse with their target mem - branes , a coupling must exist between vesicle budding and incorporation of the proper SNARE ( s ) in the bud . Accordingly , other coat adaptors with ANTH or ENTH modules , such as AP180 / CALM and epsinR were shown to bind SNAREs ( Hirst et al . , 2004 ; Miller et al . , 2007 , 2011 ; Koo et al . , 2011 ) . We hypothesize that the very low affinity of the binding of synaptobrevin 2 to epsin , which explains why such an interaction had not been described so far , may nevertheless be physiologically relevant when compounded by other interaction of epsin with the bilayer , other coat components , and ubiqui - tinated cargo proteins . Recently , it was proposed that a main role of epsin in endocytic clathrin - coated pit dynamics is to help mediate membrane fission via the membrane remodeling properties of the amphipathic helix zero of its ENTH domain ( Boucrot et al . , 2012 ) . While this action seems plausible , our study strongly suggests that epsin becomes critically important at the earlier stages of clathrin coat maturation , as we have observed an accumulation of shallow and U - shaped endocytic clathrin - coated pits in epsin TKO cells . The diffuse , rather than punctate distribution of proteins that assemble at the necks of deeply invaginated endocytic clathrin - coated pits , such as dynamin 2 , endophilin 2 , and myosin 1E further supports a stalling of pits at an early stage in epsin TKO cells . In contrast to the observations made in epsin triple KD cells , we have not observed clearly multiheaded pits , although the dome of some pits was irregular . Multiheaded pits observed in epsin triple KD cells had been interpreted as reflecting a defect in fission . However , they do not have the constricted neck of multiheaded pits observed in dynamin mutant cells ( Ferguson et al . , 2007 , 2009 ) and at least some of them could represent a clus - tering of shallow or U - shaped pits . We note that the multiple endocytic functions of epsin are con - sistent with its localization throughout the coat ( Hawryluk et al . , 2006 ; Sochacki et al . , 2014 ) , rather than a selective localization at the bud neck to mediate fission . Finally , our work also demonstrates a dramatic impact of the lack of epsin on cell division , extend - ing previous observations made in epsin 1 and 2 double KD cells , where abnormal spindles were detected ( Liu and Zheng , 2009 ) . Surprisingly , we did not find a major effect on cell division in cells derived from epsin 1 and 2 double KO embryos possibly because of long - term adaptation to decreased levels of epsin . For example , we have detected an upregulation of epsin 3 in epsin 1 and 2 double KO cells ( our unpublished observations ) . Interestingly , epsin undergoes phosphorylation and mono - or oligo - ubiquitination ( Stukenberg et al . , 1997 ; Chen et al . , 1999 , 2003 ; Polo et al . , 2002 ) in mito - sis and these covalent modifications impair its binding to clathrin and AP - 2 , suggesting a switch of its function in the mitotic cytosol ( Chen et al . , 1999 ) . A role of epsin in the dynamics of the cytoskeleton may underlie its role in mitosis , for example by affecting its anchoring to the cell cortex . Several other endocytic proteins , including clathrin , are also implicated in mitosis , and , in particular , in cytokinesis ( Royle , 2013 ) . In conclusion , our results reveal new aspects of the collective function of the three epsin genes and point to epsin as an important coordinator of the dynamics of endocytic clathrin coats from early to late stages . Our findings , along with previous studies suggest the models depicted in Figure 7 . Epsin is an early component of endocytic clathrin - coated pits , which interacts directly with PI ( 4 , 5 ) P 2 , AP - 2 , clathrin , and other early clathrin - coated pit components such as intersectin and Eps15 . The partial bilayer penetration of its ENTH domain makes it optimally suited to function at a site where membrane buckling occurs ( as a generator , sensor or stabilizer of curvature ) , while its low affinity binding to syn - aptobrevin 2 / VAMP2 may collaborate with AP180 / CALM to ensure the coupling of bud nucleation to a membrane patch containing SNAREs . It may additionally participate in the recruitment to the pit of ubiquitinated cargo proteins via its UIMs , a function that does not apply to Dictyostelium epsin , Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 16 of 25 Research article which expresses a single epsin family protein lacking UIMs . As the bud expands , epsin along with Hip1R provide a link between actin nucleation and coat maturation , which is required for deep invag - ination . Finally , the membrane insertion properties of its helix zero may assist dynamin in the fission Figure 7 . Functions of epsin in clathrin - mediated endocytosis . ( A ) Schematic representation of the role of epsin in the coupling of the endocytic clathrin coat to actin in cooperation with Hip1R ( left ) . This coupling helps invaginate the pit . Only epsin localized at the equator of the bud is shown to emphasize its actin - related function in pit invagination , but epsin is not restricted to the equator . Higher magnification representation of the interactions of epsin ( right ) : the N - terminal ENTH domain binds ( 1 ) the PI ( 4 , 5 ) P 2 - rich membrane bilayer of the plasma membrane ( and partially penetrates it ) , ( 2 ) the ANTH domain of Hip1R , and ( 3 ) synaptobrevin2 / VAMP2 ; its tail binds ( 1 ) ubiquitinated cargo proteins ( via UIMs ) , ( 2 ) AP - 2 and clathrin heavy chain ( via DPF / DPW motifs and \u2018clathrin boxes\u2019 , respectively ) , and ( 3 ) the EH domain region of Eps15 ( via NPF motifs ) . Epsin ' s tail also binds F - actin and co - operates with the THATCH domain of Hip1R in the coupling of the actin cytoskeleton to the clathrin - coated pit . In both fields Hip1R is depicted as a monomer but functions as a dimer , or heterodimer with Hip1 . ( B ) Schematic represen - tation of the multiple functions of epsin in clathrin - mediated endocytosis . Epsin participates in the early stages of the reaction as a coat nucleator , curvature inducer / sensor and also helps coupling bud formation to SNARE incorporation . As the coat matures , ubiquitinated cargo is recruited . Additionally , the link between epsin / Hip1R and actin is required for the deep invagination of the pit , and this is the process at which the action of epsin becomes essential . Subsequently , the force produced by actin and the bilayer destabilizing properties of epsin ' s ENTH domain may co - operate with dynamin in fission . DOI : 10 . 7554 / eLife . 03311 . 013 Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 17 of 25 Research article reaction . As we show here , however , its most critical function is at the transition between shallow / U - shaped and deep invaginated pits . Materials and methods Antibodies and reagents Antibodies were obtained from the following commercial sources : mouse anti - tubulin , rabbit anti - FLAG , and mouse anti - Actin ( Sigma - Aldrich , St . Louis , MO , USA ) ; mouse anti - transferrin receptor and rabbit anti - GFP ( Life Technologies , Carlsbad , CA , USA ) ; goat anti - epsin 1 and mouse anti - GST ( Santa Cruz Biotechnology , Santa Cruz , CA , USA ) ; rabbit anti - clathrin light chain , mouse anti - Arp 2 / 3 , rabbit anti - Hip1R , and mouse anti - Hip1 ( EMD - Millipore , Billerica , MA , USA ) ; rat HRP - conjugated anti - HA ( Roche , Mannheim , Germany ) ; rabbit anti - His tag ( GenScript , Piscataway , NJ , USA ) ; mouse anti - Aurora B , rabbit anti - caveolin - 1 , and mouse anti - adaptin \u03bc 2 subunit ( BD Transduction Laboratories , San Jose , CA , USA ) . The following antibodies were kind gifts : mouse anti - epsin 1 and mouse anti - epsin 3 ( Pier - Paolo Di Fiore , IFOM , Milan , Italy ) ; rabbit anti - Hip1R ( David Drubin , University of California , Berkeley , CA , USA ) ; rabbit anti - Myosin 1E ( Mark Mooseker , Yale University , New Haven , CT , USA ) and rabbit anti - synaptophysin ( Paul Greengard , The Rockefeller University , New York , NY , USA ) . Rabbit anti - epsin 2 , mouse anti - clathrin heavy chain , mouse anti - adaptin \u03b1 subunit , rabbit anti - dynamin 2 and rabbit anti - endophillin 2 were generated in our lab . Alexa594 - phalloidin , Alexa594 - mouse anti - HA and Alexa488 , Alexa594 , and Alexa647 conjugated secondary antibodies were from Life Technologies ; Alexa405 - phalloidin was from AAT Bioquest ( Sunnyvale , CA , USA ) while DAPI was from Sigma - Aldrich . Plasmids C - terminal GFP - tagged epsin 1 was obtained by PCR amplification of the epsin 1 coding sequence from brain cDNA library ( Clontech Laboratory Inc . , Mountain View , CA , USA ) followed by ligation into the pEGFP \u2013 N3 plasmid ( Clontech Laboratory Inc . ) . For the generation of the GST tagged ENTH , the rat epsin 1 ENTH coding sequence was PCR amplified and ligated in the pGEX4T - 1 vector ( GE Healthcare , Pittsburgh , PA , USA ) . The GST - tagged epsin fragments ( ENTH , DPW , NPF ) were generated by the ligation of the corre - sponding PCR products to pGEX4T - 1 ( GE Healthcare ) , while for the GST - tagged DPW \u2013 NPF tandem fragment the ligation was with pGEX6 ( GE Healthcare ) . The 6xHis - tagged ENTH of rat epsin 1 was produced by cloning ENTH coding sequence in pET21a ( + ) ( EMD - Millipore ) , and the resulting coding plasmids was mutagenized by site directed mutagenesis ( QuikChange II XL , Agilent Technology , Santa Clara , CA , USA ) to remove the helix zero coding sequence in order to obtain \u0394 Helix0 - ENTH His6 . GST - tagged Syb2 was created by PCR amplification of the cytosolic portion of human Syb2 , the corresponding SNARE motif or the N - terminal extra SNARE region . PCR products were ligated in the pGEX4T - 1 vector ( GE Healthcare ) . FLAG - tagged full - length epsin 1 was generated by cutting epsin 1 coding sequence from the above reported epsin1 - GFP followed by ligation in the pFLAG - CMV4 ( gift from Angus Nairn , Yale University , New Haven , CT , USA ) . For the generation of a cell line stably expressing Syb2 - HA , human Syb2 coding sequence in fusion with HA was cut from pCI neo Syb2 - HA and cloned into pBABE puro . Syb2 - HA fragments ( \u0394 1 \u2013 31 , \u0394 1 \u2013 50 , \u0394 1 \u2013 70 ) were generated by site directed mutagenesis ( QuikChange II XL , Agilent Technology ) of pCI neo Syb2 - HA . For study of yeast Ent1 ENTH domain \u2013 Snc1 interaction , the following plasmids were used : GST control ( pBW1546 pGEX - 5X - 1 ) , WT ENTH1 ( pBW1800 pGEX - 5X - 1 : ENTH1 ) , and Snc1 cytosolic tail ( pBW1916 pET28a : Scn1 [ aa1 - 93 ] ) . pGST 29 / 60 and pGST 61 / 93 Syb2 SNARE were a gift from Karin Reinisch ( Yale University , New Haven , CT , USA ) ; Hip1R \u2013 GFP was a gift from David Drubin ( University of California , Berkeley , CA , USA ) ; and clathrin light chain \u2013 GFP was a gift from James Keen ( Thomas Jefferson University , Philadelphia , PA , USA ) . Adaptin \u03bc 2 subunit - GFP , mCherry - tagged utrophin , and GFP - tagged human transferrin receptor were previously described ( Merrifield et al . , 2005 ; Zoncu et al . , 2007 ; Ferguson et al . , 2009 ) . Generation of epsin conditional triple knock out mice Conditional Epn1 KO mouse ( Epn1 fl / fl ) was generated by homologous recombination at the Epn1 locus with a targeting construct containing loxP sequence and a Frt - flanked neomycin cassette for selection Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 18 of 25 Research article ( removed by breeding with Flp mice , see also [ Pasula et al . , 2012 ] ) . Epn1 fl / fl mice were interbred with Epn2 \u2212 / \u2212 ( Chen et al . , 2009 ) and Epn3 \u2212 / \u2212 ( Ko et al . , 2010 ) mice and with either 4 - hydroxy - tamoxifen ( OHT ) - inducible Cre strain mice ( Badea et al . , 2003 ) , expressing Cre recombinase under the control of the estrogen receptor promoter , or nestin - Cre mice ( Tronche et al . , 1999 ) to generate OHT - inducible and brain - specific epsin triple KO mice , respectively . All mice were of a C57BL / 6J congenic background . Animal care and use was carried out in accordance with our institutional guidelines . Fibroblast cultures Primary fibroblasts were isolated form conditional KO and WT mice from E18 embryos or P1 pups and cultured by standard methods . To achieve gene recombination and obtain epsin triple KO cells and their controls , cells were cultured for 7 days with addition of OHT ( Sigma - Aldrich ) to the culture medium at day 1 ( 3 \u03bc M ) and day 4 ( 1 \u03bc M ) , a two - step 7 - day long process . For cell counting , cells were plated at 50 , 000 cells per 35 - mm dish . Two counts were performed for each of the three dishes of WT and epsin TKO cells using a hemocytometer , both at 1 and 3 days post - plating . The cell proliferation index represents the ratio between cell numbers at these two time points . Transfection Plasmids were transfected by electroporation ( Nucleofector , Amaxa , Cologne , Germany ) for imaging experiments , or using Lipofectamine 2000 ( Life Technologies ) for immunoprecipitation experiments . siRNA oligos were transfected by using Lipofectamine RNAi MAX ( Life Technologies ) , and cells were cultured for 48 \u2013 72 hours before analysis . Double - stranded siRNAs were derived from the following references : Hip1R ( mouse Hip1R , MMC . RNAI . N145070 . 12 . 1 , and 12 . 2 from IDT , Coralville , IA , USA ) ; control ( NC1 negative control duplex from IDT ) . Immunofluorescence Cells were plated and grown on 5 \u03bc g / ml human fibronectin ( EMD - Millipore ) - coated glass coverslips and fixed with 4 % paraformaldehyde \u2013 4 % sucrose in 0 . 1 M sodium phosphate buffer ( pH 7 . 2 ) at room temperature . Coverslips were washed with 50 mM NH 4 Cl pH 7 . 2 , then blocked and permeabilized in 0 . 1 % Triton X - 100 and 3 % bovine serum albumin in PBS . Primary and secondary antibody incubations were performed using the same buffer . Coverslips were finally rinsed in PBS and mounted in Prolong Gold + DAPI ( Life Technologies ) . For surface staining , cells were incubated with an Alexa594 - conjugated rat anti - HA antibody ( Life Technologies ) for 60 min on ice at 4\u00b0C . After extensive washing with cold PBS , cells were fixed as described above and counterstained with DAPI ( Sigma - Aldrich ) before coverslip mounting . Immunofluorescence data acquisition was performed using either a Zeiss Axioplan2 micro - scope ( for the epifluorescence images ) or a spinning disk confocal microscope ( see below ) . Live imaging For live imaging , cells were incubated at 37\u00b0C in the following buffer : 136 mM NaCl , 2 . 5 mM KCl , 2 mM CaCl 2 , 1 . 3 mM MgCl 2 , 3 mM D - glucose , and 10 mM Hepes - Na pH 7 . 4 . Spinning - disk con - focal microscopy was performed using the Improvision UltraVIEW VoX system ( Perkin - Elmer , Walthman , MA , USA ) built on a Nikon Ti - E inverted microscope , equipped with PlanApo objec - tives ( 60 \u00d7 1 . 45 - NA ) and controlled by Volocity ( Improvision , Coventry , UK ) software . Total internal reflection fluorescent ( TIRF ) microscopy was carried out on a Nikon TiE microscope equipped with 60 \u00d7 1 . 49 - NA and 100 \u00d7 1 . 49 - NA objectives . Excitation lights were provided by 488 nm ( for GFP ) and 561 nm ( mCherry ) diode - pumped solid - state lasers . An optical fiber coupled the lasers to the TIRF illuminator and an acousto - optic tunable filter controlled the output from the lasers . Fluorescent signals were detected with an EM - CCD camera ( DU - 887 ; Andor , Belfast , NIR ) and acquisition was controlled by Andor iQ software . Images typically were sampled at 0 . 2 Hz with exposure times in the 4 to 6 s range . Transferrin uptake and transferrin receptor localization For transferrin uptake , cells were electroporated with a plasmid encoding GFP - tagged human trans - ferrin receptor ( Merrifield et al . , 2005 ) . Cells were then starved for 2 hours with serum - free DMEM , incubated at 4\u00b0C for 30 min with 10 \u03bc g / ml Alexa594 - conjugated human transferrin ( Life Technologies ) , and then warmed up to 37\u00b0C for 15 min to allow internalization . Uptake was ended by a brief rinse with ice - cold PBS followed by fixation with 4 % parafolmaldehyde \u2013 4 % sucrose in 0 . 1 M sodium phosphate buffer pH 7 . 2 at room temperature . Cells were counterstained with DAPI to visualize nuclei . Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 19 of 25 Research article Cell surface protein biotinylation Cells were rinsed with PBS and labeled on ice for 60 min with 1 mg / ml EZ - link Sulfo - NHS - SS - Biotin ( Thermo Fisher , Walthman , MA , USA ) , rinsed on ice with 50 mM glycine in PBS pH 8 . 0 , and then in PBS only . Cells were then lysed in modified RIPA buffer ( 1 % Triton X - 100 , 0 . 1 % SDS in 20 mM TRIS pH 7 . 5 , 50 mM NaCl , EDTA 0 . 5 mM with protease and phosphatase inhibitors cocktails ) and lysates were cen - trifuged at 16 . 200\u00d7 g , at 4\u00b0C for 5 min . Biotinylated proteins were recovered on NeutrAvidin beads ( Thermo Fisher ) . After washing with modified RIPA buffer , proteins were eluted with Laemmli SDS - PAGE sample buffer and boiling . Evaluation of protein levels in starting material , biotinylated ( cell surface ) and non - biotinylated ( intracellular ) fractions was assessed by Western blotting . Electron microscopy Control and epsin TKO cells ( \u223c 80 % confluent ) in 60 - mm dishes were fixed in 2 % glutaraldehyde \u2013 0 . 1 M sodium cacodylate . They were post - fixed with 1 % OsO 4 in 1 . 5 % K 4 Fe ( CN ) 6 and 0 . 1 M sodium caco - dylate , en bloc stained with 0 . 5 % uranyl magnesium acetate , dehydrated and embedded in Embed 812 . For morphometric analysis , cells were selected at a low magnification allowing confirmation that the entire outer perimeter was intact , but not visualizing clathrin - coated pits to make the selection process blind with respect to the phenotype of interest . Once selected , higher magnification images were taken around the periphery of the cell , and clathrin - coated structures within the categories defined in Figure 2T were counted all the way around the perimeter of each cell . Electron microscopy reagents were purchased from Electron Microscopy Sciences ( Hatfield , PA , USA ) . Immunoprecipitation HeLa cells co - expressing FLAG - tagged epsin1 , or FLAG alone and HA - tagged Syb2 or HA alone , were washed in ice - cold PBS and lysed in the lysis buffer ( 150 mM NaCl , 0 . 1 % SDS , 0 . 5 % Triton X - 100 , 10 mM EDTA , 20 mM Hepes pH 7 . 4 , and protease inhibitor cocktail [ Roche ] ) . Cell lysates were then centrifuged at 16 , 000\u00d7 g for 20 min at 4\u00b0C and supernatants were incubated , under rotation , with agarose - conjugated anti - HA beads ( Roche ) for 1 hour at 4\u00b0C . Beads were then extensively washed in cold lysis buffer and bound proteins were eluted in Laemmli sample buffer and boiled for 5 min . Mouse brain cytosol preparation 4 weeks - old WT , epsin TKO or Hip1 / Hip1R double KO mouse brains were homogenized in homogeni - zation buffer ( 25 mM Tris pH 8 . 0 , 500 mM KCl , 250 mM sucrose , 2 mM EGTA , 1 mM dithiothreitol ) in the presence of a protease inhibitor cocktail ( Roche ) . The lysate was centrifuged at 160 , 000\u00d7 g for 2 hours in a TL100 . 2 rotor ( Beckman Coulter , Indianapolis , IN , USA ) . The resulting supernatant was buffer - exchanged on PD - 10 columns ( GE Healthcare ) at room temperature into cytosolic buffer ( 25 mM Hepes pH 7 . 4 , 120 mM potassium glutamate , 20 mM potassium chloride , 2 . 5 mM magnesium acetate , 5 mM EGTA ) . Protease inhibitors were added and aliquots of cytosol were immediately frozen in liquid nitrogen and stored at \u221280\u00b0C for up to 2 months . Plasma membrane sheets preparation PTK2 cells stably expressing PM \u2013 GFP ( Chen and De Camilli , 2005 ) were cultured at 37\u00b0C in 10 % CO 2 in minimum essential Eagle medium ( Life Technologies ) supplemented with 10 % ( vol / vol ) fetal bovine serum , 100 \u03bc g / ml penicillin / streptomycin , and 0 . 5 mg / ml G418 . Cells were grown for 24 \u2013 48 hours until confluent in poly - d - lysine coated MatTek dishes ( MatTek , Ashland , MA , USA ) . These dishes were prepared by exposing them to 20 \u03bc g / ml poly - d - lysine ( Sigma - Aldrich ) for 30 min and washed overnight . To prepare plasma membrane sheets , cells were rinsed with PBS and then sheared in ice - cold cytosolic buffer by a brief pulse of sonication ( about 0 . 5 s ) using a cell disruptor ( VirTis Ultrasonics ) set at 20 % of output power with a 1 / 8 - inch microprobe positioned at about 15 mm above the dish . Membrane sheets were rinsed in cytosolic buffer and used for the in vitro assay within 20 min . Generation of endocytic budding intermediates from plasma membrane sheets was achieved by incubating them at 37\u00b0C with cytosol in the presence of nucleotides in the following concentrations : 2 mg / ml cytosol , 1 . 5 mM ATP and 150 \u03bc M GTP \u03b3 S . Samples were supplemented with an ATP - regenerating system consisting of 16 . 7 mM creatine phosphate and 16 . 7 U / ml creatine phosphokinase . The reaction was stopped by washing with cytosolic buffer and fixing samples with 4 % paraformaldehyde \u2013 4 % sucrose in 0 . 1 M sodium phosphate buffer pH 7 . 2 for 15 min at room temperature . Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 20 of 25 Research article Actin co - sedimentation assay 250 \u03bc g of rabbit muscle G - actin ( Sigma - Aldrich ) were polymerized for 30 min at 25\u00b0C in polymerization buffer ( 50 mM KCl , 2 mM MgCl 2 , 1 mM ATP ) . Recombinant GST - epsin fragments ( final concentration 5 \u03bc M ) were pre - cleared by centrifugation at 150 , 000\u00d7 g for 1 hour at 4\u00b0C and incubated with F - actin ( final concentration : 15 \u03bc M ) . After 30 min of incubation at room temperature , samples were centri - fuged at 150 , 000\u00d7 g for 1 . 5 hours at 4\u00b0C . Supernatants and pellets were brought to equal volumes of SDS / PAGE Laemmli buffer , and samples were analyzed by SDS / PAGE using 4 \u2013 12 % Bis - Tris NuPAGE gel in MES buffer ( Life Technologies ) followed by anti - GST immunoblotting . Recombinant protein purification For GST - fusion protein production , BL21 - DE3 E . coli cells were transformed by heat - shock at 42\u00b0C with pGEX - 4T alone or pGEX - 4T constructs . Large - scale cultures were grown to log phase ( about 3 hours ) in Luria - Bertani broth containing ampicillin ( 100 mg / ml ) at 37\u00b0C , under shaking at 200 \u2013 250 rpm . Cultures were then induced with 500 mM isopropyl - \u03b2 - D - 1 - thiogalactopyranoside ( IPTG ) at 37\u00b0C for 3 hours and pelleted by centrifugation at 3000\u00d7 g for 15 min at 4\u00b0C . Bacteria were lysed in 150 mM NaCl , 4 mM DTT , 10 mM Hepes pH 7 . 4 , and protease inhibitor cocktail ( Roche ) in the presence of 1 % Triton X - 100 . Lysates were cleared by centrifugation at 22 , 000\u00d7 g for 20 min at 4\u00b0C , and the resulting supernatants were incubated with Glutathione - Sepharose 4 Fast Flow beads ( GE Healthcare ) for 1 hour at 4\u00b0C . After thorough washings with ice - cold buffer ( without Triton X - 100 ) , proteins were eluted by incubating the beads with 30 mM glutathione , 150 mM NaCl , 4 mM DTT , and 10 mM HEPES pH 8 . For His - tagged protein generation , BL21 - DE3 pLys - S E coli were transformed as described above with a pET21a ( + ) encoding a 6xHis - tagged fusion protein . After reaching the log phase at 37\u00b0C , bac - teria were induced overnight at 18\u00b0C with 100 mM IPTG . Bacterial pellet was lysed in buffer ( 150 mM NaCl , 1 mM \u03b2 - mercaptoethanol , 20 mM HEPES pH 7 . 4 , 1 % Triton X - 100 ) containing 15 mM imidazole , and the cleared supernatants were incubated with Ni 2 + \u2013 NTA agarose beads for 1 hour at 4\u00b0C . Samples were washed in the same buffer , but containing 30 mM imidazole , and finally recombinant proteins were eluted in the same buffer containing 300 mM imidazole . Both GST - and His - tagged proteins were subsequently purified by gel filtration using a Superdex S200 or a Superdex S75 column ( GE Healthcare ) on the Akta Pure FPLC system ( GE Healthcare ) . Finally , proteins were buffer exchanged in 150 mM NaCl , 4 mM DTT , 10 mM Hepes pH 7 . 4 by High - Prep 26 / 10 Desalting Column ( GE Healthcare ) . Protein purity was assessed by Coomassie Blue stain - ing of samples run on 4 \u2013 20 % gradient Tris / Glycine or 16 . 5 % Tris - Tricine MiniProtean gels ( Biorad , Hercules , CA , USA ) . GST pull - down assays to assess protein \u2013 protein interactions GST - tagged bait proteins bound to glutathione beads were incubated overnight at 4\u00b0C with constant agitation with 6xHis - tagged prey proteins in 150 mM NaCl , 4 mM DTT , 4 mM \u03b2 - mercaptoethanol , 0 . 1 % NP - 40 , 0 . 05 % BSA , 20 mM HEPES pH 7 . 4 . Reaction mixtures were subsequently cleared by centrifugation at 3000\u00d7 g for 5 min at 4\u00b0C ; supernatants were removed and beads were washed three times with ice - cold buffer . Bound proteins were eluted in Laemmli buffer and analyzed by separation on 16 . 5 % Tris - Tricine gradient Mini - Protean gels ( Biorad ) followed by Western blotting with anti - His antibody . In experiments where the effect of phosphoinositol groups was tested , the incubation mixture was supplemented with 50 \u03bc M final of diC8PI ( 4 , 5 ) P 2 , IP 3 ( 1 , 4 , 5 ) ( Avanti Polar Lipids , Alabaster , AL , USA ) or IP 6 ( 1 , 4 , 5 ) ( phytic acid dipotassium salt , Sigma - Aldrich ) . Generation of Syb2 \u2013 HA stable cell line Mouse fibroblasts stably expressing Syb2 \u2013 HA were generated by transducing WT and conditional epsin double KO mouse fibroblasts with a Syb2 \u2013 HA encoding retrovirus . The virus was produced in the Phoenix helper - free retrovirus producer line ( 293T cell line transformed with adenovirus E1a and carrying a temperature sensitive T antigen co - selected with neomycin ) . Briefly 10 \u03bc g of pBABE puro - Syb2 \u2013 HA were transfected by Lipofectamine 2000 ( Life Technologies ) into Phoenix cells . In parallel WT and conditional epsin double KO fibroblasts were plated . 24 hours after transfection , the medium from the Phoenix cell dishes was removed , supplemented with 8 \u03bc g / ml polybrene ( cationic polymer used to increase the retroviral infection efficiency in cell culture ) and filtered to remove cell debris . The filtrate was added to the mouse fibroblasts and removed after 4 \u2013 6 hours incubation at 37\u00b0C . 48 hours after the initial viral incubation fibroblasts were splitted in Dulbecco ' s Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 21 of 25 Research article Modified Eagle Medium ( Life Technologies ) . pBABE puro vector carries a puromycin resistance gene that was used to select the Syb2 \u2013 HA expressing cells by supplementing the culture medium with 4 \u03bc g / ml puromicyn ( [ Ferguson et al . , 2009 ] , Sigma - Aldrich ) . Quantification and statistical analysis Fluorescent signals were quantified with Fiji ImageJ ( http : / / fiji . sc / wiki / index . php / Fiji ) or the Volocity 3D Image Analysis software ( Improvision ) . Immunoblots were analyzed by Fiji ImageJ or Image Studio ( Licor Bioscience , Lincoln , NE , USA ) . Graphical presentations were made using Graph Pad Prism ( Graph Pad Software , La Jolla , CA , USA ) . Statistical analyses were performed by Graph Pad Prism using Student ' s t test for independent samples or one - way Anova . Miscellaneous procedures Bicinchoninic acid protein quantification ( Thermo Fisher ) , SDS - PAGE electrophoresis , immunoblotting , and GST - pulldowns from brain homogenates were performed by standard procedures ( Ferguson et al . , 2009 ) . Acknowledgements We thank Shawn Ferguson ( Yale University ) and Yixian Zheng ( Carnegie Institution for Science , Baltimore , MD ) for discussion and suggestions ; Frank Wilson , Lijuan Liu , and Louise Lucast for superb technical assistance and Alanna Coughran and Abigail Soyombo for help with the Hip1 / Hip1R DKO mice . Generous gifts of key reagents are acknowledged in the Materials and methods section . This work was supported in part by grants R37NS036251 , DK082700 , DK45735 and DA018343 to PDC , R01GM60979 to BW , R01CA098730 to TR and a Grant from the American Heart Association ( # 0835544N ) to HC . RF - B was the recipient of a Feodor Lynen postdoctoral fellowship from the Alexander von Humboldt Foundation , EWS was supported by NIH 5T32GM007223 - 38 / 39 and KW by NIGMST32007231 . Additional information Funding Funder Grant reference number Author Howard Hughes Medical Institute Mirko Messa , Rub\u00e9n Fern\u00e1ndez - Busnadiego , Elizabeth Wen Sun , Hong Chen , Heather Czapla , Yumei Wu , Genevieve Ko , Pietro De Camilli National Institute of Neurological Disorders and Stroke R37NS036251 Pietro De Camilli National Institute on Drug Abuse DA018343 Pietro De Camilli National Institute of Diabetes and Digestive and Kidney Diseases DK082700 , DK45735 Pietro De Camilli National Institute of General Medical Sciences R01GM60979 Beverly Wendland National Cancer Institute R01CA098730 Theodora Ross American Heart Association 0835544N Hong Chen Alexander von Humboldt - Stiftung Rub\u00e9n Fern\u00e1ndez - Busnadiego National Institute of General Medical Sciences NIGMST32007231 Kristie Wrasman National Institutes of Health 5T32GM007223 - 38 / 39 Elizabeth Wen Sun The funders had no role in study design , data collection and interpretation , or the decision to submit the work for publication . Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 22 of 25 Research article Author contributions MM , RF - B , PDC , Conception and design , Acquisition of data , Analysis and interpretation of data , Drafting or revising the article , Contributed unpublished essential data or reagents ; EWS , Acquisition of data , Analysis and interpretation of data , Drafting or revising the article ; HC , TR , Drafting or revising the article , Contributed unpublished essential data or reagents ; HC , YW , Acquisition of data , Analysis and interpretation of data ; KW , Conception and design , Acquisition of data , Analysis and interpreta - tion of data ; GK , Acquisition of data , Contributed unpublished essential data or reagents ; BW , Conception and design , Analysis and interpretation of data , Drafting or revising the article , Contributed unpublished essential data or reagents Ethics Animal experimentation : The institutional animal care and use committee ( IACUC ) of the Yale University and the approved animal protocol is 2012 - 07422 . The institutional guidelines for the care and use of laboratory animals were followed . References Aguilar RC , Longhi SA , Shaw JD , Yeh LY , Kim S , Schon A , Freire E , Hsu A , McCormick WK , Watson HA , Wendland B . 2006 . Epsin N - terminal homology domains perform an essential function regulating Cdc42 through binding Cdc42 GTPase - activating proteins . Proceedings of the National Academy of Sciences of USA 103 : 4116 \u2013 4121 . doi : 10 . 1073 / pnas . 0510513103 . Aguilar RC , Watson HA , Wendland B . 2003 . The yeast Epsin Ent1 is recruited to membranes through multiple independent interactions . The Journal of Biological Chemistry 278 : 10737 \u2013 10743 . doi : 10 . 1074 / jbc . M211622200 . Badea TC , Wang Y , Nathans J . 2003 . A noninvasive genetic / pharmacologic strategy for visualizing cell morphol - ogy and clonal relationships in the mouse . The Journal of Neuroscience 23 : 2314 \u2013 2322 . Baggett JJ , D ' Aquino KE , Wendland B . 2003 . The Sla2p talin domain plays a role in endocytosis in Saccharomyces cerevisiae . Genetics 165 : 1661 \u2013 1674 . Banerjee B , Kestner CA , Stukenberg PT . 2014 . EB1 enables spindle microtubules to regulate centromeric recruitment of Aurora B . The Journal of Cell Biology 204 : 947 \u2013 963 . doi : 10 . 1083 / jcb . 201307119 . Banfalvi G . 2011 . Overview of cell synchronization . In : Cell Cycle Synchronization Methods in Molecular Biology . Totowa , NJ : Humana Press . Vol 761 . p . 1 \u2013 23 . doi : 10 . 1007 / 978 - 1 - 61779 - 182 - 6 _ 1 . Boucrot E , Pick A , \u00c7amdere G , Liska N , Evergren E , McMahon HT , Kozlov MM . 2012 . Membrane fission is promoted by insertion of amphipathic helices and is restricted by crescent BAR domains . Cell 149 : 124 \u2013 136 . doi : 10 . 1016 / j . cell . 2012 . 01 . 047 . Boulant S , Kural C , Zeeh JC , Ubelmann F , Kirchhausen T . 2011 . Actin dynamics counteract membrane tension during clathrin - mediated endocytosis . Nature Cell Biology 13 : 1124 \u2013 1131 . doi : 10 . 1038 / ncb2307 . Bradley SV , Hyun TS , Oravecz - Wilson KI , Li L , Waldorff EI , Ermilov AN , Goldstein SA , Zhang CX , Drubin DG , Varela K , Parlow A , Dlugosz AA , Ross TS . 2007 . Degenerative phenotypes caused by the combined deficiency of murine HIP1 and HIP1r are rescued by human HIP1 . Human Molecular Genetics 16 : 1279 \u2013 1292 . doi : 10 . 1093 / hmg / ddm076 . Brady RJ , Wen Y , O\u2019Halloran TJ . 2008 . The ENTH and C - terminal domains of Dictyostelium epsin cooperate to regulate the dynamic interaction with clathrin - coated pits . Journal of Cell Science 121 : 3433 \u2013 3444 . doi : 10 . 1242 / jcs . 032573 . Brady RJ , Damer CK , Heuser JE , O ' Halloran TJ . 2010 . Regulation of Hip1r by epsin controls the temporal and spatial coupling of actin filaments to clathrin - coated pits . Journal of Cell Science 123 : 3652 \u2013 3661 . doi : 10 . 1242 / jcs . 066852 . Brett TJ , Legendre - Guillemin V , McPherson PS , Fremont DH . 2006 . Structural definition of the F - actin - binding THATCH domain from HIP1R . Nature Structural & Molecular Biology 13 : 121 \u2013 130 . doi : 10 . 1038 / nsmb1043 . Cadavid AL , Ginzel A , Fischer JA . 2000 . The function of the Drosophila fat facets deubiquitinating enzyme in limiting photoreceptor cell number is intimately associated with endocytosis . Development 127 : 1727 \u2013 1736 . Chen H , De Camilli P . 2005 . The association of epsin with ubiquitinated cargo along the endocytic pathway is negatively regulated by its interaction with clathrin . Proceedings of the National Academy of Sciences of USA 102 : 2766 \u2013 2771 . doi : 10 . 1073 / pnas . 0409719102 . Chen H , Fre S , Slepnev VI , Capua MR , Takei K , Butler MH , Di Fiore PP , De Camilli P . 1998 . Epsin is an EH - domain - binding protein implicated in clathrin - mediated endocytosis . Nature 394 : 793 \u2013 797 . doi : 10 . 1038 / 29555 . Chen H , Ko G , Zatti A , Di Giacomo G , Liu L , Raiteri E , Perucco E , Collesi C , Min W , Zeiss C , De Camilli P , Cremona O . 2009 . Embryonic arrest at midgestation and disruption of notch signaling produced by the absence of both epsin 1 and epsin 2 in mice . Proceedings of the National Academy of Sciences of USA 106 : 13838 \u2013 13843 . doi : 10 . 1073 / pnas . 0907008106 . Chen H , Polo S , Di Fiore PP , De Camilli P . 2003 . Rapid Ca2 + - dependent decrease of protein ubiquitination at synapses . Proceedings of the National Academy of Sciences of USA 100 : 14908 \u2013 14913 . doi : 10 . 1073 / pnas . 2136625100 . Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 23 of 25 Research article Chen H , Slepnev VI , Di Fiore PP , De Camilli P . 1999 . The interaction of epsin and Eps15 with the clathrin adaptor AP - 2 is inhibited by mitotic phosphorylation and enhanced by stimulation - dependent dephosphorylation in nerve terminals . The Journal of Biological Chemistry 274 : 3257 \u2013 3260 . doi : 10 . 1074 / jbc . 274 . 6 . 3257 . Chen X , Zhang B , Fischer JA . 2002 . A specific protein substrate for a deubiquitinating enzyme : liquid facets is the substrate of fat facets . Genes & Development 16 : 289 \u2013 294 . doi : 10 . 1101 / gad . 961502 . Cheng J , Grassart A , Drubin DG . 2012 . Myosin 1E coordinates actin assembly and cargo trafficking during clathrin - mediated endocytosis . Molecular Biology of the Cell 23 : 2891 \u2013 2904 . doi : 10 . 1091 / mbc . E11 - 04 - 0383 . Drake MT . 2000 . Epsin binds to clathrin by associating directly with the clathrin - terminal domain . Evidence for cooperative binding through two discrete sites . Journal of Biological Chemistry 275 : 6479 \u2013 6489 . doi : 10 . 1074 / jbc . 275 . 9 . 6479 . Engqvist - Goldstein AE , Kessels MM , Chopra VS , Hayden MR , Drubin DG . 1999 . An actin - binding protein of the Sla2 / Huntingtin interacting protein 1 family is a novel component of clathrin - coated pits and vesicles . The Journal of Cell Biology 147 : 1503 \u2013 1518 . doi : 10 . 1083 / jcb . 147 . 7 . 1503 . Engqvist - Goldstein AE , Warren RA , Kessels MM , Keen JH , Heuser J , Drubin DG . 2001 . The actin - binding protein Hip1R associates with clathrin during early stages of endocytosis and promotes clathrin assembly in vitro . The Journal of Cell Biology 154 : 1209 \u2013 1223 . doi : 10 . 1083 / jcb . 200106089 . Engqvist - Goldstein AE , Zhang CX , Carreno S , Barroso C , Heuser JE , Drubin DG . 2004 . RNAi - mediated Hip1R silencing results in stable association between the endocytic machinery and the actin assembly machinery . Molecular Biology of the Cell 15 : 1666 \u2013 1679 . doi : 10 . 1091 / mbc . E03 - 09 - 0639 . Ferguson SM , Brasnjo G , Hayashi M , W\u00f6lfel M , Collesi C , Giovedi S , Raimondi A , Gong L - W , Ariel P , Paradise S , O ' toole E , Flavell R , Cremona O , Miesenb\u00f6ck G , Ryan TA , De Camilli P . 2007 . A selective activity - dependent requirement for dynamin 1 in synaptic vesicle endocytosis . Science 316 : 570 \u2013 574 . doi : 10 . 1126 / science . 1140621 . Ferguson SM , De Camilli P . 2012 . Dynamin , a membrane - remodelling GTPase . Nature Reviews Molecular Cell Biology 13 : 75 \u2013 88 . doi : 10 . 1038 / nrm3266 . Ferguson SM , Raimondi A , Paradise S , Shen H , Mesaki K , Ferguson A , Destaing O , Ko G , Takasaki J , Cremona O , O Toole E , De Camilli P . 2009 . Coordinated actions of actin and BAR proteins upstream of dynamin at endocytic clathrin - coated pits . Developmental Cell 17 : 811 \u2013 822 . doi : 10 . 1016 / j . devcel . 2009 . 11 . 005 . Ford MG , Mills IG , Peter BJ , Vallis Y , Praefcke GJ , Evans PR , McMahon HT . 2002 . Curvature of clathrin - coated pits driven by epsin . Nature 419 : 361 \u2013 366 . doi : 10 . 1038 / nature01020 . Gottfried I , Ehrlich M , Ashery U . 2010 . The Sla2p / HIP1 / HIP1R family : similar structure , similar function in endocytosis ? Biochemical Society Transactions 38 : 187 \u2013 191 . doi : 10 . 1042 / BST0380187 . Hawryluk MJ , Keyel PA , Mishra SK , Watkins SC , Heuser JE , Traub LM . 2006 . Epsin 1 is a Polyubiquitin \u2010 Selective Clathrin \u2010 Associated Sorting Protein . Traffic 7 : 262 \u2013 281 . doi : 10 . 1111 / j . 1600 - 0854 . 2006 . 00383 . x . Hirst J , Miller SE , Taylor MJ , Mollard von GF , Robinson MS . 2004 . EpsinR is an adaptor for the SNARE protein Vti1b . Molecular Biology of the Cell 15 : 5593 \u2013 5602 . doi : 10 . 1091 / mbc . E04 - 06 - 0468 . Itoh T , Erdmann KS , Roux A , Habermann B , Werner H , De Camilli P . 2005 . Dynamin and the actin cytoskeleton cooperatively regulate plasma membrane invagination by BAR and F - BAR proteins . Developmental Cell 9 : 791 \u2013 804 . doi : 10 . 1016 / j . devcel . 2005 . 11 . 005 . Itoh T , Koshiba S , Kigawa T , Kikuchi A , Yokoyama S , Takenawa T . 2001 . Role of the ENTH domain in phosphati - dylinositol - 4 , 5 - bisphosphate binding and endocytosis . Science 291 : 1047 \u2013 1051 . doi : 10 . 1126 / science . 291 . 5506 . 1047 . Kaksonen M , Sun Y , Drubin DG . 2003 . A pathway for association of receptors , adaptors , and actin during endocytic internalization . Cell 115 : 475 \u2013 487 . doi : 10 . 1016 / S0092 - 8674 ( 03 ) 00883 - 3 . Kaksonen M , Toret CP , Drubin DG . 2006 . Harnessing actin dynamics for clathrin - mediated endocytosis . Nature Reviews Molecular Cell Biology 7 : 404 \u2013 414 . doi : 10 . 1038 / nrm1940 . Kaur S , Fielding AB , Gassner G , Carter NJ , Royle SJ . 2014 . An unmet actin requirement explains the mitotic inhibition of clathrin - mediated endocytosis . eLife 3 : e00829 . doi : 10 . 7554 / eLife . 00829 . Ko G , Paradise S , Chen H , Graham M , Vecchi M , Bianchi F , Cremona O , Di Fiore PP , De Camilli P . 2010 . Selective high - level expression of epsin 3 in gastric parietal cells , where it is localized at endocytic sites of apical canaliculi . Proceedings of the National Academy of Sciences of USA 107 : 21511 \u2013 21516 . doi : 10 . 1073 / pnas . 1016390107 . Koo SJ , Markovic S , Puchkov D , Mahrenholz CC , Beceren - Braun F , Maritzen T , Dernedde J , Volkmer R , Oschkinat H , Haucke V . 2011 . SNARE motif - mediated sorting of synaptobrevin by the endocytic adaptors clathrin assembly lymphoid myeloid leukemia ( CALM ) and AP180 at synapses . Proceedings of the National Academy of Sciences of USA 108 : 13540 \u2013 13545 . doi : 10 . 1073 / pnas . 1107067108 . Le Clainche C , Pauly BS , Zhang CX , Engqvist - Goldstein AEY , Cunningham K , Drubin DG . 2007 . A Hip1R \u2013 cortactin complex negatively regulates actin assembly associated with endocytosis . The EMBO Journal 26 : 1199 \u2013 1210 . doi : 10 . 1038 / sj . emboj . 7601576 . Legendre - Guillemin V , Metzler M , Charbonneau M , Gan L , Chopra V , Philie J , Hayden MR , McPherson PS . 2002 . HIP1 and HIP12 display differential binding to F - actin , AP2 , and clathrin . Identification of a novel interaction with clathrin light chain . The Journal of Biological Chemistry 277 : 19897 \u2013 19904 . doi : 10 . 1074 / jbc . M112310200 . Liu Z , Zheng Y . 2009 . A requirement for epsin in mitotic membrane and spindle organization . The Journal of Cell Biology 186 : 473 \u2013 480 . doi : 10 . 1083 / jcb . 200902071 . Meloty - Kapella L , Shergill B , Kuon J , Botvinick E , Weinmaster G . 2012 . Notch ligand endocytosis generates mechanical pulling force dependent on dynamin , epsins , and actin . Developmental Cell 22 : 1299 \u2013 1312 . doi : 10 . 1016 / j . devcel . 2012 . 04 . 005 . Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 24 of 25 Research article Merrifield CJ , Perrais D , Zenisek D . 2005 . Coupling between clathrin - coated - pit invagination , cortactin recruitment , and membrane scission observed in live cells . Cell 121 : 593 \u2013 606 . doi : 10 . 1016 / j . cell . 2005 . 03 . 015 . Metzler M , Legendre - Guillemin V , Gan L , Chopra V , Kwok A , McPherson PS , Hayden MR . 2001 . HIP1 functions in clathrin - mediated endocytosis through binding to clathrin and adaptor protein 2 . The Journal of Biological Chemistry 276 : 39271 \u2013 39276 . doi : 10 . 1074 / jbc . C100401200 . Miller SE , Collins BM , McCoy AJ , Robinson MS , Owen DJ . 2007 . A SNARE - adaptor interaction is a new mode of cargo recognition in clathrin - coated vesicles . Nature 450 : 570 \u2013 574 . doi : 10 . 1038 / nature06353 . Miller SE , Sahlender DA , Graham SC , H\u00f6ning S , Robinson MS , Peden AA , Owen DJ . 2011 . The molecular basis for the endocytosis of small R - SNAREs by the clathrin adaptor CALM . Cell 147 : 1118 \u2013 1131 . doi : 10 . 1016 / j . cell . 2011 . 10 . 038 . Musse AA , Meloty - Kapella L , Weinmaster G . 2012 . Notch ligand endocytosis : mechanistic basis of signaling activity . Seminars in Cell & Developmental Biology 23 : 429 \u2013 436 . doi : 10 . 1016 / j . semcdb . 2012 . 01 . 011 . Overstreet E , Chen X , Wendland B , Fischer JA . 2003 . Either part of a Drosophila epsin protein , divided after the ENTH domain , functions in endocytosis of delta in the developing eye . Current Biology 13 : 854 \u2013 860 . doi : 10 . 1016 / S0960 - 9822 ( 03 ) 00326 - 9 . Pasula S , Cai X , Dong Y , Messa M , McManus J , Chang B , Liu X , Zhu H , Mansat RS , Yoon S - J , Hahn S , Keeling J , Saunders D , Ko G , Knight J , Newton G , Luscinskas F , Sun X , Towner R , Lupu F , Xia L , Cremona O , De Camilli P , Min W , Chen H . 2012 . Endothelial epsin deficiency decreases tumor growth by enhancing VEGF signaling . The Journal of Clinical Investigation 122 : 4424 \u2013 4438 . doi : 10 . 1172 / JCI64537 . Polo S , Sigismund S , Faretta M , Guidi M , Capua MR , Bossi G , Chen H , De Camilli P , Di Fiore PP . 2002 . A single motif responsible for ubiquitin recognition and monoubiquitination in endocytic proteins . Nature 416 : 451 \u2013 455 . doi : 10 . 1038 / 416451a . Rosenthal JA , Chen H , Slepnev VI , Pellegrini L , Salcini AE , Di Fiore PP , De Camilli P . 1999 . The epsins define a family of proteins that interact with components of the clathrin coat and contain a new protein module . The Journal of Biological Chemistry 274 : 33959 \u2013 33965 . doi : 10 . 1074 / jbc . 274 . 48 . 33959 . Ross\u00e9 C , L ' Hoste S , Offner N , Picard A , Camonis J . 2003 . RLIP , an effector of the Ral GTPases , is a platform for Cdk1 to phosphorylate epsin during the switch off of endocytosis in mitosis . The Journal of Biological Chemistry 278 : 30597 \u2013 30604 . doi : 10 . 1074 / jbc . M302191200 . Roux A , Uyhazi K , Frost A , De Camilli P . 2006 . GTP - dependent twisting of dynamin implicates constriction and tension in membrane fission . Nature 441 : 528 \u2013 531 . doi : 10 . 1038 / nature04718 . Royle SJ . 2013 . Protein adaptation : mitotic functions for membrane trafficking proteins . Nature Reviews Molecular Cell Biology 14 : 592 \u2013 599 . doi : 10 . 1038 / nrm3641 . Schmid SL . 1997 . Clathrin - coated vesicle formation and protein sorting : an integrated process . Annual Review of Biochemistry 66 : 511 \u2013 548 . doi : 10 . 1146 / annurev . biochem . 66 . 1 . 511 . Shergill B , Meloty - Kapella L , Musse AA , Weinmaster G , Botvinick E . 2012 . Optical tweezers studies on notch : single - molecule interaction strength is independent of ligand endocytosis . Developmental Cell 22 : 1313 \u2013 1320 . doi : 10 . 1016 / j . devcel . 2012 . 04 . 007 . Shih SC , Katzmann DJ , Schnell JD , Sutanto M , Emr SD , Hicke L . 2002 . Epsins and Vps27p / Hrs contain ubiquitin - binding domains that function in receptor endocytosis . Nature Cell Biology 4 : 389 \u2013 393 . doi : 10 . 1038 / ncb790 . Shupliakov O , Bloom O , Gustafsson JS , Kjaerulff O , Low P , Tomilin N , Pieribone VA , Greengard P , Brodin L . 2002 . Impaired recycling of synaptic vesicles after acute perturbation of the presynaptic actin cytoskeleton . Proceedings of the National Academy of Sciences of USA 99 : 14476 \u2013 14481 . doi : 10 . 1073 / pnas . 212381799 . Sigismund S , Woelk T , Puri C , Maspero E , Tacchetti C , Transidico P , Di Fiore PP , Polo S . 2005 . Clathrin - independent endocytosis of ubiquitinated cargos . Proceedings of the National Academy of Sciences of USA 102 : 2760 \u2013 2765 . doi : 10 . 1073 / pnas . 0409817102 . Skruzny M , Brach T , Ciuffa R , Rybina S , Wachsmuth M , Kaksonen M . 2012 . Molecular basis for coupling the plasma membrane to the actin cytoskeleton during clathrin - mediated endocytosis . Proceedings of the National Academy of Sciences of USA 109 : E2533 \u2013 E2542 . doi : 10 . 1073 / pnas . 1207011109 . Slepnev VI , De Camilli P . 2000 . Accessory factors in clathrin - dependent synaptic vesicle endocytosis . Nature Reviews Neuroscience 1 : 161 \u2013 172 . doi : 10 . 1038 / 35044540 . Sochacki KA , Shtengel G , van Engelenburg SB , Hess HF , Taraska JW . 2014 . Correlative super - resolution fluores - cence and metal - replica transmission electron microscopy . Nature Methods 11 : 305 \u2013 308 . doi : 10 . 1038 / nmeth . 2816 . Stukenberg PT , Lustig KD , McGarry TJ , King RW , Kuang J , Kirschner MW . 1997 . Systematic identification of mitotic phosphoproteins . Current Biology 7 : 338 \u2013 348 . doi : 10 . 1016 / S0960 - 9822 ( 06 ) 00157 - 6 . Taylor MJ , Lampe M , Merrifield CJ . 2012 . A feedback loop between dynamin and actin recruitment during clathrin - mediated endocytosis . PLOS Biology 10 : e1001302 . doi : 10 . 1371 / journal . pbio . 1001302 . Taylor MJ , Perrais D , Merrifield CJ . 2011 . A high precision survey of the molecular dynamics of mammalian clathrin - mediated endocytosis . PLOS Biology 9 : e1000604 . doi : 10 . 1371 / journal . pbio . 1000604 . Tronche F , Kellendonk C , Kretz O , Gass P , Anlag K , Orban PC , Bock R , Klein R , Sch\u00fctz G . 1999 . Disruption of the glucocorticoid receptor gene in the nervous system results in reduced anxiety . Nature Genetics 23 : 99 \u2013 103 . doi : 10 . 1038 / 12703 . Wendland B . 1999 . Yeast epsins contain an essential N - terminal ENTH domain , bind clathrin and are required for endocytosis . The EMBO Journal 18 : 4383 \u2013 4393 . doi : 10 . 1093 / emboj / 18 . 16 . 4383 . Wendland B . 2002 . Opinion : epsins : adaptors in endocytosis ? Nature Reviews Molecular Cell Biology 3 : 971 \u2013 977 . doi : 10 . 1038 / nrm970 . Cell biology Messa et al . eLife 2014 ; 3 : e03311 . DOI : 10 . 7554 / eLife . 03311 25 of 25 Research article Wilbur JD , Chen C - Y , Manalo V , Hwang PK , Fletterick RJ , Brodsky FM . 2008 . Actin binding by Hip1 ( huntingtin - interacting protein 1 ) and Hip1R ( Hip1 - related protein ) is regulated by clathrin light chain . The Journal of Biological Chemistry 283 : 32870 \u2013 32879 . doi : 10 . 1074 / jbc . M802863200 . Wu M , De Camilli P . 2012 . Supported native plasma membranes as platforms for the reconstitution and visualization of endocytic membrane budding . Methods in Cell Biology 108 : 1 \u2013 18 . doi : 10 . 1016 / B978 - 0 - 12 - 386487 - 1 . 00001 - 8 . Wu M , Huang B , Graham M , Raimondi A , Heuser JE , Zhuang X , De Camilli P . 2010 . Coupling between clathrin - dependent endocytic budding and F - BAR - dependent tubulation in a cell - free system . Nature Cell Biology 12 : 902 \u2013 908 . doi : 10 . 1038 / ncb2094 . Xie X , Cho B , Fischer JA . 2012 . Drosophila Epsin ' s role in Notch ligand cells requires three Epsin protein functions : the lipid binding function of the ENTH domain , a single Ubiquitin interaction motif , and a subset of the C - terminal protein binding modules . Developmental Biology 363 : 399 \u2013 412 . doi : 10 . 1016 / j . ydbio . 2012 . 01 . 004 . Yang S , Cope MJ , Drubin DG . 1999 . Sla2p is associated with the yeast cortical actin cytoskeleton via redundant localization signals . Molecular Biology of the Cell 10 : 2265 \u2013 2283 . doi : 10 . 1091 / mbc . 10 . 7 . 2265 . Zoncu R , Perera RM , Sebastian R , Nakatsu F , Chen H , Balla T , Ayala G , Toomre D , De Camilli PV . 2007 . Loss of endocytic clathrin - coated pits upon acute depletion of phosphatidylinositol 4 , 5 - bisphosphate . Proceedings of the National Academy of Sciences of USA 104 : 3793 \u2013 3798 . doi : 10 . 1073 / pnas . 0611733104 .",
    "arasada2018highspeed": "Volume 29 February 1 , 2018 295 MBoC | ARTICLE High - speed superresolution imaging of the proteins in fission yeast clathrin - mediated endocytic actin patches ABSTRACT To internalize nutrients and cell surface receptors via clathrin - mediated endocy - tosis , cells assemble at least 50 proteins , including clathrin , clathrin - interacting proteins , actin filaments , and actin binding proteins , in a highly ordered and regulated manner . The mole - cular mechanism by which actin filament polymerization deforms the cell membrane is un - known , largely due to lack of knowledge about the organization of the regulatory proteins and actin filaments . We used high - speed superresolution localization microscopy of live fis - sion yeast cells to improve the spatial resolution to \u223c 35 nm with 1 - s temporal resolution . The nucleation promoting factors Wsp1p ( WASp ) and Myo1p ( myosin - I ) define two independent pathways that recruit Arp2 / 3 complex , which assembles two zones of actin filaments . Myo1p concentrates at the site of endocytosis and initiates a zone of actin filaments assembled by Arp2 / 3 complex . Wsp1p appears simultaneously at this site but subsequently moves away from the cell surface as it stimulates Arp2 / 3 complex to assemble a second zone of actin fila - ments . Cells lacking either nucleation - promoting factor assemble only one , stationary , zone of actin filaments . These observations support our two - zone hypothesis to explain endocytic tubule elongation and vesicle scission in fission yeast . INTRODUCTION Clathrin - mediated endocytosis recycles membrane receptors and takes up nutrients . Studies of budding yeast , fission yeast , and ani - mal cells identified many proteins that assemble and disassemble at endocytic sites . Recruitment of membrane proteins that recognize the endocytic cargo initiates the process at nascent endocytic sites . These sites mature with the assembly of a clathrin coat and recruit - ment of nucleation promoting factors and Arp2 / 3 complex that stim - ulate actin polymerization . Yeast cells use mechanical force provided by actin polymerization to overcome the internal turgor pressure and deform the membrane ( Aghamohammadzadeh and Ayscough , 2009 ; Minc et al . , 2009 ; Schaber et al . , 2010 ; Basu et al . , 2014 ) . However , it is still unclear how actin polymerization generates force to reshape the membrane , because conventional fluorescence microscopy has not resolved the organization of actin filaments or established the sites of actin polymerization . One model based on the localization of the nucleation promoting factors , myosin - I and WASp ( Las17 ) in budding yeast , proposes that the actin filaments polymerize against the plasma membrane at the neck of endocytic invaginations . Consequently , the barbed ends of the elongating ac - tin filaments are oriented toward the plasma membrane at the base of the membrane invagination with their pointed ends anchored to the clathrin coat via the adapter proteins at the invagination tip ( Kaksonen et al . , 2005 ; Skruzny et al . , 2012 ; Picco et al . , 2015 ) . Actin polymerization primarily at the plasma membrane together with some myosin motor activity is proposed to generate force to elon - gate the membrane tubule . On the basis of the localization of the nucleation promoting fac - tors in fission yeast , we proposed another model to explain how actin polymerization drives endocytosis ( Arasada and Pollard , 2011 ) . In our model , actin polymerization occurs in two distinct zones : one Monitoring Editor David G . Drubin University of California , Berkeley Received : Jun 21 , 2017 Revised : Nov 28 , 2017 Accepted : Nov 28 , 2017 This article was published online ahead of print in MBoC in Press ( http : / / www . molbiolcell . org / cgi / doi / 10 . 1091 / mbc . E17 - 06 - 0415 ) on December 6 , 2017 . * Address correspondence to : Thomas D . Pollard ( thomas . pollard @ yale . edu ) . \u00a9 2018 Arasada et al . This article is distributed by The American Society for Cell Biology under license from the author ( s ) . Two months after publication it is avail - able to the public under an Attribution \u2013 Noncommercial \u2013 Share Alike 3 . 0 Unported Creative Commons License ( http : / / creativecommons . org / licenses / by - nc - sa / 3 . 0 ) . \u201cASCB \u00ae , \u201d \u201cThe American Society for Cell Biology \u00ae , \u201d and \u201cMolecular Biology of the Cell \u00ae \u201d are registered trademarks of The American Society for Cell Biology . Abbreviations used : ARPC5 , Arp2 / 3 complex subunit 5 ; CHD , calponin homology domain ; DIC , differential interference contrast ; EMM5S , Edinburgh minimal me - dium 5 supplements ; Fim , fimbrin ; FPALM , fluorescence photoactivation localiza - tion microscopy ; mEGFP , monomeric enhanced green fluorescent protein ; Myo1p , myosin - I ; PALM , photoactivated localization microscopy ; sCMOS , scien - tific complementary metal - oxide - semiconductor ; STORM , stochastic optical re - construction microscopy ; Wsp1p , Wiskott - Aldrich syndrome protein . Rajesh Arasada a , Wasim A . Sayyad a , Julien Berro b , c , d , and Thomas D . Pollard a , b , c , * a Departments of Molecular Cellular and Developmental Biology , b Departments of Molecular Biophysics and Biochemistry , c Department of Cell Biology , and d Nanobiology Institute , Yale University , New Haven , CT 06520 - 8103 296 | R . Arasada et al . Molecular Biology of the Cell Over time , all of the fluorescent proteins within the \u223c 400 - nm - thick imaging plane were localized with an average radial precision of 21 \u00b1 4 nm , and the centroids of each molecule were plotted as two - dimensional histograms ( Figure 1B , middle panel ) . Software rejected molecules outside the imaging plane during image processing . To aid visualization the two - dimensional histograms of centroids of at the base of the membrane invagination and the other traveling with the tip of the invagination . This model postulates that the ex - panding actin networks in the two zones repel each other as they grow , driving invagination of the plasma membrane tubule . This two - zone model is based on quantitative measurements by confocal fluorescence microscopy of fission yeast actin patch pro - teins fused to mEGFP or mYFP ( Sirotkin et al . , 2010 ; Arasada and Pollard , 2011 ) . Two activators of Arp2 / 3 complex , type - I myosin Myo1p and WASp homologue Wsp1p , localize and initiate actin po - lymerization together but then separate with Myo1p at the base of plasma membrane invagination and Wsp1p at the tip of the invagi - nation ( Figure 1A ) . These nucleation promoting factors recruit very high densities ( \u223c 20 , 000 molecules per \u00b5m 2 ) of two F - BAR ( Fes / CIP4 homology Bin - Amphiphysin - Rvs ) proteins , Cdc15p and Bzz1p , onto the invaginating membrane tubule . Cdc15p interacts with Myo1p at the base of the membrane tubule and Bzz1p interacts with Wsp1p at the invagination tip . The F - BAR proteins interact with the nucle - ation promoting factors to stimulate Arp2 / 3 complex in two zones along the invaginating tubule . However , in yeast cells expressing GFP - actin or GFP - tagged proteins that bind actin filaments , the dense network of \u223c 5000 polymerized actin molecules appears as a single mass of homogeneous fluorescence \u223c 500 nm in diameter at sites of endocytosis rather than as two distinct polymerization zones ( Arasada and Pollard , 2011 ) . Better spatial resolution was required to determine whether actin polymerizes in two zones . Since the two yeast cells diverged from a common ancestor \u223c 400 million years ago and have adapted differently during their subse - quent evolution , they may control actin assembly during endocyto - sis in different ways . On the other hand , it is worth considering whether endocytosis in the two yeasts has more in common than suggested by these two models . We used high - speed fluorescence photoactivation localization microscopy ( FPALM ) of live Schizosaccharomyces pombe cells ex - pressing photoconvertible fluorescent proteins ( Huang et al . , 2013 ; Laplante et al . , 2016a , b ) to study the organization of proteins in ac - tin patches . We obtained 1 - s temporal resolution and \u223c 35 - nm spa - tial resolution and observed that actin filaments assemble in two zones : one at the base of the membrane invagination and the other at the tip of the invagination . Polymerization of actin in both zones depends on Arp2 / 3 complex , which is activated by type - I myosin Myo1p at the base of the invagination and WASp homologue Wsp1p along the path of the invagination from the cell surface to the tip of the membrane tubule . These observations support the two - zone hypothesis for the assembly of actin in association with endocytosis in fission yeast . RESULTS Observations of protein turnover in actin patches of live fission yeast cells by FPALM We examined actin filaments and four components of the actin as - sembly machinery in live fission yeast using a single - molecule , su - perresolution method , FPALM ( also called STORM or PALM ) ( Huang et al . , 2013 ; Laplante et al . , 2016a ) . We tagged proteins with the photoconvertible fluorescent proteins mMaple3 or mEOS3 . 2 at their gene loci unless indicated otherwise . The cell in Figure 1 , B and C expressed the actin patch component capping protein Acp2p tagged with mEOS3 . 2 . We used epifluorescence illumination to ex - pose live S . pombe cells expressing photoconvertible fluorescent proteins continuously with both a near UV laser ( 405 nm ) to photo - convert the fluorescent proteins randomly to the state that emits red light and a yellow laser ( 564 nm ) to image the red light emitted by individual , spatially separated , photoconverted fluorescent proteins . FIGURE 1 : FPALM improves the spatial resolution of actin patches over conventional fluorescence microscope . ( A ) Two - zone model based on quantitative confocal fluorescence microscopy . The cartoon represents an actin patch at \u2013 2 s when the membrane tubule elongates and at 0 s when the vesicle pinches away from the plasma membrane . The plasma membrane is a black line , the clathrin coat is gray , nucleation - promoting factors Wsp1p and Myo1p are yellow , and zones with actin filaments are blue . Growth of the branched filaments from two distinct zones of NPFs is postulated to push the tip of the invaginating tubule away from the cell surface contributing to the elongation of the tubule and scission of the coated vesicle . ( B , C ) Fluorescence micrographs of an S . pombe cell expressing capping protein Acp2p - mEOS3 . 2 with focusing in the middle plane of the cell . We used continuous epifluorescence illumination to photoconvert mEOS3 . 2 with 405 - and 564 - nm lasers to excite the photoconverted mEOS3 . 2 through the entire cell . Top panels , wide - field epifluorescence images reconstructed from the total fluorescence emission . Middle panels , raw FPALM images constructed from the localizations of single molecules . Bottom panels , each localized emitter in the raw data set was convolved with a Gaussian kernel ( \u03c3 = 1 . 5 pixel ) and color coded for density in a heat map . ( B ) Whole cell during a 1 - s interval . Scale bar is 1 \u00b5m . ( C ) Time series of images of one actin patch at 1 - s intervals each reconstructed from 200 sequential frames . Top panel , inverted contrast wide - field epifluorescence images . Scale bar is 250 nm . Volume 29 February 1 , 2018 Organization of S . pombe a ctin patches | 297 ( Figure 2B , center micrograph ) . The clusters of mMaple3 - Wsp1p maintained a constant size with a width of 100 \u00b1 38 nm and length of 200 nm ( Figure 2 , C and D , center graphs ) as they moved from the membrane proximal zone to the membrane distal zone ( a box \u223c 250 nm wide located in the cytoplasm between 200 and 550 nm from the cell edge ) ( Figure 2E , middle panel , and Supplemental Figure S1B ) . mMaple3 - Wsp1p appeared in the membrane proximal zone and peaked there at \u223c 50 localizations per second . The peak of \u223c 50 mMaple3 - Wsp1p localizations per second in the membrane distal zone was \u223c 6 s later . The localizations of Arc5p - mMaple3 ( the marker for Arp2 / 3 com - plex ) appeared first at the plasma membrane and then deeper in the cytoplasm at later times ( Figure 2A , bottom panel ) , as illustrated by composite images with temporal color coding ( Figure 2B , right micrograph ) . Throughout this shift in the localizations over time , clusters of Arc5p - mMaple3 were 122 \u00b1 62 nm wide and 300 nm long ( Figure 2 , C and D , right panels ) . Like mMaple3 - Wsp1p , the localiza - tions of Arc5p - mMaple3 molecules first peaked at \u223c 150 per second in the membrane proximal zone followed \u223c 5 s later by a peak in membrane distal zone ( Figure 2E , right panel , and Supplemental Figure S1C ) . Actin assembles in two distinct zones in an actin patch We localized actin filaments indirectly with two different probes : 1 ) an actin filament binding protein fimbrin ( Fim1p ) tagged to mEOS3 . 2 at its C - terminus and expressed from the endoge - nous locus ; and 2 ) calponin homology domain ( CHD ) from the S . pombe IQGAP protein Rng2p fused to mEOS3 . 2 at its N termi - nus and expressed ectopically from a repressible promoter at the leu1 + locus ( Figure 3 ) . Fim1p - mEOS3 . 2 expressed from the native promoter has minimal impact on the organization of actin filaments at endocytic sites ( Berro and Pollard , 2014b , a ) . CHD tagged with a fluorescent protein binds actin filaments , but it also appears to bundle actin filaments and may influence cell mor - phology when overexpressed . The time course of the appear - ance and disappearance of Fim1p - mEOS3 . 2 in patches ( Figure 3H ) was similar to a trace of Fim1p - mGFP or Fim1p - mCherry ob - served by confocal microscopy ( Sirotkin et al . , 2010 ; Berro and Pollard , 2014a ) . Both actin filament probes appeared in two distinct zones at sites of endocytosis ( Figure 3 , A and B ) , first close to the plasma mem - brane and then in a second zone further from the plasma mem - brane . Throughout their existence , the clusters of single molecule localizations of mEOS3 . 2 - CHD in patches were 110 \u00b1 55 nm wide ( along the cell long axis ) and 350 nm long , while those of Fim1p - mEOS3 . 2 were 101 \u00b1 36 nm wide and 300 nm long , both with two distinct peaks ( Figure 3 , C \u2013 F ) . Reconstructions of series of images ( Figure 3A ) with temporal color - coding emphasize that actin assem - bly begins at the plasma membrane and shifts to a site deeper in the cytoplasm over time ( Figure 3 , G and H ) . Like Arp2 / 3 complex , the mEOS3 . 2 - CHD and Fim1p - mEOS3 . 2 first peaked at \u223c 150 localiza - tions per second in the membrane proximal zone followed \u223c 4 s later by a peak in the membrane distal zone ( Figure 3 , I and J , and Sup - plemental Figure S1 , D and E ) . Starting at time 0 s , the localizations of each actin patch protein declined and appeared further from the plasma membrane . At the onset of patch assembly and up to time zero , the localizations of the patch proteins were tightly packed in a column normal to the plasma membrane . After time zero , the positions of these localizations be - came more variable and more spread out . The localizations colored in red at the end of the disassembly phase were more spread out and noisy due to fewer localizations . localized molecules were convolved with a two - dimensional Gauss - ian kernel ( \u03c3 = 7 . 5 nm ) and color coded for localization density ( Figure 1B , bottom panel ) . It is important to note that photoactivation localization mi - croscopy depends on irreversibly photobleaching each fluores - cent protein after it is imaged and localized , so a time series of FPALM images reveals the position of each molecule when it is photoconverted . Photobleaching occurs in less than 2 s under our conditions ( Laplante et al . , 2016b ) . However , the method does not reveal the subsequent history of the photobleached protein , such as its persistence at the site of localization or any motion . Actin patches concentrate at the poles of the cells but are spread around the surface , so they are viewed from many differ - ent angles . Therefore , we used differential interference contrast ( DIC ) microscopy to identify the midplane of the cells , where ac - tin patches moving in the XY - plane of the microscope were con - tained in the optical section throughout their lifetimes . To obtain enough single - molecule localizations to form an image , we com - bined groups of 200 sequential 5 - ms frames . These 1 - s compos - ite images were linked into movies or montages to show actin patch dynamics . These FPALM images highlight structural details of actin patches that appear as blurred , pixelated spots of fluo - rescence by wide - field fluorescence micro scopy ( Figure 1C , top panel ) . Localization of Arp2 / 3 complex activators and Arp2 / 3 complex We tagged three proteins that control the assembly of actin fila - ments at sites of endocytosis and established their distributions over time in actin patches . We inserted the open reading frame for mMaple3 ( Wang et al . , 2014 ) into the genome so the fluorescent protein was fused to the N - termini of Arp2 / 3 complex activators myosin - I ( Myo1p ) and Wiskott - Aldrich syndrome protein ( Wsp1p ) and on the C - terminus of the Arc5p ( ARPC5 ) subunit of Arp2 / 3 complex , so all of the tagged proteins were expressed from their endogenous promoters . Schizosaccharomyces pombe cells ex - pressing these fusion proteins were viable and had wild - type morphologies at 25\u00b0 and 36\u00b0C . In wild - type cells , localizations of mMaple3 - Myo1p appeared in a small , stationary region 65 \u00b1 18 nm ( mean \u00b1 SD ) wide and extending 85 \u00b1 18 nm from the plasma membrane ( Figure 2 , A , C , and D , left graphs ) as actin patches assembled and disassembled over time ( Figure 2A , top panel ) . A composite image with tempo - ral color coding according to the time of localization shows the time course of the whole process ( Figure 2B , left micrograph ) . Counts of mMaple3 - Myo1p localizations in actin patches over time established that 82 % of Myo1p localizations appeared and disappeared within membrane proximal zone ( defined as box \u223c 250 nm wide , extending 200 nm into the cytoplasm from the cell edge ) ( left panels in Figure 2B and and Supplemental Figure S1A ) . These observations confirmed at much higher resolution the lack of mobility of Myo1p tagged with mEGFP ( monomeric enhanced green fluorescent protein ) and imaged by confocal microscopy ( Sirotkin et al . , 2005 , 2010 ) . Over \u223c 15 s , mMaple3 - Wsp1p appeared at the plasma mem - brane and subsequently was localized along a trajectory normal to the membrane ( middle column of Figure 2 , B \u2013 E ) , similarly to lower - resolution confocal micrographs of cells expressing mEGFP - Wsp1p ( Sirotkin et al . , 2010 ) . Reconstructions of time series with temporal color - coding illustrate that mMaple3 - Wsp1p molecules were local - ized up to 300 nm from the plasma membrane at late time points 298 | R . Arasada et al . Molecular Biology of the Cell Assembly of two zones of actin is required for endocytic tubule elongation and vesicle scission from the plasma membrane Both Myo1p and Wsp1p contribute to ac - tivating Arp2 / 3 complex and the assembly actin filaments at sites of endocytosis in fis - sion yeast . Both \u2206 wsp1 and \u2206 myo1 mutant cells are viable with defects in endocytosis , but double mutant \u2206 wsp1 \u2206 myo1 cells are not viable ( Lee et al . , 2000 ; Sirotkin et al . , 2005 ) . We investigated the relationships among Myo1p , Wsp1p , Arp2 / 3 complex , and actin filaments by analyzing \u2206 wsp1 yeast cells lacking Wsp1p or \u2206 myo1 cells lacking Myo1p ( Figure 4 ) . We compared the distribution of Arc5p - mMaple3 and mEOS3 . 2 - CHD in the mutant strains with wild - type cells . Both single deletion strains differed from wild - type cells in two regards : 1 ) they accumulated far less Arc5p - mMaple3 than FIGURE 2 : Localization of nucleation promoting factors and Arp2 / 3 complex in actin patches by FPALM in wild - type cells . ( A ) Time series of FPALM images color coded for the density of localizations : top , mMaple3 - Myo1p ; middle , mMaple3 - Wsp1p ; and bottom , Arc5p - mMaple3 a subunit of Arp2 / 3 complex . The micrographs were aligned to time 0 s when the vesicle pinches away from the plasma membrane . For mMaple3 - Myo1p the peak localizations were aligned to the peak localizations of mMaple3 - Wsp1p . Yellow dashed lines show the position of the plasma membrane . Scale bar is 100 nm . ( B ) Composite images with temporal color coding of single molecule localizations over the lifetimes of single actin patches : left , 18 s from a cell expressing mMaple3 - Myo1p ; center , 16 s from a cell expressing mMaple3 - Wsp1p ; and right , 21 s from a cell expressing Arc5p - mMaple3 a subunit of Arp2 / 3 complex . The blue and red lines indicate the membrane - proximal and membrane - distal regions . Scale bar is 200 nm . ( C , D ) Spatial distributions of single molecule localizations in FPALM images of actin patches . FWHM is full - width of the distributions at half - maximums . ( C ) Width distributions and ( D ) length distributions of localizations of left , mMaple3 - Myo1p in 20 actin patches ; center , mMaple3 - Wsp1p in 12 actin patches ; and right , Arc5p - mMaple3 in 18 actin patches . The data from separate patches were aligned to their peaks and averaged . Full - widths at half - maximum were calculated from the distributions of the localizations along the x - axis . The lengths of the patches are given as the distances that include 90 % of the localizations from the cell edge . The gray area in the graphs represents the localizations detected outside the cell . ( E ) Time courses of the average number of localizations detected in ( blue line ) the membrane proximal zone ( 200 \u00d7 250 - nm region next to the cell edge ) and ( red line ) the membrane distal zone ( 350 \u00d7 250 - nm region 200 nm away from the cell edge ) : left , mMaple3 - Myo1p ; middle , mMaple3 - Wsp1p ; and right , Arc5p - mMaple3 . Raw time courses of the number of emitters localized in 15 actin patches were aligned to a reference patch and averaged to produce time courses for the average number of localizations . Mean total localizations were used to align the localizations in the membrane proximal and distal zone on the same time scale as the numbers of GFP molecules assembled in actin patches . Time 0 s indicates vesicle scission . Error bars are SDs . Every third error bar is shown . Volume 29 February 1 , 2018 Organization of S . pombe a ctin patches | 299 wild - type cells ( Figure 4A ) ; and 2 ) the localizations of Arc5p - mMaple3 appeared and remained next to the plasma membrane ( Figure 4 , A and B ) as actin patches assembled and disassembled over time . Thus , few Arc5p - mMaple3 molecules were localized in the membrane distal zone at any time ( Figure 4E and Supplemen - tal Figure S2A ) . In cells lacking either Myo1p or Wsp1p , the actin filament marker mEOS3 . 2 - CHD remained close to the plasma membrane and per - sisted there longer than in wild - type cells ( Figure 4F and Suppleme - mental Figure S2B ) . The peak numbers of mEOS3 . 2 - CHD localiza - tions in membrane proximal zones of actin patches were similar in wild - type cells and cells without Myo1p or Wsp1p , but few of the localizations appeared in the membrane distal zone . DISCUSSION High - speed FPALM increases the spatial resolution nearly 10 - fold compared with confocal microscopy and is fast enough to follow the assembly and disassembly of actin patches . We used the estab - lished timing of GFP - tagged proteins ( Sirotkin et al . , 2010 ; Arasada and Pollard , 2011 ) to align the timing of the superresolution micro - scopy data , with 0 s as the time when some of the patch compo - nents start to move away from the cell surface after vesicle scission ( Figure 5 ) . High - resolution images aligned in this way at 1 - s intervals helped us distinguish elongation of the membrane tubule from ves - icle scission and support the two - zone hypothesis . Note that after random photoconversion FPALM irreversibly bleaches each fluorescent protein in the imaging beam in \u223c 2 s ( Laplante et al . , 2016b ) , so the method does not reveal how long each molecule persists at its detected location . Thus , the absence of localizations does not necessarily mean that the tagged protein has turned over . For example , actin may persist longer in the membrane proximal zone than the detection of newly localized proteins . Status of the two - zone hypothesis The key feature of the two - zone hypothesis is that the two major nucleation promoting factors , Myo1p and Wsp1p , are distributed differently on the membrane tubules , so they can drive the reactions FIGURE 3 : Localization of actin filaments in two zones in endocytic actin patches in wild - type S . pombe cells . ( A , B ) Time series of fluorescence micrographs at 1 - s intervals of representative actin patches in cells expressing proteins that bind actin filaments . The images are regions of interest of \u223c 500 \u00d7 500 nm around one actin patch for each protein . Scale bars are 250 nm . ( A ) GFP - CHD and mEOS3 . 2 - CHD and ( B ) Fim1p - GFP and Fim1p - mEOS3 . 2 . Top panels , maximum intensity projections of five confocal z - slices in the XY - plane of individual actin patches from cells expressing GFP - CHD or Fim1p - GFP . Bottom panels , FPLAM images of cells expressing mEOS3 . 2 - CHD or Fim1p - mEOS3 . 2 . Each localized emitter was convolved with a Gaussian kernel ( \u03c3 = 1 . 5 pixel ) and color coded for density in a heat map . Yellow dashed lines show the location of the plasma membrane ( PM ) . ( C \u2013 F ) Spatial distributions of single molecule localizations of mEOS3 . 2 - CHD and Fim1p - mEOS3 . 2 in FPALM images of actin patches . FWHM is full - width of the distributions at half - maximums . ( C ) Width distribution and ( E ) length distribution of mEOS3 . 2 - CHD in 20 actin patches . ( D ) Width distribution and ( F ) length distribution of Fim1p - mEOS3 . 2 in 14 actin patches prepared as in Figure 2 . ( G , H ) Composite images with temporal color coding of single molecule localizations over the lifetimes of single actin patches . The micrographs are aligned to time 0 s when the vesicle pinches off from the plasma membrane . Scale bars are 200 nm . The blue and red lines indicate the membrane - proximal and membrane - distal regions . ( G ) Twenty - one seconds from a cell expressing mEos3 . 2 - CHD . ( H ) Seventeen seconds from a cell expressing Fim1p - mEos3 . 2 . ( I , J ) Time courses of the average number of localizations of ( I ) CHD - mEOS3 . 2 and ( J ) Fim1p - mEOS . 2 detected in ( blue line ) the membrane proximal zone ( 200 \u00d7 250 - nm region next to the cell edge ) and ( red line ) the membrane distal zone ( 350 \u00d7 250 - nm region 200 nm away from the cell edge ) . Raw time courses of the number of emitters localized in 15 actin patches were aligned to a reference patch and averaged to produce time courses for the average number of localizations . Time zero seconds indicates vesicle scission . Mean total localizations were used to align the localizations in the membrane proximal and distal zone on the same time scale as the numbers of GFP molecules assembled in actin patches . Error bars are SDs . Every third error bar is shown . 300 | R . Arasada et al . Molecular Biology of the Cell that produce two zones of actin polymeriza - tion . Confocal microscopy showed that Myo1p and Wsp1p have lifetimes of \u223c 15 s in actin patches . They appear at the same time , peak together at time \u2013 3 s with \u223c 165 molecules of Myo1p and \u223c 140 molecules of Wsp1p , and disappear together around time + 6 s ( Sirotkin et al . , 2010 ; Arasada and Pollard , 2011 ) . Superresolution images show that Myo1p is stationary at the base of the membrane invagination , while Wsp1p first appears at the same place as Myo1p followed by new localizations further from the plasma membrane . This relocation of FIGURE 4 : Actin patch assembly in cells lacking nucleation promoting factors Myo1p or Wsp1p . ( A , B ) Time series of FPALM images color coded for localization density showing the ( A ) Arc5p - mMaple3 a subunit of Arp2 / 3 complex or ( B ) mEOS3 . 2 - CHD to visualize actin filaments . Top panels , wild - type cells ; middle panels , \u2206 myo1 cells ; and bottom panels , \u2206 wsp1 cells . Yellow dotted lines indicate the cell membrane . The micrographs are aligned to time 0 s when the vesicle pinches off from the plasma membrane . Scale bars are 250 nm . ( C , D ) Composite images with temporal color coding of single molecule localizations over the lifetimes of single actin patches . The blue and red lines indicate the membrane - proximal and membrane distal regions . ( C ) Cells expressing Arc5p - mMaple3 : left , 16 s of an actin patch in a wild - type cell ; center , 18 s of an actin patch in a \u2206 myo1 cell ; and right , 18 s of an actin patch in a \u2206 wsp1 cell . ( D ) Cells expressing mEOS3 . 2 - CHD : left , 15 s of an actin patch in a wild - type cell ; center , 27 s of an actin patch in a \u2206 myo1 cell ; and right , 28 s of an actin patch in a \u2206 wsp1 cells . ( E , F ) Time courses of the average number of localizations of ( E ) Arc5p - mMaple3 and ( F ) mEOS . 2 - CHD detected in actin patches in ( blue line ) the membrane proximal zone ( 200 \u00d7 250 - nm region next to the cell edge ) and ( red line ) the membrane distal zone ( 350 \u00d7 250 - nm region 200 nm away from the cell edge ) : left , wild - type cells ; middle , \u2206 myo1 cells ; and right , \u2206 wsp1 cells . Raw time courses of the number of emitters localized in 15 actin patches were aligned to a reference patch and averaged to produce time courses for the average number of localizations . Mean total localizations were used to align the localizations in the membrane proximal and distal zone on the same time scale as the numbers of GFP molecules assembled in actin patches . Time 0 s indicates the vesicle scission . Error bars are SDs . Every third error bar is shown . Volume 29 February 1 , 2018 Organization of S . pombe a ctin patches | 301 the ability of Myo1p to stimulate actin polymerization by Arp2 / 3 complex in fission yeast ( Arasada and Pollard , 2011 ) and phosphory - lation of the Myo1p TEDS site ( Serine 361 ) by Pak1p appears to be the upstream signal that localizes Myo1p at clathrin - coated pits ( Attanapola et al . , 2009 ) . In both yeasts , WASp initially concentrates near the plasma membrane and recruits the F - BAR protein Bzz1p , which increases the ability of Wsp1p to activate Arp2 / 3 complex to assemble actin filaments ( Padrick et al . , 2008 ; Arasada and Pollard , 2011 ) . The main difference is the subsequent behavior of WASp . The current work confirms that fission yeast Wsp1p moves away from the cell surface along with Arp2 / 3 complex and actin filaments ( Arasada and Pollard , 2011 ) , while multiple papers ( Kaksonen et al . , 2005 ; Sun et al . , 2006 , 2017 ; Rajmohan et al . , 2009 ; Picco et al . , 2015 ) show that budding yeast Las17 - GFP remains close to the cell sur - face during endocytosis . Consistent with a single zone of nucleation promoting factors in budding yeast , elegant photobleaching ex - periments showed that actin filaments appear at the base of the membrane invagination ( Picco et al . , 2015 ) . The accumulated work suggests that the behavior of WASp dif - fers in the two yeasts , supporting a single zone of actin assembly in budding yeast and two zones in fission yeast . Exploring this differ - ence is worthwhile for two reasons . First , tagging Las17 or Wsp1p can influence their behavior . Growth and the time course of patch assembly are normal in fission yeast cells dependent on mGFP - Wsp1p ( Sirotkin et al . , 2005 ) , rather than growing slowly like mis - shapen , temperature - sensitive \u2206 wsp1 cells . However , the GFP - Wsp1p strain has defects in mating and sporulation , as observed in \u2206 wsp1 cells . Tagging either end of Las17 with GFP may compro - mise patch mobility ( Galletta et al . , 2008 ) . Second , electron micros - copy of sections of chemically fixed budding yeast cells treated with gold - labeled antibodies showed that Las17 with a small C - terminal HA peptide tag distributed significantly further than myo - sin - I along endocytic membrane invaginations as they elongated ( p < 0 . 001 ) . However , chemical fixation may have compromised the preservation . High - pressure freezing followed by freeze substitu - tion improves the preservation and is compatible with antibody staining for electron microscopy ( Buser and Drubin , 2013 ) or cor - relative fluorescence and electron tomography ( Kukulski et al . , 2012 ) , but neither Las17 nor myosin - I have been localized by these methods . Wsp1p begins at time \u2013 3 s coincident with the peak of Wsp1p local - izations and the accumulation of localizations in the membrane dis - tal zone of Arp2 / 3 complex ( Arc5p ) and actin filaments ( CHD and Fim1p ) . Localizations of both Arp2 / 3 complex and actin filaments peak at time 0 s . Both Arp2 / 3 complex ( Figure 3B ) and actin ( Figure 2 , H and I ) form two spatially and temporally separated networks of about the same size . Judging from their sudden increase in mobility , patches pinch off from the membrane with most of the actin network at time 0 s ( Berro and Pollard , 2014b , a ) . After time 0 , patches diffuse in the cytoplasm at increasing rates as they disassemble their actin filament coat over 6 s ( Berro and Pollard , 2014b , a ) . Localizations of Arc5p , CHD , and Fim1p decline more rapidly in the membrane proximal zone than the membrane distal region . These proteins may dissociate from the patch or persist after they are localized . The sudden collapse of the membrane tubule after vesicle scission might be a factor in dispers - ing the actin . How much does endocytosis in fission and budding yeast have in common ? In spite of differences in a few observations , most evidence sup - ports the conclusion that the mechanisms of endocytosis in bud - ding yeast and fission yeast are fundamentally the same . The two yeasts have a common ancestor and use most of the same pro - teins to form a membrane invagination \u223c 150 nm long surrounded by actin filaments ( Kaksonen et al . , 2005 ; Sirotkin et al . , 2010 ; Arasada and Pollard , 2011 ; Goode et al . , 2015 ; Lu et al . , 2016 ) . Actin polymerization in both yeasts depends on Arp2 / 3 complex and two types of nucleation promoting factors , myosin - I ( MYO3 and MYO5 in budding yeast and Myo1p in fission yeast ) and WASp homologues ( LAS17 in budding yeast and Wsp1p in fission yeast ) . In both the yeasts , myosin - I appears near the plasma membrane at the base of the tubular invagination ( Kaksonen et al . , 2005 ; Sun et al . , 2006 ; Galletta et al . , 2008 ; Idrissi et al . , 2008 ; Sirotkin et al . , 2010 ) and recruits Arp2 / 3 complex to nucleate branched actin fila - ments that give rise to the membrane - proximal network of fila - ments . In both yeasts , interactions of myosin - Is with adapter and coat proteins ( Sun et al . , 2006 ; Barker et al . , 2007 ; Arasada and Pollard , 2011 ) are important for efficient endocytosis . For example , recruitment of F - BAR protein Cdc15p to the endocytic site enhances FIGURE 5 : Hypothesis for the contributions of the nucleation promoting factors Myo1p and Wsp1p to endocytosis . Different stages of endocytosis with time zero defined as vesicle scission from the plasma membrane . The plasma membrane is a black line , clathrin is gray , nucleation promoting factors Wsp1p and Myo1p are blue , and actin filaments are red , green , and yellow . Clathrin is recruited 2 min prior to invagination . Nucleation promoting factors are recruited beginning at \u2013 6 s and peak prior to patch movement at \u2013 3 s . Myo1p and Wsp1p both stimulate actin polymerization at the base of the invagination near the cell surface . As the actin zones around Myo1p and Wsp1p grow , they repel each other , driving membrane invagination and redistribution of Wsp1p away from the base along the membrane invagination creating new sites of actin polymerization at the invagination tip . Myo1p remains associated with the plasma membrane at the base of the invagination . The expansion of the two zones triggers vesicle scission between the two zones of actin at time 0 s . 302 | R . Arasada et al . Molecular Biology of the Cell Models of actin assembly Understanding the forces produced by actin assembly during endo - cytosis depends on knowing the sites of assembly and the orienta - tions of the filaments . Electron microscopy of chemically fixed and freeze - substituted yeast cells ( Idrissi et al . , 2008 ) , platinum replicas of fixed and extracted animal cells ( Collins et al . , 2011 ) , and quanti - tative fluorescence microscopy of live cells ( Kaksonen et al . , 2005 ; Sirotkin et al . , 2005 , 2010 ; Berro et al . , 2010 ; Berro and Pollard , 2014b , a ; Arasada and Pollard , 2011 ; Picco et al . , 2015 ) have sug - gested three possible orientations of filaments in actin patches as follows : 1 ) actin filaments are assembled from the plasma mem - brane with barbed ends oriented toward the plasma membrane ( Kaksonen et al . , 2005 ; Picco et al . , 2015 ) ; 2 ) actin filaments are oriented with the barbed ends pointed toward the tip of the mem - brane invagination ( Collins et al . , 2011 ) ; and 3 ) very short actin filaments ( Young et al . , 2004 ; Kaksonen et al . , 2005 ; Sirotkin et al . , 2005 , 2010 ; Berro et al . , 2010 ; Berro and Pollard , 2014b , a ; Arasada and Pollard , 2011 ; Picco et al . , 2015 ) are oriented randomly in one zone ( Young et al . , 2004 ; Kaksonen et al . , 2005 ; Sirotkin et al . , 2005 , 2010 ; Berro et al . , 2010 ; Berro and Pollard , 2014b , a ; Arasada and Pollard , 2011 ; Picco et al . , 2015 ) or two zones ( Arasada and Pollard , 2011 ) . Model 3 with two zones is simpler and more appealing to us , because it does not require mechanisms to orient the filaments . Dis - tinguishing these hypotheses definitively will require methods to document the arrangement of individual filaments . MATERIALS AND METHODS Strain construction , growth conditions , and yeast methods Supplemental Table S1 lists the S . pombe strains used in this study . Fluorescent protein ( FP ) tag sequences were integrated into the genome by PCR - based gene tagging ( Bahler et al . , 1998 ) . All ge - nomic integrations were confirmed by PCR and fluorescence mi - croscopy . We used plasmid pFA6a - mEos3 . 2 - kanMX6 to tag pro - teins with mEOS3 . 2 at their endogenous genomic loci ( Laplante et al . , 2016b ) . We constructed plasmid pFA6a - mMaple3 - KanMX6 to tag proteins at their endogenous loci with mMaple3 . mMaple3 cDNA was PCR amplified and replaced mEOS3 . 2 in pFA6a - mEos2 - kanMX6 . We constructed pFA6a - kanMX6 - Pmyo1 - mMaple3 and pFA6a - kanMX6 - Pwsp1 - mMaple3 plasmids for N - terminal tagging of Myo1p and Wsp1p . We PCR amplified the coding sequence of mMaple3 ( Wang et al . , 2014 ) and replaced the mEGFP coding sequence of pFA6a - kanMX6 - Pmyo1 - mEGFP and pFA6a - kanMX6 - Pmyo1 - mEGFP using the restriction sites Pac I and Asc I ( Sirotkin et al . , 2005 ) . Actin filaments were visualized indirectly with either with Fim1p - mEOS3 . 2 or with Rng2p CHD . To express mEOS3 . 2 - CHD or GFP - CHD , cells expressing mEOS3 . 2 - CHD or GFP - CHD under the control of a 41x nmt1 promoter were grown in Edinburgh minimal media ( EMM5S ) for 18 h before imaging . Preparation of cells for imaging All S . pombe cells expressing a fluorescently tagged protein were grown in YE5S at 25\u00b0C to OD 595 0 . 2 \u2013 0 . 5 , washed in EMM5S filtered through a 0 . 45 - \u00b5m filter , and mounted on 25 % gelatin pads in fil - tered EMM5S . DIC images were collected prior to superresolution imaging . Superresolution data acquisition , data analysis , and display Superresolution imaging was performed with a custom - built two - dimensional FPALM system ( Huang et al . , 2013 ; Laplante et al . , 2016a , b ) equipped with a scientific complementary metal - oxide - semiconductor ( sCMOS ) camera ( ORCA - Flash4 . 0 ; Hamamatsu ) . Wide - field epifluorescence optics were used to illuminate cells simultaneously with a near - UV laser ( 405 nm ) to photoconvert small numbers of random fluorescent proteins and 564 - nm laser to excite the fluorophores through the entire cell . The average intensity of the 564 - nm laser used to excite photoconverted mEOS3 . 2 and mMaple3 for imaging was \u223c 1 . 2 kW / cm 2 . To main - tain an optimal density of photoconverted mEOS3 . 2 or mMaple3 for single molecule localization , the power of the 405 - nm photo - conversion laser was increased every 5 s during data acquisition with a computer - controlled ramp to compensate for irreversible photobleaching of molecules that were photoconverted and im - aged . Single - molecule images were acquired at a frame rate of 200 frames / s for 40 \u2013 60 s from the medial focal plane of a cell with an sCMOS camera ( ORCA - Flash4 . 0 ; Hamamatsu ) using HCimage software ( Hamamatsu ) . Acquired frames were analyzed using a custom sCMOS - specific localization algorithm based on a maximum likelihood estimator ( Huang et al . , 2013 ; Laplante et al . , 2016a , b ) . Localizations were plotted to construct a two - dimensional histogram FPALM image . To aid visualization of the two - dimensional histogram images , the im - ages were convolved with a two - dimensional Gaussian kernel ( \u03c3 = 7 . 5 nm ) . Time - lapse images showing dynamics were built by com - bining sequential 200 frame reconstructions ( 1 - s intervals ) and color coded for localization density . To display temporal information in a single composite image , the localizations in each 1 - s time point was assigned a Matlab Jet Color over the lifetime of the patch . We ex - tracted from the Matlab images of individual patches the numbers of localized emitters detected each second overall and from two spatial zones . Dimensions of an immobile mMaple3 - Myo1p patch that is confined the plasma membrane is used as a reference for the membrane proximal zone . Most of the mMaple3 - Myo1p localiza - tions are confined to a region 250 nm wide and 200 nm deep from the cell edge . Localizations that were detected within a 250 - nm - wide and 200 - nm - deep box from the cell edge were classified as membrane proximal zone , and any localizations detected outside the region are defined to be in the membrane distal region . The membrane distal region is limited to 250 nm wide and 350 nm deep starting 200 nm away from the cell edge . Images acquired in each data set ( typically 38 \u2013 40 s at 200 frames / s ) were sum projected , and the boundary of the localizations was identified as the cell edge . The raw time courses were aligned to the time course of a reference patch using the temporal superresolution realignment method ( Berro and Pollard , 2014a ) and averaged to obtain the average time course of localized emitters in the membrane proximal and mem - brane distal regions . The time courses were aligned to the vesicle scission event at time 0 s . To align the data , we first obtained the time courses of the total number of localizations detected in the actin patch ( sum localizations in the membrane proximal region and in the membrane distal regions ) . We then aligned the peak number of localizations detected to the peak number of GFP / YFP molecules assembled in an actin patch reported previously ( Sirotkin et al . , 2010 ; Arasada and Pollard , 2011 ) . To measure the dimensions of the actin patch , a 100 \u00d7 100 pixel region with an actin patch was selected , and the coordinates of the localizations for the patch were calculated from the 100 \u00d7 100 two - dimensional matrix using Matlab . The patches in all of the figures were oriented with the cell surface at the top with the cell interior below . DIC images acquired prior to FPALM imaging were used to align patches in this coordinate system . To calculate the width of the patch , localization densities in each column of the two - dimensional matrix were summed and projected onto an x - axis . The width of the patch represents the width of the localization density distribution at half the maximum . To calculate the length of the patch , localization Volume 29 February 1 , 2018 Organization of S . pombe a ctin patches | 303 Goode BL , Eskin JA , Wendland B ( 2015 ) . Actin and endocytosis in budding yeast . Genetics 199 , 315 \u2013 358 . Huang F , Hartwich TM , Rivera - Molina FE , Lin Y , Duim WC , Long JJ , Uchil PD , Myers JR , Baird MA , Mothes W , et al . ( 2013 ) . Video - rate nanoscopy using sCMOS camera - specific single - molecule localization algorithms . Nat Methods 10 , 653 \u2013 658 . Idrissi FZ , Grotsch H , Fernandez - Golbano IM , Presciatto - Baschong C , Riezman H , Geli MI ( 2008 ) . Distinct acto / myosin - I structures associate with endocytic profiles at the plasma membrane . J Cell Biol 180 , 1219 \u2013 1232 . Kaksonen M , Toret CP , Drubin DG ( 2005 ) . A modular design for the clathrin - and actin - mediated endocytosis machinery . Cell 123 , 305 \u2013 320 . Kukulski W , Schorb M , Kaksonen M , Briggs JA ( 2012 ) . Plasma membrane reshaping during endocytosis is revealed by time - resolved electron tomography . Cell 150 , 508 \u2013 520 . Laplante C , Huang F , Bewersdorf J , Pollard TD ( 2016a ) . High - speed super - resolution imaging of live fission yeast cells . Methods Mol Biol 1369 , 45 \u2013 57 . Laplante C , Huang F , Tebbs IR , Bewersdorf J , Pollard TD ( 2016b ) . Molecular organization of cytokinesis nodes and contractile rings by super - reso - lution fluorescence microscopy of live fission yeast . Proc Natl Acad Sci USA 113 , E5876 \u2013 E5885 . Lee WL , Bezanilla M , Pollard TD ( 2000 ) . Fission yeast myosin - I , Myo1p , stimulates actin assembly by Arp2 / 3 complex and shares functions with WASp . J Cell Biol 151 , 789 \u2013 800 . Lu R , Drubin DG , Sun Y ( 2016 ) . Clathrin - mediated endocytosis in budding yeast at a glance . J Cell Sci 129 , 1531 \u2013 1536 . Minc N , Boudaoud A , Chang F ( 2009 ) . Mechanical forces of fission yeast growth . Curr Biol 19 , 1096 \u2013 1101 . Padrick SB , Cheng HC , Ismail AM , Panchal SC , Doolittle LK , Kim S , Skehan BM , Umetani J , Brautigam CA , Leong JM , Rosen MK ( 2008 ) . Hierarchical regulation of WASP / WAVE proteins . Mol Cell 32 , 426 \u2013 438 . Picco A , Mund M , Ries J , Nedelec F , Kaksonen M ( 2015 ) . Visualizing the functional architecture of the endocytic machinery . Elife 4 , doi : 10 . 7554 / eLife . 04535 . Rajmohan R , Wong MH , Meng L , Munn AL , Thanabalu T ( 2009 ) . Las17p - Vrp1p but not Las17p - Arp2 / 3 interaction is important for actin patch polarization in yeast . Biochim Biophys Acta 1793 , 825 \u2013 835 . Schaber J , Adrover MA , Eriksson E , Pelet S , Petelenz - Kurdziel E , Klein D , Posas F , Goksor M , Peter M , Hohmann S , Klipp E ( 2010 ) . Biophysical properties of Saccharomyces cerevisiae and their relationship with HOG pathway activation . Eur Biophys J 39 , 1547 \u2013 1556 . Sirotkin V , Beltzner CC , Marchand JB , Pollard TD ( 2005 ) . Interactions of WASp , myosin - I , and verprolin with Arp2 / 3 complex during actin patch assembly in fission yeast . J Cell Biol 170 , 637 \u2013 648 . Sirotkin V , Berro J , Macmillan K , Zhao L , Pollard TD ( 2010 ) . Quantitative analysis of the mechanism of endocytic actin patch assembly and disas - sembly in fission yeast . Mol Biol Cell 21 , 2894 \u2013 2904 . Skruzny M , Brach T , Ciuffa R , Rybina S , Wachsmuth M , Kaksonen M ( 2012 ) . Molecular basis for coupling the plasma membrane to the actin cyto - skeleton during clathrin - mediated endocytosis . Proc Natl Acad Sci USA 109 , E2533 \u2013 E2542 . Sun Y , Leong NT , Jiang T , Tangara A , Darzacq X , Drubin DG ( 2017 ) . Switch - like Arp2 / 3 activation upon WASP and WIP recruitment to an apparent threshold level by multivalent linker proteins in vivo . Elife 6 , e29140 . Sun Y , Martin AC , Drubin DG ( 2006 ) . Endocytic internalization in budding yeast requires coordinated actin nucleation and myosin motor activity . Dev Cell 11 , 33 \u2013 46 . Wang S , Moffitt JR , Dempsey GT , Xie XS , Zhuang X ( 2014 ) . Characterization and development of photoactivatable fluorescent proteins for single - molecule - based superresolution imaging . Proc Natl Acad Sci USA 111 , 8452 \u2013 8457 . Young ME , Cooper JA , Bridgman PC ( 2004 ) . Yeast actin patches are net - works of branched actin filaments . J Cell Biol 166 , 629 \u2013 635 . ACKNOWLEDGMENTS Research reported in this publication was supported by the National Institute of General Medical Sciences of the National Institutes of Health under award numbers R01GM026132 and R01GM026338 to T . D . P . and grant R01GM115636 to J . B . The content is solely the re - sponsibility of the authors and does not necessarily represent the official views of the National Institutes of Health . We thank Steven Wang for providing a plasmid with mMaple3 and especially Jeorg Bewersdorf of Yale University for the use of his superresolution microscope . REFERENCES Aghamohammadzadeh S , Ayscough KR ( 2009 ) . Differential requirements for actin during yeast and mammalian endocytosis . Nat Cell Biol 11 , 1039 \u2013 1042 . Arasada R , Pollard TD ( 2011 ) . Distinct roles for F - BAR proteins Cdc15p and Bzz1p in actin polymerization at sites of endocytosis in fission yeast . Curr Biol 21 , 1450 \u2013 1459 . Attanapola SL , Alexander CJ , Mulvihill DP ( 2009 ) . Ste20 - kinase - dependent TEDS - site phosphorylation modulates the dynamic localisation and endocytic function of the fission yeast class I myosin , Myo1 . J Cell Sci 122 , 3856 \u2013 3861 . Bahler J , Wu JQ , Longtine MS , Shah NG , McKenzie A 3rd , Steever AB , Wach A , Philippsen P , Pringle JR ( 1998 ) . Heterologous modules for ef - ficient and versatile PCR - based gene targeting in Schizosaccharomyces pombe . Yeast 14 , 943 \u2013 951 . Barker SL , Lee L , Pierce BD , Maldonado - Baez L , Drubin DG , Wendland B ( 2007 ) . Interaction of the endocytic scaffold protein Pan1 with the type I myosins contributes to the late stages of endocytosis . Mol Biol Cell 18 , 2893 \u2013 2903 . Basu R , Munteanu EL , Chang F ( 2014 ) . Role of turgor pressure in endocyto - sis in fission yeast . Mol Biol Cell 25 , 679 \u2013 687 . Berro J , Pollard TD ( 2014a ) . Local and global analysis of endocytic patch dynamics in fission yeast using a new \u201ctemporal superresolution\u201d realignment method . Mol Biol Cell 25 , 3501 \u2013 3514 . Berro J , Pollard TD ( 2014b ) . Synergies between Aip1p and capping protein subunits ( Acp1p and Acp2p ) in clathrin - mediated endocytosis and cell polarization in fission yeast . Mol Biol Cell 25 , 3515 \u2013 3527 . Berro J , Sirotkin V , Pollard TD ( 2010 ) . Mathematical modeling of endocytic actin patch kinetics in fission yeast : disassembly requires release of actin filament fragments . Mol Biol Cell 21 , 2905 \u2013 2915 . Buser C , Drubin DG ( 2013 ) . Ultrastructural imaging of endocytic sites in Saccharomyces cerevisiae by transmission electron microscopy and im - munolabeling . Microsc Microanal 19 , 381 \u2013 392 . Collins A , Warrington A , Taylor KA , Svitkina T ( 2011 ) . Structural organiza - tion of the actin cytoskeleton at sites of clathrin - mediated endocytosis . Curr Biol 21 , 1167 \u2013 1175 . Galletta BJ , Chuang DY , Cooper JA ( 2008 ) . Distinct roles for Arp2 / 3 regula - tors in actin assembly and endocytosis . PLoS Biol 6 , e1 . densities in each row of the two - dimensional matrix were summed and projected onto a y - axis . The length of the patch is the distance that includes 90 % of the localizations from the cell edge . To align data from multiple patches , we defined the cell edge as the position with 10 % the peak number of localizations . The length distributions of the aligned patches were averaged .",
    "francis2015endocytic": "RESEARCH ARTICLE Endocytic membrane turnover at the leading edge is driven by a transient interaction between Cdc42 and GRAF1 Monika K . Francis 1 , 2 , Mikkel R . Holst 1 , Maite Vidal - Quadras 1 , Sara Henriksson 2 , 3 , Rachel Santarella - Mellwig 4 , Linda Sandblad 3 and Richard Lundmark 1 , 2 , * ABSTRACT Changes in cell morphology require coordination of plasma membrane turnover and cytoskeleton dynamics , processes that are regulated by Rho GTPases . Here , we describe how a direct interaction between the Rho GTPase Cdc42 and the GTPase - activating protein ( GAP ) GRAF1 ( also known as ARHGAP26 ) , facilitates rapid cell surface turnover at the leading edge . Both Cdc42 and GRAF1 were required for fluid - phase uptake and regulated the generation of transient GRAF1 - coated endocytic carriers , which were distinct from clathrin - coated vesicles . GRAF1 was found to transiently assemble at discrete Cdc42 - enriched punctae at the plasma membrane , resulting in a corresponding decrease in the microdomain association of Cdc42 . However , Cdc42 captured in its active state was , through a GAP - domain - mediated interaction , localised together with GRAF1 on accumulated internal structures derived from the cell surface . Correlative fluorescence and electron tomography microscopy revealed that these structures were clusters of small membrane carriers with defective endosomal processing . We conclude that a transient interaction between Cdc42 and GRAF1 drives endocytic turnover and controls the transition essential for endosomal maturation of plasma membrane internalised by this mechanism . KEY WORDS : Clathrin - independent endocytosis , GRAF1 , Cdc42 , Cell surface , Actin INTRODUCTION Cell surface dynamics are fundamental to a variety of basic biological processes , such as migration , polarisation and division , as well as protein and lipid trafficking . Intimate coupling to an underlying meshwork of cortical actin keeps the plasma membrane under high tension , providing the cell surface with its vital strength to withstand mechanical stress ( Gauthier et al . , 2012 ) . Consequently , structural alterations of the cell surface can only be achieved through the intricate coordination of membrane remodelling events and cytoskeletal rearrangements ( de Curtis and Meldolesi , 2012 ) . Small G - proteins of the Rho GTPase family are known as master regulators of the cytoskeleton and have been shown to greatly influence membrane tension and plasma membrane turnover by regulating endocytic and exocytic events ( Jaffe and Hall , 2005 ; Ridley , 2001 ) . Rho GTPases are peripherally attached to the membrane through lipid modifications and are active in their GTP - loaded state , where they interact with various effector molecules and function as molecular switches through rounds of GTP hydrolysis . The activity , which is stimulated by external cues , is strictly regulated by GTPase - activating proteins ( GAPs ) and guanine - nucleotide - exchange factors ( GEFs ) , which promote GTP hydrolysis and facilitate loading of GTP , respectively ( Jaffe and Hall , 2005 ) . However , the intricate mechanisms that allow for a cyclical nucleotide exchange in small G - proteins to synchronise membrane and cytoskeletal dynamics are still not understood . The clathrin - independent carrier ( CLIC ) pathway , a major pinocytic endocytic route in fibroblasts , facilitates polarised and rapid uptake of fluid , bacterial toxins , glycosylphosphatidylinositol ( GPI ) - linked receptors and receptors involved in cell adhesion ( Howes et al . , 2010a ) . CLICs were originally defined by their dependence on the Rho GTPase Cdc42 and tubular morphology , separating them from clathrin - coated vesicles and caveolae ( Hansen and Nichols , 2009 ; Howes et al . , 2010b ; Mayor and Pagano , 2007 ) . Carriers derived from this pathway have been shown to fuse into GPI - enriched endosomal compartments ( GEECs ) , which subsequently merge with early endosomes or recycling endosomes ( Kalia et al . , 2006 ) . The formation of CLICs depends on a poorly characterised molecular system , involving the small G - proteins Cdc42 and Arf1 , and their regulatory proteins ( Chadda et al . , 2007 ; Kumari and Mayor , 2008 ; Lundmark et al . , 2008 ; Sabharanjak et al . , 2002 ) . Although CLICs account for a major portion of the endocytic turnover of the plasma membrane , very little mechanistic detail is available regarding their formation . It has previously been shown that GRAF1 ( also known as ARHGAP26 ) , a GAP active against Cdc42 ( Hildebrand et al . , 1996 ; Jelen et al . , 2009 ; Longenecker et al . , 2000 ) , is essential for CLIC uptake ( Lundmark et al . , 2008 ) . GRAF1 is a dimeric multidomain protein composed of Bin , amphiphysin , RVS161 / 167 ( BAR ) , pleckstrin homology ( PH ) , GAP and Src homology 3 ( SH3 ) domains . The BAR and PH domains can generate and / or stabilise highly curved endocytic membranes from phosphatidylinositol 4 , 5 - bisphosphate ( PIP 2 ) - enriched regions , whereas the C - terminal GAP and SH3 domains modulates Cdc42 activity and interacts with regulatory proteins , respectively ( Doherty et al . , 2011a ; Hildebrand et al . , 1996 ; Lundmark et al . , 2008 ; Okada et al . , 2011 ) . Loss of GRAF1 downregulates endocytosis and affects cellular processes such as cortical actin remodelling , integrin trafficking , adhesion , spreading , migration and fusion , suggesting that this protein is an important regulator of cell surface dynamics ( Doherty et al . , 2011a , b ; Lundmark et al . , 2008 ; Nonnenmacher and Weber , 2011 ; Simpson et al . , 2008 ; Stergiou et al . , 2013 ) . In this work , we studied the interplay between GRAF1 and Cdc42 to decipher its importance for coordination of membrane and Received 13 May 2015 ; Accepted 28 September 2015 1 Integrative Medical Biology , Ume\u00e5 University , Ume\u00e5 901 87 , Sweden . 2 Medical Biochemistryand Biophysics , Ume\u00e5 University , Ume\u00e5 901 87 , Sweden . 3 Molecular Biology , Ume\u00e5 University , Ume\u00e5 901 87 , Sweden . 4 European Molecular Biology Laboratory , Meyerhofstr . 1 , Heidelberg 69117 , Germany . * Author for correspondence ( richard . lundmark @ umu . se ) This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( http : / / creativecommons . org / licenses / by / 3 . 0 ) , which permits unrestricted use , distributionandreproductioninanymediumprovidedthattheoriginalworkisproperlyattributed . 4183 \u00a9 2015 . Published by The Company of Biologists Ltd | Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e actin dynamics during cell surface turnover at the leading edge . We characterise GRAF1 as a molecular marker of CLICs and demonstrate how this protein , through a direct interaction , regulates Cdc42 activity during endocytosis . This study also defines a temporal and spatial restriction of active Cdc42 as a prerequisite for the internalisation and further trafficking of cell surface components . RESULTS Correlative light and electron microscopy reveals highly curved membrane carriers decorated by GRAF1 and GTPase - deficient Cdc42 Consistent with the proposed involvement of the GTPase Cdc42 and the GAP GRAF1 in the generation of CLICs , GRAF1 localised with a dominant - active form of Cdc42 ( Cdc42 - Q61L , which is deficient in GTP hydrolysis ) to discrete assemblies in fixed cells ( Fig . 1A ) . As a step towards determining their identity , correlative light and electron tomography microscopy ( CLEM ) was used to resolve the ultrastructure of these assemblies . High - pressure frozen cells transfected with GFP \u2013 GRAF1 and mCherry \u2013 Cdc42 - Q61L were further prepared for CLEM as described in the Materials and Methods ( Kukulski et al . , 2011 ; Nixon et al . , 2009 ) . By overlaying the captured epifluorescence micrograph with high - resolution electron tomograms acquired from the same sample and region of interest ( ROI ) , it was apparent that all the fluorescent spots containing both proteins overlapped with small membranous structures ( Fig . 1B ) . Cdc42 - Q61L fluorescence alone , commonly seen lining the cell edge , was instead associated with the plasma membrane . Three - dimensional ( 3D ) models of the membranous structures , reconstructed from the tomograms , revealed that the GRAF1 - and Cdc42 - Q61L - positive structures detected by epifluorescence microscopy actually corresponded to tightly packed clusters of membrane carriers ( Fig . 1B ) . Most of the carriers were separated from the plasma membrane , but examples of tubular structures connected to the cell surface were also found ( Fig . 1B ) . Although they were quite morphologically diverse , the modelled carriers could roughly be divided into being either disc - like or tube - like on the basis of their shape ( Fig . 1C ; compare to the GRAF1 - and Cdc42 - Q61L - negative clathrin - coated vesicle ) . The measurement of the minor axis of each structure further indicated a surprisingly conserved diameter , spanning 20 \u2013 60 nm in 97 % of the carriers ( Fig . 1D ) . The close proximity of the GRAF1 - and Cdc42 - Q61L - positive carriers to the basal cell surface , evident using confocal microscopy ( Fig . 1A ) , was further confirmed by the reconstructed 3D models . The carrier clusters were detected within 500 nm of the basal membrane , and their distance to the cell edge , as measured in the x - y plane , ranged between 360 and 2200 nm . Taken together , the information collected from light and electron microscopy shows that GRAF1 and GTPase - deficient Cdc42 localise to carriers formed near the cell surface that have a morphology consistent with the membrane compartments previously defined as CLICs . GRAF1 is a marker of CLICs that are rapidly forming at the leading edge of cells To enable the dissection of a potential molecular interplay between Cdc42 and GRAF1 during CLIC formation , a cell model system was established utilising GRAF1 as a pathway marker . In a Flp - In T - REx HeLa cell line , untagged or GFP - tagged GRAF1 was placed under the control of a tet promoter modulated by doxycycline , resulting in an inducible system for the expression of a controlled amount of protein ( Fig . 2A ) . The protein expression level adapted for all subsequent experiments was chosen as to enable the detection of fluorescently tagged GRAF1 by microscopy , inevitably resulting in protein levels above the comparatively low endogenous level in this specific HeLa cell line . However , in contrast to using transient overexpression of GRAF1 , the developed Flp - In T - REx cell lines importantly allowed more reproducible and homogenous protein levels within the cultures ( Fig . S1A ) . In fixed cells , confocal microscopy showed that GRAF1 , independent of the presence of a fluorescent tag , assembled into punctate and , much less frequently , tubular structures at the cell surface ( Fig . S1B ) . In agreement with previous results detecting endogenous and transiently expressed GRAF1 in HeLa cells ( Lundmark et al . , 2008 ) , these assemblies colocalised with a fraction of the promiscuous cargo cholera toxin B - subunit ( CTxB ) , as assayed after 2 min internalisation at 37\u00b0C ( Fig . S2A ) . Moreover , these assemblies were only on very rare occasions found to overlap with clathrin - associated markers like transferrin ( Tfn ) and AP - 2 ( the adaptin for clathrin - mediated endocytosis ) , or the early endosomal marker EEA1 ( Fig . S2A ) . Analysis by live - cell spinning disc confocal microscopy revealed the formation of abundant short - lived GRAF1 \u2013 GFP - positive punctae at the leading edge of cells ( Fig . 2B ; Movie 1 ) . Software was used to create tracks corresponding to each detected structure , and from these the duration time of each track was derived . The average lifetime ( track duration ) of the dynamic GRAF1 assemblies was thereby determined to be \u223c 10 s ( Fig . 2C ) , showing that the protein was transiently assembled at the leading edge . This is in agreement with GRAF1 being associated with CLICs , which are dynamically formed at the cell periphery during cell spreading and migration ( Doherty et al . , 2011a ; Howes et al . , 2010a ) . To determine whether GRAF1 assembled at the plasma membrane , cells were followed by live - cell total internal reflection fluorescence ( TIRF ) microscopy . Structure tracks were created from the acquisitions and the duration of each track was calculated . The lifetimes of GRAF1 \u2013 GFP and GFP \u2013 GRAF1 assemblies were comparable to that detected by spinning disc confocal microscopy , suggesting that the burst in GRAF1 fluorescence corresponded to activity very near or at the cell surface ( Fig . 2D ; Fig . S1C ) . In agreement with this , detecting the surface binding and uptake of fluorescently tagged CTxB by performing live - cell TIRF microscopy revealed the recruitment of GFP \u2013 GRAF1 to CTxB clusters and the subsequent disappearance of both markers from the surface , indicative of endocytic carrier formation ( Fig . S2B ) . Importantly , in GRAF1 \u2013 GFP cells co - expressing mCherry - tagged clathrin light chain , live - cell TIRF microscopy followed by track analysis showed no overlap between the two proteins ( Fig . 2E ; Movie 2 ) , implying that GRAF1 does not aid the process of clathrin - mediated endocytosis . It should be noted that GRAF1 assemblies were not exclusively observed at the leading edge of cells and that an increased fraction of the assemblies in the centre and rear of cells displayed duration times longer than 30 s ( Fig . S1D , E ) . LossofCdc42andactinpolymerisation impairGRAF1 carrier processing and the fluid - phase endocytic capacity Cdc42 was identified as a key regulator of CLIC - mediated endocytosis on the basis of the effects that overexpressed Cdc42 activity mutants exerted on the uptake through this pathway ( Sabharanjak et al . , 2002 ) . To test the importance of Cdc42 and GRAF1 for clathrin - independent endocytosis in our established Flp - In T - REx HeLa cell lines , fluid - phase internalisation was quantified in cells depleted of either of these two proteins , or of AP - 2 by small interfering RNA ( siRNA ) . To control for effects on clathrin - mediated endocytosis , internalisation of transferrin was 4184 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e also assayed . The fluorescence of transferrin - S - S - CW800 or dextran - S - S - CW800 was measured after MesNa - reduction to remove surface - exposed fluorophores , and the internalisation was quantified in relation to that of cells transfected with control siRNA ( Fig . 3A , B ) . Although dextran uptake was impaired in all three samples , transferrin uptake was only significantly reduced in AP - 2 - depleted cells ( Fig . 3A ) , verifying the importance of Cdc42 and GRAF1 in clathrin - independent endocytosis . The \u223c 50 % reduction of dextran uptake seen in all samples further indicated that clathrin - independent and - dependent pathways , under these conditions , contribute to roughly equal amounts of internalised fluid volume . To investigate the importance of Cdc42 function on GRAF1 - mediated carrier formation , the localisation and behaviour of GRAF1 \u2013 GFP was followed by live - cell spinning disc confocal microscopy in cells depleted of Cdc42 . Although GRAF1 in these cells could still be found in dynamic assemblies at the leading edge , there was a marked increase of GRAF1 - decorated tubular structures associated with the cell surface ( Fig . 3C ; Fig . S3A , B ) . This implies that GRAF1 in cells with a decreased available pool of Cdc42 can exert its sculpting and / or stabilising function on the plasma membrane , but in the absence of the GTPase that this activity is not properly regulated . Interestingly , active Cdc42 has been shown to be important for recruitment of the Wiskott \u2013 Aldrich syndrome protein ( WASP ) - dependent actin machinery to aid CLIC formation ( Chadda et al . , 2007 ) . Similar to the effect seen after Cdc42 depletion , cells treated with the WASP - inhibitor wiskostatin showed a time - dependent increase of GFP \u2013 GRAF1 - decorated tubes associated with the cell surface ( Fig . 3D , E ) . The same phenotype was also observed after treatment with the actin - depolymerising drugs latrunculin A and cytochalasin D ( Fig . S3C ) , previously shown to inhibit uptake through CLICs ( Chadda et al . , 2007 ) . The altered processing of GRAF1 - mediated carriers in Cdc42 - depleted cells is likely to be at least a contributing factor to their impaired endocytic capacity , and a result of dysregulated Cdc42 - dependent actin dynamics . Fig . 1 . GRAF1 and GTPase - deficient Cdc42 localise to surface - detached pleomorphic membrane compartments . ( A ) Confocal stack of a Myc \u2013 GRAF1 - and GFP \u2013 Cdc42 - Q61L - expressing HeLa cell , presented as a maximum intensity projection and a 90\u00b0 tilted 3D volume . Scale bar : 10 \u03bc m . ( B ) Correlative microscopy analysis of a GFP \u2013 GRAF1 and mCherry \u2013 Cdc42 - Q61L - expressing HeLa cell . The left panel depicts the epifluorescence and electron micrographs . White arrowheads mark the analysed structures positive for both proteins . Colour - coded 3D models derived from the reconstructed electron tomograms and magnifications of two of the analysed clusters are visualised in the right panel . Thewhite arrow points at atube connected to the cell surface . Scale bars : 100 nm . ( C ) Representative examples of the two types of membrane compartments recognised within the analysed clusters in B ( disc - like and tube - like ) and a clathrin - coated vesicle ( CCV ) . Scale bars : 50 nm . ( D ) Widest diameterofminoraxis measurementsfromthe analysed membrane compartments inB . 136 structuresfrom 11 clusters in one cell were included in the analysis . The mean\u00b1s . d . is indicated . 4185 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e GRAF1 assembly at Cdc42 - enriched microdomains antagonises the local plasma membrane association of the GTPase To study the dynamics of the interplay between Cdc42 and GRAF1 during carrier formation , mCherry - tagged wild - type Cdc42 was co - expressed in GRAF1 \u2013 GFP Flp - In T - REx HeLa cells , and the two proteins followed over time with live - cell spinning disc microscopy ( Fig . 4A ; Movie 3 ) . Remarkably , 94 % of the GRAF1 assemblies that transiently appeared at the leading edge overlapped with membrane - associated Cdc42 . Moreover , 88 % of these assemblies were recruited to existing microdomains enriched with the GTPase . Further dissection of the indicated interdependency between the two proteins was accomplished by detailed analysis of fluorescence intensity plots created for each detected structure and channel over time ( Fig . 4B ) . The lack of major lateral movements in the detected GRAF1 punctae allowed the intensity value calculations to be restricted to circular ROIs ( Fig . 4B ) . The peak range and maximum was determined for each structure and channel , as described in the Materials and Methods . An average GRAF1 assembly lifetime of 18 s was approximated on the basis of the determined peak ranges , which was slightly longer , but in the same range , as the \u223c 10 s calculated from the track durations ( Fig . 4C , compare to Fig . 2C ) . Quantification of the lifetime overlap between GRAF1 and Cdc42 assemblies averaged 99 % , consistent with the visual assessment above ( Fig . 4D ) . Interestingly , there was a clear trend of Cdc42 reaching the maximum intensity value before GRAF1 ( on average 5 s before the GRAF1 intensity maximum , and 3 s after the start of the GRAF1 peak ) ( Fig . 4E ) . Consistent with this , for 49 % of the analysed structures , it was evident that the Cdc42 intensity decreased as the GRAF1 intensity increased ( compared to 82 % when analysing the GRAF1 intensity decrease ) ( Fig . 4F ) . This suggests that GRAF1 somehow antagonises the existing cell surface enrichment of Cdc42 as it assembles on the membrane , likely by mediating its internalisation and / or transiently interacting with the GTPase to switch it off and thereby destabilising its association with the microdomain . An interaction between the GAP domain and transition state Cdc42 enforces membrane assembly of GRAF1 To investigate the impact of the Cdc42 activity state on GRAF1 membrane assembly , GFP \u2013 GRAF1 Flp - In T - REx HeLa cells co - expressing CoralHue - tagged dominant - negative ( nucleotide - binding deficient T17N ) , wild - type and dominant - active Cdc42 were analysed by confocal microscopy . In comparison to the other two samples , Cdc42 - Q61L - transfected cells showed a striking accumulation of GRAF1 in bright punctate and tubular Fig . 2 . Dynamic assembly of GRAF1 at the leading edge . ( A ) Immunoblots detecting endogenous and recombinant GRAF1 in Flp - In T - REx HeLa GFP \u2013 GRAF1 , GRAF1 and GRAF1 \u2013 GFP cell lysates after induction with the stated doxycycline ( Doxy ) concentrations for 24 h . A doxycycline concentration of 1 ng / ml was chosen for all subsequent experiments . ( B ) Fluorescence micrograph detecting GRAF1 \u2013 GFP , corresponding to the last frame of a live - cell spinning disc confocal microscopy acquisition . At the leading edge , tracks of structure movement and duration over time are illustrated as colour - coded lines . Representative frames from the same acquisition are presented as a time series . Scale bar : 10 \u03bc m . ( C ) Duration time of GRAF1 \u2013 GFP structure tracks derived from acquisitions exemplified in B . 19 cells from three independent experiments were analysed . The bar graph depicts the distribution of the mean\u00b1s . d . values derived from each evaluated cell . ( D ) Duration time of GRAF1 \u2013 GFP and GFP \u2013 GRAF1 structure tracks derived from live - cell TIRF microscopy acquisitions . Bar graphs depict the distribution of the mean\u00b1s . d . values derived from each cell included in the respective analysis . Three independent experiments were analysed ( Mann \u2013 Whitney U test , n = 11 \u2013 12 cells , \u03b1 = 0 . 05 , ns , not significant ) . ( E ) Fluorescent micrograph showing GRAF1 \u2013 GFP and mCherry - tagged clathrin light chain ( CLC ) at a protrusion , corresponding to one frame from a live - cell TIRF microscopy acquisition . Tracksvisualise thedetectedstructuremovementin the respective channel . The edge of the cell is outlined in the upper panels . Arrowheads highlight GRAF1 assemblies . Scale bar : 2 \u03bc m . 4186 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e structures ( Fig . 5A , B ) , reminiscent of the clustered carriers formed in HeLa cells transiently co - expressing the two proteins ( compare with Fig . 1A ) . Moreover , this phenotype was dependent on the membrane localisation of the GTPase because it was abolished after introducing a point mutation of the cysteine residue ( C188A ) responsible for anchoring Cdc42 to the membrane ( Fig . 5A , B ) . Notably , no accumulation of GRAF1 structures was seen in cells co - expressing dominant - active RhoA - Q63L or Rac1 - Q61L ( Fig . 5A ) , two other GTPases within the same Rho family of actin regulators , confirming the specificity of the induced phenotype . In vitro pulldown experiments were used to determine whether the Cdc42 - Q61L mutation somehow influenced the otherwise inherently transient interaction between GRAF1 and Cdc42 , to thereby so profoundly affect GRAF1 membrane assembly . Using purified GST - tagged GRAF1 GAP domains as bait and Cdc42 or Cdc42 - Q61L loaded with GTP \u03b3 S ( a more stable GTP analogue ) as prey , revealed a preference in binding to the mutant ( Fig . 5C ) . However , repeating the assay with GDP - loaded GTPase in the presence of AlF x , to mimic the transition state conformation ( Mittal et al . , 1996 ) , resulted , in addition , to detectable binding to wild - type Cdc42 ( Fig . 5C ) . The Q61L mutation in Cdc42 increased the affinity to GRAF1 , likely by imposing an active site conformation , suggesting that the accumulation of GRAF1 assemblies in cells overexpressing this mutant is the effect of an enforced prolonged direct interaction . To test this hypothesis in vivo , HeLa cells transiently co - expressing CoralHue - tagged Cdc42 - Q61L and a GFP - tagged GAP domain containing or lacking truncates of GRAF1 , were analysed by confocal microscopy . Like full - length GRAF1 , the BAR - PH - GAP truncate showed an extensive accumulation in punctate and tubular structures together with the mutated GTPase ( Fig . 5D , E ) . Importantly , this effect was not seen with GFP \u2013 BAR - PH , confirming that it was dependent on the presence of the GAP domain . To avoid potential averse consequences of overexpressing truncated proteins , a Flp - In T - REx HeLa cell line was established with an inducible expression of GFP \u2013 GRAF1 - R412D , a point mutant disrupted in the so - called arginine finger known to be vital for the GAP activity ( Jelen et al . , 2009 ) ( Fig . S3D , E ) . Co - expression of Cdc42 - Q61L in this cell line did not result in any accumulation of GRAF1 - R412D structures ( Fig . 5A ; Fig . S3G ) , verifying that the enforced GRAF1 membrane accumulation induced by GTPase - deficient Cdc42 is a consequence of a direct interaction between the two proteins . GTP hydrolysis deficiency traps Cdc42 together with GRAF1 on endocytic vesicles The confirmed interaction between GRAF1 and Cdc42 in vivo , begged the question as to whether GRAF1 - mediated inactivation of Cdc42 was important during endocytic carrier formation . To resolve how Cdc42 activity influenced the formation of endocytic carriers , the dynamics of GFP \u2013 GRAF1 assemblies in Flp - In Fig . 3 . Inhibition of Cdc42 function and actin polymerisation affects GRAF1 carrier processing . ( A , B ) Relative internalisation of aminodextran - S - S - CW800 and transferrin - S - S - CW800 after 30 min uptake in Flp - In T - REx HeLa GFP \u2013 GRAF1 cells depleted of AP - 2 , Cdc42 or GRAF1 by siRNA transfection ( A ) , as confirmed by the immunodetection of the respective protein in comparison to clathrin heavy chain ( CHC ) in the cell lysates ( B ) . The cargo uptake was normalised to the corresponding control siRNA - transfected sample and the significance in relation to this sample was determined by analyses of at least three independent experiments ( mean\u00b1s . d . ; Mann \u2013 Whitney U tests , n = 3 \u2013 4 samples , \u03b1 = 0 . 05 ; ns , not significant ; * P < 0 . 05 ) . ( C ) Fluorescent micrographs depicting GRAF1 \u2013 GFP in control cells and cells depleted of Cdc42 . The ratio of cells with GRAF1 in at least one tubular structure ( i . e . a structure with a length > 2 \u03bc m ) was quantified on the basis of three independent experiments ( mean\u00b1s . d . ; Chi - square test , n > 450 cells , \u03b1 = 0 . 05 ; * * * P < 0 . 0001 ) . ( D ) Ratio of cells with GFP \u2013 GRAF1 in tubular structures after treatment with DMSO ( Vehicle ) or wiskostatin ( Wisk ) , for the indicated time intervals . Cells from three independent experiments were included in the analysis ( mean\u00b1s . d . ; Chi - square tests , n > 300 cells , \u03b1 = 0 . 05 , * * * P < 0 . 0001 ) . ( E ) Confocal stack of a GFP \u2013 GRAF1 - expressing cell after 20 min wiskostatin treatment , with presented top and slice views positioned as indicated ( white arrows ) . Scale bar : 10 \u03bc m . 4187 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e T - REx HeLa cells co - expressing CoralHue - tagged wild - type Cdc42 , Cdc42 - T17N or Cdc42 - Q61L were analysed by live - cell TIRF microscopy . Individual GRAF1 structures detected over the whole basal membrane were tracked over time in the acquisitions ( Fig . 6A ) and used to calculate the number of assemblies , their lifetime ( track duration ) , the distance between their appearance and disappearance ( track displacement ) , and the maximum speed of the structures ( Fig . 6B \u2013 E ) . Consistent with the results from fixed cells , co - expression of Cdc42 - Q61L resulted in four times more GRAF1 assemblies than detected in the other samples ( Fig . 6B ) . Notably , no clear difference in lifetime ( track duration ) of GRAF1 assemblies was found between Cdc42 - and Cdc42 - Q61L - expressing cells ( Fig . 6C ) , showing that the GRAF1 membrane accumulation enforced by Cdc42 - Q61L was not trapping the protein at the cell surface . However , the GRAF1 structures in cells co - expressing the GTPase - deficient mutant exhibited both an increased lateral mobility ( track displacement ) and a higher maximum speed ( Fig . 6D , E ) . Although the directional movement revealed by the track pattern of GRAF1 structures in Cdc42 - Q61L co - expressing cells ( Fig . 6A ) was unlikely to represent diffusion within the membrane , it could be explained by enforced GRAF1 assembly on intracellular vesicles trafficked along cytoskeletal tracks near the cell surface that are in close enough proximity to the plasma membrane to be detected within the TIRF field . To further explore the mobility of the GRAF1 assemblies enforced by co - expression of GTPase - hydrolysis - deficient Cdc42 , live - cell spinning disc microscopy was employed to increase the depth of the fluorescence probe detection in the focal plane of interest ( i . e . a plane close to , but just above , the basal membrane ) . In GFP \u2013 GRAF1 cells co - expressing mCherry \u2013 Cdc42 - Q61L , the proteins localised together both on budding carriers and intracellular laterally mobile vesicles ( Fig . 6F ; Movie 4 ) . The appearance of the latter type of structures in the TIRF and captured confocal fields indicates that carriers formed from the cell surface , positive for both proteins , later stay or reappear in a plane near the plasma membrane . From live - cell spinning disc microscopy acquisitions of cells co - expressing either wild - type Cdc42 or Cdc42 - Q61L , the detected structures of GRAF1 and respective GTPase protein were tracked and the time range of colocalisation was calculated . The colocalisation of GRAF1 and Cdc42 was significantly more transient than that of GRAF1 and Cdc42 - Q61L ( Fig . 6G ) , suggesting that the prolonged interaction between GRAF1 and GTP - hydrolysis - deficient Cdc42 results in a longer co - existence of the proteins on internalised vesicles ( i . e . an entrapment of both proteins on budded vesicles ) . Fig . 4 . GRAF1 recruitment to Cdc42 microdomains coincides with a local decreasing plasma membrane association of the GTPase . ( A ) Time series from a live - cell spinning disc confocal microscopy acquisition , visualising a protrusion from a Flp - In T - REx HeLa GRAF1 \u2013 GFP cell co - expressing mCherry \u2013 Cdc42 . White arrows mark detected GRAF1 assemblies . ( B ) Intensity profiles of the green and red channels , calculated over time as described in the Materials and Methods , for the first - appearing GRAF1 assembly in A . The grey area highlights the peak duration for the green channel . The insert visualises the ROI used to derive the plotted fluorescence intensity curves . ( C \u2013 F ) Parameters calculated for single GRAF1 assemblies on the basis of the corresponding intensity peak recorded within the intensity profiles derived from live - cell acquisitions , as exemplified in A and B . The time points fortheintensitypeakstart , maximumand end were defined for each structure and channel as described in the Materials and Methods . 34 \u2013 36 GRAF1 structures from three cells were included in the analysis and results are mean\u00b1s . d . 4188 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e Inactivation of Cdc42 by GRAF1 is not required for fluid - phase uptake but for the intracellular maturation of CLICs The potential effects on endocytosis inferred by overexpression of Cdc42 - Q61L were furtherassayed byanalysing the uptake of dextran . Dextran \u2013 Alexa - Fluor - 555 was added to GFP \u2013 GRAF1 Flp - In T - REx HeLa cells co - expressing Myc \u2013 Cdc42 - Q61L as a 5 - min pulse , then a quick wash was used to remove the excess extracellular marker and probes were followed by live - cell confocal microscopy . Immediately after the pulse , dextran was enriched in approximately one third of the accumulated GRAF1 structures ( Fig . 7A ) , verifying that these correspond to plasma - membrane - derived and budded endocytic carriers . The same protocol was used to prepare fixed samples for epifluorecence microscopy . Dextran could also be visualised in GRAF1 - positive carriers under fixed conditions ( Fig . 7B ) , but it should be noted that permeabilisation of fixed samples was avoided , given that the detection of this marker deteriorates with detergent treatment . Using software to define dextran compartments from the captured images , it was determined that the total uptake of fluid - phase was not significantly altered in cells affected by the Cdc42 - Q61L transfection ( Fig . 7C ) . Also in line with this , GFP \u2013 GRAF1 - R412D cells showed a higher prevalence of membrane assemblies in comparison to GFP \u2013 GRAF1 cells ( Fig . S3H ) , without any detectable effect on dextran endocytosis ( Fig . 7E ; Fig . S3F ) , further verifying that the GRAF1 GAP activity against Cdc42 was not required for endocytic carrier formation . The entrapment of GRAF1 and Cdc42 - Q61L on erratically behaving internalised vesicles , pointed towards a post - fission trafficking defect . Strikingly , in GRAF1 and Cdc42 - Q61L co - Fig . 5 . Interaction with transition state Cdc42 enforces membrane assembly of GRAF1 . ( A ) Ratio of Flp - In T - REx HeLa cells showing a phenotype of abundant ( \u2265 15 ) GFP \u2013 GRAF1 or GFP \u2013 GRAF1 - R412D structures in the absence ( \u2212 ) or presence of co - expressed CoralHue - tagged small GTPases and mutants thereof . For statistical analysis , cells from at least three independent experiments were included . The results were compared to the GRAF1 samples expressingwild - typeCdc42 , orinthecaseofGRAF1 - R412Dcells , wererelatedtothecorrespondingGRAF1samples ( mean\u00b1s . d . ; Chi - squaretests , n \u2265 300cells , \u03b1 = 0 . 05 ; ns , not significant ; * * * P < 0 . 0001 ) . ( B ) Maximum intensity projections of confocal stacks detecting GFP \u2013 GRAF1 in cells co - expressing the indicated CoralHue - tagged Cdc42 proteins . ( C ) Coomassie - stained SDS - PAGE gel from a pulldown experiment using purified GST ( GST ) or a GST - tagged GRAF1 GAP domain ( GAP ) as bait , and purified Cdc42 or Cdc42 - Q61L loaded with GTP \u03b3 S or GDP in the presence of AlF x ( GDP / AlF x ) as prey . The bound ( B ) and unbound GTPases ( U ) weredetected . ( D ) Maximumintensity projectionsofconfocalstacksdetectingGFP \u2013 GRAF1orindicatedtruncates ( theBAR - PH - GAPandBAR - PH domains ) and CoralHue \u2013 Cdc42 - Q61L in transfected HeLa cells . ( E ) Percentage volume of GFP \u2013 GRAF1 proteins colocalised with CoralHue \u2013 Cdc42 - Q61L , as visualised in D . Three independent experiments were analysed ( mean\u00b1s . d . ; Kruskal - Wallis test , Dunn \u2019 s post test , n = 15 cells , \u03b1 = 0 . 05 ; ns , not significant ; * * * P > 0 . 0001 ) . Scale bars : 10 \u03bc m . 4189 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e expressing cells with accumulated GRAF1 assemblies , significantly more , but smaller , dextran - positive compartments were detected ( Fig . 7D ) . This indicated that these internal carriers were stalled in their endosomal maturation or that early endosomes were fragmented under these conditions . However , there was no significant effect on the number of early endosomes or the total intensity of the early Fig . 6 . Cdc42 GTPase deficiency affects GRAF1 carrier dynamics after their formation . ( A ) Fluorescence micrographs depicting the start frames from live - cell TIRF microscopy acquisitions of GFP \u2013 GRAF1 structures in Flp - In T - REx HeLa cells transfected with the indicated CoralHue - tagged Cdc42 constructs ( upper panels ) . The corresponding structure tracks over time , derived from the acquisitions , are represented as colour - coded lines ( lower panels ) . Structures detected over the entire basal cell surface were included in the track analysis . ( B \u2013 E ) Quantification of the number and dynamic parameters of GFP \u2013 GRAF1 structure tracks from acquisitions exemplified in A . Three independent experiments were analysed [ Student \u2019 s t - tests on log 10 - transformed data , n \u2265 136 structure tracks ( fromfourcells ) , \u03b1 = 0 . 05 ; ns , notsignificant ; * * * P > 0 . 0001 ] . BargraphsinC \u2013 Edepictthedistributionofthemean\u00b1s . d . valuesderivedforeachcellincludedin the respective analysis . ( F ) Fluorescence micrograph showing the first frame from a live - cell spinning disc confocal microscopy acquisition of GFP \u2013 GRAF1 structures in cells transfected with the indicated mCherry - tagged Cdc42 - Q61L . Speeds measured over the detected GRAF1 tracks are represented as colour - codedlines . Magnificationsofareas1 and2arevisualised asatime seriestohighlight themobile membranestructurespositiveforbothproteins . Thewhitearrow marks a structure at the cell surface , the yellow arrow follows the lateral movement of an internal structure and the red arrow points out the fission of a tubular structure from the cell surface . ( G ) Quantification of duration of colocalisation between GFP \u2013 GRAF1 membrane assemblies and respective indicated mCherry - tagged Cdc42 protein , based upon track analyses performed on live - cell spinning disc confocal microscopy acquisitions like that exemplified in F . Cells from two independent experiments were analysed and the mean\u00b1s . d . is indicated [ Student \u2019 s t - test on log 10 - transformed data , n \u2265 29 structure tracks ( from three cells ) , \u03b1 = 0 . 05 ; * * * P > 0 . 0001 ) . Scale bars : 10 \u03bc m . 4190 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e endosomal marker EEA1 ( Fig . 7F ) . Furthermore , carriers decorated by GRAF1 and Cdc42 - Q61L were never positive for EEA1 ( Fig . 7G ) , suggesting that these more likely represented a trapped intermediate membrane compartment . To explore this in more detail , dextran trafficking was captured in GFP \u2013 GRAF1 cells co - expressing BFP \u2013 Rab5 alone , or together with Myc \u2013 Cdc42 - Q61L , by live - cell spinningdiscmicroscopy . Cells were pulsedwith the cargofor 5 min , quickly washed and the dynamics of the probes were monitored over 5 min . At the start of the acquisition , there was a significantly lower ratio of dextran - filled Rab5 vesicles in cells affected by the co - expression of GTPase - deficient Cdc42 ( Fig . 7H ; Fig . S4A ) , showing that trafficking of internalised dextran to this early endosomal compartment was negatively affected . It should also be noted that therewasa highnumberofdextran - containing vesiclesthat , at the end of the acquisition , were still devoid of Rab5 , even in cells not co - expressing Cdc42 - Q61L , likely corresponding to alternative trafficking routes . Although fusion events could be distinguished between dextran - filled Rab5 compartments ( Fig . S4B ) , and between Fig . 7 . See next page for legend . 4191 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e separate GRAF1 compartments in cells co - expressing Cdc42 - Q61L ( Fig . 7I ) , no example of fusion between GRAF1 and Rab5 compartments was found despite frequent transient engagements observed between trapped GRAF1 compartments and labelled early endosomes ( Fig . 7I ) . As reflected by the Cdc42 - Q61L - enforced clustering of GRAF1 carriers resolved by CLEM , taken together , our study shows that GRAF1 - mediated inactivation of Cdc42 is not vital for the formation and budding of endocytic carriers , but for their further intracellular maturation . DISCUSSION Cell spreading and migration involve dynamic alterations in cell surface turnover supporting local membrane rigidity and flexibility through directed trafficking of lipids and surface receptors . This is reflected in the polarised activities of endocytic pathways including caveolae concentrated at the lagging edge , and fast endophilin - mediated endocytosis ( FEME ) and CLICs enriched at the leading edge ( Boucrot et al . , 2015 ; Chaudhary et al . , 2014 ) . Here , we show that direct interplay between GRAF1 and Cdc42 coordinates maturation of CLICs at the cell front and hence constitutes an efficient machinery for aiding in the turnover of membrane constituents . We found that GRAF1 dynamically assembled at the plasma membrane in discrete areas that did not overlap with the formation of clathrin - coated pits . The assemblies of GRAF1 were short lived and preceded by local membrane enrichment of Cdc42 , suggesting that GRAF1 is transiently recruited to membrane microdomains imposed by upstream Cdc42 activity . The concentration of GRAF1 will increase the local curvature of the membrane through the BAR domain , promoting membrane invagination . At the same time , this results in high spatially confined GAP activity to stimulate GTP hydrolysis by Cdc42 . The inactivation of Cdc42 has been shown to reduce the membrane domain association and thereby results in a more dynamic pool of Cdc42 ( Bendez\u00fa et al . , 2015 ; Chadda et al . , 2007 ) . Indeed , the progressive assembly of GRAF1 correlated with a decrease in local Cdc42 enrichment . Given that we could not detect Cdc42 on internal carriers together with GRAF1 , we suggest that the transient interaction between the two proteins results in inactivation and a subsequent decrease in the membrane association of Cdc42 . This spatial restriction of Cdc42 activity could promote the identity transition from plasma membrane to an internal membrane compartment and facilitate further maturation . Based on this , we propose that the key role of GRAF1 in regulating turnover at the leading edge explains the previously described defect in the fusion of myoblasts and impaired migratory behaviour of cells lacking GRAF1 ( Doherty et al . , 2011a , b ; Lenhart et al . , 2014 ) . The initiation of the budding process and activation of Cdc42 might be facilitated by GEFs in response to extracellular signals , cellular polarisation or alterations in membrane tension . There are likely to be additional factors that are involved in the initiation and progression of CLIC formation . Previously , Arf1 has been shown to be essential for CLIC uptake , and the activation of Arf1 at the cell surface has been proposed to recruit ARHGAP10 to modulate Cdc42 activity ( Kumari and Mayor , 2008 ) . It is not yet known whether GRAF1 is directly coupled to these components or whether they regulate similar or distinct subtypes of CLICs . Actin polymerisation appears to play a central role in CLIC formation . Local actin - driven organisation of the cell surface has been suggested to promote endocytosis through clustering of lipids and proteins ( Gowrishankar et al . , 2012 ; Romer et al . , 2010 ) . In addition , actin polymerisation has been proposed to facilitate the release of carriers from the surface in a process controlled by Cdc42 ( Chadda et al . , 2007 ) . Interestingly , we found that Cdc42 and the downstream WASP - mediated actin polymerisation were vital for processing of GRAF1 - positive carriers . Depletion of Cdc42 or the acute inhibition of actin polymerisation resulted in extended GRAF1 - decorated tubules associated with the cell surface . This shows that Cdc42 is not essential for GRAF1 membrane assembly , but suggests that Cdc42 - driven actin polymerisation plays a central role for processing of membrane invaginations generated by this mechanism . The effect on carriers could be due to impaired scission from the cell surface or reduced support from local cortical actin that relieves membrane tension and thereby promotes tubulation of the cell surface . Interestingly , it has recently been shown that perturbation of the cortical actin by inhibition or depletion of Cdc42 results in reduced membrane tension ( Bretou et al . , 2014 ) . We show that the depletion of either Cdc42 or GRAF1 results in a striking reduction in fluid uptake , consistent with that the cycling activities of Cdc42 substantially regulate the dynamics and turnover of the plasma membrane ( Kumari et al . , 2010 ) . However , the precise influence and effects of GTP - binding and hydrolysis has so far not been understood . Here , we show that the inactivation of Cdc42 was not required for the generation or the internalisation of GRAF1 - Fig . 7 . Cdc42 GTPase deficiency disrupts the maturation of internalised GRAF1carriers . ( A ) Fluorescencemicrographrepresentingthefirstframeofa live - cell confocal acquisition detecting dextran \u2013 Alexa - Fluor - 555 and GFP \u2013 GRAF1 in Flp - In T - REx HeLa cells co - expressing Myc \u2013 Cdc42 - Q61L ( upper panels ) . The acquisition was started after a 5 - min incubation with the cargo followed by medium exchange . Analysis on the basis of four captured cells revealed dextran enrichment in 31\u00b116 % ( mean\u00b1s . e . m . ) of the identified GRAF1 structures . Areas 1 and 2 from the acquisition are presented as time series ( lower panels ) . Red stars indicate cells affected by Myc \u2013 Cdc42 - Q61L . Scale bar : 10 \u03bc m . ( B ) Epifluorescence micrograph of GFP \u2013 GRAF1 and dextran structures after a 5 - min uptake in two cells , one defined as phenotypically unaffected ( control ) and the other defined as affected by Myc \u2013 Cdc42 - Q61L ( red star ) . Scale bar : 10 \u03bc m . ( C , D ) Quantification of the total dextran uptake and descriptive parameters for dextran - positive compartments for comparison of GFP \u2013 GRAF1 cells unaffected and affected by the Cdc42 - Q61L transfection , as defined in B , to circumvent inter - sample variation of cargo fluorescence . Error bars represent the s . e . m . from three independent experiments ( Mann \u2013 Whitney U tests , n = 12 \u2013 14 cells , \u03b1 = 0 . 05 ; ns , not significant ; * * P < 0 . 01 ) . ( E ) Relative internalisation of aminodextran - S - S - CW800 and transferrin - S - S - CW800 after a 30 - min uptake in Flp - In T - REx HeLa GFP \u2013 GRAF1 - R412D cells . The cargo uptake of the mutant - expressing cells collected from at least three independent experiments was normalised to the corresponding wild - type sample and the significance in relation to this samplewas assessed ( mean\u00b1s . d . ; Mann \u2013 WhitneyU tests , n = 3 \u2013 4 , \u03b1 = 0 . 05 ; ns , not significant ) . ( F ) EEA1 intensity per endosome and number of EEA1 - positive endosomes per GFP \u2013 GRAF1 cell transfected with mCherry - tagged Cdc42 or Cdc42 - Q61L . Three independent experiments wereanalysed [ mean \u00b1s . d . ; left diagram , Student \u2019 s t - test , n > 13 , 500 EEA1 - stained endosomes ( from 50 cells ) , \u03b1 = 0 . 05 ; ns , not significant ; right diagram , Mann \u2013 Whitney U test , n = 3 cells , \u03b1 = 0 . 05 ; ns , not significant ) . Bars in the left graph depict thedistribution of the mean values derived from each experiment . ( G ) Confocal fluorescence micrographs of GFP \u2013 GRAF1 cells transfected with mCherry \u2013 Cdc42 - Q61L and immunostained with anti - EEA1 antibody . Scale bar : 5 \u03bc m . ( H ) Ratio of dextran - filled Rab5 compartments in BFP \u2013 Rab5 - transfected GFP - GRAF1 cells without or with co - expressed Myc \u2013 Cdc42 - Q61L after a 5 - min cargo uptake . Cells from three independent experiments were included in the analysis [ mean\u00b1s . d . ; Mann \u2013 Whitney U test , n \u2265 5 captured frames ( at least 5 cells ) , \u03b1 = 0 . 05 ; * P < 0 . 05 ] . ( I ) Representative time series from a live - cell spinning disc confocal microscopy acquisition visualising the dynamics of dextran - filled internal compartments in GFP \u2013 GRAF1 cells co - expressing BFP \u2013 Rab5 and Myc \u2013 Cdc42 - Q61L . The yellow arrow marks a Rab5 vesicle devoid of dextran , adjacent to a cargo - filled GRAF1 compartment pointed out by the magenta arrow . The turquoise arrow marks a second GRAF1 - labelled structure devoid of dextran , which over the depicted time range fuses with the first GRAF1 compartment ( specifically note the redistribution of cargo between these two compartments ) . Scale bar : 2 \u03bc m . 4192 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e positive CLICs given that neither the duration time of GRAF1 assemblies at the surface nor the total amount of internalised fluid was influenced by expression of Cdc42 - Q61L or GRAF1 mutated in the GAP domain . However , we show that inability of Cdc42 to hydrolyse GTP results in the accumulation of rapidly moving internal carriers coated by both GRAF1 and Cdc42 . Analysis of these carriers by CLEM showed that they were composed of multiple small vesicular and tubular structures with a mean diameter of 50 nm . This is in agreement with the previously described morphology of CLICs , and the proposed membrane curvature promoted by the BAR domain of GRAF1 ( Howes et al . , 2010a ; Lundmark et al . , 2008 ) . We show that the Q61L mutation increases the affinity of Cdc42 for the GAP domain of GRAF1 , which could explain the prolonged time of residence of GRAF1 and Cdc42 on the carrier membrane . Our results suggest that this affects the maturation of carriers , possibly due to altered recruitment of endosomal proteins and lipid turnover . We found that the stalled carriers were able to fuse with each other but not with Rab5 - positive endosomes , suggesting that they were trapped in their membrane identity . CLIC and GEEC carriers have been shown to undergo homotypic and endosomal fusion ( Kalia et al . , 2006 ) . We did not detect GRAF1 in EEA1 - positive endosomes , suggesting that this protein likely comes off the nascent vesicle to allow fusion with endosomal compartments . Taken together , the transient interaction between GRAF1 and Cdc42 will terminate Cdc42 activity and promote transition of the local PIP 2 - enriched membrane environment to enable release of GRAF1 and promote endosomal maturation . This might be facilitated by the phosphoinositide 3 - kinase - driven conversion of the lipid environment . Indeed GRAF1 , but not Cdc42 , has been detected on CLICs stalled by wortmannin treatment , which inhibits phosphoinositide 3 - kinase ( Howes et al . , 2010a ) . We have previously shown that GRAF1 is detected in pleomorphic structures ( Lundmark et al . , 2008 ) . Because there is no definite cargo known for CLICs , it is difficult to know whether these structures represent distinct compartments or different steps of maturation of carriers . Characterisation of GRAF1 - positive carriers using the Flp - In T - Rex HeLa cell system revealed that the majority of GRAF1 structures detected at the cell periphery were small and transient . We also observed longer tubes , likely corresponding to the previously described endogenous GRAF1 - containing tubes ( Lundmark et al . , 2008 ) , suggesting that GRAF1 is involved in further membrane maturation events under some conditions . The stabilisation of CLIC and GEEC intermediates upon the expression of Cdc42 - Q61L phenotypically resembles the previously described trapping of an endosomal compartment induced by constant Arf6 activity ( Naslavsky et al . , 2004 ) . Interestingly , the analogous Arf GAPs of the ASAP family are involved in regulating cytoskeletal and membrane reorganisation ( Randazzo et al . , 2007 ) , and might influence Arf activity and membrane curvature through a similar co - operative regulatory mechanism as proposed here . We believe that these types of systems should not be considered canonical means for receptor sorting and vesicle generation . Instead , they are optimised to regulate cell surface dynamics and the recycling of membrane reservoirs in response to external cues such as changes in adhesion or membrane tension . MATERIALS AND METHODS Antibodies , probes and plasmids The polyclonal anti - GRAF1 rabbit antibody RA - 83 was produced as previously described ( Lundmark et al . , 2008 ) . Commercially acquired antibodies included rabbit anti - Myc ( 2272S , Cell Signaling Technology ) , mouse anti - clathrin heavy chain ( clone 23 , 610499 , BD Transduction Laboratories ) , mouse anti - AP50 ( clone 31 , 611351 , BD Transduction Laboratories ; an AP - 2 subunit ) , rabbit anti - Cdc42 ( ab109553 , Abcam ) , mouse anti - EEA1 ( 610456 , Clone 14 , BD Transduction Laboratories ) , mouse anti - GFP ( JL - 8 , 632381 , Living Colours ) and goat anti - aldolase ( AB1809 , Chemicon International , Inc . ) antibodies . Secondary antibodies conjugated to Alexa Fluor molecules ( Molecular Probes ) , horseradish peroxidase ( HRP ; Sigma - Aldrich and Agrisera ) and IRDye 800CW or 680RD ( LI - COR Biosciences ) were used for immunofluorescence and western blot detections . The fluorescent DNA probe DRAQ5 was obtained from BioStatus Limited , and the 10 kDa dextran \u2013 Alexa - Fluor - 555 and CTxB \u2013 Alexa - Fluor - 647 were purchased from Molecular Probes . Cy3 \u2013 transferrin , 800CW - S - S - transferrin and 800CW - S - S - aminodextran were produced by covalently coupling respective cargo ( Sigma - Aldrich ) to amine - reactive Cy3 ( Amersham ) and 800CW - S - S ( LI - COR Biosciences ) according to the manufacturer \u2019 s recommendations . The plasmids used in this study are summarised in Table S1 . Constructs were created by PCR amplification of inserts with flanking restriction sites for ligation into corresponding vectors . Amino acid mutations were generated through site - directed PCR mutagenesis with PCR primers specified in Table S2 . Protein purification and pulldown experiments GST fusions of wild - type GRAF1 , its arginine finger mutated GAP truncate and Cdc42 proteins were purified as previously described ( Doherty et al . , 2011a ) . Thrombin ( GE Healthcare Life Sciences ) was used for tag removal . GST \u2013 GAP and GAP - R412D fusions were purified in 25 mM HEPES pH 7 . 4 and 150 mM NaCl , and Cdc42 proteins in the same buffer supplemented with 5 mM MgCl 2 . The nucleotide loading of GTPases was performed at 30\u00b0C for 20 min in the presence of EDTA ( twofold molar excess to MgCl 2 ) and nucleotide ( 18 - fold molar excess to protein ) ( GTP \u03b3 S and GDP from Sigma - Aldrich ) . MgCl 2 ( twofold molar excess to EDTA ) was added to stop the reaction and the experimental buffer conditions were reached by buffer exchange in Micro Bio - Spin P - 6 Gel columns ( Bio - Rad ) . For the in vitro protein interaction experiments , glutathione \u2013 Sepharose - 4Bbead - bound GST - tagged protein baits were incubated with prey proteins for 1 \u2013 3 h at 4\u00b0C under rotation . The unbound protein was collected by centrifugation ( 400 g for 2 min ) , the bound protein was washed , and both fractions were boiled in sample buffer for analysis by SDS - PAGE . All assays including GDP - loaded GTPase were performed in the presence of AlF x ( 1 mM AlCl 3 , and 25 mM NaF ) . Cell culture and transfection HeLa cells ( ATCC - CRM - CCL - 2 ) were cultured in Dulbecco \u2019 s modified Eagle \u2019 s medium ( DMEM ; low glucose , L - glutamine , sodium pyruvate , HEPES and Phenol Red ) , supplemented with 10 % fetal bovine serum ( FBS ; Gibco ) . Flp - In T - REx HeLacell lines with tetracycline - inducible expression of GRAF1 , GRAF1 \u2013 GFP , GFP \u2013 GRAF1 and GFP \u2013 GRAF1 - R412D were generated as previously described ( Mohan et al . , 2015 ) . Established cultures were grown in the mentioned DMEM , exceptions being during 800CW - S - S - cargo internalisation assays and live - cell acquisitions when basic medium containing 4 . 5 g / l glucose was used ( Gibco ) . The culturing media were further supplemented with 100 \u03bc g / ml hygromycin B and 5 \u03bc g / ml blasticidin S HCl ( Gibco ) for plasmid selection , and recombinant protein expression was induced by incubation in doxycycline hyclate ( Sigma - Aldrich ) for 20\u00b14 h . Targeted protein silencing was accomplished with siRNA against GRAF1 ( \u2018 siRNAb \u2019 Stealth RNAi , Invitrogen ) ( Lundmark et al . , 2008 ) , Cdc42 # 1 ( J - 005057 - 05 - 0020 ON - TARGET plus siRNA , Dharmacon ) , Cdc42 # 2 ( VHS40393 , Stealth siRNA , Life Technologies ) or AP - 2 ( AP2M1HSS101955 Stealth RNAi , Invitrogen ) , and a Medium GC Duplex RNA ( Stealth RNAi , Invitrogen ) served as the negative control . Transfections for transient expression of plasmids ( 16\u00b14 h ) and siRNA ( 60 \u2013 96 h ) were performed using Lipofectamine 2000 ( Invitrogen ) according to the manufacturer \u2019 s recommendations . Protein expression levels were analysed in cleared cell lysates by western blotting and HRP or IRDye antigen detection using a Medical X - ray Processor ( Kodak ) or an Odyssey Sa reader ( LI - COR Biosciences ) , respectively . Cellular uptake and drug treatments To capture cargo uptake after a given time interval , cells were incubated at 37\u00b0C with fluorescently labelled CTxB ( 3 \u03bc g / ml ) , transferrin ( 2 \u03bc g / ml ) or 4193 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e dextran ( 2 mg / ml ) and washed in preheated medium before fixation . For the quantification of internalised IRDye - S - S - linked cargo , cells seeded in 24 - well plates were incubated for 30 min at 37\u00b0C with 8 \u03bc g / ml 800CW - S - S - transferrin or 0 . 18 mg / ml 800CW - S - S - aminodextran and washed in preheated medium before fixation . At room temperature , the samples were washed in stripping buffer ( 50 mM Tris - HCl pH 8 . 7 , 100 mM NaCl , and 2 . 5 mM CaCl 2 ) , reduced for 1 . 5 h in the same buffer supplemented with 60 mM MesNa and further washed in phosphate - buffered saline ( PBS ) . Cells were finally stained with 1 \u03bc M DRAQ5 for 30 min and washed in PBS before recording probe fluorescence using the Odyssey Sa reader . The uptake was defined as the ratio between the cell number ( the measured DRAQ5 fluorescence ) and the amount of internalised cargo ( the measured 800CW fluorescence ) . Cargo trafficking was also monitored by live - cell fluorescent microscopy directly after the addition of 3 \u03bc g / ml CTxB \u2013 Alexa - Fluor - 647 or after 5 min of incubation with 0 . 5 \u2013 2 mg / ml dextran \u2013 Alexa - Fluor - 555 followed by a wash with preheated medium . The effects of actin polymerisation drugs on GRAF1 structures were analysed by fluorescence microscopy of fixed cells after treatment with 2 \u03bc M cytochalasin D , 2 \u03bc M latrunculin A ( Sigma - Aldrich ) , 5 \u03bc M wiskostatin ( Calbiochem ) or DMSO . Fluorescence microscopy Cellswerefixedandpreparedforimmunofluorescence analysisaspreviously described ( Lundmark et al . , 2008 ) . Images of fixed cell samples were captured using an epifluorescence Axioimager Z1 system ( AxioCam MRm camera ) ( Zeiss ) with the ZEN Software and a 63\u00d7 lens ( Plan - Apochromat 1 . 40 Oil DIC 0 . 17 ) or an A1 R Laser Scanning Confocal Microscope system ( ANDOR iXon EMCCD camera ) ( Nikon Instruments ) under control of the NIS - Elements Microscope Imaging Software and a 60\u00d7 lens ( Apochromat 1 . 40 Oil \u03bb S 0 . 17 WD 0 . 14 , Nikon ) , at the appropriate excitation and emission wavelengths . Live - cell confocal movies were recorded using the 63\u00d7 lens in the Nikon system or the 63\u00d7 lens ( Plan - Apochromat 1 . 40 Oil DIC M27 ) in a Cell Observer Spinning Disc Confocal Microscope system ( ANDOR iXon Ultra ) ( Zeiss ) controlled by the ZEN Software . Real - time TIRF acquisitions were captured byemploying the 100\u00d7 lens in respective system [ Apochromat 1 . 49 Oil 0 . 13 - 0 . 20 DIC N2 ( Nikon ) or Plan - Apochromat 1 . 46 Oil DIC M27 ( Zeiss ) ] . Micrographs and the acquired movies were prepared ( cropped , rotated , linearly adjusted for intensity and converted ) using Adobe Photoshop or ImageJ . Correlative microscopy For correlative fluorescence and electron microscopy , cells were high - pressure frozen on sapphire discs ( 0 . 7 % low - melt agarose and DMEM ; HPM100 ) and processed by freeze substitution . 250 - \u00b5m sections were placed on 200 mesh copper grids with a carbon support film ( TAAB ) and immediately imaged by epifluorescence microscopy , with all above steps performed according to Kukulski et al . ( 2011 , 2012 ) . Grids were further dried , coated with 15 - nm gold particles conjugated to protein A ( BioCell ) and post - stained with 2 % uranyl acetate and Reynolds lead citrate . Grids were placed in a high - tilt dual - axis holder ( Fischione Instruments ) and electron tomograms corresponding to the fluorescent structures of interest ( determined from recorded grid positions ) were acquired using a Tecnai F30 electron microscope ( FEI ) at 300 kV , equipped with an Eagel 4 k CCD camera ( FEI ) using serial electron microscopy ( Mastronarde , 2005 ) . The 9400\u00d7 magnification tomograms were acquired as single - axis and the 15 , 500\u00d7 magnification tomograms as dual - axis tilt series over \u2212 60\u00b0 to 60\u00b0 ( 1 . 5\u00b0 increment ) . All tomograms were reconstructed , modelled and quantified with the IMOD package version 4 . 7 . 10 ( Kremer et al . , 1996 ) . The 15 , 500\u00d7 magnification tomograms used for the detailed models have a voxel size of 7 . 67 \u00c5 . Fluorescent micrographs , models and tomograms were overlaid using Adobe Photoshop . Image analysis and quantifications Protein bands on western blots detected with HRP - conjugated antibodies were determined from scanned images in ImageJ to derive the relative sample protein content . Estimations of the fraction of fixed cells with GFP \u2013 GRAF1 proteins in abundant or tubular structures were performed by epifluorescent visual assessment . The Imaris Software V7 . 5 ( Bitplane ) was utilised to analyse micrographs captured from fixed and live - cell samples as specified in Table S3 . The recruitment of GRAF1 \u2013 GFP to mCherry \u2013 Cdc42 membrane domains was analysed by comparing the fluorescence intensity profiles of each channel over time , from the live - cell confocal spinning disc acquisitions . The mean intensity value within a five - unit - diameter circular ROI , spanning each structure of interest , was obtained for each channel and frame using the \u2018 Plot Z - axis profile \u2019 tool in ImageJ . The start and end of each intensity peak were defined as the time points where the calculated mean intensity values rose above and declined below , respectively , the background . Parameters of GFP \u2013 GRAF1 and mCherry \u2013 Cdc42 - Q61L carriers calculated from the correlative microscopy were derived from measurements on the constructed 3D image using the IMOD package . Statistics Statistical tests were performed using Prism 5 ( GraphPad Software ) with the indicated sample size and number of independent experiments . All quantifications are visualised as the mean\u00b1s . d . unless otherwise stated . Acknowledgements The authors would like to thank the Biochemical Imaging Centre Ume\u00e5 , Ume\u00e5 Core facility for Electron Microscopy and the electron microscopy facility at EMBL - Heidelberg . Competing interests The authors declare no competing or financial interests . Author contributions M . K . F . , M . R . H . , M . V . Q . , S . H . , R . S . M . , L . S . and R . L . designed and performed experiments . M . K . F . and R . L . wrote the manuscript . All authors commented on the final manuscript . Funding This work was supported by the Swedish Cancer Society [ grant number 2012 / 751 ] ; the Swedish Research Council [ grant number 821 - 2013 - 2241 ] ; the Swedish Foundation for Strategic Research [ grant number FFL09 - 0181 ] ; the Kempe foundations ; the Baltic Group foundations ; Molecular Infection Medicine Sweden ( MIMS ) ; and Ume\u00e5 Centre for Microbial Research ( UCMR ) . Deposited in PMC for immediate release . Supplementary information Supplementary information available online at http : / / jcs . biologists . org / lookup / suppl / doi : 10 . 1242 / jcs . 174417 / - / DC1 References Bendezu \u0301 , F . O . , Vincenzetti , V . , Vavylonis , D . , Wyss , R . , Vogel , H . and Martin , S . G . ( 2015 ) . Spontaneous Cdc42 polarization independent of GDI - mediated extraction and actin - based trafficking . PLoS Biol . 13 , e1002097 . Boucrot , E . , Ferreira , A . P . A . , Almeida - Souza , L . , Debard , S . , Vallis , Y . , Howard , G . , Bertot , L . , Sauvonnet , N . and McMahon , H . T . ( 2015 ) . Endophilinmarks and controls a clathrin - independent endocytic pathway . Nature 517 , 460 - 465 . Bretou , M . , Jouannot , O . , Fanget , I . , Pierobon , P . , Larochette , N . , Gestraud , P . , Guillon , M . , Emiliani , V . , Gasman , S . , Desnos , C . et al . ( 2014 ) . Cdc42 controls the dilation of the exocytotic fusion pore by regulating membrane tension . Mol . Biol . Cell 25 , 3195 - 3209 . Chadda , R . , Howes , M . T . , Plowman , S . J . , Hancock , J . F . , Parton , R . G . and Mayor , S . ( 2007 ) . Cholesterol - sensitive Cdc42 activation regulates actin polymerization for endocytosis via the GEEC pathway . Traffic 8 , 702 - 717 . Chaudhary , N . , Gomez , G . A . , Howes , M . T . , Lo , H . P . , McMahon , K . - A . , Rae , J . A . , Schieber , N . L . , Hill , M . M . , Gaus , K . , Yap , A . S . et al . ( 2014 ) . Endocytic crosstalk : cavins , caveolins , and caveolae regulate clathrin - independent endocytosis . PLoS Biol . 12 , e1001832 . de Curtis , I . and Meldolesi , J . ( 2012 ) . Cell surface dynamics - how Rho GTPases orchestrate the interplay between the plasma membrane and the cortical cytoskeleton . J . Cell Sci . 125 , 4435 - 4444 . Doherty , G . J . , Ahlund , M . K . , Howes , M . T . , Moren , B . , Parton , R . G . , McMahon , H . T . andLundmark , R . ( 2011a ) . TheendocyticproteinGRAF1isdirectedtocell - matrix adhesion sites and regulates cell spreading . Mol . Biol . Cell 22 , 4380 - 4389 . Doherty , J . T . , Lenhart , K . C . , Cameron , M . V . , Mack , C . P . , Conlon , F . L . and Taylor , J . M . ( 2011b ) . Skeletal muscle differentiation and fusion are regulated by the BAR - containing Rho - GTPase - activating protein ( Rho - GAP ) , GRAF1 . J . Biol . Chem . 286 , 25903 - 25921 . Gauthier , N . C . , Masters , T . A . and Sheetz , M . P . ( 2012 ) . Mechanical feedback between membrane tension and dynamics . Trends Cell Biol . 22 , 527 - 535 . 4194 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e Gowrishankar , K . , Ghosh , S . , Saha , S . , C , R . , Mayor , S . and Rao , M . ( 2012 ) . Active remodeling of cortical actin regulates spatiotemporal organization of cell surface molecules . Cell 149 , 1353 - 1367 . Hansen , C . G . and Nichols , B . J . ( 2009 ) . Molecular mechanisms of clathrin - independent endocytosis . J . Cell Sci . 122 , 1713 - 1721 . Hildebrand , J . D . , Taylor , J . M . and Parsons , J . T . ( 1996 ) . An SH3 domain - containing GTPase - activating protein for Rho and Cdc42 associates with focal adhesion kinase . Mol . Cell . Biol . 16 , 3169 - 3178 . Howes , M . T . , Kirkham , M . , Riches , J . , Cortese , K . , Walser , P . J . , Simpson , F . , Hill , M . M . , Jones , A . , Lundmark , R . , Lindsay , M . R . et al . ( 2010a ) . Clathrin - independent carriers form a high capacity endocytic sorting system at the leading edge of migrating cells . J . Cell Biol . 190 , 675 - 691 . Howes , M . T . , Mayor , S . and Parton , R . G . ( 2010b ) . Molecules , mechanisms , and cellularrolesofclathrin - independentendocytosis . Curr . Opin . CellBiol . 22 , 519 - 527 . Jaffe , A . B . and Hall , A . ( 2005 ) . Rho GTPases : biochemistry and biology . Annu . Rev . Cell Dev . Biol . 21 , 247 - 269 . Jelen , F . , Lachowicz , P . , Apostoluk , W . , Mateja , A . , Derewenda , Z . S . and Otlewski , J . ( 2009 ) . Dissecting the thermodynamics of GAP \u2013 RhoA interactions . J . Struct . Biol . 165 , 10 - 18 . Kalia , M . , Kumari , S . , Chadda , R . , Hill , M . M . , Parton , R . G . andMayor , S . ( 2006 ) . Arf6 - independent GPI - anchored protein - enriched early endosomal compartments fuse with sorting endosomes via a Rab5 / phosphatidylinositol - 3 \u2032 - kinase - dependent machinery . Mol . Biol . Cell 17 , 3689 - 3704 . Kremer , J . R . , Mastronarde , D . N . and McIntosh , J . R . ( 1996 ) . Computer visualization of three - dimensional image data using IMOD . J . Struct . Biol . 116 , 71 - 76 . Kukulski , W . , Schorb , M . , Welsch , S . , Picco , A . , Kaksonen , M . and Briggs , J . A . G . ( 2011 ) . Correlated fluorescence and 3D electron microscopy with high sensitivity and spatial precision . J . Cell Biol . 192 , 111 - 119 . Kukulski , W . , Schorb , M . , Welsch , S . , Picco , A . , Kaksonen , M . and Briggs , J . A . G . ( 2012 ) . Precise , correlated fluorescence microscopy and electron tomography of lowicryl sections using fluorescent fiducial markers . Methods Cell Biol . 111 , 235 - 257 . Kumari , S . andMayor , S . ( 2008 ) . ARF1isdirectlyinvolvedindynamin - independent endocytosis . Nat . Cell Biol . 10 , 30 - 41 . Kumari , S . , Mg , S . and Mayor , S . ( 2010 ) . Endocytosis unplugged : multiple waysto enter the cell . Cell Res . 20 , 256 - 275 . Lenhart , K . C . , Becherer , A . L . , Li , J . , Xiao , X . , McNally , E . M . , Mack , C . P . and Taylor , J . M . ( 2014 ) . GRAF1 promotes ferlin - dependent myoblast fusion . Dev . Biol . 393 , 298 - 311 . Longenecker , K . L . , Zhang , B . , Derewenda , U . , Sheffield , P . J . , Dauter , Z . , Parsons , J . T . , Zheng , Y . and Derewenda , Z . S . ( 2000 ) . Structure of the BH domain from graf and its implications for Rho GTPase recognition . J . Biol . Chem . 275 , 38605 - 38610 . Lundmark , R . , Doherty , G . J . , Howes , M . T . , Cortese , K . , Vallis , Y . , Parton , R . G . andMcMahon , H . T . ( 2008 ) . TheGTPase - activatingproteinGRAF1regulatesthe CLIC / GEEC endocytic pathway . Curr . Biol . 18 , 1802 - 1808 . Mastronarde , D . N . ( 2005 ) . Automated electron microscope tomography using robust prediction of specimen movements . J . Struct . Biol . 152 , 36 - 51 . Mohan , J . , More \u0301 n , B . , Larsson , E . , Holst , M . R . and Lundmark , R . ( 2015 ) . Cavin3 interacts with cavin1 and caveolin1 to increase surface dynamics of caveolae . J . Cell Sci . 128 , 979 - 991 . Mayor , S . and Pagano , R . E . ( 2007 ) . Pathways of clathrin - independent endocytosis . Nat . Rev . Mol . Cell Biol . 8 , 603 - 612 . Mittal , R . , Ahmadian , M . R . , Goody , R . S . and Wittinghofer , A . ( 1996 ) . Formation of a transition - state analog of the Ras GTPase reaction by Ras GDP , tetrafluoroaluminate , and GTPase - activating proteins . Science 273 , 115 - 117 . Naslavsky , N . , Weigert , R . and Donaldson , J . G . ( 2004 ) . Characterization of a nonclathrin endocytic pathway : membrane cargo and lipid requirements . Mol . Biol . Cell 15 , 3542 - 3552 . Nixon , S . J . , Webb , R . I . , Floetenmeyer , M . , Schieber , N . , Lo , H . P . and Parton , R . G . ( 2009 ) . A single method for cryofixation and correlative light , electron microscopy and tomography of zebrafish embryos . Traffic 10 , 131 - 136 . Nonnenmacher , M . and Weber , T . ( 2011 ) . Adeno - associated virus 2 infection requires endocytosis through the CLIC / GEEC pathway . Cell Host Microbe 10 , 563 - 576 . Okada , H . , Uezu , A . , Mason , F . M . , Soderblom , E . J . , Moseley , M . A . , III and Soderling , S . H . ( 2011 ) . SH3 domain - based phototrapping in living cells reveals Rho family GAP signaling complexes . Sci . Signal . 4 , rs13 . Randazzo , P . A . , Inoue , H . and Bharti , S . ( 2007 ) . Arf GAPs as regulators of the actin cytoskeleton . Biol . Cell 99 , 583 - 600 . Ridley , A . J . ( 2001 ) . Rho proteins : linking signaling with membrane trafficking . Traffic 2 , 303 - 310 . Ro \u0308 mer , W . , Pontani , L . - L . , Sorre , B . , Rentero , C . , Berland , L . , Chambon , V . , Lamaze , C . , Bassereau , P . , Sykes , C . , Gaus , K . et al . ( 2010 ) . Actin dynamics drivemembranereorganizationandscissioninclathrin - independentendocytosis . Cell 140 , 540 - 553 . Sabharanjak , S . , Sharma , P . , Parton , R . G . and Mayor , S . ( 2002 ) . GPI - anchored proteins are delivered to recycling endosomes via a distinct cdc42 - regulated , clathrin - independent pinocytic pathway . Dev . Cell 2 , 411 - 423 . Simpson , K . J . , Selfors , L . M . , Bui , J . , Reynolds , A . , Leake , D . , Khvorova , A . and Brugge , J . S . ( 2008 ) . Identification of genes that regulate epithelial cell migration using an siRNA screening approach . Nat . Cell Biol . 10 , 1027 - 1038 . Stergiou , L . , Bauer , M . , Mair , W . , Bausch - Fluck , D . , Drayman , N . , Wollscheid , B . , Oppenheim , A . and Pelkmans , L . ( 2013 ) . Integrin - mediated signaling induced by simian virus 40 leads to transient uncoupling of cortical actin and the plasma membrane . PLoS ONE 8 , e55799 . 4195 RESEARCH ARTICLE Journal of Cell Science ( 2015 ) 128 , 4183 - 4195 doi : 10 . 1242 / jcs . 174417 J o u r n a l o f C e ll S c i e n c e",
    "rohatgi1999interaction": "Cell , Vol . 97 , 221 \u2013 231 , April 16 , 1999 , Copyright \u00aa 1999 by Cell Press The Interaction between N - WASP and the Arp2 / 3 Complex Links Cdc42 - Dependent Signals to Actin Assembly 1999 ) . Because of their central role in signal transduc - tion , a great deal of interest has focused on the identifi - cation of downstream pathways that link these G pro - teins to the actin cytoskeleton ( Hall , 1998 ) . Progress has also been made in identifying compo - Rajat Rohatgi , * (cid:107) Le Ma , * (cid:107) Hiroaki Miki , \u2021 Marco Lopez , * \u2020 Tomas Kirchhausen , * \u2020 Tadaomi Takenawa , \u2021 and Marc W . Kirschner * \u00a7 * Department of Cell Biology \u2020 The Center for Blood Research Harvard Medical School nents closer to the process of actin polymerization at the plasma membrane . Actin polymerization can be reg - Boston , Massachusetts 02115 \u2021 Department of Biochemistry ulated by uncapping filament barbed ends ( Hartwig et al . , 1995 ) , by severing filaments ( Arber et al . , 1998 ; Yang Institute of Medical Science University of Tokyo et al . , 1998 ) , or by de novo nucleation ( Zigmond , 1998 ) . The Arp2 / 3 complex is the best candidate to generate Tokyo 108 Japan new barbed ends by stimulating nucleation ( Bi and Zig - mond , 1999 ; Machesky and Gould , 1999 ) . The complex is composed of seven polypeptides and has been puri - fied from Acanthamoeba castellanii ( Machesky et al . , Summary 1994 ) , Saccharomyces cerevisiae ( Winter et al . , 1997 ) , humans ( Welch et al . , 1997b ) , and Xenopus laevis ( Ma Although small GTP - binding proteins of the Rho family et al . , 1998b ) . The Arp2 / 3 complex is required for actin havebeenimplicatedinsignaling totheactincytoskel - assembly stimulated by the motile pathogen Listeria eton , the exact nature of the linkage has remained monocytogenes ( Welch et al . , 1997b ) , implicated in the obscure . We describe a novel mechanism that links polarized organization and motility of actin patches of one Rho family member , Cdc42 , to actin polymeriza - yeast ( Winter et al . , 1997 ) and required for Cdc42 - induced tion . N - WASP , a ubiquitously expressed Cdc42 - inter - actin assembly in Xenopus egg extracts ( Ma et al . , acting protein , is required for Cdc42 - stimulated actin 1998b ) . The complexnucleatesfilamentsin vitro ( Mullins polymerization in Xenopus egg extracts . The C terminus et al . , 1998 ) and is localized to actin - rich structures ( Kel - of N - WASP binds to the Arp2 / 3 complex and dramati - leher et al . , 1995 ; Welch et al . , 1997a ; Winter et al . , 1997 ) . cally stimulatesits abilityto nucleateactin polymeriza - Although the actin nucleation activity of the Arp2 / 3 tion . Although full - length N - WASP is less effective , its complex is weak ( Mullins et al . , 1998 ) , it can be stimu - activity can be greatly enhanced by Cdc42 and phos - lated by the Listeria protein ActA ( Welch et al . , 1998 ) . phatidylinositol ( 4 , 5 ) bisphosphate . Therefore , N - WASP This result points to the presence of analogous cellular and the Arp2 / 3 complex comprise a core mechanism mechanisms , presumably tied to signal transduction that directly connects signal transduction pathways pathways , for regulating the complex ( Beckerle , 1998 ) . to the stimulation of actin polymerization . Our demonstration that the Arp2 / 3 complex is required for Cdc42 - induced actin assembly in Xenopus extracts Introduction suggests that the Cdc42 signaling pathway might be such a cellular mechanism ( Ma et al . , 1998b ) . However , The actin cytoskeleton is a dynamic filament network intermediate components between Cdc42 and Arp2 / 3 that is essential for cell movement , polarization , mor - are clearly required because the complex does not di - phogenesis , and cell division ( Drubin and Nelson , 1996 ; rectly bind to Cdc42 , and it is not sufficient to mediate Mitchison and Cramer , 1996 ; Field et al . , 1999 ) . To en - Cdc42 - induced actin polymerization ( Ma et al . , 1998b ) . gage in these complex behaviors , cells must direct actin Among all the Cdc42 - interacting proteins identified to assembly with a high degree of spatial and temporal date , the WASP family proteins , especially WASP and resolution in response to extracellular signals . Over the N - WASP , are the best candidates for mediating the last 10 years , extensive research using genetic , cell bio - effects of Cdc42 on the actin cytoskeleton . WASP logical , and biochemical approaches has focused on ( W iskott - A ldrich s yndrome p rotein ) , which is only ex - the signal transduction pathways that link cell surface pressed in hematopoietic cells , was originally identified receptors to actin assembly at the plasma membrane . as a protein mutated in patients with Wiskott - Aldrich In the early 1990s , members of the Rho family of small syndrome ( WAS ) . N - WASP , which is ubiquitously ex - GTP - binding proteins ( G proteins ) , including Rac , Rho , pressed , shares (cid:122) 50 % homology with WASP ( Fukuoka and Cdc42 , were found to cause distinct morphological et al . , 1997 ) . Both proteins possess several domains : a changes in the actin cytoskeleton upon injection into pleckstrin homology ( PH ) domain that binds phosphati - cultured cells ( Ridley and Hall , 1992 ; Ridley et al . , 1992 ; dylinositol ( 4 , 5 ) bisphosphate ( PI ( 4 , 5 ) P 2 ) , a Cdc42 - bind - Nobes et al . , 1995 ) . Since then , Rho family G proteins ing ( GBD ) domain , a proline - rich region , a G - actin - bind - have been shown to regulate actin - dependent pro - ing verprolin homology ( V ) domain , a domain ( C ) with cesses in many different systems ( Hall , 1998 ; Johnson , homology to the actin - depolymerizing protein cofilin , and finally a C - terminal acidic segment ( A ) ( Figure 1A ) ( Miki et al . , 1996 ) . Both WASP and N - WASP induce actin \u00a7 To whom correspondence should be addressed ( e - mail : marc @ polymerization when overexpressed in fibroblasts ( Miki hms . harvard . edu ) . (cid:107) These authors contributed equally to this work . et al . , 1996 ; Symons et al . , 1996 ) . Cell222 Figure 1 . N - WASP Is Required for Cdc42 - Induced Actin Assembly in High - Speed Su - pernatants of Xenopus Egg Extracts ( A ) Schematic diagram of N - WASP showing its domain structure . Abbreviations for the domains used throughout the text are indi - cated inside the boxes . Numbers refer to amino acid residues . The H208D and (cid:68) cof mutantsareshownbelowalongwiththesitesofthemutations . ( B ) Immunodepletion of N - WASP from Xeno - pus HSS . Affinity - purified (cid:97) - N - WASP anti - body or antibody buffer ( mock ) was used to immunodeplete N - WASP , and the material left in the supernatant or captured on the beadswasanalyzedbyimmunoblottingusingtheaffinity - purified (cid:97) - N - WASP antibody . 0 . 6 % of the input material ( HSS ) , 0 . 7 % of the supernatant material , and 2 % of the material captured on beads was loaded on the gel . ( C ) Comparison of actin assembly stimulated by250nMGTP (cid:103) S - chargedGST - Cdc42inun - treated HSS , mock - depleted HSS , (cid:97) - N - WASP depleted HSS , and in depleted HSS reconsti - tutedwith50nMrecombinantN - WASP . Poly - merization kinetics were monitored in the HSS using the pyrene actin assay in which fluorescence increase indicates filament for - mation . ( D ) Comparison of actin assembly stimulated by250nMGTP (cid:103) S - chargedGST - Cdc42in (cid:97) - N - WASP depleted HSS to which the indicated concentrations of recombinant N - WASP were added . The initial rate of actin assembly and the maximum level of F - actin attained were quantitated from curves of the type shown in ( C ) . ( E ) Relative ability of wild - type N - WASP , the H208D mutant , and the (cid:68) cof mutant to re - storeactinassemblyactivitytoHSSdepletedofendogenousN - WASP . In all cases , actin assembly was initiated by the addition of 250 nM GTP (cid:103) S - charged GST - Cdc42 . N - WASP is a particularly good candidate for mediat - Results ing the effects of Cdc42 on the cytoskeleton because it can interact with a Cdc42 mutant ( Y40C ) that still N - WASP Is Required for Cdc42 - Induced Actin Polymerization in Xenopus Egg Extracts induces filopodia but which cannot bind to a variety of other targets , including WASP ( Miki et al . , 1998a ) . In We have previously described the reconstitution of Cdc42 - stimulated actin assembly in high - speed super - addition , dominant - negative mutations in N - WASP block Cdc42 - induced filopodium formation , and wild - type natants from Xenopus egg extracts ( \u201c Xenopus HSS\u201d or \u201c Xenopus extract\u201d in the rest of the text ) ( Ma et al . , N - WASP potentiates Cdc42 - induced microspike forma - tion ( Miki et al . , 1998a ) . Finally , in search for other com - 1998a ) . Thekineticsofactinpolymerizationcanbemoni - tored by supplementing the extracts with pyrene actin , a ponents ( Ma et al . , 1998b ) required for Cdc42 - induced actin assembly in vitro , we found that N - WASP partially fluorescent derivative of actin that exhibits much higher fluorescence intensity when actin is assembled into fila - copurified with MCAP2 ( m ediator of C dc42 - induced a ctin p olymerization ; unpublished observation ) . ments . Using this cell - free system , we have demon - strated a requirement for the Arp2 / 3 complex in Cdc42 - We show here that N - WASP can activate the Arp2 / 3 complex to stimulate actin polymerization in vitro using induced actin polymerization ( Ma et al . , 1998b ) . In order to test the requirement for N - WASP in this purified components . This N - WASP \u2013 Arp2 / 3 interaction is also required for Cdc42 - stimulated actin assembly in system , we raised a polyclonal rabbit antiserum against a full - length , recombinant rat N - WASP protein that is Xenopus egg extracts . Although full - length N - WASP is only weakly active , a C - terminal fragment confers a dra - (cid:46) 90 % identical to the bovine and human forms . After affinity purification , the antibody detected a single (cid:122) 65 matic 30 - fold acceleration of actin assembly kinetics . The activity of full - length N - WASP can be increased to kDa band in Xenopus extracts on immunoblots and also immunoprecipitated it ( Figure 1B ) . Using this affinity - levels similar to the C - terminal fragment in the presence of two putative regulators of N - WASP , Cdc42 and purified antibody , we immunodepleted (cid:46) 95 % of N - WASP from Xenopus HSS ( Figure 1B ) . The depletion of N - WASP PI ( 4 , 5 ) P 2 . Thus , the N - WASP \u2013 Arp2 / 3 interaction seems to be a critical link in Cdc42 - dependant signaling path - eliminates the ability of the extract to support Cdc42 - stimulated actin polymerization ( Figure 1C ) . The activity ways that control actin assembly . N - WASP Interaction with the Arp2 / 3 Complex 223 Figure 2 . Full - Length N - WASP Cooperates with the Arp2 / 3 Complex to Accelerate Actin Polymerization In Vitro ( A ) The composition of the Arp2 / 3 complex purified from bovine brain is shown on a 12 % polyacrylamidegelstainedwithGelcodeBlue ( Coomassie G - 250 ) . The thin , unlabeled lines on the right indicate the seven subunits of the complex . Listed in descending order of molecular weight , the subunits are Arp3 , Arp2 , p41 - ARC , p34 - ARC , p21 - ARC , p20 - ARC , and p16 - ARC . ( B ) Thepyrene actin assaywas usedto moni - tor the polymerization of 2 . 5 (cid:109) M G - actin ( 1 . 5 (cid:109) M unlabeled actin (cid:49) 1 (cid:109) M pyrene actin ) in the presence of 250 nM Arp2 / 3 complex alone , 400 nM N - WASP alone , or both com - ponents added together . In all the curves shown here and in the remainder of the fig - ures , polymerizationwasinitiatedattime (cid:53) 0 . ( C ) Comparison of the abilities of wild - type N - WASP ( 400 nM ) and the (cid:68) cof mutant ( 400 nM or 800 nM ) to stimulate actin polymeriza - tionin thepresence of60 nMArp2 / 3 complex underconditionsdescribedin ( B ) . Thecontrol curve depicts actin polymerization in the presence of Arp2 / 3 alone . can be rescued by adding back purified recombinant Full - Length N - WASP Cooperates with the Arp2 / 3 Complex to Produce a Modest Acceleration N - WASP protein at approximately physiological con - of Actin Polymerization with Purified centrations ( (cid:122) 20 \u2013 50 nM ) ( Suzuki et al . , 1998 ) . Adding Components In Vitro back increasing amounts of N - WASP leads to a dose - Since we had identified requirements for both N - WASP dependant increase both in the initial rate and in the and the Arp2 / 3 complex in Cdc42 - induced actin assem - final level of actin polymerization induced by a fixed bly , we wanted to determine whether N - WASP could amount of Cdc42 ( Figure 1D ) . synergize with the Arp2 / 3 complex to accelerate actin Thespecificity oftheadd - backexperiment wastested polymerization as observed for the ActA protein of Liste - using two previously characterized mutants of N - WASP , ria ( Welch et al . , 1998 ) . For these assays , we used the H208D and (cid:68) cof ( Figure 1A ) . H208D cannot bind to Arp2 / 3 complex purified from bovine brain extracts ( Fig - Cdc42 due to a point mutation in its Cdc42 - binding ure 2A ) . When we used pyrene actin fluorescence to domain and does not potentiate Cdc42 - induced filo - monitor actin assembly in the presence of either pure podia formation when introduced into cells ( Miki et al . , N - WASP or pure Arp2 / 3 complex , the kinetics of assem - 1998a ) . (cid:68) cof , containing a four \u2013 amino acid deletion in bly were not significantly different from those of actin the C domain of N - WASP , does not induce actin reorga - alone ( Figure 2B ) . Actin assembly kinetics demonstrate nization when overexpressed in Cos - 7 cells and acts in an initial lag phase , reflecting the kinetic barrier to nucle - a dominant - negative fashion to block Cdc42 - induced ation , followed by a rapid , linear phase during which filopodium formation ( Miki et al . , 1998a ) . In Xenopus filaments elongate by monomer addition . In the pres - HSS depleted of N - WASP , these mutants cannot rescue ence of both Arp2 / 3 and N - WASP , the slope of the elon - Cdc42 - induced actin assembly even when added back gation phase is increased by (cid:122) 4 - fold and the lag phase at concentrations 20 - fold above the wild - type protein is shortened , although not completely eliminated ( Figure ( Figure 1E ) . These results show that regions of N - WASP 2B ) . Addition of increasing amounts of N - WASP to a implicated in interactions with both upstream regulators fixed concentration of Arp2 / 3 caused a dose - dependant ( Cdc42 ) and downstream targets are necessary for its acceleration of actin polymerization ( Figure 5C ) . ability to support Cdc42 - induced actin assembly in The (cid:68) cof mutantdescribed abovedoes notaccelerate actin assembly in the presence of the Arp2 / 3 complex Xenopus HSS . Cell224 Figure 3 . N - WASP Binds to and Activates the Arp2 / 3 Complex via Its C Terminus ( A ) Schematic showing the domain structure of full - length N - WASP and the various C - terminal fragments of N - WASP constructed as GST fusion proteins . All fusion proteins are named according to the C - terminal segments ( V , C , or A ) of N - WASP that they contain . ( B ) The GST fusion proteins shown in ( A ) were immobilized on glutathione - Sepharose beads and tested for their abilities to pull down either theArp2 / 3complexorG - actinfromsolution . ThepresenceoftheArp2 / 3complexorG - actinboundtothebeadswasassayedbyimmunoblotting with (cid:97) - Arp2 or (cid:97) - actin , respectively . ( C ) Actin polymerization ( 2 . 5 (cid:109) M actin and 60 nM Arp2 / 3 complex ) was followed in the presence of GST - VCA ( 50 nM ) , GST - V ( 200 nM ) , GST - VC ( 200 nM ) , GST - CA ( 200 nM ) , or in the absence of any additions ( control ) . ( D ) The initial phases of actin polymerization ( 2 . 5 (cid:109) M actin and 60 nM Arp2 / 3 complex ) , including the lag phase , are shown in the presence of indicated concentrations of GST - VCA . ( E ) Summary of the G - actin binding , Arp2 / 3 binding , and Arp2 / 3 activation properties of full - length N - WASP and the fragments shown in ( A ) . Binding of full - length N - WASP to Arp2 / 3 was determined by coimmunoprecipitation rather than GST pull down . Binding of GST - V and GST - VCA was determined by both methods . ( Figure 2C ) , suggesting a biochemical mechanism for its the p21ARC subunit of the Arp2 / 3 complex ( Machesky and Insall , 1998 ) . phenotypes in tissue culture cells as well as in Xenopus extracts ( Figure 1E ) . Unexpectedly , the H208D mutant , We expressed and purified GST fusion proteins con - taining various portions of this VCA region ( diagrammed which has only one point mutation in the GBD domain , does not accelerate actin assembly ( data not shown ) , in Figure 3A ) and tested their abilities to interact with the Arp2 / 3 complex in a GST pull - down assay . The complex suggesting that this mutation leads to defects in func - tions other than just Cdc42 binding . bound to GST - CA and GST - VCA but not to GST , GST - V , and GST - VC ( Figure 3B ) . As a control , the fragments were tested for binding to G - actin and , as expected , the The C Terminus of N - WASP Strongly Accelerates Arp2 / 3 - Mediated Actin Assembly presence of the V domain perfectly correlated with this property ( Figure 3B ) . We conclude that the C - terminal To define the region of N - WASP required for activity , we concentrated on a C - terminal region of N - WASP 55 amino acids of N - WASP ( 450 \u2013 505 ) are necessary and sufficient to bind to the Arp2 / 3 complex and that the ( amino acids 392 \u2013 505 ) , which contains a G - actin - bind - ing V domain , a cofilin homology sequence ( C ) , and an ability to bind G - actin through the V domain is not re - quired . We were unable to detect any interaction be - acidic segment ( A ) ( Figures 1A and 3A ) . The VCA region is thought to mediate interactions of N - WASP with tween full - length N - WASP and Arp2 / 3 by coimmuno - precipitation under conditions where the binding of downstream effectors because mutations in this region abrogate its ability to induce actin rearrangements upon GST - VCA to Arp2 / 3 is readily detectable ( data not shown ) . injection into cells andoften function as dominant - nega - tive mutations ( Miki et al . , 1998a ) . In addition , a homolo - Since the C - terminal region of N - WASP could bind to the Arp2 / 3 complex and actin , we wanted to determine gous region from other WASP family proteins ( Scar1 / WAVE and WASP ) was recently found to interact with whether it could confer the activity of the full - length N - WASP Interaction with the Arp2 / 3 Complex 225 Figure 4 . The GST - CA Fragment Can Inhibit the Interaction between N - WASP and the Arp2 / 3 Complex ( A ) Stimulation of actin polymerization ( 2 . 5 (cid:109) M actin and 60 nM Arp2 / 3 complex ) by N - WASP ( 200 nM ) is inhibited by GST - CA but not GST - V . ( B ) Stimulation of actin polymerization ( 2 . 5 (cid:109) M actin and 60 nM Arp2 / 3 complex ) by GST - VCA ( 1 nM ) is inhibited by GST - CA in a dose - dependent manner . ( C ) Cdc42 - stimulated actin polymerization in Xenopus HSS is inhibited by GST - CA , but not GST - V or GST - VC ( 200 nM each ) . ( D ) Dose - dependent inhibition of Cdc42 - stimulated actin polymerization in Xenopus HSS by GST - CA as compared using the initial rate and maximal level of F - actin assembly . The control depicts assembly in the absence of any GST - CA . protein to accelerate actin polymerization . When the N - WASP \u2013 Arp2 / 3 interaction ( Figure 3E ) . When this model was tested using purified components , GST - CA purified fragments were mixed with pure Arp2 / 3 com - plex , only the GST - VCA fusion protein accelerated as - ( but not GST - V , GST - VC , or GST ) inhibits the ability of both full - length N - WASP and the GST - VCA fragment to semblykinetics , shorteningthelagphaseandincreasing the elongation rate by 30 - fold ( Figure 3C ) . Figure 3D stimulate actin polymerization in the presence of the Arp2 / 3 complex ( Figures 4A and 4B ) . clearly shows that the GST - VCA fragment shortens the lag phase to polymerization in a dose - dependent man - The efficacy of the GST - CA fragment as a dominant negative was tested in Xenopus HSS for its effects on ner , suggesting that it can lower the kinetic barrier to nucleation . These effects are dependent on the pres - Cdc42 - induced actin assembly . When used in extracts at concentrations roughly equivalent to the estimated ence of the Arp2 / 3 complex and can be inhibited by cytochalasin D ( data not shown ) . The GST - V , GST - VC , endogenousArp2 / 3concentration ( (cid:122) 100 \u2013 300nM ) , GST - CA specifically inhibited Cdc42 - induced actin assembly and GST - CA fragments have no effect , even when used at 4 - fold higher concentrations ( Figure 3C ) . The GST - ( Figures 4C and 4D ) . This result strongly suggests that the kind of interaction between N - WASP and the Arp2 / 3 VCA fragment is both much more potent and effective in stimulating actin polymerization when compared to complex that we have observed using purified compo - nents in vitro is also required for Cdc42 - induced actin the full - length N - WASP , a point we will return to later . Thus , the requirements for binding to the Arp2 / 3 com - polymerization in a more complete and physiological system . plex are different from the requirements for activation , as summarized in Figure 3E . The C - Terminal Domain of N - WASP Is at Least 100 - Fold More Potent Than A C - Terminal Fragment of N - WASP Can Act as a Dominant Inhibitor and Block the Function the Full - Length Protein Using purified components , N - WASP ( 200 nM ) can stim - of N - WASP and the Arp2 / 3 Complex The observation that the GST - CA fragment can bind to ulate actin polymerization in the presence of only the Arp2 / 3 complex and G - actin ( Figure 2 ) . However , when the Arp2 / 3 complex but cannot activate it suggests that it can be used as a competitive inhibitor to block the added to Xenopus HSS at concentrations ( 400 nM ) as Cell226 Figure 5 . Full - Length N - WASP and the GST - VCA Fragment Activate Actin Polymerization with Different Potencies ( A ) Comparion of spontaneous ( Cdc42 - independent ) actin polymerization in Xenopus HSS treated with the indicated concentrations of full - length N - WASP , GST - CA , or GST - VCA . ( B ) Actin polymerization ( 2 . 5 (cid:109) M actin and 60 nM Arp2 / 3 complex ) stimulated by full - length N - WASP ( 200 nM ) or GST - VCA ( 10 nM ) . ( C ) Dose \u2013 response curves showing the variation in the maximum rate of filament elongation as a function of increasing concentrations of either GST - VCA or full - length N - WASP in the presence of 60 nM Arp2 / 3 complex . The elongation rate was calculated from the linear phase of polymerization curves of the type shown in ( B ) . The main graph depicts the concentrations ( abscissa ) on a log scale , whereas the inset depicts them on a linear scale . At higher concentrations of GST - VCA , the elongation rate saturates and is no longer a good indicator of activity . Instead , the shortening of the lag phase becomes more dramatic ( Figure 3D ) . high as 8 - fold above the estimated endogenous N - WASP difference between the full - length protein and the GST - VCA fragment in their ability to stimulate actin polymer - concentration ( 50 nM ) , full - length N - WASP did not stim - ulate spontaneous actin assembly in the absence of ization in the presence of the Arp2 / 3 complex . When full - length N - WASP was directly compared to the GST - activated Cdc42 ( Figure 5A ) . To test whether this differ - ence in the behavior of N - WASP in the purified system VCA fragment , a large difference in activity was observed : 10 nM GST - VCA stimulates polymerization in vitro far compared to the extract is due to inhibitory factors that interact with its N terminus , we added the GST - VCA more effectively than 200 nM N - WASP ( Figure 5B ) . To obtain a more quantitative measure of the activity fragment to Xenopus HSS . We reasoned that GST - VCA would bypass the Cdc42 requirement and act in a domi - difference between full - length N - WASP and the GST - VCA fragment , we generated dose \u2013 response curves for nant active fashion in the extract . Indeed , when GST - VCA was added to the HSS ( in the absence of any acti - both proteins , using elongation rates as an indicator of activation ( Figure 5C ) . The concentration required for vated Cdc42 ) even at very low concentrations , actin polymerization was dramatically stimulated : the effect half maximal activation is (cid:122) 250 nM for the full - length protein and (cid:122) 2 nM for GST - VCA ( Figure 5C ) . Based on of 1 nM GST - VCA was comparable to that of 800 nM N - WASP ( Figure 5A ) . This observation suggests that elongation rates , the C - terminal VCA domain is at least 100 times more potent than full - length N - WASP . This N - WASP is regulated in extracts , a property that is con - ferred by the N terminus of the molecule . potency difference is also demonstrated by the (cid:122) 20 - fold higher molar ratio of GST - CA needed to inhibit GST - How might the activity of N - WASP be regulated ? It hasbeen proposedthatN - WASPis intrinsicallyinhibited VCA compared to the full - length protein ( Figures 4A and 4B ) . The efficacy of GST - VCA at inducing actin polymer - and is activated when the VCA region is unmasked , perhaps by the binding of upstream regulators to the N ization is also about 5 - fold higher than that of N - WASP ( Figure 5C ) . The difference in activity between GST - VCA terminus ( Miki et al . , 1998a ) . This model predicts a large N - WASP Interaction with the Arp2 / 3 Complex 227 Figure 6 . Cdc42 and PI ( 4 , 5 ) P 2 Coordinately Activate Full - Length N - WASP In Vitro ( A ) The effect of GTP (cid:103) S - Cdc42 ( 500 nM ) , PI ( 4 , 5 ) P 2 - containing vesicles ( 100 (cid:109) M , PC : PI : PI ( 4 , 5 ) P 2 , 48 : 48 : 4 ) , or both on actin polymerization ( 2 . 5 (cid:109) M actin and 60 nM Arp2 / 3 complex ) stimulated by N - WASP ( 200 nM ) . The solid line shows stimulation of actin polymerization under the same conditions by GST - VCA ( 200 nM ) alone . ( B ) Dose \u2013 response curves showing the maximum elongation rate of actin polymerization ( 2 . 5 (cid:109) M actin and 60 nM Arp2 / 3 complex ) as a function of the following molecules added : GST - VCA , full - length N - WASP , and full - length N - WASP stimulated with GTP (cid:103) S - Cdc42 ( 50 nM ) and PI ( 4 , 5 ) P 2 - containing vesicles ( 100 (cid:109) M , PC : PI : PI ( 4 , 5 ) P 2 , 48 : 48 : 4 ) . Note that the curve of N - WASP activated by PI ( 4 , 5 ) P 2 and GTP (cid:103) S - Cdc42 starts to taper off between 8 and 80 nM because the Cdc42 is at a relatively low concentration in the experiment ( 50 nM ) ; addition of higher concentrations of Cdc42 will drive activation to higher levels approaching those of GST - VCA ( Figure 6A ) . ( C ) Effects of Cdc42 bound to different nucleotides on N - WASP . Actin polymerization ( 2 . 5 (cid:109) M actin and 60 nM Arp2 / 3 complex ) was stimulated by N - WASP at the indicated concentrations in the presence of Cdc42 ( 50 nM , GTP (cid:103) S or GDP (cid:98) S charged ) and PI ( 4 , 5 ) P 2 - containing vesicles ( 10 (cid:109) M , PC : PI : PI ( 4 , 5 ) P 2 , 48 : 48 : 4 ) . ( D ) Actin polymerization ( 2 . 5 (cid:109) M actin and 60 nM Arp2 / 3 complex ) was stimulated by N - WASP ( 5 nM ) in the presence of Cdc42 ( 50 nM , GTP (cid:103) S charged ) and the indicated concentrations of lipid vesicles containing PC / PI ( 50 : 50 ) or PC / PI / PI ( 4 , 5 ) P 2 ( 48 : 48 : 4 ) . and full - length N - WASP is not due to the presence of and PH domains of N - WASP , respectively ( Figure 1A ) . Both domains are necessary forN - WASP to induce actin the GST domain fused to VCA . In the presence of 60 nM Arp2 / 3 and 2 . 5 (cid:109) M actin , 50 nM GST - VCA and 50 reorganization in fibroblasts ( Miki et al . , 1996 , 1998a ) . Addition of GTP (cid:103) S - charged Cdc42 had only a small nM VCA ( when cleaved from the GST moiety ) stimulated the elongation rate by 35 - fold and 25 - fold , respectively , effect on the activity of full - length N - WASP , increasing the elongation rate by (cid:122) 2 - fold and having no effect on compared to the 2 . 5 - fold stimulation obtained with a 4 - fold higher concentration ( 200 nM ) of full - length the duration of the lag phase ( Figure 6A ) . Likewise , lipid vesicles containing phosphatidylcholine ( PC ) , phospha - N - WASP . tidylinositol ( PI ) , and PI ( 4 , 5 ) P 2 in a 48 : 48 : 4 ratio failed to activate N - WASP ( Figure 6A ) . However , the addition of Cdc42 and Phosphatidylinositol ( 4 , 5 ) Bisphosphate Can Synergistically Activate N - WASP In Vitro both PI ( 4 , 5 ) P 2 and GTP (cid:103) S - Cdc42 dramatically stimu - lated actin assembly ( Figure 6A ) , an effect that de - Theabovedatastronglysupport thehypothesisthatfull - length N - WASP is inhibited both in vitro and in Xenopus pended on N - WASP and the Arp2 / 3 complex ( data not shown ) . In fact , the ability of 200 nM N - WASP to activate extracts ( Miki et al . , 1998a ) . To identify potential up - stream regulators that may activate N - WASP in vitro , actin assembly in the presence of PI ( 4 , 5 ) P 2 and GTP (cid:103) S - Cdc42 was equal to that achieved by 200 nM GST - VCA we focused on Cdc42 and PI ( 4 , 5 ) P 2 , which have been implicated in actin assembly in Xenopus extracts ( Ma ( Figure 6A ) . To more precisely determine the activation potency of PI ( 4 , 5 ) P 2 and GTP (cid:103) S - Cdc42 , we compared et al . , 1998a ) . Cdc42 and PI ( 4 , 5 ) P 2 interact with the GBD Cell228 Figure 7 . Signal - DependentActinNucleation A speculative model for the mechanism by which signals , such as Cdc42 and lipids , may regulate actin assembly at membrane - proxi - mal sites by recruitment and activation of the Arp2 / 3 complex via N - WASP - like proteins . the elongation rate as a function of the concentration N - WASP and Actin Nucleation Although N - WASP has been implicated in many actin - of N - WASP , GST - VCA , or N - WASP activated by GTP (cid:103) S - Cdc42 and PI ( 4 , 5 ) P 2 ( Figure 6B ) . In the presence of dependent processes , such as filopodium formation ( Miki et al . , 1998a ) and the actin - based motility of Shi - Cdc42 and PI ( 4 , 5 ) P 2 , the potency of N - WASP was in - creased to a level approaching that of GST - VCA , sug - gella ( Suzuki et al . , 1998 ) , the mechanism by which it mediates actin polymerization has remained elusive . In gesting that N - WASP was fully activated by the two molecules . this report , we provide evidence that N - WASP stimu - lates actin nucleation , a kinetically unfavored step of The coordinate stimulation by Cdc42 and PI ( 4 , 5 ) P 2 favors GTP (cid:103) S - over GDP (cid:98) S - charged Cdc42 , but it is actin polymerization , by interacting with the Arp2 / 3 complex . not completely specific for the GTP form ( Figure 6C ) . Selectivity is most apparent under conditions where the An N - WASP fragment consisting of the C - terminal 55 amino acids ( CA region ) directly binds to the Arp2 / 3 concentration of N - WASP is below that of Cdc42 ( Figure 6C ) . At these low concentrations of N - WASP , the po - complex in vitro ( Figure 3 ) . This association is consistent with the observation that C - terminal segments of Scar1 / tency differenceof GTP (cid:103) S - overGDP (cid:98) S - charged Cdc42 is (cid:122) 5 - fold . Since the PH domain of N - WASP has been WAVE and WASP , both of which share homology with the corresponding region in N - WASP , can bind to the reported to bind specifically to PI ( 4 , 5 ) P 2 , we tested Arp2 / 3 complex ( Machesky and Insall , 1998 ) . However , whether the stimulation seen above is specific for a longer segment of N - WASP , including both the Arp2 / PI ( 4 , 5 ) P 2 - containing vesicles ( Miki et al . , 1996 ) . While 3 - binding CA region and the actin - binding V region , is control lipid vesicles containing only PC and PI ( 50 : 50 ) required for the activation of the Arp2 / 3 complex . The gave only an (cid:122) 2 - fold activation above the no lipid con - homologous VCA region of Scar1 / WAVE has recently trol , similarvesicles containing 4 % PI ( 4 , 5 ) P 2 can activate been shown to disrupt Arp2 / 3 localization and actin as - by (cid:46) 10 - fold ( Figure 6D ) . Finally , GTP (cid:103) S - charged Rho is sembly at the cell periphery when overexpressed in ineffective at mediating stimulation ( data not shown ) . Swiss 3T3cells ( Macheskyand Insall , 1998 ) . The authors Thus , the coordinate activation of N - WASP is Cdc42 and also observed an increase in diffuse perinuclear phalloi - PI ( 4 , 5 ) P 2 specific and selective for the GTP form of Cdc42 . din staining , consistent with our observation that the VCA region should both bind to and stimulate the Arp2 / 3 Discussion complex ( Machesky and Insall , 1998 ) . Disruption of pe - ripheral actin assembly is likely to be a consequence of The study of complex signal transduction networks competition for the Arp2 / 3 complex or actin or both . comprising many direct and indirect pathways requires The Listeria protein ActA can dramatically stimulate an integrated use of biochemical , genetic , and cell bio - the actin nucleation activity of the Arp2 / 3 complex , as logical approaches . Biochemical reconstitution experi - evidenced by a shortening of the lag phase and an in - ments can isolate the minimal set of components essen - crease in the elongation rate ( Welch et al . , 1998 ) . The tialfor biologicalfunctionandhelp elucidatethedetailed VCA region of N - WASP has a very similar effect on actin mechanism by which these components interact . Be - polymerization kinetics in the presence of the Arp2 / 3 cause small G proteins and phosphoinositides have complex , suggesting that N - WASP - like proteins might many targets in the cell ( Toker and Cantley , 1997 ; Van be the cellular counterparts of ActA ( Figures 2 and 3 ) . and D\u2019Souza , 1997 ) , the pathways that connect them to Other than activation of nucleation , these kinetic changes actin polymerization have been difficult to delineate . In in actin assembly could be produced by filament sev - this paper , we have demonstrated one such pathway in ering or an increased rate of elongation . We can essen - vitro , using both a cell - free system and purified compo - tially rule out both these alternative explanations be - nents . WehaveshownthattheArp2 / 3complex , required causewefailedtodetectanysignificantseveringactivity for actin nucleation in a variety of systems , is regulated or any change in the elongation rate of actin filaments by N - WASP . This interaction mediates Cdc42 - induced in the presence of N - WASP , the Arp2 / 3 complex , or both actin polymerization in extracts and can be modulated ( data not shown ) . Although ActA and N - WASP share no significant sequence homology , they do share a short by both PI ( 4 , 5 ) P 2 and Cdc42 . N - WASP Interaction with the Arp2 / 3 Complex 229 segment of similarity in the cofilin homology domain , N - WASP ( Miki et al . , 1996 ) . We have previously demon - strated thatboth Cdc42and PI ( 4 , 5 ) P 2 canstimulate actin which is also present in WASP , Scar1 / WAVE , and Bee1 ( Bi and Zigmond , 1999 ) . The (cid:68) cof mutant , which is de - polymerization in Xenopus egg extracts ( Ma et al . , 1998a ) . Our in vitro data suggests that PI ( 4 , 5 ) P 2 may fective in Arp2 / 3 activation ( Figure 2C ) , contains a four \u2013 amino acid deletion ( (cid:68) 473 KRSK 476 ) within this segment . play an important role in N - WASP activation . HowisN - WASPregulatedby both Cdc42andPI ( 4 , 5 ) P 2 ? How do N - WASPand the Arp2 / 3 complexact together to stimulate actin nucleation ? Since the entire VCA do - Miki et al . have suggested that N - WASP is regulated by an allosteric mechanism ( 1998a ) . Full - length N - WASP main of N - WASP is required for its ability to stimulate actin polymerization , we favor a model ( Machesky and may exist in two conformations that are in equilibrium : an inactive conformation , in which the VCA domain is Insall , 1998 ) in which actin monomers and the Arp2 / 3 complex could be brought into proximity as a result of masked , and an active one in which the VCA domain is exposed and able to interact with the Arp2 / 3 complex . their interactions with the V and CA regions , respectively ( Figure 7 ) . In fact , the V domain is required for the ability At rest , N - WASP largely populates the inactive confor - mation . This is supported by two observations : ( 1 ) full - of N - WASP to remodel the actin cytoskeleton when overexpressed in cells ( Miki and Takenawa , 1998 ) . lengthN - WASPonlymodestlyactivatestheArp2 / 3com - plex ( in comparison to the isolated VCA fragment ) ; and ( 2 ) the full - length protein does not bind to the Arp2 / 3 Cdc42 Signaling and the N - WASP \u2013 Arp2 / 3 complex under conditions where the association be - Complex Interaction tween its VCA fragment and Arp2 / 3 is readily detectable . Our results also shed light on the detailed mechanism However , since full - length N - WASP is completely inac - by which Cdc42 regulates actin polymerization . We pre - tive when added to extracts , we believe that other trans - viously demonstrated a requirement for the Arp2 / 3 com - acting factors may further suppress even the modest plex in Cdc42 - induced actin assembly ( Ma et al . , 1998b ) . activity we have observed with the purified protein in Here , we extend this work to show that N - WASP relays vitro . Indeed , the majority of N - WASP from bovine brain Cdc42 signals and directly activates the Arp2 / 3 complex extracts is present in high molecular weight complexes in Xenopus extracts and in vitro with purified compo - ( unpublished observation ) . In any case , binding of up - nents . Most importantly , using purified N - WASP and stream regulators , such as Cdc42 and PI ( 4 , 5 ) P 2 , to the purified Arp2 / 3 complex , we find that actin polymeriza - N terminus would activate N - WASP by stabilizing the tion can be coordinately stimulated by Cdc42 and active conformation of the molecule ( Figure 7 ) . In addi - PI ( 4 , 5 ) P 2 ( Figure 6 ) . tion to Cdc42 and PI ( 4 , 5 ) P 2 , the N - WASP protein binds This novel finding strongly suggests that the Cdc42 to calmodulin and to SH3 domain \u2013 containing proteins signaling pathway represents one cellular mechanism and may be regulated in other ways ( Miki et al . , 1996 ) . that stimulates actin polymerization by regulating the Unmasked in this active conformation , the VCA domain nucleation activity of the Arp2 / 3 complex ( Ma et al . , would promote nucleation by recruiting Arp2 / 3 via its 1998b ) . The intracellular bacteria Shigella and Listeria CA region and G - actin via its V region . These speculative appear to hijack the cellular machinery for actin assem - models of N - WASP activation and filament nucleation bly by intersecting this pathway at different points . The ( summarized in Figure 7 ) await rigorous testing using ActA protein of Listeria bypasses N - WASP and interacts kinetic analysis as well as by higher resolution mutagen - with the Arp2 / 3 complex directly ( Welch et al . , 1998 ) , esis . However , the results presented here establish whereas the VirG protein of Shigella functions by re - N - WASP and the Arp2 / 3 complex as a core for signal - cruiting N - WASP ( Suzuki et al . , 1998 ) . However , the regulated actin assembly . Cdc42 \u2013 N - WASP pathway probably represents only one of several cellular mechanisms that control actin assem - bly via the Arp2 / 3 complex . N - WASP belongs to a family Experimental Procedures of proteins , including WASP , Scar1 / WAVE ( Bear et al . , Preparation of Xenopus Egg Extracts 1998 ; Miki et al . , 1998b ) , and Bee1 ( Li , 1997 ) , which and Phospholipid Vesicles share sequence homology at their C termini . Since the Xenopus HSSwaspreparedasdescribedpreviously ( Maetal . , 1998b ) . N termini of these proteins are quite different ( Scar1 / Phospholipid vesicles containing PC : PI ( 50 : 50 ) or PC : PI : PI ( 4 , 5 ) P 2 WAVE and Bee1 do not have the GBD or PH domains ) , ( 48 : 48 : 4 ) were prepared as described previously ( Ma et al . , 1998a ) they may transduce diverse upstream signals to a com - with the following modifications . Chloroform - dissolved phospholip - ids ( PC , PI from Avanti Polar Lipids [ Alabaster , AL ] ; PI [ 4 , 5 ] P 2 from moneffector , theArp2 / 3 complex , throughtheirhomolo - Calbiochem ) were mixed in the appropriate ratios and dried under gous C termini . More generally , our results firmly estab - nitrogen . The dried lipid mixture was resuspended in lipid buffer ( 10 lish that de novo actin nucleation is a step at which mMHEPES [ pH7 . 7 ] , 100mMNaCl , 5mMEGTA , and50mMsucrose ) signaling pathways can regulate the actin cytoskeleton . to a final concentration of 10 mM and then extruded through a 100 nm pore polycarbonate filter ( Avanti ) using the Mini - Extruder . Signal - Dependent Actin Polymerization In Vitro The difference in potency between the full - length Preparation of Recombinant Proteins The GST - Cdc42 and GST - Rho fusion proteins were prepared and N - WASP and its C - terminal VCA fragment both in vitro loaded with different nucleotides ( GTP (cid:103) S or GDP (cid:98) S ) while still and in extracts suggests that N - WASP\u2019s activity is sup - bound to glutathione - Sepharose beads according to established pressed by 2 \u2013 3 orders of magnitude . However , using procedures ( Ma et al . , 1998a ) . All excess unbound nucleotide was purified components , we can maximally activate full - washed away , and the G proteins were eluted in 20 mM HEPES ( pH length N - WASP in the presence of two signaling mole - 7 . 6 ) , 5 mM MgCl 2 , 100 mM NaCl , 0 . 1 % sodium cholate , and 10 mM cules , Cdc42 and PI ( 4 , 5 ) P 2 , that have previously been glutathione . Recombinant forms of full - length rat N - WASP , the H208D mutant , shown to interact with the GBD and PH domains in Cell230 and the (cid:68) cof mutant ( deletion of amino acids 473 \u2013 476 ) were pre - were eluted with SDS sample buffer . G - actin pull downs were per - formed in the same manner except that G - actin was used at a pared according to a published protocol , with modifications ( Miki et al . , 1998a ) . Proteins were expressed in insect ( Sf9 ) cells and concentration of 0 . 05 (cid:109) M , below the barbed end critical concen - tration . purified sequentially over heparin , Q - Sepharose , and Superdex 200 columns . GST fusion proteins containing various C - terminal regions of bovine N - WASP ( shown in Figure 3 ) were prepared as described Actin Polymerization Assays previously ( Miki et al . , 1996 ) . The VCA fragment was cleaved from Pyrene actin was used to follow actin polymerization in Xenopus GST - VCA on glutathione - Sepharose beads using Factor Xa ( Phar - extracts as described previously ( Ma et al . , 1998b ) . Polymerization macia ) , and the Factor Xa was separated from VCA using Benzami - wasinitiatedbyaddingGTP (cid:103) S - charged GST - Cdc42ataconcentra - dine Sepharose ( Pharmacia ) . tion of 250 nM ( unless otherwise noted ) . All proteins were quantitated by the Bradford assay ( Bio - Rad , To follow actin polymerization using purified components , pyrene Hercules , CA ) or by scanning densitometry of Gelcode Blue ( Coo - G - actin or unlabeled G - actin was isolated by incubating freshly massie G - 250 ) stained gels , using BSA as a standard in both cases . thawed proteins in G buffer ( 5 mM Tris - HCl [ pH 8 . 0 ] , 0 . 2 mM CaCl 2 , Proteins were stored at concentrations (cid:36) 1 mg / ml at (cid:50) 80 (cid:56) C and 0 . 2 mM ATP , 0 . 2 mM DTT ) for (cid:122) 10 hr at 4 (cid:56) C and then removing thawed and diluted immediately before use . residualF - actinbycentrifugationat400 , 000 (cid:51) gfor1hr . Polymeriza - tion reactions contained 1 . 5 (cid:109) M unlabeled actin , 1 (cid:109) M pyrene actin Preparation and Purification of Antibodies ( 50 % labeled ) , 0 . 2 mM ATP , and various proteins in 80 (cid:109) l of XB . All Affinity - purified peptide antibodies against Arp3 and p34ARC were reaction components except the actin were first mixed together in kindly provided by Matt Welch ( University of California , San Fran - XB and preincubated for 5 min . The reaction was started by adding cisco , CA ) . The anti - actin monoclonal antibody was purchased a mixture of actin and pyrene actin to the preincubated protein mix , from Amersham ( Arlington Heights , IL ) and the anti - Glutathione and fluorescence changes were measured in a fluorometer . In all S - Transferase polyclonal antibody from Santa Cruz Biotech ( Santa the curves shown in the figures , actin polymerization was initiated Cruz , CA ) . at time (cid:53) 0 ; the curves are shifted along the abscissa to account Full - length , Sf9 - expressed rat N - WASP purified according to the for the delay time between the addition of actin and the first fluores - proceduresgivenabovewasusedtoraiseantiserainrabbits ( Zymed cence reading . Labs , SouthSanFrancisco , CA ) . Theantibodieswereaffinitypurified The curves shown in Figure 2B indicate an increase in the steady - essentially as described ( Harlow and Lane , 1988 ) . state pyrene fluorescence in the presence of both Arp2 / 3 and Anti - Arp2 serum was generated by injecting rabbits ( Zymed ) with N - WASP . This is not due to a change in filaments mass at steady a Maltose - binding protein ( MBP ) fusion of human Arp2 , which was state because the addition of phalloidin after the steady state is expressed in E . coli andpurified onan amylose affinitycolumn ( New reached did not change the signal ( data not shown ) . In cases where England Biolabs , Beverly , MA ) . The specificity of the antiserum was a clear steady state was attained , steady - state signals have been confirmedby itsability torecognize ( whenused forimmunoblotting ) normalized to avoid underestimating the slopes of the control the Arp2 subunit from the purified bovine Arp2 / 3 described below . curves . Purification of the Arp2 / 3 Complex from Bovine Brain Extracts Data Analysis The Arp2 / 3 complex was purified from bovine brain extracts by All kinetic analyses were performed using either the software pro - modifying a previously described protocol for the purification of vided with the fluorescence spectrometer ( SLM - Aminco ) or using Arp2 / 3 from Xenopus egg extracts ( Ma et al . , 1998b ) . High - speed Origin ( Microcal Software , Northampton , MA ) . Maximum elongation extracts from calf brain were sequentially fractionated on Butyl rates from the polymerization curves using the purified system have Sepharose , DEAE Sepharose , S Sepharose , Phenyl Sepharose , and been calculated by performing a linear regression on the linear , Superdex 200 columns , and the Arp2 / 3 complex was followed by elongation phase of the actin assembly curve . For pyrene assays immunoblottingwithanti - Arp3oranti - p34ARC . Thebovinecomplex in extracts , the initial rate and the maximum F - actin were calculated runs as a symmetric peak at the expected ( (cid:122) 200 kDa ) molecular as described previously ( Ma et al . , 1998a ) . In all the curves shown weight on a gel filtration column and contains all seven polypeptide in the figures , solid lines are drawn through all the data points subunits ( Figure 2A ) . The identities of four of these polypeptides collected ; a symbol is only placed on every fifth or tenth point for ( Arp2 , Arp3 , p41 - ARC , and p34 - ARC ) were confirmed by immu - clarity . All data shown in the figures were taken from experiments noblotting . The bovine complex is also functional because it can performed at least twice . When normalized to controls , data varied support Cdc42 - induced actin assembly when substituted for a Xen - by (cid:35) 20 % between independent experiments . opus Arp2 / 3 - containing fraction ( Ma et al . , 1998b ) . Immunodepletion Acknowledgments For immunodepletion of N - WASP , high - speed supernatants of Xen - opus egg extracts were mixed with affinity - purified anti - N - WASP We thank Jeff Peterson for his help in the purification of the Arp2 / 3 antibodies at a concentration of 75 (cid:109) g / ml and incubated on ice for complex from bovine brain extracts and for his preparation of lipid 2 hr . The immune complexes were captured by adding one - tenth vesicles . We thank Teresita Bernal , Louise Evans , and Ann Georgi volume of packed protein - A Affiprep beads ( Bio - Rad ) and gently for their help in expressing recombinant proteins in Sf9 cells . We rockingat 4 (cid:56) Cfor 2hr . Mockdepletions wereperformed inthe same thank Tim Mitchison and Phil Allen for many helpful discussions way except that antibody buffer ( PBS (cid:49) 40 % glycerol ) was used in - and Matt Welch for providing antibodies against Arp3 and p34ARC . stead of the anti - N - WASP . The efficacy of the depletion was con - We also thank Jeff Peterson , Susannah Rankin , and Todd Stuken - firmed to be (cid:36) 95 % by immunoblotting with the anti - N - WASP an - berg for comments on the manuscript . R . R . is a member of the tibody . Medical Scientist Training Program at Harvard Medical School . This work was supported in part by grants from the National Institutes of Health to M . W . K . ( GM26875 ) and T . K . ( 5PO1 HL 59561 - 02 ) . Binding Assays GST pull - down assays were performed by immobilizing equal amounts of the GST fusion proteins on glutathione - Sepharose Received February 25 , 1999 ; revised March 24 , 1999 . beads ( Pharmacia ) at a concentration of approximately 5 mg / ml packed beads . For the Arp2 / 3 complex pull downs , (cid:122) 5 (cid:109) l of beads were incubated with 200 (cid:109) l of bovine Arp2 / 3 complex ( 0 . 8 (cid:109) M ) in References XB ( 10 mM HEPES [ pH 7 . 6 ] , 100 mM KCl , 1 mM MgCl 2 , 0 . 1 mM EDTA , 1 mM DTT ) plus 0 . 1 % ( v / v ) Tween - 20 , 0 . 2 mM ATP , and 0 . 5 Arber , S . , Barbayannis , F . A . , Hanser , H . , Schneider , C . , Stanyon , C . A . , Bernard , O . , and Caroni , P . ( 1998 ) . Regulation of actin dynam - mg / ml chicken albumin ( binding buffer ) . After gently rocking for (cid:122) 1 hr at 4 (cid:56) C , the beads were washed two times in 100 (cid:109) l of XB plus ics through phosphorylation of cofilin by LIM - kinase . Nature 393 , 805 \u2013 809 . 0 . 1 % ( v / v ) Tween - 20 and 0 . 2 mM ATP . Proteins bound to the beads N - WASP Interaction with the Arp2 / 3 Complex 231 Bear , J . E . , Rawls , J . F . , and Saxe , C . L . , III ( 1998 ) . SCAR , a WASP - regulates the assembly of focal adhesions and actin stress fibers in response to growth factors . Cell 70 , 389 \u2013 399 . related protein , isolated as a suppressor of receptor defects in late Dictyostelium development . J . Cell Biol . 142 , 1325 \u2013 1335 . Ridley , A . J . , Paterson , H . F . , Johnston , C . L . , Diekmann , D . , and Hall , A . ( 1992 ) . The small GTP - binding protein rac regulates growth fac - Beckerle , M . C . ( 1998 ) . Spatial control of actin filament assembly : lessons from Listeria . Cell 95 , 741 \u2013 748 . tor - induced membrane ruffling . Cell 70 , 401 \u2013 410 . Suzuki , T . , Miki , H . , Takenawa , T . , and Sasakawa , C . ( 1998 ) . Neural Bi , E . , and Zigmond , S . H . ( 1999 ) . Where the WASP stings . Curr . Biol . 9 , R160 \u2013 R163 . Wiskott - Aldrich syndrome protein is implicated in the actin - based motility of Shigella flexneri . EMBO J . 17 , 2767 \u2013 2776 . Drubin , D . G . , and Nelson , W . J . ( 1996 ) . Origins of cell polarity . Cell 84 , 335 \u2013 344 . Symons , M . , Derry , J . M . , Karlak , B . , Jiang , S . , Lemahieu , V . , Mccor - mick , F . , Francke , U . , and Abo , A . ( 1996 ) . Wiskott - Aldrich syndrome Field , C . , Li , R . , and Oegema , K . ( 1999 ) . Cytokinesis in eukaryotes : protein , a novel effector for the GTPase CDC42Hs , is implicated in a mechanistic comparison . Curr . Opin . Cell Biol . 11 , 68 \u2013 80 . actin polymerization . Cell 84 , 723 \u2013 734 . Fukuoka , M . , Miki , H . , and Takenawa , T . ( 1997 ) . Identification of Toker , A . , andCantley , L . C . ( 1997 ) . Signallingthrough thelipid prod - N - WASP homologs in human and rat brain . Gene 196 , 43 \u2013 48 . ucts of phosphoinositide - 3 - OH kinase . Nature 387 , 673 \u2013 676 . Hall , A . ( 1998 ) . Rho GTPases and the actin cytoskeleton . Science Van , A . L . , and D\u2019Souza , S . C . ( 1997 ) . Rho GTPases and signaling 279 , 509 \u2013 514 . networks . Genes Dev . 11 , 2295 \u2013 2322 . Harlow , E . , and Lane , D . ( 1988 ) . Antibodies : A Laboratory Manual Welch , M . D . , DePace , A . H . , Verma , S . , Iwamatsu , A . , and Mitchison , ( Cold Spring Harbor , NY : Cold Spring Harbor Press ) , pp . 283 \u2013 318 . T . J . ( 1997a ) . The human Arp2 / 3 complex is composed of evolution - Hartwig , J . H . , Bokoch , G . M . , Carpenter , C . L . , Janmey , P . A . , Taylor , arily conserved subunits and is localized to cellular regions of dy - L . A . , Toker , A . , and Stossel , T . P . ( 1995 ) . Thrombin receptor ligation namic actin filament assembly . J . Cell Biol . 138 , 375 \u2013 384 . and activated Rac uncap actin filament barbed ends through phos - Welch , M . D . , Iwamatsu , A . , and Mitchison , T . J . ( 1997b ) . Actin poly - phoinositide synthesis in permeabilized human platelets . Cell 82 , merization is induced by Arp2 / 3 protein complex at the surface of 643 \u2013 653 . Listeria monocytogenes . Nature 385 , 265 \u2013 269 . Johnson , D . I . ( 1999 ) . Cdc42 : an essential Rho - type GTPase control - Welch , M . D . , Rosenblatt , J . , Skoble , J . , Portnoy , D . A . , andMitchison , ling eukaryotic cell polarity . Microbiol . Mol . Biol . Rev . 63 , 54 \u2013 105 . T . J . ( 1998 ) . Interaction of human Arp2 / 3 complex and the Listeria Kelleher , J . F . , Atkinson , S . J . , and Pollard , T . D . ( 1995 ) . Sequences , monocytogenes ActA protein in actin filament nucleation . Science structural models , and cellular localization of the actin - related pro - 281 , 105 \u2013 108 . teins Arp2 and Arp3 from Acanthamoeba . J . Cell Biol . 131 , 385 \u2013 397 . Winter , D . , Podtelejnikov , A . V . , Mann , M . , and Li , R . ( 1997 ) . The Li , R . ( 1997 ) . Bee1 , ayeastprotein withhomology toWiscott - Aldrich complexcontainingactin - relatedproteinsArp2andArp3isrequired syndrome protein , is critical for the assembly of cortical actin cy - for the motility and integrity of yeast actin patches . Curr . Biol . 7 , toskeleton . J . Cell Biol . 136 , 649 \u2013 658 . 519 \u2013 529 . Ma , L . , Cantley , L . C . , Janmey , P . A . , and Kirschner , M . W . ( 1998a ) . Yang , N . , Higuchi , O . , Ohashi , K . , Nagata , K . , Wada , A . , Kangawa , Corequirement of specific phosphoinositides and small GTP - bind - K . , Nishida , E . , and Mizuno , K . ( 1998 ) . Cofilin phosphorylation by ing protein Cdc42 in inducing actin assembly in Xenopus egg ex - LIM - kinase 1 and its role in Rac - mediated actin reorganization . Na - tracts . J . Cell Biol . 140 , 1125 \u2013 1136 . ture 393 , 809 \u2013 812 . Ma , L . , Rohatgi , R . , and Kirschner , M . W . ( 1998b ) . The Arp2 / 3 com - Zigmond , S . H . ( 1998 ) . Actin cytoskeleton : the Arp2 / 3 complex gets plex mediates actin polymerization induced by the small GTP - bind - to the point . Curr . Biol . 8 , R654 \u2013 R657 . ing protein Cdc42 . Proc . Natl . Acad . Sci . USA 95 , 15362 \u2013 15367 . Machesky , L . M . , and Gould , K . L . ( 1999 ) . The Arp2 / 3 complex : a multifunctional actin organizer . Curr . Opin . Cell Biol . 11 , 117 \u2013 121 . Machesky , L . M . , and Insall , R . H . ( 1998 ) . Scar1 and the related Wiskott - Aldrich syndrome protein , WASP , regulate the actin cy - toskeleton through the Arp2 / 3 complex . Curr . Biol . 8 , 1347 \u2013 1356 . Machesky , L . M . , Atkinson , S . J . , Ampe , C . , Vandekerckhove , J . , and Pollard , T . D . ( 1994 ) . Purification of a cortical complex containing two unconventional actins from Acanthamoeba by affinity chroma - tography on profilin - agarose . J . Cell Biol . 127 , 107 \u2013 115 . Miki , H . , and Takenawa , T . ( 1998 ) . Direct binding of the verprolin - homology domain in N - WASP to actin is essential for cytoskeletal reorganization . Biochem . Biophys . Res . Commun . 243 , 73 \u2013 78 . Miki , H . , Miura , K . , and Takenawa , T . ( 1996 ) . N - WASP , a novel actin - depolymerizing protein , regulates the cortical cytoskeletal re - arrangement in a PIP2 - dependent manner downstream of tyrosine kinases . EMBO J . 15 , 5326 \u2013 5335 . Miki , H . , Sasaki , T . , Takai , Y . , and Takenawa , T . ( 1998a ) . Induction of filopodium formation by a WASP - related actin - depolymerizing protein N - WASP . Nature 391 , 93 \u2013 96 . Miki , H . , Suetsugu , S . , and Takenawa , T . ( 1998b ) . WAVE , a novel WASP - family protein involved in actin reorganization induced by Rac . EMBO J . 17 , 6932 \u2013 6941 . Mitchison , T . J . , and Cramer , L . P . ( 1996 ) . Actin - based cell motility and cell locomotion . Cell 84 , 371 \u2013 379 . Mullins , R . D . , Heuser , J . A . , and Pollard , T . D . ( 1998 ) . The interaction of Arp2 / 3 complex with actin : nucleation , high affinity pointed end capping , and formation of branching networks of filaments . Proc . Natl . Acad . Sci . USA 95 , 6181 \u2013 6186 . Nobes , C . D . , Hawkins , P . , Stephens , L . , and Hall , A . ( 1995 ) . Activa - tion of the small GTP - binding proteins rho and rac by growth factor receptors . J . Cell Sci . 108 , 225 \u2013 233 . Ridley , A . J . , and Hall , A . ( 1992 ) . The small GTP - binding protein rho",
    "serwas2022mechanistic": "Article Mechanistic insights into actin force generation during vesicle formation from cryo - electron tomography Graphical abstract Highlights d Native - state description of force - producing actin networks during endocytosis d Branched actin \ufb01lament assembly is triggered from multiple mother \ufb01laments d Actin force production is robust despite considerable network variability d Filament anchorage points are key to pulling force generation and ef\ufb01ciency Authors Daniel Serwas , Matthew Akamatsu , Amir Moayed , . . . , Karen M . Davies , Padmini Rangamani , David G . Drubin Correspondence daniel . serwas @ berkeley . edu ( D . S . ) , drubin @ berkeley . edu ( D . G . D . ) In brief Actin \ufb01lament polymerization generates forces essential for numerous cellular processes including vesicle formation , cell motility , and cytokinesis . Serwas et al . combined cryo - electron tomography of intact mammalian cells with mathematical modeling to gain mechanistic insights into how actin assembly forces pull on the plasma membrane to support endocytic vesicle formation . Serwas et al . , 2022 , Developmental Cell 57 , 1132 \u2013 1145 May 9 , 2022 \u00aa 2022 Elsevier Inc . https : / / doi . org / 10 . 1016 / j . devcel . 2022 . 04 . 012 ll Article Mechanistic insights into actin force generation during vesicle formation from cryo - electron tomography Daniel Serwas , 1 , * Matthew Akamatsu , 1 , 4 Amir Moayed , 1 Karthik Vegesna , 1 Ritvik Vasan , 2 , 5 Jennifer M . Hill , 1 Johannes Scho\u00a8neberg , 1 , 6 Karen M . Davies , 1 , 3 , 7 Padmini Rangamani , 2 and David G . Drubin 1 , 8 , * 1 Department of Molecular and Cell Biology , University of California , Berkeley , Berkeley , CA , USA 2 Department of Mechanical and Aerospace Engineering , University of California , San Diego , La Jolla , CA , USA 3 Molecular Biophysics and Integrative Bioimaging Division , Lawrence Berkeley National Laboratory , Berkeley , CA , USA 4 Present address : Department of Biology , University of Washington , Seattle , WA , USA 5 Present address : Allen Institute of Cell Science , Seattle , WA , USA 6 Present address : Department of Pharmacology and Department of Chemistry & Biochemistry , University of California , San Diego , La Jolla , CA , USA 7 Present address : Electron Bio - Imaging Centre , Diamond Light Source , Harwell Science and Innovation Campus , Didcot , UK 8 Lead contact * Correspondence : daniel . serwas @ berkeley . edu ( D . S . ) , drubin @ berkeley . edu ( D . G . D . ) https : / / doi . org / 10 . 1016 / j . devcel . 2022 . 04 . 012 SUMMARY Actin assembly provides force for a multitude of cellular processes . Compared to actin - assembly - based force production during cell migration , relatively little is understood about how actin assembly generates pulling forces for vesicle formation . Here , cryo - electron tomography identi\ufb01ed actin \ufb01lament number , orga - nization , and orientation during clathrin - mediated endocytosis in human SK - MEL - 2 cells , showing that force generation is robust despite variance in network organization . Actin dynamics simulations incorporating a measured branch angle indicate that suf\ufb01cient force to drive membrane internalization is generated through polymerization and that assembly is triggered from (cid:1) 4 founding \u2018\u2018mother\u2019\u2019 \ufb01laments , consistent with tomography data . Hip1R actin \ufb01lament anchoring points are present along the entire endocytic invagi - nation , where simulations show that it is key to pulling force generation , and along the neck , where it targets \ufb01lament growth and makes internalization more robust . Actin organization described here allowed direct translation of structure to mechanism with broad implications for other actin - driven processes . INTRODUCTION Actin \ufb01laments are structurally polarized linear polymers that preferentially grow at one end , called the plus or barbed end ( Pollard , 2016 ) . Polymerization of individual actin \ufb01laments can produce forces in the range of 1 \u2013 9 pN ( Dmitrieff and Nedelec , 2016 ) . These \ufb01laments can organize into higher order assemblies that facilitate a multitude of essential cellular functions including clathrin - mediated endocytosis ( CME ) ( Rottner et al . , 2017 ) . During CME , the plasma membrane is deformed to produce cargo - containing clathrin - coated vesicles ( CCVs ) . This mem - brane remodeling is promoted by assembly of a clathrin - contain - ing protein coat and forces provided by the actin cytoskeleton ( Kaksonen and Roux , 2018 ) . Although how a growing actin network can push on a cellular membrane , for example , during cell migration , is well understood , how assembly can aid in mem - brane pulling during endocytosis and intracellular traf\ufb01cking is much less well understood . CME is well suited to studies of membrane pulling through actin assembly as nearly complete lists of the components involved and detailed information on their dynamics exist ( Kaksonen and Roux , 2018 ) . However , mo - lecular - scale positional information about these components in their native state , which is essential for attaining a mechanistic understanding of their activities , is lacking . During CME , clathrin coat assembly initiation is followed by recruitment of the actin \ufb01lament nucleating Arp2 / 3 complex ( Tay - lor et al . , 2011 ) . This complex can bind to the sides of existing \u2018\u2018mother\u2019\u2019 \ufb01laments to induce assembly of new \u2018\u2018daughter\u2019\u2019 \ufb01la - ments , leading to formation of branched actin networks ( Rottner et al . , 2017 ) . The clathrin and plasma membrane - binding coat protein Hip1R can tether actin \ufb01laments to CME sites to harness \ufb01lament polymerization forces for plasma membrane deforma - tion ( Akamatsu et al . , 2020 ; Engqvist - Goldstein et al . , 1999 , 2001 ) . Using agent - based models , we previously found that actin self - assembles into a branched network during CME , with \ufb01la - ments oriented orthogonal to the plasma membrane , surround - ing , and attached to , the clathrin coat , and their growing plus ends oriented toward the plasma membrane . This geometry was dependent on the experimentally constrained spatial distri - bution of activated Arp2 / 3 complexes and actin \ufb01lament - binding 1132 Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 \u00aa 2022 Elsevier Inc . ll Hip1R linkers embedded in the clathrin coat ( Akamatsu et al . , 2020 ; Mund et al . , 2018 ; Sochacki et al . , 2017 ) . The resulting network geometry was consistent with previously proposed models and predicted the range for the numbers of \ufb01laments that could be involved in force generation in CME ( Akamatsu et al . , 2020 ; Kaksonen et al . , 2006 ) . In contrast , platinum replica electron microscopy ( EM ) of unroofed mammalian cells showed branched actin \ufb01laments surrounding only the neck region of CME sites in a collar - like fashion with \ufb01laments aligned parallel to the plasma membrane ( Collins et al . , 2011 ) . However , this method , likeotherclassicEMmethods , mightnotpreservenative actin cytoskeleton organization as it might result in partial removal of some of the actin network during unroo\ufb01ng and might not allow entire actin \ufb01laments in dense networks to be traced ( Collins et al . , 2011 ; Maupin - Szamier and Pollard , 1978 ; Resch et al . , 2002 ; Small , 1981 ) . In addition , in silico experiments sug - gest that both parallel and orthogonal \ufb01lament arrangements can facilitate CME progression ( Hassinger et al . , 2017 ) . These predictions emphasize the need to determine the orientation and numbers of actin \ufb01laments at CME sites before a quantitative understandingof theprecisemechanism ofactin - mediated force generationduringCMEcanbeachieved . TheplatinumreplicaEM study further suggested that the branched actin networks origi - natefromasinglefoundingmother\ufb01lament , butitsoriginremains unclear ( Collins et al . , 2011 ) . In addition to the orientation of branched actin \ufb01laments and the origin of mother \ufb01laments , the precise localization of the critical actin - CME linker Hip1R is ambiguous since partially contradicting data have been pub - lished ( Clarke and Royle , 2018 ; Engqvist - Goldstein et al . , 2001 ; Sochacki et al . , 2017 ) . Obtaining the structural information described above requires a method that allows visualization of native three - dimensional ( 3D ) actin networks at the single - \ufb01la - ment level , determination of each \ufb01lament\u2019s orientation , and the precise localization of the aforementioned linker protein Hip1R . In recent years , in situ cryo - electron tomography ( cryo - ET ) has been shown to be an extremely powerful approach to visualize the 3D organization of cellular features , including actin networks and protein - coated vesicles , at near - native - state conditions with unprecedented detail ( Bykov et al . , 2017 ; F \u20ac a\u00dfler et al . , 2020 ; Jas - nin and Crevenna , 2016 ; Mahamid et al . , 2016 ; Vinzenz et al . , 2012 ) . Here , to elucidate the mechanism of pulling force gener - ation through actin network assembly , we integrated cryo - ET on organization of actin networks involved in mammalian cell CME with mathematical modeling . RESULTS Clathrin coat identi\ufb01cation in cryo - electron tomograms of intact cells Toinvestigatethestructural organization ofactinduring CME , we used cryo - ET of vitri\ufb01ed intact human - derived SK - MEL - 2 cells growing on EM grids ( Serwas and Davies , 2021 ) . We identi\ufb01ed honeycomb - like arrangements in our tomography data , which were reminiscent of clathrin coats seen in previous EM studies ( Figure 1A ) ( Avinoam et al . , 2015 ; Cheng et al . , 2007 ; Heuser , 1980 ) . To test whether these arrangements are indeed clathrin coats , we applied correlative cryo - \ufb02uorescence lightmicroscopy ( cryo - FLM ) andcryo - ETusingSK - MEL - 2cellsthatendogenously express GFP - tagged clathrin light chain A ( CLTA - GFP ) and were incubated with \ufb02uorescent transferrin ( TF ) cargo to precisely pinpoint CME events ( Figures S1A and S1B ) . Tomograms ob - tained by this method showed the same structural features as in the randomly collected datasets ( Figure S1B ) . The randomly collected tomograms were of higher quality than the correlative data , likely due to fewer manual handling steps , and were there - fore used for the further analysis . Subtomogram averaging of clathrin coat vertices from a total of 8 CME sites and CCVs iden - ti\ufb01ed additional details of the native hub structure ( Figure 1B ; Video S1 ) . The resolution of our density map was 2 . 7 nm based on the 0 . 5 Fourier shell correlation criterion ( Figure 1C ) . Compu - tational \ufb01ttingoftherecently published structural model obtained by single particle cryo - EM ( PDB : 6SCT ; Morris et al . , 2019 ) into our in situ map resulted in a cross - correlation score of 0 . 94 ( Fig - ure 1D ) . Our structural analysis together with our correlative mi - croscopy results thus veri\ufb01es the identity of the protein coats in our tomograms as clathrin coats . Actin organization at different CME stages DuringCME , therelatively\ufb02atplasmamembraneinvaginatesand thenaconstrictedneckformsatthebaseoftheformedpitpriorto CCV scission ( Avinoam et al . , 2015 ; Roth and Porter , 1964 ) . We classi\ufb01ed the clathrin structures in our tomograms according to the shape of their underlying membrane as early \ufb02at , early , and late invaginated clathrin - coated pits ( CCPs ) and as CCVs ( Figures 2A , 2B , S2A , and S2B ; Videos S2 , S3 , S4 , and S5 ) . CCVs had a mean membrane and a clathrin coat diameter of 46 . 7 \u00b1 7 . 5 and 93 . 4 \u00b1 4 . 2 nm , respectively , whereas the CCP mean membrane and coat diameter were 90 . 2 \u00b1 10 . 4 and 136 . 1 \u00b1 17 . 7 nm ( Figure S1C ) . We then generated segmentation models to visualize the spatial relationship between actin \ufb01la - ments , the membrane , and the clathrin coat in 3D ( Figures 2B and S2B ) . Our analysis showed that individual clathrin triskelia tended to be somewhat disconnected in early clathrin coats compared with late CME sites and CCVs , indicating \ufb02exibility of earlyclathrincoats , whichisconsistentwitharecentstudyonun - roofed , chemically \ufb01xed cells ( Sochacki et al . , 2021 ) . Using live cell microscopy , it was previously shown that actin assembly oc - curs at (cid:1) 87 % of CME events in SK - MEL - 2 cells ( Grassart et al . , 2014 ) . Of 13 clathrin - coated structures in our tomograms ( Table S1 ) , only two did not show actin \ufb01lament association , and these two were released CCVs . It is important to note that duetotechnicallimitationsincryo - ET , ouranalysiswasrestricted tothinperipheralcellregions . Perhaps , inthesecellregions , CME sites are more likely to assemble actin . It will be important to determine whether CME events in thicker cell regions are more likely to be free of actin \ufb01laments or if they have amounts of actin \ufb01laments that are below the detection limit of light microscopy . Actin branch junctions could clearly be identi\ufb01ed based on the presence of an arc - like density at the junctions , most likely repre - senting the Arp2 / 3 complex branch nucleator ( Figures 3A and S3A ) ( Rouiller et al . , 2008 ; Vinzenz et al . , 2012 ) . To our surprise , CME - associatedactin\ufb01lamentsdidnotconsistofbranchedactin \ufb01lamentsexclusivelyaspreviouslyproposed ( Collinsetal . , 2011 ) . Instead , we observed a mixture of branched and unbranched \ufb01l - aments at all stages of CME ( Figures 2B and S2B ) . While branched \ufb01laments accumulated directly adjacent to CME sites , unbranched \ufb01laments were distributed across the entire tomographic volume . Unbranched \ufb01laments were organized in ll Article Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 1133 bundles close to CME sites , in dense meshworks of cortical actin \ufb01laments covering and surrounding CME sites , or as separated individual \ufb01laments . Branched actin \ufb01laments appeared asym - metrically distributedaround CMEsites , as suggested previously ( Collins et al . , 2011 ; Yarar et al . , 2005 ) . Themechanicalproperties ofindividual actin \ufb01lamentsdepend ontheirlength , withshorter\ufb01lamentsbeingmorerigidthanlonger ones ( De La Cruz and Gardel , 2015 ) . Our previous simulations predicted that branched actin \ufb01laments grow to a median length of 64 nm during CME and that branched \ufb01lament length was exponentially distributed , which can be explained by stochastic capping ( Figure S2C ; Tables S2 and S3 ; Akamatsu et al . , 2020 ) . Here , we measured a median branched \ufb01lament length in tomo - grams of 59 nm , indicating that these \ufb01laments could have assembled over the course of a CME event ( Figure 2C ; Table S3 ) . This possibility is further supported by the relatively low number of branched \ufb01laments in the early - stage CME tomo - gram shown in Figures 2A and 2B . As predicted by our previous simulations , branched \ufb01lament length in our tomograms was exponentially distributed ( Table S2 ) . Unbranched \ufb01laments were longer and had a median length of 108 nm ( Figure 2C ; Table S3 ) . Some of the unbranched \ufb01laments likely represent \ufb01l - aments that existed before the onset of the CME event , for example , as part of the actin cortex . However , a subset of these unbranched \ufb01laments could have been newly polymerized and alsocontributedtoforcegenerationtosupportCMEprogression . CME is robust to variations in the actin \ufb01lament branch angle As described above , branch junctions were identi\ufb01ed based on the presence of an additional density , likely the Arp2 / 3 complex , at the connection between the mother and daughter \ufb01lament . Figure 1 . Native clathrin coat architecture and structure ( A ) TomographicslicesofaCCVatZ - positionsrelativetocentralslice ( Z = 0nm ) . Topandbottomsliceshowcharacteristiccage - likeclathrincoatarchitecturewith threeleg - like extensions emanating fromeachvertex ( yellowarrowheads ) . Yellowand whitecirclesincentralslice indicate vesicleandcoatedarea , respectively . ( B ) Density map of the clathrin coat determined within intact cells . ( C ) Resolution determination of the clathrin density map based on Fourier shell correlation ( FSC ) . ( D ) Structuralmodeloftheclathrinhub ( PDB : 6SCT , inrainbowcolors , fromMorrisetal . [ 2019 ] ) \ufb01ttedintosubtomogramaverageisosurfacemodelofcytoplasmic clathrin vertex . Scale bars , 50 nm in ( A ) and 10 nm in ( B ) and ( D ) . See also Figure S1 . ll Article 1134 Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 Figure 2 . CME - actin networks consist of branched and unbranched \ufb01laments ( A ) Tomographic slices at Z - positions relative to central slice ( Z = 0 nm ) of CME events at indicated stages . Clathrin coat and individual actin \ufb01laments are high - lighted with yellow and red arrowheads , respectively . ( legend continued on next page ) ll Article Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 1135 We used subtomogram averaging to further test whether the identi\ufb01ed junctions were indeed Arp2 / 3 complex - based branch junctions . A low - resolution density map was obtained , which was suf\ufb01cient to computationally \ufb01t the recently published in situ branch junction structure , indicating that the junctions were Arp2 / 3 complex based ( Figures S3B \u2013 S3D ; F \u20ac a\u00dfler et al . , 2020 ) . Branch angles in a range of about 60 (cid:3) \u2013 80 (cid:3) , obtained from various sample types and methods , have been reported previously for Arp2 / 3 complex - nucleated actin \ufb01lament net - works ( Blanchoin et al . , 2000 ; F \u20ac a\u00dfler et al . , 2020 ; Jasnin et al . , 2019 ; Mueller et al . , 2014 ; Mullins et al . , 1998 ; Rouiller et al . , 2008 ; Svitkina and Borisy , 1999 ; Vinzenz et al . , 2012 ) . The branch angle of the structure that we used to \ufb01t into our subtomogram average is 71 (cid:3) . However , this average value does not capture the range of branch angle values at CME sites . Therefore , we measured the angles of individual branch junctions . The average branch angle from these measurements was 68 (cid:3) \u00b1 9 (cid:3) , which is in close agreement with recently published in situ structures ( Fig - ure 3B ) ( F \u20ac a\u00dfler et al . , 2020 ; Jasnin et al . , 2019 ) . Physical limita - tions ( e . g . , limited tilt range due to sample holder geometry and sample thickness ) restrict electron tomography data recording , resulting in a wedge - shaped lack of information in 3D Fourier space ( Lu (cid:2) ci (cid:3) c et al . , 2013 ; Wan and Briggs , 2016 ) . The so - called \u2018\u2018missing wedge\u2019\u2019 results in anisotropic resolution such that the tomography data are elongated in the electron beam direction ( z direction ) ( Lu (cid:2) ci (cid:3) c et al . , 2013 ; Wan and Briggs , 2016 ) . To test whether the missing wedge strongly affected our branch angle measurements , we plotted the measured branch angles against the orientation of the branch junctions in our to - mography data ( Figure S3E ; see STAR Methods for details ) . Branch angle values were independent of branch orientation . We therefore concluded that within the precision of our measure - ments , our branch angle measurements were not affected by the missing wedge . The native branch angles were on average smaller and the junctions slightly stiffer ( indicated by the stan - dard deviation ) compared with the average branch angle of 78 (cid:3) \u00b1 16 (cid:3) used in our previous mathematical model ( Akamatsu et al . , 2020 ) . To test whether the smaller branch angles affect Figure 3 . CME is robust against branch angle variation ( A ) Tomographic slicesof an individual branch junction atZ - positions , left toright , relative tocentral slice ( Z = 0nm ) . Positions of mother , daughter \ufb01laments , and plus and minus ends are indicated . Arrow points to the arc - like density of the Arp2 / 3 complex . ( B ) Branch angle distribution measured in tomograms . Mean value and standard deviation are displayed . ( C ) Resultofsimulations showing theeffect ofbranchangle variationonthenumberofplusendsatthebase ofCME invaginations ( greenand orangecurves ) and the internalization rate ( cyan and purple curves ) . Mean and SD are shown . Scale bar , 5 nm in ( A ) . See also Figure S3 . ( B ) Segmentation models of the tomograms in ( A ) , bottom row shows branched actin \ufb01laments only . Color - coded legend describes elements shown in the models . ( C ) Filament length distribution of unbranched and branched \ufb01laments across all CME events shown in this publication . Median \ufb01lament length for unbranched and branched actin \ufb01laments are highlighted . Scale bars , 50 nm in ( A and B ) . See also Figure S2 . ll Article 1136 Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 the force production capability , we modi\ufb01ed our computational model of branched actin \ufb01lament assembly at CME sites to re\ufb02ect the newly measured branch angle ( Figure S3F ) . In this model , an endocytic pit is modeled as a solid hemispherical ob - ject that resists internalization elastically ( Hassinger et al . , 2017 ) . Activated Arp2 / 3 complex resides in a ring around the base of the pit ( Almeida - Souza et al . , 2018 ; Mund et al . , 2018 ) , whereas Hip1R linkers are embedded in the surface of the curved pit ( So - chacki et al . , 2017 ) . Linear \ufb01laments diffusing near the pit serve as mother \ufb01laments that , when in proximity to the active Arp2 / 3 complex , generate new actin \ufb01laments branching from the mother \ufb01lament ( Blanchoin et al . , 2000 ) . 3D stochastic simu - lations of the model showed that actin \ufb01laments self - assembles into a polarized branched network that internalizes the pit against physiological values of plasma membrane tension ( Aka - matsu et al . , 2020 ; Kaplan et al . , 2021 ; Nedelec and Foethke , 2007 ) . We found that for both ranges of branch angles , the num - ber of plus ends polymerizing against the base of the CME site remained the same ( 10 \u00b1 4 ) , leading to similar internalization rates ( Figure 3C ) . In light of this result , we next tested the effect of a wider range of branch angles ( 10 (cid:3) \u2013 170 (cid:3) ) on internalization ef\ufb01ciency . We found that endocytosis remained robust ( Figures S3G and S3H ) . Although the branch angle might affect some architectural features , we conclude that overall polymeri - zation - based force production for CME is robust against branch angle variation , which contrasts with results reported for simula - tions of lamellipodia formation where a branch angle of (cid:1) 70 (cid:3) \u2013 80 (cid:3) was predicted to be optimal ( Garner and Theriot , 2020 ) . Branched actin assembly is nucleated from multiple mother \ufb01laments Next , we tested the previously postulated notion that branched actin network assembly at CME sites is initiated from a single Figure 4 . Branched actin \ufb01laments are orga - nized into clusters ( A ) Simulation snapshot highlighting clustered or - ganization of branched actin \ufb01laments in mathe - matical model . Individual clusters are color coded . ( B ) Simulation of mean cluster number during inter - nalization . ( C ) Segmentation model of late CME site from Fig - ure 2A highlighting clustered organization of branched actin \ufb01laments . Individual clusters are color coded . ( D ) Number of clusters found at individual CME events . Mean and SD are shown in ( B ) and ( D ) . Scale bar , 50 nm in ( C ) . See also Figure S4 . \u2018\u2018founding\u2019\u2019 mother \ufb01lament ( Collins et al . , 2011 ) . Simulations using the conditions we de\ufb01ned for our model predicted that branched actin network assembly can be initiated from several \u2018\u2018founding\u2019\u2019 mother \ufb01l - aments , each giving rise to a distinct branchedactin\ufb01lamentcluster . Thenumber of clusters and number of \ufb01laments per cluster increased during CME progression to an average of 4 \u00b1 2 clusters with 49 \u00b1 21 \ufb01laments in each cluster ( Figures 4A , 4B , and S4A ) . The predicted clustered branched actin \ufb01lament orga - nizationwasconsistentwiththearrangementofbranchedactin\ufb01l - aments in the tomograms , where theaverage cluster numberwas 8 \u00b1 6 ( Figures 3C and 3D ) . However , individual clusters only con - sisted of an average of 2 . 2 \u00b1 0 . 5 \ufb01laments in the tomography data ( FigureS4B ) . Thediscrepancyinthe\ufb01lamentnumberperclusterin the model versus experiment can be understood as follows : although the model shows 49 \u00b1 21 \ufb01laments per cluster at the end of a CME event , the number of \ufb01laments at the base is only 10 \u00b1 4 , suggesting that not all \ufb01laments in the model contribute directly to force generation ( Figure 3C ) . Since the model only cap - tures select actin interactions , it is also possible that the excess \ufb01lament number per cluster in the model re\ufb02ects certain limita - tions in the model . For example , it has been shown previously that in lamellipodia individual branch junctions are separated by multiples of the actin helix repeat of 36 nm along individual \ufb01la - ments ( Vinzenz et al . , 2012 ) . In the current version of the model , the spacing between individual branches along a \ufb01lament cannot be controlled . As a result , branching occurred at a higher fre - quency , with a median inter - branch distance of 6 . 5 nm , than is possible in cells due to geometrical and crowding - related spatial restrictions ( Figure S4C ) . It will be important for future models to controlthemaximumspacingbetweenbranchesinordertobetter understand the relationship between local branch and cluster ge - ometryandoverallendocyticactinnetworkarchitecture . Theearly CME site in the tomogram in Figure 2 showed the lowest number of branched actin \ufb01lament clusters , suggesting that more branched actin clusters are initiated and assembled during the plasma membrane internalization phase . Actin \ufb01lament orientation at CME sites Next , we set out to analyze the polarity and orientation of the actin \ufb01laments in the tomograms to assess the direction of force ll Article Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 1137 production through polymerization . The polarity of branched \ufb01l - aments can be determined based on branch junction geometry ( Figure 3A ) ( Narita et al . , 2012 ) . To analyze the polarity of un - branched \ufb01laments , we adapted a previously published method , which is based on cross - correlation analysis of the \ufb01laments in tomograms against simulated reference \ufb01laments with known polarity ( Figures S5A \u2013 S5E ) ( Narita et al . , 2012 ) . We developed an analysis software package that allowed us to calculate and plot the orientation of the \ufb01laments relative to the normal vector of a simulated reference plane representing the position of the plasma membrane or individual CCVs . Orientation of the vector was de\ufb01ned such that 0 (cid:3) orientation indicates actin \ufb01lament plus end pointing toward the reference plane , and 180 (cid:3) orientation in - dicates actin \ufb01lament plus end pointing away from the plane . The average \ufb01lament orientation across all tomograms was 86 (cid:3) \u00b1 31 (cid:3) for unbranched and 81 (cid:3) \u00b1 36 (cid:3) for branched \ufb01laments , indicating that a large proportion of \ufb01laments were oriented parallel to the plasma membrane ( Figures S5F and S5G ) . Further examination of individual CME sites identi\ufb01ed two CME archetypes for branched actin \ufb01lament organization , one in which \ufb01laments were oriented orthogonal to the plasma membrane , similar to the predictions from our simulations , and a second in which \ufb01la - ments were oriented largely parallel to the plasma membrane , similar to what was observed in previous work using platinum replica EM ( Figures 5A \u2013 5D , S5H , and S5I ) ( Akamatsu et al . , 2020 ; Collins et al . , 2011 ) . On average , 11 \u00b1 8 branched \ufb01lament plus ends and 36 \u00b1 7 unbranched \ufb01lament plus ends were ori - ented toward the base of individual CME sites ( relative \ufb01lament orientation angle < 90 (cid:3) ) during the membrane internalization phase ( Figure 5E ) . In contrast to our simulations , we also observed up to two branched \ufb01lament plus ends pointing toward the neck of CME invaginations for both types of actin arrange - ments , where they could potentially produce a squeezing force to support neck constriction and scission ( Figures 5B , 5C , and 5E ) . We also found branched actin \ufb01laments oriented with their plus ends toward CCVs , suggesting a role for \ufb01lament polymer - ization in CCV transportation inside cells ( Figures S5J and S5K ) . Hip1R localizes to the CME neck , where it is predicted to increase internalization ef\ufb01ciency Our previous model identi\ufb01ed the localization and distribution of the actin - binding protein Hip1R as a main determinant for actin organization and function ( Akamatsu et al . , 2020 ) . There , we had assumed that Hip1R exclusively localizes to the tip of CME invaginations , based on Sochacki et al . ( 2017 ) , in which averaging of FLM data were used ( Akamatsu et al . , 2020 ; So - chacki et al . , 2017 ) . However , individual images from the work of Sochacki et al . ( 2017 ) and other published work show variable Hip1R localization , including at the neck of CME invaginations ( Clarke and Royle , 2018 ; Engqvist - Goldstein et al . , 2001 ; So - chacki et al . , 2017 ) . Importantly , only the methodology in the study of Clarke and Royle ( 2018 ) allowed visualization of the entire CME invagination , and it is likely that the other localization data gave an incomplete picture . Therefore , we asked if we could identify Hip1R in our tomograms and whether its localization might play a role in targeting actin \ufb01laments to the neck of CME sites . Hip1R forms dimers that by quick - freeze deep - etch EM appear as (cid:1) 60 nm rod - shaped densities with two globular actin - binding domains at one end and two globular mem - brane - binding domains at the other ( Engqvist - Goldstein et al . , 2001 ) . We identi\ufb01ed densities resembling Hip1R in size and structure over the invagination surface including the neck of CME invaginations as well as at the plasma membrane adjacent to the pit , which was consistent with the work from Clarke and Royle ( 2018 ) ( Figures 6A and S6A ) . To further test whether these densities could be Hip1R dimers , we subjected the putative cytoplasmic actin - binding domains to subtomogram averaging , yielding a low - resolution density map ( Figures 6B , S6B , and S6C ) . The size and shape of the density was suf\ufb01cient to house two actin - binding domains plus the dimerization domains of the homologous protein talin and to \ufb01t onto an actin \ufb01lament ( Fig - ure 6B ) ( Gingras et al . , 2008 ) . This analysis provides further sup - port for the likelihood that these densities are Hip1R dimers . We then tested in our model the consequence of Hip1R localization at the neck by conducting simulations , wherein Hip1R molecules were positioned along the neck surface in addition to the invag - ination tip . The number of \ufb01lament plus ends at the base ( 10 \u00b1 4 ) remained the same with or without additional Hip1R dimers at the neck . However , an additional 4 . 5 \u00b1 2 plus ends were now found in the neck region , which is similar to the branched \ufb01la - ment plus end number in our tomograms ( Figures 5E , 6C , S6D , and S6E ) . Strikingly , in our simulations , Hip1R neck localization not only directed \ufb01laments to the neck region but also strongly improved internalization ef\ufb01ciency , indicating that Hip1R at the neck can help with actin - mediated force generation ( Figure 6D ) . DISCUSSION Forces produced by actin \ufb01lament assembly are harnessed for many membrane remodeling processes , but much remains to be learned about the underlying assembly regulation and force - producing mechanisms . CME is a particularly attractive process for elucidating these mechanisms because it consists of a predictable series of morphological changes , each of which is coupled to recruitment of speci\ufb01c proteins . This work set out to de\ufb01ne the 3D organization of individual actin \ufb01laments at different CME stages because this information directly informs the force - producing mechanism . What follows is an attempt to synthesize our cryo - ET and simulation results into a harmonious model for how actin assembly forces are generated and har - nessed during CME . Our \ufb01nding that branched actin networks are organized in branched actin \ufb01lament clusters at individual CME sites supports the conclusion that these clusters originate from multiple found - ing mother \ufb01laments ( 4 \u00b1 2 in the simulations and 8 \u00b1 6 in the to - mograms ) , which is a different conclusion from the single mother \ufb01lament model that was suggested previously ( Collins et al . , 2011 ) . This clustered organization may provide \ufb02exibility for branched actin \ufb01laments to assemble in a crowded environment like the cell cortex , which also suggests that excluded volume ef - fects may affect branched actin network assembly ( Schreiber et al . , 2010 ) . Individual branch clusters consisted of an average of 2 . 2 \u00b1 0 . 5 \ufb01laments . Accordingly , most clusters in our tomograms were ar - rangements of a mother and one daughter \ufb01lament . This number of branch junctions seems low at \ufb01rst glance . As described above , cryo - ET data are affected by the missing wedge problem . We therefore cannot completely exclude the possibility that ll Article 1138 Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 Figure 5 . Actin \ufb01lament orientation at CME sites ( A ) Color code indicating \ufb01lament orientation relative to the normal ( 0 (cid:3) ) of a reference plane representing the plasma membrane . At 0 (cid:3) , \ufb01lament plus ends point toward the membrane . ( B and C ) Color - coded \ufb01lament orientation at two late - stage CME invaginations ( green ) , ( C ) shows a site toward the end of vesicle scission . Note the presence of \ufb01laments with their plus ends oriented toward the neck region ( red arrowheads ) . ( D ) Relative \ufb01lament to reference plane orientation of late CME sites 1 and 2 shown in ( B ) and ( C ) , respectively . ( E ) Number of \ufb01lament plus ends pointing toward base and neck region of CME sites . Mean and SD are shown in ( D ) and ( E ) . See also Figure S5 . ll Article Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 1139 some \ufb01laments were not detected in our tomograms ( i . e . , when they are parallel to the electron beam direction and are posi - tioned completely in the missing wedge ) . However , given that the branch number per individual mother \ufb01lament across multi - ple tomograms is very consistent , we are convinced that our data reliably represent the branched actin \ufb01lament organization at CME sites in cells . A possible biological explanation for the sparse number of branch junctions concerns the spacing of Arp2 / 3 branch junctions . Previous work showed that the inter - branch spacing along individual mother \ufb01laments in lamellipodia corresponds to multiples of the actin helix repeat of 36 nm ( Vin - zenz et al . , 2012 ) . The reason for this inter - branch spacing is not clear but might be because more branch junctions cannot be initiated due to structural or steric restrictions or it might be due to the interplay between \ufb01lament growth and Arp2 / 3 binding dynamics ( i . e . , the mother \ufb01lament grows faster than new Arp2 / 3 complexes can bind ) . The median branched actin \ufb01lament length in our data is 59 nm . Based on the work by Vinzenz et al . ( 2012 ) , we would not expect that more than one branch junction would form along such short \ufb01laments . The number of branched actin \ufb01laments and the number of branched actin \ufb01lament clusters were greater during the Figure 6 . Hip1R - like linkers observed at the neck of CME invaginations increase internalization ef\ufb01ciency ( A ) Tomographic slices of the regions indicated in the segmentation model of the tomogram shown in Figures 2A and 2B . Slices highlight putative Hip1R dimers ( cyanboxes ) thatareassociatedwiththecoatedmembraneoftheCMEinvagination ( greenarrowheads ) andtheirpositionsarehighlightedbycyanspheresinthe segmentation model correspond . Red arrowheads in slice 1 point at an actin \ufb01lament associated with a putative Hip1R dimer . ( B ) Superimposition of the surface model of the density map in ( Figure S6B ) with previously published atomic models of two actin - binding domains ( ABD ) ( PDB : 2JSW , Gingras et al . [ 2008 ] ) and two dimerization domains ( PDB : 2QDQ , Gingras et al . [ 2008 ] ) of the homologous protein talin . Model of actin \ufb01lament ( PDB : 6DJN , Chouand Pollard [ 2019 ] ) was added asa reference . Mathematical model readoutshowing\ufb01lament distribution between CME invagination base and neck in the presence or absence of Hip1R molecules at the neck . ( C ) Result of simulations showing the effect of Hip1R neck localization on the number of plus ends at the base ( orange and green curves ) and neck regions ( cyan and purple curves ) of CME invaginations . Mean and SD are shown . ( D ) Simulated internalization ef\ufb01ciency with and without Hip1R molecules at the neck . Mean and SD are shown . Scale bars , 50 nm in ( A ) and 5 nm in ( B ) . See also Figure S6 . ll Article 1140 Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 internalization phase compared with early stage , consistent with ongoing assembly ( Figures 2B and S2B ) . In addition , the branched actin \ufb01lament length distribution indicates that these \ufb01laments could have polymerized over the course of a CME event and are thus capable of providing assembly force ( Figures 2B , 2C , S2B , and S2C ; Akamatsu et al . , 2020 ) . Variation in branched actin \ufb01lament organization and in the number of clus - ters and \ufb01laments between individual CME sites likely re\ufb02ects the stochastic nature of \ufb01lament assembly as well as local adap - tive responses to variable conditions . High variability in actin network organization and density from one CME site to the next might result from such factors as variance in mother \ufb01la - ment number and orientation , Hip1R localization , active Arp2 / 3 complex , N - WASP position , and differences in plasma mem - brane tension ( this study ; Akamatsu et al . , 2020 ; Clarke and Royle , 2018 ; Engqvist - Goldstein et al . , 2001 ; Kaplan et al . , 2021 ; Sochacki et al . , 2017 ) . Work in the CME \ufb01eld has mainly focused on assembly of branched actin networks and did not take the presence of un - branched \ufb01laments at CME sites into account when investigating actin function . Also , unbranched actin \ufb01laments were not identi - \ufb01ed by the previous EM work , likely due to the inability to trace every individual \ufb01lament along its entire length by platinum replica EM ( Collins et al . , 2011 ) . Given the high number of un - branched actin \ufb01laments observed in close proximity to CME sites in our tomograms , two important questions arise : where do they come from and what are their functions ? The presence of unbranched \ufb01laments at early stages of CME suggests that they might represent pre - existing cortical actin \ufb01laments ( early - stage site in Figure 2 , CME invaginations in Figure S2 ) or \ufb01la - ments originating from other nearby actin structures like the bundle in the late - stage tomogram in Figure 2 . These dense cortical actin arrangements might represent a physical barrier that needs to be cleared by \ufb01lament severing , disassembly , or repositioning for CME progression . Shorter \ufb01laments might be ones that are diffusing through the cytoplasm ( Chen and Pollard , 2013 ; Aroush et al . , 2017 ) . Importantly , some of the unbranched \ufb01laments might be actively producing assembly forces at CME sites as well as acting as mother \ufb01laments for new branch forma - tion . This possibility is supported by the presence of the Arp2 / 3 nucleation promoting factor SPIN90 at CME sites . SPIN90 pro - motes nucleation of unbranched actin \ufb01laments and plays a role in epidermal growth factor receptor endocytosis ( Kim et al . , 2006 ; Luan et al . , 2018 ; Oh et al . , 2013 ) . In addition , un - branched actin \ufb01laments might dissociate from pre - existing actin structures that were involved in other processes and have undergone dynamic remodeling , representing an additional potential source of mother \ufb01laments required for Arp2 / 3 - medi - ated , SPIN90 - independent , branched actin \ufb01lament nucleation ( Pollard , 2007 ) . The abundance of potential sources of mother \ufb01l - aments and pathways for actin \ufb01lament assembly may allow for a safety net that allows actin assembly and adaptation to ensure robust CME under variable conditions within a cell . We acknowl - edge that not all of these unbranched \ufb01laments necessarily have a CME - speci\ufb01c function . Because our cryo - ET approach allowed individual actin \ufb01la - ments to be fully traced and oriented in the volume surrounding CME sites , we were able to use simulations to assess the force - producing capabilities of these networks . We found in our cryo - ET data that on average , during the invagination formation and neck constriction stages , 11 \u00b1 8 branched and 36 \u00b1 7 un - branched \ufb01lament plus ends were pointing toward the base of the CME sites ( relative angle < 90 (cid:3) ) , and up to two additional branched actin \ufb01lament plus ends were pointing toward the neck region ( Figure 5 ) . In the simulations , we found that 10 \u00b1 4 \ufb01lament plus ends assembled at the base of CME invaginations , which was suf\ufb01cient for successful internalization ( Figures 3C , 6C , and 6D ; Akamatsu et al . , 2020 ) . In previous simulations , the number of plus ends varied between 2 at low and 22 at high membrane tension conditions ( Akamatsu et al . , 2020 ) . The force requirement to support the transition from a U - shaped to an omega - shaped invagination with a constricted neck through actin polymerization was predicted to be < 1 pN when the force is applied in parallel to the plasma membrane and 15 pN when applied orthogonally ( Hassinger et al . , 2017 ) . Polymerization of individual actin \ufb01laments can provide between 1 and 9 pN of force ( Dmitrieff and Nedelec , 2016 ) . Despite the variability in actin network organization observed in the tomograms , our sim - ulations predicted that the internalization ef\ufb01ciency was robust under different conditions . This conclusion suggests that actin \ufb01lament assembly can harness the local variability and stochas - ticity at the \ufb01lament level to generate ef\ufb01cient , robust force - generating machineries at the network level for CME . The actin - binding linker Hip1R might provide a spatial constraint for actin assembly to ensure robust internalization . Here , we \ufb01nd that Hip1R is not only at the tip surface of the CME invagination but also at the neck region . In simulations , Hip1R neck localiza - tion directs \ufb01laments toward the neck constriction and improves internalization ef\ufb01ciency . Actin \ufb01lament growth toward the neck also supports a possible role for actin polymerization during vesicle scission . Our modeling approach currently does not allow us to test whether an effective neck constriction force can be produced by the observed actin arrangements . Previous work showed that the total force requirement for constriction can be less than 1 pN when the force is applied in a collar - like fashion . In principle , this force could be provided by a single polymerizing actin \ufb01lament ( Dmitrieff and Nedelec , 2016 ) . How - ever , additional experiments and modeling developments will be required to test under which cellular conditions an effective constriction force can be generated . The principles identi\ufb01ed here are expected to apply to other actin - driven processes where cellular membranes are being pushed , pulled , or squeezed ( Figure 7 ) . The \ufb01nding that branched actin \ufb01lament networks at individual CME sites are organized in multiple discrete clusters is similar to force - produc - ing branched actin assemblies in lamellipodia , which push the plasma membrane outward during cell migration ( Vinzenz et al . , 2012 ) . However , the presence of actin \ufb01lament anchoring points at CME sites allows conversion of pushing force into pull - ing and squeezing forces . Internalization ef\ufb01ciency and \ufb01lament orientation strongly depend on the distribution of these anchor points ( this study ; Akamatsu et al . , 2020 ) . The same mechanism might facilitate budding and \ufb01ssion at intracellular membranes , for example , during vesicle budding from the trans - Golgi , where Hip1R anchoring points are important , or endosomes where the actin - and lipid - binding protein moesin could mediate anchorage ( Carreno et al . , 2004 ; Fehon et al . , 2010 ; Muriel et al . , 2016 ) . Actin \ufb01laments in \ufb01lopodia are also anchored to the plasma ll Article Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 1141 membrane , which might be important for \ufb01lopodia formation and maintenance ( Medalia et al . , 2007 ) . However , instead of being pulled inward , the plasma membrane is pushed outward . Be - sides of the position of anchoring points , we identi\ufb01ed the posi - tion of actin \ufb01lament assembly factors as a second constraint that de\ufb01nes the geometry of CME - actin networks , which is also likely to be the case for \ufb01lopodia extension ( Akamatsu et al . , 2020 ) . The nucleation promoting factor N - WASP arranges in a ring around CME sites ( Almeida - Souza et al . , 2018 ; Mund et al . , 2018 ) . In contrast , actin assembly factors accumulate at the tip of \ufb01lopodia ( Rottner et al . , 2017 ) . Interestingly , previous structural work also suggests that , similar to our results , about 10 \ufb01laments are involved in \ufb01lopodia extension ( Medalia et al . , 2007 ) . Moreover , analysis of baculovirus - induced actin comet tails by electron tomography of membrane extracted cells suggests that the intracellular pathogen is pushed through the cytoplasm by the simultaneous assembly of 2 \u2013 6 branched actin \ufb01laments ( Mueller et al . , 2014 ) . We found that endocytic vesicles are pushed through the cytoplasm by a comparable number of branched actin \ufb01laments ( Figures S5J and S5K ) . In conclusion , we provided a comprehensive description of the complex native actin architecture during mammalian CME in unprecedented detail . We have precisely quanti\ufb01ed the \ufb01la - ment number and their orientation at CME events allowing for un - ambiguous predictions on their force production capabilities , which has not been achieved for other actin - driven processes in unperturbed cells . Our experimental and modeling data both highlight the remarkable \ufb02exibility and robustness of productive actin network organization . We conclude that assembly of actin networks in human cells produces suf\ufb01cient force to account for robust endocytic membrane internalization and neck constric - tion . By combining structural analysis and mathematical modeling , we gained previously inaccessible mechanistic in - sights into CME - actin regulation and function , which has impli - cations for many other actin - driven processes . Limitations of the study Manual segmentation of actin \ufb01laments and identi\ufb01cation of actin \ufb01lament branch junctions , as we did in this study , are well - established methods for negative staining electron tomo - graphy and for cryo - ET data ( Mueller et al . , 2014 , 2017 ; Vinzenz et al . , 2012 ) . In addition to manual approaches , automatic actin \ufb01lament segmentation and branch junction detection ap - proaches have also been developed ( F \u20ac a\u00dfler et al . , 2020 ; Jasnin and Crevenna , 2016 ; Jasnin et al . , 2019 ; Rigort et al . , 2012 ) . Automatic approaches can speed up the data analysis pro - cedure but are not necessarily more precise compared to manual approaches executed by expert users ( Rigort et al . , 2012 ) . Additionally , these approaches still require some level of user interference and , likemanual approaches , are not completely bias free . Although we are convinced that we did not miss a signi\ufb01cant number of authentic branch junctions , we cannot completely exclude the existence of additional junc - tions that might have been picked up using automatic ap - proaches . Special care needs to be taken , either manually or computationally , to ensure that identi\ufb01ed branches are indeed bona \ufb01de branch junctions . For a cell - biology - based project like ours , false - positive branch junctions would result in misleading conclusions . Two promising , freely available actin convolutional neural - network - based segmentation tools were published after we \ufb01nalized our analysis ( Dimchev et al . , 2021 ; Martins et al . , Figure 7 . Harnessing actin polymerization force for the pushing , pulling , and squeezing of cellular membranes Actin polymerization can generate a pushing force to extend the plasma membrane during the formation of \ufb01lopodia or lamellipodia and to push vesicles and intracellular pathogens through the cytoplasm . Proteins with both actin - and membrane - binding abilities ( e . g . , Hip1R ) can anchor actin \ufb01laments to the mem - brane , converting pushing force into pulling and squeezing forces based on their spatial localization . In addition to the distribution of anchoring points , actin network morphology and polymerization force direction are de\ufb01ned by a second geometrical constraint , the position of actin assembly factors . Note that the precise localization of assembly factors and actin - membrane linkers is not clear for most of the displayed processes . ll Article 1142 Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 2021 ) . One of these tools also allows actin \ufb01lament polarity to be determined , possibly with greater precision than was possible using our approach ( Martins et al . , 2021 ) . Thus , this software tool might have allowed \ufb01lament polarity determination for some of the unbranched \ufb01laments in our study , for which a clear polarity assignment was not possible ( see STAR Methods ) , though a head - to - head comparison of the two approaches would be required to de\ufb01nitively test this possibility . Given that we already detected an excess of \ufb01lament plus ends at CME sites over what is required for CME progression , this tool in all likelihood would not have changed our overall conclusions , though it might have yielded a more precise quanti\ufb01cation . Given the prospects for new convolutional neural - network - based tech - nologies , we are eager to implement such approaches in our future studies . STAR + METHODS Detailed methods are provided in the online version of this paper and include the following : d KEY RESOURCES TABLE d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND SUBJECT DETAILS B Cell lines B Cell culture B Cell line authentication d METHOD DETAILS B Cell culture B Cryo - sample preparation B Cryo - \ufb02uorescence light microscopy B Cryo - electron tomography data acquisition B Tomogram reconstruction , segmentation model gen - eration and subtomogram averaging B Size measurement of clathrin - coated features B Filament length analysis B Branch angle measurements B Filament polarity and orientation analysis B Mathematical modeling using Cytosim d QUANTIFICATION AND STATISTICAL ANALYSIS B Image analysis and quanti\ufb01cation B Distribution analysis SUPPLEMENTAL INFORMATION Supplemental information can be found online at https : / / doi . org / 10 . 1016 / j . devcel . 2022 . 04 . 012 . ACKNOWLEDGMENTS We would like to thank Elizabeth Montabana , Daniel Toso , and Jonathan Re - mis for technical assistance with cryo - ET method development and data collection ; Paul Tobias for computation support ; and Sun Hae Hong for CLTA - GFP cell line generation and validation . Cryo - ET sample preparation was optimized at the LBNL cryo - EM resource , and image data were collected at the Bay Area Cryo - EM facility at the University of California Berkeley . Daniel Fletcherprovidedvaluableremarks onthemanuscript . Funding : thisworkwas supported through National Institutes of Health MIRA grant R35GM118149 to D . G . D . , Human Frontier Science Program long term fellowship LT000234 / 2018 - L to D . S . , Arnold and Mabel Beckman Foundation fellowship and Na - tional Institutes of Health grant 1 K99 GM132551 - 01 to M . A . , and National In - stitutes of Health grant R01 - GM132106 to P . R . . Method development was partially supported through Departments of Energy Early Career Award DE - AC02 - O5CH11231 to K . M . D . The funders had no role in study design , data collectionandinterpretation , orthedecisiontosubmittheworkforpublication . AUTHOR CONTRIBUTIONS Conceptualization , D . S . , M . A . , P . R . , and D . G . D . ; methodology , D . S . and M . A . ; investigation , D . S . , M . A . , A . M . , K . V . , and J . M . H . ; visualization , D . S . ; funding acquisition , D . S . , M . A . , K . M . D . , P . R . , and D . G . D . ; project administration , D . S . and D . G . D . ; software , D . S . , R . V . , M . A . , and J . S . ; supervision , D . G . D . and P . R . ; writing\u2014original draft , D . S . ; writing\u2014review & editing , D . S . , M . A . , P . R . , and D . G . D . DECLARATION OF INTERESTS The authors declare no competing interests . Received : August 13 , 2021 Revised : January 18 , 2022 Accepted : April 7 , 2022 Published : May 2 , 2022 REFERENCES Akamatsu , M . , Vasan , R . , Serwas , D . , Ferrin , M . A . , Rangamani , P . , andDrubin , D . G . ( 2020 ) . Principles of self - organization and load adaptation by the actin cytoskeleton during clathrin - mediated endocytosis . eLife 9 , e49840 . Almeida - Souza , L . , Frank , R . A . W . , Garc\u0131\u00b4a - Nafr\u0131\u00b4a , J . , Colussi , A . , Gunawardana , N . , Johnson , C . M . , Yu , M . , Howard , G . , Andrews , B . , Vallis , Y . , and McMahon , H . T . ( 2018 ) . A \ufb02at BAR protein promotes actin polymeriza - tion at the base of clathrin - coated pits . Cell 174 , 325 \u2013 337 . e14 . Avinoam , O . , Schorb , M . , Beese , C . J . , Briggs , J . A . G . , and Kaksonen , M . ( 2015 ) . ENDOCYTOSIS . Endocyticsitesmaturebycontinuousbendingandre - modeling of the clathrin coat . Science 348 , 1369 \u2013 1372 . Blanchoin , L . , Amann , K . J . , Higgs , H . N . , Marchand , J . B . , Kaiser , D . A . , and Pollard , T . D . ( 2000 ) . Direct observation of dendritic actin \ufb01lament networks nucleated by Arp2 / 3 complex and WASP / Scar proteins . Nature 404 , 1007 \u2013 1011 . Bykov , Y . S . , Schaffer , M . , Dodonova , S . O . , Albert , S . , Plitzko , J . M . , Baumeister , W . , Engel , B . D . , and Briggs , J . A . G . ( 2017 ) . The structure of the COPI coat determined within the cell . eLife 6 , e32493 . Carreno , S . , Engqvist - Goldstein , A . E . , Zhang , C . X . , McDonald , K . L . , and Drubin , D . G . ( 2004 ) . Actin dynamics coupled to clathrin - coated vesicle forma - tion at the trans - Golgi network . J . Cell Biol . 165 , 781 \u2013 788 . Chen , Q . , andPollard , T . D . ( 2013 ) . Actin\ufb01lamentseveringbyco\ufb01lindismantles actin patches and produces mother \ufb01laments for new patches . Curr . Biol . 23 , 1154 \u2013 1162 . Cheng , Y . , Boll , W . , Kirchhausen , T . , Harrison , S . C . , and Walz , T . ( 2007 ) . Cryo - electron tomography of clathrin - coated vesicles : structural implications for coat assembly . J . Mol . Biol . 365 , 892 \u2013 899 . Chou , S . Z . , and Pollard , T . D . ( 2019 ) . Mechanism of actin polymerization re - vealed by cryo - EM structures of actin \ufb01laments with three different bound nu - cleotides . Proc . Natl . Acad . Sci . USA 116 , 4265 \u2013 4274 . Clarke , N . I . , and Royle , S . J . ( 2018 ) . FerriTag is a new genetically - encoded inducible tag for correlative light - electron microscopy . Nat . Commun . 9 , 2604 . Collins , A . , Warrington , A . , Taylor , K . A . , and Svitkina , T . ( 2011 ) . Structural or - ganization of the actin cytoskeleton at sites of clathrin - mediated endocytosis . Curr . Biol . 21 , 1167 \u2013 1175 . De La Cruz , E . M . , and Gardel , M . L . ( 2015 ) . Actin mechanics and fragmenta - tion . J . Biol . Chem . 290 , 17137 \u2013 17144 . ll Article Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 1143 Dimchev , G . , Amiri , B . , F \u20ac a\u00dfler , F . , Falcke , M . , and Schur , F . K . M . ( 2021 ) . Computational toolbox for ultrastructural quantitative analysis of \ufb01lament net - works in cryo - ET data . J . Struct . Biol . 213 , 107808 . Dmitrieff , S . , and Ne\u00b4de\u00b4lec , F . ( 2016 ) . Ampli\ufb01cation of actin polymerization forces . J . Cell Biol . 212 , 763 \u2013 766 . Dominguez , R . , and Holmes , K . C . ( 2011 ) . Actin structure and function . Annu . Rev . Biophys . 40 , 169 \u2013 186 . Doyon , J . B . , Zeitler , B . , Cheng , J . , Cheng , A . T . , Cherone , J . M . , Santiago , Y . , Lee , A . H . , Vo , T . D . , Doyon , Y . , Miller , J . C . , etal . ( 2011 ) . Rapidand ef\ufb01cient cla - thrin - mediated endocytosis revealed in genome - edited mammalian cells . Nat . Cell Biol . 13 , 331 \u2013 337 . Engqvist - Goldstein , A . E . , Kessels , M . M . , Chopra , V . S . , Hayden , M . R . , and Drubin , D . G . ( 1999 ) . An actin - binding protein of the Sla2 / huntingtin interacting protein 1 family is a novel component of clathrin - coated pits and vesicles . J . Cell Biol . 147 , 1503 \u2013 1518 . Engqvist - Goldstein , A . E . , Warren , R . A . , Kessels , M . M . , Keen , J . H . , Heuser , J . , and Drubin , D . G . ( 2001 ) . The actin - binding protein Hip1R associates with cla - thrin during early stages of endocytosis and promotes clathrin assembly in vitro . J . Cell Biol . 154 , 1209 \u2013 1223 . F \u20ac a\u00dfler , F . , Dimchev , G . , Hodirnau , V . - V . , Wan , W . , and Schur , F . K . M . ( 2020 ) . Cryo - electrontomography structureofArp2 / 3complexincellsrevealsnewin - sights into the branch junction . Nat . Commun . 11 , 6437 . Fehon , R . G . , McClatchey , A . I . , and Bretscher , A . ( 2010 ) . Organizing the cell cortex : the role of ERM proteins . Nat . Rev . Mol . Cell Biol . 11 , 276 \u2013 287 . Garner , R . M . , and Theriot , J . A . ( 2020 ) . Leading edge stability in motile cells is an emergent property of branched actin network growth . eLife 11 , e74389 . Gingras , A . R . , Bate , N . , Goult , B . T . , Hazelwood , L . , Canestrelli , I . , Grossmann , J . G . , Liu , H . , Putz , N . S . M . , Roberts , G . C . K . , Volkmann , N . , et al . ( 2008 ) . The structure of the C - terminal actin - binding domain of talin . EMBO J . 27 , 458 \u2013 469 . Grassart , A . , Cheng , A . T . , Hong , S . H . , Zhang , F . , Zenzer , N . , Feng , Y . , Briner , D . M . , Davis , G . D . , Malkov , D . , andDrubin , D . G . ( 2014 ) . Actinanddynamin2dy - namics and interplay during clathrin - mediated endocytosis . J . Cell Biol . 205 , 721 \u2013 735 . Hassinger , J . E . , Oster , G . , Drubin , D . G . , and Rangamani , P . ( 2017 ) . Design principles for robust vesiculation in clathrin - mediated endocytosis . Proc . Natl . Acad . Sci . USA 114 , E1118 \u2013 E1127 . Heumann , J . M . , Hoenger , A . , and Mastronarde , D . N . ( 2011 ) . Clustering and variance maps for cryo - electron tomography using wedge - masked differ - ences . J . Struct . Biol . 175 , 288 \u2013 299 . Heuser , J . ( 1980 ) . Three - dimensional visualization of coated vesicle formation in \ufb01broblasts . J . Cell Biol . 84 , 560 \u2013 583 . Jasnin , M . , Beck , F . , Ecke , M . , Fukuda , Y . , Martinez - Sanchez , A . , Baumeister , W . , and Gerisch , G . ( 2019 ) . The architecture of traveling actin waves revealed by cryo - electron tomography . Structure 27 , 1211 \u2013 1223 . e5 . Jasnin , M . , andCrevenna , A . H . ( 2016 ) . Quantitativeanalysisof\ufb01lamentbranch orientation in Listeria actin comet tails . Biophys . J . 110 , 817 \u2013 826 . Kaksonen , M . , andRoux , A . ( 2018 ) . Mechanismsofclathrin - mediatedendocy - tosis . Nat . Rev . Mol . Cell Biol . 19 , 313 \u2013 326 . Kaksonen , M . , Toret , C . P . , and Drubin , D . G . ( 2006 ) . Harnessing actin dy - namicsforclathrin - mediatedendocytosis . Nat . Rev . Mol . CellBiol . 7 , 404 \u2013 414 . Kaplan , C . , Kenny , S . J . , Chen , X . , Scho\u00a8neberg , J . , Sitarska , E . , Diz - Mun\u02dcoz , A . , Akamatsu , M . , Xu , K . , and Drubin , D . G . ( 2021 ) . Load adaptation of endocytic actin networks . Mol . Biol . Cell . Published online April 7 , 2022 . https : / / doi . org / 10 . 1091 / mbc . E21 - 11 - 0589 . Kim , S . H . , Choi , H . J . , Lee , K . W . , Hong , N . H . , Sung , B . H . , Choi , K . Y . , Kim , S . - M . , Chang , S . , Eom , S . H . , and Song , W . K . ( 2006 ) . Interaction of SPIN90 with syndapin is implicated in clathrin - mediated endocytic pathway in \ufb01bro - blasts . Genes Cells 11 , 1197 \u2013 1211 . Kremer , J . R . , Mastronarde , D . N . , and McIntosh , J . R . ( 1996 ) . Computer visual - ization of three - dimensional image data using IMOD . J . Struct . Biol . 116 , 71 \u2013 76 . Luan , Q . , Liu , S . - L . , Helgeson , L . A . , and Nolen , B . J . ( 2018 ) . Structure of the nucleation - promoting factor SPIN90 bound to the actin \ufb01lament nucleator Arp2 / 3 complex . EMBO J . 37 , e100005 . Lu (cid:2) ci (cid:3) c , V . , Rigort , A . , and Baumeister , W . ( 2013 ) . Cryo - electron tomography : the challenge of doing structural biology in situ . J . Cell Biol . 202 , 407 \u2013 419 . Mahamid , J . , Pfeffer , S . , Schaffer , M . , Villa , E . , Danev , R . , Cuellar , L . K . , Fo\u00a8rster , F . , Hyman , A . A . , Plitzko , J . M . , and Baumeister , W . ( 2016 ) . Visualizing the mo - lecular sociology at the HeLa cell nuclear periphery . Science 351 , 969 \u2013 972 . Martins , B . , Sorrentino , S . , Chung , W . - L . , Tatli , M . , Medalia , O . , and Eibauer , M . ( 2021 ) . Unveiling the polarity of actin \ufb01laments by cryo - electron tomogra - phy . Structure 29 , 488 \u2013 498 . e4 . Mastronarde , D . N . ( 2005 ) . Automated electron microscope tomography using robust prediction of specimen movements . J . Struct . Biol . 152 , 36 \u2013 51 . Maupin - Szamier , P . , and Pollard , T . D . ( 1978 ) . Actin \ufb01lament destruction by osmium tetroxide . J . Cell Biol . 77 , 837 \u2013 852 . Medalia , O . , Beck , M . , Ecke , M . , Weber , I . , Neujahr , R . , Baumeister , W . , and Gerisch , G . ( 2007 ) . Organization of actin networks in intact \ufb01lopodia . Curr . Biol . 17 , 79 \u2013 84 . Morris , K . L . , Jones , J . R . , Halebian , M . , Wu , S . , Baker , M . , Armache , J . - P . , Avila Ibarra , A . , Sessions , R . B . , Cameron , A . D . , Cheng , Y . , and Smith , C . J . ( 2019 ) . Cryo - EM of multiple cage architectures reveals a universal mode of clathrin self - assembly . Nat . Struct . Mol . Biol . 26 , 890 \u2013 898 . Mueller , J . , Pfanzelter , J . , Winkler , C . , Narita , A . , Le Clainche , C . , Nemethova , M . , Carlier , M . - F . , Maeda , Y . , Welch , M . D . , Ohkawa , T . , et al . ( 2014 ) . Electron tomography andsimulation ofbaculovirusactin comet tailssupportatethered \ufb01lament model of pathogen propulsion . PLoS Biol . 12 , e1001765 . Mueller , J . , Szep , G . , Nemethova , M . , de Vries , I . , Lieber , A . D . , Winkler , C . , Kruse , K . , Small , J . V . , Schmeiser , C . , Keren , K . , et al . ( 2017 ) . Load adaptation of Lamellipodial actin networks . Cell 171 , 188 \u2013 200 . e16 . Mullins , R . D . , Heuser , J . A . , and Pollard , T . D . ( 1998 ) . The interaction of Arp2 / 3 complex with actin : nucleation , high af\ufb01nity pointed end capping , and forma - tion of branching networks of \ufb01laments . Proc . Natl . Acad . Sci . USA 95 , 6181 \u2013 6186 . Mund , M . , van der Beek , J . A . , Deschamps , J . , Dmitrieff , S . , Hoess , P . , Monster , J . L . , Picco , A . , Ne\u00b4de\u00b4lec , F . , Kaksonen , M . , and Ries , J . ( 2018 ) . Systematic nanoscale analysis of endocytosis links ef\ufb01cient vesicle formation to patterned actin nucleation . Cell 174 , 884 \u2013 896 . e17 . Muriel , O . , Tomas , A . , Scott , C . C . , and Gruenberg , J . ( 2016 ) . Moesin and cor - tactincontrolactin - dependentmultivesicularendosome biogenesis . Mol . Biol . Cell 27 , 3305 \u2013 3316 . Narita , A . , Mueller , J . , Urban , E . , Vinzenz , M . , Small , J . V . , andMae\u00b4da , Y . ( 2012 ) . Direct determination of actin polarity in the cell . J . Mol . Biol . 419 , 359 \u2013 368 . Nedelec , F . , and Foethke , D . ( 2007 ) . Collective Langevin dynamics of \ufb02exible cytoskeletal \ufb01bers . New J . Phys . 9 , 427 . Nicastro , D . , Schwartz , C . , Pierson , J . , Gaudette , R . , Porter , M . E . , and McIntosh , J . R . ( 2006 ) . The molecular architecture of axonemes revealed by cryoelectron tomography . Science 313 , 944 \u2013 948 . Oh , H . , Kim , H . , Chung , K . - H . , Hong , N . H . , Shin , B . , Park , W . J . , Jun , Y . , Rhee , S . , and Song , W . K . ( 2013 ) . SPIN90 knockdown attenuates the formation and movement of endosomal vesicles in the early stages of epidermal growth fac - tor receptor endocytosis . PLoS One 8 , e82610 . Pollard , T . D . ( 2007 ) . Regulation of actin \ufb01lament assembly by Arp2 / 3 complex and formins . Annu . Rev . Biophys . Biomol . Struct . 36 , 451 \u2013 477 . Pollard , T . D . ( 2016 ) . Actin and actin - binding proteins . Cold Spring Harb . Perspect . Biol . 8 , a018226 . Raz - Ben Aroush , D . , Ofer , N . , Abu - Shah , E . , Allard , J . , Krichevsky , O . , Mogilner , A . , and Keren , K . ( 2017 ) . Actin turnover in Lamellipodial fragments . Curr . Biol . 27 , 2963 \u2013 2973 . e14 . Resch , G . P . , Goldie , K . N . , Hoenger , A . , and Small , J . V . ( 2002 ) . Pure F - actin networks are distorted and branched by steps in the critical - point drying method . J . Struct . Biol . 137 , 305 \u2013 312 . Rigort , A . , G \u20ac unther , D . , Hegerl , R . , Baum , D . , Weber , B . , Prohaska , S . , Medalia , O . , Baumeister , W . , and Hege , H . - C . ( 2012 ) . Automated segmentation of ll Article 1144 Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 electron tomograms for a quantitative description of actin \ufb01lament networks . J . Struct . Biol . 177 , 135 \u2013 144 . Roth , T . F . , and Porter , K . R . ( 1964 ) . Yolk protein uptake in the oocyte of the mosquito Aedes aegypti . L . J . Cell Biol . 20 , 313 \u2013 332 . Rottner , K . , Faix , J . , Bogdan , S . , Linder , S . , and Kerkhoff , E . ( 2017 ) . Actin as - sembly mechanisms at a glance . J . Cell Sci . 130 , 3427 \u2013 3435 . Rouiller , I . , Xu , X . - P . , Amann , K . J . , Egile , C . , Nickell , S . , Nicastro , D . , Li , R . , Pollard , T . D . , Volkmann , N . , andHanein , D . ( 2008 ) . Thestructuralbasisofactin \ufb01lament branching by the Arp2 / 3 complex . J . Cell Biol . 180 , 887 \u2013 895 . Schreiber , C . H . , Stewart , M . , and Duke , T . ( 2010 ) . Simulation of cell motility that reproduces the force \u2013 velocity relationship . Proc . Natl . Acad . Sci . USA 107 , 9141 \u2013 9146 . Serwas , D . , and Davies , K . M . ( 2021 ) . Getting started with in situ cryo - electron tomography . In Methods in Moecular Biology ( Springer US ) , pp . 3 \u2013 23 . Small , J . V . ( 1981 ) . Organization of actin in the leading edge of cultured cells : in\ufb02uence of osmium tetroxide and dehydration on the ultrastructure of actin meshworks . J . Cell Biol . 91 , 695 \u2013 705 . Sochacki , K . A . , Dickey , A . M . , Strub , M . - P . , and Taraska , J . W . ( 2017 ) . Endocyticproteinsarepartitionedattheedgeoftheclathrinlatticeinmamma - lian cells . Nat . Cell Biol . 19 , 352 \u2013 361 . Sochacki , K . A . , Heine , B . L . , Haber , G . J . , Jimah , J . R . , Prasai , B . , Alfonzo - Me\u00b4ndez , M . A . , Roberts , A . D . , Somasundaram , A . , Hinshaw , J . E . , and Taraska , J . W . ( 2021 ) . The structure and spontaneous curvature of clathrin lat - tices at the plasma membrane . Dev . Cell 56 , 1131 \u2013 1146 . e3 . Svitkina , T . M . , and Borisy , G . G . ( 1999 ) . Arp2 / 3 complex and actin depolyme - rizing factor / co\ufb01lin in dendritic organization and treadmilling of actin \ufb01lament array in lamellipodia . J . Cell Biol . 145 , 1009 \u2013 1026 . Taylor , M . J . , Perrais , D . , and Merri\ufb01eld , C . J . ( 2011 ) . A high precision survey of the molecular dynamics of mammalian clathrin - mediated endocytosis . PLoS Biol . 9 , e1000604 . Vinzenz , M . , Nemethova , M . , Schur , F . , Mueller , J . , Narita , A . , Urban , E . , Winkler , C . , Schmeiser , C . , Koestler , S . A . , Rottner , K . , et al . ( 2012 ) . Actin branching in the initiation and maintenance of lamellipodia . J . Cell Sci . 125 , 2775 \u2013 2785 . Wan , W . , and Briggs , J . A . G . ( 2016 ) . Chapter thirteen\u2014cryo - electron tomogra - phy and subtomogram averaging . In Methods Enzymol , R . A . Crowther , ed . ( Academic Press ) , pp . 329 \u2013 367 . Yarar , D . , Waterman - Storer , C . M . , and Schmid , S . L . ( 2005 ) . A dynamic actin cytoskeleton functions at multiple stages of clathrin - mediated endocytosis . Mol . Biol . Cell 16 , 964 \u2013 975 . ll Article Developmental Cell 57 , 1132 \u2013 1145 , May 9 , 2022 1145 STAR + METHODS KEY RESOURCES TABLE RESOURCE AVAILABILITY Lead contact Further information and requests for resources and reagents should be directed to and will be ful\ufb01lled by the lead contact , David G Drubin . REAGENT or RESOURCE SOURCE IDENTIFIER Chemicals , peptides , and recombinant proteins Alexa Fluor 647 ChromPure human transferrin Jackson Immuno Research Inc . 009 - 600 - 050 ; RRID : AB _ 2337114 Deposited data Tomogram This paper EMDB - 26483 EM density map clathrin hub This paper EMDB - 26484 Experimental models : Cell lines SK - MEL - 2 ( sex : male ) ( endogenously expressing Dynamin2 tagged with eGFP clathrin light chain A ( CLTA ) tagged with TagRFP ) Drubin lab N / A SK - MEL - 2 ( sex : male ) ( endogenously expressing clathrin light chain A ( CLTA ) tagged with eGFP ) Drubin lab N / A Software and algorithms SerialEM https : / / bio3d . colorado . edu / SerialEM / download . html ( Mastronarde , 2005 ) N / A IMOD 4 . 9 . 12 https : / / bio3d . colorado . edu / imod / ( Kremer et al . , 1996 ) N / A PEET 1 . 13 . 0 https : / / bio3d . colorado . edu / PEET / , ( Heumann et al . , 2011 ; Nicastro et al . , 2006 ) N / A Leica Application Suite 3 . 7 Leica N / A actinPolarity This paper ( https : / / github . com / JohSchoeneberg / actinPolarity ) https : / / doi . org / 10 . 5281 / zenodo . 6410857 Actincme This paper ( https : / / github . com / kvegesna / polarity - analysis / tree / paper _ submission ) https : / / doi . org / 10 . 5281 / zenodo . 6413181 Other Leica EM Cryo CLEM ( Leica DM6 FS ) Leica N / A 50x Leica EM Cryo CLEM ceramic tipped objective Leica N / A GFP ET \ufb01lter ( Excitation : 470 / 40 , Dichroic : 495 , Emission 525 / 50 ) Leica 11504164 Y5 ( Excitation : 620 / 60 , Dichroic : 660 , Emission 700 / 75 ) Leica 11504171 Leica DFC9000 GT Leica 11547006 Titan Krios G2 FEI ( now Thermo Fisher ) N / A K2 direct electron detector Gatan N / A K3 direct electron detector Gatan N / A BioQuantum energy \ufb01lter Gatan N / A Vitrobot Mark IV ( FEI , now Thermo Fisher ) N / A Holey carbon grids ( R2 / 1 , 200 mesh , gold ) Quantifoil Q35430 ll Article e1 Developmental Cell 57 , 1132 \u2013 1145 . e1 \u2013 e5 , May 9 , 2022 Materials availability Mammalian cell lines that were used in this study are available from the lead contact upon request . Data and code availability d One representative tomogram and the EM density map of the clathrin hub have been deposited in the EMDB and are publicly available as of the date of publication . Accession numbers are listed in the key resources table . d Additional data reported in this paper will be shared by the lead contact upon reasonable request . d All original code has been deposited at GitHub and is publicly available as of the date of publication . URLs are listed in the key resources table . EXPERIMENTAL MODEL AND SUBJECT DETAILS Cell lines All cell lines used in this study are derivates of commercially available SK - MEL - 2 cells ( https : / / www . atcc . org / products / htb - 68 ) . These are human melanoma cells . Constructions of recombinant cell lines and culture conditions are described in detail below . Cell culture Double tagged SK - MEL - 2 cells endogenously expressing clathrin light chain CLTA - TagRFP and dynamin2 - eGFP , and single tagged SK - MEL - 2 cells endogenously expressing clathrin light chain CLTA - eGFP , were cultured in DMEM / F12 ( Gibco (cid:2) ) supplemented with 10 % fetal bovine serum ( premium grade , VWR Life Science Seradigm ) and 1 % Penicillin / Streptomycin ( Gibco (cid:2) ) . Cell line authentication Genome - edited cells were authenticated by short tandem repeat pro\ufb01ling . METHOD DETAILS Cell culture The double - tagged CLTA - TagRFP and dynamin2 - eGFP cell line was generated using zinc - \ufb01nger nuclease - mediated genome editing and was published previously ( Doyon et al . , 2011 ; Grassart et al . , 2014 ) . The same reagents were used for the single - tagged cell line , except that the TagRFP in the donor plasmid used to construct the double tagged cell line was replaced with eGFP ( Doyon et al . , 2011 ; Grassart et al . , 2014 ) . SK - MEL - 2 cell lines were used because of their robust endocytic dynamics , good adherence and spreading behavior on EM grids . Cryo - sample preparation Holey carbon grids ( Quantifoil R2 / 1 , 200 mesh , gold ) were washed in acetone while agitating on a rocker for 15 to 30 min to remove potential residual plastic backing , washed in H 2 O and dried . Grids were then glow discharged using a Pelco SC - 6 sputter coater , sterilized in 70 % ethanol in H 2 O . Grids were placed in a 6 - well plate containing cell culture medium ( one grid per well ) under sterile conditions in a biosafety cabinet . Grids were incubated in cell culture medium at 37 (cid:3) C and 5 % CO 2 in a cell culture incubator over - night . Cell culture medium was replaced the next day . SK - MEL - 2 cells from a 75 \u2013 90 % con\ufb02uent 10 cm plastic cell culture dish were transferred to wells using a 1 : 2 \u2013 1 : 3 dilution . Cells were further incubated overnight at 37 (cid:3) C and 5 % CO 2 in a cell culture incubator . Before vitri\ufb01cation , grids were removed from the 6 - well plate , washed by continuously dripping a total of 10 m l of 10 nm BSA Gold Tracer ( Electron Microscopy Sciences ) solution and removing the drops with \ufb01lter paper . Next , 3 m l 10 nm BSA Gold Tracer were pipetted onto the sample side . Using a Vitrobot Mark IV ( FEI , now Thermo Fisher ) , samples were blotted either from both sides or from the backside of the grid by replacing the \ufb01lter paper on the sample side with a same - sized Te\ufb02on sheet , and vitri\ufb01ed by plunging into liquid ethane . See ( Serwas and Davies , 2021 ) for a step - by - step protocol . For cryo - correlative light and electron microscopy im - aging , SK - MEL - 2 cells expressing CLTA - eGFP were serum starved in medium without fetal bovine serum for 10 \u2013 30 min . Grids were then incubated in a 3 m l drop of medium containing 25 m g / ml Alexa Fluor 647 ChromPure human transferrin ( Jackson Immuno Research Inc . ) for 2 \u2013 6 min prior to initiating the washing steps in BSA Gold Tracer solution . Chamber conditions for the Vitrobot Mark IV were set to 37 (cid:3) C and 90 % humidity . To ensure that suitable grids were produced for each imaging session , a range of plot forces between 0 and 10 , blotting times of 2 . 5 to 4 s , and a 1 s drain time , were used . Grids were \ufb01xed into AutoGrid carrier ( Thermo Fisher ) and stored in liquid nitrogen until they were imaged . Cryo - \ufb02uorescence light microscopy Cryo - samples prepared from single - tagged CLTA - GFP expressing SK - MEL - 2 cells and incubated with Alexa Fluor 647 ChromPure human transferrin ( Jackson Immuno Research Inc . ) were imaged on a Leica EM Cryo CLEM system ( Leica ) . The system consists of a Leica DM6 FS wide\ufb01eld microscope that is equipped with a motorized Leica EM Cryo stage and a short working distance ( < 0 . 28 mm ) 50x Leica EM Cryo CLEM ceramic tipped objective ( numerical aperture = 0 . 90 ) . These speci\ufb01cations allow sample imaging at liquid nitrogen temperatures . A halogen lamp powered by a CTR6 halogen supply unit was used as light source . We used GFP ll Article Developmental Cell 57 , 1132 \u2013 1145 . e1 \u2013 e5 , May 9 , 2022 e2 ET ( Excitation : 470 / 40 , Dichroic : 495 , Emission 525 / 50 ) , Y5 ( Excitation : 620 / 60 , Dichroic : 660 , Emission 700 / 75 ) \ufb01lter cubes for imaging . For cryo - correlative light and electron microscopy , a grid overview map was recorded using transmitted light and GFP channels . The map was used to identify suf\ufb01ciently spread cells in regions with good ice quality . The CLTA - GFP signal appeared diffuse in regions of thick or crystalline ice . Z - stacks ( total Z = (cid:1) 6 m m in 0 . 6 m m steps ) of cells with clear CLTA - GFP foci were recorded using the transmitted light , GFP ( CLTA ) and Y5 ( transferrin ) channels . The same imaging conditions were used for cells on individual grids . Imaging conditions were varied between grids to obtain suf\ufb01cient signal to allow correlation in later steps . Note that some cells or CME sites did not show a transferrin signal most likely either because not all CME sites were loaded with transferrin cargo or due to inconsistent labeling . Images for panel generation in Adobe Illustrator were prepared using FIJI and Adobe Photoshop . No nonlinear gamma correction was applied during image processing . Cryo - electron tomography data acquisition Samples were imaged on a Titan Krios transmission electron microscope ( FEI ) operated equipped with a X - FEG electron source , a BioQuantum energy \ufb01lter ( Gatan ) and a K2 or K3 direct electron detecting device ( Gatan ) ( see Table S1 for details on the used de - tector ) and operated at 300 kV . Samples were visually inspected for ice quality using the FEI \ufb02u cam . Overview grid maps were ac - quired at (cid:1) 0 . 2 m m pixel size . If samples were imaged by cryo - \ufb02uorescence light microscopy before , these grid maps were used to identify cells of interest from the cryo - \ufb02uorescence light microscopy overview maps . For random data collection , cells with thin cell regions were located . For both types of data collection , polygon maps of the regions of interest with an overlap of 20 - 25 % between individual images were recorded . The polygon maps were used to pick tilt series acquisition points either at random or based on \ufb02uorescent signal from cryo - \ufb02uorescence light microscopy . The hole pattern of the carbon \ufb01lm was used as a guide . Acquisition points were chosen with adequate distance between individual points to prevent electron dose exposure damage prior to data collection . Only cellular regions without obvious sample damage ( e . g . , through blotting ) were used for data collection . SerialEM ( Mastronarde , 2005 ) in low - dose mode was used for automatic tilt - series recording using a bidirectional tilt scheme starting at + 20 (cid:3) and with a typical tilt range from + 60 (cid:3) to - 60 (cid:3) and base increment of 2 (cid:3) . Pixel sizes between 2 . 97 and 3 . 72 A\u02da were used . Target defocus was varied between - 2 and - 8 m m and the targeted total electron dose was 100 e - / A\u02da 2 ( see Table S1 for acquisition parameter details of individual tomograms presented in here ) . Data was collected in super resolution mode on the K2 detector or in 0 . 5 binning mode on the K3 detector . Frame time was 0 . 2 \u2013 0 . 25 s . Note that at the beginning of the project , the Leica cryo - CLEM microscope was not available to us and we therefore used the double - tagged \ufb02uorescent SK - MEL - 2 cell line to verify that CME occurs in cells growing on holey carbon grids by \ufb02uorescence microscopy . Tomogram reconstruction , segmentation model generation and subtomogram averaging Tilt series alignment and tomogram reconstruction were done using the freely available IMOD software package ( Kremer et al . , 1996 ) . Tilt series were aligned using 10 nm gold particles as \ufb01ducials and tomograms were reconstructed using the backprojection algorithm combined with the SIRT - like \ufb01lter . Tomograms were then \ufb01ltered using the Nonlinear Anisotropic Diffusion \ufb01lter and binned by a fac - tor of 2 using the binvol function to further increase contrast . Tomograms from a total of 7 independent datasets were analyzed ( see Table S1 for details ) . Subtomogram averaging was performed using PEET . The tomograms used are indicated in Table S1 ( Heumann et al . , 2011 ; Ni - castro et al . , 2006 ) . Subtomograms were picked manually and initial motive lists with starting orientations for alignment were gener - ated using the spikeInit or the slicer2MOTL functions . A single subtomogram was used as the starting reference . Subtomograms were iteratively aligned and averaged and an updated coarse aligned average was used after each iteration . For the subtomogram average of the clathrin hub , 3 - fold rotational symmetry was applied . Fourier shell correlation was performed by using the calcFSC function and plotted using the plotFSC function . ChimeraX ( https : / / www . rbvi . ucsf . edu / chimerax / ) was used for subtomogram average visualization and docking of PDB structures . The clathrin hub isosurface model that was used in the segmentation models was generated in IMOD . Segmentation model generation in IMOD was done by manually tracing actin \ufb01laments and membrane shapes and plotting the clathrin hub isosurface model to the original subtomogram positions that were re\ufb01ned during the subtomogram averaging procedure . The \ufb01ltered and binned tomograms were used for this procedure . Filament segmentation and branch junction identi\ufb01cation were car - ried out by two individuals ( DS and AM ) to reduce bias . Branch junctions were modeled as a scattered point object ( spherical object ) that was placed in the center of the junction density . Branch junctions were identi\ufb01ed manually in a two - step process : First , candidate branch junctions were marked during the initial manual actin \ufb01lament segmentation . Then , we searched our segmentation models for actin \ufb01lament pairs where one \ufb01lament end was close to another \ufb01lament and resembled branch junction geometry . Tomograms were re - inspected in these positions and we re - evaluated whether these \ufb01lament pairs were connected through branch junctions . Tomograms and corresponding segmentation models were cropped to a volume that was centered on the CME site or CCV and had a size of about 500 nm x 500 nm in x and y dimension while the zdimension was kept at its original value . Actin \ufb01laments were only considered for further analysis when they were contained in the cropped volume and thus in close proximity to the clathrin - coated feature . For \ufb01lament length analysis ( see below ) , the original length of the uncropped \ufb01lament was used if only a part of that \ufb01lament was contained in the cropped volume . Images for panel preparation in Adobe Illustrator were generated in IMOD and further processed in Adobe Photoshop . No nonlinear gamma correction was applied during image processing . ll Article e3 Developmental Cell 57 , 1132 \u2013 1145 . e1 \u2013 e5 , May 9 , 2022 Size measurement of clathrin - coated features CCV and CCP membrane and coat diameters were measured in the slicer view mode in IMOD . Rotation angles were adjusted to bring the CCV / CCP into full view as a symmetric object . A total of 3 measurements per CCV / CCP were performed on the central tomo - graphic slice and averaged to compensate for imprecision . Measurements for CCPs with a constricted neck were performed at the widest point of the invagination . Further analysis and graph generation were done in Prism8 . Filament length analysis Filament length was extracted from segmentation models of the \ufb01ltered and binned tomograms using the imodinfo function in IMOD . Further analysis and graph generation were done in Prism8 . Branch angle measurements Subtomograms of the branch junctions were extracted using the junction model points and the boxstartend function in IMOD . Sub - tomograms were displayed in the slicer view mode in IMOD . Rotational angles were adjusted to bring both \ufb01laments at the junction into view as shown in Figures 3A and S3A . Tiff images of these views were saved and opened in FIJI . Branch angle measurement was performed using the Angle tool . To minimize the effect of \ufb01lament \ufb02exibility , measurements were performed close to the junction and minimal parts of the \ufb01laments were included . To assess the effect of the missing wedge , which is most prominent in the z direction , on the branch angle measurements , we plotted the measured branch angle against the x and y rotations that had to be applied to bring the junction into view ( mother and daughter \ufb01lament visible ) . The equal distribution of branch angles across the rotation angles shows that our branch angle measurements were not strongly affected by the missing wedge at this level of precision . Filament polarity and orientation analysis Branched actin \ufb01lament polarity was analyzed based on branch junction geometry as indicated in Figure 3A . For analysis of un - branched \ufb01lament polarity , we adapted a previously published cross - correlation based method ( Narita et al . , 2012 ) . For the cross - correlation analysis , we \ufb01rst generated an arti\ufb01cial 3D actin \ufb01lament from a previously published structure ( PDB : 6DJM ; ( Chou and Pollard , 2019 ) ) using ChimeraX and IMOD . We then used the relion _ image _ handler function in Relion to rescale and low - pass \ufb01lter the arti\ufb01cial \ufb01lament to the pixel size of our binned tomography data ( 5 . 943 A\u02da ) and a resolution of 30 A\u02da . Next , a tiff image stack of 2D projection images of the arti\ufb01cial actin \ufb01lament was generated using the xyzproj function in IMOD . Each of the projection images was taken after rotating the \ufb01lament by 5 (cid:3) around the \ufb01lament axis ( F ) relative to the previous image . The complete image stack contained a total of 36 images covering a 180 (cid:3) rotation of the arti\ufb01cial actin \ufb01lament . The image stack was then cropped to 69 pixels x 23 pixels , corresponding to the length of a 13 - subunit actin \ufb01lament . This size was picked because of the structural or - ganization of actin \ufb01laments . The symmetry of an actin \ufb01lament can be described as a single strand left - handed helix with about 13 monomers repeating every six turns ( Dominguez and Holmes , 2011 ) . Accordingly , a 13 - subunit reference projection series contains suf\ufb01cient information about the helical organization of actin \ufb01laments to serve as a reference structure . We then generated a second reference image stack with opposing polarity by rotating the original stack by 180 (cid:3) ( Figure S4A ) . Hereafter we refer to the reference with the plus pointing up as ref + angle - i and the one with the minus end pointing up as ref - angle - i . Filaments for polarity analysis were extracted from the tomograms using the corresponding segmentation models and the imodmop function in IMOD . 2D projection im - ages of these \ufb01laments were generated in FIJI . Images were oriented as shown as in Figure S4B and cropped so the resulting image was 23 pixels wide ( = width of the reference images ) . Note was taken on the rotation that was applied to the images , and only straight parts of the \ufb01laments were included . From these test \ufb01laments ( test ) , sub - images of the size of the reference images ( 69 pixels x 23 pixels ) were extracted with a frequency of 5 pixels ( test i ) , corresponding to about one actin subunit in the \ufb01lament ( Figure S4B ) . The cross - correlation coef\ufb01cient ( R ) was calculated for each of the test sub - images test i with each of images in both of the reference stacks ref + angle - i and ref - angle - i ( Figure S4C ) . Then , the correlation curves with ref + angle - i and ref - angle - i were subtracted from each other and a difference curve was generated ( Figure S4D ) . The mean average difference was calculated from these values . If the value was negative , the tested \ufb01lament was determined to have the same polarity as ref - angle - i , if the value was positive , the tested \ufb01lament was determined to have the same polarity as ref + angle - i . The cross - correlation calculation algorithm was prototyped on ImageJ with the Image CorellationJ plugin ( https : / / www . gcsca . net / IJ / ImageCorrelationJ . html ) . To increase throughput and ease of use , the software was automated and reimplemented into a custom python - based command line tool called \" actinPolarity\u2019\u2019 . The accuracy of this method depends on test \ufb01lament length . Therefore , only unbranched \ufb01laments of at least 80 nm length ( > two 13 - subunit repeats ) were included in the analysis . Branched \ufb01laments were included without length consideration because the \ufb01lament junction identi\ufb01es the polarity . To calculate and display the orientation of the \ufb01laments in the tomograms based on the determined polarity ( either by branch junc - tion geometry or cross - correlation analysis ) relative to a user - de\ufb01ned reference plane , we developed an analysis pipeline with Jupyter notebooks ( Python 3 . 7 ) called \u2018\u2018actincme\u2019\u2019 . In general , the reference plane was parallel to the plasma membrane for CME invagina - tions or tangential to the surface for CCVs . The reference plane was manually segmented using IMOD . We calculated a vector normal to the reference plane , such that 0 (cid:3) orientation was de\ufb01ned as the vector pointing from the \ufb01laments toward the reference plane , and 180 (cid:3) was identi\ufb01ed as pointing away from the plane . Filaments with model points crossing within 10 pixels ( 6 nm ) from the reference plane were excluded from the analysis . The directionality of the \ufb01laments ( from minus end to plus end ) was calculated as the arcco - sine of the dot product between the \ufb01lament vector and the vector normal to the reference plane . Filament positions and orientations were plotted using matplotlib ( 3 . 0 . 2 ) . Further analysis and graph generation were done in Prism8 . ll Article Developmental Cell 57 , 1132 \u2013 1145 . e1 \u2013 e5 , May 9 , 2022 e4 Mathematical modeling using Cytosim We used the agent - based model in Akamatsu et al . 2020 to run 3D stochastic simulations of the mammalian endocytic actin network ( Akamatsu et al . , 2020 ) . We used similar parameters and initial conditions with the following modi\ufb01cations : To test effects of the branch angle variability on force production and internalization ef\ufb01ciency , we changed the average branching angle of the Arp2 / 3 complex to be 70 (cid:3) rather than 77 (cid:3) , to closer match the measured branch angle in the tomograms ( Akamatsu et al . , 2020 ) . In both cases the branch \ufb02exibility was set to 0 . 076 pN m m / rad . In general , to average multiple stochastic simulations , we ran 24 simulations per condition . To add Hip1R molecules at the neck , we repurposed Matlab code from Akamatsu et al . ( 2020 ) in order to distribute Hip1R molecules uniformly around a cylinder of radius 30 nm to match the shape and diameter of a typical CME neck . Sixty Hip1R molecules were distributed along 45 nm of the surface of the neck . This conservative value for the number of Hip1R molecules at the neck corresponds to a lower molecular surface density relative to the bud , and results in more tractable simulation runs that proceed to completion . QUANTIFICATION AND STATISTICAL ANALYSIS Image analysis and quanti\ufb01cation A total of 12 CME events from 11 tomograms were analyzed ( Table S1 ) , which is the range of similar studies done previously ( Jasnin et al . , 2019 ; Mueller et al . , 2017 ) . Image analysis and quanti\ufb01cation was done as detailed in the Method details section . Statistical analysis was done in GraphPad Prism8 unless otherwise noted . Statistical details can be found in the Results section , \ufb01gure legends , STAR Methods , Tables S2 and S3 . Distribution analysis The Prism8 curve \ufb01tting function was used for distribution analysis . We tested for Gaussian , lognormal and exponential ( one phase decay ) distributions ( Table S2 ) . R 2 values were used to determine the best \ufb01t . For Gaussian distributed data , we used mean and stan - dard deviation values for description in the main manuscript text , and we used the median for non - Gaussian distributed data . ll Article e5 Developmental Cell 57 , 1132 \u2013 1145 . e1 \u2013 e5 , May 9 , 2022",
    "pedersen2023endocytic": "REPORT Endocytic myosin - 1 is a force - insensitive , power - generating motor Ross T . A . Pedersen 1 * \ue840 , Aaron Snoberger 2 * \ue840 , Serapion Pyrpassopoulos 2 \ue840 , Daniel Safer 2 \ue840 , David G . Drubin 1 \ue840 , and E . Michael Ostap 2 \ue840 Myosins are required for clathrin - mediated endocytosis , but their precise molecular roles in this process are not known . This is , in part , because the biophysical properties of the relevant motors have not been investigated . Myosins have diverse mechanochemical activities , ranging from powerful contractility against mechanical loads to force - sensitive anchoring . To better understand the essential molecular contribution of myosin to endocytosis , we studied the in vitro force - dependent kinetics of the Saccharomyces cerevisiae endocytic type I myosin called Myo5 , a motor whose role in clathrin - mediated endocytosis has been meticulously studied in vivo . We report that Myo5 is a low - duty - ratio motor that is activated \u223c 10 - fold by phosphorylation and that its working stroke and actin - detachment kinetics are relatively force - insensitive . Strikingly , the in vitro mechanochemistry of Myo5 is more like that of cardiac myosin than that of slow anchoring myosin - 1s found on endosomal membranes . We , therefore , propose that Myo5 generates power to augment actin assembly - based forces during endocytosis in cells . Introduction During clathrin - mediated endocytosis ( CME ) , the plasma mem - brane invaginates and undergoes scission to become a cytoplas - mic vesicle . Coat proteins like clathrin can deform membranes under low tension ( Dannhauser and Ungewickell , 2012 ; Busch et al . , 2015 ; Cail et al . , 2022 ) , but when bending is resisted by membrane tension ( Hassinger et al . , 2017 ) , the actin cytoskeleton drives membrane invagination ( Boulant et al . , 2011 ; Kaplan et al . , 2022 ) . In yeasts , including Saccharomyces cerevisiae and Schizo - saccharomyces pombe , turgor pressure opposes plasma mem - brane invagination , so actin is required at every CME site ( Aghamohammadzadeh and Ayscough , 2009 ; Basu et al . , 2014 ) . The actin cytoskeleton can produce pushing and pulling force , both of which are required for CME in S . cerevisiae ( Sun et al . , 2006 ) . When actin filament ends grow against a surface , they push the surface forward ( Mogilner and Oster , 1996 , 2003 ) . During CME , actin filaments , bound by coat proteins , grow against the plasma membrane to drive invagination ( Picco et al . , 2015 ; Kaksonen et al . , 2005 , 2003 ; Skruzny et al . , 2012 , Fig . 1 ) . Modeling of the homologous CME machinery in mammalian cells demonstrated that such actin networks generate sufficient power for CME ( Akamatsu et al . , 2020 ) , but whether actin as - sembly alone can overcome turgor pressure in yeast cells is debated ( Nickaeen et al . , 2019 ; Carlsson , 2018 ) . Additional power may be provided by myosins , which gen - erate tension on actin filaments . The myosins critical for CME \u2014 Myo3 and Myo5 in budding yeast and Myo1e in vertebrates \u2014 are type I myosins ( Geli and Riezman , 1996 ; Cheng et al . , 2012 ; Krendel et al . , 2007 ) . Some type I myosins are suited to generate power \u2014 i . e . , they carry out mechanical work over time by consuming ATP to execute a power stroke . Other type I myosins are suited to serve as force - sensitive anchors in that mechanical load locks them in a low energy \u2013 requiring , tension - maintaining state ( Greenberg and Ostap , 2013 ) . The possible roles of type I myosins in CME depend on whether endocytic myosins are power generators or force - sensitive anchors . If endocytic type I myosins are acutely force sensitive , they might organize the actin filaments of the endocytic actin net - work , while if they are less force sensitive , they could power plasma membrane invagination ( Evangelista et al . , 2000 ; Fig . 1 ) . Myosin - 1 motors form a ring at the base of CME sites , where the invaginated membrane meets the plasma membrane ( Mund et al . , 2018 ; Fig . 1 ) . Yeast type I myosins serve at least one or - ganizational function as a membrane anchor for the actin as - sembly machinery , a function associated with the non - motor tail of the molecules ( Lewellyn et al . , 2015 ) , but motor activity is required in addition to membrane anchorage ( Pedersen and . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Department of Molecular and Cell Biology , University of California , Berkeley , Berkeley , CA , USA ; 2 Pennsylvania Muscle Institute , Perelman School of Medicine , University of Pennsylvania , Philadelphia , PA , USA . Correspondence to David G . Drubin : drubin @ berkeley . edu ; E . Michael Ostap : ostap @ mail . med . upenn . edu * R . T . A . Pedersen and A . Snoberger contribution equally to this paper . R . T . A . Pedersen \u2019 s current affiliation is Department of Embryology , Carnegie Institution for Science , Baltimore , MD , USA . \u00a9 2023 Pedersen et al . This article is available under a Creative Commons License ( Attribution4 . 0 International , as describedat https : / / creativecommons . org / licenses / by / 4 . 0 / ) . Rockefeller University Press https : / / doi . org / 10 . 1083 / jcb . 202303095 1 of 12 J . Cell Biol . 2023 Vol . 222 No . 10 e202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 Drubin , 2019 ) . If endocytic myosin - 1s are force - sensitive an - chors , they may serve a further organizational role by holding growing filaments in an optimal orientation for force generation ( Fig . 1 , left ) . If the myosins are power - generating motors , they may pull actin filament ends away from the plasma membrane , deepening the plasma membrane invagination and creating space for monomer addition and filament elongation ( Fig . 1 , right ) , a model supported by the observation that the actin as - sembly rate at CME sites depends on type I myosin motors in a dose - dependent manner ( Manenschijn et al . , 2019 ) . To distinguish between these possibilities , we measured the force sensitivity of the endocytic myosin Myo5 ( not to be confused with the vertebrate type V myosin ) . Myo5 is insensitive to resis - tive force compared to related myosins . We , therefore , propose that Myo5 actively powers CME . Because actin and myosin col - laborate in a variety of membrane remodeling processes , we ex - pect that these results will be instructive beyond CME . Results and discussion Heavy chain phosphorylation activates Myo5 ATPase activity To determine Myo5 force sensitivity , we first needed to measure its unloaded kinetics . We purified a Myo5 construct containing the motor and lever domains from S . cerevisiae ( Fig . 2 A ) . Because phosphorylation of Myo5 at the TEDS site is required for most CME events and is thought to regulate Myo5 \u2019 s motor activity ( Grosshans et al . , 2006 ; Sun et al . , 2006 ; Bement and Mooseker , 1995 ) , we purified a phosphory - lated version and an unphosphorylated version of the protein ( see Materials and methods ) . The p21 - activated kinase was used to phosphorylate the myosin at the TEDS site ( S357 ) , as determined by control experiments with an S357A mutant ( Fig . S1 ) . The phosphorylation state of preparations was judged to be uniform when ATP - induced actoMyo5 dissociation transients were well fit by single exponential functions ( see below ) . The yeast light chain for Myo5 , calmodulin ( Cmd1 , Geli et al . , 1998 ) , was purified from E . coli and included in excess in all experiments ( Fig . 2 A ) . We measured the steady - state actin - activated ATPase activ - ities of phosphorylated and unphosphorylated Myo5 using the NADH - coupled assay ( De La Cruz and Ostap , 2009 ) in the presence of 0 \u2013 80 \u00b5M phalloidin - stabilized actin filaments . Un - phosphorylated Myo5 ATPase activity was largely insensitive to actin filaments : the ATPase rate at 0 \u00b5M actin was 0 . 14 s \u2013 1 , while the maximum ATPase rate measured was 0 . 39 s \u2013 1 at 40 \u00b5M actin . ( Fig . 2 B ) . Phosphorylation activated Myo5 ATPase activity by \u223c 10 - fold ( Fig . 2 B ) . The actin concentration dependence of the phosphorylated Myo5 ATPase rate ( k obs ) was well fit by : k obs (cid:1) v o V max [ Actin ] K ATPase + [ Actin ] . ( 1 ) From the fit , the actin concentration at half - maximum of the ATPase rate ( K ATPase ) was determined to be 5 . 1 \u00b1 0 . 88 \u00b5M , and the maximum ATPase rate ( V max ) was found to be 3 . 3 \u00b1 0 . 15 s \u2013 1 ( Fig . 2 B ; Table 1 ) . ATP binding and ADP release are non - rate limiting for Myo5 ATPase activity Resistive force impacts the rate of myosin detachment from actin and two biochemical transitions , ADP release and subsequent ATP binding , determine the detachment rate . Therefore , we used stopped - flow kinetics to measure ADP release from ( Fig . 2 C , k + 5 9 ) and ATP binding to ( Fig . 2 C , K 1 9 and k + 2 9 ) actoMyo5 . We found that yeast Myo5 does not quench the fluorescence of actin labeled at cys - 374 with pyrene iodoacetamide , which is the probe most used to measure myosin binding ( De La Cruz and Ostap , 2009 ) . Thus , we measured actoMyo5 detachment by monitoring light scattering , which decreases as myosin unbinds actin filaments . To determine the rate constant for ATP binding , we mixed nucleotide - free actoMyo5 ( 100 nM ) with varying concentrations of ATP and monitored 90\u00b0 light scattering . Time courses fol - lowed single exponential functions ( Fig . 2 D ) . For phosphorylated Myo5 , the observed rates determined from the fits increased linearly with ATP concentration ( Fig . 2 E ) . At concentrations of > 1 mM ATP , the actomyosin complex dissociated within the re - sponse time of the instrument , precluding measurement . For unphosphorylated Myo5 , the observed rates fit a rectangular hyperbola with increasing ATP concentration ( Fig . 2 E ) . Scheme 1 The mechanism was modeled as in Scheme 1 ( De La Cruz and Ostap , 2009 ) , where K 1 9 is a rapid equilibrium binding step , k 2 9 is Figure 1 . Models for the functions of actin assembly and myosin ac - tivity during membrane deformation for clathrin - mediated endocytosis . Cartoon diagram illustrating the organization of actin filaments and Myo5 molecules at endocytic sites . Actin filaments are bound by coat proteins at the tip of the growing membrane invagination and oriented with their growing ends toward the plasma membrane , powering membrane invagi - nation . The type I myosin Myo5 could either anchor the actin network in a favorable orientation ( left ) or provide an assisting force ( right ) . Pedersen et al . Journal of Cell Biology 2 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 Figure 2 . In - solution , population biochemical characterization of Myo5 . ( A ) Coomassie - stained SDS - polyacrylamide gels showing example preparations of the purified Myo5 motor / lever construct and calmodulin ( Cmd1 , light chain ) used in all experiments . ( B ) The actin concentration dependence of the steady - state ATPase activity of 100 nM unphosphorylated ( gray circles ) and phosphorylated Myo5 ( black circles ) . Each data point represents the average of 6 \u2013 7 time courses , which were 100 s each . The orange line is the best fit of the phosphorylated Myo5 data to a rectangular hyperbola . ( C ) Schematic pathway for the Myo5 ATPase cycle . Blue motors are in tightly bound conformations and green motors are weakly bound / unbound . ( D ) Example of light scattering transients reportingon ATP - induced dissociation of phosphorylated ( left , k obs = 17 s \u2212 1 ) and unphosphorylated ( right , k obs = 64 . 1 s \u2212 1 ) actoMyo5 , obtained by mixing 100 nM actoMyo5 ( AM ) with94and 72 \u00b5M ATP , respectively , as shown intheinset schematic . Theblacklineisthefit of a singleexponential function to thedata . ( E ) ATP Pedersen et al . Journal of Cell Biology 3 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 a rate - limiting isomerization to the AM . ATP state , and k diss is the rapid actin dissociation step . The apparent second - order rate constant for ATP binding to phosphorylated actoMyo5 was de - termined by a linear fit to the data ( K 1 9 k 2 9 = 0 . 39 \u00b1 0 . 017 \u00b5m \u2013 1 s \u2013 1 ) . The unphosphorylated actoMyo5 data were fit by k obs (cid:1) K 1 \u2019 ATP [ ] 1 + K 1 \u2019 ATP [ ] (cid:1) (cid:3) k 2 \u2019 ( 2 ) and the maximum rate of isomerization ( k 2 9 = 290 \u00b1 24 s \u2013 1 ) and ATP affinity ( K 1 9 = 0 . 006 \u00b1 0 . 0016 \u00b5M \u2013 1 ) were determined . The apparent second - order rate constant for ATP binding ( K 1 9 k 2 9 ) was determined from a linear fit of the observed rates below 100 \u00b5M ATP to be 1 . 1 \u00b1 0 . 28 \u00b5M \u2013 1 s \u2013 1 ( Table 1 ) . The rate constant for ADP dissociation ( k + 5 9 ) was measured by preincubating 100 \u00b5M ADP with 200 nM actoMyo5 and then rapidly mixing with 2 . 5 mM ATP , as shown in Scheme 2 . Scheme 2 When myosin active sites are saturated with ADP , the rate of ATP - induced dissociation of actomyosin is limited by ADP \u2019 s slow dissociation . Light scattering transients were fitted by single exponential functions , yielding rates for ADP release for phosphorylated actoMyo5 ( k + 5 9 = 74 \u00b1 2 . 0 s \u2013 1 ) and for un - phosphorylated actoMyo5 ( k + 5 9 = 107 \u00b1 5 . 9 s \u2013 1 ; Fig . 2 F and Table 1 ) . The signal - to - noise ratio of the fast light scattering transients is low , resulting in large uncertainties on these fits . However , these rates are substantially faster than the steady - state ATPase values but slower than the maximum rate of ATP - induced actomyosin dissociation . ADP release for actoMyo5 ADP is much faster than ADP release for vertebrate Myo1b and Myo1c ( Greenberg et al . , 2012 ; Lewis et al . , 2006 ) . It is more similar to the vertebrate en - docytic myosin - 1 , Myo1e ( El Mezgueldi et al . , 2002 ) . Because ADP release is rate limiting for detachment of Myo5 and Myo1e from actin , fast ADP release by these molecules means that the unloaded actin - attachment lifetimes for endocytic type I myosins are < 15 ms . This property may make these motors particularly well - suited to function in dynamic actin networks like those at CME sites , where actin filaments elongate and \u201c treadmill \u201d into the cytoplasm ( Kaksonen et al . , 2003 , 2005 ) . Actin gliding is dependent on Myo5 phosphorylation state Our results suggest that both phosphorylated and unphosphorylated Myo5 have low duty ratios ( i . e . , the motor spends a small fraction of its ATPase cycle bound to actin ) . Since ADP release limits the rate of phosphorylated Myo5 detachment from actin at saturating ATP ( k + 5 9 = 74 \u00b1 2 . 0 s \u2013 1 ) and since we have measured the overall ATPase rate ( V max = 3 . 3 \u00b1 0 . 15 s \u2013 1 ) , we can estimate the duty ratio as follows : DutyRatio (cid:1) 1 k + 5 \u2019 (cid:4) (cid:5) 1 V max (cid:4) (cid:5) ( 3 ) The calculated duty ratio of phosphorylated Myo5 is 0 . 045 . Unphosphorylated Myo5 has a lower duty ratio ( < 0 . 004 ) . To assess the effect of phosphorylation on Myo5 motility , we performed in vitro motility assays at 1 mM ATP . Motors were attached site - specifically to coverslips coated with anti - His 6 antibody . Coverslips were incubated with a range of concentrations of phosphorylated and unphosphorylated Myo5 , creating a titration series of surface densities . At low Myo5 surface densities ( incubation with \u2264 30 nM phosphory - lated Myo5 , \u2264 150 nM unphosphorylated Myo5 ) , actin fila - ments failed to bind the coverslip ( Fig . 2 G ; and Videos 1 and 2 ) . At higher concentrations , phosphorylated Myo5 moved actin filaments at velocities ranging from 720 \u00b1 40 nm / s ( 100 nM phosphorylated Myo5 ) to 880 \u00b1 90 nm / s ( 40 nM ; Fig . 2 G and Video 1 ) . These gliding velocities are considerably higher than those reported by Sun et al . , ( 2006 ) , possibly reflecting differences in the phosphorylation state of the purified Myo5 protein ( see below ) or differences in other motility assay conditions , such as light chain availability . Higher ( greater concentration dependence of dissociation of 100 nM unphosphorylated ( gray circles ) and phosphorylated actoMyo5 ( black circles ) . Each data point represents 3 \u2013 6 time courses averaged and fit to a single exponential decay function . The orange line is a linear best fit of the phosphorylated Myo5 data . The purple line is the best fit of the unphosphorylated Myo5 data to a rectangular hyperbola . ( F ) Example light scattering transients reporting ATP - induced dissociation of ADP - saturated phosphorylated ( left ) and unphosphorylated ( right ) actoMyo5 , obtained by preincubating 200 nM actoMyo5 ( AM ) with 100 \u00b5M ADP , then mixing rapidly with 2 . 5 mM ATP , as shown in the inset schematic . The black line is the fit of a single exponential function to the data . ( G ) Velocity of actin filament gliding , measured at varying surface densities of Phospho - Myo5 ( black circles , orange line ) and unphosphorylated Myo5 ( gray circles , purple line ) in in vitro motility assays . Myosin concentrations indicate the quantity of protein incubated in the flow chamber before washing . Each data point represents the average velocity of 30 \u2013 60 filaments , and the error bars are standard deviations . Source data are available for this figure : SourceData F2 . Table 1 . Rate and equilibrium constants of the Myo5 ATPase cycle Phosphorylated Myo5 Unphosphorylated Myo5 Steady - state actin - activated ATPase V max ( s \u2212 1 ) 3 . 3 ( \u00b10 . 15 ) ND K ATPase ( \u00b5M ) 5 . 1 ( \u00b10 . 88 ) ND ATP binding K 1 9 ( \u00b5M \u2212 1 ) ND 0 . 006 ( \u00b10 . 0016 ) k 2 9 ( s \u2212 1 ) \u2265 335 290 ( \u00b124 ) K 1 9 k 2 9 ( \u00b5M \u2212 1 s \u2212 1 ) a 0 . 39 ( \u00b10 . 017 ) b 1 . 1 ( \u00b10 . 28 ) c ADP release k + 5 9 ( s \u2212 1 ) 74 \u00b1 2 . 0 107 ( \u00b15 . 9 ) Summary ofrateand equilibriumconstants measuredforMyo5inthisstudy . Errors are standard errors of the fits . a Determined from a linear fit of the unbinding rates . b Linear fit of all data for phosphorylated Myo5 in Fig . 2 E . c Linear fit of observed rates below 100 \u00b5M ATP for unphosphorylated Myo5 in Fig . 2 E . ND : Not determined . Pedersen et al . Journal of Cell Biology 4 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 than fivefold ) surface densities of unphosphorylated Myo5 were required to achieve smooth motility , but this motility occurred at a substantially slower speed , \u223c 120 nm / s ( Fig . 2 G and Video 2 ) . While it is possible that residual phosphorylated Myo5 in the unphosphorylated preparation contributed to this motility , Sun et al . , 2006 similarly reported that Myo5 harboring TEDS site mutations moved actin filaments much more slowly . The slower actin gliding speed for unphosphorylated myosin was unexpected given the similar rates of ADP release between phosphorylated and unphosphorylated Myo5 ( Table 1 ) . It is possible that our kinetics experiments have not determined the rate - limiting step for detachment , but it is more likely that the motility of the unphosphorylated myosin is limited by the slow attachment rate of the motor ( Stewart et al . , 2021 ) , as suggested by the slow actin - activated ATPase rate . The activation of Myo5 motility by phosphorylation could explain why fast , cargo - induced endocytosis , which involves rapid and dynamic actin turnover , requires phosphorylated Myo5 , while slower constitutive endo - cytosis does not ( Grosshans et al . , 2006 ) . Myo5 \u2019 s working stroke comprises two substeps that are consistent with unloaded kinetic measurements The kinetics of actin attachment and mechanics of single myosin molecules were measured by optical trapping ( Woody et al . , 2018 ; Snoberger et al . , 2021 ) . We used the three - bead optical trapping geometry in which a biotinylated actin filament is held between two laser - trapped polystyrene beads coated with neu - travidin , creating a bead - actin - bead dumbbell ( Fig . 3 A ) . Dumb - bells were lowered onto pedestal beads that were sparsely coated with phosphorylated Myo5 - His 9 bound to a surface - adsorbed anti - His 6 tag antibody . The positions of trapped beads were detected , and single actomyosin binding events were identified by a de - crease in the covariance of the bead positions ( Fig . 3 , B \u2013 D ) . Traces acquired at 1 , 10 , and 1 , 000 \u00b5M ATP reveal clear displacements and drops in covariance during actomyosin binding events . Event durations decreased with increasing ATP concentrations ( Fig . 3 , B \u2013 D , blue lines ) . The myosin - 1 working stroke occurs in two discrete substeps , with the first substep occurring with phosphate release and the second with ADP release ( Jontes et al . , 1995 ; Veigel et al . , 1999 , Fig . 3 E ) . The substeps can be characterized in optical trapping assays by ensemble averaging single interactions ( Veigel et al . , 1999 ; Chen et al . , 2012 ; Laakso et al . , 2008 ) , where the detected events are aligned at their beginnings and forward - averaged in time ( Fig . 3 , F \u2013 H , left ) , or aligned at their ends and reverse - averaged in time ( Fig . 3 , F \u2013 H , right ) . Ensemble averages of Myo5 interactions showed a two - step working stroke at the three ATP concentrations but the step size was most accurately resolved at 10 \u00b5M ATP ( see Materials and methods ) . In this condition , an initial substep of 4 . 8 nm was followed by a second substep of 0 . 2 nm ( Fig . 3 G ) . We deter - mined the lifetimes of the substeps by fitting the ensemble averages with single exponential functions . At 1 \u00b5M ATP ( Fig . 3 F , left trace ) , the measured rate ( > 30 s \u2212 1 ) of the time - forward average was limited by the covariance smoothing window , but at 10 and 1 , 000 \u00b5M ATP ( Fig . 3 , G and H , left traces ) , the rates were 49 \u00b1 1 . 6 and 50 \u00b1 0 . 2 s \u2212 1 , respectively ( Fig . 3 K ) , which are similar to the measured ADP release rate ( k + 5 9 , 74 \u00b1 2 . 0 s \u2212 1 , Table 1 ) supporting the model that the tran - sition from State 1 to State 2 accompanies ADP release . The kinetics of time - reversed averages reveal the lifetime of State 2 ( Fig . 3 , F \u2013 H , right traces ) . Fitting single exponential functions to these traces reveals rates of 0 . 59 \u00b1 0 . 003 and 7 . 34 \u00b1 0 . 1 s \u2212 1 at 1 and 10 \u00b5M ATP , respectively ( Fig . 3 K ) . At 1 , 000 \u00b5M ATP , the observed rate ( > 187 s \u2212 1 ) was limited by the covariance smoothing window ( 5 . 25 ms ; Fig . 3 K ) . The observed rates at 1 and 10 \u00b5M ATP are consistent with the second - order rate constant for ATP binding of 0 . 39 \u00b1 0 . 017 \u00b5M \u2212 1 s \u2212 1 measured by stopped - flow kinetics ( K 1 9 k 2 9 , Table 1 ) . We determined the detachment rates of actomyosin events by plotting the cumulative frequency of individual attachment durations and fitting a single exponential function to the data by maximum likelihood estimation ( MLE ; Fig . 3 I ) . Data from 1 and 10 \u03bc M ATP were well fit by single exponentials with rates of 0 . 88 and 6 . 87 s \u2212 1 , respectively ( Fig . 3 , I \u2013 K ) . These rates match well with the observed rate of ATP binding ( Table 1 ) , as well as the fits for the reverse ensemble averages , indicating that at sub - saturating ATP ( 1 and 10 \u00b5M ) , detachment is limited by ATP binding ( Fig . 3 , I \u2013 K ) . Data from 1 , 000 \u03bc M ATP were best described as the sum of two exponentials , with the major rate of 67 . 8 s \u2212 1 comprising 92 . 1 % of the total , and a minor rate of 11 . 6 s \u2212 1 comprising 7 . 9 % of the total ( Fig . 3 , I and K ) . The major rate is consistent with both the observed ADP release rate and the measured forward ensemble average rates , indicating that at saturating ATP , ADP release limits detachment of actomyosin interactions ( Fig . 3 , J \u2013 K ) . Myo5 is a relatively force - insensitive motor We measured how the actin detachment rate of Myo5 was af - fected by mechanical force opposing the power stroke using an isometric feedback system ( Takagi et al . , 2006 ) . The initial force applied to Myo5 in this system depends on where along the actin filament Myo5 stochastically binds , allowing measurement of attachment durations at a range of resistive forces ( Fig . 4 A ) . Plotting attachment durations as a function of force revealed a trend of longer attachment durations at higher resisting forces . At each force , attachment durations are exponentially distributed and , as expected , the data appear noisy when plotted ( Fig . 4 A ) . Con - verting these data to detachment rates by binning by force , aver - aging , and taking the inverse clearly reveals the trend ( Fig . 4 B ) . The force dependence of the Myo5 detachment rate was fit by the Bell Equation : k ( F ) (cid:1) k 0 e \u2212 Fd kB \u2219 T , ( 4 ) where k ( F ) is the detachment rate at force F , k 0 is the detachment rate in the absence of load , d is the distance parameter ( the distance to the force - dependent transition state and a measure of force sensitivity ) , k B is Boltzmann \u2019 s constant , and T is the temperature . Best fit parameters for k 0 and d were determined by MLE of the unaveraged data from Fig . 4 A , incorporating the instrument response time ( Woody et al . , 2016 ) . The esti - mated detachment rate in the absence of force is 67 . 6 s \u2212 1 , in agreement with the measured detachment rate under low load Pedersen et al . Journal of Cell Biology 5 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 Figure 3 . Single molecule , optical trap analysis of Myo5 step size and kinetics . ( A ) Cartoon schematic of the three - bead optical trapping setup . A bi - otinylated actin filament is tethered between two neutravidin - coated beads that are trapped in a dual - beam optical trap . This bead - actin - bead \u201c dumbbell \u201d is lowered onto pedestal beads that have been sparsely coated with His 6 antibody to attach Myo5 - motor / lever - Avi - Tev - His 9 . ( B \u2013 D ) Single Myo5 displacements of a single bead position and covariance traces , calculated using both beads , showing single molecule interactions acquired in the presence of 1 \u00b5M ( B ) 10 \u00b5M ( C ) , and 1 , 000 \u00b5M ATP ( D ) . Blue bars indicate attachment events as identified by covariance ( gray ) decreases . The threshold of event detection by the co - variance traces are indicated by dashed gray lines . ( E ) Schematic of displacement traces depicting the two - step nature of actomyosin displacements in the optical trap . ( F \u2013 H ) Binding events were synchronized at their beginnings ( left ) or ends ( right ) and averaged forward or backward in time , respectively . The Pedersen et al . Journal of Cell Biology 6 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 conditions at 1 , 000 \u00b5M ATP ( 67 . 8 s \u2212 1 , Fig . 3 K ) , and the value for d was 1 . 14 nm . To put Myo5 \u2019 s force sensitivity in context , we replotted the function describing the force - dependent actin detachment rate of Myo5 alongside the same curves for vertebrate Myo1b , Myo1c , and \u03b2 - cardiac myosin , which were determined by the same experimental approach ( Fig . 4 C , Laakso et al . , 2010 ; Greenberg et al . , 2012 ; Woody et al . , 2018 ) . The mechanochemistry of Myo5 ( d = 1 . 14 nm ) is most like that of \u03b2 - cardiac ( muscle ) myosin ( d = 1 . 3 nm ) , suggesting that it is well - suited for generating power . The difference between Myo5 and acutely force - sensitive Myo1b , a tension - sensitive anchor myosin ( d = 15 nm ) , is dramatic . From 0 to 2 pN of resistance , Myo1b at - tachment lifetimes slow from \u223c 600 ms to \u223c 45 s , resulting in negligible power generation ( Fig . 4 D ) . Over the same in - terval , Myo5 attachment lifetimes slow modestly from \u223c 15 to \u223c 25 ms , allowing it to generate considerable power ( Fig . 4 D ) . Thus , Myo5 is unlikely to act as a force - sensitive anchor in measured total displacement of Myo5 was 5 . 0 nm at 10 \u00b5M ATP , with the first substep contributing a 4 . 8 nm displacement ( arrow 1 in G ) and the second substep contributing a 0 . 2 - nm displacement ( arrow 2 in G ) . ( F \u2013 H ) Left : Forward - averaged ensembles synchronized at the beginnings of events . Right : Reverse - averaged ensembles synchronized at the ends of events . Black and gray lines are single exponential fits in the forward and reverse ensembles , respectively . ( I ) Cumulative distributions of attachment durations for Myo5 at 1 , 10 , and 1 , 000 \u00b5M ATP . Blue lines show the cumulative frequency of attachment durations at the indicated ATP concentrations , and the red , yellow , and green lines indicate fitted exponential distributions at 1 , 10 , and 1 , 000 \u00b5M ATP , respectively . 1 and 10 \u00b5M ATP were fit well to single exponentials , and the 1 , 000 \u00b5M ATP data were best described by the sum of two exponentials . ( J ) Summary of rates at 1 , 10 , and 1 , 000 \u00b5M ATP calculated from F \u2013 H . Blue boxes are the fitted exponential distributions from I , black diamonds are forward ensemble fits from F \u2013 H ( left ) , and gray diamonds are reverse ensemble fits from F \u2013 H ( right ) . At lower concentrations of ATP ( 1 and 10 \u00b5M ) , the rate of detachment is limited by ATP association , corresponding to the reverse ensemble fits , while at saturating ATP concentration ( 1 , 000 \u00b5M ) , the detachment rate is limited by the rate of ADP dissociation , corresponding to the forward ensemble fits . ( K ) Summary of rates determined via single - molecule optical trapping . Errors for detachment rates are 95 % confidence intervals . Errors for forward and reverse ensemble fits are standard errors of the fits . * Detachment rates at 1 , 000 \u00b5M ATP were best fit to the sum of two exponents . The major component of the fit ( 67 . 8 s \u2013 1 ) comprises 92 . 1 % of the total with the remaining 7 . 9 % having a rate of 11 . 6 s \u2013 1 . Figure 4 . Myo5 attachment lifetimes are substantially less force - dependent than other known type I myosins . An isometric optical force clamp was utilized to determine the force sensitivity of the detachment of Myo5 from actin . ( A ) Durations of individual actomyosin attachments as a function of force , plotted on a semi - log scale . ( B ) The solid black line shows the force dependence of the detachment rates determined by MLE fitting of unaveraged points in A . For illustration purposes , attachment durations from A were binned by force at every 10 points , averaged , and converted to rates . Best - fit parameters were determined by MLE fitting and 95 % confidence intervals were calculated via bootstrapping . The solid black line is calculated from best - fit parameters ( k = 67 . 6 s \u2013 1 , d = 1 . 14 nm ) , while the gray shaded region is the 95 % confidence interval ( k = 62 . 4 \u2013 72 . 9 s \u2013 1 , d = 1 . 03 \u2013 1 . 26 nm ) . All MLE fitting was performed on unaveraged data and was corrected for instrument deadtime . ( C ) The force dependent detachment rate of Myo5 ( from B ) plotted alongside the force de - pendent detachment rates for Myo1b , Myo1c , and \u03b2 - cardiac muscle myosin , Myh7 . ( D ) Power output for the same four myosins is calculated over a range of forces by multiplying the functions from C by the applied force F , and the step size and duty ratios of each myosin . Pedersen et al . Journal of Cell Biology 7 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 cells and is more likely to power movements against a resisting load . Proposed function of type I myosin in clathrin - mediated endocytosis Myo5 is one of the best - studied myosin - 1 proteins in vivo . Ex - tensive imaging has revealed that it is recruited to CME sites at the initiation of actin assembly , where it concentrates at the base of the site during membrane invagination ( Jonsdottir and Li , 2004 ; Idrissi et al . , 2008 ) . Although it was known that the mechanochemical activity of type - 1 myosins is required for CME ( Geli and Riezman , 1996 ; Goodson et al . , 1996 ; Sun et al . , 2006 ) , the mechanistic contribution of motor activity was unknown . When it was discovered that some type I myosins are acutely force - sensitive ( Laakso et al . , 2008 ) , it became apparent that these motors could have mechanochemical activities that range from force - dependent anchoring to power generation . It is clear that mutant Myo5 molecules with disrupted mechanochemistry block CME , but these results do not reveal the molecular role of Myo5 ( Lewellyn et al . , 2015 ; Idrissi et al . , 2012 ) . Perhaps , the most informative finding in cells has been the observation that varying the number of type I myosins at CME sites results in altered actin assembly rates ( Manenschijn et al . , 2019 ) . How - ever , because load influences growing branched actin networks in complex ways ( Bieling et al . , 2016 ) , even this finding did not clarify the molecular roles of endocytic myosin - 1s . Here , we have shown that Myo5 \u2019 s motor generates power rather than forming force - sensitive catch bonds . The overall ATPase rate of Myo5 is slow relative to other power - generating myosins , but its power stroke and detachment from actin are fast , and they slow only modestly under load ( Fig . 4 C ) . Myo5 \u2019 s relative force insensitivity means it generates power against resistance ( Fig . 4 D ) . Because Myo3 and Myo5 can each support CME in the absence of the other , we suspect that Myo3 is sim - ilarly force - insensitive . Given the homology between Myo5 and vertebrate Myo1e , together with the close agreement of their unloaded kinetics ( El Mezgueldi et al . , 2002 ) , we also predict that Myo1e generates biologically relevant power . Our finding that Myo5 \u2019 s kinetics are relatively force insen - sitive led us to interpret the previously described dose depen - dence of actin assembly on the number of myosin - 1s at endocytic sites to mean that this motor moves actin filaments at CME sites to power plasma membrane invagination and create space for new monomers to assemble ( Manenschijn et al . , 2019 ; Fig . 1 , right ) . On the order of 300 myosin molecules ( Myo3 and Myo5 combined ) are present at CME sites , mostly where the invagi - nating membrane meets the plasma membrane ( Mund et al . , 2018 ; Idrissi et al . , 2008 ; Sun et al . , 2019 ; Picco et al . , 2015 ) . Myo5 \u2019 s diffusion is likely to be impeded by the many proteins at the base of CME sites and it may move actin filaments at an angle to the membrane to which it is bound , both conditions that allow a related myosin to generate and sustain sub - piconewton forces ( Pyrpassopoulos et al . , 2016 ) . Actin networks grow at the plasma membrane at \u223c 50 \u2013 100 nm / s ( Kaksonen et al . , 2003 , 2005 ) , so Myo5 \u2019 s motility rate of 700 \u2013 900 nm / s ( Fig . 2 G ) , which we would expect resistance to slow only modestly , is fast enough to do work on the actin network as it assembles . We therefore expect that the myosins power membrane invagination and relieve the load to accelerate actin assembly during CME . Type I myosins are involved in a variety of membrane - reshaping events in cells , where they often interact with growing branched actin networks ( Sokac et al . , 2006 ; Almeida et al . , 2011 ; Joensuu et al . , 2014 ; Krendel et al . , 2007 ; Cheng et al . , 2012 ) , but the relative contributions of myosin motor activity and actin assembly have rarely been resolved . Here , we dem - onstrated that a type I myosin critical for CME , a process well - known to be driven by actin assembly , generates power . The implication of endocytic type I myosin as a force - insensitive motor suggests that actin assembly and myosin power genera - tion can be coordinated to do coherent work in membrane re - modeling processes . Materials and methods Reagents , proteins , and buffers ATP concentrations were determined spectrophotometrically after each experiment by absorbance at 259 nm , 2 259 = 15 , 400 M \u2212 1 cm \u2212 1 . For all ATP solutions , 1 M equivalent of MgCl 2 was included to make MgATP . Rabbit skeletal muscle actin was prepared and gel - filtered ( Spudich and Watt , 1971 ) . Actin con - centrations were determined spectrophotometrically by absor - bance at 290 nm , 2 290 = 26 , 600 M \u2212 1 cm \u2212 1 . All actin was stabilized with 1 M equivalent of phalloidin ( Sigma - Aldrich ) . Steady - state , transient , and single molecule experiments were performed at 20\u00b0C in KMg25 buffer ( 60 mM MOPS pH 7 , 25 mM KCl , 1 mM EGTA , 1 mM MgCl 2 , 1 mM DTT ) . Apyrase VII was obtained from Sigma - Aldrich . The purity and concentration of purified pro - teins were determined by comparing in - gel Coomassie blue staining to staining of known amounts of bovine serum albumin ( Pierce ) . Expression and purification of Cmd1 The S . cerevisiae calmodulin gene CMD1 was cloned from genomic DNA into a bacterial expression plasmid with a sequence en - coding His 6 - TEV situated at the 5 9 end to generate pDD2743 . pDD2743 was transformed into Rosetta E . coli , optimized for expression ( Novagen ) . A saturated overnight culture in LB ( 10 g / L Bacto tryptone , 5 g / L Bacto yeast extract , 10 g / L NaCl ) was used to inoculate a 1 - L culture in LB to OD 600 = 0 . 1 . Cells were grown to OD 600 = 0 . 6 \u2013 1 , induced with 0 . 5 mM IPTG for 5 h at 37\u00b0C , pelleted at 4 , 225 \u00d7 g for 20 min at 4\u00b0C in a Sorvall SLA - 3000 ( fixed angle ) rotor , washed with cold 20 mM HEPES pH 7 . 5 , and re - pelleted at 2 , 250 \u00d7 g for 10 min at 4\u00b0C in a Jouan CR3i ( swinging bucket ) centrifuge . Cell pellets were flash - frozen in 45 ml lysis buffer ( 20 mM HEPES pH 7 . 5 , 1 M KCl , 20 mM Im - idazole ) . Upon thawing , cells were lysed by sonication , 2 mg DNase I ( Roche ) and Triton X - 100 to 1 % were added , and the resulting lysate was incubated on ice for 30 min , then spun at 92 , 000\u00d7 g for 25 min in a Beckman Type 70 Ti rotor . The su - pernatant was loaded onto a 1 - ml HisTrap HP column ( GE healthcare ) pre - equilibrated with binding buffer ( 20 mM HEPES pH 7 . 5 , 500 mM KCl , 20 mM imidazole ) . The column was washed with 20 ml binding buffer , and Cmd1 was eluted using a 30 ml linear gradient from 0 \u2013 100 % elution buffer ( 20 mM Pedersen et al . Journal of Cell Biology 8 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 HEPES pH 7 . 5 , 500 mM KCl , 500 mM imidazole ) . Fractions containing Cmd1 were pooled , Cmd1 was cleaved from His 6 with TEV protease , and dialyzed overnight at 4\u00b0C into a low salt buffer ( 10 mM Tris pH 7 , 25 mM NaCl , 2 mM MgCl 2 , 5 mM DTT ) . Fol - lowing dialysis , purified , cleaved Cmd1 was bound to a MonoQ column and eluted using a 10 ml linear gradient from 0 \u2013 70 % high salt buffer ( 10 mM Tris pH 7 , 1 M NaCl , 2 mM MgCl 2 , 5 mM DTT ) . Fractions containing Cmd1 were pooled , dialyzed into KMg50 buffer ( 60 mM MOPS pH 7 , 50 mM KCl , 1 mM MgCl 2 , 1 mM EGTA , 1 mM DTT , 5 % glycerol ) , and stored at \u2212 80\u00b0C . Expression and purification of Myo5 Myo5 was coexpressed with the myosin chaperone She4 in S . cerevisiae . The MYO5 open reading frame ( ORF ) from S . cerevisiae was cloned from genomic DNA and truncated at Gly 763 , gener - ating a construct containing the motor domain and both Cmd1 - binding IQ motifs of the lever arm . The SHE4 ORF was cloned in its entirety from S . cerevisiae genomic DNA . Both ORFs were li - gated into a 2 - \u00b5m expression plasmid with a partially defective LEU2 gene ( leu2d ) to ensure a high copy number , creating plasmid pDD2744 ( parent vector described in Roy et al . , 2011 ) . The MYO5 ORF was situated with a sequence encoding AviTag - TEV - His 9 at the 3 9 end . Expression of the MYO5 and SHE4 ORFs was driven by a bidirectional Gal 1 / 10 promotor . pDD2744 was transformed into D1074 yeast ( Roy et al . , 2011 ) . Saturated overnight cultures in synthetic minimal medium ( 1 . 5 g / L Difacto yeast nitrogen base , 5 g / L ammonium sulfate , supplemented with 2 % glucose 20 \u00b5g / ml adenine , L - histadine , L - methionine , and 30 \u00b5g / ml L - lysine ) were used to inoculate 1 . 5 liters of cultures in the same media with raffinose substituted for glucose to OD 600 = 0 . 1 . After 18 h of growth at 30\u00b0C , cultures were induced with 2 % galactose , Bacto yeast extract was added to 10 g / L , and Bacto peptone to 20 g / L . After 8 h of expression , the cells were harvested at 4 , 225 \u00d7 g for 20 min at 4\u00b0C in a Sorvall SLA - 3000 rotor , washed with 25 ml cold Milli - Q water , re - pelleted at 2 , 250 \u00d7 g for 10 min at 4\u00b0C in a Jouan CR3i centri - fuge , resuspended in 0 . 2 vol of cold Milli - Q water , and drop frozen into liquid nitrogen . Lysis was achieved through cry - omilling ( 10 cycles of 3 min grinding with one min cooldown ) in the large vials of a 6870 freezer / mill ( SPEX Sample Prep ) . Cell powders were thawed in binding buffer ( 10 mM Tris pH 7 , 500 mM NaCl , 4 mM MgCl 2 , 2 mM ATP , 20 mM imidazole , 5 mM DTT ) supplemented with 1 mM PMSF , 1 \u00d7 cOmplete protease inhibitor cocktail without EDTA ( Roche ) , and 1 \u00b5M Cmd1 . For purification of phosphorylated Myo5 , 1 \u03bc g Pak1 ( Sigma - Aldrich , Brzeska et al . , 1997 ; Fig . S1 ) was included in the lysis buffer and 10 mM \u03b2 - glycerophosphate , 5 mM sodium py - rophosphate , and 50 mM sodium fluoride were included in all purification buffers . For the purification of unphosphorylated Myo5 , 4 , 000 units of lambda phosphatase ( NEB ) and 1 mM MnCl 2 were included in the lysis buffer . The lysate was then spun at 345 , 000 \u00d7 g for 10 min at 4\u00b0C in a Beckman TLA100 . 3 rotor , filtered through a 0 . 22 - \u00b5m filter , and loaded onto a 1 ml HisTrap HP column . The column was washed with wash buffer ( binding buffer with only 200 mM NaCl ) , and Myo5 was eluted using a 20 ml linear gradient from 0 - 100 % elution buffer ( wash buffer with 1 M imidazole ) . Fractions containing Myo5 were pooled and supplemented with Cmd1 to 1 \u00b5M . For unphosphorylated Myo5 purification , a further 20 , 000 units of lambda phosphatase were added along with MnCl 2 to 1 mM , and the fractions were incubated at 30\u00b0C for 30 min . Purified protein was dialyzed through a 3 . 5 KD MWCO membrane into 1 L storage buffer ( KMg50 with 50 % glycerol ) overnight at 4\u00b0C and again into 500 ml of the same buffer for 2 h at 4\u00b0C and then stored at \u2212 20\u00b0C . Kinetic measurements Steady - state actin - activated ATPase activity was measured us - ing the NADH enzyme - linked assay in an Applied Photophysics SX . 18 MV stopped - flow apparatus ( De La Cruz and Ostap , 2009 ) . In one reaction , the syringe contained the ATP mix ( 200 \u00b5M NADH , 20 U / ml lactic dehydrogenase , 100 U / ml pyruvate ki - nase , 500 \u00b5M phopho ( enol ) pyruvate , 2 mM MgCl 2 , 2 mM ATP in KMg25 ) and the other syringe contained the mixture of actin ( 0 \u2013 80 \u00b5M ) and myosin ( 100 nM ) in KMg25 . The concentrations above are after mixing . After mixing , the concentration of NADH loss due to ATP hydrolysis was monitored by absorbance at 340 nm ( 2 340 = 6 , 220 M \u2212 1 cm \u2212 1 ) , and the linear regions of the curve were fitted to a straight line to determine ATPase activity . ATP - induced dissociation of actoMyo5 was measured and analyzed as described ( De La Cruz and Ostap , 2009 ) . Briefly , one reaction syringe contained ATP ( 0 \u2013 2 . 7 mM ) in KMg25 and the other syringe contained 100 nM Myo5 and 100 nM phalloidin - stabilized actin . Reactants were rapidly mixed by the instru - ment , and light scattering at 90\u00b0 was acquired using a 450 - nm excitation light and a 400 - nm emission filter . Experimental transients were fit by single exponentials using the software provided with the stopped - flow apparatus . One to seven traces were averaged together to generate each data point . 0 . 04 U / ml apyrase was added to solutions of actoMyo5 before mixing to remove contaminating ADP and ATP . Unphosphorylated acto - Myo5 required prolonged treatment with apyrase to achieve sufficient signal , presumably because a larger fraction of the population was bound to ATP left over from purification and because the actin - activated ATPase rate of unphosphorylated Myo5 is slow . ADP release transients were acquired and ana - lyzed as above by preincubating 100 \u00b5M ADP with 200 nM actoMyo5 and then rapidly mixing with 2 . 5 mM ATP . The concentrations reported are after mixing . Motility assays Motility assays were carried out essentially as in Lin et al . ( 2005 ) . Double - sided Scotch tape and vacuum grease were used to create flow chambers from a clean glass coverslip ( 22 mm \u00d7 40 mm , # 1 . 5 ; Thermo Fisher Scientific ) and a glass coverslip coated with 20 \u03bc l nitrocellulose ( catalog number 11180 ; Ernest F . Fullam , Inc . ) . A mouse monoclonal antibody against His 6 ( Sigma - Aldrich ) , made with 0 . 2 mg / ml in motility buffer ( 10 mM MOPS pH 7 , 25 mM KCl , 1 mM EGTA , 1 mM MgCl 2 , 1 mM DTT ) , was first added to the flow chamber and incubated there for 5 min to coat the nitrocellulose - coated coverslip with the antibody . The flow chamber was then blocked for 2 min with 2 mg / ml casein . Blocking coverslips with bovine serum albumin ( BSA ) led to in - ferior gliding . Phosphorylated or unphosphorylated Myo5 in Pedersen et al . Journal of Cell Biology 9 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 motility buffer with 2 mg / ml casein , diluted to a range of con - centrations as indicated in Fig . 2 G , was added to the flow chamber and incubated for 2 min , then the chamber was washed once with motility buffer containing 1 mM ATP and 5 \u00b5M Cmd1 and three more times with the same buffer without ATP . Motility was initiated by loading the chambers with 5 nM rhodamine - phalloidin - labeled actin filaments in motility buffer with 1 \u00b5M ATP , 5 \u00b5M Cmd1 , 2 mg / ml casein , 0 . 4 mg / ml glucose oxidase , 0 . 08 mg / ml catalase , and 5 mg / ml glucose . Movies of actin mo - tility in the flow chambers were recorded at room temperature ( \u223c 20\u00b0C ) on a Leica DMI3000 B microscope outfitted with a 100\u00d7 , 1 . 4 NA plan apo objective and a Retiga R6 CCD camera ( Teledyne ) , controlled by Metamorph software . The rate of actin filament gliding was determined using the manual tracking plugin in Fiji . Optical trapping Flow chambers for optical trapping were constructed with double - sided tape and vacuum grease as previously described ( Snoberger et al . , 2021 ; Greenberg et al . , 2017 ) . Briefly , the coverslip was coated with a 0 . 1 % mixture of nitrocellulose and 2 . 47 \u03bc m diameter silica beads . Coverslips were dried for at least 30 min and were used within 24 h of preparation . To define the walls of the flow cell , two strips of double - sided tape were placed \u223c 5 mm apart , and a 1 - mm thick glass slide was placed on top and carefully sealed with vacuum grease after the addition of the final buffer . Trapping buffer ( KMg25 with 1 mM DTT freshly added ) was used for all trapping assays . A 100\u00d7 stock of glucose oxidase + catalase ( GOC ) was freshly prepared by centrifuging catalase ( > 30 , 000 U \u00b7 ml \u2212 1 ; Sigma - Aldrich ) at 15 , 000 \u00d7 g for 1 min and adding 2 \u03bc l of catalase supernatant to 20 \u03bc l of 19 . 1 U / \u03bc l glucose oxidase ( Sigma - Aldrich ) . 0 . 01 mg / ml anti - His 6 antibody ( Sigma - Aldrich ) was flowed in the chamber and incubated between 30 s and 3 min and then immediately blocked with 2 , 3 - min incubations of 1 \u2013 2 mg / ml BSA . Stocks of phosphorylated His 9 - tagged Myo5 were diluted to 1 nM in trapping buffer with 300 mM added KCl and incubated in the flow cell for 2 min . The number of myosins bound to the surface was limited by the surface concentration of anti - His 6 antibody and the incubation time of anti - His 6 antibody was adjusted daily between 30 s and 3 min such that one of the three to five pedestals tested showed clear myosin interactions with the actin dumbbell . Following incubation with Myo5 , a second blocking step with two 3 - min incubations of 1 \u2013 2 mg / ml BSA was performed . The final buffer added to the flow cell contained trapping buffer with the indicated amount of ATP , 1 \u03bc l of GOC added immediately prior to addition to the chamber , and 0 . 1 \u2013 0 . 25 nM rabbit skeletal muscle actin polymerized with 15 % biotinylated actin ( Cyto - skeleton ) stabilized by rhodamine - phalloidin ( Sigma - Aldrich ) at a 1 . 1 \u2013 1 . 2 molar ratio with G - actin concentration . Neutravidin - coated beads were prepared by incubating 0 . 4 ng of 0 . 8 \u03bc m di - ameter polystyrene beads ( Polysciences ) and coated with 5 mg / ml neutravidin ( Thermo Fisher Scientific ) . 3 \u03bc l of neutravidin - coated beads were added to one side of the chamber prior to sealing . All trapping data were acquired within 90 min of the addition of the final buffer to the chamber . Optical trapping experiments were performed at room tem - perature ( 20 \u00b1 1\u00b0C ) using a dual beam 1 , 064 - nm trapping laser as previously described ( Woody et al . , 2017 , 2018 ) . A single laser beam was split into two beams using polarizing beam splitters and steered into a 60\u00d7 water immersion objective ( Nikon ) . Laser light was projected through an oil immersion condenser and into quadrant photodiodes ( JQ - 50P ; Electro Optical Components , Inc . ) , each of which were conjugate to the back focal plane of the objective . Direct force detection from the quadrant photodiodes was achieved using a custom - built high - voltage reverse bias and an amplifier . Data acquisition , beam position control output , and isometric feedback calculations were controlled with custom - built virtual instruments ( Labview ; Matlab ) . Individual 0 . 8 - \u03bc m - diameter neutravidin - coated beads were caught in the two traps and held \u223c 5 \u03bc m apart . Trap stiffnesses were adjusted to 0 . 05 \u2013 0 . 1 pN / nm for each trap . A biotinylated actin filament visualized by rhodamine - phalloidin was bound to the two trapped beads , creating a bead - actin - bead dumbbell . The dumbbell was pretensioned ( 3 \u2013 5 pN ) by steering one beam using a piezo - controlled mirror conjugate to the back focal plane of the objective , and the surface of the pedestal beads was probed for myosins . Putative myosin interactions were detected via drops in the variance of the two beads , and the three - dimensional position of the dumbbell relative to the myosin was refined further by maximizing the rate and size of the observed power stroke deflections . Every 30 \u2013 60 s , the dumbbell was moved axially along the actin filament in \u223c 6 - nm steps between trace acquisition to ensure even accessibility of actin - attachment target zones . Stage drift was corrected via a feedback system using a nano - positioning stage and imaging the position of the pedestal bead with nm precision ( Woody et al . , 2017 ) . In ex - periments using 1 \u03bc M ATP , due to the longer actomyosin in - teractions , stage drift was still observed even with the stage feedback engaged , leading to a presumed underestimation of the displacement size . All data were acquired at a sampling rate of 250 kHz . Isometric optical clamping experiments were performed as previously described ( Woody et al . , 2018 ; Takagi et al . , 2006 ) using a digital feedback loop and a 1 - D electro - optical deflector ( EOD , LTA4 - Crystal ; Conoptics ) to steer the beam position using input from a high - voltage source ( Model 420 Amplifier ; Con - optics ) . Briefly , the position of one bead ( the \u201c transducer \u201d bead ) was maintained at a constant position by adjusting the position of the other bead ( referred to as the \u201c motor \u201d bead ) during ac - tomyosin interactions . The response time of the feedback loop during actomyosin interactions was \u223c 15 \u2013 30 ms . Optical trap data analysis Actomyosin interactions for non - isometric optical clamping experiments were detected using the single - molecule compu - tational tool SPASM ( Software for Precise Analysis of Single Molecules , Blackwell et al . , 2021 ) , which uses a calculation of the dumbbell bead covariances and a change - point algorithm . Data collected at 1 , 000 \u03bc M ATP were analyzed at 250 kHz , while data collected at 1 and 10 \u03bc M ATP were downsampled to 2 kHz by averaging every 125 points to enhance analysis speed . Events were detected by calculating the covariance of the two beads Pedersen et al . Journal of Cell Biology 10 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 using a smoothing window of 33 . 3 , 15 , and 5 . 25 ms and an av - eraging window 60 , 36 , and 12 ms at 1 , 10 , and 1 , 000 \u03bc M ATP , respectively . The instrument deadtime was calculated to be two times the covariance averaging window . For each 15 - s trace , the detected covariance was plotted and fit to double Gaussian dis - tributions , with the smaller mean Gaussian corresponding to the actomyosin \u201c bound \u201d portion and the larger mean Gaussian corresponding to the \u201c unbound \u201d portion of events . A putative event was defined as an event where the covariance starts above the unbound peak mean , drops below the bound peak mean , and remains below the unbound peak mean for at least the length of the instrument deadtime prior to returning back above the unbound peak mean . Event starts and ends were further refined using a changepoint algorithm as described ( Blackwell et al . , 2021 ) . Attachment durations and ensemble averages of single events were determined using built - in features in the SPASM software . Exponential fits for forward and reverse ensemble averages were performed in Origin 2019 graphing and analysis software ( OriginLab ) . Events detected in isometric optical clamping experiments were detected as described in ( Takagi et al . , 2006 ) using a zero crossing analysis via custom MATLAB scripts . Briefly , when a myosin is actively engaged with the dumbbell , force is applied to the transducer bead , a feedback loop is engaged , and opposing force is applied to the motor bead until the position of the trans - ducer bead is restored . Beginnings of events are defined at the point at which the feedback signal increases from the baseline in the motor bead and ends of events are defined when the feedback signal decreases back below the baseline in the motor bead . Online supplemental material Fig . S1 shows the results of a kinase assay demonstrating that Pak1 , used in purifications of phosphorylated Myo5 , specifically phosphorylates Myo5 serine - 357 . Video 1 shows motility assays with phosphorylated Myo5 . Video 2 shows motility assays with unphosphorylated Myo5 . Acknowledgments We thank M . Greenberg , E . Lewellyn , and A . Kunibe , each of whom played important roles at the inception of this study . We thank Y . E . Goldman for optical trap instrumentation . We thank M . Ferrin for his helpful comments on the manuscript . This work was funded by the National Institute of General Medical Sciences ( NIGMS ) grant R35 GM118149 to D . G . Drubin and grant 5R37GM057247 to E . M . Ostap . R . T . A . Pedersen is currently funded by NIGMS F32 GM142145 . Author contributions : R . T . Pedersen , A . Snoberger , S . Pyr - passopoulos , D . G . Drubin , and E . M . Ostap conceived of the ex - periments . R . T . Pedersen , A . Snoberger , S . Pyrpassopoulos , and D . Safer generated the reagents . R . T . Pedersen , A . Snoberger , S . Pyrpassopoulos , D . Safer , and E . M . Ostap performed the ex - periments and analyzed the data . R . T . Pedersen , A . Snoberger , D . G . Drubin , and E . M . Ostap wrote the manuscript . D . G . Drubin and E . M . Ostap secured funding . Disclosures : The authors declare no competing interests exist . Submitted : 23 March 2023 Revised : 17 May 2023 Accepted : 24 July 2023 References Aghamohammadzadeh , S . , and K . R . Ayscough . 2009 . Differential require - ments for actin during yeast and mammalian endocytosis . Nat . Cell Biol . 11 : 1039 \u2013 1042 . https : / / doi . org / 10 . 1038 / ncb1918 Akamatsu , M . , R . Vasan , D . Serwas , M . A . Ferrin , P . Rangamani , and D . G . Drubin . 2020 . Principles of self - organization andload adaptation by the actin cytoskeleton during clathrin - mediated endocytosis . Elife . 9 : 1 \u2013 40 . https : / / doi . org / 10 . 7554 / eLife . 49840 Almeida , C . G . , A . Yamada , D . Tenza , D . Louvard , G . Raposo , and E . Coudrier . 2011 . Myosin 1b promotes the formation of post - Golgi carriers by reg - ulating actin assembly and membrane remodelling at the trans - Golgi network . Nat . Cell Biol . 13 : 779 \u2013 789 . https : / / doi . org / 10 . 1038 / ncb2262 Basu , R . , E . L . Munteanu , and F . Chang . 2014 . Role of turgor pressure in en - docytosis in fission yeast . Mol . Biol . Cell . 25 : 679 \u2013 687 . https : / / doi . org / 10 . 1091 / mbc . E13 - 10 - 0618 Bement , W . M . , and M . S . Mooseker . 1995 . TEDS rule : A molecular rationale for differential regulation of myosins by phosphorylation of the heavy chain head . Cell Motil . Cytoskeleton . 31 : 87 \u2013 92 . https : / / doi . org / 10 . 1002 / cm . 970310202 Bieling , P . , T . D . Li , J . Weichsel , R . McGorty , P . Jreij , B . Huang , D . A . Fletcher , and R . D . Mullins . 2016 . Force feedback controls motor activity and mechanicalproperties of self - assembling branched actinnetworks . Cell . 164 : 115 \u2013 127 . https : / / doi . org / 10 . 1016 / j . cell . 2015 . 11 . 057 Blackwell , T . , W . T . Stump , S . R . Clippinger , and M . J . Greenberg . 2021 . Com - putational tool for ensemble averaging of single - molecule data . Biophys . J . 120 : 10 \u2013 20 . https : / / doi . org / 10 . 1016 / j . bpj . 2020 . 10 . 047 Boulant , S . , C . Kural , J . - C . Zeeh , F . Ubelmann , and T . Kirchhausen . 2011 . Actin dynamics counteract membrane tension during clathrin - mediated en - docytosis . Nat . Cell Biol . 13 : 1124 \u2013 1131 . https : / / doi . org / 10 . 1038 / ncb2307 Brzeska , H . , U . G . Knaus , Z . Y . Wang , G . M . Bokoch , and E . D . Korn . 1997 . p21 - activated kinase has substrate specificity similar to Acanthamoeba myosin I heavy chain kinase and activates Acanthamoeba myosin I . Proc . Natl . Acad . Sci . USA . 94 : 1092 \u2013 1095 . https : / / doi . org / 10 . 1073 / pnas . 94 . 4 . 1092 Busch , D . J . , J . R . Houser , C . C . Hayden , M . B . Sherman , E . M . Lafer , and J . C . Stachowiak . 2015 . Intrinsically disordered proteins drive membrane curvature . Nat . Commun . 6 : 7875 . https : / / doi . org / 10 . 1038 / ncomms8875 Cail , R . C . , C . R . Shirazinejad , and D . G . Drubin . 2022 . Induced nanoscale membrane curvature bypasses the essential endocytic function of clathrin . J . Cell Biol . 221 : e202109013 . https : / / doi . org / 10 . 1083 / jcb . 202109013 Carlsson , A . E . 2018 . Membrane bending by actin polymerization . Curr . Opin . Cell Biol . 50 : 1 \u2013 7 . https : / / doi . org / 10 . 1016 / j . ceb . 2017 . 11 . 007 Chen , C . , M . J . Greenberg , J . M . Laakso , E . M . Ostap , Y . E . Goldman , and H . Shu - man . 2012 . Kinetic schemes for post - synchronized single molecule dy - namics . Biophys . J . 102 : L23 \u2013 L25 . https : / / doi . org / 10 . 1016 / j . bpj . 2012 . 01 . 054 Cheng , J . , A . Grassart , and D . G . Drubin . 2012 . Myosin 1E coordinates actin assembly and cargo trafficking during clathrin - mediated endocytosis . Mol . Biol . Cell . 23 : 2891 \u2013 2904 . https : / / doi . org / 10 . 1091 / mbc . E11 - 04 - 0383 Dannhauser , P . N . , and E . J . Ungewickell . 2012 . Reconstitution of clathrin - coated bud and vesicle formation with minimal components . Nat . Cell Biol . 14 : 634 \u2013 639 . https : / / doi . org / 10 . 1038 / ncb2478 Evangelista , M . , B . M . Klebl , A . H . Tong , B . A . Webb , T . Leeuw , E . Leberer , M . Whiteway , D . Y . Thomas , and C . Boone . 2000 . A role for myosin - I in actin assembly through interactions with Vrp1p , Bee1p , and the Arp2 / 3 complex . J . Cell Biol . 148 : 353 \u2013 362 . https : / / doi . org / 10 . 1083 / jcb . 148 . 2 . 353 Geli , M . I . , and H . Riezman . 1996 . Role of type I myosins in receptor - mediated endocytosis in yeast . Science . 272 : 533 \u2013 535 . https : / / doi . org / 10 . 1126 / science . 272 . 5261 . 533 Geli , M . I . , A . Wesp , and H . Riezman . 1998 . Distinct functions of calmodulin are required for the uptake step of receptor - mediated endocytosis in yeast : the type Imyosin Myo5p isone ofthe calmodulintargets . EMBOJ . 17 : 635 \u2013 647 . https : / / doi . org / 10 . 1093 / emboj / 17 . 3 . 635 Goodson , H . V . , B . L . Anderson , H . M . Warrick , L . A . Pon , andJ . A . Spudich . 1996 . Synthetic lethality screen identifies a novel yeast myosin I gene ( MYO5 ) : myosin I proteins are required for polarization of the actin cytoskeleton . J . Cell Biol . 133 : 1277 \u2013 1291 . https : / / doi . org / 10 . 1083 / jcb . 133 . 6 . 1277 Pedersen et al . Journal of Cell Biology 11 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 Greenberg , M . J . , T . Lin , Y . E . Goldman , H . Shuman , and E . M . Ostap . 2012 . Myosin IC generates power over a range of loads via a new tension - sensing mechanism . Proc . Natl . Acad . Sci . USA . 109 : E2433 \u2013 E2440 . https : / / doi . org / 10 . 1073 / pnas . 1207811109 Greenberg , M . J . , and E . M . Ostap . 2013 . Regulation and control of myosin - I by the motor and light chain - binding domains . Trends Cell Biol . 23 : 81 \u2013 89 . https : / / doi . org / 10 . 1016 / j . tcb . 2012 . 10 . 008 Greenberg , M . J . , H . Shuman , andE . M . Ostap . 2017 . Measuringthe kinetic and mechanical properties of non - processive myosins using optical Tweez - ers . Methods Mol . Biol . 1486 : 483 \u2013 509 . https : / / doi . org / 10 . 1007 / 978 - 1 - 4939 - 6421 - 5 _ 19 Grosshans , B . L . , H . Gr\u00a8otsch , D . Mukhopadhyay , I . M . Fern\u00b4andez , J . Pfannstiel , F . - Z . Idrissi , J . Lechner , H . Riezman , and M . I . Geli . 2006 . TEDS site phosphorylation of the yeast myosins I is required for ligand - induced but not for constitutive endocytosis of the G protein - coupled receptor Ste2p . J . Biol . Chem . 281 : 11104 \u2013 11114 . https : / / doi . org / 10 . 1074 / jbc . M508933200 Hassinger , J . E . , G . Oster , D . G . Drubin , and P . Rangamani . 2017 . Design principles for robust vesiculation in clathrin - mediated endocytosis . Proc . Natl . Acad . Sci . USA . 114 : E1118 \u2013 E1127 . https : / / doi . org / 10 . 1073 / pnas . 1617705114 Idrissi , F . - Z . , A . Blasco , A . Espinal , and M . I . Geli . 2012 . Ultrastructural dy - namics of proteins involved in endocytic budding . Proc . Natl . Acad . Sci . USA . 109 : E2587 \u2013 E2594 . https : / / doi . org / 10 . 1073 / pnas . 1202789109 Idrissi , F . Z . , H . Gr\u00a8otsch , I . M . Fern\u00b4andez - Golbano , C . Presciatto - Baschong , H . Riezman , and M . I . Geli . 2008 . Distinct acto / myosin - I structures asso - ciate with endocytic profiles at the plasma membrane . J . Cell Biol . 180 : 1219 \u2013 1232 . https : / / doi . org / 10 . 1083 / jcb . 200708060 Joensuu , M . , I . Belevich , O . R\u00a8am\u00a8o , I . Nevzorov , H . Vihinen , M . Puhka , T . M . Witkos , M . Lowe , M . K . Vartiainen , and E . Jokitalo . 2014 . ER sheet persistence is coupled to myosin 1c - regulated dynamic actin filament arrays . Mol . Biol . Cell . 25 : 1111 \u2013 1126 . https : / / doi . org / 10 . 1091 / mbc . E13 - 12 - 0712 Jonsdottir , G . A . , and R . Li . 2004 . Dynamics of yeast myosin I : Evidence for a possible role in scission of endocytic vesicles . Curr . Biol . 14 : 1604 \u2013 1609 . https : / / doi . org / 10 . 1016 / j . cub . 2004 . 08 . 055 Jontes , J . D . , E . M . Wilson - Kubalek , and R . A . Milligan . 1995 . A 32 degree tail swing in brush border myosin I on ADP release . Nature . 378 : 751 \u2013 753 . https : / / doi . org / 10 . 1038 / 378751a0 Kaksonen , M . , Y . Sun , and D . G . Drubin . 2003 . A pathway for association of receptors , adaptors , andactinduring endocyticinternalization . Cell . 115 : 475 \u2013 487 . https : / / doi . org / 10 . 1016 / S0092 - 8674 ( 03 ) 00883 - 3 Kaksonen , M . , C . P . Toret , and D . G . Drubin . 2005 . A modular design for the clathrin - and actin - mediated endocytosis machinery . Cell . 123 : 305 \u2013 320 . https : / / doi . org / 10 . 1016 / j . cell . 2005 . 09 . 024 Kaplan , C . , S . J . Kenny , X . Chen , J . Sch\u00a8oneberg , E . Sitarska , A . Diz - Mu\u00f1oz , M . Akamatsu , K . Xu , and D . G . Drubin . 2022 . Load adaptation by endocytic actin networks . Mol . Biol . Cell . 33 : ar50 . https : / / doi . org / 10 . 1091 / mbc . E21 - 11 - 0589 Krendel , M . , E . K . Osterweil , and M . S . Mooseker . 2007 . Myosin 1E interacts with synaptojanin - 1 and dynamin and is involved in endocytosis . FEBS Lett . 581 : 644 \u2013 650 . https : / / doi . org / 10 . 1016 / j . febslet . 2007 . 01 . 021 De La Cruz , E . M . , and E . M . Ostap . 2009 . Kinetic and equilibrium analysis of the myosin ATPase . Methods Enzymol . 455 : 157 \u2013 192 . https : / / doi . org / 10 . 1016 / S0076 - 6879 ( 08 ) 04206 - 7 Laakso , J . M . , J . H . Lewis , H . Shuman , and E . M . Ostap . 2008 . Myosin I can act as a molecular force sensor . Science . 321 : 133 \u2013 136 . https : / / doi . org / 10 . 1126 / science . 1159419 Laakso , J . M . , J . H . Lewis , H . Shuman , andE . M . Ostap . 2010 . Controlofmyosin - I force sensing by alternative splicing . Proc . Natl . Acad . Sci . USA . 107 : 698 \u2013 702 . https : / / doi . org / 10 . 1073 / pnas . 0911426107 Lewellyn , E . B . , R . T . Pedersen , J . Hong , R . Lu , H . M . Morrison , andD . G . Drubin . 2015 . An engineered minimal WASP - myosin fusion protein reveals essential functions for endocytosis . Dev . Cell . 35 : 281 \u2013 294 . https : / / doi . org / 10 . 1016 / j . devcel . 2015 . 10 . 007 Lewis , J . H . , T . Lin , D . E . Hokanson , and E . M . Ostap . 2006 . Temperature de - pendence of nucleotide association and kinetic characterization of myo1b . Biochemistry . 45 : 11589 \u2013 11597 . https : / / doi . org / 10 . 1021 / bi0611917 Lin , T . , N . Tang , and E . M . Ostap . 2005 . Biochemical and motile properties of Myo1b splice isoforms . J . Biol . Chem . 280 : 41562 \u2013 41567 . https : / / doi . org / 10 . 1074 / jbc . M508653200 Manenschijn , H . E . , A . Picco , M . Mund , A . S . Rivier - Cordey , J . Ries , and M . Kaksonen . 2019 . Type - I myosins promote actin polymerization to drive membrane bending in endocytosis . Elife . 8 : 490011 . https : / / doi . org / 10 . 7554 / eLife . 44215 El Mezgueldi , M . , N . Tang , S . S . Rosenfeld , and E . M . Ostap . 2002 . The kinetic mechanism of Myo1e ( human myosin - IC ) . J . Biol . Chem . 277 : 21514 \u2013 21521 . https : / / doi . org / 10 . 1074 / jbc . M200713200 Mogilner , A . , and G . Oster . 1996 . Cell motility driven by actin polymerization . Biophys . J . 71 : 3030 \u2013 3045 . https : / / doi . org / 10 . 1016 / S0006 - 3495 ( 96 ) 79496 - 1 Mogilner , A . , andG . Oster . 2003 . Forcegeneration by actin polymerization II : The elastic ratchet and tethered filaments . Biophys . J . 84 : 1591 \u2013 1605 . https : / / doi . org / 10 . 1016 / S0006 - 3495 ( 03 ) 74969 - 8 Mund , M . , J . A . van der Beek , J . Deschamps , S . Dmitrieff , P . Hoess , J . L . Monster , A . Picco , F . N\u00b4ed\u00b4elec , M . Kaksonen , and J . Ries . 2018 . Sys - tematic nanoscale analysis of endocytosis links efficient vesicle for - mation to patterned actin nucleation . Cell . 174 : 884 \u2013 896 . e17 . https : / / doi . org / 10 . 1016 / j . cell . 2018 . 06 . 032 Nickaeen , M . , J . Berro , T . D . Pollard , and B . M . Slepchenko . 2019 . Actin as - sembly produces sufficient forces for endocytosis in yeast . Mol . Biol . Cell . 30 : 2014 \u2013 2024 . https : / / doi . org / 10 . 1091 / mbc . E19 - 01 - 0059 Pedersen , R . T . A . , and D . G . Drubin . 2019 . Type I myosins anchor actin as - sembly to the plasma membrane during clathrin - mediated endocytosis . J . Cell Biol . 218 : 1138 \u2013 1147 . https : / / doi . org / 10 . 1083 / jcb . 201810005 Picco , A . , M . Mund , J . Ries , F . N\u00b4ed\u00b4elec , and M . Kaksonen . 2015 . Visualizing the functional architecture of the endocytic machinery . Elife . 4 : 1 \u2013 29 . https : / / doi . org / 10 . 7554 / eLife . 04535 Pyrpassopoulos , S . , G . Arpa \u02d8 g , E . A . Feeser , H . Shuman , E . T\u00fczel , and E . M . Ostap . 2016 . Force generation by membrane - associated myosin - I . Sci . Rep . 6 : 25524 . https : / / doi . org / 10 . 1038 / srep25524 Roy , M . A . , N . Siddiqui , and D . D \u2019 Amours . 2011 . Dynamic and selective DNA - binding activity of Smc5 , a core component of the Smc5 - Smc6 complex . Cell Cycle . 10 : 690 \u2013 700 . https : / / doi . org / 10 . 4161 / cc . 10 . 4 . 14860 Skruzny , M . , T . Brach , R . Ciuffa , S . Rybina , M . Wachsmuth , andM . Kaksonen . 2012 . Molecular basis for coupling the plasma membrane to the actin cytoskeletonduring clathrin - mediated endocytosis . Proc . Natl . Acad . Sci . USA . 109 : E2533 \u2013 E2542 . https : / / doi . org / 10 . 1073 / pnas . 1207011109 Snoberger , A . , B . Barua , J . L . Atherton , H . Shuman , E . Forgacs , Y . E . Goldman , D . A . Winkelmann , and E . M . Ostap . 2021 . Myosin with hypertrophic cardiac mutation R712L has a decreased working stroke which is res - cued by omecamtiv mecarbil . Elife . 10 : 1 \u2013 24 . https : / / doi . org / 10 . 7554 / eLife . 63691 Sokac , A . M . , C . Schietroma , C . B . Gundersen , and W . M . Bement . 2006 . Myosin - 1c couples assembling actin to membranes to drive compen - satory endocytosis . Dev . Cell . 11 : 629 \u2013 640 . https : / / doi . org / 10 . 1016 / j . devcel . 2006 . 09 . 002 Spudich , J . A . , and S . Watt . 1971 . The regulation of rabbit skeletal muscle contraction . I . Biochemical studies of the interaction of the tropomyosin - troponin complex with actin and the proteolytic frag - ments of myosin . J . Biol . Chem . 246 : 4866 \u2013 4871 . https : / / doi . org / 10 . 1016 / S0021 - 9258 ( 18 ) 62016 - 2 Stewart , T . J . , V . Murthy , S . P . Dugan , and J . E . Baker . 2021 . Velocity of myosin - basedactin sliding dependsonattachment and detachment kinetics and reaches a maximum when myosin - binding sites on actin saturate . J . Biol . Chem . 297 : 101178 . https : / / doi . org / 10 . 1016 / j . jbc . 2021 . 101178 Sun , Y . , A . C . Martin , and D . G . Drubin . 2006 . Endocytic internalization in budding yeast requires coordinated actin nucleation and myosin motor activity . Dev . Cell . 11 : 33 \u2013 46 . https : / / doi . org / 10 . 1016 / j . devcel . 2006 . 05 . 008 Sun , Y . , J . Sch\u00a8oneberg , X . Chen , T . Jiang , C . Kaplan , K . Xu , T . D . Pollard , and D . G . Drubin . 2019 . Direct comparison of clathrin - mediated endocytosis in budding and fission yeast reveals conserved and evolvable features . Elife . 8 : e50749 . https : / / doi . org / 10 . 7554 / eLife . 50749 Takagi , Y . , E . E . Homsher , Y . E . Goldman , and H . Shuman . 2006 . Force gen - eration in single conventional actomyosin complexes under high dy - namic load . Biophys . J . 90 : 1295 \u2013 1307 . https : / / doi . org / 10 . 1529 / biophysj . 105 . 068429 Veigel , C . , L . M . Coluccio , J . D . Jontes , J . C . Sparrow , R . A . Milligan , and J . E . Molloy . 1999 . The motor protein myosin - I produces its working stroke in two steps . Nature . 398 : 530 \u2013 533 . https : / / doi . org / 10 . 1038 / 19104 Woody , M . S . , M . J . Greenberg , B . Barua , D . A . Winkelmann , Y . E . Goldman , and E . M . Ostap . 2018 . Positive cardiac inotrope omecamtiv mecarbil acti - vates muscle despite suppressing the myosin working stroke . Nat . Commun . 9 : 3838 . https : / / doi . org / 10 . 1038 / s41467 - 018 - 06193 - 2 Woody , M . S . , J . H . Lewis , M . J . Greenberg , Y . E . Goldman , andE . M . Ostap . 2016 . MEMLET : An easy - to - use tool for data fitting and model comparison using maximum - likelihood estimation . Biophys . J . 111 : 273 \u2013 282 . https : / / doi . org / 10 . 1016 / j . bpj . 2016 . 06 . 019 Woody , M . S . , E . M . Ostap , Y . E . Goldman , and M . Capitanio . 2017 . An ultra - fast EOD - based force - clamp detects rapid biomechanical transitions . SPIE - Intl Soc Optical Eng . 26 . Pedersen et al . Journal of Cell Biology 12 of 12 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 Supplemental material Figure S1 . P21 - activated Kinase 1 ( Pak1 ) phosphorylates Myo5 on S357 . Crude preparations of wild type and S357A Myo5 motor / lever constructs were mixed with 250 \u00b5M ATP including 20 \u00b5Ci of ATP \u03b3 P32 in kinase assay buffer ( 5 mM MOPS pH 7 , 2 . 5 mM \u03b2 - glycerophosphate , 5 mM MgCl 2 , 400 \u00b5M EDTA , 1 mM EGTA , and 50\u00b5M DTT ) in eitherthe presenceor absence of Pak1 . Reactions were incubated at 25\u00b0C for 60min , thenquenchedby adding an equal volumeof 2\u00d7 Tris urea sample buffer ( 125 mM Tris pH 6 . 8 , 6 M urea , 2 % SDS , 0 . 1 % bromophenol blue , 10 % \u03b2 - mercaptoethanol ) and resolved on a 10 % polyacrylamide gel . The gel was stained with Coomassie , then dried onto Whatman paper and exposed to a storage phosphor screen ( Amersham ) . The Coomassie - stained gel was imaged on a standard photo scanner and the phosphor screen on a Typhoon gel imager ( Amersham ) . Note that there are differences in baseline labeling in the absence of added kinase between the two different protein preps , but the addition of Pak1 clearly results in radiolabeling of wild type but not mutant Myo5 . Pedersen et al . Journal of Cell Biology S1 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023 Video 1 . Motility assays with phosphorylated Myo5 . Rhodamine - phalloidin - labeled actin filaments gliding over coverslips coated with a concentration series of phosphorylated Myo5 protein in motility buffer with 1 mM ATP . Videos were collected at one frame per second and are played back at 16 frames per second . Video 2 . Motility assays with unphosphorylated Myo5 . Rhodamine - phalloidin - labeled actin filaments gliding over coverslips coated with a concentration series of unphosphorylated Myo5 protein in motility buffer with 1 mM ATP . Short movies of motility at 100 and 150 nMunphosphorylated Myo5 were collected because no motility was observed . Movies at all other concentrations were collected at one frame every 4 s and are played back at 16 frames per second . The playback rate of Video 2 is four times faster than the playback rate of Video 1 . Pedersen et al . Journal of Cell Biology S2 An endocytic myosin - 1 generates power https : / / doi . org / 10 . 1083 / jcb . 202303095 D o w n l oaded f r o m h t t p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 222 / 10 / e202303095 / 1916174 / j c b _ 202303095 . pd f b y gue s t on 18 A ugu s t 2023",
    "kosmalska2015physical_supplement": "Supplementary figures Supplementary Figure 1 . Stretch system . a , Schematic of stretch system . A ring clamping a PDMS membrane ( shown in red ) is placed on a support containing a central loading circular post and an external ring , with an opening in between . Imaging objectives can then be placed either below ( inverted microscope ) or above ( upright microscope ) . Cells are then cultured on the membrane after coating with fibronectin . b , Once vacuum is applied through the opening , it deforms and stretches the membrane . In some cases , cells were seeded on polyacrylamide and soft silicone gels previously attached to the membrane ( see methods ) . c , Level of strain in the X and Y axis obtained after applying different vacuum suction pressures . Stretch was biaxial . Supplementary Figure 2 . Response of cells to 2 % strain . Response of a peYFP - mem transfected cell to the application of 2 % strain for three minutes . No visible effects were observed either upon stretch application or upon stretch release . Insets display magnifications of the areas in red , showing that membrane ruffles were not flattened upon stretch application , and reservoirs were not formed upon stretch release . Scale bar is 20 \u03bcm . Supplementary Figure 3 . Dynamic formation and resorption of membrane structures . a , Quantification of VLD fluorescence after re - application of iso - osmotic medium ( 1 : maximum fluorescence , 0 : background ) . n = 20 VLDs from 2 cells . Imaging was carried out using an inverted microscope ( 60x objective ) which allowed to visualize the initial VLD formation period . b , Quantification of reservoir fluorescence after stretch release ( 1 : initial fluorescence , 0 : background ) . n = 100 / 80 / 30 reservoirs from 10 / 3 / 3 cells . Imaging was carried out using an upright microscope ( 60x objective ) . Both decreasing temperature ( pink symbols ) and increasing stretch magnitude ( black symbols ) slowed the process of reservoir formation . Images to the right show corresponding examples of membrane structures at the times indicated in the graph . c , Quantification of VLD fluorescence after re - application of iso - osmotic medium ( 1 : initial fluorescence , 0 : background ) . n = 100 / 50 / 60 VLDs from 10 / 5 / 4 cells . Imaging was carried out using an upright microscope ( 60x objective ) . Both decreasing temperature ( pink symbols ) and increasing the magnitude of the hypo - osmotic shock ( black symbols ) slowed the process of VLD formation . Images to the right show corresponding examples of membrane structures at the times indicated in the graph . All insets have a size of 10x10 \u03bcm . Supplementary Figure 4 . Co - localization of actin and membrane structures before , during and after stretch and hypo - osmotic shocks . a , cells transfected with pEYFP - mem and Lifeact - Ruby before , during , and after application of 6 % stretch for 3 minutes . b , equivalent images obtained before , during , and after application of a 50 % hypo - osmotic shock for 3 minutes . Insets ( 10x10 \u03bcm ) show zoomed views of membrane and actin structures . Scale bars are 20\u03bcm . Supplementary Figure 5 . Poroelasticity of polyacrylamide gels . a , graph showing the height of polyacrylamide gels during and after application of 6 % stretch for three minutes . Values are shown normalized to the height before stretch application . As gels are stretched , they first decrease their height to maintain volume , but they progressively incorporate water and swell . Once stretch is released , the progressive reduction in thickness is indicative of water expulsion flow . b , As a control , gels submitted to a 50 % decrease in media osmolarity did not modify their height . ( n = 5 gels ) Supplementary Figure 6 . Reservoirs and VLDs are unrelated to caveolae . a , Time sequence of pEYFP - mem and Cav1 - mcherry transfected cells before , during , and after application of 6 % stretch for three minutes . Image to the right shows the lack of co - localization between pEYFP - mem ( green ) and Cav1 - mcherry ( red ) . b , pEYFP - mem transfected cell ( green in merged image ) fixed right after restoring iso - osmotic media and stained for caveolin 1 ( red in merged image ) . No co - localization was observed between caveolin and VLDs . c , Time sequence of pEYFP - mem and Cav1 - mcherry transfected cells before , during , and after application of 50 % hypo - osmotic media for three minutes . Image to the right shows the lack of co - localization between peYFP - mem ( green ) and Cav1 - mcherry ( red ) . d , pEYFP - mem transfected cell ( green in merged image ) fixed right after releasing 6 % stretch and stained for caveolin 1 ( red in merged image ) . No co - localization was observed between caveolin and reservoirs . e , Time sequence of pEYFP - mem transfected cells before , during , and after application of 50 % hypo - osmotic media during three minutes . Zoomed insets show the formation and evolution of VLDs . Caveolin 1 knock - out cells were reconstituted either with caveolin 1 - GFP or an empty control vector ( IRES - GFP ) . f , Quantification of VLD fluorescence after re - application of iso - osmotic media ( 1 : initial fluorescence , 0 : background ) . n = 50 / 25 VLDs from 5 / 3 cells . g - h , Corresponding quantification of mean VLD diameter ( g ) ( n = 100 / 60 VLDs from 5 / 3 cells ) and density ( h ) ( n = 30 / 20 regions from 5 / 3 cells ) . No significant differences were observed . i , Time sequence of peYFP - mem transfected cells before , during , and after application of constant 6 % stretch during three minutes . Zoomed insets show the formation and evolution of membrane reservoirs . Caveolin 1 knock - out cells were reconstituted either with caveolin 1 - GFP or an empty control vector ( IRES - GFP ) . j , Quantification of reservoir fluorescence after stretch release ( 1 : initial fluorescence , 0 : background ) ( n = 30 / 60 reservoirs from 3 / 4 cells ) . k , Corresponding quantification of mean reservoir diameter ( n = 250 / 100 reservoirs from 3 / 3 cells ) . No significant differences were observed . l , Corresponding quantification of mean reservoir density ( n = 40 / 20 regions from 3 / 3 cells ) . No significant differences were observed . Scale bars are 20\u03bcm . Supplementary Figure 7 . Actin and temperature dependence of reservoirs and VLDs . a , Time sequence of peYFP - mem transfected cells before , during , and after application of constant 6 % stretch during three minutes for control cells and cells treated with 0 . 5 \u03bcM cytochalasin D . Zoomed insets show the formation and evolution of membrane reservoirs . b , Quantification of reservoir fluorescence after stretch release ( 1 : initial fluorescence , 0 : background ) ( n = 100 / 100 reservoirs from 10 / 4 cells ) . c , Corresponding quantification of mean reservoir diameter ( n = 250 / 100 reservoirs from 8 / 4 cells ) . No significant differences were observed . d , Corresponding quantification of mean reservoir density ( n = 30 / 30 regions from 5 cells ) . e , Time sequence of peYFP - mem transfected cells before , during , and after application of 50 % hypo - osmotic media during three minutes for control cells and cells treated with 0 . 5 \u03bcM cytochalasin D . Zoomed insets show the formation and evolution of membrane VLDs . f , Quantification of VLD fluorescence after re - application of iso - osmotic media ( 1 : initial fluorescence , 0 : background ) . n = 100 / 100 VLDs from 10 / 5 cells . g , Corresponding quantification of mean VLD diameter ( n = 100 / 60 VLDs from 10 / 3 cells ) . h , Corresponding quantification of mean VLD density ( n = 50 / 30 regions from 8 / 3 cells ) . i , Time sequence of peYFP - mem transfected cells before , during , and after application of constant 6 % stretch during three minutes for control cells at 37\u00ba and cells at 26\u00ba . Zoomed insets show the formation and evolution of membrane reservoirs . j , Quantification of reservoir fluorescence after stretch release ( 1 : initial fluorescence , 0 : background ) n = 100 / 60 reservoirs from 10 / 3 cells . k , Corresponding quantification of mean reservoir diameter ( n = 250 / 140 reservoirs from 8 / 3 cells ) . No significant differences were observed . l , Corresponding quantification of mean reservoir density ( n = 30 / 30 zones from 5 / 3cells ) . No significant differences were observed . m , Time sequence of peYFP - mem transfected cells before , during , and after application of 50 % hypo - osmotic media during three minutes for control cells at 37\u00ba and cells at 26\u00ba . Zoomed insets show the formation and evolution of membrane VLDs . n , Quantification of VLD fluorescence after re - application of iso - osmotic media ( 1 : initial fluorescence , 0 : background ) . n = 100 / 24 VLDs from 10 / 3 cells . o , Corresponding quantification of mean VLD diameter ( n = 100 / 45 VLDs from 10 / 3 cells ) . No significant differences were observed . p , Corresponding quantification of mean VLD density ( n = 50 / 25 zones from 8 / 3 cells ) . No significant differences were observed . Scale bars indicate 20 \u03bcm . N . s . , non - significant , * , p < 0 . 05 , * * * , p < 0 . 001 . Scale bars are 20 \u03bcm . Supplementary Figure 8 . Membrane reservoirs and VLDs are observed across cell types and species . Images of Chinese Hamster Ovary ( CHO ) cells , A431 human squamous carcinoma cells , and human keratinocytes ( HaCaT ) transfected with pEYFP - mem before and after being submitted to biaxial stretch ( a ) or hypo - osmotic shock ( b ) . All cell types consistently showed the formation of reservoirs or VLDs , respectively . Zoomed insets ( 10x10 \u03bcm ) show a magnification of both membrane structures . Scale bar indicates 20 \u03bcm . Supplementary Figure 9 . Adhesion dependence of reservoirs and VLDs . a , Time sequence of pEYFP - mem transfected cells before , during , and after application of constant 6 % stretch during three minutes for control cells and cells treated with 10 \u00b5g mL - 1 \u03b15\u03b21 antibody . b , Quantification of reservoir fluorescence after stretch release ( 1 : initial fluorescence , 0 : background ) ( n = 100 / 150 reservoirs from 10 / 6 cells ) . c , Corresponding quantification of mean reservoir diameter ( n = 250 / 200 reservoirs from 8 / 6 cells ) . d , Corresponding quantification of mean reservoir density ( n = 30 / 60 regions from 5 / 5 cells ) . e , Time sequence of pEYFP - mem transfected cells before , during , and after application of 50 % hypo - osmotic media during three minutes for control cells and cells treated with 10 \u00b5g mL - 1 \u03b15\u03b21 antibody . f , Quantification of VLD fluorescence after re - application of iso - osmotic media ( 1 : initial fluorescence , 0 : background ) . n = 100 / 40 VLDs from 10 / 5 cells . g , Corresponding quantification of mean VLD diameter ( n = 100 / 150 VLDs from 10 / 7 cells ) . h , Corresponding quantification of mean VLD density ( n = 50 / 70 regions from 8 / 7 cells ) . Scale bars indicate 20 \u03bcm . N . s . , non - significant , * , p < 0 . 05 , * * * , p < 0 . 001 . Scale bars are 20 \u03bcm . In all cases , Zoomed insets ( 10x10 \u03bcm ) show the formation and evolution of membrane structures . Supplementary Figure 10 . Model predictions . a , Model prediction for reservoir length upon stretch release as a function of applied stretch . b , model prediction for VLD diameters formed after restoring iso - osmotic medium from osmotic shocks of different magnitude . In a and b , red line corresponds to the bending modulus used throughout the work , dashed black line shows the prediction for a 5 - fold increase in bending modulus . c , VLD shape after restoring osmolarity from hypo - osmotic shocks of different magnitude and d , VLD shape after restoring osmolarity from a 50 % hypo - osmotic shock in a pre - stretched membrane ( top ) and then de - stretching the membrane by 6 % ( top ) . Supplementary Notes Supplementary Note 1 \u2013 Dynamics of reservoir and VLD formation . Throughout the figures , images of stretch - induced reservoirs and osmolarity - induced VLDs generally show fully formed structures at the first frame after their appearance . In the case of reservoirs , this is because their formation was faster than the few seconds required to re - position and re - focus cells after being moved by the stretch device . In the case of VLDs , a few seconds to re - focus after changing the medium were also required . This is because we used an upright microscope with a water - immersion objective that imaged through the medium , which lost focus after medium changes . Whereas the time lag required after stretch operations was unavoidable , we could eliminate the lag required after osmolarity changes by imaging with an inverted microscope . In this case , we observed a VLD formation time of approximately 20 s ( Supplementary Figure 3a ) . Interestingly , the formation period of both reservoirs and VLDs became observable even with the upright microscope after either increasing the magnitude of stretch / hypo - osmotic shock or after decreasing cell activity by reducing temperature ( Supplementary Figure 3b - c ) . This shows that the dynamics of VLD / reservoir formation are governed both by the magnitude of the applied stimulus and the active ability of the cell to recover its original state . Supplementary Note 2 \u2013 Driving forces behind VLD formation and maintenance . It is important to distinguish that VLDs form due to the need to store water volume at the cell - substrate interface , not due to changes in the volume of cells themselves . In earlier work , VLDs had been hypothesized to initiate as membrane invaginations driven by cellular shrinking , then enlarge due to hydrostatic pressure transients , and finally become stabilized by a spectrin cytoskeletal lining 1 , 2 , 3 . However , we believe that hydrostatic pressure of confined water at the cell - substrate interface accounts for all stages of the VLD cycle for different reasons . First , exposure to 50 % osmotic shocks for 3 minutes only generates small membrane compressions of approximately 2 % , which do not form any visible membrane invaginations ( Supplementary Figure 2 ) . Second , elimination of water confinement by using polyacrylamide gels abrogates VLD formation altogether . Finally , decreasing osmolarity after VLD formation collapses VLDs ( Fig . 7e ) , showing that water flows are essential to their maintenance . Similarly , VLDs in cells with ATP depletion ( Fig . 5e - f ) , cytochalasin D treatment ( Supplementary Figure 7e - f ) , or reduced temperature ( Supplementary Figure 7m - n ) gradually collapse , but do not disappear entirely . This suggests that progressive flow from VLDs to the exterior bath reduces pressure and collapses the structure , while active resorption of membrane accumulation is impaired by the different treatments . Thus , hydrostatic pressure driven by confined water generates and maintains VLDs . A detailed analysis of the process would nevertheless require the consideration of the specific and non - specific adhesion between cells and their substrate , and the complex dynamics of confined flows through this interface . Another important aspect to clarify is that the formation of VLDs , while driven by water volume storage , also requires the recruitment of membrane area . This is exemplified by the collapsed VLDs observed after either decreasing osmolarity ( Fig . 7e ) , depleting ATP ( Fig . 5e - f ) , treating with cytochalasin D treatment ( Supplementary Figure 7e - f ) , or reducing temperature ( Supplementary Figure 7m - n ) . In fact , collapsed VLDs are membrane accumulations akin to reservoirs , which like reservoirs can be eliminated through application of stretch ( Fig . 7e ) . Supplementary Note 3 \u2013 Model assumptions , parameters , and predictions . We compare the experimental observations with the predictions of the model described in detail in 4 . In this model , a circular membrane patch with one protrusion at its center is considered , and the optimal shape is obtained by minimizing the energy of the system , consisting of elastic ( stretching and bending ) and adhesive contributions . Depending on the excess membrane area , given by the applied strain , and the interstitial volume , or alternatively the osmotic gradient across the membrane , various optimal shapes emerge , ranging from shallow caps when very little membrane area is available to very thin tubules when very little interstitial volume is available . The nominal separation between membrane structures is modeled through the size of this circular patch , and can be obtained from experiments . For instance , consistent with a density of between 1 and 2 VLDs per 100 \u03bcm 2 , we choose a domain size of 5 \u03bcm . The higher density of reservoirs is modeled with a domain size of 1 . 25 \u03bcm . For the adhesion energy , we use a cell - measured value of 4 mJ / m 2 , reflecting the adhesion strength of a cell membrane to an RGD - coated substrate 5 . We use the same parameters as in 4 for the lipid bilayer stretching ( K s = 0 . 12 J / m 2 ) and bending moduli ( k = 10 - 19 J ) . We note that unlike the model bilayer membranes in 4 , cell membranes are lined with an actin cortex . This cortex has a stretching modulus of the order of 10 - 4 J / m 2 , as obtained from a Young\u2019s modulus and thickness of the order of 10 3 Pa and 100 nm , respectively 6 , 7 . This is three orders of magnitude below the value of the membrane itself , indicating that the contribution of the cortex to the overall stretching modulus is negligible . In contrast , the cortex would be expected to significantly increase membrane bending modulus . Indeed , it has been shown that cells with impaired membrane - actin links have a membrane bending stiffness comparable to that of lipid bilayers , whereas this bending stiffness increases 5 - fold in cells with a normal membrane - cortex connection 8 . However , a 5 - fold increase in bending stiffness did not significantly alter model predictions of the experimental observables , such as tube length or VLD diameter ( Supplementary Figure 10a - b ) . Given the very high membrane stretching modulus and the strong osmolarity differences across the bilayer , the main factors determining the shape of tubes or VLDs were simply the excess area ( compressive strain ) and the volume enclosed in the interstitial space . Consistently with a negligible effect of bending stiffness , depolymerizing the actin cytoskeleton with cytochalasin did not impair reservoir or VLD formation ( Supplementary Figure 7e and 1 ) , and barely affected reservoir diameter ( Supplementary Figure 7c ) . The more marked effect of cytochalasin on VLDs ( reduction in diameter of 30 % , and a 2 - fold increase in density , Supplementary Figure 7g ) is probably due not to the cortex but to stress fibers and focal adhesions , through which VLDs have to make their way ( Supplementary Figure 4 ) . Cytochalasin treatment would reduce cytoskeletal resistance from focal adhesions and stress fibers , leading to the generation of many small VLDs throughout the cell . In contrast , increased cytoskeletal resistance in control cells would allow VLD formation only in specific sites with low resistance , which would then have to accommodate a larger water volume . The only parameter that we adjust to fit our experimental results is the equilibrium separation of the adhesion potential , which in 4 was set to 3 nm . Here , best results are obtained with a value of 30 nm , reflecting the additional separation likely provided by integrins ( extending approximately 20 nm from the membrane 9 ) and the fibronectin coat . With these parameters , the model closely replicates the observations . For instance , the model predicts tubule length to increase linearly with applied destretch , see Fig . 6b . The VLD diameter predicted by the model for a 2 % excess area and various degrees of osmolarity reduction from baseline osmolarity ( M 0 = 300 mOsm ) agrees well with the experimental quantification in Fig . 6c ( see Supplementary Figure 10a ) . Mimicking Fig . 7g , j , increasing the membrane area by de - stretch after VLD formation in pre - stretched cells results in bud - like protrusions with a small neck and a slightly smaller apparent diameter , which store the same volume as the hemispherical VLDs but significantly more membrane area , see Supplementary Figure 10b . In agreement with Fig . 5p , we find that a three - fold increase in the density of VLDs ( modelled through a corresponding decrease in the size of the membrane patch ) results in a decrease of VLD diameter from 2 . 2 \u03bcm to 1 . 6 \u03bcm . Supplementary references 1 . Herring TL , Cohan CS , Welnhofer EA , Mills LR , Morris CE . F - actin at newly invaginated membrane in neurons : implications for surface area regulation . J Membr Biol 1999 , 171 ( 2 ) : 151 - 169 . 2 . Herring TL , Juranka P , McNally J , Lesiuk H , Morris CE . The spectrin skeleton of newly - invaginated plasma membrane . J Muscle Res Cell Motil 2000 , 21 ( 1 ) : 67 - 77 . 3 . Morris CE , Homann U . Cell surface area regulation and membrane tension . J Membr Biol 2001 , 179 ( 2 ) : 79 - 102 . 4 . Staykova M , Arroyo M , Rahimi M , Stone HA . Confined Bilayers Passively Regulate Shape and Stress . Physical Review Letters 2013 , 110 ( 2 ) : 028101 . 5 . Rico F , Roca - Cusachs P , Sunyer R , Farre R , Navajas D . Cell dynamic adhesion and elastic properties probed with cylindrical atomic force microscopy cantilever tips . Journal of Molecular Recognition 2007 , 20 ( 6 ) : 459 - 466 . 6 . Clark AG , Dierkes K , Paluch EK . Monitoring actin cortex thickness in live cells . Biophys J 2013 , 105 ( 3 ) : 570 - 580 . 7 . Salbreux G , Charras G , Paluch E . Actin cortex mechanics and cellular morphogenesis . Trends Cell Biol 2012 , 22 ( 10 ) : 536 - 545 . 8 . Simson R , Wallraff E , Faix J , Niewohner J , Gerisch G , Sackmann E . Membrane bending modulus and adhesion energy of wild - type and mutant cells of Dictyostelium lacking talin or cortexillins . Biophys J 1998 , 74 ( 1 ) : 514 - 522 . 9 . Eng ET , Smagghe BJ , Walz T , Springer TA . Intact \u03b1IIb\u03b23 Integrin Is Extended after Activation as Measured by Solution X - ray Scattering and Electron Microscopy . J Biol Chem 2011 , 286 ( 40 ) : 35218 - 35226 .",
    "vojnovic2024combining": "royalsocietypublishing . org / journal / rsob Research Cite this article : Vojnovic I , Caspari OD , Ho \u015f kan MA , Endesfelder U . 2024 Combining single - molecule and expansion microscopy in fission yeast to visualize protein structures at the nanostructural level . Open Biol . 14 : 230414 . https : / / doi . org / 10 . 1098 / rsob . 230414 Received : 8 November 2023 Accepted : 4 December 2023 Subject Area : cellular biology / microbiology / structural biology / molecular biology / biophysics Keywords : single - molecule localization microscopy , photoactivated localization microscopy , expansion microscopy , Schizosaccharomyces pombe , protein retention yield , correlative expansion microscopy Author for correspondence : Ulrike Endesfelder e - mail : endesfelder @ uni - bonn . de \u2020 Present address : Institut f\u00fcr Mikrobiologie und Biotechnologie , Rheinische Friedrich - Wilhelms - Universit\u00e4t , Bonn , Germany . \u2021 These authors contributed equally to this study . \u00b6 Present address : Department of Medical Microbiology and Immunology , University of Wisconsin - Madison , Madison , WI , USA . One contribution to a Special Feature \u2018 Bioimaging \u2019 . Electronic supplementary material is available online at https : / / doi . org / 10 . 6084 / m9 . figshare . c . 7007803 . Combining single - molecule and expansion microscopy in fission yeast to visualize protein structures at the nanostructural level Ilijana Vojnovic 1 , 2 , \u2020 , \u2021 , Oliver D . Caspari 1 , 3 , \u2020 , \u2021 , Mehmet Ali Ho \u015f kan 1 , \u00b6 and Ulrike Endesfelder 1 , 2 , \u2020 1 Department of Systems and Synthetic Microbiology , Max Planck Institute for Terrestrial Microbiology and LOEWE Center for Synthetic Microbiology ( SYNMIKRO ) , Marburg , Germany 2 Department of Physics , Carnegie Mellon University , Pittsburgh , PA , USA 3 Department of Microbiology , Institute Pasteur , Paris , France ODC , 0000 - 0001 - 8235 - 0503 ; UE , 0000 - 0002 - 7801 - 6278 In this work , we have developed an expansion microscopy ( ExM ) protocol that combines ExM with photoactivated localization microscopy ( ExPALM ) for yeast cell imaging , and report a robust protocol for single - molecule and expansion microscopy of fission yeast , abbreviated as SExY . Our optimized SExY protocol retains about 50 % of the fluorescent protein signal , doubling the amount obtained compared to the original protein retention ExM ( proExM ) protocol . It allows for a fivefold , highly isotropic expansion of fis - sion yeast cells , which we carefully controlled while optimizing protein yield . We demonstrate the SExY method on several exemplary molecular tar - gets and explicitly introduce low - abundant protein targets ( e . g . nuclear proteins such as cbp1 and mis16 , and the centromere - specific histone protein cnp1 ) . The SExY protocol optimizations increasing protein yield could be ben - eficial for many studies , when targeting low abundance proteins , or for studies that rely on genetic labelling for various reasons ( e . g . for proteins that cannot be easily targeted by extrinsic staining or in case artefacts introduced by unspecific staining interfere with data quality ) . 1 . Introduction The resolution of optical microscopy is constrained by the diffraction limit of light at about 200 nm [ 1 ] . However , most of molecular organization occurs in dimensions below this limit and therefore cannot be resolved by conventional light microscopic techniques . Super - resolution fluorescence microscopy ( SRM ) can be used to circumvent the diffraction limit by modulating the fluorescence of fluorescently labelled molecules at a sub - diffraction level , thereby discrimi - nating between them [ 2 ] . This is achieved either by precisely defined illumination patterns ( e . g . in STED [ 3 ] and SIM [ 4 ] imaging ) or by single - mol - ecule localization microscopy ( SMLM ) methods , in which the fluorescence of individual , on - and off - blinking or on - and off - binding fluorophores is separ - ated in time ( e . g . in PALM [ 5 ] , ( d ) STORM [ 6 , 7 ] or PAINT [ 8 ] techniques ) . More recently , expansion microscopy ( ExM ) techniques have been developed that physically expand the biological sample , increasing resolution by practi - cally a factor of 3 \u2013 20 - fold as a result of inflating the structures [ 9 \u2013 13 ] . In contrast to the \u2018 classical \u2019 SRM methods , where preparing a good sample mostly relies on choosing the right fluorophores ( e . g . bright ( all ) and either photostable ( SIM , STED ) or well - controlled in photoswitching ( SMLM ) ) and optimizing for efficient and specific labelling ( e . g . small labels for dense label - ling , non - sticking labels for high specificity or genetic , covalent or high affinity labels for high efficiency ) [ 2 ] , ExM poses additional demands to the sample preparation due to the desired physical expansion : in ExM , the sample is \u00a9 2024 The Authors . Published by the Royal Society under the terms of the Creative Commons Attribution License http : / / creativecommons . org / licenses / by / 4 . 0 / , which permits unrestricted use , provided the original author and source are credited . D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 anchored to a gel mesh and expanded upon incubation in aqueous media [ 9 ] . To achieve an isotropic expansion of the sample and thus a preservation of the underlying biological ultrastructure , all physical connections within the structure must be efficiently removed by , e . g . cell wall removal [ 14 \u2013 17 ] and protein digestion steps [ 10 , 18 \u2013 21 ] , and all target molecules must be properly linked to the gel mesh before expansion by , for example , anchoring [ 22 \u2013 24 ] or linkage by fixation [ 22 ] . In this context , implementing ExM in organisms with rigid cell walls [ 14 \u2013 17 , 25 \u2013 27 ] or maintaining isotropic expansion within samples of heterogeneous \u2018 rigidity \u2019 [ 20 , 28 \u2013 31 ] is a particular challenge . Recent work has shown that samples can show different macro - and micro - expansion factors ( e . g . factors that vary within a gel or tissue or differ between different organelles or domains within organelles [ 20 , 28 \u2013 31 ] ) . Finally , protocols are often not directly transfer - able between different samples . For example , an ExM method that preserved isolated centrioles from Chlamydomo - nas reinhardtii failed to isotopically expand chromatin in root tips of barley [ 28 , 32 ] . A second challenge in ExM is to retain the molecular information of the sample while breaking physical connec - tions for expansion . The protein retention yield is typically measured by how many target molecules can still be detected via fluorescent markers in an expanded sample [ 23 ] . It is mainly affected by the protocol steps of protein digestion and gelation . Protein digestion involves homogenization of the sample by cleavage of the proteins \u2019 peptide chains with proteinase K or collagenase type II , heat denaturation or homogenizing agents such as SDS [ 9 , 10 , 18 , 19 , 21 ] . Gelation involves the generation of radicals during the polymerization chain reaction of monomers to form a polyacrylamide gel ( PAA ) . Only recently , a radical - free gelation method using novel custom - synthesized monomers was developed [ 33 ] . Both , protein digestion and gelation , thus have potential to reduce the protein retention yield by degrading target sites for staining or by degrading genetically encoded markers . Protocols combining ExM with SRM techniques to date have achieved an overall resolution of 1 \u2013 30 nm [ 34 \u2013 37 ] . While a combination of ExM and SRM yields superior resol - ution , fluorophore choice and labelling strategy must suit both techniques , i . e . withstand and perform under the com - bined sample preparation and imaging protocols . To date , ExM was successfully combined with SIM [ 34 ] , STED [ 35 ] , dSTORM [ 36 , 37 ] and fluctuation - based techniques [ 38 , 39 ] . These ExM \u2013 SRM protocols rely on extrinsic labelling tech - niques of the targets using labels with organic dyes , mainly immunofluorescence [ 34 \u2013 39 ] . Furthermore , dSTORM uses ionic switching buffers needed for the on - and off - blinking of dyes . These buffers however shrink expanded gels from deionized water . Thus , ExSTORM relies on expansion in ionic buffers , which achieves only lower expansion of about threefold [ 36 ] . An alternative is Ex - SMLM [ 37 ] . Here , gels expanded in deionized water are embedded in an uncharged secondary gel . These double - layer gels achieve an about three - fold increase and tolerate incubation in ionic switching buffers . While current ExM \u2013 SRM protocols work well for abun - dant targets and larger ( polymeric ) structures in various samples , staining background signal from non - specific adher - ent labels severely compromises imaging of sparse targets . Additionally , factors that hinder efficient extrinsic labelling , such as a cell wall or a highly crowded and charged cytosol like in microbial organisms , can complicate labelling [ 40 ] and can introduce artefacts [ 41 ] . To date , there are only a handful of publications on combined ExM and SRM strat - egies for microbiology , namely two studies using ExSIM of fungi [ 16 ] and bacteria [ 42 ] , and one has combined ExM with SRRF imaging for viral SARS - Cov - 2 particles [ 38 ] . In this work , we set out to establish a protocol for com - bined ExM and SMLM imaging in yeast . Here , our SMLM method of choice was PALM microscopy . In contrast to dSTORM , PALM microscopy makes use of genetic labelling using fluorescent proteins ( FP ) [ 2 ] . Thus , PALM samples intrinsically possess high labelling efficiency and specificity and do not rely on switching buffers which makes an implementation of ExPALM attractive for microbiological studies ( at the price that current FPs offer lower photostabil - ity and photon yield compared to dyes ) . Nevertheless , as fluorescent proteins are affected and at least partially degraded by the ExM sample preparation protocols , so far , no protocols that combine ExM with PALM ( ExPALM ) exist . Here , we combine PALM imaging with a proExM protocol [ 23 ] optimized for isotropyand protein retention for the fission yeast Schizosaccharomyces pombe . Using our optimized sample preparation protocols , we have established an ExPALM proto - col that preserves about 50 % of the FP signal , doubling the retained amount compared to applying the protein retention ExM ( proExM ) protocol to the tested S . pombe cell samples . Our optimized protocol achieves a fivefold , highly isotropic expansion of fission yeast cells which we tested for several molecular targets . Taken together , this protocol is the first demonstration of combined single - molecule and expansion microscopy of yeast ( SExY microscopy ) . 2 . Material and methods 2 . 1 . Strain construction C - terminal tagging of cbp1 with mEos2 [ 43 ] and sad1 with mScarlet - I [ 44 ] was adapted from [ 45 ] . First , two intermediate plasmids were created . For creating cbp1 - mEos2 , a pBlue - script II SK + plasmid containing the Saccharomyces cerevisiae ADH1 terminator , a kanamycin resistance gene and the mEos2 gene was constructed using primers with 20 bp over - lap sequences ( primers 9 \u2013 14 ; electronic supplementary material , table S1 ) [ 45 ] . Similarly , for creating sad1 - mScarlet - I , a pFA6 - mScarlet - I - ADH1 - hphMX6 plasmid was con - structed amplifying mScarlet - I from pFA6a - mScarlet - I - hphNT1 [ 46 ] ( primers 24 \u2013 25 ) and pFA6 - ADH1 - hphMX6 ( primers 15 \u2013 16 ) from pFA6 - hphMX6 [ 47 ] . DNA fragments were combined by Gibson Assembly ( New England Biolabs , cat . no . E5510S ) , transformed into competent DH5 \u03b1 cells , streaked onto LB Amp R plates and incubated overnight at 37 \u00b0C . Single colonies were picked , grown in LB Amp R cul - tures for 2 h at 37 \u00b0C and plasmids were extracted ( ZymoPure II Plasmid Midiprep kit , Zymo Research , cat . no . D4200S ) and checked by sequencing ( Eurofins Genomics Germany GmbH ) . In a second step , 200 \u2013 600 bp DNA frag - ments up - and downstream of the respective insertion site were amplified from isolated genomic DNA from h \u2212 S . pombe cells with 20 bp overlaps to the flanking genes using primers 1 \u2013 4 for cbp1 and 17 , 19 & 22 \u2013 23 for sad1 . The DNA fragments were combined with the corresponding FP - ADH1 - antibiotic resistance fragment using overlap - extension PCR [ 48 ] for sad1 - mScarlet - I or SLiCE [ 49 ] for cbp1 - mEos2 . r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 2 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 10 \u00b5l of the PCR mix were transformed into either competent h + ade6 - 210 leu1 - 32 ura4 - D18 mEos2 : cnp1 or h - WT strain using the Frozen - EZ Yeast Transformation II Kit ( Zymo Research , cat . no . T2001 ) . Cells were streaked onto YES agar plates , incubated overnight at 32 \u00b0C , replica plated onto either YES Kan R or YES Hyg R agar plates and grown at 32 \u00b0C for another two days . Successful genomic integration was checked by colony - PCR and DNA sequencing ( Eurofins Genomics Germany GmbH ) . For N - terminal tagging of mis16 , DNA fragments up - and downstream of the insertion site where generated using primers 5 \u2013 8 . The DNA fragments were combined with mEos2 - mis16 using SLiCE [ 49 ] . 10 \u00b5l of the PCR mix were transformed into either competent h + ade6 - M210 leu1 - 32 ura4 - D18 ura4nm41 : mis16 cells using the Frozen - EZ Yeast Transformation II Kit ( Zymo Research , cat . no . T2001 ) . Cells were streaked out onto YES agar plates containing 1 g l \u2212 1 5 - Fluoroorotic Acid and 50 mg ml \u2212 1 uracil and grown at 32 \u00b0C for 2 days . Successful genomic integration was checked by colony - PCR and DNA sequencing ( Eurofins Genomics Germany GmbH ) . 2 . 2 . Protease enzyme activity assay The colorimetric assay to measure protease activities was adapted from [ 50 , 51 ] . In short , casein ( Sigma - Aldrich , cat . no . C7078 - 500G ) was digested to release tyrosine . After stop - ping the digestion with tricholoroacetic acid ( Sigma - Aldrich , cat . no . T0699 - 100ML ) , released tyrosine was measured by absorption at 660 nm using Folin & Ciocalteu \u2019 s Phenol reagent ( Sigma - Aldrich , cat . no . F9252 - 100ML ) . Enzymes were tested at their respective optimal working temperature and digestion buffer for 10 min . Concentration ranges were 0 . 2 \u2013 1 . 0 U ml \u2212 1 for proteinase K , 0 . 2 \u2013 1 . 0 , 10 U ml \u2212 1 and 50 U ml \u2212 1 for Zymo - lyase and \u03b2 - glucoronidase and 0 . 02 \u2013 0 . 1 U ml \u2212 1 for lysing enzyme . Controls with 20 mM of the protease inhibitor phe - nylmethylsulfonylfluorid ( PMSF , ThermoFisher Scientific , cat . no . 36978 ) were conducted for proteinase K and lysing enzyme and a standard curve of free L - tyrosine ( Sigma - Aldrich , cat . no . T3754 - 50G ) was measured . 2 . 3 . Schizosaccharomyces pombe cell culture S . pombe strains were grown on YES or EMM - leu agar plates at 32 \u00b0C for two days . Single colonies were picked and inocu - lated overnight in YES or EMM - leu medium at 25 \u00b0C . Overnight cultures were diluted to OD 0 . 05 for YES and OD 0 . 1 for EMM - leu cultures and grown at 25 \u00b0C to OD 0 . 5 \u2013 0 . 8 . For imaging non - expanded cells expressing cytosolic mEos2 , the EMM - leu overnight cultures were diluted to OD 0 . 2 in EMM containing 5 \u00b5M thiamine to repress mEos2 expression and were grown for 3 h at 25 \u00b0C . 2 . 4 . Sample preparation using the SExY protocol The SExY protocol was adapted from the proExM method [ 23 ] and optimized for PALM and expansion microscopy in fission yeast . The ideal cell density for sample preparation was deter - mined to be 6 OD ml \u2212 1 for all fluorescently - labelled protein of interest ( POI ) strains ( electronic supplementary material , table S2 ) and 12 OD ml \u2212 1 for SP16 cells which express cytosolic mEos2 and were added to all samples for drift correction . The appropriate volumes were transferred to fresh Erlenmeyer flasks and fixed with 3 . 2 % paraformaldehyde ( Sigma - Aldrich , cat . no . F8775 - 4X25ML ) and 0 . 1 % glutaraldehyde ( Sigma - Aldrich , cat . no . G5882 - 50ML ) at 25 \u00b0C for 10 min . The cells were harvested by centrifugation ( 358 g , 2 min , room tempera - ture ) , resuspended in 1 ml 1x PBS and washed twice in 1 ml 1x PBS for 5 min . The cells were washed twice with 200 \u00b5l 1x S - c / pH 5 . 8 ( 1 . 2 M D - Sorbitol ( Sigma - Aldrich , cat . no . S1876 - 500G ) in citrate / phosphate buffer ( Sigma - Aldrich , cat . no . C0759 - 1KG and cat . no . S7907 - 500G ) , pH 5 . 8 ) for 10 min each . The pellet after centrifugation was resuspended in 1 ml per 6 OD / ml Lallzyme MMX mix ( 100 mg ml \u2212 1 Lallzyme MMX ( Lallemand , cat . no . ElO11 - 2240 - 15 ) , 1 . 2 M D - Sorbitol in citrate / phosphate buffer , pH 5 . 8 ) and incubated at 30\u00b0C for 2 h for cell wall digestion . Subsequently , the cells were washed twice in 1 ml 1x S - PBS ( 1 . 2 M D - Sorbitol in 1x PBS , pH 7 . 4 ) . To anchor the amine groups of the proteins to the PAA gel mesh , the cell pellets were resuspended in 200 \u00b5l per 6 OD ml \u2212 1 AcX solution ( 0 . 1 mg ml \u2212 1 Acryloyl X ( Invitro - gen , cat . no . A20770 ) , 1 % ( v / v ) DMSO ( Carl Roth , cat . no . A994 . 1 ) in 1x S - PBS ) . Each POI strain was mixed with SP16 cells in a 1 : 2 ratio and incubated at 25 \u00b0C overnight . The next day , the sample was washed twice with 500 \u00b5l 1x S - PBS for 15 min . Round coverslips were cleaned in 1 M KOH for 30 min and rinsed with MilliQ water . Air - dried coverslips were incubated with 100 \u00b5l poly - L - lysine ( Sigma - Aldrich , cat . no . P8920 ) for 20 min at room temperature ( RT ) and then assembled into custom made imaging gel cassettes made of polyoxymethylene ( POM ) . Since a previous study found that an initial incubation of the monomer solution with the sample without the catalysator yielded higher expansion factors [ 31 ] , the cell pellets were resuspended in half of the 1 . 06x Monomer solution ( 8 . 625 % ( w / w ) Sodium Acrylate ( Sigma - Aldrich , cat . no . 408220 - 25G ) , 2 . 5 % ( w / w ) 37 : 1 Acrylamide : Bisacrylamide ( Sigma - Aldrich , cat . no . A6050 - 100ML ) , 0 . 15 % ( w / w ) N , N 0 - methylenebisacrylamide ( Sigma - Aldrich , cat . no . M1533 - 25ML ) , 2 M NaCl ( Sigma - Aldrich , cat . no . S3014 - 500G ) in 1x PBS ) together with 0 . 2 % TEMED ( Research Products Int , cat . no . T18000 - 0 . 05 ) and 0 . 05 % of freshly prepared L - glutathione ( Carl Roth , cat . no . 6832 . 3 ) . This pre - gelation mix was pipetted onto the coverslip in the imaging gel cassette and incubated for 10 min at 37 \u00b0C in the dark . Gelation was initiated by addition of 0 . 2 % APS ( Carl Roth , cat . no . 9178 . 3 ) and the rest of the 1 . 06x Monomer solution to the pre - gelation mix and thoroughly mixed . The gel cassettes were placed into a humid environment using Petri dishes with wet paper towels wrapped in aluminium foil and incubated at 37 \u00b0C for 1 h . The gels were then incu - bated in 2 ml digestion buffer ( 50 mM Tris ( Carl Roth , cat . no . AE15 . 2 ) pH 7 . 5 , 1 mM EDTA ( Sigma - Aldrich , cat . no . EDS - 100G ) , 0 . 2 % Triton X - 100 ( Sigma - Aldrich , cat . no . T8787 - 50ML ) , 1 M NaCl ) with 3 U ml \u2212 1 effective proteinase K ( ThermoFisher , cat . no . EO0491 ) at 37 \u00b0C for 30 min in the aluminium foil wrapped mini Petri dishes ( Sarstedt , cat . no . 82 . 1135 . 500 ) . Thereafter , gels were washed with MilliQ water , transferred to Petri dishes , wrapped in aluminium foil and fully expanded in 100 ml MilliQ water for 4 h at RT , exchan - ging MilliQ water every hour . Ibidi 8 - well glass bottom slides ( Ibidi , cat . no . 80807 ) , which were previously cleaned with 1 M KOH ( Sigma - Aldrich , cat . no . 221473 - 2 . 5KG - M ) for 30 min and washed twice with MilliQ water , were incubated with poly - L - lysine for 20 min and air dried . The expanded gel was cut using a scalpel , transferred into an Ibidi well , excess water was removed and the gel was incubated for r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 3 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 10 min . After that , the gel was secured in place by embedding it in 2 % agarose ( Carl Roth , cat . no . 3810 . 3 ) and allowing the agarose to dry for 10 min . The gel was submerged in 300 \u00b5l MilliQ water and then imaged . 2 . 5 . Tests of various sample preparation steps while optimizing the protocol 2 . 5 . 1 . Cell wall digestion To lyse the cell wall , the chemically fixed cells were treated with various enzymes and tested for a homogeneous macro - scale expansion via fluorescence microscopy . The used enzymes and conditions for cell wall digestion can be found in electronic supplementary material , table S3 . All incubations were done agitating the samples in a thermo block . 2 . 5 . 2 . Permeabilization For permeabilizing the cell membrane , the fixed and in 1xPBS washed cells were resuspended in 200 \u00b5l 1xPBS containing 0 . 1 % Triton X - 100 and incubated at 37 \u00b0C for 30 min . After - wards the cell wall was digested and the initial proExM protocol for fission yeast was continued . 2 . 5 . 3 . Protein digestion by heat denaturation The initial proExM protocol adapted for fission yeast was performed until the gelation step , where 0 . 05 % glutathione were added to the gelation as described above . Afterwards , the gel was submerged into 2 ml of the 50 mM Tris . HCl buffer at pH 8 . 0 and incubated at 65\u00b0C for 15 min . Then the gel was incubated in the renaturing buffer ( 35 mM KCl 2 , 1 mM MgCl 2 , 1 mM DTT , 30 % glycerol , 50 mM Tris . HCl , pH 8 . 0 ) at RT for 30 min . 2 . 5 . 4 . Protein digestion by homogenization with SDS and heat denaturation The sample was prepared the same way as for the heat dena - turation until protein digestion . The gel was then submerged in 2 ml freshly prepared SDS solution ( 200 mM SDS , 200 mM NaCl , 50 mM Tris . HCl , pH 8 . 0 ) and incubated at 37 \u00b0C for 30 min with a subsequent incubation at 65 \u00b0C for 15 min . After washing the gel with deionized water , it was fully sub - merged in renaturing buffer and incubated at RT for 30 min . 2 . 5 . 5 . Gelation including glutathione To determine the effect of glutathione , 0 \u2013 0 . 5 % freshly prepared glutathione was added to the pre - gelation mix and incubated with the sample for 10 min at 37 \u00b0C in the dark . After initiation of the gelation , samples were taken for fluorescence intensity measurement , and the rest of the sample was digested and expanded as described above . The gel rigidity of the expanded gels was determined visually and haptically . 2 . 6 . Microscope set - up For all imaging experiments , a custom build set - up based on an automated Nikon Ti Eclipse microscopy body with suitable dichroic and emission filters ( ET dapi / Fitc / cy3 dichroic , ZT405 / 488 / 561rpc rejection filter , ET525 / 50 for GFP and the green form of mEos2 or ET610 / 75 for mScar - let - I and the red photoconverted form of mEos2 , all AHF Analysentechnik , Germany ) and a CFI Apo TIRF 100 \u00d7 oil objective ( NA 1 . 49 , Nikon ) was used . A perfect focus system ( Nikon , Germany ) was used for z - axis control , except for ExPALM imaging , where the sample was imaged at 10 \u2013 80 \u00b5m depth . All lasers ( 405 nm OBIS , 488 nm Sapphire LP and 561 nm OBIS , Coherent Inc . USA ) were modulated by an acousto - optical tunable filter ( AOTF ; Gooch and Housego , USA ) . Fluorescence detection was performed by an emCCD camera ( iXON Ultra 888 , Andor , UK ) at a pixel size of 129 nm . Image acquisition was controlled by a customized version of \u00b5Manager [ 52 ] . 2 . 7 . Fluorescence imaging of GFP - nup132 Expanded and non - expanded GFP - nup132 cells were ident - ified using a 488 nm laser ( 21 W cm \u2212 2 at the sample level ) . A fluorescence video of 50 frames with an exposure time of 100 ms per frame was acquired using 81 W cm \u2212 2 at the sample level . 2 . 8 . Fluorescence imaging and analysis of sad1 - mScarlet - I While optimizing the ExPALM protocol , a sample was taken after every step of the protocol and placed on a cleaned and poly - L - lysine coated Ibidi 8 - well glass bottom slide . After 15 min of settling time , the sample was either washed twice ( 1x PBS for fixation / permeabilization steps or 1x S - PBS for cell wall digestion / anchoring steps ) or prevented from drying / shrinking by adding one to two drops of MilliQ water in gelation / expansion steps . Only cells in mitosis with two sad1 - mScarlet - I spots were imaged . A video of 10 imaging frames of the sad1 - mScarlet - I fluorescence was taken with an exposure time of 100 ms per frame and in epi - illumination using a 561 nm laser . Laser intensities as measured at the sample level were 3 W cm \u2212 2 ( live - anchoring ) and 6 W cm \u2212 2 ( gelation - protein digestion ) . The integrated intensity of the fluorescent spots and a close - by background area were measured in a 14 \u00d7 22 pixel ROI in the first three imaging frames using ImageJ [ 53 ] . The final intensity measure was obtained by calculating the mean of the three frames for both ROIs and subtracting the background . 2 . 9 . PALM imaging and analysis of expanded cells Expanded cbp1 - mEos2 and mis16 - mEos2 cells were ident - ified in the gel by their fluorescence in epi - illumination using the 488 nm laser ( 21 W cm \u2212 2 at the sample level ) and PALM videos of 3000 - 5000 frames were acquired with an exposure time of 60 \u2013 80 ms per frame photoconverting and imaging mEos2 by continuous illumination of 405 nm laser ( 3 \u2013 12 W cm \u2212 2 at the sample level ) and 561 nm laser ( 1 . 2 kW cm \u2212 2 at the sample level ) . For imaging the DNA , the DNA was stained after mEos2 read - out using 100 nM TO - PRO - 3 Iodide ( Invitrogen , cat . no . T3605 ) which was added to the imaging well on the microscope stage and incu - bated for 30 min . A fluorescence video of 50 frames with an exposure time of 100 ms per frame was acquired using the 561 nm laser ( 0 . 5 kW cm \u2212 2 at the sample level ) . Expanded mEos2 - cnp1 / sad1 - mScarlet - I cells where identified by r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 4 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 mScarlet - I fluorescence using the 561 nm laser . A fluor - escence movie of 50 frames with an exposure time of 100 ms per frame was acquired using the 561 laser ( 0 . 5 kW cm \u2212 2 at the sample level ) . Remaining mScarlet - I flu - orescence was photobleached ( 1 . 2 kW cm \u2212 2 at the sample level , 561 nm laser ) and the mEos2 signal and the TO - PRO - 3 Iodide DNA stain were imaged as described above . All samples contained cells expressing cytosolic mEos2 , which served as marker for drift correction . Each field of view contained at least one cell expressing cytosolic mEos2 and was drift - corrected by cross - correlation or by relative frame - to - frame shift ( supplementary code and https : / / github . com / Endesfelder - Lab ) . Data with unreliable drift cor - rection , which typically occurred at the end of PALM videos at low localization densities , were discarded . Images were reconstructed in Rapidstorm 3 . 2 with a pixel size of 10 nm [ 54 ] . Localization precision was determined using the NeNA method [ 55 , 56 ] . For image representation , images were blurred by a Gaussian using their NeNA value as the sigma value . PALM images were overlaid with corre - sponding fluorescence images of sad1 - mScarlet - I and DNA in ImageJ 1 . 52p . 2 . 10 . PALM imaging and analysis of non - expanded cells Cells expressing cytosolic mEos2 were fixed with 3 . 2 % paraf - ormaldehyde ( Sigma - Aldrich , cat . no . F8775 - 4X25ML ) and 0 . 1 % glutaraldehyde ( Sigma - Aldrich , cat . no . G5882 - 50ML ) and embedded in 0 . 5 % agarose gels in an Ibidi 8 - well glass bottom slide together with a 1 : 500 dilution of sonicated , dark - red , 0 . 2 \u00b5m sized FluoSpheres ( Molecular probes , cat . no . F8807 ) . PALM videos of 3000 - 5000 frames were acquired with an exposure time of 60 ms per frame photoconverting and imaging mEos2 by pulsed illumination of 405 nm laser ( 3 \u2013 12 W cm \u2212 2 at the sample level ) and 561 nm laser ( 1 . 2 kW cm \u2212 2 at the sample level ) . Each field of view was corrected by fiducial drift correc - tion using a custom script ( Python 3 . 7 ) . Images were reconstructed in Rapidstorm 3 . 2 with a pixel size of 10 nm [ 54 ] . Localization precision was determined using the NeNA method [ 55 , 56 ] . For image representation , images were tracked in Rapidstorm 3 . 2 with an allowed blinking interval of five frames and a sigma set to the NeNA value . After filtering out all trajectories of three steps and less , a PALM image was reconstructed in Rapidstorm 3 . 2 and blurred by a Gaussian using the NeNA value as sigma value . 2 . 11 . Determination of expansion factor The expansion factor was measured by the cytosol width and the nuclear size of expanded and non - expanded cells . For the cytosol width , mean and standard error of cells in recon - structed PALM images of fixed and expanded cells expressing cytosolic mEos2 were measured . Raw SMLM data were used for this purpose , as cell peripheries appear sparser post - processing ( e . g . due to combining multiple local - izations of the same fluorescence event into one ) . For measuring the nuclear expansion factor , the mean and stan - dard error of the nuclear diameter were determined in fluorescence images of expanded and non - expanded GFP - nup132 cells were taken . Cells with non - spherical nuclei were excluded from the analysis . The expansion factor was calculated by dividing the expanded cytosol width or nuclear size by the non - expanded cytosol width or nuclear size . 2 . 12 . Assessment of microscale isotropy of individual expanded cells by evaluating the distribution of cytosolic mEos2 molecules Nearest - neighbour distances between cytosolic mEos2 local - izations in expanded cells were compared with cytosolic mEos2 localizations in non - expanded cells in R , v3 . 6 . 1 ( https : / / www . r - project . org / ) , using RStudio , 2022 . 07 . 2 + 576 ( https : / / www . rstudio . com / ) and the spatstat package [ 57 ] . To be able to compare the distances , first , point densities ( number of points per cell area ) were evaluated for all cells . Next , for a given expanded cell , areas of the 30 non - expanded cells were enlarged in silico until point densities matched . Then , the average nearest - neighbour distances were calcu - lated . Finally , the mean ( \u03bc ) across non - expanded cells was subtracted from the value of the expanded cell ( x i ) , and nor - malized to the standard deviation ( \u03c3 ) of the distribution of non - expanded cells : ( x i \u2212 \u03bc ) / \u03c3 . A two - sided Student \u2019 s t - test was performed . 2 . 13 . Data reproducibility statement All experiments were performed at least in duplicate for negative results that did not lead anywhere further and at least in triplicate for forward - leading results from different days of sample preparation . 3 . Results and discussion In this work , we established a sample preparation and imaging protocol which combines the super - resolution tech - nique PALM with a proExM protocol optimized for isotropy and protein retention for the fission yeast S . pombe , which we named Single - molecule and Expansion microscopy in fis - sion Yeast ( SExY ) . Our final protocol ( figure 1 a ) consists of six steps : chemical fixation of the cells and subsequent cell wall digestion , protein anchoring and gelation of the PAA , which is then followed by protein digestion and finally expansion . SExY achieves an about fivefold expansion , as demonstrated by cells that cytosolically express the FP mEos2 whose positions were super - resolved by PALM ima - ging ( figure 1 b ) . While cytosolic , free diffusive mEos2 molecules are generally distributed throughout the cell , SExY also clearly resolves their ( expected ) non - random , small - scale substructured distribution ( e . g . small omitted regions of round vesicles and a maximum of FPs near the nucleus ) . Essential aspects of the SExY protocol are , first , an isotro - pic and satisfactory expansion , and second , a high protein retention yield to preserve protein yield ( i . e . the FP signal for PALM read - out ) . Importantly , neither was the case when using the original proExM protocol on S . pombe cells . For verifying the isotropic expansion of the cells , we measured the cytosol width of cells expressing cytosolic mEos2 as seen in PALM images ( figure 1 c ( i ) ) . For verifying the isotropic expansion of organelles , we measured the nuclear diameter of cells in interphase that expressing GFP - tagged nuclear pore protein Nup132 ( figure 1 c ( ii ) ) . Non - expanded cells yielded a cytosol width of 2 . 4 \u00b5m \u00b1 0 . 19 \u00b5m r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 5 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 and a nuclear diameter of 2 . 36 \u00b5m \u00b1 0 . 34 \u00b5m , cells expanded with the SExY protocol showed a cytosol width of 11 . 8 \u00b5m \u00b1 1 . 8 \u00b5m and a nuclear diameter of 11 . 6 \u00b5m \u00b1 1 . 4 \u00b5m . This resulted in a macroscale expansion factor of 4 . 9 ( standard error ( s . e . ) 0 . 2 , standard deviation ( s . d . ) 0 . 8 ) as measured by the cytosol width and a factor of 4 . 9 ( s . e . 0 . 2 , s . d . 0 . 9 ) for 5 . protein digestion 6 . expansion in MQ water 1 . fixation 2 . cell wall digestion 3 . protein anchoring 4 . gelation of PAA gel O macroscale expansion ( d ) microscale expansion 2 \u00b5m isotropic ( e ) FP survival 2 \u00b5m non - expanded expanded 0 2 4 6 8 10 12 14 non - expanded expanded 0 2 4 6 8 10 12 14 post optimization 2 \u00b5m non - isotropic pre optimization 0 500 1000 1500 2000 n = 143 n = 19 n = 68 n = 28 n = 21 n = 41 n = 35 c y t o s o l w i d t h ( \u00b5 m ) nu c l e u s d i a m e t e r ( \u00b5 m ) f l uo r e s ce n ce s po t i n t e n s it y ( a r b . un it s ) fixedcells pre post optimization ( a ) ( b ) ( c ) ( i ) ( ii ) 2 \u00b5m 5 \u00d7 expansion Figure 1 . Principle and performance of the SExY method . ( a ) Principle of SExY sample preparation , a correlative ExPALM protocol for expanding yeast . Shown are cell wall ( black ) , cell membrane ( grey ) , inner and outer nuclear membrane ( magenta ) , target protein ( blue ) , FP marker ( red ) and the PAA gel mesh ( orange ) . The acrylate group of the protein anchoring agent AcX is highlighted in orange . ( b ) Fission yeast cells expand approximately fivefold using the SExY protocol . ( c ) SExY provides a macroscale isotropic expansion as determined by two different samples , measuring ( i ) cytosol widths and ( ii ) nuclear diameter ( Nup132 ) of non - expanded and expanded cells imaged over a total of nine different experiments . The expansion factor was determined to be 4 . 9 ( s . e . 0 . 2 , s . d . 0 . 8 ) ( cytosol width ) and 4 . 9 ( s . e . 0 . 2 , s . d . 0 . 9 ) ( nuclear diameter ) , respectively . Note that nuclear widths were estimated from diffraction limited epifluorescence data and cannot be compared directly to cytosolic widths derived from SMLM data . ( d ) Additionally , the isotropic microscale expansion of the SExY protocol was probed investigating cell borders . Here , cells expanded with the SExY protocol show no mEos2 localizations being pulled outside of the cell area and hence an isotropic expansion on the microscale . ( e ) SExY retains 46 % ( s . e . 3 . 1 % , s . d . 21 . 8 % ) of FP signal retention , whereas the original proExM protocol retained 22 % ( s . e . 2 . 0 % , s . d . 12 . 4 % ) . r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 6 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 the nuclear diameter . Notably , the achieved expansion is higher than in current protocols that combine ExM with dSTORM and achieve an about threefold expansion [ 36 , 37 ] . To assess isotropic expansion on the microscale , we examined the cell boundaries of cells expressing cytosolic mEos2 which were expanded using either the initial proExM protocol ( adding the step of using Lallzyme MMX for cell wall diges - tion ) or the optimized SExY protocol . When applying the original protocol , we found a substantial fraction of mEos2 molecules located outside the main cell area within the poly - acrylamide mesh network , apparently pulled outward due to heterogeneous expansion and resulting in \u2018 fuzzy \u2019 cell bound - aries . By contrast , cell boundaries were clearly defined when the SExY protocol was applied ( figure 1 d ) , indicating a more defined and homogeneous microscale expansion . Further - more , statistical testing for changes in cytosolic mEos2 FP distribution upon expansion did not show any changes ( elec - tronic supplementary material , figure S1 ) . While prior to optimization of the protocol , individual expanded cells show a relatively high variation in cytosolic mEos2 distri - bution relative to non - expanded cells ( up to 1 . 88 s . d . ) , post - optimization all expanded cells lie well within the range expected for isotropic expansion ( up to 0 . 5 s . d . in relation to the mean of non - expanded cells ) . To measure the protein retention yield , we quantified the fluorescent signal retained from the spindle pole protein sad1 tagged with mScarlet - I from cells during anaphase B in mitosis . For our initial proExM protocol , we found that only 22 % ( s . e . 2 . 0 % , s . d . 12 . 4 % ) of the fluorescence signal was retained when compared to fixed cells ( figure 1 e ) . However , using the SExY protocol , about 46 % ( s . e . 3 . 1 % , s . d . 21 . 8 % ) of the sad1 - mScarlet - I signal was preserved , doubling the retained amount . To gauge whether an increase in resolution could be achieved by expansion , localization precisions were estimated using the NeNA method [ 55 ] to be 13 . 9 nm ( s . e . 0 . 8 , s . d . 3 . 5 , n = 22 ) for non - expanded cells and 26 . 5 nm ( s . e . 0 . 7 , s . d . 5 . 4 , n = 55 ) for expanded cells ( electronic supplementary material , figure S2 ) . This implies a 1 . 9x ( s . e . 0 . 1 , s . d . 0 . 6 ) worse localiz - ation precision in SExY compared to conventional PALM . This change can be attributed to the reduced brightness of individual spots due to imaging in a larger depth of several micrometres to which our set - up is not adapted to . However , the 1 . 9x worse localization precision is compensated by the 4 . 9x expansion , which leaves us with an overall increase in optical resolution of 2 . 5x when going from conventional PALM imaging to SExY on our set - up . By using a set - up better suited for imaging at depth , this loss could be compensated further . Altogether , these quality measures indicate that the SExY protocol is a promising approach to map protein structures in fission yeast . To achieve this , we have optimized the cell wall digestion , gelation and protein digestion steps in the develop - ment of the final SExY protocol . We report on these optimizations and discuss their rationale in detail in the following sections . 3 . 1 . Efficient and complete cell wall digestion of fission yeast We tested various cell wall removal enzymes commonly used in fission yeast studies [ 58 \u2013 60 ] and visualized expanded S . pombe cells that cytosolically expressed mEos2 . We found that zymolyase , zymolyase in combination with lysing enzyme , snail enzyme and lyticase caused partial expansion of yeast cells visible by \u2018 hour - glass \u2019 like shapes failing to expand the septum or earlier cytokinesis sites ( figure 2 a ) . Lallzyme MMX , chitinase and \u03b2 - glucoronidase completely expanded S . pombe cells and showed comparable mEos2 signal to the non - expanded cells ( figure 1 b ; electronic supplementary material , table S3 ) . The fact that we were able to fully expand fission yeast cells using a chitinase for cell wall digestion may indicate that chitin is present at the septum . The fission yeast septum consists of a primary septum composed predomi - nantly of \u03b2 - ( 1 , 3 ) - glucan and a flanking secondary septum composed of \u03b1 ( 1 , 3 ) - glucan and branched \u03b2 ( 1 , 6 ) - glucan [ 61 , 62 ] . Chitin synthase genes ( chs1 , chs2 ) are present , but their function is still controversial [ 63 , 64 ] . On the one hand , chs2 has been shown to be localized to the growing septal edges in vegetative S . pombe cells , yet chs2 possesses several mutations at sites critical for chitin synthesis [ 64 \u2013 66 ] . We decided not to further investigate the \u2018 hourglass - like pheno - types \u2019 and the potential localization of chitin at the septum and instead proceeded to the next steps of protein digestion and gelation to be optimized , using Lallzyme MMX as our standard for cell wall digestion , as it was the cheapest and most reliable alternative among the three enzymes that resulted in complete expansion in our hands . In a recent study , S . pombe was successfully expanded using zymolyase for cell wall removal [ 67 ] . We speculate that this may be explained by the generally harsher conditions of this proto - col , e . g . an additional fixation step with ice - cold acetone and replacement of proteinase K digestion with SDS treat - ment , which may be at the expense of protein retention yield , but this was not specified . 3 . 2 . Increasing the protein retention yield To evaluate how much fluorescent signal of the fluorescent proteins is retained after each step of sample preparation , we measured the fluorescent spot intensity of the mScarlet - I - tagged spindle pole protein sad1 , whose structure and protein copy numbers is well defined during anaphase B in mitosis [ 68 ] . For our initial protocol , following the proExM protocol with an added step for cell wall digestion using Lall - zyme MMX , only 22 % ( s . e . 2 . 0 % , s . d . 12 . 4 % ) of the signal was retained compared to chemical fixed samples ( figure 1 e ) . This overall loss of fluorescent protein signal was mainly caused by three processing steps : A decrease of 18 % ( s . e . 2 . 8 % , s . d . 29 . 6 % ) after permeabilization with 0 . 1 % Triton X - 100 , a decrease of another 22 % ( s . e . 2 . 5 % , s . d . 23 . 9 % ) after cell wall digestion using Lallzyme MMX compared to the pre - vious step in the protocol and a decrease of another 46 % ( s . e . 2 . 0 % , s . d . 12 . 4 % ) after protein digestion ( electronic sup - plementary material , figure S3 ) . We therefore targeted these three steps for optimization of the protocol . In a first optimization , we excluded a separate permeabi - lization step using Triton X - 100 as we hypothesized that the cell membrane is sufficiently disintegrated combining the steps of fixation with 3 . 2 % paraformaldehyde and 0 . 1 % glu - taraldehyde and of protein digestion in a buffer containing Triton X - 100 . Cells expanded with a protocol excluding this step indeed expanded normally ( examples in figure 1 ) . Fur - thermore , we tested proteinase K as well as the different enzymes for cell wall removal for their proteinase activity r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 7 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 in a colorimetric assay measuring tyrosine release in casein digestion ( figure 2 b and electronic supplementary material , figure S4 ) . For proteinase K , we found that the proteinase activity for 1 U ml \u2212 1 of our stock was reduced to only 64 % for freshly bought proteinase K and lower for older stocks . A control with protease inhibitor PMSF showed no residual proteinase activity . Therefore , for all further experiments , we determined the effective proteinase K activity using the colorimetric assay and adjusted the used concentrations accordingly . Among the cell wall removal enzymes , only the lysing enzyme and Lallzyme MMX showed proteinase activity , whereas zymolyase and \u03b2 - glucoronidase showed no activity at 1 U ml \u2212 1 and even at higher concentrations of 50 U ml \u2212 1 . Compared to proteinase K at a 10 - fold higher con - centration , lysing enzyme at a concentration of 0 . 1 U ml \u2212 1 showed rather strong activity . For Lallzyme MMX , we detected a ( low ) proteinase activity at 10 mg ml \u2212 1 . Impor - tantly , we found no significant decrease in fluorescence intensity of sad1 - mScarlet - I after the exclusion of a separate permeabilization step with Triton X - 100 ( electronic sup - plementary material , figure S3b ) . We thus speculate that Lallzyme MMX without the detergent step at this early stage of the protocol might be hindered to freely access the cell , in particular the nucleus , and thus does not show visible proteinase activity on sad1 - mScarlet - I fluorescence . When adjusting the proteinase K concentration for protein digestion ( see below ) , we nevertheless could measure an effect when blocking the proteinase activity of Lallzyme MMX by PMSF . In general , different proteinase K activities in different laboratories could be one cause for inhomogeneous expansions [ 20 , 28 \u2013 31 ] . On the basis of excluding a separate permeabilization step and using 10 mg ml \u2212 1 of Lallzyme MMX for cell wall digestion , we then optimized protein digestion and tested different methods that have been reported in the ExM literature : protein digestion using proteinase K [ 9 ] , heat denaturation [ 10 ] and denaturation using SDS [ 19 ] . Hom - ogenization by heat denaturation or treatment with SDS resulted in expanded fission yeast cells , but no or very low fluorescent signal for sad1 - mScarlet - I or cytosolically expressed mEos2 was subsequently detected ( data not shown ) . Using the proteinase K concentration from the proExM protocol , we were able to expand and image fluores - cently labelled fission yeast , albeit with severe signal loss . As sufficient sample homogenization is critical for isotropic expansion , we tested different proteinase K concentrations to determine the minimal needed concentration of use . We found that the lowest effective concentration that allowed iso - tropic expansion on both , the macro - and the microscale of the sample was 3 U ml \u2212 1 ( controlled for efficiency by the snail enzyme lyticase zymolyase + lysing enzyme 5 \u00b5m released tyrosine ( \u00b5mol ) (cid:2) 0 . 5 0 0 . 5 1 . 0 1 U ml \u2013 1 proteinase K 1 U ml \u2013 1 proteinase K + 20 mM PMSF 1 U ml \u2013 1 zymolyase 50 U ml \u2013 1 zymolyase 1 U ml \u2013 1 (cid:3) - glucoronidase 50 U ml \u2013 1 (cid:3) - glucoronidase 0 . 1 U ml \u2013 1 lysing enzyme 0 . 1 U ml \u2013 1 lysing enzyme + 20 mM PMSF 10 mg ml \u2013 1 lallzyme MMX ( a ) cell wall digestion zymolyase ( b ) effective proteinase activity 0 0 . 05 0 . 10 0 . 20 0 . 30 glutathione ( % ) 0 . 40 0 . 50 0 500 1000 1500 2000 f l uo r e s ce n ce s po t i n t e n s it y ( a r b . un it s ) gel rigidity ( c ) gelation n o g e l a ti o n n = 34 n = 143 n = 34 n = 35 n = 35 Figure 2 . Protocol optimizations for SExY ( a ) Examples of incomplete cell wall digestion using zymolyase , snail enzyme , a combination of zymolyase and lysing enzyme and lyticase . All tested conditions lead to a partial and non - isotropic expansion constrained by remaining cell wall . ( b ) The proteinase activity of proteinase K ( red ) in comparison to different cell wall digesting enzymes measured by the release of tyrosine during casein digestion . Zymolyase ( blue ) , \u03b2 - glucoronidase ( green ) , lysing enzyme ( yellow ) and Lallzyme MMX ( magenta ) . Adding proteinase inhibitor PMSF successfully suppressed tyrosine release . ( c ) sad1 - mScarlet - I signal shows an increase of 27 % ( s . e . 12 . 8 % , s . d . 85 % ) in FP retention when adding 0 . 05 % glutathione during gelation . Gel rigidity decreases with increasing glutathione concentration . Thus , no gel formation when adding 0 . 4 % glutathione or higher was achieved . r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 8 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 activity assay ) . Lower concentrations of proteinase K or a cell wall digest of Lallzyme MMX in the presence of the protein - ase inhibitor PMSF followed by 3 U ml \u2212 1 proteinase K during protein digestion resulted in heterogeneous samples with a symptomatically increasing fraction of non - expanded cells cumulating at the bottom of the gel , only semi - expanded ( two - to threefold expansion ) or non - isotropic expanded cells , and only partially expanded gels ( data not shown ) . For the final SExY protocol , we thus use the protease activity of 3 U ml \u2212 1 proteinase K combined with the protease activity of Lallzyme MMX which represents the minimum protein digestion level for homogeneous expansion of fission yeast to maintain a robust ExM protocol while retaining the maximum possible fluorescence signal . In a final optimization , we then focused on the gelation step of the protocol as the free radicals that emerge during the gelation of the PAA gel have been shown to reduce the protein retention yield as well [ 23 , 33 ] . While we were not able to detect this loss in the initial protocol , we could detect the negative effect of the gelation process on protein retention after having optimized the other protocol steps ( figure 2 c ) . To minimize the loss , we tested whether a pre - sumably milder gelation by adding the antioxidant glutathione ( GSH ) , an efficient radical scavenger , would yield a higher retention yield , since glutathione can accept a radical at its sulfur moiety by hydrogen atom transfer ( GS ) , adding a hydrogen at the former radical site of proteins [ 69 , 70 ] . To test this hypothesis , we measured the fluorescence intensity of sad1 - mScarlet - I after PAA gelation for different glutathione concentrations and found a fluorescence increase of about 27 % ( s . e . 12 . 8 % , s . d . 85 % ) when adding 0 . 05 % gluta - thione ( figure 2 c ) . At higher glutathione concentrations , no further increase in fluorescence spot intensity but a decrease in PAA gel rigidity was observed until gelation was comple - tely abolished at concentrations of 0 . 4 % glutathione and higher ( figure 2 c ) . Presumably , the lower PAA gel rigidity stems from overall shorter PAA chains in the gel mesh , e . g . by glutathione reacting with growing PAA radical chains and thus terminating the chains early or by glutathione rad - icals ( GS ) forming glutathione disulfide ( GSSG ) , decreasing the overall radical concentration . In the final SExY protocol , we thus use 0 . 05 % glutathione during gelation . Finally , during this work , we tested several commonly used FPs for their fluorescence retention in ExM sample prep - aration as it is a known problem that chromophores often do not withstand treatments in correlative imaging protocols , e . g . when combining fluorescence microscopy with electron microscopy [ 71 ] . For conventional fluorophores , we tested GFP and mCherry , commonly used FPs of different origin from jellyfish and coral , in a dual - colour strain expressing two nuclear pore protein fusions [ 72 ] . While GFP - Nup132 was clearly visible , we could not detect mCherry - Nup131 in the expanded cells ( figure 3 d ) . Interestingly , we then found that mScarlet - I , a synthetic FP constructed by rational design based on several red FPs but with a high sequence identity of 86 % to mCherry , survives the protocol and we could obtain images of expanded cells with mScarlet - I labelled spindle pole protein sad1 ( figure 3 e , f ) . Owing to mScarlet - I \u2019 s superior brightness and red colour ( allowing to image deeper into the samples than when using GFP ) , we decided to use it for the characterization of protein retention ( figure 2 ) [ 73 ] . For PALM imaging , mainly FPs from Anthozoan corals , either DsRed - type like PAmCherry or Kaede - type like the Eos / Dendra / Maple FP family , are used [ 40 ] . We tested photoactivatable PAmCherry which showed no residual fluorescence already after fixation ( data not shown ) . From previous studies , it is known that PAmCherry photoactivation is substantially reduced for fixation protocols including glutaraldehyde [ 74 ] . Within the ExM protocol , a comparably high glutaraldehyde concentration is used . We thus hypothesize that PAmCherry was fully quenched by glutaraldehyde when fixing the samples . By contrast , the green - to - red photoconvertible FP mEos2 survives the ExM protocol just fine . We therefore tagged different low and high abundant proteins with different cellular localizations with mEos2 to validate our imaging sensitivity and FP survival during SExY imaging . Using mEos2 , we could suc - cessfully super - resolve proteins in expanded cells such as the nuclear DNA binding protein cbp1 ( figure 3 b ) [ 75 ] and the nuclear protein mis16 ( figure 3 c ) [ 76 ] as well as rather low - abundant proteins such as the centromere - specific his - tone protein cnp1 ( figure 3 e , f ) , which is only present in several tens to a few hundred copies per cell [ 41 , 77 ] . Using TO - PRO - 3 Iodide , we co - visualized DNA ( figure 3 c , e ) . On a small technical note , we also explored different drift correction strategies for PALM imaging of expanded gels . Drift correction of PALM data is commonly needed , as the recording of a PALM movie can take up to several minutes in which the sample can physically drift . This drift is often corrected by adding fiducial markers to the sample , e . g . flu - orescent polystyrene beads or gold nanoparticles [ 78 ] . For expanded gels , we found that fiducial markers ( even when covalently linking them to the gel mesh by amine groups ) to a large extend either ended up at the bottom of the expanded gel and in case of polystyrene beads , high amounts of individual dyes leached out into the gel and resulted in cells coated by fuzzy highly fluorescent background signal ( data not shown ) . Thus , we decided to use cross - correlation drift correction [ 79 ] , which was previously also used in ExSTORM and Ex - SMLM [ 36 , 37 ] . While cross - correlation approaches work well for large statistics , thus directly on samples with high abundant proteins such as e . g . microtu - bules ( as imaged in the ExSTORM and Ex - SMLM works ) , the correction of low - abundant protein signal fails due to insufficient statistics . Therefore , we mixed all samples with cells that cytosolically expressed mEos2 at high amounts and used the latter as markers for cross - correlation . Even at this for microbial PALM studies rather high level of several thousands of localizations per cell per movie , available cross - correlation approaches that cross - correlate sub - stacks of data fail to work reliably in many cases due to still too low statistics . Therefore , we implemented an additional drift correction method that determines the relative frame - to - frame shift between frames using tracked data of the cyto - solically expressed mEos2 ( supplementary code and https : / / github . com / Endesfelder - Lab ) . 4 . Conclusion With the combination of ExM with PALM ( ExPALM ) , we close a gap in available super - resolution and ExM correlative protocols . We report the homogeneous expansion of fission yeast S . pombe cells to a fivefold expanded size while retaining about 50 % of FP signal by several optimization steps of the original proExM protocol . The final protocol , which we r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 9 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 termed SExY ( Single - molecule and Expansion Microscopy of Yeast ) is robust and reproducible , and extends the ExM tool - box by a correlative protocol for ExPALM imaging in microbiology . SExY explicitly excludes a separate permeabili - zation step with a detergent , uses Lallzyme MMX for digestion of the cell wall as an alternative to pricier enzymes and / or lysing components that only show heterogeneous or partial expansion ( i . e . most striking \u2018 hour - glass \u2019 shaped phe - notypes ) , and includes a fine - tuned protein digest using proteinase K at a controlled effective potency of 3 U ml \u2212 1 and a gentler PAA gelation by adding 0 . 05 % glutathione . We suppose that the improved microscale isotropy in expan - sion as seen in SExY ( figure 1 d ) stems from the dual effect of optimized homogeneous protein digestion and overall shorter PAA chains due to the milder gelation . SMLM imaging has high demands for labelling specificity and efficiency . SExY as a correlative PALM imaging tech - nique is especially suitable for low - abundant proteins as it does not rely on indirect target labelling via organic dyes using immunofluorescence and covalent bioconjugation methods ( which are prone to unspecific staining due to the charge and molecular crowding in microbes [ 41 ] ) and is opti - mized for a high protein retention yield retaining 50 % of the FP signal . Cells expanded by the SExY protocol show a high fivefold expansion compared to Ex - SMLM and ExSTORM that both report only a threefold expansion , which can be attributed to the needed photoswitching buffers . Thus , for some research , using SExY as an ExPALM method might diminish the need for additional expansion steps of the sample in iterative ExM protocols . Possible future optimizations should be to implement a non - radical , highly defined lattice - like gel into the method , as recently introduced by the Tetragel [ 33 , 80 ] . This might lower loss of FP signal during gelation . Second , the range of suitable FPs should be extended . As the selection of FPs suitable for high - quality PALM imaging is rather small [ 2 , 40 ] , we believe that engineering optimized FP variants for correlative ExM approaches based on current FPs might be an effective strategy . Engineering FPs that are not further affected by ExM protocol steps ( e . g . that are less sensitive ( a ) ( b ) ( c ) ( d ) ( e ) ( f ) ( i ) GFP - Nup132 cytosolic mEos2 mEos2 - mis16 cbp1 - mEos2 1 \u00b5 m sad1 - mScarlet - I mEos2 - cnp1 DNA i ) sad1 - mScarlet - I mEos2 - cnp1 5 \u00b5m 5 \u00b5m 5 \u00b5m 5 \u00b5m 1 \u00b5m 1 \u00b5m DNA Figure 3 . Examples of SExY microscopy SExY imaging of high ( a \u2013 c ) and low ( d \u2013 f ) abundant proteins in fission yeast . Super - resolved targets are marked in italic , targets imaged by conventional epifluorescence in regular script . ( a ) mEos2 expressed in the cytosol , ( b ) nuclear DNA binding protein cbp1 , ( c ) nuclear protein mis16 and DNA ( TO - PRO - 3 Iodide ) , ( d ) nuclea pore protein Nup132 , ( e , f ) centromeric histone protein cnp1 relative to the spindle pole body protein sad1 ; combined with visualizing DNA ( TO - PRO - 3 Iodide ) ( e , inset i ) ) . r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 10 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 to a specific protein digestion method ) offers a significant advantage [ 81 ] . This strategy has been shown successful for other correlative methods ( e . g . in the development of mEos4 FPs ) . Here , nucleophilic amino acid residues that cross - link with aldehydes were replaced so that mEos4 var - iants tolerate osmium ( OsO4 ) fixation \u2014 a treatment needed for electron microscopy [ 82 ] . Third , microscopes should be set - up for imaging at 50 \u2013 80 \u00b5m depths to limit losses in local - ization precision that run counter to the gain in resolution provided by sample expansion . Overall , the new correlative protocol is valuable for all yeast researchers who wish to combine ExM imaging with super - resolution imaging . In particular , we believe that the optimizations in protein retention yield , encompassing reten - tion yield for fluorescent proteins , will also be of interest to a larger audience in microbial cell biology research and will extend to even more distant research areas if their interest involves studying low abundance proteins or for any reason relies on genetic labelling , e . g . when proteins are not readily accessible by extrinsic staining . Ethics . This work did not require ethical approval from a human subject or animal welfare committee . Data accessibility . Raw image data can be obtained from the authors on request . Custom relative frame - to - frame drift correction code is added as supplementary code and provided in the electronic sup - plementary material [ 83 ] as well as on github https : / / github . com / Endesfelder - Lab / frametoframedrift . Declaration of AI use . We have not used AI - assisted technologies in creating this article . Authors \u2019 contributions . I . V . : data curation , investigation , methodology , supervision , validation , visualization , writing \u2014 original draft , writ - ing \u2014 review and editing ; O . D . C . : conceptualization , data curation , formal analysis , funding acquisition , investigation , methodology , software , supervision , validation , visualization , writing \u2014 original draft , writing \u2014 review and editing ; M . A . H . : data curation , investi - gation , methodology ; U . E . : conceptualization , formal analysis , funding acquisition , investigation , methodology , project adminis - tration , resources , software , supervision , validation , visualization , writing \u2014 original draft , writing \u2014 review and editing . All authors gave final approval for publication and agreed to be held accountable for the work performed therein . Conflict of interest declaration . Authors declare no competing interests . Funding . This work was supported by funds from the Max Planck Society , the SYNMIKRO Post - Doc short - term fellowship to O . C . and start - up funds at Carnegie Mellon University . Acknowledgements . We thank Dr David Lando , Dr David Virant and Prof . Dr Haruhiko Asakawa for sharing their strains for this study . References 1 . Abbe E . 1873 Beitr\u00e4ge zur Theorie des Mikroskops und der mikroskopischen Wahrnehmung . Archiv f . mikrosk . Anatomie 9 , 413 \u2013 468 . ( doi : 10 . 1007 / BF02956173 ) 2 . Turkowyd B , Virant D , Endesfelder U . 2016 From single molecules to life : microscopy at the nanoscale . Anal . Bioanal . Chem . 408 , 6885 \u2013 6911 . ( doi : 10 . 1007 / s00216 - 016 - 9781 - 8 ) 3 . Klar TA , Hell SW . 1999 Subdiffraction resolution in far - field fluorescence microscopy . Opt . Lett . 24 , 954 \u2013 956 . ( doi : 10 . 1364 / ol . 24 . 000954 ) 4 . Gustafsson MG . 2000 Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy . J . Microsc . 198 , 82 \u2013 87 . ( doi : 10 . 1046 / j . 1365 - 2818 . 2000 . 00710 . x ) 5 . Betzig E , Patterson GH , Sougrat R , Lindwasser OW , Olenych S , Bonifacino JS , Davidson MW , Lippincott - Schwartz J , Hess HF . 2006 Imaging intracellular fluorescent proteins at nanometer resolution . Science ( New York , N . Y . ) 313 , 1642 \u2013 1645 . ( doi : 10 . 1126 / science . 1127344 ) 6 . Heilemann M , van de Linde S , Sch\u00fcttpelz M , Kasper R , Seefeldt B , Mukherjee A , Tinnefeld P , Sauer M . 2008 Subdiffraction - resolution fluorescence imaging with conventional fluorescent probes . Angew . Chem . ( International ed . in English ) 47 , 6172 \u2013 6176 . ( doi : 10 . 1002 / anie . 200802376 ) 7 . Rust MJ , Bates M , Zhuang X . 2006 Sub - diffraction - limit imaging by stochastic optical reconstruction microscopy ( STORM ) . Nat . Methods 3 , 793 \u2013 795 . ( doi : 10 . 1038 / nmeth929 ) 8 . Sharonov A , Hochstrasser RM . 2006 Wide - field subdiffraction imaging by accumulated binding of diffusing probes . Proc . Natl Acad . Sci . USA 103 , 18 911 \u2013 18 916 . ( doi : 10 . 1073 / pnas . 0609643104 ) 9 . Chen F , Tillberg PW , Boyden ES . 2015 Optical imaging : expansion microscopy . Science ( New York , N . Y . ) 347 , 543 \u2013 548 . ( doi : 10 . 1126 / science . 1260088 ) 10 . Ku T et al . 2016 Multiplexed and scalable super - resolution imaging of three - dimensional protein localization in size - adjustable tissues . Nat . Biotechnol . 34 , 973 \u2013 981 . ( doi : 10 . 1038 / nbt . 3641 ) 11 . Chang J - B et al . 2017 Iterative expansion microscopy . Nat . Methods 14 , 593 \u2013 599 . ( doi : 10 . 1038 / nmeth . 4261 ) 12 . Truckenbrodt S , Maidorn M , Crzan D , Wildhagen H , Kabatas S , Rizzoli SO . 2018 X10 expansion microscopy enables 25 - nm resolution on conventional microscopes . EMBO Rep . 19 , e45836 . ( doi : 10 . 15252 / embr . 201845836 ) 13 . Cheng S , Zhao Y . 2019 Nanoscale imaging of E . coli cells by expansion microscopy . Discoveries ( Craiova , Romania ) 7 , e98 . ( doi : 10 . 15190 / d . 2019 . 11 ) 14 . Chen L , Yao L , Zhang L , Fei Y , Mi L , Ma J . 2021 Applications of super resolution expansion microscopy in yeast . Front . Phys . 9 , 650353 . ( doi : 10 . 3389 / fphy . 2021 . 650353 ) 15 . Kunz TC , R\u00fchling M , Moldovan A , Paprotka K , Kozjak - Pavlovic V , Rudel T , Fraunholz M . 2021 The expandables : cracking the staphylococcal cell wall for expansion microscopy . Front . Cell . Infect . Microbiol . 11 , 644750 . ( doi : 10 . 3389 / fcimb . 2021 . 644750 ) 16 . G\u00f6tz R , Panzer S , Trinks N , Eilts J , Wagener J , Turr\u00e0 D , Di Pietro A , Sauer M , Terpitz U . 2020 Expansion microscopy for cell biology analysis in fungi . Front . Microbiol . 11 , 574 . ( doi : 10 . 3389 / fmicb . 2020 . 00574 ) 17 . Trinks N , Reinhard S , Drobny M , Heilig L , L\u00f6ffler J , Sauer M , Terpitz U . 2021 Subdiffraction - resolution fluorescence imaging of immunological synapse formation between NK cells and A . fumigatus by expansion microscopy . Commun . Biol . 4 , 1151 . ( doi : 10 . 1038 / s42003 - 021 - 02669 - y ) 18 . Drelich L , Aboulouard S , Franck J , Salzet M , Fournier I , Wisztorski M . 2021 Toward high spatially resolved proteomics using expansion microscopy . Anal . Chem . 93 , 12 195 \u2013 12 203 . ( doi : 10 . 1021 / acs . analchem . 0c05372 ) 19 . M \u2019 Saad O , Bewersdorf J . 2020 Light microscopy of proteins in their ultrastructural context . Nat . Commun . 11 , 3850 . ( doi : 10 . 1038 / s41467 - 020 - 17523 - 8 ) 20 . Campbell LA , Pannoni KE , Savory NA , Lal D , Farris S . 2021 Protein - retention expansion microscopy for visualizing subcellular organelles in fixed brain tissue . J . Neurosci . Methods 361 , 109285 . ( doi : 10 . 1016 / j . jneumeth . 2021 . 109285 ) 21 . Yu C - CJ et al . 2020 Expansion microscopy of C . elegans . eLife 9 , e46249 . ( doi : 10 . 7554 / eLife . 46249 ) 22 . Chozinski TJ , Halpern AR , Okawa H , Kim H - J , Tremel GJ , Wong ROL , Vaughan JC . 2016 Expansion microscopy with conventional antibodies and fluorescent proteins . Nat . Methods 13 , 485 \u2013 488 . ( doi : 10 . 1038 / nmeth . 3833 ) 23 . Tillberg PW et al . 2016 Protein - retention expansion microscopy of cells and tissues labeled using standard fluorescent proteins and antibodies . Nat . Biotechnol . 34 , 987 \u2013 992 . ( doi : 10 . 1038 / nbt . 3625 ) 24 . Damstra HG , Mohar B , Eddison M , Akhmanova A , Kapitein LC , Tillberg PW . 2021 Visualizing cellular and tissue ultrastructure using Ten - fold robust expansion microscopy ( TREx ) . eLife 11 , e73775 . ( doi : 10 . 7554 / eLife . 73775 ) 25 . Lim Y , Shiver AL , Khariton M , Lane KM , Ng KM , Bray SR , Qin J , Huang KC , Wang B . 2019 Mechanically resolved imaging of bacteria using expansion microscopy . PLoS Biol . 17 , e3000268 . ( doi : 10 . 1371 / journal . pbio . 3000268 ) r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 11 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 26 . Fan Y , Lim Y , Wyss LS , Park S , Xu C , Fu H , Fei J , Hong Y , Wang B . 2021 Mechanical expansion microscopy . Methods Cell Biol . 161 , 125 \u2013 146 . ( doi : 10 . 1016 / bs . mcb . 2020 . 04 . 013 ) 27 . Kao P , Nodine MD . 2021 Application of expansion microscopy on developing Arabidopsis seeds . Methods Cell Biol . 161 , 181 \u2013 195 . ( doi : 10 . 1016 / bs . mcb . 2020 . 06 . 004 ) 28 . Kubalov\u00e1 I , Schmidt \u010c ernohorsk\u00e1 M , Huranov\u00e1 M , Weisshart K , Houben A , Schubert V . 2020 Prospects and limitations of expansion microscopy in chromatin ultrastructure determination . Chrom . Res . 28 , 355 \u2013 368 . ( doi : 10 . 1007 / s10577 - 020 - 09637 - y ) 29 . B\u00fcttner M , Lagerholm CB , Waithe D , Galiani S , Schliebs W , Erdmann R , Eggeling C , Reglinski K . 2021 Challenges of using expansion microscopy for super - resolved imaging of cellular organelles . Chembiochem Eur . J . Chem . Biol . 22 , 686 \u2013 693 . ( doi : 10 . 1002 / cbic . 202000571 ) 30 . Zhu C et al . 2021 Measurement of expansion factor and distortion for expansion microscopy using isolated renal glomeruli as landmarks . J . Biophot . 14 , e202100001 . ( doi : 10 . 1002 / jbio . 202100001 ) 31 . Pernal SP et al . 2020 Nanoscale imaging using differential expansion microscopy . Histochem . Cell Biol . 153 , 469 \u2013 480 . ( doi : 10 . 1007 / s00418 - 020 - 01869 - 7 ) 32 . Gambarotto D et al . 2019 Imaging cellular ultrastructures using expansion microscopy ( U - ExM ) . Nat . Methods 16 , 71 \u2013 74 . ( doi : 10 . 1038 / s41592 - 018 - 0238 - 1 ) 33 . Gao R , Yu C - CJ , Gao L , Piatkevich KD , Neve RL , Munro JB , Upadhyayula S , Boyden ES . 2021 A highly homogeneous polymer composed of tetrahedron - like monomers for high - isotropy expansion microscopy . Nat . Nanotechnol . 16 , 698 \u2013 707 . ( doi : 10 . 1038 / s41565 - 021 - 00875 - 7 ) 34 . Halpern AR , Alas GCM , Chozinski TJ , Paredez AR , Vaughan JC . 2017 Hybrid structured illumination expansion microscopy reveals microbial cytoskeleton organization . ACS Nano 11 , 12 677 \u2013 12 686 . ( doi : 10 . 1021 / acsnano . 7b07200 ) 35 . Li R , Chen X , Lin Z , Wang Y , Sun Y . 2018 Expansion enhanced nanoscopy . Nanoscale 10 , 17 552 \u2013 17 556 . ( doi : 10 . 1039 / C8NR04267E ) 36 . Xu H et al . 2019 Molecular organization of mammalian meiotic chromosome axis revealed by expansion STORM microscopy . Proc . Natl Acad . Sci . USA 116 , 18 423 \u2013 18 428 . ( doi : 10 . 1073 / pnas . 1902440116 ) 37 . Zwettler FU , Reinhard S , Gambarotto D , Bell TDM , Hamel V , Guichard P , Sauer M . 2020 Molecular resolution imaging by post - labeling expansion single - molecule localization microscopy ( Ex - SMLM ) . Nat . Commun . 11 , 3388 . ( doi : 10 . 1038 / s41467 - 020 - 17086 - 8 ) 38 . Shaib AH et al . 2022 Expansion microscopy at one nanometer resolution . 39 . Wang B , Yao L , Jing Y , Fei Y , Bai Q , Mi L , Ma J . 2020 Multicomposite super - resolution microscopy : enhanced Airyscan resolution with radial fluctuation and sample expansions . J . Biophoton . 13 , e2419 . ( doi : 10 . 1002 / jbio . 201960211 ) 40 . Vojnovic I , Winkelmeier J , Endesfelder U . 2019 Visualizing the inner life of microbes : practices of multi - color single - molecule localization microscopy in microbiology . Biochem . Soc . Trans . 47 , 1041 \u2013 1065 . ( doi : 10 . 1042 / BST20180399 ) 41 . Virant D , Vojnovic I , Winkelmeier J , Endesfelder M , Turkowyd B , Lando D , Endesfelder U . 2021 Unraveling the kinetochore nanostructure in Schizosaccharomyces pombe using multi - color SMLM imaging . J . Cell . Biol . 222 , e202209096 . ( doi : 10 . 1083 / jcb . 202209096 ) 42 . G\u00f6tz R , Kunz TC , Fink J , Solger F , Schlegel J , Seibel J , Kozjak - Pavlovic V , Rudel T , Sauer M . 2020 Nanoscale imaging of bacterial infections by sphingolipid expansion microscopy . Nat . Commun . 11 , 6173 . ( doi : 10 . 1038 / s41467 - 020 - 19897 - 1 ) 43 . McKinney SA , Murphy CS , Hazelwood KL , Davidson MW , Looger LL . 2009 A bright and photostable photoconvertible fluorescent protein . Nat . Methods 6 , 131 \u2013 133 . ( doi : 10 . 1038 / nmeth . 1296 ) 44 . Bindels DS et al . 2017 mScarlet : a bright monomeric red fluorescent protein for cellular imaging . Nat . Methods 14 , 53 \u2013 56 . ( doi : 10 . 1038 / nmeth . 4074 ) 45 . Hayashi A , Ding D - Q , Da - Qiao D , Tsutsumi C , Chikashige Y , Masuda H , Haraguchi T , Hiraoka Y . 2009 Localization of gene products using a chromosomally tagged GFP - fusion library in the fission yeast Schizosaccharomyces pombe . Genes to Cells Mol . Cell . Mech . 14 , 217 \u2013 225 . ( doi : 10 . 1111 / j . 1365 - 2443 . 2008 . 01264 . x ) 46 . Abella M , Andruck L , Malengo G , Skruzny M . 2021 Actin - generated force applied during endocytosis measured by Sla2 - based FRET tension sensors . Dev . Cell 56 , 2419 \u2013 2426 . e4 . ( doi : 10 . 1016 / j . devcel . 2021 . 08 . 007 ) 47 . Hentges P , van Driessche B , Tafforeau L , Vandenhaute J , Carr AM . 2005 Three novel antibiotic marker cassettes for gene disruption and marker switching in Schizosaccharomyces pombe . Yeast ( Chichester , England ) 22 , 1013 \u2013 1019 . ( doi : 10 . 1002 / yea . 1291 ) 48 . Bryksin AV , Matsumura I . 2010 Overlap extension PCR cloning : a simple and reliable way to create recombinant plasmids . BioTechniques 48 , 463 \u2013 465 . ( doi : 10 . 2144 / 000113418 ) 49 . Zhang Y , Werling U , Edelmann W . 2012 SLiCE : a novel bacterial cell extract - based DNA cloning method . Nucleic Acids Res . 40 , e55 . ( doi : 10 . 1093 / nar / gkr1288 ) 50 . Cupp - Enyard C . 2008 Sigma \u2019 s non - specific protease activity assay - casein as a substrate . J . Visual . Exp . 19 , 899 . ( doi : 10 . 3791 / 899 ) 51 . Lowry O , Rosebrough N , Farr AL , Randall R . 1951 Protein measurement with the folin phenol reagent . J . Biol . Chem . 193 , 265 \u2013 275 . ( doi : 10 . 1016 / S0021 - 9258 ( 19 ) 52451 - 6 ) 52 . Edelstein A , Amodaj N , Hoover K , Vale R , Stuurman N . 2010 Computer control of microscopes using \u00b5Manager . Curr . Protoc . Mol . Biol . ch . 14 , Unit14 . 20 . ( doi : 10 . 1002 / 0471142727 . mb1420s92 ) 53 . Schindelin J et al . 2012 Fiji : an open - source platform for biological - image analysis . Nat . Methods 9 , 676 \u2013 682 . ( doi : 10 . 1038 / nmeth . 2019 ) 54 . Wolter S , L\u00f6schberger A , Holm T , Aufmkolk S , Dabauvalle M - C , van de Linde S , Sauer M . 2012 rapidSTORM : accurate , fast open - source software for localization microscopy . Nat . Methods 9 , 1040 \u2013 1041 . ( doi : 10 . 1038 / nmeth . 2224 ) 55 . Endesfelder U , Malkusch S , Fricke F , Heilemann M . 2014 A simple method to estimate the average localization precision of a single - molecule localization microscopy experiment . Histochem . Cell Biol . 141 , 629 \u2013 638 . ( doi : 10 . 1007 / s00418 - 014 - 1192 - 3 ) 56 . Malkusch S , Heilemann M . 2016 Extracting quantitative information from single - molecule super - resolution imaging data with LAMA - LocAlization Microscopy Analyzer . Sci . Rep . 6 , 34486 . ( doi : 10 . 1038 / srep34486 ) 57 . Baddeley A , Rubak E , Turner R . 2016 Spatial point patterns . Methodology and applications with R . Chapman & Hall / CRC interdisciplinary statistics series . Boca Raton , FL : CRC Press . 58 . Flor - Parra I , Zhurinsky J , Bernal M , Gallardo P , Daga RR . 2014 A Lallzyme MMX - based rapid method for fission yeast protoplast preparation . Yeast 31 , 61 \u2013 66 . ( doi : 10 . 1002 / yea . 2994 ) 59 . Forsburg SL , Rhind N . 2006 Basic methods for fission yeast . Yeast 23 , 173 \u2013 183 . ( doi : 10 . 1002 / yea . 1347 ) 60 . Po \u017e gajov\u00e1 M , Navratilova A , Trakovicka A . 2017 Determination of the efficient enzyme concentration for lytic digestion of vegetative cells but not spores in Schizosaccharomyces pombe . Acta Fytotechn Zootechn 20 , 20 \u2013 22 . ( doi : 10 . 15414 / afz . 2017 . 20 . 01 . 20 - 22 ) 61 . P\u00e9rez P , Cort\u00e9s JCG , Cansado J , Ribas JC . 2018 Fission yeast cell wall biosynthesis and cell integrity signalling . Cell Surf . ( Amsterdam , Netherlands ) 4 , 1 \u2013 9 . ( doi : 10 . 1016 / j . tcsw . 2018 . 10 . 001 ) 62 . Garc\u00eda Cort\u00e9s JC , Ramos M , Osumi M , P\u00e9rez P , Ribas JC . 2016 The cell biology of fission yeast septation . Microbiol . Mol . Biol . Rev . 80 , 779 \u2013 791 . ( doi : 10 . 1128 / MMBR . 00013 - 16 ) 63 . Bowen AR , Chen - Wu JL , Momany M , Young R , Szaniszlo PJ , Robbins PW . 1992 Classification of fungal chitin synthases . Proc . Natl Acad . Sci . USA 89 , 519 \u2013 523 . ( doi : 10 . 1073 / pnas . 89 . 2 . 519 ) 64 . Ding DQ , Tomita Y , Yamamoto A , Chikashige Y , Haraguchi T , Hiraoka Y . 2000 Large - scale screening of intracellular protein localization in living fission yeast cells by the use of a GFP - fusion genomic DNA library . Genes to Cells Dev . Mol . Cell . Mech . 5 , 169 \u2013 190 . ( doi : 10 . 1046 / j . 1365 - 2443 . 2000 . 00317 . x ) 65 . Matsuo Y , Tanaka K , Nakagawa T , Matsuda H , Kawamukai M . 2004 Genetic analysis of chs1 + and chs2 + encoding chitin synthases from Schizosaccharomyces pombe . Biosci . Biotechnol . Biochem . 68 , 1489 \u2013 1499 . ( doi : 10 . 1271 / bbb . 68 . 1489 ) 66 . Mart\u00edn - Garc\u00eda R , Dur\u00e1n A , Valdivieso M - H . 2003 In Schizosaccharomyces pombe chs2p has no chitin synthase activity but is related to septum formation . FEBS Lett . 549 , 176 \u2013 180 . ( doi : 10 . 1016 / S0014 - 5793 ( 03 ) 00812 - 3 ) 67 . Hinterndorfer K et al . 2022 Ultrastructure expansion microscopy reveals the nanoscale cellular architecture of budding and fission yeast . J . Cell . Sci . 135 , jcs260240 . ( doi : 10 . 1242 / jcs . 260240 ) r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 12 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024 68 . Bestul AJ , Yu Z , Unruh JR , Jaspersen SL . 2017 Molecular model of fission yeast centrosome assembly determined by superresolution imaging . J . Cell Biol . 216 , 2409 \u2013 2424 . ( doi : 10 . 1083 / jcb . 201701041 ) 69 . Osburn S , Berden G , Oomens J , Gulyuz K , Polfer NC , O \u2019 Hair RAJ , Ryzhov V . 2013 Structure and reactivity of the glutathione radical cation : radical rearrangement from the cysteine sulfur to the glutamic acid \u03b1 - carbon atom . ChemPlusChem 78 , 970 \u2013 978 . ( doi : 10 . 1002 / cplu . 201300057 ) 70 . Ramis R , Casasnovas R , Ortega - Castro J , Frau J , \u00c1lvarez - Idaboy JR , Mora - Diez N . 2019 Modelling the repair of carbon - centred protein radicals by the antioxidants glutathione and Trolox . New J . Chem . 43 , 2085 \u2013 2097 . ( doi : 10 . 1039 / C8NJ05544K ) 71 . Watanabe S , Punge A , Hollopeter G , Willig KI , Hobson RJ , Davis MW , Hell SW , Jorgensen EM . 2011 Protein localization in electron micrographs using fluorescence nanoscopy . Nat . Methods 8 , 80 \u2013 84 . ( doi : 10 . 1038 / nmeth . 1537 ) 72 . Asakawa H et al . 2019 Asymmetrical localization of Nup107 - 160 subcomplex components within the nuclear pore complex in fission yeast . PLoS Genet . 15 , e1008061 . ( doi : 10 . 1371 / journal . pgen . 1008061 ) 73 . Hagan I , Yanagida M . 1995 The product of the spindle formation gene sad1 + associates with the fission yeast spindle pole body and is essential for viability . J . Cell Biol . 129 , 1033 \u2013 1047 . ( doi : 10 . 1083 / jcb . 129 . 4 . 1033 ) 74 . Endesfelder U , Finan K , Holden SJ , Cook PR , Kapanidis AN , Heilemann M . 2013 Multiscale spatial organization of RNA polymerase in Escherichia coli . Biophys . J . 105 , 172 \u2013 181 . ( doi : 10 . 1016 / j . bpj . 2013 . 05 . 048 ) 75 . Baum M , Clarke L . 2000 Fission yeast homologs of human CENP - B have redundant functions affecting cell growth and chromosome segregation . Mol . Cell . Biol . 20 , 2852 \u2013 2864 . ( doi : 10 . 1128 / MCB . 20 . 8 . 2852 - 2864 . 2000 ) 76 . Hayashi T , Fujita Y , Iwasaki O , Adachi Y , Takahashi K , Yanagida M . 2004 Mis16 and Mis18 are required for CENP - A loading and histone deacetylation at centromeres . Cell 118 , 715 \u2013 729 . ( doi : 10 . 1016 / j . cell . 2004 . 09 . 002 ) 77 . Lando D et al . 2012 Quantitative single - molecule microscopy reveals that CENP - A ( Cnp1 ) deposition occurs during G2 in fission yeast . Open Biol . 2 , 120078 . ( doi : 10 . 1098 / rsob . 120078 ) 78 . Balinovic A , Albrecht D , Endesfelder U . 2019 Spectrally red - shifted fluorescent fiducial markers for optimal drift correction in localization microscopy . J . Phys . D : Appl . Phys . 52 , 204002 . ( doi : 10 . 1088 / 1361 - 6463 / ab0862 ) 79 . Mlodzianoski MJ , Schreiner JM , Callahan SP , Smolkov\u00e1 K , Dlaskov\u00e1 A , Santorov\u00e1 J , Je \u017e ek P , Bewersdorf J . 2011 Sample drift correction in 3D fluorescence photoactivation localization microscopy . Opt . Express 19 , 15 009 \u2013 15 019 . ( doi : 10 . 1364 / OE . 19 . 015009 ) 80 . Lee H , Yu C - C , Boyden ES , Zhuang X , Kosuri P . 2021 Tetra - gel enables superior accuracy in combined super - resolution imaging and expansion microscopy . Sci . Rep . 11 , 16944 . ( doi : 10 . 1038 / s41598 - 021 - 96258 - y ) 81 . Campbell BC , Paez - Segala MG , Looger LL , Petsko GA , Liu CF . 2022 Chemically stable fluorescent proteins for advanced microscopy . Nat . Methods 19 , 1612 \u2013 1621 . ( doi : 10 . 1038 / s41592 - 022 - 01660 - 7 ) 82 . Kopek BG et al . 2017 Diverse protocols for correlative super - resolution fluorescence imaging and electron microscopy of chemically fixed samples . Nat . Protocols 12 , 916 \u2013 946 . ( doi : 10 . 1038 / nprot . 2017 . 017 ) 83 . Vojnovic I , Caspari OD , Ho \u015f kan MA , Endesfelder U . 2024 Combining single - molecule and expansion microscopy in fission yeast to visualize protein structures at the nanostructural level . Figshare . ( doi : 10 . 6084 / m9 . figshare . c . 7007803 ) r o y a l s o c i e t y pub l i s h i ng . o r g / j o u r n a l / r s o b O p e n B i o l . 14 : 230414 13 D o w n l o a d e d fr o m h tt p s : / / r oy a l s o c i e t ypub li s h i ng . o r g / o n 28 A p r il 2024",
    "wilsonEffectsBiologicalExamples2010": "The e\ufb00ects of biological examples in idea generation Jamal O . Wilson 1 and David Rosen , Woodru\ufb00 School of Mechanical Engineering , Georgia Institute of Technology , Atlanta , GA 30332 - 0405 , USA Brent A . Nelson 2 and Jeannette Yen , Center for Biologically Inspired Design , Georgia Institute of Technology , Atlanta , USA In engineering design , analogies can be used in conceptual design to aid in generating new and novel design ideas . In this paper , a cognitive study was performed to assess the impact of biological examples , which serve as surface dissimilar analogies , in the idea generation process during conceptual design . In this study , participants were exposed to biological examples during the idea generation process . These results were then compared to those of participants receiving no examples and to those receiving human - engineered examples . The results suggest that exposure to biological examples in idea generation can increase the novelty of design ideas generated after exposure without inhibiting the variety of the design ideas generated , unlike human - engineered examples which resulted in decreased variety . (cid:2) 2009 Elsevier Ltd . All rights reserved . Keywords : innovation , design techniques , conceptual design , creative design , evaluation I n the conceptual design phase of the engineering design process ( Pahl & Beitz , 1996 ) , the designer is tasked with searching for novel and innovative solutions to engineering problems . However , humans are imperfect search engines ( Busby & Lloyd , 1999 ) and tend to focus on a narrow range of solution approaches and overlook many valuable solutions ( Perttula , 2006 ) . According to the theory of bounded rationality ( Simon , 1976 , 1983 , 1996 ) , the space searched is bounded by the limited cognitive abilities of the designer ( Schild , Herstatt , & Luthje , 2004 ) . To overcome this limitation , designers often employ various techniques to aid in idea generation , which are listed elsewhere ( Pahl & Beitz , 1996 ) . These techniques aid the designer in expanding and exploring her / his design space more e\ufb03ciently . In engineering design , analogies are often used in idea generation to transfer knowledge through analogical mapping from a source domain containing the analogous phenomena to a target domain containing the problem to be solved by the analogy ( Mak & Shu , 2004 ) . Analogies can be classi\ufb01ed by the similarity , or the conceptual distance , between the source and target Corresponding author : Jamal O . Wilson jamal _ o _ wilson @ whirlpool . com www . elsevier . com / locate / destud 0142 - 694X $ - see front matter Design Studies 31 ( 2010 ) 169 e 186 doi : 10 . 1016 / j . destud . 2009 . 10 . 003 169 (cid:2) 2009 Elsevier Ltd . All rights reserved . domain . Local analogies are analogies where the source domain is similar to that of the target , where surface - level attributes and relations between these attributes can be easily mapped ( Dahl & Moreau , 2002 ) . Surface - level attributes are easily retrievable aspects of representation , such as color and shape ( Gentner & Markman , 1997 ) . In distant analogies , the source and target domains are very di\ufb00erent and are surface dissimilar . In this case , few surface - level attributes can be mapped and structural similarity must be relied upon ( Dahl & Moreau , 2002 ) . For illustration , consider the design of an air condi - tioning system for a hotel . A local , or surface similar , analog would be consid - ered commercial air conditioning systems from other companies , whereas a distant , or structurally similar , analogy would be self - cooling termite mounds . In the distant case , the termite mounds and commercial air condi - tioning systems share no surface - level attributes and are considered surface dissimilar . However , at the structural level , they both are analogous in provid - ing a cooling function . Tseng and co - authors ( Tseng , Moss , Cagan , & Kotovsky , 2008 ) studied the role of timing and similarity in the idea generation process . The authors found that distantly related information presented after participants had begun problem solving increased both the novelty and number of functionally dis - tinct solutions . This agrees with other work in the literature where the number of distant analogies used during design was positively correlated with the originality ( novelty ) of the resulting design ( Dahl & Moreau , 2002 ) and these distant analogies are considered the main drivers of innovative thought ( Ward , 1998 ) . However , the use of distant analogies does have challenges . Because distant analogies are surface dissimilar and involve analogous relationships at the structural level , it is often di\ufb03cult to transfer knowledge or solutions ( Johnson - Laird , 1989 ) . These analogies are more di\ufb03cult to access than local analogies because the nature of the similarity is at a more abstract , relational level , thus causing an increase in cognitive e\ufb00ort ( Dahl & Moreau , 2002 ) . In this work , we extend the work of Tseng and co - authors ( Tseng et al . , 2008 ) and prior literature by ( 1 ) testing the e\ufb00ect of surface dissimilar analogies in the form of biological systems and ( 2 ) more completely evaluating the e\ufb00ect of these analogs on the ideation process . In this research , we believe that biological examples provide a source of structurally similar ( and surface dis - similar ) analogies . Biologically inspired design ( BID ) describes the process of transferring solution approaches embodied in biological adaptations into other domains , such as engineering , materials science , and design . Because millions of years of evolution re\ufb01ned these locally optimized adaptations , such solutions may be useful for engineers in their search for energy - e\ufb03cient answers to similar problems raised by their technologies ( Vincent , 2002 ) . By adapting principles from these natural solutions , engineers have developed many innovative and novel products , including Velcro based on the hooked seeds of the burdock plant ( Velcro , 1955 ) , dry reusable tape based on Gecko 170 Design Studies Vol 31 No . 2 March 2010 feet ( Autumn et al . , 2002 ) and self - cleaning paint based on the leaves of the lotus plant ( Barthlott & Neinhuis , 1997 ) . Several researchers have begun de - veloping systematic approaches for leveraging biological examples in engineer - ing design , including functional keyword searches through biological literature ( Chiu & Shu , 2004 , 2005 ) , data repositories containing biological examples ( Bruck et al . , 2006 ; Chakrabarti , Sarkar , Leelavathamma , & Nataraju , 2005 ; Vincent , 2002 ; Vincent , Bogatyreva , Bogatyrev , Bowyer , & Pohl , 2006 ; Vincent & Mann , 2002 ; Yim , Wilson , & Rosen , 2008 ) , BID methods ( Benyus & Baumeister , 2007 ; Wilson & Rosen , 2007 , 2008 ) , and cognitive analyses of BID ( Helms , Vattam , & Goel , 2009 ) . However , there has been little empirical evidence on the value of bio - inspired techniques and the cognitive mechanisms behind these techniques in idea generation . In this work , we evaluate the e\ufb00ect of surface dissimilar analogies in the form of biological examples on the ideation process . Theories on the e\ufb00ects of expo - sure to examples in idea generation have converged on a dual in\ufb02uence model of both negative ( design \ufb01xation ) and positive ( cognitive stimulation ) e\ufb00ects ( Perttula , 2006 ) . Design \ufb01xation refers to design conformity that results from exposing designers to example solutions . Cognitive stimulation refers to the stimulation of new ideas that occurs as a result of exposure to the ideas of others . Jansson and Smith found that participants generated a larger amount of ideas containing features of examples to which they were exposed than did a control group not exposed to the examples ( Jannson & Smith , 1991 ) . Similar results were found in studies by Purcell and co - authors ( Purcell & Gero , 1996 ; Purcell , Williams , Gero , & Colbron , 1993 ) , but only when the principles from the example problem were considered to involve the same knowledge base as the expertise of the mechanical engineering participants ( Purcell & Gero , 1996 ; Purcell et al . , 1993 ) . In contrast , students in industrial design did not show these conformity e\ufb00ects ( Purcell & Gero , 1996 ; Purcell et al . , 1993 ) . Several other studies have shown similar conformity e\ufb00ects ( Marsh , Bink , & Hicks , 1999 ; Marsh , Landau , & Hicks , 1996 ; Ward , 1993 ) . Exposure to examples can also result in cognitive stimulation , thus having a positive in\ufb02uence on idea generation . In cognitive stimulation , the knowl - edge embodied in the examples stimulate ideas that designers would not other - wise have accessed ( Perttula , 2006 ) . In a study on idea generation in groups , Paulus and Yang found enhanced performance in group brainstorming due to cognitive stimulation ( Paulus & Yang , 2000 ) . Nijstad found that partici - pants exposed to ideas from a wide variety of categories surveyed more cate - gories of ideas than participants not exposed to any ideas or participants exposed to ideas from fewer categories ( Nijstad , Stroebe , & Lodewijkx , 2002 ) . The authors also argued that design problems with larger design spaces were more likely to show these stimulation e\ufb00ects than ones with smaller spaces . Dugosh also found stimulation e\ufb00ects in group brainstorming and The e\ufb00ects of biological examples in idea generation 171 increased amounts in cases where participants were asked to attend to the ideas of others ( Dugosh , Paulus , Roland , & Yang , 2000 ) . The main goal of this research was to examine the e\ufb00ect of surface dissim - ilar analogies in the form of biological examples on the ideation process . Speci\ufb01cally , the impact of analogies on design space exploration and expan - sion was quanti\ufb01ed . De\ufb01ned metrics ( Nelson , Wilson , Rosen , & Yen , 2009 ; Shah , Vargas - Hernandez , & Smith , 2002 ) relating the novelty and variety of ideas generated in ideation to the expansion and exploration of the design space were used . These metrics for novelty and variety were used to assess the value of surface dissimilar analogies in the form of biological examples in ideation . To examine this , an experimental study was conducted in which participants were exposed to biological examples in the idea generation pro - cess . These results were then compared to that of participants receiving no examples and to those receiving surface similar examples . 1 Hypotheses This study tested the following speci\ufb01c hypotheses : ( 1 ) exposure to surface dissimilar analogies in the form of biological design examples in ideation will increase the novelty of design ideas generated and ( 2 ) exposure to surface dissimilar analogies in the form of biological design examples in ideation will increase the variety of design ideas generated . 2 Experimental methods To test the hypotheses presented in Section 2 , a study was performed in which participants generated conceptual designs to solve con\ufb02icting design require - ments . In testing hypotheses 1 and 2 , participants were placed into one of three potential conditions : ( 1 ) a condition in which participants received a surface dissimilar biological example after some time in ideation , ( 2 ) a condition in which participants received a surface similar human - engineered example after some time in ideation , and ( 3 ) an un - aided condition where no example was presented . The un - aided condition established a baseline for un - aided concept generation . The surface similar condition established a baseline for concept generation using easily accessible , human made design examples , and the sur - face dissimilar condition assessed the e\ufb00ect of biological design examples in concept generation . At the onset of the study , participants were given a design problem , several sheets of paper , and a single pen color by the experimenters . Participants were randomly assigned to one of the experimental conditions . The assigned condition was designated by the color paper the participants received , al - though the participants were unaware of this . After instruction and time for questions , the participants were given 20 min of un - aided concept gener - ation time , then given their respective design aids if applicable , and then given another 20 min for further concept generation . At the beginning of 172 Design Studies Vol 31 No . 2 March 2010 the study , the participants were verbally instructed to generate and label distinct design ideas , but no instructions were given regarding quantity or creativity of the design ideas so as not the bias the participants . Participants were also given a new color of pen for the second 20 min of idea generation to enable the researchers to distinguish between designs generated during the initial 20 min and those from the second 20 min . When the participants were given their respective design aids , they were verbally instructed that the design aids may or may not prove useful to them and that they were not re - quired to use the materials . To test our hypotheses , comparisons were made both within groups and be - tween groups . The within - group analysis compared the design ideas generated before and after exposure to the design example , while the between - group analysis compared the ideas generated after exposure to the design example across each of the di\ufb00erent conditions . 3 Leg immobilization versus portability 3 . 1 Participants Twenty - six undergraduate mechanical engineering students from our institu - tion participated in Study 1 . The students were recruited from a senior level capstone design course . To be eligible for the course , students needed at least 3 years of undergraduate engineering coursework and at least one semester of a formal undergraduate design course . All participants were volunteers and did not receive payment or additional course credit for participation in the study . 3 . 2 Materials All participants were asked to individually solve the same design problem . The design problem given in this study , displayed in Figure A - 1 of the Appendix , asked participants to solve a design problem with con\ufb02icting design require - ments of leg immobilization and portability . 3 . 3 Experimental procedure The design study was carried out in a classroom setting in the second meeting of the senior capstone design course . At the onset of the study , participants were given the design problem , several sheets of paper , and a black pen by the experimenters . Participants were assigned to either the biological ( N \u00bc 9 ) , the human - engineered ( N \u00bc 9 ) , or the un - aided ( N \u00bc 8 ) condition in an alternating fashion based on seating location . Speci\ufb01c procedures for each condition were as follows . 3 . 3 . 1 Biological condition Participants in the biological condition were given a biological design ex - ample at the midpoint of the study that described the variable - sti\ufb00ness be - havior of the mutable connective tissue of the sea cucumber . The design The e\ufb00ects of biological examples in idea generation 173 example included a pictorial and textual description of the biological sys - tem and is displayed in Figure A - 2 of the Appendix . 3 . 3 . 2 Human - engineered condition Participants in the human - engineered condition received a design example with a pictorial and textual description of a human - engineered system with similar behavior to that of the biological system . The design example described the variable - sti\ufb00ness behavior of electro - rheological \ufb02uids and is displayed in Figure A - 3 of the Appendix . The human - engineered and biological examples were intentionally chosen to match each other very closely in functional principles . 3 . 3 . 3 Un - aided condition After the initial 20 min of concept generation , participants in the un - aided condition received no design example and were instructed to continue with concept generation after receiving a new pen color . After the allotted 40 min , the students were stopped and given a post survey to evaluate their prior knowledge and experience as well as their understanding of the design examples given to them . Figure 1 displays some examples of design solutions generated in this study . 3 . 4 Data analysis The impact of surface dissimilar examples on idea generation was measured by evaluating the novelty and variety of the conceptual designs generated by the participants . The novelty indicated the degree to which exposure to examples aided designers in expanding their design space , while the va - riety indicated how examples a\ufb00ected the degree to which designers ex - plored their design space ( Shah et al . , 2002 ) . Metrics for novelty and variety were assessed using the drawings and descriptions generated by the participants . 3 . 4 . 1 Novelty Novelty was assessed as de\ufb01ned in the literature ( Nelson et al . , 2009 ; Shah et al . , 2002 ) . First , the design problem was decomposed into its key functions . Next , each design idea generated by the participants was categorized on the basis of the principles used to address the key functions and characteristics of the design problem . Finally , the used principles of the designs were given a score corresponding to the degree to which they were unusual . For our study , the score was determined by the frequency of occurrence of the given design principle among the designs generated by all the participants in the \ufb01rst 20 min of the study , before exposure to design examples . This initial group of ideas was considered to be the original design space of the participants since it corresponded to ideas generated before design aids were given , and most stu - dents had stopped generating new ideas by the end of the initial 20 min . 174 Design Studies Vol 31 No . 2 March 2010 Such a measure of novelty by frequency of occurrence was shown to be similar to subjective novelty scores assigned by external judges ( Shah et al . , 2002 ) . The novelty score , N , for an idea with m functions or characteristics was calculated for each design idea using the equation . N \u00bc X m j \u00bc 1 f j T j (cid:2) C j T j \u00f0 1 \u00de where T j is the total number of participants in the study , C j is the total num - ber of participants that utilized the particular design principle in their initial designs , and f j is an optional weighting factor for function j and was set to unity in this study . The average novelty score was then calculated for each in - dividual by averaging the novelty scores for their designs . For example , con - sider the sample set of design concepts displayed in Figure 1 . In the initial idea generation session ( before exposure to examples ) , if many participants gener - ated ideas with the working principle \u2018in\ufb02atable\u2019 , the novelty score would be low ( C j (cid:3) T j ) . Alternatively , if no one generated a design idea containing a working principle of a variable sti\ufb00ness , composite \ufb02uid ( electro - rheological \ufb02uid ) , then the idea would be considered fully novel ( C j (cid:3) 0 ) and get a novelty score of 1 . 3 . 4 . 2 Variety Variety indicates the total design space explored , and was calculated using a genealogy tree of the design ideas ( Shah et al . , 2002 ) . The genealogy tree Figure 1 Sample of design ideas generated by students . ( a ) In\ufb02atable ( b ) multiple - part snap ( c ) electro - rheolog - ical \ufb02uid chambers with power source ( d ) chemically - rigidizable The e\ufb00ects of biological examples in idea generation 175 was constructed based on di\ufb00erentiation between designs at four levels of cat - egorization : physical principles , working principles , embodiment , and detail . While most of the concept of the variety metric was generated by Shah et al . ( Shah et al . , 2002 ) a re\ufb01nement of that metric was used in this study ( Nelson et al . , 2009 ) . Variety , V , was calculated using the equation . V \u00bc X m j (cid:2) 1 f j S 1 \u00f0 b 1 (cid:2) 1 \u00de \u00fe X 4 k \u00bc 2 S k X b k (cid:2) 1 l \u00bc 1 d l ! \u00f0 2 \u00de where f j is a weighting factor for function j , S k is the score given for di\ufb00er - entiation at level k , b k is the number of branches in the tree at level k , and m is the total number of functions . Di\ufb00erentiation between designs at higher levels of the genealogy tree receive a higher weighting , S k , than di\ufb00erentia - tion at lower levels of hierarchy . Weightings of 10 , 5 , 2 , and 1 were used at the physical principle , working principle , embodiment , and detail hierarchy levels , respectively , to ensure that di\ufb00erentiation at higher levels of hierar - chy were valued at least twice that of the level beneath it ( Nelson et al . , 2009 ) . For example , consider again the design solutions from Figure 1 as a single set of ideas . Figure 2 displays the genealogy tree that would represent this set . De - sign ideas ( a ) , ( c ) , and ( d ) have the same physical principle of using \u2018variable sti\ufb00ness\u2019 to satisfy the immobilization vs . portability con\ufb02ict . However , they have di\ufb00erent working principles to achieve variable sti\ufb00ness : ( a ) in\ufb02ation , ( c ) controllable , composite \ufb02uid , and ( d ) chemical hardening . Idea ( b ) uses a separate physical principle , \u2018assembly of multiple parts\u2019 , to solve the con\ufb02ict . The variety score would be calculated as 10 (cid:4) ( 1 node separating the two phys - ical principles ) \u00fe 5 (cid:4) ( 2 nodes separating the 3 working principles having a \u2018variable - sti\ufb00ness\u2019 physical principle ) \u00bc 10 (cid:4) ( 1 ) \u00fe 5 (cid:4) ( 2 ) \u00bc 20 . Note that unlike novelty , the variety score is not normalized by the total number of de - signs , which yields a truer measure of total design space exploration ( Nelson et al . , 2009 ) since variety is a metric that inherently only applies to a set of multiple ideas . ( b ) Assemblage ( a , c , d ) Variable Stiffness ( d ) Chemical hardening ( c ) controllable , composite fluid ( a ) inflation Physical Principle Working Principle Figure 2 Genealogy tree for sample design solutions in Figure 1 176 Design Studies Vol 31 No . 2 March 2010 For this study , two judges , who were also authors on this study , independently coded genealogy trees for the four level categorization scheme based on the idea set generated by the participants . These trees were then compared and showed mostly similar groupings , with one of the trees di\ufb00ering primarily in that it had fewer branches but most of the same descriptions . Unaligned points of the independent trees were discussed and reconciled by discussion and con - sensus between the judges . This helped to assure accuracy in the \ufb01nal master genealogy tree , which was used to calculate the variety metric . The partici - pants\u2019 ideas were then allocated within the genealogy tree and the variety score was calculated . The \ufb01nal genealogy tree used to calculate variety in the study is displayed in Figure 3 . 4 Results The novelty and variety of the design ideas generated by the participants were calculated using the metrics de\ufb01ned in Section 3 . 3 . The results were analyzed using nonparametric Wilcoxon signed - ranked tests and Mann e Whitney U tests at an alpha level of 0 . 05 ( a \u00bc 0 . 05 ) . The more conservative nonparametric tests were chosen because the small sample sizes resulted in non - normally distrib - uted results . 4 . 1 Novelty The \ufb01rst hypothesis stated that exposure to surface dissimilar analogies in the form of biological design examples would lead to design ideas of greater nov - elty . This hypothesis was tested by comparing the novelty scores of the partic - ipants in the biological condition before and after the design example was introduced . These results were then compared to that of the human - engineered and un - aided conditions . The mean and the standard deviation ( SD ) of the novelty scores for the three conditions are displayed in Table 1 . Immobolize Leg / Storage Elastic helix brace wrap Collapsible Rigid Structure tent - pole Sliding Telescoping Other sliding Folding / Ratcheting / Hinging Wooden plates and adjustable rings Shape memory Detachable multiple / segmented crutch rigid / soft elements Sleeve : soft w / rigid inserts Splint : rods / straps Thin / rollable Aluminum tape rolled splint ( Slap / tape meas . ) Variable Stiffness Inflatable Inflatable splint inflatable sleeve Chemical stiff 2 - part fiberglass / bondo epoxy thermal - set foaming cement crystal structure Composite fluid ER stiff Fibrils Dual purpose materials tie legs together gear backpack frame hiking stick scavenge environmental wood splint makeshift stretcher makeshift crutch Medical paralyze leg anesthetic Figure 3 Genealogy tree for the entire set of designs produced in the study The e\ufb00ects of biological examples in idea generation 177 Analysis of the results from Table 1 yields the following : (cid:5) Participants in both the biological and human - engineered condition showed statistically signi\ufb01cant increases in novelty ( W \u00bc 2 , p < 0 . 01 and W \u00bc 0 , p < 0 . 01 , respectively ) after being exposed to the example ; participants in the un - aided condition showed no signi\ufb01cant change in the novelty of their design ideas . (cid:5) Participants in both the biological and human - engineered conditions gener - ated design ideas with higher novelty ( U \u00bc 68 , p < 0 . 01 and U \u00bc 68 , p < 0 . 01 , respectively ) than participants in the un - aided condition . The dif - ference between the novelty scores of participants in the biological and human - engineered condition approached signi\ufb01cance ( U \u00bc 55 , p < 0 . 07 ) . These results supported the hypothesis that exposure to surface dissimilar analogies in the form of biological examples leads to design ideas of greater novelty . The same e\ufb00ect also occurred with participants exposed to the surface similar ( human - engineered ) example . 4 . 2 Variety The second hypothesis examined was that exposure to surface dissimilar anal - ogies in the form of biological examples would lead to a greater variety of de - sign ideas . This hypothesis was tested by comparing the variety of the design ideas produced by the participants before and after the design examples were introduced . Table 2 displays the variety scores of participants in the three conditions . The results from Table 2 are summarized as follows : (cid:5) Participants in the biological and un - aided conditions showed no statisti - cally signi\ufb01cant change in the variety of design ideas during the two phases of the study , while participants in the human - engineered condition showed a signi\ufb01cant decrease ( W \u00bc 0 , p < 0 . 01 ) in the variety of design ideas gener - ated after receiving the example . (cid:5) The variety of design ideas of participants in the biological and un - aided conditions did not di\ufb00er signi\ufb01cantly . The hypothesis that exposure to surface dissimilar analogies in the form of bi - ological examples would lead to a greater variety of design ideas was Table 1 Novelty scores Unaided ( SD ) Bio - inspired ( SD ) Human - engineered ( SD ) Before 0 . 32 ( 0 . 26 ) 0 . 36 ( 0 . 08 ) 0 . 28 ( 0 . 19 ) After 0 . 41 ( 0 . 33 ) 0 . 76 ( 0 . 24 ) 0 . 94 ( 0 . 11 ) 178 Design Studies Vol 31 No . 2 March 2010 unsupported with these results . However , while the biological example did not increase the variety of design ideas relative to the un - aided condition , it did not signi\ufb01cantly decrease the variety of the design ideas , as had resulted in the hu - man - engineered ( surface similar ) condition . 5 Discussion The results from the studies presented in this paper supported the hypoth - eses that exposure to surface dissimilar analogies in the form of biological examples in the idea generation process increases the novelty of design ideas . The results from the study did not support the hypothesis that ex - posure to surface dissimilar analogies in the form of biological examples in the idea generation process would increase the variety of design ideas generated . However , the surface similar examples negatively a\ufb00ected the observed variety , and this did not occur for the surface dissimilar condition . In this work , novelty was de\ufb01ned as the uniqueness of an idea with respect to the un - aided set of ideas generated by the group of participants . Hypothesis 1 stated that exposure to surface dissimilar analogies in the idea generation process would increase the novelty of ideas generated , and the results indi - cated that exposure to biological examples did indeed increase the novelty of ideas generated . This increase in novelty signi\ufb01es that participants were able to generate ideas that otherwise would not have been accessed without exposure to the biological example . A signi\ufb01cant increase in the novelty of participants receiving the surface similar example was also found . The in - crease in novelty cannot be attributed solely to increased ideation time , since participants not receiving any examples showed no increase in the novelty of their design ideas during the second ideation phase . The increase in novelty can be attributed to the use of a non - obvious solution approach in the stimulation examples given at the midpoint of the study . Most students had limited familiarity with variable - sti\ufb00ness materials , and those with \ufb01rst aid training would have been taught about rigid splints already , so the ma - jority of their pre - example design concepts did not utilize the idea of variable sti\ufb00ness . Because the baseline for the novelty calculation was determined from only those concepts generated before the example was given , the later ideas utilizing variable - sti\ufb00ness materials yielded a high novelty score . This implies that merely the presence of external stimulation in the ideation process increases the novelty of the design ideas generated . Shah et al . Table 2 Variety scores Unaided ( SD ) Bio - inspired ( SD ) Human - engineered ( SD ) Before 5 . 8 ( 5 . 6 ) 12 . 1 ( 7 . 8 ) 13 . 3 ( 9 . 3 ) After 5 . 6 ( 6 . 8 ) 9 . 4 ( 7 . 6 ) 2 . 4 ( 3 . 7 ) The e\ufb00ects of biological examples in idea generation 179 ( Shah et al . , 2002 ) correlated this increase in novelty to a broadening of the design space of the designer . The type of analogy ( surface similar vs . surface dissimilar ) was also found to have an e\ufb00ect on novelty . Participants receiving surface similar examples produced more novel ( near signi\ufb01cance ) designs than those receiving surface dissimilar examples . One possible explanation lies in the manner in which participants used the design examples . Because surface dissimilar analogies have a larger conceptual distance to the design problem , participants are less likely to view them as relevant and thus , less likely to utilize them . In this case , participants receiving the surface similar example transferred some attribute of the human - engineered example to 100 % of their design ideas . In contrast , participants receiving the surface dissimilar example only transferred some attribute of the biological example 52 % of the time . Since the use of external stimuli helps increase the novelty of design ideas , this only holds when these external stimuli are utilized . In this case , the larger conceptual distance of the surface dissimilar examples leads to a lower utilization rate versus that of the surface similar examples where attributes can be easily transferred . Hypothesis 2 stated that exposure to surface dissimilar analogies would in - crease the variety of the design ideas generated . This hypothesis was un - supported by the results . The variety before and after exposure to the biological example showed no statistically signi\ufb01cant di\ufb00erence in variety . However , even though exposure to the biological examples did not in - crease the variety of ideas generated , such exposure also did not signi\ufb01 - cantly decrease the variety of the generated ideas . Exposure to examples in idea generation has been shown to cause \ufb01xation ( Jannson & Smith , 1991 ; Purcell & Gero , 1996 ; Purcell et al . , 1993 ) , which decreases the va - riety of the generated ideas . When comparing the variety scores between the bio - inspired and human - engineered condition in Table 2 , participants exposed to surface dissimilar analogies had a signi\ufb01cantly higher variety ( U \u00bc 17 . 5 , p < . 05 ) than participants exposed to surface similar examples . This suggests that the participants exposed to the surface similar example \ufb01xated more so than those exposed to the surface dissimilar example . A possible explanation for this phenomenon can be found by examining the \ufb01xation e\ufb00ect in \ufb01ner detail . As mentioned previously , participants exposed to the surface dissimilar example transferred less attributes from the example problem to their de - sign ideas than participants exposed to the surface similar example . This lower utilization rate leaves way for ideas that are on di\ufb00erent ideation pathways , thus keeping the design space open and fostering variation in the ideas . An additional explanation for the larger variety in participants exposed to surface dissimilar examples can also be found by looking at the 180 Design Studies Vol 31 No . 2 March 2010 level at which attributes were transferred . Knowledge can be transferred from design examples at varying levels of abstraction . Higher levels of ab - straction when transferring from an analogical source would correspond to utilizing the principles and mechanisms from the source ( i . e . structural relationships ) , whereas lower levels of abstraction would correspond to utilizing speci\ufb01c materials and devices from the source ( i . e . surface - level attributes ) . To assess the level of transfer of ideas , the design examples given to the students were \ufb01rst decomposed into their key characteristics ( i . e . physical principles , working principles , etc . ) by the evaluators . Next , the ideas generated by the participants were compared to the char - acteristics of the design example . The number of design ideas that utilized the physical principle and working principle of the example was then de - termined for each participant . The results are displayed in Table 3 . The values in Table 3 represent the average percentage of the participants\u2019 ideas generated after exposure to the example that utilized characteristics of the example at either the physical principle or working principle level of abstraction . As seen in Table 3 , and as mentioned above , participants in the surface sim - ilar condition transferred more characteristics at both the physical princi - ple , and the more basic working principle level of abstraction . The signi\ufb01cance of the results in Table 3 lies in that participants exposed to the surface similar examples transferred signi\ufb01cantly more characteristics ( U \u00bc 61 . 5 , p < 0 . 01 ) at the working principle level of abstraction than those exposed to biological examples . For example , participants in the surface similar condition primarily suggested designs speci\ufb01cally utilizing electro - rheological \ufb02uids whereas participants in the biological condition suggested designs that utilized the general principle of a variable - sti\ufb00ness material . This means that those exposed to the surface similar example transferred more surface - level ( working principle ) characteristics . In this case , \ufb01xation on surface - level characteristics could have constrained the variety of ideas in idea generation . In the surface dissimilar case , transferring principles at a higher level of abstraction ( structural level ) allowed a greater variety of ideas to still be generated . An additional reason that biological analogical sources encourage higher level abstraction is that biological analogies typically use mechanisms and Table 3 Level of transfer scores Level of transfer Bio - inspired Human - engineered Physical principle 52 % 100 % Working principle 19 % 80 % The e\ufb00ects of biological examples in idea generation 181 materials that are not commercially available products . A designer utilizing sea cucumber connective tissue as an analogical source cannot plausibly im - plement sea cucumber connective tissue within a design , and thus the de - signer is forced to understand the design at higher functional levels . In contrast , designers presented with electro - rheological \ufb02uids can implement such \ufb02uids within their designs and are not forced to determine the higher level physical principle e that of a variable - sti\ufb00ness material . The greater do - main distance of the biological example forces analogical transfer at higher levels of abstraction . Participants in both the surface similar and dissimilar cases \ufb01xated on their respective examples , but at di\ufb00erent levels of abstraction . Those receiving surface similar examples \ufb01xated and transferred characteristics from both the surface and structural levels , while those receiving surface dissimilar ex - amples transferred characteristics at the structural level . This \ufb01xation at lower levels constrained the searchable design space and the \ufb01xation e\ufb00ect was greater than that of stimulation . In the case of the surface dissimilar ex - ample , while the participants still \ufb01xated , the higher level of abstraction still allowed stimulation of a broad range of new ideas and outweighing the \ufb01x - ation e\ufb00ect . 6 Conclusions Currently , there is little empirical evidence as to the e\ufb00ects of surface dis - similar analogies from the biological domain in the idea generation process , despite the existence several research e\ufb00orts devoted to the process of ana - logical mapping from the biological domain . The aim of this work was to quantify the value of these techniques in ideation as it relates to design space expansion and exploration . In this study , we describe experimental re - sults for the concept generation demonstrated by senior mechanical engi - neering students exposed to biological examples during the idea generation process . We compared these results to participants receiving no examples as well as to those receiving human - engineered ( surface simi - lar ) examples . The results indicated that , in this case , exposure to biological examples increased the novelty of design ideas generated after exposure without decreasing the variety of design ideas generated . The results also indicated that the type of example used ( surface dissimilar vs . surface sim - ilar ) had a signi\ufb01cant impact on the novelty and variety of design ideas gen - erated . When compared to surface similar examples , the biological examples showed a lower utilization rate , which negatively impacted the novelty of the design idea set . However , this lower utilization rate coupled with the transfer of attributes from the biological example at a high level of abstraction positively impacted the variety of design ideas generated . For this reason , biological examples thus show potential as a resource for iden - tifying surface dissimilar analogies . 182 Design Studies Vol 31 No . 2 March 2010 The results of the work presented in this paper have a number of implications in engineering design . One of the primary goals of engineering design is to dis - cover new and innovative solutions to encountered problems . In the develop - ment of these solutions , idea generation is paramount . Dylla has demonstrated signi\ufb01cant correlation between the amount of design space considered in idea generation and the quality of the \ufb01nal design ( Dylla , 1991 ) . It follows that methods for aiding designers in expanding and exploring their design space should yield better designs . Search strategies that include surface dissimilar analogies in the form of biological examples may help expand the design space of the designer while also limiting some of the restrictive e\ufb00ects of \ufb01xation typ - ically associated with exposure to examples . These results have been found for senior mechanical engineering students , but are anticipated to be extendable beyond the current sample group . The scope of this research was limited to the investigation of biological anal - ogies as a potential source of surface dissimilar analogies . Future research in this area should include an investigation and comparison amongst a variety of sources of surface dissimilar analogies . A larger sample of participants would also be useful in the expanding the validity space of this research . Acknowledgements JOW acknowledges \ufb01nancial support from the O\ufb03ce of Naval Research and the David and Lucille Packard Foundation . BAN acknowledges \ufb01nancial support from an NAE CASEE postdoctoral fellowship . Appendix Figure A - 1 Design problem for Study 1 The e\ufb00ects of biological examples in idea generation 183 Figure A - 2 Biological design example for Study 1 Figure A - 3 Human - engineered design example for Study 1 184 Design Studies Vol 31 No . 2 March 2010 1 . Present address : Department of Mechanical Engineering , Northern Arizona University , Flagsta\ufb00 , AZ 86011 - 5600 , USA . 2 . Present address : Whirlpool Corporation , 750 Monte Rd . , MD 5155 , Benton Harbor , MI 49022 , USA . References Autumn , K . , Sitti , M . , Lang , Y . A . , Peattie , A . M . , Hansen , W . R . , Sponberg , S . , et al . ( 2002 ) . Evidence of van der Waals adhesion in gecko setae . Proceedings of the National Academy of Sciences of the United States of America , 99 , 12252 e 12256 . Barthlott , W . , & Neinhuis , C . ( 1997 ) . Purity of the sacred lotus , or escape from contamination in biological surfaces . Planta , 202 ( 1 ) , 1 e 8 . Benyus , J . , & Baumeister , D . ( 2007 ) . The designspiral . www . biomimicryinstitute . org . Bruck , H . A . , Gershon , A . L . , Golden , I . , Gupta , S . K . , Lawrence , S . , Gyger Jr . , et al . ( 2006 ) . New educational tools and curriculum enhancements for motivat - ing engineering students to design and realize bio - inspired products . In C . A . Brebbia ( Ed . ) , Design and Nature III : Comparing Design in Nature with Science and Engineering . WIT Press . Busby , J . A . , & Lloyd , P . A . ( 1999 ) . In\ufb02uences on solution search processes in design organisations . Research in Engineering Design , 11 , 158 e 171 . Chakrabarti , A . , Sarkar , P . , Leelavathamma , B . , & Nataraju , B . S . ( 2005 ) . Afunctional representationforaidingbiomimeticandarti\ufb01cialinspirationofnewideas . Arti\ufb01cial Intelligence for Engineering Design , Analysis and Manufacturing , 19 , 113 e 132 . Chiu , I . , & Shu , L . H . ( 2004 ) . Natural Language Analysis for Biomimetic Design . 2004 DETC Design Theory and Methodology . Salt Lake City , Utah : ASME . Chiu , I . , & Shu , L . H . ( 2005 ) . Bridging Cross - Domain Terminology for Biomimetic Design . 2005 IDETC Design Theory and Methodology . Long Beach , CA : ASME . Dahl , D . W . , & Moreau , P . ( 2002 ) . The in\ufb02uence and value of analogical thinking during new product ideation . Journal of Marketing Research , XXXIX , 47 e 60 . Dugosh , K . L . , Paulus , P . B . , Roland , E . J . , & Yang , H . - C . ( 2000 ) . Cognitive stimula - tion in brainstorming . Journal of Experimental Social Psychology , 79 ( 5 ) , 722 e 735 . Dylla , N . ( 1991 ) . Thinking methods and procedures in mechanical design . Me - chanical design , technical university of Munich , PhD . Gentner , D . , & Markman , A . B . ( 1997 ) . Structure mapping in analogy and sim - ilarity . American Psychologist , 52 ( 1 ) , 45 e 56 . Helms , M . , Vattam , S . , & Goel , A . K . ( 2009 ) . Biologically inspired design : process and products . Design Studies , 30 , 606 e 622 . Jannson , D . G . , & Smith , S . M . ( 1991 ) . Design \ufb01xation . Design Studies , 12 ( 1 ) . Johnson - Laird , P . N . ( Ed . ) . ( 1989 ) . Analogy and the Exercise of Creativity . Simi - larity and Analogical Reasoning . New York , NY : Cambridge University Press . Mak , T . W . , & Shu , L . H . ( 2004 ) . Use of Biological Phenomena in Design by Anal - ogy . ASME 2004 Design Engineering Technical Conference . ASME . Salt Lake City , UT : ASME . Marsh , R . L . , Bink , M . L . , & Hicks , J . L . ( 1999 ) . Conceptual priming in a gener - ative problem - solving task . Memory & Cognition , 27 ( 2 ) , 355 e 363 . Marsh , R . L . , Landau , J . D . , & Hicks , J . L . ( 1996 ) . How examples may ( and may not ) constrain creativity . Memory & Cognition , 24 ( 3 ) , 669 e 680 . Nelson , B . A . , Wilson , J . , Rosen , D . , & Yen , J . ( 2009 ) . Re\ufb01ned metrics for measuring ideation e\ufb00ectiveness . Design Studies , 30 , 737 e 743 . Nijstad , B . A . , Stroebe , W . , & Lodewijkx , H . F . ( 2002 ) . Cognitive stimulation and interference in groups : exposure e\ufb00ects in an idea generation task . Journal of Experimental Social Psychology , 38 ( 6 ) , 535 e 544 . The e\ufb00ects of biological examples in idea generation 185 Pahl , G . , & Beitz , W . ( 1996 ) . Engineering Design : a Systematic Approach . London : Springer - Verlag . Paulus , P . B . , & Yang , H . - C . ( 2000 ) . Idea generation in groups : a basis for crea - tivity in organizations . Organizational Behavior and Human Decision Processes , 82 ( 1 ) , 76 e 87 . Perttula , M . K . ( 2006 ) . Idea generation in engineering design : application of memory searchperspectiveandsomeexperimentalstudies , DepartmentofMechanicalEn - gineering , Helsinki University of Toronto , Doctor of Science in Technology , 62 . Purcell , A . T . , & Gero , S . J . ( 1996 ) . Design and other types of \ufb01xation . Design Studies , 17 ( 4 ) , 363 e 383 . Purcell , A . T . , Williams , P . , Gero , J . S . , & Colbron , B . ( 1993 ) . Fixation e\ufb00ects : do they exist in design problem solving ? Environment and Planning B : Planning and Design , 20 ( 3 ) , 333 e 345 . Schild , K . , Herstatt , C . , & Luthje , C . ( 2004 ) . How to Use Analogies for Break - through Innovations . Hamburg , Germany : Technical University of Hamburg . Shah , J . J . , Vargas - Hernandez , N . , & Smith , S . M . ( 2002 ) . Metrics for measuring ideation e\ufb00ectiveness . Design Studies , 24 , 111 e 134 . Simon , H . A . ( 1976 ) . Administrative Behavior . New York , N . Y : The Free Press . Simon , H . A . ( 1983 ) . Reason in Human A\ufb00airs . Stanford , CA : Stanford University Press . Simon , H . A . ( 1996 ) . The Sciences of the Arti\ufb01cial . Cambridge , MA : MIT Press . Tseng , I . , Moss , J . , Cagan , J . , & Kotovsky , K . ( 2008 ) . The role of timing and an - alogical similarity in the stimulation of idea generation in design . Design Stud - ies , 29 , 203 e 221 . Velcro , S . A . ( 1955 ) . \u2018\u2018Improvements in or relating to a method and a device for producing a velvet type fabric , \u2019\u2019 Switzerland , Patent no 721338 . Vincent , J . F . V . ( 2002 ) . Stealing ideas from nature . In S . Pellegrino ( Ed . ) , Deploy - able Structures ( pp . 51 e 58 ) . Italy : Springer Wein New York . Vincent , J . F . V . , Bogatyreva , O . A . , Bogatyrev , N . R . , Bowyer , A . , & Pahl , A . - K . ( 2006 ) . Biomimetics : its practice and theory . Journal of the Royal Society Inter - face , 3 , 471 e 482 . Vincent , J . F . V . , & Mann , D . L . ( 2002 ) . Systematic technology transfer from bi - ology to engineering . Philosophical Transactions of the Royal Society of London A , 360 , 159 e 173 . Ward , T . B . ( 1993 ) . Structured imagination : the role of category structure in ex - emplar generation . Cognitive Psychology , 27 ( 1 ) , 1 e 40 . Ward , T . B . ( 1998 ) . Analogical distance and purpose in creative thought : mental leaps versus mental hops . In K . Holyoak , D . Gentner , & B . Kokinov ( Eds . ) , Ad - vances in Analogy Research : Integration of Theory and Data from the Cognitive , Computational , and Neural Sciences . So\ufb01a : New Bulgarian University Press . Wilson , J . , & Rosen , D . ( 2007 ) . Systematic Reverse Engineering of Biological Sys - tems . 2007 IDETC Design Theory and Methodology . Las Vegas , NV : ASME . Wilson , J . , & Rosen , D . ( 2008 ) . A systematic method for transferring biological strat - egiesinengineeringdesign : applicationtothedesignofvariable - sti\ufb00nessmaterials , Biological Approaches for Engineering Conference , Southampton , UK . Yim , S . , Wilson , J . , & Rosen , D . ( 2008 ) . Development of an Ontology for Bio - In - spired Design Using Description Logics , International Conference on Product Lifecycle Management , Seoul , Korea . 186 Design Studies Vol 31 No . 2 March 2010",
    "sitarska2023sensing": "Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Sensing their plasma membrane curvature allows migrating cells to circumvent obstacles Ewa Sitarska 1 , 2 , Silvia Dias Almeida 1 , 8 , Marianne Sandvold Beckwith 1 , Julian Stopp 3 , Jakub Czuchnowski 1 , Marc Siggel 4 , 5 , Rita Roessner 4 , 5 , Aline Tschanz 1 , 2 , Christer Ejsing 1 , 6 , Yannick Schwab 1 , Jan Kosinski 4 , 5 , 7 , Michael Sixt 3 , Anna Kreshuk 1 , Anna Erzberger 1 & Alba Diz - Mu\u00f1oz 1 To navigate through diverse tissues , migrating cells must balance persistent self - propelled motion with adaptive behaviors to circumvent obstacles . We identify a curvature - sensing mechanism underlying obstacle evasion in immune - like cells . Speci \ufb01 cally , we propose that actin polymerization at the advancing edge of migrating cells is inhibited by the curvature - sensitive BAR domain protein Snx33 in regions with inward plasma membrane curvature . The genetic perturbation of this machinery reduces the cells \u2019 capacity to evade obstructions combined with faster and more persistent cell migration in obstacle - free environments . Our results show how cells can read out their surface topography and utilize actin and plasma membrane biophysics to interpret their environment , allowing them to adaptively decide if they should move ahead or turn away . On the basis of our \ufb01 ndings , we propose that the natural diversity of BAR domain proteins may allow cells to tune their curva - ture sensing machinery to match the shape characteristics in their environment . Cell migration drives many developmental , physiological , and patho - logical processes . While the mechanisms underlying propulsion are largely known , it remains unclear how cells steer their movement to navigate complex and dynamic environments 1 and circumvent obsta - cles within tissues 2 , 3 . These adaptive behaviors are particularly important for fast and dynamic cell types , such as immune cells and disseminating tumor cells . Switching between actin - based protrusions and blebbing has been shown to aid cell steering in development 4 , but cells often display only actin - rich protrusions such as lamellipodia and ruf \ufb02 es . These need to be long - lived enough for cells to explore and then persistently move through their surroundings , while \u2018 unstable \u2019 enough to allow them to adapt and change direction when confronted with an obstacle 5 , 6 . During migration , the outmost boundary of the cell \u2014 the plasma membrane \u2014 is deformed by the changing microenvironment . How - ever , it remains unclear whether membrane curvature encodes infor - mation that cells use to choose their migration path 7 , 8 , or if membrane topography is only a side - effect of the forces acting at the cell surface . Many migrating cells indeed express a variety of curvature - sensing proteins , which can directly interact with both the plasma membrane and the subjacent actin cytoskeleton 9 . In particular , the family of BAR domain proteins could facilitate membrane curvature sensing . These Received : 18 October 2022 Accepted : 22 August 2023 Check for updates 1 Cell Biology and Biophysics Unit , European Molecular Biology Laboratory , 69117 Heidelberg , Germany . 2 Collaboration for joint PhD degree between EMBL andHeidelbergUniversity , FacultyofBiosciences , EMBLandHeidelbergUniversity , Heidelberg , Germany . 3 InstituteofScienceandTechnologyAustria , 3400 Klosterneuburg , Austria . 4 EMBLHamburg , EuropeanMolecularBiologyLaboratory , 22607Hamburg , Germany . 5 CentreforStructuralSystemsBiology , 22607 Hamburg , Germany . 6 Department of Biochemistry and Molecular Biology , Villum Center for Bioanalytical Sciences , University of Southern Denmark , Cam - pusvej 55 , 5230 Odense , Denmark . 7 Structural and Computational Biology Unit , European Molecular Biology Laboratory , 69117 Heidelberg , Germany . 8 Present address : Division of Medical Image Computing , German Cancer Research Center ( DKFZ ) , 69120 Heidelberg , Germany . e - mail : diz @ embl . de Nature Communications | ( 2023 ) 14 : 5644 1 1 2 3 4 5 6 7 8 9 0 ( ) : , ; 1 2 3 4 5 6 7 8 9 0 ( ) : , ; proteins form crescent - shaped membrane - binding dimers that can sense and generate curvature 9 \u2013 11 , and are known to interact with reg - ulators of the actin cytoskeleton such as the Arp2 / 3 complex , formins , or Rho GTPases through several auxiliary domains 9 , 12 . In fact , changes in surface topography induce recruitment of some cytosolic BAR domain proteins to regions of highly positive ( inward ) curvature 13 , and can trigger the formation of actin structures and endocytosis hotspots 14 , 15 . Here , we show that the curvature - sensing BAR domain protein Snx33 inhibits actin polymerization and rearranges the locali - zation of WAVE2 to limit the persistence of the leading edge and steer migration . By this mechanism Snx33 allows cells to adaptively decide if they should move ahead or turn away . Results Snx33 is preferentially excluded from outward - curved mem - brane regions at the leading edge of migrating cells During the differentiation process immune - like HL - 60 cells undergo substantial changes in gene expression , initiate rapid migration , and acquire a complex membrane topography , resembling what occurs in the bone marrow in vivo 16 ( Supplementary Fig . 1a ) . The terminally differentiated , motile cells ( dHL - 60 cells ) display both actin - rich lamellipodia and membrane ruf \ufb02 es at the leading edge ( Fig . 1a , b ) . To analyze membrane topography in more detail , we used scanning electron microscopy ( SEM ) , for ultrastructural detail of natively \ufb01 xed cell membranes , and polarized total internal re \ufb02 ection \ufb02 uorescence microscopy ( p - TIRFM ) , which enables probing membrane curvature dynamics in live cells . We observed curved membrane patterns , as seen by SEM in the upper plasma membrane ( Fig . 1c ) , that are very dynamic , as visualized by p - TIRFM in the basal plasma membrane ( Fig . 1d , Supplementary Fig . 1b \u2013 d , and Supplementary Video 1 ) . These curved membrane patterns likely constitute binding sites for curvature - sensitive proteins , such as those from the BAR domain family . To identify candidates from this large family that could sense indentingobstacles andthusbe relevant foradaptive behaviorsduring cell migration , we measured their expression pro \ufb01 le before and after HL - 60 differentiation to a migratory state , i . e . , in a curvature - poor ( blasts are uniformly round ) versus a curvature - rich state ( Supple - mentary Fig . 1a ) . The expression of the BAR domain protein Snx33 increased 16 - fold during this differentiation process ( Fig . 1e ) . We imaged its subcellular localization by confocal microscopy after \ufb02 uorescent tagging ( eGFP - Snx33 ) and found that Snx33 is enriched throughout membrane ruf \ufb02 es ( Fig . 1f , g ) , the highly curved structures at the leading edge of dHL - 60 cells . Based on the similarity within its protein family , Snx33 is predicted to bind shallow curvatures in com - parison to many other BAR domain proteins from the BAR subgroup 9 . The membrane - binding domain of Snx33 ( PX - BAR ) contains a posi - tivelychargedpatchontheconvexsurface , whichstronglysuggestsan inward curvature - dependent membrane - binding on this surface ( Supplementary Fig . 2a \u2013 d ) . To assess its curvature - sensing properties in more detail , we performed long coarse - grained molecular dynamics ( MD ) simulations of the PX - BAR domain ( Fig . 1h \u2013 j ) on buckled mem - branes with plasma membrane - derived lipid composition ( Table S1 ; see Methods for details ) . This computational assay has been used extensively to characterize curvature - sensing properties of a wide variety of proteins 17 , 18 , including other BAR domains 19 . During the 30 \u00b5 s long simulation , the PX - BAR domain of Snx33 remained tightly bound to the membrane and showed a strong preference for membrane regions with inward local curvature ( Fig . 1i , j , Supplementary Fig . 2e , f ) . Snx33 binds the plasma membrane with a surface of basic residues such as lysine and arginine that mediate electrostatic interactions ( Supplementary Figs . 2a \u2013 d , 3a , b , 4a \u2013 g , see Supplemental Discussion for details ) , which is common for plasma membrane binding 20 . To further test the curvature sensitivity of Snx33 in cells , we imaged its subcellular localization by lattice light sheet microscopy and , sup - porting our MDsimulations , we found that Snx33 isexcluded from the ruf \ufb02 e tips and lamellipodium edge , both structures with highly nega - tive ( outward ) curvature ( Fig . 1k \u2013 m , Supplementary Fig . 5 , and Sup - plementary Video 2 , see Methods for details ) . To corroborate the tip exclusionofSnx33 , wecompareditslocalizationtothatofIRSp53 ( also known as BAIAP2 ) , acanonicaloutward curvature - bindingprotein 21 . As previously reported , in neutrophil - like cells IRSp53 accumulates at the tip of the advancing front and retracting \ufb01 bers 22 , 23 , known outward - curved structures ( Supplementary Fig . 6a \u2013 c and Supplementary Video 3 ) . Altogether , we show that the BAR domain - containing protein Snx33 has a strong preferencefor membrane regions with inward local curvature and is enriched at the leading edge but excluded from outward - curved membrane structures . Snx33 controls leading edge growth and modulates plasma membrane tension TostudytheroleofSnx33inadaptivemigration , wegeneratedaSnx33 knockout ( Snx33 - / - ) cell line using CRISPR / Cas9 genome editing ( SupplementaryFig . 7a , b ) . Propulsioninimmune - likecellsdependson the active polymerization of actin at the leading edge of the cell 24 , 25 . Given that cell shape re \ufb02 ects changes in motion - driving actin - rich protrusions 26 , we performed a quantitative and unbiased comparison of selected cell morphometric parameters , i . e . , cell spreading , cell eccentricity , and leading - edge characteristics . As immune cells radi - cally change their morphology in short periods of time , we averaged time - lapse movies to capture their dynamics . To this end , we trained and used machine - learning - based algorithms in ilastik 27 to analyze movies of cells imaged by total internal re \ufb02 ection \ufb02 uorescence microscopy ( TIRFM ) ( see Methods for details ) . Snx33 - / - cells spread to a larger extent , and showed a more elongated morphology and bigger leading edge ( Fig . 2a \u2013 f , Supplementary Fig . 8a \u2013 e , and Supplementary Videos 4 and 5 ) , independent of adhesion to the substrate ( Supple - mentary Fig . 8f \u2013 i ) . Moreover , the increase in spread area and leading edge area could be rescued by expressing \ufb02 uorescently tagged Snx33 proving the speci \ufb01 city of the observed phenotypes ( Supplementary Fig . 8b , c ) . Altogether , these results indicate that Snx33 - / - cells have a more stable leading edge during migration . Leading edge growth and increased cell spreading during persis - tent migration is known to increase plasma membrane tension 28 , 29 . To test if the more stable leading edge induced by the loss of Snx33 leads to higher membrane tension , we measured it in the leading edge by static tether pulling using atomic force spectroscopy , where a plasma membrane tether is heldwith a constant length until itbreaks ( Fig . 2g ) . We found that the static tether force was signi \ufb01 cantly increased in Snx33 - / - dHL - 60 cells ( from 61 . 58 to 75 . 25 pN , Fig . 2h ) , which corre - sponds to an almost 50 % increase in apparent membrane tension ( from 177 . 87 to 265 . 62 \u03bc N / m ; see Methods for details ) . Moreover , the increase in membrane tension could be rescued by stably expressing \ufb02 uorescently tagged Snx33 , excluding Snx33 - independent functions as the origin of this phenotype ( Fig . 2h ) . Notably , overexpression of Snx33 - GFP on the wild - type background did not decrease membrane tension , suggesting that a gain of function is not suf \ufb01 cient to alter the leading edge or its effects on membrane mechanics ( Supplementary Fig . 9a ) . Last , these phenotypes were not a consequence of defective differentiation , as the neutrophil differentiation marker CD11b was unperturbed in Snx33 - / - cells , nor of potentialoff - target effects caused by cell line generation by CRISPR / Cas9 technology ( Supplementary Fig . 9b , c ) . Altogether , these results con \ufb01 rm that Snx33 - / - cells have a more stable leading edge showing all the characteristics expected for persistent migration . Snx33 regulates actin polymerization Our \ufb01 ndings suggest that the curvature - sensing protein Snx33 nega - tively regulates actin polymerization and thereby limits leading edge size . Thus , we quanti \ufb01 ed the amount of \ufb01 lamentous actin ( F - actin ) by phalloidin staining upon chemoattractant ( fMLP ) addition by \ufb02 ow Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 2 cytometry and observed a signi \ufb01 cant increase in the total amounts of F - actin in Snx33 - / - cells when compared with their wild - type counter - parts ( Fig . 2i ) . Thissuggests thatSnx33 plays an inhibitoryrole onactin polymerization , particularly during the initial burst after stimulation . But how does Snx33 inhibit actin polymerization ? Some BAR domain proteins remodel actin by binding to it directly or by the recruitment of NucleationPromoting Factors ( NPFs ) that activate the Actin Related Protein 2 / 3 ( Arp2 / 3 ) complex 9 , 30 \u2013 34 . To understand how Snx33 inhibits actinpolymerization , wecoimmunoprecipitatedfull - lengthGFP - Snx33 and its bindingpartnersinthe knockout background followed by mass spectrometry . As it is common among BAR domain proteins to have intramolecular autoinhibitory interactions , we also performed coim - munoprecipitation of two truncations of Snx33 containing or missing the PX - BAR domain , but altogether spanning the full - length protein PX - BAR Membrane t = 30\u03bcs t = 0 . 0\u03bcs 10nm Mean curvature H [ \u03bcm - 1 ] Probability 0 . 02 0 . 01 0 . 04 0 . 03 0 . 05 0 . 00 0 50 - 50 0 . 06 8s 16s 0s b r i gh t f i e l d 0s 8s 16s p - po l a r i z a t i on a b c d Pearson correlation ( Snx33 , membrane ) R e l a t i v e po s i t i on z / h 0 . 8 0 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 2 0 . 4 0 . 6 0 . 0 h upper plane bottom plane S n x 33 M e m b r ane f z - 4 4 0 4 3 2 1 0 - 4 - 3 - 2 - 1 e SNX33 ARHGAP26FNBP1LSRGAP1 SH3GLB2 GAS7 ARFIP1FCHSD1 SNX18PSTPIP2ACAP2PSTPIP1DNMBP ASAP1ARHGAP44 APPL2 SNX9SNX30 ARHGAP4SH3GLB1FESFCHO2 OPHN1FNBP1 SH3BP1 PACSIN2BIN3SNX2FCHO1SNX1SH3GL1SNX6 FERARHGAP42 SNX8FCHSD2SRGAP2BACAP3 SRGAP2C SRGAP2PICK1TRIP10 SRGAP3 ARHGAP17ASAP2 BIN1 APPL1ARFIP2 ARHGAP10 SH3GL2 ACAP1 BAIAP2L2 BAIAP2L1 BAIAP2 SNX4 SNX5 MTSS1L - 8 - 7 - 6 - 5 h i j Membrane Snx33 Membrane Snx33 Snx33 Membrane k m z z lines top membrane postion in z always 0 Snx33 signal 0 . 00 0 . 05 0 . 10 on ruffle off ruffle 0 . 00 0 . 05 0 . 0 0 . 1 0 . 2 on ruffles off ruffles S n x 33 d i s t an c e i n z f r o m c e ll m e m b r ane [ no r m a li z ed ] * top membrane postion 0 . 00 1 . 00 BAR domain proteins Fold change [ log2 ] H > 0 inward H < 0 outward l t = 15\u03bcs g Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 3 ( Fig . 2j ) . Several members of the Arp2 / 3 complex as well as two sub - units of capping proteins coimmunoprecipitated with Snx33 ( limma p - value \u2266 0 . 01 , fold - change \u2265 50 % ) ( Fig . 2j ) . In particular , we identi \ufb01 ed positive fold - changes for ARPC2 , ARPC4 - TTLL3 , ACTR2 , ACTR3 , CAPZA1 and CAPZB1 in the full - length GFP - Snx33 or its truncations , suggesting that Snx33 binds to the Arp2 / 3 complex and to capping proteins . The Arp2 / 3 complex is the actin nucleator responsible for lamellipodia and ruf \ufb02 es formation at the leading edge of neutrophil cells 35 , while capping proteins terminate the growth of actin \ufb01 laments by binding to them , and potentiate their branching by the Arp2 / 3 complex 36 \u2013 39 . Interestingly , we could observe that some Arp2 / 3 com - ponents and capping protein subunits were substantially enriched with either domain compared to the full - length protein ( Supplemen - taryFig . 10a ) , suggestingthatdiversedomainsofSnx33bindthemwith different af \ufb01 nities , possiblybecause ofautoinhibitory interactionsthat mayoccludebindingsitesoraffectmembranebinding , as reportedfor other BAR domain proteins 40 \u2013 42 . In summary , Snx33 negatively reg - ulatesactinpolymerization , likely throughtheregulationof theArp2 / 3 complex and / or capping proteins . Snx33 affects WAVE2 localization and ruf \ufb02 e morphology WAVE2 isthe only knownactin NPF upstreamof the Arp2 / 3 complex in neutrophils 24 , 35 . Thus , to further characterize the leading edge of Snx33 - / - cells we imaged a component of the WAVE2 complex ( Hem - 1 ) during migration by TIRFM ( hereafter referred to as WAVE2 ) ( Fig . 3a and Supplementary Video 6 and 7 ) . WAVE2 earned its name because it localizes in waves on the basal membrane during cell migration 24 . To generate such patterns , WAVE2 waves deposit an inhibitor in their wake that transiently inhibits WAVE2 recruitment 24 . Interestingly , polymerized actin is a key component of this inhibitory feedback but what regulates WAVE2 binding to the membrane and determines its patch morphology remains poorly understood 43 . Thus , we quanti \ufb01 ed WAVE2 pattern characteristics using machine - learning - based seg - mentation . Strikingly , Snx33 - / - cells showed a 40 % increase in the WAVE2 area and the size of its patches at the plasma membrane ( Fig . 3b \u2013 d ) . To determine whether this is due to a difference in the expression levels of WAVE2 components or due to increased mem - brane binding , we performed RNAseq . We observed none or a very minor difference in expression in Snx33 - / - cells when compared with their control counterparts ( Supplementary Fig . 10b ) , suggesting that Snx33 affects the localization of the WAVE2 complex rather than its expression . This is particularly interesting as , to date , only very drastic WAVE2 patch phenotypes have been reported ( complete abrogation or increase in number ) , which lead to a severely disrupted capacity to migrate 44 , 45 . We next sought to test the functional relevance of the change inWAVE2 patch morphology for leading - edge morphology . To this end , we quanti \ufb01 ed the effective ruf \ufb02 e wavelength from SEM ima - ges and observed a signi \ufb01 cantly less tight arrangement of ruf \ufb02 es for Snx33 - / - cells when compared with their wild - type counterparts ( Fig . 3e , f , Supplementary Fig . 8j , Supplementary Fig . 11a ) . Last , to determinewhethertheincreaseineffectiveruf \ufb02 ewavelengthismerely the consequence of increase membrane tension in Snx33 - / - cells , we performed SEM on PLD2 knock - down ( KD ) cells , which also display increased membrane tension 44 . Notably , PLD2 KD cells showed no differences in effective ruf \ufb02 e wavelength compared to their Nonsense counterparts ( SupplementaryFig . 11b , c ) . These \ufb01 ndingsidentifySnx33 as a link between membrane shape and the actin polymerization fac - tors that modulate it . Snx33 enables curvature - dependent object evasion Our data thus far reveal that the membrane curvature - binding protein Snx33 regulates actin polymerization , leading - edge morphology , and stability . Leading edge patterning may facilitate adaptive motility by reducing persistence and promoting random evasive maneuvers . But the particular form of Snx33 - mediated membrane - actin coupling suggests an additional , more direct effect , in which propulsion is reduced speci \ufb01 cally where the presence of an obstacle in the path of thecellis likely , i . e . , whereexternalindentationsgenerateanincreased fraction of regions with inward curvature . Thus , Snx33 could facilitate cell steering , becausemembranedeformationsgenerated by collisions would locally down - regulate propulsion and thereby reorient the leading edge . To test this hypothesis , we \ufb01 rst positioned the cells in micro \ufb02 uidic devices 46 , 47 where they migrated through channels and encountered obstacles in the form of differently sized pores . Snx33 - / - dHL60 cells required almost 20 % more time than their wild - type counterparts to navigate these decision - points and to \ufb01 nd the path of leastresistance ( Fig . 4a , Supplementary Fig . 12a , b , and Supplementary Video 8 ) . As the nucleus has been shown to function as a mechanical guide along the path of least resistance 46 , we measured nuclear stiff - nessby atomic forceindentation and ruled out that nuclear mechanics wasaffected in Snx33 - / - cells ( Supplementary Fig . 12c , d ) , inagreement with their retained ability to read out pore size ( Supplementary Fig . 12b ) . However , in decision - free channels , Snx33 - / - cells migrated 80 % faster than theirwild - type counterparts , including during passage through a single constriction ( Fig . 4b , Supplementary Fig . 12e , and Supplementary Video 9 ) . Next , we quanti \ufb01 ed uncon \ufb01 ned migration on 2D planar substrates with a homogeneous chemoattractant , promot - ing cell motility without requiring the cells to adapt to any obstacle ( Fig . 4c , see Methods for details ) . As in decision - free channels , Snx33 - / - cells migrated faster than wild - type cells ( Fig . 4d ) . Moreover , while the motility of Snx33 - / - cells was more persistent , wild - type cells were more prone to perform large spontaneous turns ( Fig . 4e ) . Thus , the lossofSnx33appearstorendercellslesspronetospontaneousturning and less effective in circumnavigating an object while displaying more persistent migration in decision - free environments . These migratory phenotypes are consistent with the increase in leading edge size and plasma membrane tension we have observed in Snx33 - / - cells ( Fig . 2d \u2013 h ) . Together , these observations suggest that Snx33 - / - cells Fig . 1 | Curvature patterning in the lamellipodium and the curvature - sensitive protein Snx33 . a The leading edge of migrating cells is characterized by intricate curvature patterns . The arrow indicates the direction of cell movement . Lines depict a complex 3D environment . b Time - lapse bright - \ufb01 eld imaging shows migrating immune - like differentiated HL - 60 cells . c Scanning electron microscopy ( SEM ) image of a wild - type cell with zoom - in at the leading edge . n = 175 . d Time - lapse p - polarization of p - TIRFM imaging reveals dynamic membrane waves at the leading edge . e Up - regulated ( orange ) and down - regulated ( blue ) BAR domain genesbetweendifferentiated ( migratory ) andundifferentiated ( non - migratory ) HL - 60cells . f FluorescentlytaggedSnx33 ( eGFP - Snx33 ) andmembranemarker ( CAAX - mcherry ) inupper ( close - to - the - cell - top ) and lower ( near - surface plane ) z - planesin a cell . g Pearson correlation coef \ufb01 cient of \ufb02 uorescently tagged Snx33 and mem - brane marker at different heights acquired by confocal microscopy . n = 10 . Error bars denote the standard error of the mean . h Structure of the Snx33 PX - BAR domain as modeled with AlphaFold - multimer ( green ) compared with the incomplete crystal structure ( PDB ID : 4AKV ; gray ) . i Probability histogram of the mean curvature ( H ) sampled at the center of mass of PX - BAR ( green , mean curvature PX - BAR = 16 \u03bc m \u2212 1 ) and at a random lipid phosphate position ( yellow ) . Ker - nel density estimates to smooth the distributions are shown . The mean values are indicated as dashed vertical lines . j Top and side views from snapshots of the coarse - grained simulation of a buckled membrane with the PX - BAR domain . Pro - tein backbone beads ( green ) , phosphate beads ( yellow spheres ) , and lipid tails ( gray sticks ) are shown . Water and ions are omitted for clarity . A top and side view are shown . Time points are indicated for the respective frames . k , l Cross - sections ( xz ) of migrating cells using lattice - light - sheet microscopy visualizing eGFP - Snx33 and CAAX - mcherry with a zoom - in on a membrane ruf \ufb02 e ( k ) , and on a lamellipo - dium ( l ) . n = 6 . m Quanti \ufb01 cation of the average z - position of Snx33 relative to the plasma membrane on - and off - ruf \ufb02 es ( p TopMembrane = 0 . 01635 , t = \u2212 3 . 5524 , df = 5 , paired , two - sided ) . n = 6cellswhichdata wereobtained from two - color3D movies . Each point represents the average for 1 cell . Scale bars = 10 \u03bc m . p < 0 . 05 ( * ) . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 4 haveimprovedpropulsionbutlacktheabilitytoadaptwhenfacedwith an obstacle . To gain further insights into the object - evasion response of Snx33 - de \ufb01 cient cells , we devised a reductionistic assay to test contact inhibition of locomotion ( CIL ) , a common phenomenon in many cell types , including dHL - 60s , where cells stop moving in a particular direction when they come into contact with another cell or an object 24 , 48 . To test if CIL as a response to cellular obstacles is controlled by Snx33 , we seeded a higher density of dHL - 60 cells and imaged their 2D migration in the presence of cell \u2013 cell interactions by TIRFM . Snx33 - / - cells formed larger cell \u2013 cell contacts compared to wild - type cells ( Fig . 4f , g ) , in agreement with their increased decision time when faced with an inert obstacle ( Fig . 4a ) . Our computational results together with the observed localization of Snx33 ( Fig . 1h \u2013 m ) suggest preferential binding to inward - curved regions . To assess what changes in plasma membrane geometry , Fig . 2 | Snx33 knockout cellsdisplayalteredcelland leading edge morphology due to an increase in actin polymerization . a TIRFM images of a wild - type and a Snx33 - / - cell . b Cellspreadarea ( p = 1 . 505e - 7 , Mann \u2013 Whitney - U - test , two - sided ) and c eccentricity differ between wild - type and Snx33 - / - cells during movement ( p = 0 . 004989 , Mann \u2013 Whitney - U - test , two - sided ) . d Leading edge segmentationof a wild - type and a Snx33 - / - cell . Time frames ( 5s ) are color - coded . e Leading edge area ( p = 1 . 634e - 5 , Mann \u2013 Whitney - U - Test , two - sided ) and f length ( p = 0 . 2111 , Mann \u2013 Whitney - U - test , two - sided ) . n = 82 ( wt ) , n = 78 ( Snx33 - / - ) . g Schematic of statictether pulling experiments . h Meanstatic tetherforce of wt ( n = 24 ) , Snx33 - / - ( n = 26 ) and Snx33 - / - with overexpressed eGFP - Snx33 ( n = 25 ) from 3 independent experiments ( p wt vs . Snx33 - / - = 7 . 883e - 6 , t = \u2212 5 . 0144 , df = 47 . 376 ; p wt vs . Snx33 - / - + eGFP - Snx33 = 0 . 2141 , t = \u2212 1 . 2601 , df = 45 . 515 ; p Snx33 - / - vs . Snx33 - / - + eGFP - Snx33 = 0 . 2141 , t = \u2212 1 . 2601 , df = 45 . 515 , two - sided ) . i Median value of phalloidin \ufb02 uorescence intensity in \ufb01 xed wt and Snx33 - / - dHL - 60 cells quanti \ufb01 ed by \ufb02 ow cytometry . Data from 3 independent experiments ( p t00 = 0 . 184 , Mann \u2013 Whitney - U - test , two - sided ; p t01 = 0 . 0373 , t = \u2212 4 . 26 , df = 2 . 375 , two - sided ; p t30 = 0 . 1298 , t = \u2212 1 . 9728 , df = 3 . 5035 , two - sided ) . j Schematic of full - length Snx33 with its domains and two Snx33 trun - cations ( \u0394 PXBAR and PXBAR ) . Volcano plots from co - immunoprecipitation experiments comparing Snx33 - / - with Snx33 - / - + GFP - Snx33 , Snx33 - / - + GFP - Snx33 \u0394 PXBAR and Snx33 - / - + GFP - Snx33PXBAR . Enriched hits ( limma p - value \u2266 0 . 01 , fold - change \u2267 50 % ) are shown green , and the rest gray . Scale bars = 10 \u03bc m . p < 0 . 001 ( * * * ) , p < 0 . 01 ( * * ) , p < 0 . 05 ( * ) . Box - plots : lower and upper hinges corre - spondtothe25thand75thpercentile . Theupperwhiskerextendsfromthehingeto the largest value , but no further than 1 . 5 * IQR . The lower whisker extends from the hingetothesmallestvalue , butnolowerthan1 . 5 * IQRofthehinge . Databeyondthe whiskers : black dots . Black line : median . Black dot : mean . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 5 occur upon collisions between cells , we quanti \ufb01 ed the curvature differences at the leading edge of freely moving cells and of pairs after collision . We performed serial block - face scanning electron microscopy ( SBEM ) , manually segmented the plasma membrane , and quanti \ufb01 ed the curvatures at free and contacting areas in the front of colliding and single cells in the xz plane ( Fig . 4h \u2013 k ) . Addi - tionally , to rule out potential \ufb01 xation artefacts , we analyzed leading edges in freely migrating and colliding live cells using a membrane marker ( CAAX ) and lattice light sheet ( LLS ) imaging in the xy plane ( Fig . 4l \u2013 n and Supplementary Fig . 13f \u2013 i ) . In both datasets , we observed a change in the curvature distribution at the front between single and collided cells . While single cells displayed a broader distribution biased towards outward curvatures , the con - tact surfaces were characterized by narrower distributions centered at zero , in line with an overall \ufb02 attening at the contact site . Cell \u2013 cell contacts suppressed both positive and negative curvatures ( Sup - plementary Fig . 14a \u2013 d ) . To further dissect the molecular role of Snx33 in CIL , we simulta - neously imaged the neutrophil WAVE2 complex ( using Hem1 ) and Snx33 at the leading edge of live cells . While both proteins largely colocalized in most parts of the cell , they were anticorrelated in the highly negatively curved rim of the leading edge ( Supplementary Fig . 15a , b ) . Here , Snx33 levels decreased while WAVE2 accumulated ( Supplementary Fig . 15c , d ) , consistent with Snx33 being excluded from outward - curved regions and restricting WAVE - binding to such areas . Moreover , when cells collided , Snx33 localized to the cell contact area , whereupon WAVE2 disappeared and relocated to contact - free zones of the plasma membrane ( Fig . 4o , p ) . Subsequently , thecellrepolarizedby forming a new leading edge ( Fig . 4o , p ) . To assess whether the dysfunctional CIL response in Snx33 - / - cells is due to impaired WAVE2 inhibition at the contact area , we followed WAVE2 localization before and after collision in Snx33 - / - cells ( Supplementary Videos 10 ) . In con - trast to wild - type cells , Snx33 - / - cells indeed failed to remove WAVE2 from the contact site ( Fig . 4q and Supplementary Fig . 15e ) . Altogether , weshowakeyroleforthecurvature - sensingproteinSnx33inregulating actin polymerization during CIL . Speci \ufb01 cally , Snx33 displaces WAVE from cell \u2013 cell contact areas and thus leads to cell repolarization towards a contact - free zone upon collision with an obstacle . Discussion Object evasion is key not only for the migration of immune and cancer cells in complex tissue environments , but also fundamental during embryogenesis and collective migration in vivo 49 . Combin - ing genetic perturbations , microscopy , and micro \ufb02 uidics we iden - ti \ufb01 ed a curvature - sensing mechanism underlying adaptive motility in immune - like cells . We show that the BAR domain protein Snx33 reads out changes in the membrane curvature landscape and feeds back to the actin polymerization machinery to regulate complex patterns at the leading edge of migrating cells . Speci \ufb01 cally , Snx33 inhibits actin polymerization and tunes the localization of WAVE2 , the main nucleation - promoting factor in neutrophils . Thereby , Snx33 limits the persistence of the leading edge and reduces migration directionality . This mechanism for promoting sponta - neous reorientation during cell migration complements the one described for abundantly expressed I - BAR domain proteins , where outward curvature - sensing coupled to local activation of actin polymerization enhances exploratory cell behaviors 22 . While few other BAR domain proteins contain domains with inhibitory effects Fig . 3 | WAVE2 pattern width and ruf \ufb02 e wavelength are increased in Snx33 knockout cells . a TIRFM images of a wild - type and Snx33 - / - cell show the dis - tribution of WAVE2 , the only actin nucleator promoting factor upstream of the Arp2 / 3 complex in neutrophils . b The total area occupied by WAVE2 patches in the leading edge increases upon loss of Snx33 ( p = 0 . 000408 , Mann \u2013 Whitney - U - test , two - sided ) . n = 82 ( wt ) , n = 78 ( Snx33 - / - ) . c Segmentation of dynamic WAVE2 patches in a wild - type and Snx33 - / - cell . Time frames ( 5s ) are color - coded . d The size of WAVE2 patches increases upon loss of Snx33 ( p = 0 . 000464 , Mann \u2013 Whitney - U - test , two - sided ) . n = 82 ( wt ) , n = 78 ( Snx33 - / - ) . e SEM images withruf \ufb02 esegmentations ( red ) showtheleadingedgesofawild - typeandSnx33 - / - cells fromthe50 th percentileofthe distribution . f Theeffectiveruf \ufb02 e wavelength increases upon loss of Snx33 ( p = 3 . 171e - 7 , Mann \u2013 Whitney - U - test , two - sided ) . n = 175 ( wt ) , n = 170 ( Snx33 - / - ) . Statistics : t - testornon - parametricMann \u2013 Whitney - U - test . Scale bars = 10 \u03bc m . p < 0 . 001 ( * * * ) , p < 0 . 01 ( * * ) , p < 0 . 05 ( * ) . Box - plots : the lower and upper hinges correspond to the 25th and 75th percentile . The upper whisker extends from the hinge to the largest value , but no further than 1 . 5 * IQR . Thelowerwhiskerextendsfromthehingetothesmallestvalue , butnolowerthan 1 . 5 * IQR of the hinge . Data beyond the whiskers : black dots . Black line : median . Black dot : mean . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 6 on actin polymerization , they are likely not fully redundant . We envision that their mechanism of action will be the result of several properties , including curvature sensitivity , strength of binding , binding partners , and what other domains are present . Thus , BAR domain proteins emerge as versatile tools that locally activate or inhibit actin polymerization and thus endow cells with functional control of their leading edge . The ability to spontaneously reorient and explore the micro - environment with dynamic protrusions in and of itself contributes to more ef \ufb01 cient navigation in crowded contexts 6 , 22 , 44 , 46 , 50 , 51 . A curvature - dependent downregulation of protrusive activity can furthermore aid object evasion more directly , by steering the cell away from the direction in which the presence of an obstacle on the outside is most likely . Indeed , we show that Snx33 is key for cells to circumnavigate both inert and cellular obstacles and thus migrate through complex three - dimensional environments . To date , little is known about the molecular machinery orchestrating CIL . srGAP2 , a BAR domain protein that binds outward membrane curvature , was Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 7 shown to induce protrusions during CIL 52 \u2013 55 . Our study shows a key role for the inward curvature - sensing protein Snx33 in inhibiting actin polymerization during CIL . Speci \ufb01 cally , we show that the multi - domain protein Snx33 localizes to cell \u2013 cell contacts and redistributes WAVE2 to the free surfaces , thereby reorienting the cell . This likely occurs in a curvature - dependent manner as Snx33 preferentially localizes to inward - curved membrane regions while WAVE2 is restricted to outward - curved areas at the leading edge 23 . At cell \u2013 cell contacts we identify a distinct shift in the curvature distribution using two datasets that differ in measured curvature values by an order of magnitude . Speci \ufb01 cally , upon collision , the curvature distributions are narrower compared to that of the free leading edge of migrating single cells , consistent with membrane \ufb02 attening upon collision . The expected curvature sensitivity of an Snx33 PX - BAR dimer corresponds to a higher curvature than that measured at cell \u2013 cell collisions . However , it is important to note that BAR domain proteins have auxiliary domains and can form oligomers as well as linear aggregates that bind membranes with considerably lower curvature 56 \u2013 58 . Furthermore , our observations in cells are restricted to radii of curvature with magnitudes above the respective image resolution limits and limited by detectable cur - vature length scales . Thus , in the future , it will be important to explore how forming assemblies as well as the presence / absence of auxiliary domains may permit BAR domain proteins to span larger radii of curvatures to respond to larger - scale changes in membrane shape within the molecular complexity of cells 19 , 59 . Here , we show that regulation of actin polymerization by Snx33 has important consequences for cell migration . This , in conjunction with the inherent curvature - sensing properties of BAR domain pro - teins and the observed shift in membrane curvature distribution upon cell \u2013 cellcollisionduringCILsuggestamechanismbywhichmembrane curvature changes direct adaptive cellular behaviors . Together , our study supports the notion that cells use their membrane topography to encode information about the external environment they encounter . Given the diversity of BAR domain proteins present in cells , we expect that this regulatory principle is likely to be used in many bio - logical functions that have to react to shape changes , from the sub - cellular level ( organelle homeostasis , membrane traf \ufb01 cking ) , via the single - cell level ( immune surveillance , tumor dissemination ) , to the multicellular level ( gastrulation , tissue folding ) . Methods Cell culture HL - 60 cells were grown in RPMI 1640 media with 10 % heat - inactivated FBS ( # 10500 - 064 , Gibco ) and 1 % Penicillin - Streptomycin ( # 15140 - 122 , Gibco ) in a humidi \ufb01 ed incubator at 37\u00b0C with 5 % CO 2 . Cells were differentiated by adding 1 . 5 % DMSO ( # D2438 , Sigma Aldrich ) and used after 5 days . Each independently differentiated batch was treated as a biological replicate . For starvation , cells were kept for 1h in FBS - free RPMI 1640 media with 0 . 3 % fatty acid - free BSA ( # A7030 - 10G , Sigma Aldrich ) . For imaging or \ufb01 xation , dHL - 60 cells were plated on \ufb01 bronectin - coated ( 0 . 01 mg / ml , # 356008 , Corning ) glass - bottom dishes ( # 627860 , Greiner bio - one ) and allowed to adhere for 10min in growth media . Next , cells were washed and stimulated with 10nM fMLP ( # F3506 - 5MG , Sigma Aldrich ) . For lowering adhesion , the coat - ing was supplemented with 5 % molar BSA ( # A7030 - 10G , Sigma Aldrich ) . To generate stable cell lines with \ufb02 uorescently tagged Snx33 and its truncations ( PXBAR and \u0394 PXBAR ) , as well as Hem1 and CAAX lentiviral transduction was used as described previously 44 . Cells were sorted on a BD FACS Aria \u2122 Fusion at the EMBL \ufb02 ow cytometry core facility . Generation of knockout cell line by CRISPR / Cas9 CRISPR / Cas9 generation in HL - 60 cells was performed as described previously 45 . Brie \ufb02 y , cloning of the target guide sequence to target Snx33 was performed 60 , 61 ( Forward : CACCGctgggacgacGGATGC ACAG ; Reverse : aaacCTGTGCATCCgtcgtcccagC ) . Cells expressing BFP - tagged Cas9 were single - cell sorted in 96 - well plates on a BD FACS Aria TM Fusion at the EMBL \ufb02 ow cytometry core facility . Single - cell clones were veri \ufb01 ed by genomic DNA ampli \ufb01 cation by Touch - down PCR 62 and sequencing , followed by a Western blot of selected clonal lines . RNA sequencing Total RNA samples obtained from 3 biological replicates were pur - i \ufb01 ed using a RNeasy Mini Kit ( # 74104 , Qiagen ) according to the manufacturer \u2019 s instructions with a DNase digestion step ( # 79254 , Qiagen ) . To ensure high quality , samples were analyzed on an Agi - lent 2100 Bioanalyzer ( Agilent Technologies ) . RNA sequencing was performed on an Illumina NextSeq 500 platform as NextSeqHigh - 75 SE at the EMBL genomics core facility . For sequence alignment , the hg19 reference genome was used . Differential expression ana - lysis was performed with a custom - made Galaxy pipeline using a DESeq2 package . The RNAseq data have been deposited to the ArrayExpress collections from BioStudies with the accession num - ber E - MTAB - 12436 . Imaging TIRFM images of live cells were acquired on a Nikon TiEclipse inverted microscopewithaCFIPlanApoLambda100x Oil ( # MRD01905 , Nikon ) silicone objective and a sCMOS camera controlled by NIS - Elements ( Nikon ) . Sample drift was reduced using an autofocus system ( Perfect Focus , Nikon ) for time - lapse imaging . Fig . 4 | Snx33 steers cell movement in single - cell 2D and 3D migration by inhibiting the WAVE2 complex . a Decision channel passage time ( n wt = 159 , mean = 1 . 67min ; n Snx33 - / - = 9 ; mean = 1 . 96min ) in cells ( p = 0 . 02 , Mann \u2013 Whitney - U - t est , two - sided ) . b Constriction passage time ( n wt = 234 , n Snx33 - / - = 158 ) in cells ( p = 2 . 2e - 16 , Mann \u2013 Whitney - U - test , two - sided ) . c Cellbody displacement over time in wt and Snx33 - / - cells . Time frames ( 5s ) are color - coded . d Cell speed in wt and Snx33 - / - cells ( p = 7 . 661e - 5 , Mann \u2013 Whitney - U - test , two - sided ) e Distribution of anglesatwhichcellsturnduringmigration ( p = 0 . 0007583 , Mann \u2013 Whitney - U - Test , two - sided ) . N wt = 82 , n Snx33 - / - = 78 . Data from 3 independent biological replicates . f Segmentation of a contact event between two wt or Snx33 - / - dHL - 60 cells . Shade highlights the contact duration . g Percentage ofcellcircumference in contact with another cell in wt ( n = 18 ) and Snx33 - / - ( n = 20 ) dHL - 60 cells ( p = 0 . 0001704 , Mann \u2013 Whitney - U - test , two - sided ) . h Visualization of curvature magnitudes mea - sured using various techniques . i Cross - section of a single and collided migrating dHL - 60 cell by SBEMimaging . N single = 6 , n collided = 6 . j Visualization of the absolute curvature value in the leading edge of single and collided migrating cell ( from i ) . k Histogram showing the distribution of absolute curvature values in the leading edge of single and collided cells ( n single = 6 , n collided = 6 ) . Data points denote an average value per cell . l xy plane of a membrane marker ( CAAX ) of a single and collided migrating dHL - 60 cells by lattice - light - sheet . m Visualization of the abso - lute curvature value in the leading edge of single and collided migrating cell ( from l ) . n Histogram showing absolute curvature in the leading edge of single and col - lidedcellsfromlatticelightsheetimages ( n single = 5 , n collided = 5 ) . Datapointsdenote an average per cell . o Bright - \ufb01 eld and p TIRFM imaging of cell \u2013 cell contact in wt dHL - 60 cells with \ufb02 uorescently tagged Snx33 and Hem1 . Arrow points towards cell \u2013 cellcontact . n = 3 . q WAVE2fold - changeaftercell \u2013 cellcontactinwt ( n = 9 ) and Snx33 - / - ( n = 10 ) dHL - 60cells ( p = 0 . 01816 , t = \u2212 2 . 5877 , df = 18 . 789 , two - sided ) . Scale bars = 10 \u03bc m . p < 0 . 001 ( * * * ) , p < 0 . 01 ( * * ) , p < 0 . 05 ( * ) . Box - plots : the lower and upper hinges correspond to the 25th and 75th percentile . The upper whisker extends from the hinge to the largest value , but no further than1 . 5 * IQR . The lower whisker extends from the hinge to the smallest value , but no lower than 1 . 5 * IQR of the hinge . Data beyond the whiskers : black dots . Black line : median . Black dot : mean . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 8 Confocal images of \ufb01 xed cells were obtained with a UPLSAPO 60X S ( NA 1 . 3 ; WD 0 . 3 mm ) silicone objective on an Olympus FV3000 inverted microscope at the EMBL advanced light micro - scopy facility . Epi \ufb02 uorescent and bright - \ufb01 eld imaging of \ufb01 xed cells was per - formed using a 40x objective ( # MRD00405 , Nikon ) , a SOLA SE II , and 100W halogen lamps ( Nikon ) using appropriate \ufb01 lter sets . Polarized TIRFM ( pTIRFM ) modality was implemented based on previous work 63 \u2013 68 . For imaging , dHL - 60 cells were stained before plating with carbocyanine dye DiI ( # D3911 , ThermoFisher Scienti \ufb01 c ) . Lattice light sheet imaging of live cells was performed on a Zeiss Lattice Light sheet 7 ( Zeiss , Oberkochen , Germany ) using appropriate \ufb01 lter sets and a 10\u00d7550 \u03bc m base beam . Image analysis For confocal images ( Fig . 1g and Supplementary Fig . 15b ) , only the z - planes that contained the top 80 % intensity of mCherry - CAAX were considered based on line scans covering the entire resliced maximum intensity z projection . A channel of interest ( ChoF1 ) was used for mask generation based on automatic Otsu segmentation . A custom - made ImageJ script allowed us to calculate the Pearson correlation coef \ufb01 - cient ( PCC ) forevery z - planeofChoF1withChoF2basedonthemaskof ChoF1usingthein - builtColoc2ImageJplugin . Z - sliceswereassignedto 10binsandthemeanwithstandarderrorofthemeanforevery binwas calculated . For analysis of migrating cells imaged by TIRFM ( Fig . 2b \u2013 f , Fig . 3b \u2013 d , and Fig . 4c \u2013 e ) a segmentation of the cell mask ( based on mCherry - CAAX signal ) and of the WAVE2 mask ( based on eGFP - Hem1 signal ) were acquired using the machine - learning - based ilastik soft - ware ( Versions 1 . 3 . 1b1 and 1 . 3 . 3post3 ) 27 . Further image analysis was achieved using an in - house built program implemented in Python . The angle at which cells are moving was calculated based on the center of mass for three consecutive frames . The leading edge was de \ufb01 ned as the difference between two consecutive frames where at least one pixel of WAVE2 mask is present per cluster . The leading edge length was de \ufb01 ned as the number of pixels in the outside perimeter of the leading edge . For the analysis of cell \u2013 cell contacts the same segmen - tation strategy was used to segment individual cells . Cell \u2013 cell contact was de \ufb01 ned as the perimeters \u2019 intersection of both cells . For membrane topography analysis of SEM data ( Fig . 3e , f , Sup - plementary Fig . 8j , Supplementary Fig . 11a \u2013 c ) , the leading edge area with ruf \ufb02 es was manually segmented on median \ufb01 ltered images . Next , ridges were detected within the segmented regions using the Meijer - ing \ufb01 lter 69 . Ridges were later segmented using automatic Otsu thresholding and skeletonized using a custom Python script . Inversion of the number of pixels within the skeletonization per leading edge area corresponds to the effective ruf \ufb02 e wavelength . For analysis of membrane curvature of SBEM and lattice light sheet images ( Fig . 4i \u2013 n and Supplementary Fig . 13b \u2013 e , g , i ) , the plasma membrane was manually segmented at the leading edge ( de \ufb01 ned as the area in front of the nucleus ) or segmented using Otsuthreshold on the \ufb02 uorescence channel with membrane signal followed by a manual curation , respectively . Next , the curvature was quanti \ufb01 ed by a circle \ufb01 tting to a window with a user - de \ufb01 ned length ( 2 times the speci \ufb01 ed half - width + 1 pixel ) sliding along the segmented membrane , using a custom Python script . For determining the sign of the curvature , \ufb01 ducialsweremanuallyaddedtotheimagestopointtothecellinterior ( Supplementary Fig . 14a \u2013 d ) . For analysis of protein position in relation to membrane based on lattice light sheet images ( Fig . 1m and Supplementary Fig . 5 ) only cells with visible membrane ruf \ufb02 es in MIP wereincluded in the analysis . The ruf \ufb02 es were segmented in the MIP by Mejiering \ufb01 ltering followed by Otsu thresholding . For all pixels in the user - de \ufb01 ned ROI z - lines of the membraneandSnx33channelswereextractedandnormalizedtotheir peakintensity . To account for changes incell thicknessat different x \u2013 y positions the z - lines were normalized by localizing the peaks in the membrane channel corresponding to the top and bottom cell mem - branes ( all z - lines with more than 2 peaks detected are excluded from further analysis ) . The normalized z - lines were then separated into two halves corresponding to the bottom and top cell membrane , each of the halves was then assessed for Snx33 colocalization by calculating the overlap between the Snx33 and membrane channels . All z - line halves showing low overlap are excluded from further analysis . Next , all z - lines corresponding to one cell were pooled and separated into the in - ruf \ufb02 e and off - ruf \ufb02 e groups as well as the top - and bottom - membrane subgroups . To allow unbiased comparison between the in - ruf \ufb02 e and off - ruf \ufb02 e groups the distributionof z - linesin eachgroup was adjusted to containthe same frequencyof cellthicknesses . In practice , as the off - ruf \ufb02 e group contains more z - lines than the in - ruf \ufb02 e group the off - ruf \ufb02 e group was randomly subsampled to reconstitute the distribution in the in - ruf \ufb02 e group . Finally , as the SNR of single z - lines does not allow for accurate assessment of Snx33 to membrane dis - tance we employ averaging by bootstrap where random 100 z - lines from each group were averaged and renormalized to their peak intensity . The distance was then measured between the positions where the signal \ufb01 rst reaches 20 % in the membrane and Snx33 chan - nels . This process was then repeated multiple times to generate a distribution of distances that captures the heterogeneity of raw data . For each cell , the mean value for each of the 4 subgroups was extracted and they were compared to validate the robustness of the approach . Cell migration assays in PDMS - based devices PDMS - based micro \ufb02 uidic devices were prepared as previously described 46 , 70 , 71 . Thedevicesused forthemigrationofdHL - 60cellshad heights of 2 . 8 and 3 . 13 \u03bc m for channels with decision points and channels with constriction , respectively . The decision channels had constrictions of 2 , 3 , 4 , and 5 \u03bc m in two arrangements . The channels with single constrictions were 2 \u03bc m . To visualize nuclei and cell body , Hoechst 33342 ( # 62249 , Thermo Fisher Scienti \ufb01 c ) and TAMRA ( Invi - trogen ) were added before the introduction of cells into the PDMS device . Cell migration towards chemoattractant ( fMLP ) was imaged on an inverted wide - \ufb01 eld Nikon Eclipse microscope using a 20x / 0 . 5 PH1 airobjective , equipped with aLumencorlight source ( 390nm , 475 nm , 542 / 575 nm ) , an incubation chamber and a heated stage with CO 2 . The acquired data were analyzed using ImageJ and manually curated . Only single cells that moved through the entire channel were con - sideredforanalysis . Allparameterswerequanti \ufb01 edbasedonthenuclei signal . Tether extrusion using atomic force spectroscopy Apparent membrane tension was measured by extruding plasma membrane tethers . For measurements , a Olympus BioLever ( k = 60 pN / nm ) from Bruker was mounted on a CellHesion 200 AFM ( Bruker ) with JPK SPM Software 6 . 1 . 183 , which is integrated into an Eclipse Ti inverted light microscope ( Nikon ) . Cantilevers were calibrated using the thermal noise method and coated with 2 . 5mg / ml Concanavalin A ( # C5275 , Sigma Aldrich ) . Prior to the measurements , cantilevers were rinsed in dPBS . For tether measurement , the canti - lever was positioned over the cell , preferably over the leading edge . Measurements parameters for static tether pulling experiments were as follows : approach velocity was set to 1 \u03bc m / s , contact force to 100 \u2013 300 pN , contact time to 5 \u2013 10s , and retraction speed to 10 \u03bc m / s . After a 10 \u03bc m tether was pulled , the cantilever position was held con - stant until it broke , but no longer than 30s . In every experimental repetition , the conditions \u2019 order was randomized . For every cell at least 3 different tether measurements were taken . The data analysis was performed using the JPK Data Processing Software 6 . 1 . 183 . For assessing the magnitude of membrane tension based on tether force measurements , the following formula was Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 9 used 44 : T = F 20 8 B \u03c0 2 \u00f0 1 \u00de where F 0 is the tether force measured by the AFM and B is the bending rigidity of the plasma membrane , which we assume to be invariable between different experimental conditions ( 2 . 7\u00d7 10 \u2212 19 Nm based on previous measurements 72 , 73 . F - actin staining of non - adherent dHL - 60 cells upon fMLP stimulation Cells were starved and \ufb01 xed by adding an equal volume of 2x \ufb01 xation buffer to 10 6 cells in growth media or stimulated with 10 nM fMLP and \ufb01 xed 0 - , 1 - , and 30 - min post - stimulation . Fixation buffer ( 1x ) contains 3 . 7 % paraformaldehyde ( # 28908 , Thermo Scienti \ufb01 c ) , 1x intracellular buffer ( 140mM KCL , 1mM MgCl 2 , 2 mM EGTA , 20mM HEPES , pH 7 . 5 ) , 320 mM sucrose ( # S0389 - 500G , Sigma Aldrich ) and 0 . 2 % BSA ( # A7030 - 10G , SigmaAldrich ) . Next , cellswere washed carefully with1x intracellular buffer by plate centrifugation and stained with phalloidin coupled with TRITC ( # P1951 , Sigma Aldrich ) for 1 h in intracellular buffer ( 1x ) containing 0 . 2 % of Triton X - 100 ( # T8787 , Sigma Aldrich ) . Cellswerethen again washed with intracellularbuffer ( 1x ) . Finally , they were re - suspended in 1 ml of 0 . 1 % BSA , 2 . 5mM EDTA in dPBS , and analyzed using a Cytek\u00ae Aurora ( Cytek ) at the EMBL Flow Cytometry Core Facility . Data were further analyzed and plotted using the FlowJo software ( Version 10 . 9 . 0 ) . Co - immunoprecipitation For coimmunoprecipittion , GFP - tagged Snx33 and its truncation were added to the Snx33 - / - cell line by viral transduction followed by cell sorting . Harvesting and lysis of dHL - 60 cells was performed as recom - mended for cytoplasmic proteins following the ChromoTek GFP - Trap Magentic Particles M - 270 with protease inhibitor supplementation ( # gtd , Chromotek ) . GFP - nanotrap beads were used to precipitate GFP - tagged proteins from the lysate ( full - length Snx33 and two truncations : PXBAR and \u0394 PXBAR ) afterovernightrotation . Elutionwasperformedin 2xSDS - sample buffer and submitted for mass spectrometry . Protein mass spectrometry Sample preparation . Reduction and alkylation were performed with dithiothreitol ( 56 \u00b0C , 30min , 10 mM in 50 mM HEPES , pH 8 . 5 ) and 2 - chloroacetamide ( room temperature , in the dark , 30min , 20mM in 50 mM HEPES , pH 8 . 5 ) . Samples were prepared according to the SP3 protocol ( https : / / doi . org / 10 . 15252 / msb . 20145625 , https : / / doi . org / 10 . 1038 / s41596 - 018 - 0082 - x ) . In short , sequencing - grade trypsin ( Pro - mega ) was added in an enzyme - to - protein ratio of 1 : 50 for overnight digestion at 37\u00b0C . Peptide recovery was performed in 50 mM HEPES , pH 8 . 5 by collecting the supernatant on the magnet and combining it with a second elution . Peptides were labeled with TMT16plex Isobaric Label Reagent ( ThermoFisher ) according to the manufacturer \u2019 s instructions . Inshort , 0 . 8 mg reagent was dissolved in 42 \u00b5 l acetonitrile ( 100 % ) and 8 \u00b5 l of stock was added and incubated for 1 h at room temperature . The reaction was quenched with 5 % hydroxylamine for 15min at RT . Sam - ples were combined and cleaned up with an OASIS\u00ae HLB \u00b5 Elution Plate ( Waters ) . Mass spectrometry analysis . An UltiMate 3000 RSLC nano - LC system ( Dionex ) \ufb01 tted with a trapping cartridge ( \u00b5 - Precolumn C18 PepMap 100 , 5 \u00b5 m , 300 \u00b5 m i . d . x 5 mm , 100 \u00c5 ) and an analytical column ( nanoEase \u2122 M / Z HSS T3 column 75 \u00b5 m x 250 mm C18 , 1 . 8 \u00b5 m , 100\u00c5 , Waters ) was coupled to an Orbitrap Fusion \u2122 Lumos \u2122 Tribrid \u2122 Mass Spectrometer ( Thermo ) using the Nanospray Flex \u2122 ion source in positive ion mode . Peptides were concentrated with a constant \ufb02 ow rateof30 \u00b5 l / min ( 0 . 05 % tri \ufb02 uoroaceticacidinwater ) ontothetrapping column for 4 min . Subsequently , peptides were eluted via the analy - tical column running using solvent A ( 0 . 1 % formic acid in water , 3 % DMSO ) with a constant \ufb02 ow of 0 . 3 \u00b5 l / min and with an increasing per - centage of solvent B ( 0 . 1 % formic acid in acetonitrile , 3 % DMSO ) from 2 % to 8 % in 4 min , from 8 % to 28 % for a further 104 min , in another 4 min . from 28 % to 40 % , and \ufb01 nally 40 % \u2013 80 % for 4 min followed by re - equilibration back to 2 % B in 4 min . MS instrument parameters were as follows : spray voltage of 2 . 4 kV , capillary temperature 275 \u00b0C , MS1 mass range 375 \u2013 1500 m / z , pro \ufb01 le mode , in the orbitrap with a resolution of 120000 . The max - imum \ufb01 ll time 50 ms , with an AGC target set to standard . Data - dependentacquisition ( DDA ) wasperformed withtheresolutionofthe Orbitrap set to 30000 , with a \ufb01 ll time of 94ms and a limitation of 1\u00d710 5 ions . A normalized collision energy of 34 was applied . MS2 data was acquired in pro \ufb01 le mode . Fixed \ufb01 rst mass at 110 m / z . Mass spectrometry data analysis \u2014 Isobarquant . IsobarQuant 74 and Mascot ( v2 . 2 . 07 ) were utilized to process the acquired data . Data was searched against the Homo sapiens proteome database ( UP000005640 ) containing common contaminants and reversed sequences . The following modi \ufb01 cations were included in the search parameters : Carbamidomethyl ( C ) andTMT10 ( K ) ( \ufb01 xedmodi \ufb01 cation ) , Acetyl ( Protein N - term ) , Oxidation ( M ) and TMT10 ( N - term ) ( variable modi \ufb01 cations ) . Forthefullscan ( MS1 ) amasserrortoleranceof10ppm and for MS / MS ( MS2 ) spectra of 0 . 02 Da was set . Further parameters were set : Trypsin as protease with an allowance of a maximum of two missed cleavages : a minimum peptide length of seven amino acids ; at leasttwo unique peptides were required for protein identi \ufb01 cation . The false discovery rate on peptide and protein levels was set to 0 . 01 . Raw output \ufb01 les of IsobarQuant ( protein . txt \u2013 \ufb01 les ) were analyzed using R programming language ( ISBN 3 - 900051 - 07 - 0 ) . Only proteins with at least two unique peptides were included in the analysis and quanti \ufb01 cation . In total , 561 proteins passed the quality control \ufb01 lters . Raw signal - sums ( signal _ sum columns ) were cleaned for batch effects using limma ( PMID : 25605792 ) and later normalized using vsn ( var - iance stabilization normalization - PMID : 12169536 ) . To test proteins for differential enrichment limma package was employed . The replicate information was appended as a factor in the design matrix given as an argument to the \u2018 lmFit \u2019 function of limma . A hit was de \ufb01 ned as a pro - tein annotated with a false discovery rate ( fdr ) smaller 5 % and a fold - change of at least 100 % and as a candidate with a fdr below 20 % and a fold - change of at least 50 % . The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE 75 partner repository with the dataset identi \ufb01 er PXD033666 . MD simulations Coarse - grained molecular dynamics simulations . Molecular dynamicssimulationswereperformedusingGROMACS2021 . 4 76 , using the coarse - grained Martini2 . 2 force \ufb01 eld 77 , 78 and applying an estab - lished scaling procedure ( alpha = 0 . 7 ) 79 to all protein beads . As a basis for the structural model , an existing crystal structure ( PDB ID : 4AKV ) of the Snx33 membrane - binding domain ( PX - BAR ) was used . Since someloops were missing inthis structure we predicted the same sequence using AlphaFold - multimer 80 , 81 with default settings but increasing the number of recycles to 6 . The resulting model was in excellent agreement ( RMSD : 2 . 3 \u00c5 ) with the crystal structure and was directly used for simulations . The structuralmodel wascoarse - grained using the \u2018 martinize . py \u2019 script 78 applying secondary structure restraints assigned by DSSP and using an elastic network 82 across both subunits with f c = 500 kJ / mol \u2212 2 and a cut - off c = 1 . 2 nm as previously reported for simulations of other extended BAR proteins 19 . For all disordered coil regions , all elastic bonds were removed . The proteinwas then placed ontoa buckled membranewhich was generated from compression of a \ufb02 at membrane according to the Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 10 following procedure . A \ufb02 at symmetric bilayer with a size of 70x 35x 20nm 3 was prepared using the \u2018 insane . py \u2019 tool 83 . A plasma membrane - derived lipid composition determined by mass spectro - metry was used as input ( Table S1 ) . The membrane was solvated and Na + - ions were added to make the system charge neutral . The mem - brane was then energy minimized using the steepest decent algorithm for 1000 steps . Subsequently , the bilayer was simulated in an NPT ensemble ( semi - isotropic pressure coupling ) for 50 ns with a timestep of 20fs using the Berendsen barostat 84 and velocity rescaling thermostat 85 . Coupling times of 1 ps and 4ps were used , respectively . The tem - perature wasset to 310K in all simulations . The Verlet neighbor search algorithm was used to update the neighbor list , with the length and update frequency being automatically determined . Lennard - Jones and Coulomb forces were cutoff at r c = 1 . 1 nm with the potential shifted to 0 using the Verlet - shift potential modi \ufb01 er 86 . Then , lateral pressure was set to 10bar to generate buckled membrane structures ( P x , y = 10bar , P z = 1 bar ) . Two membranes with different degrees of curvature were extracted from the trajectory of compression with a box size of 58 . 9\u00d7 29 . 5\u00d725 . 9 nm 3 ( excess mem - brane area = 25 . 1 nm 2 ) and 56 . 3 \u00d7 28 . 2 \u00d728 . 3 nm 3 ( excess membrane area = 31 . 4nm 2 ) . In both systems , the protein was placed above the buckled membrane and re - solvated and charge - neutralized . A brief equilibra - tion was performed using the above settings and maintaining position restraints on the protein with a f c = 1000 kJmol \u2212 1 nm \u2212 2 . The equilibra - tion was run for 250 ns . Production simulations were then carried out using anisotropic pressure coupling with \ufb01 xed x , y dimensions and P z = 1 bar . Production simulations were carried out for 30 \u00b5 s . The same simulation parameters were used as above except the pressure was controlled using instead the Parrinello - Rahman barostat with a cou - pling time of 20ps 87 . Images and visualizations were made with VMD ( Version 1 . 9 . 4 ) 88 . Analyses were carried out with MDAnalysis 1 . 1 . 1 89 , 90 and python 3 . 6 . To analyze the curvature preference of the protein on the buckled membrane a previously established protocol by Bhaskara et al . 17 was used ( see : https : / / github . com / bio - phys / MemCurv ) . A 2D Fourier expansion of a height function h ( x , y ) was \ufb01 t to the phosphate beads of the membrane every 1 ns by optimizing a least - squared \ufb01 t . Subse - quently , the meancurvature H wasderived from the shape operatorof the \ufb01 t height function h ( x , y ) . The mean curvature H was calculated for the x , y position of the center of mass of the protein as well as a ran - domly selected phosphate bead position to sample the background membrane at every considered frame . Scanning electron microscopy ( SEM ) After 30min of 10nM fMLP stimulation , cells were \ufb01 xed in 2 . 5 % GA ( # 16220 , EMS ) in 0 , 1M PHEM buffer by adding 37\u00b0C double strength \ufb01 xative ( 5 % GA in 0 , 1 M PHEM ) directly 1 : 1 to the cell medium . After 10min incubation , the \ufb01 xative was replaced by fresh single - strength \ufb01 xative , and cells were further \ufb01 xed at room temperature for 1h . After \ufb01 xation , cells were washed 2 times in 0 , 1 M PHEM and 2 times in 0 , 1 M cacodylate buffer . Next , they were post \ufb01 xed for 2 h on ice in freshly prepared and \ufb01 ltered 1 % OsO 4 ( # 19190 , EMS ) and 0 . 8 % potassium ferrocyanide ( K 4 [ Fe ( CN ) 6 ] * 3H 2 O , # 4984 , Merck ) in 0 . 1 M cacodylate buffer . Afterpost \ufb01 xation , thecellswerewashed4timesinH 2 O , andleft at 4 \u00b0C until further processing . Next , cells were treated with freshly prepared and \ufb01 ltered 1 % tannic acid ( TA , CAS # 1401 - 55 - 4 , EMS ) in water using a Pelco Bio - Wave microwave for seven 1 - min cycles alternating between 150 W and 0 W power . Steady temperature was set to 23 \u00b0C and the vacuum to on for all steps . After TA treatment , cells were washed 2x in H 2 O on the bench and 2x in H 2 O in the microwave for 40 s per step at 250 W power . Cells were then treated with 1 % UA ( # 77870 , Serva ) in H 2 O using the same microwave program as for TA . After washing once in H 2 O and twice in 25 % EtOH , cells were dehydrated in a graded series of ethanol ( 25 % \u2013 50 % \u2013 75 % \u2013 90 % \u2013 100 % \u2013 100 % ) using a microwave program with a step length of 40 s and 250 W power , with a steady temperature at 4 \u00b0C and without vacuum . Finally , the cells were in \ufb01 ltrated with a graded series of Hexamethyldisilizane ( HMDS , CAS # 999 - 97 - 3 , Sigma Aldrich ) in ethanol ( 25 % \u2013 50 % \u2013 75 % \u2013 100 % \u2013 100 % ) using a microwave program with 6 steps of 1 min each , with a power of 150 W for step 1 , 3 , 4 and 6 , and 0 W for steps 2 and 5 . After the \ufb01 nal 100 % HMDS in \ufb01 ltration , all HMDS was removed , and coverslips were left to dry overnight . Silica gel with moisture indicator ( Merck ) was added in 4 empty wells ( corners ) in the 24 - well plate to remove excess humidity . After drying , coverslips were mounted in aluminum stubs ( Agar Scienti \ufb01 c G301F ) using carbon tape , and sputter coated with a layer of gold for 180 s at 30mA current using a Quorum sputter coater model Q150RS . Imaging was performed on a Zeiss Crossbean 540 microscope , using 5kV acceleration voltage and 700pA current for the electron beam , with a working distanceof 5 mm . A secondary electron detector ( SESI ) was used for signal detection , and all images were acquired with a pixel size of 28 , 9 nm / pixel . Serial block - face scanning electron microscopy ( SBEM ) After 30 min of 10nM fMLP stimulation in a MatTek dish , cells were \ufb01 xed in 2 . 5 % GA ( # 16220 , EMS ) in 0 . 1M PHEM buffer by adding 37\u00b0C double strength \ufb01 xative ( 5 % GA in 0 . 1 M PHEM ) directly 1 : 1 to the cell medium . After 10 min incubation , the \ufb01 xative was replaced by fresh single - strength \ufb01 xative , and cells were incubated in \ufb01 xative at 4 \u00b0C overnight . After \ufb01 xation , cells were washed 2 times in 0 . 1 M PHEM and 2 times in 0 . 1 M cacodylate buffer . Next , they were post \ufb01 xed for 1 . 5h on ice in 1 % OsO 4 ( # 19190 , EMS ) and 0 . 8 % potassium ferrocyanide ( freshly prepared and \ufb01 ltered ) ( K 4 [ Fe ( CN ) 6 ] * 3H 2 O , # 4984 , Merck ) in 0 . 1 M cacodylate buffer . After post \ufb01 xation , the cells were washed 4 times in H 2 O , and left in H 2 O at 4 \u00b0C until further processing ( 5 days ) . Next , the cells were stained in three consecutive steps in the fol - lowing order : 1 % Thiocarbihydrazide ( TCH , # 21900 , EMS ) in water , 2 % OsO 4 in water , and 1 % UA ( # 77870 , Serva ) in water . For all three staining steps the cells were processed in a Pelco BioWave microwave with a 7 - step program of 2 min each , with power alternating between 100Wand0W ( startingwith100 ) , steadytemperaturewassetto23 \u00b0C and vacuum on for all steps . Between each staining step , the cells were washed 4 times in H 2 O , two times on the bench , and two times in the microwave ( 40s , 250W power , vacuum off ) . After UA staining , the cells were washed once in H 2 O and twice in 25 % EtOH , then further dehydrated in the microwave in a graded ethanol series ( 50 % \u2013 70 % \u2013 90 % \u2013 100 % \u2013 100 % ) . Microwave settings for the dehydration steps were : 40s time , 250W power , vacuum off , steady temp 4 \u00b0C . After dehydration the cells were in \ufb01 ltrated in a graded series of durcupan resin ( Durcupan ACM from Sigma , # 44611 - # 44614 ( four components ) ) in ethanol ( 25 % \u2013 50 % \u2013 75 % \u2013 100 % \u2013 100 % \u2013 100 % durcu - pan ) , using the microwave for 3 min per step at 150W power , 23 \u00b0C steady temp and vacuum cycle . Finally , the cells were put on a small amount of fresh resin ( cov - ering the center of the MatTek dish ) and left for ~ 1 \u2013 2 h to evaporate residual solvent and air bubbles . Before polymerization , most of the resin was removed ( just enough to cover the center was left ) . A drop of durcupan was added to an 18 \u00d718 mm coverslip ( to avoid trapping air bubbles ) , and the coverslip was put on top of the center of the MatTek dish . The assembly was polymerized in the oven at 60\u00b0C for 2 days . After polymerization , the MatTek dish was removed my sawing close to the coverslip , and the glass on both sides of the round center peace was removed by dipping the assembly in liquid nitrogen and warm water . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 11 A small piece was cut out from the sample using a razor blade and mounted on a SEM pin ( Micro to Nano Gatan3View pins , # 10 - 006003 - 50 ) using two - component silver conductive epoxy ( TedPella , # 16043 ) . Thesamplewasalsocoveredwithsilverepoxyonthesides , and \ufb01 nally , sputter coated with a layer of gold for 180s at 30mA current using a Quorum sputter coater model Q150RS . For curing the silver epoxy , the sample was cured in total for an additional day at 60\u00b0C . Image acquisition was performed on a Zeiss GeminiSEM 450 with 3View from Gatan ( DigitalMicrograph Version 3 . 51 . 3720 . 0 ; SmartSEM version 6 . 06 with Service Pack 4 ) , with the program SBEM Image ( 2021 . 08 . dev ) installed for controlling the acquisition 91 . For image acquisition an acceleration voltage of 1 . 5 kV was used , a beam current of 300 pA , pixel size 10 nm and a dwell time of 1 . 6 \u00b5 s . An on - point BSE detector was used for detection , with the contrast set to 99 . 9 % and brightness to 11 . 8 ( with inverted LUT ) , BSD bias \u2212 5V . Slice thickness was set to 40 nm . Between each cycle an overview image was acquired at 165 . 76nm pixel size and 0 . 8 \u00b5 s dwell time . Statistical analysis Statistical analyses were performed using R ( Version 3 . 2 . 1 ) , while data visualization by both R and Adobe Illustrator\u00ae . The normality of data distribution was tested by Shapiro \u2013 Wilk test . A two - tailed t - test was used for normal distribution . Otherwise , a non - parametric Mann \u2013 Whitney - U - test was used , if not indicated differently . In all box plots , the lower and upper hinges correspond to the \ufb01 rst and third quartiles ( the 25th and 75th percentiles ) . The upper whisker extends from the hinge to the largest value , but no further than 1 . 5 * IQR ( distance between the \ufb01 rst and third quartiles ) . The lower whisker extends from the hinge to the smallest value , but no lower than 1 . 5 * IQR of the hinge . Data beyond the end of the whiskers are plotted as black dots . The black line and dot correspond to the median and mean , respectively . All measurements were taken from distinct samples . Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article . Data availability RCSB PDB database was used in the study with accession number 4AKV . The RNAseq data have been deposited to the ArrayExpress collections from BioStudies with the accession number E - MTAB - 12436 . The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner reposi - tory with the dataset identi \ufb01 er PXD033666 . The MD simulation data are available through \ufb01 gshare [ https : / / \ufb01 gshare . com / articles / dataset / Supporting _ data _ for _ Sensing _ their _ plasma _ membrane _ curvature _ allows _ migrating _ cells _ to _ circumvent _ obstacles _ by _ Ewa _ Sitarska _ Silvia _ Dias _ Almeida _ Marianne _ Sandvold _ Beckwith _ Julian _ Stopp _ Jakub _ Czuchnowski _ Marc _ Siggel _ Rita _ Roessner _ Aline _ Tschanz / 22109204 ] . The raw numbers for charts and graphs are available in the Source Data \ufb01 le whenever possible . All other data and unique reagents that support this study are available from the correspond - ing authors upon request . Source data are provided in this paper . Code availability The source code for MD simulations and image analysis are available through \ufb01 gshare [ https : / / \ufb01 gshare . com / articles / dataset / Supporting _ data _ for _ Sensing _ their _ plasma _ membrane _ curvature _ allows _ migrating _ cells _ to _ circumvent _ obstacles _ by _ Ewa _ Sitarska _ Silvia _ Dias _ Almeida _ Marianne _ Sandvold _ Beckwith _ Julian _ Stopp _ Jakub _ Czuchnowski _ Marc _ Siggel _ Rita _ Roessner _ Aline _ Tschanz / 22109204 ] and on github 92 [ https : / / github . com / JakubCzuchnowski / Sensing - their - plasma - membrane - curvature - allows - migrating - cells - to - circumvent - obstacles . git ] . References 1 . Sarris , M . & Sixt , M . Science direct navigating in tissue mazes : chemoattractant interpretation in complex environments . Curr . Opin . Cell Biol . 36 , 93 \u2013 102 ( 2015 ) . 2 . Stoitzner , P . , St\u00f6ssel , H . , Romani , N . & Pfaller , K . A close - up view of migrating langerhans cells in the skin . J . Investig . Dermatol . 118 , 117 \u2013 125 ( 2002 ) . 3 . Weigelin , B . , Bakker , G . - J . & Friedl , P . Intravital third harmonic generation microscopy of collective melanoma cell invasion : prin - ciples of interface guidance and microvesicledynamics . Intravital 1 , 32 \u2013 43 ( 2012 ) . 4 . Diz - Mu\u00f1oz , A . et al . Steering cell migration by alternating blebs and actin - rich protrusions . BMC Biol . 1 \u2013 13 https : / / doi . org / 10 . 1186 / s12915 - 016 - 0294 - x ( 2016 ) . 5 . Fritz - Laylin , L . K . et al . Actin - based protrusions of migrating neu - trophils are intrinsically lamellar and facilitate direction changes . Elife 6 , 437 ( 2017 ) . 6 . Leithner , A . et al . Diversi \ufb01 ed actin protrusions promote environ - mental exploration but are dispensable for locomotion of leuko - cytes . Nat . Cell Biol . 18 , 1253 \u2013 1259 ( 2016 ) . 7 . Baptista , D . , Teixeira , L . , van Blitterswijk , C . , Giselbrecht , S . & Truckenm\u00fcller , R . Overlooked ? Underestimated ? Effects of sub - strate curvature on cell behavior . Trends Biotechnol . 1 \u2013 17 https : / / doi . org / 10 . 1016 / j . tibtech . 2019 . 01 . 006 ( 2019 ) . 8 . Kessels , M . M . & Qualmann , B . Interplay between membrane cur - vature and the actin cytoskeleton . Curr . Opin . Cell Biol . 68 , 10 \u2013 19 ( 2021 ) . 9 . Carman , P . J . & Dominguez , R . BAR domain proteins - a linkage between cellular membranes , signaling pathways , and the actin cytoskeleton . Biophys . Rev . 10 , 1587 \u2013 1604 ( 2018 ) . 10 . McMahon , H . T . & Gallop , J . L . Membrane curvature and mechan - isms of dynamic cell membrane remodelling . Nature 438 , 590 \u2013 596 ( 2005 ) . 11 . Simunovic , M . , Voth , G . A . , Callan - Jones , A . & Bassereau , P . When physics takes over : BAR proteins and membrane curvature . Trends Cell Biol . 25 , 780 \u2013 792 ( 2015 ) . 12 . de Kreuk , B . - J . & Hordijk , P . L . Control of Rho GTPase function by BAR - domains . Small GTPases 3 , 45 \u2013 52 ( 2014 ) . 13 . Galic , M . et al . External push and internal pull forces recruit curvature - sensing N - BAR domain proteins to the plasma mem - brane . Nat . Cell Biol . 14 , 874 \u2013 881 ( 2012 ) . 14 . Zhao , W . et al . Nanoscale manipulation of membrane curvature for probing endocytosis in live cells . 1 \u2013 9 https : / / doi . org / 10 . 1038 / nnano . 2017 . 98 ( 2017 ) . 15 . Lou , H . - Y . et al . Membrane curvature underlies actin reorganization in response to nanoscale surface topography . Proc . Natl Acad . Sci . USA 116 , 23143 \u2013 23151 ( 2019 ) . 16 . Hoogendijk , A . J . et al . Dynamic transcriptome - proteome correla - tion networks reveal human myeloid differentiation and neutrophil - speci \ufb01 c programming . Cell Rep . 29 , 2505 \u2013 2519 . e4 ( 2019 ) . 17 . Bhaskara , R . M . et al . Curvature induction and membrane remo - deling by FAM134B reticulon homology domain assist selective ER - phagy . Nat . Commun . 1 \u2013 13 https : / / doi . org / 10 . 1038 / s41467 - 019 - 10345 - 3 ( 2019 ) . 18 . Jensen , L . E . etal . Membranecurvaturesensingandstabilizationbythe autophagic LC3 lipidation machinery . Sci . Adv . 8 , eadd1436 ( 2022 ) . 19 . Mahmood , M . I . , Noguchi , H . & Okazaki , K . - I . Curvature induction and sensing of the F - BAR protein Pacsin1 on lipid membranes via molecular dynamics simulations . Sci . Rep . 9 , 14557 \u2013 11 ( 2019 ) . 20 . Bigay , J . & Antonny , B . Curvature , lipid packing , and electrostatics ofmembraneorganelles : de \ufb01 ningcellularterritoriesindetermining speci \ufb01 city . Dev . Cell 23 , 886 \u2013 895 ( 2012 ) . 21 . Mattila , P . K . etal . Missing - in - metastasisandIRSp53deformPI ( 4 , 5 ) P 2 - rich membranes by an inverse BAR domain \u2013 like mechanism . J . Cell Biol . 176 , 953 \u2013 964 ( 2007 ) . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 12 22 . Begemann , I . et al . Mechanochemical self - organization determines search pattern in migratory cells . Nat . Phys . 1 \u2013 13 https : / / doi . org / 10 . 1038 / s41567 - 019 - 0505 - 9 ( 2020 ) . 23 . Pipathsouk , A . et al . The WAVE complex associates with sites of saddle membrane curvature . J . Cell Biol . 220 , e202003086 ( 2021 ) . 24 . Weiner , O . D . , Marganski , W . A . , Wu , L . F . , Altschuler , S . J . & Kirschner , M . W . An actin - based wave generator organizes cell motility . PLoS Biol . 5 , e221 ( 2007 ) . 25 . Graziano , B . R . et al . A module for Rac temporal signal integration revealed with optogenetics . J . Cell Biol . 216 , 2515 \u2013 2531 ( 2017 ) . 26 . Lacayo , C . I . et al . Emergence of large - scale cell morphology and movement from local actin \ufb01 lament growthdynamics . PLoS Biol . 5 , e233 ( 2007 ) . 27 . Berg , S . et al . ilastik : interactive machine learning for ( bio ) image analysis . Nat . Methods 16 , 1226 \u2013 1232 ( 2019 ) . 28 . Gauthier , N . C . , Fardin , M . A . , Roca - Cusachs , P . & Sheetz , M . P . Temporary increase in plasma membrane tension coordinates the activation of exocytosis and contraction during cell spreading . Proc . Natl Acad . Sci . USA 108 , 14467 \u2013 14472 ( 2011 ) . 29 . Houk , A . R . et al . Membrane tension maintains cell polarity by con \ufb01 ning signals to the leading edge during neutrophil migration . Cell 148 , 175 \u2013 188 ( 2012 ) . 30 . Rocca , D . L . , Martin , S . , Jenkins , E . L . & Hanley , J . G . Inhibition of Arp2 / 3 - mediated actin polymerization by PICK1 regulates neuronal morphology and AMPA receptor endocytosis . Nat . Cell Biol . 10 , 259 \u2013 271 ( 2008 ) . 31 . Cao , H . et al . FCHSD1 and FCHSD2 are expressed in hair cell ste - reocilia and cuticular plate and regulate actin polymerization in vitro . PLoS ONE 8 , e56516 \u2013 11 ( 2013 ) . 32 . Kostan , J . et al . Direct interaction of actin \ufb01 laments with F - BAR protein pacsin2 . EMBO Rep . 15 , 1154 \u2013 1162 ( 2014 ) . 33 . Dr\u00e4ger , N . M . etal . Bin1directlyremodelsactindynamicsthroughits BARdomain . EMBO Rep . 18 , 2051 \u2013 2066 ( 2017 ) . 34 . Chen , P . - W . et al . The BAR domain of the Arf GTPase - activating protein ASAP1 directly binds actin \ufb01 laments . J . Biol . Chem . 295 , 11303 \u2013 11315 ( 2020 ) . 35 . Graziano , B . R . et al . Cell con \ufb01 nement reveals a branched - actin independent circuit for neutrophil polarity . PLoS Biol . 17 , e3000457 \u2013 34 ( 2019 ) . 36 . Akin , O . & Mullins , R . D . Capping protein increases the rate of actin - based motility by promoting \ufb01 lament nucleation by the Arp2 / 3 complex . Cell 133 , 841 \u2013 851 ( 2008 ) . 37 . Cooper , J . A . & Sept , D . Newinsightsintomechanismandregulation of actin capping protein . Int . Rev . Cell Mol . Biol . 267 , 183 \u2013 206 ( 2008 ) . 38 . Edwards , M . et al . Capping protein regulators \ufb01 ne - tune actin assembly dynamics . 1 \u2013 13 https : / / doi . org / 10 . 1038 / nrm3869 ( 2014 ) . 39 . Funk , J . et al . A barbed end interference mechanism reveals how capping protein promotes nucleation in branched actin networks . Nat . Commun . 1 \u2013 17 https : / / doi . org / 10 . 1038 / s41467 - 021 - 25682 - 5 ( 2021 ) . 40 . Rao , Y . et al . Molecular basis for SH3 domain regulation of F - BAR - mediated membrane deformation . Proc . Natl Acad . Sci . USA 107 , 8213 \u2013 8218 ( 2010 ) . 41 . Kast , D . J . et al . Mechanism of IRSp53 inhibition and combinatorial activation by Cdc42 and downstream effectors . 1 \u2013 11 https : / / doi . org / 10 . 1038 / nsmb . 2781 ( 2014 ) . 42 . Stanishneva - Konovalova , T . B . et al . Coordinated autoinhibition of F - BAR domain membrane binding and WASp activation by Nervous Wreck . Proc . Natl Acad . Sci . USA 113 , E5552 \u2013 E5561 ( 2016 ) . 43 . Inagaki , N . & Katsuno , H . Actin waves : origin of cell polarization and migration ? Trends Cell Biol . 27 , 515 \u2013 526 ( 2017 ) . 44 . Diz - Mu\u00f1oz , A . et al . Membrane tension acts through PLD2 and mTORC2 to limit actin network assembly during neutrophil migra - tion . PLoS Biol . 14 , e1002474 \u2013 30 ( 2016 ) . 45 . Graziano , B . R . et al . Cell con \ufb01 nement reveals a branched - actin independent circuit for neutrophil polarity . PLoS Biol . 17 , e3000457 ( 2019 ) . 46 . Renkawitz , J . et al . Nuclear positioning facilitates amoeboid migration along the path of least resistance . Nature 568 , 546 \u2013 550 ( 2019 ) . 47 . Yamada , K . M . & Sixt , M . Mechanisms of 3D cell migration . Nat . Rev . Mol . Cell Biol . 1 \u2013 15 https : / / doi . org / 10 . 1038 / s41580 - 019 - 0172 - 9 ( 2019 ) . 48 . Roycroft , A . & Mayor , R . Molecular basis of contact inhibition of locomotion . Cell . Mol . Life Sci . 73 , 1119 \u2013 1130 ( 2015 ) . 49 . Stramer , B . & Mayor , R . Mechanismsandinvivofunctionsofcontact inhibition of locomotion . Nat . Rev . Mol . Cell Biol . 18 , 43 \u2013 55 ( 2016 ) . 50 . Driscoll , M . K . et al . Cell shape dynamics : from waves to migration . PLoS Comput . Biol . 8 , e1002392 ( 2012 ) . 51 . Stankevicins , L . etal . Deterministicactinwavesasgeneratorsofcell polarization cues . Proc . Natl Acad . Sci . USA 117 , 826 \u2013 835 ( 2020 ) . 52 . Coutinho - Budd , J . , Ghukasyan , V . , Zylka , M . J . & Polleux , F . The F - BAR domains from srGAP1 , srGAP2 and srGAP3 regulate mem - brane deformation differently . J . Cell Sci . 125 , 3390 \u2013 3401 ( 2012 ) . 53 . Guerrier , S . et al . The F - BAR domain of srGAP2 induces membrane protrusions required for neuronal migration and morphogenesis . Cell 138 , 990 \u2013 1004 ( 2009 ) . 54 . Fritz , R . D . et al . SrGAP2 - dependent integration of membrane geo - metry and slit - robo - repulsive cues regulates \ufb01 broblast contact inhibition of locomotion . Dev . Cell 35 , 78 \u2013 92 ( 2015 ) . 55 . Ren , C . et al . Leukocyte cytoskeleton polarization is initiated by plasma membrane curvature from cell attachment . Dev . Cell 49 , 206 \u2013 219 . e7 ( 2019 ) . 56 . Simunovic , M . , Srivastava , A . & Voth , G . A . Linear aggregation of proteins on the membrane as a prelude to membrane remodeling . Proc . Natl Acad . Sci . USA 110 , 20396 \u2013 20401 ( 2013 ) . 57 . Simunovic , M . , \u0160 ari \u0107 , A . , Henderson , J . M . , Lee , K . Y . C . & Voth , G . A . Long - Range Organization of Membrane - Curving Proteins . ACS Cent . Sci . 3 , 1246 \u2013 1253 ( 2017 ) . 58 . Jarin , Z . et al . Unusual organization of I - BAR proteins on tubular and vesicular membranes . Biophys . J . 117 , 553 \u2013 562 ( 2019 ) . 59 . Nepal , B . , Sepehri , A . & Lazaridis , T . Mechanism of negative mem - brane curvature generation by I - BAR domains . Structure 29 , 1440 \u2013 1452 . e4 ( 2021 ) . 60 . Shalem , O . et al . Genome - scale CRISPR - Cas9 knockout screening in human cells . Science 343 , 84 \u2013 87 ( 2014 ) . 61 . Sanjana , N . E . , Shalem , O . & Zhang , F . Improved vectors and genome - wide libraries for CRISPR screening . Nat . Meth . 1 \u2013 2 https : / / doi . org / 10 . 1038 / nmeth . 3047 ( 2014 ) . 62 . Koch , B . etal . Generationandvalidationofhomozygous \ufb02 uorescent knock - in cells using CRISPR \u2013 Cas9 genome editing . Nat . Publ . Group 13 , 1465 \u2013 1487 ( 2018 ) . 63 . Axelrod , D . Evanescent excitation and emission in \ufb02 uorescence microscopy . Biophys . J . 104 , 1401 \u2013 1409 ( 2013 ) . 64 . Anantharam , A . , Onoa , B . , Edwards , R . H . , Holz , R . W . & Axelrod , D . Localized topological changes of the plasma membrane upon exocytosis visualized by polarized TIRFM . J . Cell Biol . 188 , 415 \u2013 428 ( 2010 ) . 65 . Oreopoulos , J . , Epand , R . F . , Epand , R . M . & Yip , C . M . Peptide - induced domain formation in supported lipid bilayers : direct evi - dence by combined atomic force and polarized total internal re \ufb02 ection \ufb02 uorescence microscopy . Biophys . J . 98 , 815 \u2013 823 ( 2010 ) . 66 . Oreopoulos , J . & Yip , C . M . Combined scanning probe and total internal re \ufb02 ection \ufb02 uorescence microscopy . Methods 46 , 2 \u2013 10 ( 2008 ) . 67 . Sund , S . E . , Swanson , J . A . & Axelrod , D . Cell membrane orientation visualized by polarized total internal re \ufb02 ection \ufb02 uorescence . Bio - phys . J . 77 , 2266 \u2013 2283 ( 1999 ) . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 13 68 . Axelrod , D . Chapter 7 : Total internal re \ufb02 ection \ufb02 uorescence microscopy . Methods Cell Biol . 89 , 169 \u2013 221 ( 2008 ) . 69 . Meijering , E . et al . Design and validation of a tool for neurite tracing and analysis in \ufb02 uorescence microscopy images . Cytom . A 58 , 167 \u2013 176 ( 2004 ) . 70 . Renkawitz , J . , Reversat , A . , Leithner , A . , Merrin , J . & Sixt , M . Micro - engineered \u2018 pillar forests \u2019 to study cell migration in complex but controlled 3D environments . Methods Cell Biol . 147 , 79 \u2013 91 ( 2018 ) . 71 . Kopf , A . et al . Microtubules control cellular shapeand coherence in amoeboid migrating cells . J . Cell Biol . 219 , 193 \u2013 24 ( 2020 ) . 72 . Sens , P . & Plastino , J . Membrane tension and cytoskeleton organi - zation in cell motility . J . Phys . Condens . Matter 27 , 273103 ( 2015 ) . 73 . Hochmuth , F . M . , Shao , J . Y . , Dai , J . & Sheetz , M . P . Deformation and \ufb02 ow of membrane into tethers extracted from neuronal growth cones . Biophys . J . 70 , 358 \u2013 369 ( 1996 ) . 74 . Franken , H . et al . Thermal proteome pro \ufb01 ling for unbiased identi - \ufb01 cation of direct and indirect drug targets using multiplexed quantitative mass spectrometry . Nat . Publ . Group 10 , 1567 \u2013 1593 ( 2015 ) . 75 . Perez - Riverol , Y . et al . The PRIDE database resources in 2022 : a hub for mass spectrometry - based proteomics evidences . Nucl . Acids Res . 50 , D543 \u2013 D552 ( 2022 ) . 76 . Abraham , M . J . et al . GROMACS : high performance molecular simulations through multi - level parallelism from laptops to super - computers . SoftwareX 1 - 2 , 19 \u2013 25 ( 2015 ) . 77 . Marrink , S . J . , Risselada , H . J . , Ye \ufb01 mov , S . , Tieleman , D . P . & deVries , A . H . The MARTINI force \ufb01 eld : coarse grained model for biomole - cular simulations . J . Phys . Chem . B 111 , 7812 \u2013 7824 ( 2007 ) . 78 . Monticelli , L . et al . The MARTINI coarse - grained force \ufb01 eld : exten - sion to proteins . J . Chem . Theory Comput . 4 , 819 \u2013 834 ( 2008 ) . 79 . Benayad , Z . , B\u00fclow , von , S . , Stelzl , L . S . & Hummer , G . Simulation of FUS protein condensates with an adapted coarse - grained model . J . Chem . Theory Comput . 17 , 525 \u2013 537 ( 2021 ) . 80 . Jumper , J . et al . Highly accurate protein structure prediction with AlphaFold . Nature 1 \u2013 12 https : / / doi . org / 10 . 1038 / s41586 - 021 - 03819 - 2 ( 2021 ) . 81 . Evans , R . et al . Protein complex prediction with AlphaFold - Multimer . bioRxiv https : / / doi . org / 10 . 1101 / 2021 . 10 . 04 . 463034 ( 2022 ) . 82 . Periole , X . , Cavalli , M . , Marrink , S . - J . & Ceruso , M . A . Combining an elastic network with a coarse - grained molecular force \ufb01 eld : struc - ture , dynamics , and intermolecular recognition . J . Chem . Theory Comput . 5 , 2531 \u2013 2543 ( 2009 ) . 83 . Wassenaar , T . A . , Ing\u00f3lfsson , H . I . , B\u00f6ckmann , R . A . , Tieleman , D . P . & Marrink , S . J . Computational lipidomics with insane : a versatile tool for generating custom membranes for molecular simulations . J . Chem . Theory Comput . 11 , 2144 \u2013 2155 ( 2015 ) . 84 . Berendsen , H . J . C . , Postma , J . P . M . , vanGunsteren , W . F . , DiNola , A . & Haak , J . R . Molecular dynamics with coupling to an external bath . J . Chem . Phys . 81 , 3684 ( 1998 ) . 85 . Bussi , G . , Donadio , D . & Parrinello , M . Canonical sampling through velocity rescaling . J . Chem . Phys . 126 , 014101 ( 2007 ) . 86 . de Jong , D . H . , Baoukina , S . , Ing\u00f3lfsson , H . I . & Marrink , S . J . Martini straight : Boosting performance using a shorter cutoff and GPUs . Comput . Phys . Commun . 199 , 1 \u2013 7 ( 2016 ) . 87 . Parrinello , M . & Rahman , A . Polymorphic transitions in single crys - tals : a new molecular dynamics method . J . Appl . Phys . 52 , 7182 ( 1998 ) . 88 . Humphrey , W . , Dalke , A . & Schulten , K . VMD : visual molecular dynamics . J . Mol . Graph 14 , 33 \u2013 8 \u2013 27 \u2013 8 ( 1996 ) . 89 . Gowers , R . et al . in Proceedings of the 15th Python in Science Con - ference ( https : / / conference . scipy . org / proceedings / scipy2016 / oliver _ beckstein . html ) , pp . 98 \u2013 105 ( 2016 ) . 90 . Michaud - Agrawal , N . , Denning , E . J . , Woolf , T . B . & Beckstein , O . MDAnalysis : a toolkit for the analysis of molecular dynamics simu - lations . J . Comput . Chem . 32 , 2319 \u2013 2327 ( 2011 ) . 91 . Titze , B . , Genoud , C . & Friedrich , R . W . SBEMimage : versatile acquisition control software for serial block - face electron micro - scopy . Front . Neural Circuits 12 , 54 ( 2018 ) . 92 . Czuchnowski , J . Supporting data for \u201c Sensing their plasma mem - brane curvature allows migrating cells to circumvent obstacles \u201d . GitHub , https : / / doi . org / 10 . 5281 / zenodo . 8169105 ( 2023 ) . Acknowledgements We thank Jan Ellenberg , Leanne Strauss , Anusha Gopalan , and Jia Hui Li for critical feedback on the manuscript and the Life Science Editors for editing assistance . The plasmid with hSnx33 was a kind gift from Duanqing Pei . Cell line with GFP - tagged IRSp53 was a kind gift from Orion Weiner . We thank Brian Graziano for providing protocols , reagents , and key advice to generate CRISPR knockout HL - 60 cells . We thank the EMBL \ufb02 ow cytometry core facility , the EMBL advanced light microscopy facility , the EMBL proteomics facility , and the EMBL geno - mics core facility for support and advice . We thank Anusha Gopalanand Martin Bergert for their support during mechanical measurements by AFM . We thank Estela Sosa Osorio for technical assistance for the co - immunoprecipitation . We thank the EMBL genome biology computa - tional support ( and specially Charles Girardot and Jelle Scholtalbers ) for critical assistance during RNAseq analysis . We thank Hans Kristian Hannibal \u2010 Bach for his technical assistance during the lipidomic analysis of plasma membrane isolates . We thank Steffen Burgold for their sup - port with LLS7 microscope in the ZEISS Microscopy Customer Center Europe . We acknowledge the \ufb01 nancial support of the European Mole - cular Biology Laboratory ( EMBL ) to A . D . - M . , Y . S . , A . K . , and A . E . , the EMBL Interdisciplinary Postdocs ( EIPOD ) program under Marie Sklodowska - Curie COFUND actions MSCA - COFUND - FP to M . S . B . and M . S . ( grant agreement number : 847543 ) , the BEST program funding by FCT ( SFRH / BEST / 150300 / 2019 ) to S . D . A . and the Joachim Herz Stiftung Add - on Fellowship for Interdisciplinary Science to E . S . Author contributions A . D . - M . and E . S . conceived the project and designed the experiments . E . S . , M . S . B . , and J . S . performed experiments with EM advice from Y . S . and micro \ufb02 uidic support from M . S . A . D . - M . and C . S . E . isolated plasma membranes and performed lipidomic analysis . M . S . and R . R . performed MD simulations with support from J . K . E . S . and S . D . A . analyzed the data with support from J . C . , A . E . , and A . K . A . D . - M . , E . S . , and A . E . wrote the manuscript . Allauthorscontributedtotheinterpretationofthedata , and read and approved the \ufb01 nal manuscript . Funding Open Access funding enabled and organized by Projekt DEAL . Competing interests The authors declare no competing interests . Additional information Supplementary information The online version contains supplementary material available at https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 . Correspondence and requests for materials should be addressed to Alba Diz - Mu\u00f1oz . Peer review information Nature Communications thanks Chetan Poojari and the anonymous reviewer ( s ) for their contribution to the peer review of this work . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 14 Reprints and permissions information is available at http : / / www . nature . com / reprints Publisher \u2019 s note Springer Nature remains neutral with regard to jur - isdictional claims in published maps and institutional af \ufb01 liations . Open Access This article is licensed under a Creative Commons Attribution 4 . 0 International License , which permits use , sharing , adaptation , distribution and reproduction in any medium or format , as long as you give appropriate credit to the original author ( s ) and the source , provide a link to the Creative Commons license , and indicate if changes were made . The images or other third party material in this article are included in the article \u2019 s Creative Commons license , unless indicated otherwise in a credit line to the material . If material is not included in the article \u2019 s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use , you will need to obtain permission directly from the copyright holder . To view a copy of this license , visit http : / / creativecommons . org / licenses / by / 4 . 0 / . \u00a9 The Author ( s ) 2023 Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 41173 - 1 Nature Communications | ( 2023 ) 14 : 5644 15",
    "day2015microtubule": "doi : 10 . 1111 / tra . 12269 Microtubule Motors Power Plasma Membrane Tubulation in Clathrin - Independent Endocytosis Charles A . Day 1 , 2 , Nicholas W . Baetz 1 , Courtney A . Copeland 1 , Lewis J . Kraft 3 , Bing Han 1 , Ajit Tiwari 1 , Kimberly R . Drake 1 , Heidi De Luca 4 , Daniel J . - F . Chinnapen 4 , Michael W . Davidson 5 , Randall K . Holmes 6 , Michael G . Jobling 6 , Trina A . Schroer 7 , Wayne I . Lencer 4 , 8 and Anne K . Kenworthy 1 , 3 , 9 , 10 \u2217 1 Department of Molecular Physiology and Biophysics , Vanderbilt University School of Medicine , Nashville , TN , USA 2 Current address : Hormel Institute , University of Minnesota , Austin , MN , USA 3 Chemical and Physical Biology Program , Vanderbilt University , Nashville , TN , USA 4 GI Cell Biology , Department of Pediatrics , Boston Children\u2019s Hospital , Boston , MA , USA 5 National High Magnetic Field Laboratory , The Florida State University , Tallahassee , FL , USA 6 Department of Immunology and Microbiology , University of Colorado School of Medicine , Aurora , CO , USA 7 Department of Biology , The Johns Hopkins University , Baltimore , MD , USA 8 Harvard Medical School and the Harvard Digestive Diseases Center , Boston , MA , USA 9 Department of Cell and Developmental Biology , Vanderbilt University School of Medicine , Nashville , TN , USA 10 Epithelial Biology Program , Vanderbilt University School of Medicine , Nashville , TN , USA \u2217 Corresponding author : Anne K . Kenworthy , Anne . kenworthy @ vanderbilt . edu Abstract How the plasma membrane is bent to accommodate clathrin - independent endocytosis remains uncertain . Recent studies suggest Shiga and cholera toxin induce membrane curvature required for their uptake into clathrin - independent carriers by binding and cross - linking multiple copies of their glycosphingolipid receptors on the plasma membrane . But it remains unclear if toxin - induced sphin - golipid crosslinking provides suf\ufb01cient mechanical force for deforming the plasma membrane , or if host cell factors also contribute to this process . To test this , we imaged the uptake of cholera toxin B - subunit into surface - derived tubular invaginations . We found that cholera toxin mutants that bind to only one glycosphingolipid receptor accumulated in tubules , and that toxin binding was entirely dispensable for membrane tubulations to form . Unexpectedly , the driving force for tubule extension was supplied by the combination of microtubules , dynein and dynactin , thus de\ufb01ning a novel mechanism for generating membrane curvature during clathrin - independent endocytosis . Keywords cholera toxin , clathrin - independent endocytosis , dynactin , dynein , membrane curvature , microtubules Received 30 January 2015 , revised and accepted for publication 6 February 2015 , uncorrected manuscript published online 18 February 2015 , published online 27 April 2015 Endocytosis , the process of internalizing vesicles or tubules and their associated cargo from the cell surface , serves essential functions for all cell types . Two general mechanisms of membrane uptake have been described . One involves the assembly of a rigid clathrin coat on the cytoplasmic surface of the plasma membrane that gener - ates a highly curved membrane invagination , sequesters cargo and facilitates membrane fission and budding from the cell surface for trafficking into the cell ( 1 ) . The other mechanism is clathrin - independent , and consists of multiple pathways , but how clathrin - independent carriers are generated remains poorly understood ( 2 \u2013 9 ) . The AB 5 subunit toxins cholera ( CTx ) and Shiga toxin ( STx ) have become widely used as model cargoes for study - ing clathrin - independent endocytosis , including uptake via tubular endocytic carriers ( 10 \u2013 25 ) . Both toxins bind plasma membrane glycosphingolipids , and contain five or fifteen receptor binding sites for high - avidity association with cells via their respective B - subunits ( 26 , 27 ) . Receptor 572 www . traf\ufb01c . dk \u00a9 2015 The Authors . Traf\ufb01c published by John Wiley & Sons Ltd . This is an open access article under the terms of the Creative Commons Attribution - NonCommercial License , which permits use , distribution and reproduction in any medium , provided the original work is properly cited and is not used for commercial purposes . Microtubules Tubulate the Plasma Membrane cross - linking is dispensable for CTx intoxication of host cells ( 28 , 29 ) , but it is clear that CTx and STx binding to multiple glycosphingolipid receptors affects the membrane dynamics of uptake and intracellular trafficking in ways that enhance their toxicity ( 18 , 30 ) . One way this is thought to occur is via a novel clathrin - independent mechanism in which toxin binding induces negative curvature of the plasma membrane ( 18 , 31 ) . This mechanism has been best described for the B - subunit of Shiga toxin ( STxB ) , but is also utilized by the B - subunit of cholera toxin ( CTxB ) and the glycosphingolipid - binding virus simian virus 40 ( 18 , 30 ) . According to this model , toxin binding deforms the mem - brane beneath the toxin by compacting the glycosphin - golipids and / or by reorienting them to form a curved sur - face ( 18 , 30 ) . In support of this idea , STxB or CTxB binding to giant unilamellar lipid vesicles ( GUVs ) causes inwardly directed tubulations of these artificial lipid membranes . Similarly , in live cells subjected to treatments that block membrane scission , toxins accumulate in tubular invagi - nations originating from the plasma membrane , analo - gous to those seen in GUVs ( 18 , 30 ) . These microns - long structures are devoid of markers of clathrin - dependent endocytosis and can form even under conditions where cellular ATP is depleted ( 18 ) . Toxin - induced changes in membrane organization are thought to be important for their formation , operating from outside the cell to form plasma membrane tubules in the absence of active cel - lular processes . But it remains unclear how toxin bind - ing could supply the mechanical force needed to form the microns - long membrane invaginations observed in cells under these conditions . To what extent host cell fac - tors contribute to this process is also uncertain , as few endogenous cellular regulators of this pathway have been identified . To address these questions , we investigated the mechanisms responsible for plasma membrane deforma - tion during clathrin - independent endocytosis of CTxB in live cells . Results Assay for analysis of tubulation of the plasma membrane To study mechanisms that participate in bending the plasma membrane during clathrin - independent endocytosis , we took advantage of a previously described assay that traps STxB and CTxB in surface attached tubules thought to correspond to stalled endocytic intermediates ( 18 , 30 ) . Several conditions facilitate the growth of surface attached tubular invaginations containing fluorescently labeled toxins , including ATP depletion , actin disruption , cholesterol depletion and inhibition of dynamin . These conditions are thought to be permissive for tubule growth , but prevent tubule scission , leading to the accumulation of toxin within these structures . CTxB accumulated in linear , microns - long tubules in ATP - depleted cells and / or when actin was either disrupted or stabilized ( Figures 1A \u2013 D and S1 ) . Control experiments revealed that STxB and CTxB localized to the same tubules in cells incubated with both toxins , supporting the idea they are formed by a common mechanism ( Figure S1D ) . We also found that the tubules were rapidly labeled by a second application of CTxB after they had formed , confirming that the tubules originated from the plasma membrane and were surface attached for the time course of these studies ( Figure S1E ) . In cells treated with the dynamin inhibitor Dynasore , irregular branched networks of tubules were observed ( Figure 1E ) , resembling those seen before in cells expressing a Dynamin1 K44A mutant ( 17 ) . These findings confirm that CTxB readily associates with membrane tubules under a variety of conditions that inhibit scission , thus allowing us to further study requirements for tubule formation . Binding to a single GM1 is suf\ufb01cient to support the association of cholera toxin with tubular invaginations Current models propose compaction of glycosphingolipids by toxin binding plays an essential role in initiating mem - brane curvature and tubulating the membrane to form plasma membrane invaginations ( 18 ) . A mutant form of STxB lacking the Gb 3 binding site III demonstrated greatly reduced ability to drive tubule formation in cells , suggesting multiple glycosphingolipid binding sites enhance the ability of toxins to deform membranes ( 18 ) . To test if cholera toxin binding causes plasma mem - brane tubulations by compacting or reorienting multiple glycosphingolipids into a curved surface in a similar manner , we first studied a mixture of chimeric cholera toxins ( CTx chimera ) that bind 0 , 1 or 2 glycosphingolipid Traf\ufb01c 2015 ; 16 : 572 \u2013 590 573 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Day et al . Figure 1 : Legend on next page . 574 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Microtubules Tubulate the Plasma Membrane receptors ( ganglioside GM 1 ) instead of the usual 5 ( 28 ) . Unexpectedly , the chimeric mutant toxins were readily observed in tubular invaginations in ATP - depleted cells , and the average number of invaginations per cell and length of the invaginations were similar for wild type and chimeric toxins ( Figure 1F \u2013 I ) . We next asked if a toxin with only a single GM 1 binding site ( monovalent CTx ) ( 29 ) can be directed to tubules . Strikingly , this monovalent CTx was localized within tubular invaginations as well ( Figure 1J \u2013 O ) , suggesting that toxin - induced crosslinking of GM 1 is dispensable for tubule formation . Toxin binding is dispensable for tubule formation The finding that extensive crosslinking of GM 1 is not required for tubules to form suggests that the elongated tubules containing CTx may be generated by machin - ery endogenous to the host cell . If so , cellular plasma membrane proteins should also be found in the tubu - lar invaginations in the absence of bound toxin . To test this , we first screened several cell surface markers for co - localization with CTxB in tubular invaginations . Sev - eral cell surface proteins were found to extensively label tubular structures even in the absence of CTxB binding ( Figures 2 and S2 ) . As one example , the cytosolic plasma membrane - associated protein HRas ( GFP - HRas ) strongly localized to the tubular invaginations both in the pres - ence of CTxB and independently of CTxB in response to ATP - depletion , actin disruption or actin stabilization ( Figure 2A \u2013 F ) . In ATP - depleted cells , a similar number of GFP - HRas positive tubules were found in the pres - ence ( 18 \u00b1 15 , n = 26 cells ) or absence ( 18 \u00b1 11 , n = 23 cells ) of CTxB , suggesting they form by the same mecha - nism . Tubule formation was not stimulated by the GTPase activity of Ras , because a minimal membrane targeted form of GFP , GFP - HRas tail , also labeled tubules ( Figure 2G , H ) . Thus , tubulation of the plasma membrane can occur in the absence of toxin - induced cross - linking of glycolipids , indi - cating that the driving force ( s ) for tubule extension can be generated by factors endogenous to the host . An intact microtubule network is required for the formation of extended tubular invaginations It is well known that microtubules and microtubule motors are capable of deforming membranes ( 32 \u2013 34 ) . Such mech - anisms are not currently thought to contribute to the early stages of endocytosis ( 32 ) . However , CTxB has previously been found to localize within microtubule - dependent tubular invaginations of intact BSC1 cells , suggesting a microtubule - dependent process of toxin uptake ( 13 ) . Consistent with these findings , we noticed that the tubular invaginations containing CTxB in ATP - depleted cells were often directed toward the cell center in an orientation typifying the microtubule network ( Figures 1F , J and 2A ) and that the microtubule networks remained intact after ATP depletion ( Figure 3A ) . Tubular invaginations con - taining CTxB were also often found aligned closely with taxol - stabilized microtubules ( Figure 3B , C ) . When imaged over time ( Movies S1 and S2 ) , the tubules sometimes grew smoothly ( Figure 3D \u2013 F ) , but were often observed to pause and undergo bi - directional motions ( Figure 3G \u2013 I ) and branching events ( Figure 3J ) character - istic of microtubule - dependent motions . We thus asked if the microtubule network was required for tubular invagi - nations to form . Remarkably , disruption of microtubules prior to CTxB binding led to a complete loss of tubular Figure 1 : No more than one functional GM 1 binding site is required to target cholera toxin to plasma membrane invaginations . A \u2013 E ) CTxB accumulates in either linear extended tubules ( A \u2013 D ) or branched tubules ( E ) under conditions that block scission . Bar , 10 \u03bc m . ( F \u2013 M ) Cholera toxin binding mutants accumulate in tubular invaginations . F ) Cy3 - CTx chimera labels tubules in ATP depleted COS - 7 cells . G \u2013 I ) Quanti\ufb01cation of invaginations in ATP - depleted cells labeled with Cy3 - CTx chimera or Alexa555 - CTxB . G ) Percentage of cells displaying invaginations ( mean \u00b1 SD from 117 \u2013 119 cells ) . * , p < 0 . 05 , chi - squared test . H ) Average number of invaginations per cell ( mean \u00b1 SD of 42 \u2013 46 cells ) . n . s . , p > 0 . 05 ; Student t - test . I ) Length of invaginations ( mean \u00b1 SD for 219 \u2013 332 invaginations ) . n . s . , p > 0 . 05 ; Student t - test . J \u2013 M ) Both wild type CTxB and monovalent CTx accumulate in tubular invaginations in cells subjected to Jasplakinolide pretreatment prior to ATP depletion . L ) Percentage of cells displaying invaginations . ( mean \u00b1 SD of 59 \u2013 63 cells ) . n . s . , p > 0 . 05 ; chi - squared test . M ) Average number of invaginations per cell . ( mean \u00b1 SD of 59 \u2013 63 cells ) . n . s . , p > 0 . 05 ; Student t - test . N and O ) Similar to wild type CTxB , monovalent CTx accumulates in branched tubules in Dynasore - treated cells . Bars , 10 \u03bc m . Traf\ufb01c 2015 ; 16 : 572 \u2013 590 575 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Day et al . Figure 2 : Toxin binding is not necessary for tubular invaginations to form . A , B ) EGFP - HRas ( green ) is found in plasma membrane invaginations in ATP - depleted cells in both the presence ( A ) and absence ( B ) of Alexa555 - CTxB ( red ) . C \u2013 F ) Similar results were obtained for GFP - HRas in cells subjected to actin disruption ( C and D ) or actin stabilization ( E and F ) . G and H ) A construct containing only the C - terminal 10 amino acids of HRas , EGFP - HRas - tail ( green ) , also localized to tubules in both the presence and absence of CTxB . Bars , 10 \u03bc m . invaginations containing the toxin in ATP - depleted cells ( Figure 4A , C ) . Microtubule disruption also inhibited the formation of tubules containing CTxB or monovalent CTx in cells subjected to dynamin inhibition , actin disruption or actin stabilization ( Figure 4E , F ; Figure S3A , B ) . Thus , the extended tubular invaginations are strongly microtubule dependent . Microtubule plus end dynamics are not required for the growth of tubular invaginations To elucidate how microtubules might support tubulation of the plasma membrane , we first considered a mechanism in which interactions between membranes and dynamic microtubules are mediated by plus - end binding proteins to drive endomembrane translocation ( 35 \u2013 39 ) . To test this possibility , we monitored the plus - end binding pro - tein GFP - EB3 ( 40 ) . In ATP - depleted cells , GFP - EB3 still labeled microtubules , but was no longer concentrated at their tips ( Movie S3 ) , indicating enrichment of plus - end binding proteins at microtubule ends cannot be required for tubule formation . Furthermore , pretreating cells with low doses of nocodazole ( 150 nM ) to suppress micro - tubule plus - end dynamics ( 40 \u2013 42 ) had no detectable effect on the number of ATP - depleted cells that contained CTxB - positive invaginations ( Figure 4B , D ) . Dynamic microtubule growth thus cannot explain tubule extension . 576 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Microtubules Tubulate the Plasma Membrane Figure 3 : Tubular invaginations align along microtubules and undergo complex motions including bidirectional motility and branching events . A ) Microtubules persist in RFP - \u03b1 - tubulin expressing HeLa cells following ATP depletion . B ) CTxB positive invaginations ( green ) align with taxol - stabilized microtubules ( red ) in stably expressing RFP - \u03b1 - tubulin HeLa cells under ATP depletion . C ) Percentage of CTxB - positive tubules that align with microtubules ( black ) , partially align with microtubules ( gray ) , or do not colocalize with microtubules ( light gray ) in ATP - depleted HeLa cells expressing RFP - \u03b1 - tubulin . N = 84 cells . D \u2013 J ) CTxB - enriched tubules exhibit complex motions in live cells as illustrated by representative frames from time series and corresponding kymographs . Time stamps are in minutes : seconds . D \u2013 F ) CTxB - positive invaginations often grow in \ufb02uid directed motions . G \u2013 I ) CTxB - positive invaginations also extend , retract and regrow along the same axis . J ) Occasionally , the CTxB positive tubules undergo branching events . Bars , 10 \u03bc m . Low levels of microtubule motor activity are retained in ATP - depleted cells Another way microtubules could support the growth of in - vaginations would be through the activity of microtubule - based motors ( 32 \u2013 34 , 43 ) . Microtubule - based motility involves two classes of motor proteins , dynein and mem - bers of the kinesin family ( 44 , 45 ) . Given that the tubular invaginations underwent preferential growth toward the center of the cell , where microtubule minus ends are located ( Figure 1 , Movies S1 and S2 ) , we hypothesized that the minus - end directed motor dynein might be involved . Dynein is an ATPase ( 44 ) whose activity is expected to be attenuated in ATP - depleted conditions . We thus asked if dynein still functions as a motor in cells depleted of ATP by monitoring the intracellular movement of lysosomes labeled with mCherry - tagged LAMP1 ( 46 ) . Traf\ufb01c 2015 ; 16 : 572 \u2013 590 577 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Day et al . Figure 4 : An intact microtubule network is required for the formation of tubular invaginations . A and C ) Microtubule disruption with high dose nocodazole prevents the formation of tubular invaginations in ATP - depleted cells ( mean \u00b1 SD , N = 74 cells . ) * * p < 0 . 01 , chi - squared test . B and D ) Inhibition of microtubule dynamics with low dose nocodazole has no effect on the formation of tubular invaginations in ATP - depleted cells ( mean \u00b1 SD , N = 135 \u2013 169 cells . ) n . s . p > 0 . 05 , chi - squared test . E and F ) Microtubule disruption prior to Dynasore treatment blocks the formation of branched tubules in cells labeled with either wild type CTxB ( E ) or monovalent CTx ( F ) . Bars , 10 \u03bc m . Although strongly reduced compared to control cells , some long - range motions of lysosomes persisted in ATP - depleted cells ( Figure 5A \u2013 G ) ( Movie S4 ) . These data imply that motor proteins remain active at low levels in ATP - depleted cells , consistent with dynein\u2019s known activity under low ATP conditions in vitro ( 47 , 48 ) . We also asked whether dynein is localized appropriately to assist with tubule extension . Due to incompatibility of fix - ation conditions required to preserve dynein staining and tubule morphology , we were unable to determine whether endogenous dynein was present on tubular invagina - tions . We therefore instead used a HeLa cell line stably expressing low levels of multifunctional green fluorescent protein ( mfGFP ) - tagged 74 - kDa dynein intermediate chain ( IC74 ) ( 49 ) to visualize both simultaneously . Some mfGFP - IC74 could be observed at the plasma membrane in ATP - depleted cells , and a few dynein intermediate chain - positive puncta co - localized with the tubular invaginations ( Figure S4 ) . These findings further support the possibility that at least a small number of dyneins are localized correctly to facilitate plasma membrane tubulation under these conditions . The ATPase activity of dynein and an intact dynactin complex are required for tubule extension To test if dynein contributes to the formation of extended tubular plasma membrane invaginations , we used a small molecule inhibitor of dynein , ciliobrevin A ( 50 ) . In the presence of ciliobrevin A , we observed impaired formation of tubular invaginations containing CTxB in ATP - depleted cells ( Figure 6A , B ) . This result implicates dynein in the extension of tubular invaginations . To test this hypothe - sis in another way , we inhibited the function of dynactin , a complex required for dynein function ( 44 , 51 ) . We per - turbed dynactin by overexpressing either a GFP - tagged form of p50 / dynamitin ( 51 ) ( Figure 6C ) or a dsRed - tagged form of the dynein - binding portion of the dynactin subunit 578 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Microtubules Tubulate the Plasma Membrane Figure 5 : Motor - based motions persist in ATP - depleted cells . A \u2013 C ) A subset of lysosomes labeled with mCherry - LAMP1 display long range directed motions in ATP - depleted cells . B ) Time lapse of zoomed in region of the cell in panel A . An example of a lysosome undergoing long - range directed motion is marked with the arrowhead . C ) Tracings of lysosome movement in the cell shown in A . Each track is indicated by a different color . D and E ) Frequent long - range directed motions are observed under control conditions . F and G ) mCherry - LAMP1 positive lysosomes are immobile in PFA \ufb01xed cells . Elapsed time in F , H and J = 470 s . Bars , 10 \u03bc m . p150 Glued , CC1 - dsRed ( 51 ) ( Figure 6D , E ) . Overexpression of either protein strongly inhibited the formation of tubular invaginations in response to ATP depletion ( Figure 6C \u2013 E ) . p50 expression also blocked extended branched tubules from forming in Dynasore - treated cells ( Figure S3E , F ) . Thus , dynactin is required for tubulation of the membrane , further implicating dynein as a host cell factor that under - lies tubule extension . Bulk uptake of CTxB is unaffected by disruption of microtubules or the dynactin complex , suggesting the tubular carriers de\ufb01ne a low capacity endocytic pathway Given our findings that microtubules , dynactin and dynein are required for plasma membrane tubulation , we won - dered if they are also necessary for the uptake of CTxB . CTxB can be internalized by multiple mechanisms ( 12 , 13 ) , including a high capacity pathway that involves morpho - logically distinct clathrin - independent carriers ( 16 , 17 ) and tubular endocytic intermediates that contain CTxB ( 13 ) . Interestingly , we found that uptake of CTxB into ATP replete cells was unaffected by the expression of GFP - p50 ( Figure 7A , B ) or pretreatment of cells with nocodazole ( Figure 7C , D ) . Thus , the microtubule - dependent pathway cannot be highly efficient or high capacity . However , these results do not exclude the possibility that uptake of toxin by plasma membrane tubules is physiologically relevant \u2013 as cholera toxin - induced toxicity is poorly correlated with overall levels of toxin internalization ( 13 ) . Indeed , we found that intact microtubules were required for full tox - icity of CTx in polarized human intestinal epithelial T84 cells , as measured by an electrophysiological assay that monitors toxin - induced Cl \u2212 secretion ( 52 ) ( Figure 7E ) . This is consistent with a role for microtubules in the uptake or trafficking of CT . Discussion Here , we show that microtubules , dynein and dynactin pro - vide an important source of mechanical force that tubulates Traf\ufb01c 2015 ; 16 : 572 \u2013 590 579 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Day et al . Figure 6 : The ATPase activity of dynein and an intact dynactin complex are required for the formation of tubular invaginations . A and B ) Inhibition of dynein ATPase activity with ciliobrevin - A ( Cilio - A ) signi\ufb01cantly reduces the percent of cells displaying tubular invaginations ( mean \u00b1 SD from 95 \u2013 122 cells ) . * * p < 0 . 01 , chi - squared test . C ) Expression of GFP - p50 reduced the prevalence of cells with invaginations as compared to untransfected cells ( \u2212 ) or cells expressing EGFP . ( mean \u00b1 SD from 42 \u2013 101 cells ) . n . s . , p > 0 . 05 ; * * p < 0 . 01 , chi - squared test . D and E ) Expression of CC1 - dsRed ( red ) signi\ufb01cantly reduces the percent of cells with CTxB positive invaginations ( green ) as compared to untransfected cells ( \u2212 ) or control cells expressing mCherry ( red ) . ( mean \u00b1 SD from 40 \u2013 64 cells ) . n . s . , p > 0 . 05 ; * * p < 0 . 01 , chi - squared test . Bars , 10 \u03bc m . the plasma membrane during clathrin - independent endocytosis of CTxB . Interestingly , similar tubules were observed in the presence and absence of bound toxin , sug - gesting they demarcate an endogenous , toxin - independent pathway . Very recent evidence indicates microtubules and dynein are also involved in the clathrin - independent uptake of STxB via endophilin A2 - containing tubules ( 53 ) . Formation of these endophilin A2 - positive tubules is strongly induced by toxin binding ( 53 ) . Thus , the mecha - nism of dynein - and microtubule - dependent membrane tubulation that we describe here appears to be a general one utilized by multiple classes of clathrin - independent carriers , inclusive of both constitutive and cargo - induced pathways . On the basis of our findings , we propose a model wherein dynein and dynactin can interact with the plasma mem - brane and adjacent microtubules to allow formation and extension of nascent endocytic tubules ( Figure 8 ) . Dynein likely drives tubule extension by pulling the membrane along existing microtubules , albeit only slowly under conditions where ATP is limiting as in ATP - depleted cells . Other minus end directed motors , such as kinesins - 14 might also contribute to these processes . By providing an internal pulling force that drives membrane curvature , motor - driven bending of the plasma membrane could facilitate recruitment of curvature - sensitive / generating proteins that help stabilize and elongate tubules ( 54 , 55 ) . Pulling forces could also participate in other steps in 580 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Microtubules Tubulate the Plasma Membrane Figure 7 : Legend on next page . Traf\ufb01c 2015 ; 16 : 572 \u2013 590 581 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Day et al . clathrin - independent endocytosis . For example , they could help sort cargo into tubules in a curvature - dependent manner ( 56 \u2013 58 ) , as well as contribute to the process of membrane scission ( 53 , 59 ) . The exact mechanism by which dynein and dynactin bind the plasma membrane and how this process is regulated in clathrin - independent endocytosis remains to be deter - mined . It could involve recruitment to sites of local mem - brane bending induced by endogenous cellular factors , similar to that previously described for recruitment of dynein by retromer ( 56 ) , or in the case of the toxins , by membrane bending at the site of toxin binding as discussed further below ( 18 , 30 , 53 ) . Caveolae could potentially serve as a preferential site for dynein / dynactin recruitment to the plasma membrane , as caveolin - 1 has been observed in long , toxin - positive tubules under conditions that inhibit endocytosis ( 24 , 60 ) . The mechanism of attachment of dynein / dynactin to the plasma membrane may alternatively be related to those involved in positioning of the mitotic spindle ( 61 ) . This scenario seems less likely given that stable anchor - ing of dynein at the cell cortex is thought to facilitate microtubule - dependent pulling forces ( 62 ) . Interestingly however , microtubule - dependent plasma membrane invaginations have been reported to form in C . elegans embryos at cortical sites where spindle poles are tethered ( 63 ) . These invaginations are observed at low frequency in unperturbed embryos , but are more readily evident under conditions where the acto - myosin cortex is weakened ( 63 ) . This suggests that cortical actin reorganization may repre - sent a key event that determines whether microtubule - and dynein - dependent tubulation of the plasma membrane can occur . For example , cortical actin may normally act as a physical barrier between microtubules and the cell surface . Reorganization of cortical actin , either through endogenous processes or in response to actin perturb - ing agents ( 64 ) , may permit microtubules access to the plasma membrane , facilitating the formation of the extended tubular carriers into which endocytic cargo such as CTxB enters . Changes in cortical actin organi - zation could also potentially decrease plasma membrane tension , favoring the formation of long invaginations by microtubule - dependent pulling . Actin dynamics may in turn ultimately also contribute to the tubule scission process that releases tubules from the plasma membrane ( 21 , 53 ) . Our findings also have important implications for our understanding of how glycosphingolipid - binding toxins such as cholera toxin and Shiga toxin manipulate cell mem - branes to facilitate their uptake and subsequent cellular intoxification . It is clear that CTx and STx binding to mul - tiple copies of their glycosphingolipid receptors enhances their toxicity ( 18 , 28 , 29 , 30 , 65 ) and that toxins can induce membrane bending in vitro ( 18 , 30 ) . Toxin binding can also initiate the recruitment of curvature - sensing proteins to the plasma membrane ( 53 ) . However , cross - linking of membrane lipids is dispensable for both plasma mem - brane tubulation ( this study ) and CTx intoxication of host cells ( 28 , 29 ) . While CTxB binding is not required for extended tubular invaginations to form , membrane reor - ganizations and local bending induced by the AB5 tox - ins binding to their sphingolipid receptors ( 18 , 30 ) may function cooperatively with the microtubule - based mech - anisms of membrane tubulation we propose here . Toxin binding may for example contribute to the induction or Figure 7 : Bulk endocytosis of CTxB is largely unaffected by disruption of microtubules or the dynactin complex . A and B ) Endocytosis of dextran , transferrin and CTxB in cells expressing GFP - p50 ( gray bars ) compared to cells expressing GFP ( black bars ) . B shows mean \u00b1 SD for 37 \u2013 278 cells . n . s . , p > 0 . 05 , Student\u2019s t - test . C and D ) Effect of microtubule disruption with 5 \u03bc g / mL nocodazole ( gray bars ) on the uptake of dextran , transferrin and CTxB . Control cells were treated with DMSO ( black bars ) . D shows mean \u00b1 SD for 78 \u2013 358 cells . n . s . , p > 0 . 05 , * , p < 0 . 05 , Student\u2019s t - test . Bars , 10 \u03bc m . E ) Representative time course of toxin - induced chloride secretion in T84 cells in response to treatment with 20 nM wt CTx . Cells were either pretreated with NZ ( closed symbols ) or with DMSO ( open symbols ) prior to toxin addition to either the apical ( blue ) or basolateral ( red ) surface at t = 30 min as described in the MaterialsandMethods . Forskolin was added to control monolayers ( green or gray diamonds ) at 90 min in order to demonstrate equivalency of the secretory response and monolayer viability . The error bars indicate the variance calculated as the standard deviation ( n = 3 ) . Data are representative of the results of two independent experiments . 582 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Microtubules Tubulate the Plasma Membrane Figure 8 : Working model : microtubules , dynactin and cytoplasmic dynein facilitate plasma membrane tubu - lation . Dynein and dynactin provide attachment sites or generate tugging forces on the plasma membrane , leading to microtubule - dependent tubulation . Glycolipid - binding toxins may further sense , induce or stabilize membrane curvature , enabling their ef\ufb01cient sorting into tubular structures . stabilization of membrane curvature ( 53 ) or favor sort - ing of the glycosphingolipid - toxin complex into curved structures ( 66 ) . Indeed , we observed a slight preference of wild type CTxB for sorting into tubules compared to mutant toxins containing only 1 or 2 lipid GM 1 bind - ing sites . Scission of toxin - containing tubules may also depend on cooperative actions of curvature generating pro - teins and microtubule - based motors ( 53 ) . Under physio - logical conditions , STxB and CTxB could also potentially regulate cellular machinery that controls membrane cur - vature , either by signaling - based mechanisms ( 67 ) or by directly influencing microtubule dynamics ( 68 ) . Still , the cross - linking of membrane glycosphingolipids by toxin is not absolutely required for sorting into tubules or internal - ization . Our results implicate other mechanisms of mem - brane tubulation in this pathway , one of which we propose is facilitated by microtubules and the microtubule motor dynein . Materials and Methods Cells and reagents COS - 7 and HeLa cells were acquired from ATCC ( Manassas , VA ) . Stable RFP - \u03b1 - tubulinexpressingHeLacellswereagiftfromPaulChang ( M . I . T . ) . HeLacellsstablyexpressingmultifunctionalGFP ( mfGFP ) - tagged74kDa dynein intermediate chain ( IC74 ) were kindly provided by Takashi Murayama ( Department of Pharmacology , Juntendo University School of Medicine , Tokyo , Japan ) . T84 cells were cultured as previously described ( 52 ) . COS - 7 cells were maintained in Dulbecco\u2019s modified Eagle medium ( DMEM ) containing 10 % fetal bovine serum ( Life Technologies ) at 37 \u2218 C and5 % CO 2 . HeLacellsweremaintainedinRoswellParkMemorialInsti - tute medium ( RPMI ) containing 10 % fetal bovine serum at 37 \u2218 C and 5 % CO 2 . Media for RFP - \u03b1 - tubulin HeLa cells additionally contained G418 ( Corning ) , andhygromycin ( 400 \u03bc g / mL ) ( Invitrogen ) wasaddedtomedia for the IC74 cells . Cells were plated on coverslips or into MatTek cham - bers ( MatTek Corporation ) 2 days prior to experiments . Transient trans - fections were performed 24 h prior to imaging using FuGENE 6 as per manufacturer instructions ( Roche Diagnostics ) . Alexa488 - CTxBandAlexa555 - CTxBwereobtainedfromInvitrogen . CTx chimerawasgeneratedaspreviouslydescribed ( 28 ) . MonovalentCTxwas madeessentiallyasdescribed ( 29 ) . Briefly , an E . coli expressionstraincon - taining three plasmids encoding native CTA , CTB - G33D ( non - binding mutant ) and C - terminally GS - H6 - tagged wt - CTB was induced with 0 . 0005 % L - arabinose and 400 \u03bc M IPTG and grown overnight at 30 \u2218 C . Each subunit is secreted to the periplasm where holotoxins with mixed CTB pentamers assemble from a random assortment of G33D mutant and GS - H6 - tagged wt CTB monomers with CTA . This mixture of assem - bled holotoxins and free B pentamers containing from 0 to 5 GS - H6 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 583 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Day et al . tagged B subunits ( native binding sites ) was purified from a cell extract by Talon affinity chromatography , and separated into individual species by three rounds of ion - exchange chromatography . Free pentamers were first removed by binding to a cationic resin ( HS20 ) ; the mixture of holo - toxins eluting in the unbound fraction was then bound to an HQ20 anion exchange column , and individual species were eluted with a 0 \u2013 1M NaCl gradient , peak fractions were pooled , concentrated and repuri - fied by anion exchange . Densitometry of protein bands separated by SDS - PAGEshowedtheexpectedratioof1Aand1taggedwtBsubunitto4 native - sized ( G33D ) B subunits . Purified holotoxins were stored at 4 \u2218 C in 50mM Tris \u2013 HCl , pH8 . 0 . Before labeling , several preparations of single binding site holotoxin were pooled , concentrated and buffer - exchanged into PBS , pH7 . 5 by ultrafiltration using an Amicon Ultra - 4 ( Millipore ) 10K cutoff centrifugal filter . Alexa568 - labeled monovalent CTx was prepared by reacting Alexa568 - succinimidyl ester ( Invitrogen ) to 300 \u03bc g toxin chimera ( 29 ) in 100mM sodium bicarbonate buffer pH8 . 3 for 1 h under stirring at room temperature and purified using provided size exclusion chro - matography resin . Alexa568 - labeled wild type CTxB was prepared as above using recombinant B - subunit purified from periplasmic E . coli extracts . Alexa 488 - STxB and Cy3 - STxB along with a plasmid for Gb 3 transferase ( 69 , 70 ) were gifts from Ludger Johannes ( Institut Curie ) . Plasmids encoding EGFP - HRas , EGFP - HRas - tail ( consisting of GFP fused to the C - terminal 10 amino acid residues of HRas ) , and GFP - Fyn were provided by Mark Philips ( NYU School of Medicine ) ( 71 ) ; GFP - EB3 ( 40 ) was a gift from Anna Akhmanova ( Utrecht University ) ; and CC1 - dsRed and GFP - p50 ( 72 ) were gifts from Trina Schroer ( Johns Hopkins ) . LYFPGT46 ( referred to as YFP - GT46 ) was as previously described ( 73 , 74 ) . MyrPalm - mCFP was obtained from Roger Tsien ( 75 ) . A plasmid encoding mCherry - LAMP1 was generated by standard tech - niquesbyfusingthemammalianexpressionplasmidformCherrytoDNA encoding rat lysosomal membrane glycoprotein 1 ( LAMP1 ; NM _ 012857 ; gift from George Patterson , NIH ) . A rabbit anti - caveolin - 1 antibody was obtained from BD Biosciences . A mouse anti - transferrin receptor anti - body was purchased from Life Technologies . A rabbit anti - GFP antibody was obtained from Abcam . An anti - Myc ( 9B11 ) mouse monoclonal anti - bodywas from Cell Signaling . Fluorescently labeled secondaryantibodies were from Jackson ImmunoResearch and Life Technologies . D ( + ) glucose , sodium azide , 2 - deoxyglucose , bovine serum albumin ( BSA ) , trichloroacetic acid , Nocodazole ( NZ ) , Dynasore , Latrunculin A , methyl \u03b2 - cyclodextrin and ciliobrevin A ( HPI - 4 ) were purchased from Sigma Aldrich . Taxol was from Alexis Biochemical . HEPES and TAE buffers were purchased from Mediatech , Inc . . Alexa546 - dextran was from Life Technologies . Jasplakinolide and Alexa647 - transferrin were obtained from Invitrogen . Confocal microscopy Confocal microscopy was carried out on a Zeiss LSM 510 confo - cal microscope ( Carl Zeiss MicroImaging , Inc . ) using a 40X 1 . 4 NA Zeiss Plan - Neofluar oil immersion objective or 100X 1 . 4 NA Zeiss Plan - Apochromat oil immersion objective . Images were collected using 1 Airy unit confocal slices unless otherwise indicated . Cells were maintained in media supplemented with 25mM HEPES and any indicated drugs as described above for live - cell imaging experiments . Cells were maintained at 37 \u2218 C using a stage heater and objective heater during imaging . EGFP and Alexa488 were excited using the 488nm line of a 40mW Argon laser . YFP was excited using the 514nm line of the Argon laser . Alexa546 , Alexa555 , Alexa568 , Cy3 , RFP and mCherry were excited at 543nm using a HeNe laser . Alexa647 was excited at 647nm using a HeNe laser . Fluorescence emission was detected using filter sets provided by the manufacturer . For presentation purposes , images were exported in tiff format and brightness and contrast were adjusted using ImageJ or Fiji ( 76 ) . ATP depletion ATP depletion was performed by pre - incubating cells at 37 \u2218 C and 5 % CO 2 for 15min in ATP depletion medium , composed of glucose - free DMEM containing 50mM 2 - deoxy - D - glucose , 0 . 02 % sodium azide , 25mM HEPES , and 1mg / mL BSA as described in ( 77 ) . Control cells were incubated in ATP control medium ( composed of glucose - free DMEM supplemented with 50mM D - ( + ) - glucose , 25mM HEPES , and 1mg / mL BSA ) . Cells were rinsed twice , incubated for 5 min at room temperature with CTxB ( 100 or 500 nM ) , STxB ( 99 nM ) , CTx chimera ( 500 nM ) or monovalent CTx ( 400 or 800 nM ) , rinsed twice , then imaged live at 37 \u2218 C in either ATP depletion or control media . ATP depletion also induces the formation of abundant actin - rich membrane protrusions ( 78 ) , so ATP - depleted cells were always imaged in 3 - dimensions by confocal microscopy to definitively identify invaginations ( Figure S1 ) . For some experiments , cells were pretreated with Jasplakinolide ( 250 nM ) for 30min prior to ATP depletion , then imaged in the continued presence of both Jasplakinolide and ATP depletion medium . Quanti\ufb01cation of cellular ATP levels ATPdepletionwasverifiedtodecreaseATPlevelsto < 5 % ofcontrolvalues using the commercially available ENLITEN \u00ae ATP assay kit ( Promega ) . For this assay , COS - 7 cells were split into 12 well plates . After 2 days the medium was removed and some cells were incubated directly in 500 \u03bc L ATP - extraction solution [ 1 % TCA in Tris - Acetate - EDTA ( TAE ) buffer ] to collect baseline ATP readings . The remaining cells were rinsed twice with ATP depletion media and incubated at 37 \u2218 C and 5 % CO 2 . At the indicated times after the initiation of ATP depletion , the depletion media was replaced with 500 \u03bc L ATP - extraction solution . Cells were incubated in ATP - extraction solution for 30 min at RT , as described before ( 79 ) . Aliquots of the cell extract were then moved to 96 well plates and diluted tenfoldinTAEbuffer . Reactionreagentcontainingluciferasewasaddedto each well and the chemiluminescence was read on a Synergy H4 Hybrid Multi - Mode Microplate Reader ( BioTek ) . Actin disruption To disrupt actin , cells were first washed with imaging buffer , incubated for 5 min in 1 \u03bc M Alexa546 - CTxB in imaging buffer , and washed again . 584 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Microtubules Tubulate the Plasma Membrane Actin depolymerization was then performed by incubating the cells at 37 \u2218 C for 5min in imaging buffer containing 1 \u03bc M Latrunculin A . Control cells wereincubated inimagingbuffercontaining0 . 1 % DMSO . Cells were maintained in their respective buffer during imaging and all imaging was performed within 30 min of treatment . Where indicated , cells were preincubated with 5 \u03bc g / mL NZ for 15min on ice , followed by a 1 h incubation at 37 \u2218 C prior to Latrunculin A treatment and CTxB labeling , thenimagedliveinthecontinuedpresenceofbothLatrunculinAandNZ . Actin stabilization Cell culture media was replaced with complete imaging buffer containing 250 nM Jasplakinolide or DMSO and incubated at 37 \u2218 C and 5 % CO 2 for 30 min . Cells were then labeled with 100 nM Alexa - labeled CTxB for 5 min at RT in complete imaging buffer containing 250 nM Jasplakinolide or DMSO . Cells were rinsed and imaged in complete imaging buffer containing 250 nM Jasplakinolide or DMSO . Where indicated , cells were ATP depleted following Jasplakinolide treatment as described above . In some experiments , cells were preincubated with 5 \u03bc g / mL NZ for 15min on ice followed by a 1 h incubation at 37 \u2218 C prior to Jasplakinolide treatment and CTxB labeling , then imaged live in the continued presence of both Jasplakinolide and NZ . Dynamin inhibition Dynamin 2 was inhibited using the small molecule inhibitor , Dynasore ( 80 ) . Cell culture media was replaced with DMEM supplemented with 25 \u03bc MHEPESandeither80 \u03bc MDynasoreorDMSO . Cellswereincubated in Dynasore or DMSO at 37 \u2218 C and 5 % CO 2 for 30 min . Cells were then labeled with either 100 nM Alexa568 - CTxB or 400 \u2013 800 nM Alexa568 monovalent CTx for 5 min at RT , rinsed and imaged in DMEM supple - mented with 25 \u03bc M HEPES and either 80 \u03bc M Dynasore or DMSO . Where indicated , cells were preincubated with 5 \u03bc g / mL NZ for 15min on ice fol - lowed by a 1 h incubation at 37 \u2218 C prior to Dynasore treatment and CTxB or monovalent CTx labeling , then imaged live in the continued presence of both Dynasore and NZ . In other experiments , cells were transfected with GFP or GFP - p50 prior to Dynasore treatment and toxin labeling as described above , then imaged live . Analysis of surface accessibility of tubules ATPdepletedCOS - 7werelabeledwith25nMAlexa555 - CTxBandplaced on the stage at 37 \u2218 C . A highly concentrated dose of Alexa488 - CTxB was added to bring the final Alexa488 - CTxB concentration in the imaging buffer to 50 nM while continuously imaging . Immuno\ufb02uorescence labeling For immunostaining of caveolin - 1 and transferrin receptor , cells grown in MatTek dishes were subjected to ATP depletion and labeled with CTxB as described above . After CTxB labeling they were incubated for 30min 37 \u2218 C and 5 % CO 2 in the continued presence of the ATP depletion media . Theywerethenfixedat37 \u2218 Cwith4 % PFA / 0 . 2 % glutaraldehydefor 15min . After rinsing several times they were blocked in PBS containing 10 % FBS and 0 . 1 % saponin for 15min and then labeled with transferrin receptor or caveolin - 1 antibodies for 30min . They were again washed and then labeled with secondary antibody for 30min . After several additional washes they were imaged . In IC74 cells , immunolabeling of the tagged dynein intermediate chain subunit was performed using an anti - myc antibody in order to amplify the fluorescence signal . For these experiments , cells grown on coverslips were either ATP depleted for 15min or left untreated . They were then rinsed twice and incubated for 5 min at room temperature with 100 nM Alexa488 - CTxB or Alexa555 - CTxB . The cells were then rinsed twice and fixedat37 \u2218 Cfor15mininpre - warmed4 % PFA / 0 . 2 % glutaraldehyde . Post fixation , the cells were quenched by three rinses in 100mM glycine in PBS . Permeabilization and blocking for 60min at RT was performed in blocking buffer composed of 0 . 1 % TX - 100 in PBS containing 5 % glycine and 5 % normal goat or donkey serum . Cells were incubated with rabbit anti - myc antibody for 2h at RT . After rinsing in PBS coverslips were incubatedfor1hina1 : 200dilutionoffluorescently - conjugatedsecondary antibodies , andmountedusingProLongGold ( Invitrogen , Carlsbad , CA ) . Microtubule stabilization HeLa cells expressing RFP - \u03b1 - tubulin were incubated with imaging buffer containing 1 \u03bc M taxol for 4 h at 37 \u2218 C . They were then rinsed twice with ATP depletion media containing 1 \u03bc M taxol , incubated at 37 \u2218 C for 15 min , labeled with 100 nM Alexa - labeled CTxB for 5 min at RT , rinsed and imaged in ATP depletion media containing 1 \u03bc M taxol . Suppression of microtubule plus end dynamics Cells were incubated with imaging buffer containing 150 nM NZ ( 40 \u2013 42 ) or DMSO at 37 \u2218 C for 5 min . The media was then replaced with ATP depletion media supplemented with 150 nM NZ or DMSO and incubated at 37 \u2218 C for 15 min . Cells were labeled with 100 nM Alexa - labeled CTxB for 5 min at RT , rinsed and imaged in ATP depletion media containing 150 nM NZ or DMSO . Microtubule disruption Todisruptmicrotubules , cellswereincubatedwithimagingbuffer ( phenol red - free DMEM , 10 % BSA , 25mM HEPES , and 1mg / mL BSA ) contain - ing 5 \u03bc g / mL ( 16 . 7 \u03bc M ) NZ ( 81 , 82 ) or DMSO on ice for 15 min . Cells were then shifted to 37 \u2218 C for 1 h and subjected to further drug treatments and CTx or CTxB labeling as indicated above . Cells were imaged in the continued presence of either NZ or DMSO . The efficacy of microtubule disruptionwasconfirmedbyvisualizingthedistributionoftubulinincon - trol experiments . Analysis of \ufb02uid phase and transferrin uptake in ATP - depleted cells Control experiments were carried out to verify the efficacy of ATP deple - tionbytestingitseffectsonfluidphaseandtransferrinuptake ( FigureS1 ) . For the fluid phase uptake experiments , COS - 7 cells were preincubated in ATP control media or ATP depletion media for 15 min . They were then labeled with 100 nM Alexa488 - CTxB ( to mark the position of cells ) and Traf\ufb01c 2015 ; 16 : 572 \u2013 590 585 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Day et al . Alexa546 - Dextran ( 1mg / mL ) for 20 min at 37 \u2218 C , rinsed 10 times with the respective media and imaged live . To measure transferrin uptake in ATP - depleted cells , cells were serum starved for 1h prior to incubation in ATP control media or ATP depletion media for 15min . They were then labeled with 100 nM Alexa488 - CTxB ( to mark the position of cells ) and 25 \u03bc g / mL Alexa647 transferrin for 20 min at 37 \u2218 C , rinsed 5 times with respective media , and imaged live . Inhibition of dynein motor activity In control experiments , we defined conditions in which ciliobrevin A inhibited the dynein - based delivery of CTxB to the perinuclear region in cells with physiological levels of ATP ( Figure S3C , D ) . To generate ciliobrevinAdosedependencecurves , ciliobrevinAwasdilutedinDMSO and stocks were produced by serial dilution in DMEM + 10 % fetal bovine serum . Cells were incubated in medium containing ciliobrevin A or DMSOat37 \u2218 Cfor1h . Theywerethenlabeledwith100nMAlexa - labeled CTxB in ciliobrevin A or DMSO containing media for 5 min at RT , rinsed and either fixed immediately or shifted to 37 \u2218 C and 5 % CO 2 for 30 min priortofixationwith3 . 4 % PFAatRTfor15min . Samplesweremountedin ProLong Gold ( Invitrogen , Carlsbad , CA ) and fields of cells were imaged and image analysis performed as indicated below . For experiments examining the effects of ciliobrevin A on tubule for - mation , cell culture media was replaced with imaging buffer containing 500 \u03bc M ciliobrevin A or DMSO and incubated at 37 \u2218 C and 5 % CO 2 for 1 h . Cells were rinsed twice with ATP depletion or control media con - taining ciliobrevin A or DMSO and incubated at 37 \u2218 C and 5 % CO 2 for 15 min . Cells were labeled with 100 nM Alexa - labeled CTxB for 5 min at RT , rinsed and imaged in ATP depletion or control media containing cil - iobrevin A or DMSO . Care was taken to protect ciliobrevin - A from light throughout experiment . Analysis of \ufb02uid phase uptake in cells expressing GFP - p50 COS - 7 cells transiently expressing EGFP or EGFP - p50 were loaded with Alexa - 546 dextran ( 1mg / mL ) in 100 \u03bc L of serum free imaging buffer and incubated at 37 \u2218 C for 20 min , rinsed with imaging buffer 10 times and imagedlive . Imageswerecollectedusinga1Airyunitconfocalsliceunder identical imaging conditions for cells expressing EGFP and EGFP - p50 . Analysis of transferrin and CTxB uptake in cells expressing GFP - p50 COS - 7cellsplatedoncoverslipsweretransfectedwithEGFPorGFP - p50 . To quantify the cellular uptake of CTxB , cells were rinsed with cold imaging buffer and labeled with 100 nM A555 - CTxB for 5 min on ice . For quantification of transferrin uptake , cells were serum starved using DMEM containing 25mM HEPES for 1 h at 37 \u2218 C and 5 % CO 2 . Serum starved cells were then rinsed with cold DMEM supplemented with 10 % FBS and 25mM HEPES and labeled with 5 \u03bc g / mL Alexa568 - transferrin for 5 min on ice . Cells were rinsed with imaging buffer and shifted to 37 \u2218 C and 5 % CO 2 for 20 min . Acid stripping was performed on ice by incubating cells with 2mL of 100mM glycine , pH2 . 0 , for 5 min and then with HBSS , pH7 . 4 , for 5 min . Acid - stripped samples were subsequently incubated with 37 \u2218 C HBSS , pH7 . 4 , for 10 s to promote release of remaining surface - bound toxin or transferrin ( 83 ) . This process was repeated 3 times . Cells were fixed in 3 . 4 % PFA at room temperature for 15 min and rinsed with 1x PBS . Cells were labeled with an anti - GFP primary antibody and an Alexa488 secondary antibody . Samples were mounted in ProLong Gold . Cells were imaged at 1x zoom and multiple fields were collected using a 1 Airy unit confocal slice for CTxB and a 2 Airy unit slice for transferrin . Experiments were performed at least three times . Analysis of \ufb02uid phase uptake in NZ - treated cells COS - 7 cells were transiently transfected with EGFP ( to mark the position ofcells ) thedaybeforetheexperiment . Thenextday , cellswerepre - chilled on ice in serum free imaging buffer for 5min , then incubated with serum free imaging buffer containing 5 \u03bc g / mL ( 16 . 7 \u03bc M ) NZ or DMSO on ice for 15 min . Cells were then shifted to 37 \u2218 C for 1 h . They were then labeled with Alexa546 dextran ( 1mg / mL ) for 20 min at 37 \u2218 C in the continued presence of NZ or DMSO , rinsed 10 times with the respective media and imagedlive . Imageswerecollectedusinga1Airyunitsliceunderidentical imaging conditions for the NZ and DMSO - treated samples . Analysis of transferrin and CTxB uptake in NZ - treated cells COS - 7 cells were transiently transfected with EGFP ( to mark the position ofcells ) thedaybeforetheexperiment . Thenextday , cellswerepre - chilled on ice in serum free imaging buffer for 5min , then incubated with serum free imaging buffer containing 5 \u03bc g / mL ( 16 . 7 \u03bc M ) NZ or DMSO on ice for 15 min . Cells were then shifted to 37 \u2218 C for 1 h . Cells were then labeled with either CTxB or transferrin exactly as described above except that NZ or DMSO were included in all labeling and washing steps . Acid stripping and further processing of cells was performed exactly as described above . Electrophysiology Measurements of short circuit currents ( Isc ) and resistance ( R ) were performed on confluent monolayers of human intestinal T84 cells grown on 0 . 33cm 2 filters , as previously described ( 52 ) . Monolayers were treated with 5 \u03bc g / mL nocodazole or DMSO for 30min on ice followed by further incubation for 30min at 37 \u2218 C prior to addition of 20 nM cholera toxin . Image analysis To quantify the percentage of a population of ATP - depleted cells con - taining invaginations , full field ( 512 \u00d7 512 ) z - stacks of fields of cells were taken at 1 . 7 \u00d7 zoom on a 40 \u00d7 objective , with line averaging of 4 and opti - mal overlay in z - direction . z - Stacks were collected consecutively for 50 min after labeling with toxin . Cells were then scored by hand as display - ing or not displaying invaginations . For cells expressing a plasmid , that is CC1 - dsRed , the cells were scored in the CTxB channel with the exper - imentalist blind to which cells were expressing the plasmid and which were not . 586 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Microtubules Tubulate the Plasma Membrane To determine the number and length of invaginations z - sections were taken of individual ATP - depleted cells consecutively for 50 min after labeling with toxin . The JFilament plugin ( Dimitrios Vavylonis and XiaoleiHuang ; LehighUniversity ) forImageJwasusedtotraceindividual invaginations in 2D . The automated snake tracking feature was used to align the snakes with the invaginations . Tracings were examined and cor - rected by hand . Kymographs were produced using the MultipleKymo - graph plugin for ImageJ . To quantify the effects of ATP depletion , NZ treatment and GFP - p50 expressionontheuptakeofendocyticcargo , outlinesofcellsweremadein theGFPorCTxBchannelusingImageJ . Theaveragefluorescenceintensity in the endocytic cargo channel was then recorded from the same regions . Backgroundfluorescencewasmeasuredfromregionsdevoidofcells . After subtracting background , the data were normalized to 100 % of control values . For lysosomal tracking experiments , particle detection was per - formed using a custom - written MATLAB algorithm ( available upon request ) and tracking was performed using u - track 2 . 0 software ( http : / / lccb . hms . harvard . edu / software . html ) ( 84 ) . Both programs were run using MATLAB 7 . 11 . 0 ( Mathworks ) . The ciliobrevin A dose dependence curve was generated by collecting full field ( 512 \u00d7 512 ) 16 bit images using a 40x objective with line averaging of 8 . ROIs were then drawn around the perinuclear space , flat membraneregionsandbackgroundinImageJ . Perinuclearandmembrane per pixel fluorescence were corrected for background fluorescence and then divided to produce a perinuclear to membrane ratio per cell . Statistical analysis All experiments were performed at least twice and most were carried out three or more times . Chi - square tests were performed using Excel ( Microsoft ) , and Student t - tests were performed using Excel or OriginPro 8 . 6 ( OriginLab ) . Acknowledgments We thank Dana Hardbower , Carol Landerman and Bradley Clarke for assistance with experiments ; Drs . Ludger Johannes , Ramiro Massol , Irina Kaverina , Julie Donaldson , Mark McNiven , Stephen King and Puck Ohi for stimulating discussions ; Dr . Aurelio Galli for comments on the manuscript ; Jacob Dowler for assistance with figure preparation ; and all those who contributed reagents . Supported by NIH R01 GM106720 ( AKK ) , NIH RO1 GM073846 ( AKK ) , R01 HL111259 ( AKK ) , NIH RO1 AI31940 ( RKH ) , NIH R01 GM44589 ( TAS ) , NIH grants DK48106 , DK084424 and DK090603 and the Harvard Digestive Diseases Cen - ter P30 DK34854 ( WIL ) , the Vanderbilt Molecular Biophysics training grant ( NIH T32 GM08320 ) ( CAD ) , a Vanderbilt Discovery Grant ( Dr . Todd Graham and AKK ) , and NIH grant R44 - EB008589 ( to J . Dami - ano ) . It also utilized the core ( s ) of the Vanderbilt Diabetes Research and Training Center funded by grant DK020593 from the National Insti - tute of Diabetes and Digestive and Kidney Disease . MWD was sup - ported by the National High Magnetic Field Laboratory , which is sup - ported by NSF DMR - 1157490 and the State of Florida . The funding sources had no role in the study design , collection , analysis or interpre - tation of data , writing the report or the decision to submit the paper for publication . Supporting Information Additional Supporting Information may be found in the online version of this article : Figure S1 : Control experiments characterizing the properties of the tubular invaginations and documenting the efficacy of ATP depletion . ATP depletion impairs clathrin - dependent and - independent endocyto - sis , resultingintheaccumulationofSTxBandCTxBinplasmamembrane tubular invaginations . A ) Alexa488 - STxB and Alexa488 - CTxB accumu - late in tubular invaginations in ATP depleted COS - 7 cells . To facilitate STxB labeling , cells were transfected with Gb 3 synthase . B and C ) As reported previously ( 78 ) , fluorescent CTxB labels both plasma mem - brane invaginations ( arrowheads ) and plasma membrane protrusions ( arrows ) in ATP - depleted cells . B and C are single frames from a con - focal z - stack . Dashes mark the position of the xz - sections shown below . D ) Alexa488 - STxB and Alexa555 - CTxB colocalize in invaginations in ATPdepletedCOS - 7cellsexpressingGb 3 synthase . E ) Tubules prelabeled with Alexa555 - CTxB filled at approximately the same rateas newly added Alexa488 - CTxBaccumulatesattheplasmamembrane , indicatingthatthe tubulesareopentothefluidphaseinATP - depletedcells . Timestampsare in minutes : seconds . F \u2013 H ) ATP depletion inhibits uptake of dextran and internalization of transferrin ( Tfn ) ( Mean \u00b1 SD , N = 325 \u2013 417 cells ) . * * * , p < 0 . 001 , Student\u2019s t - test . Bars , 10 \u03bc m . ( Related to Figure 1 . ) Figure S2 : Plasma membrane derived tubules are selective for mem - brane markers . A and B ) CFP - Myr - palm and GFP - Fyn accumulate in tubular invaginations in either the ( A ) presence or ( B ) absence of CTxB in ATP - depleted cells , whereas YFP - GT46 is excluded from tubules . C and D ) Endogenous transferrin receptor ( C ) and caveolin - 1 ( D ) typically did not colocalize with tubules in immunostained ATP - depleted cells , although occasionally caveolin - 1 staining was seen at the end of a tubule ( arrowhead ) . ( Related to Figure 2 ) Figure S3 : Invaginations present following Lat A , Jasplak , or Dyna - sore treatment are dependent on microtubules and dynein . A and B ) Pretreatment of cells with 16 . 6 \u03bc M nocodazole prior to LatA ( A ) or Jasplakinolide treatment ( B ) blocks tubule formation . C and D ) Pretreatment with ciliobrevin A ( Cilio - A ) disrupts delivery of CTxB to perinuclear compartments in ATP replete cells in a dose - dependent manner . n = 36 \u2013 108 cells ; error bars = SD . E and F ) Expression of GFP - p50 ( E ) , but not GFP alone ( F ) blocked the formation of long branched tubules in Dynasore - treated cells . Similar results were obtained for both wild type CTxB or monovalent CTx . Bars , 10 \u03bc m . ( Related to Figures 4 and 6 ) Figure S4 : Some dynein is associated with the tubular invaginations . A ) Distribution of mfGFP - dynein 74kDa intermediate chain in a stably expressing HeLa cell line . Cells were fixed and immunostained using Traf\ufb01c 2015 ; 16 : 572 \u2013 590 587 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Day et al . a myc antibody to enhance the fluorescence signal . B ) Following ATP depletion , dynein 74kDa intermediate chain staining is apparent at the plasma membrane ( arrowheads ) . C ) mfGFP - IC74 expressing cells were ATP depleted , labeled with CTxB , fixed and immunostained for tagged dynein intermediate chain . D ) Zoom of boxed region of cell shown in C . Some mfGFP - IC74 - positive puncta align along CTxB - containing tubular invaginations . Bars , 5 \u03bc m . ( Related to Figure 5 ) MovieS1 : Dynamics of growth of CTxB - positive tubular invaginations in ATP depleted COS - 7 cells . Correspond to cells shown in Figure 3 . Time stamps are in minutes : seconds . Bar , 10 \u03bc m . ( Related to Figure 3 ) . MovieS2 : Dynamics of growth of CTxB - positive tubular invaginations in ATP depleted COS - 7 cells . Correspond to cells shown in Figure 3 . Time stamps are in minutes : seconds . Bar , 10 \u03bc m . ( Related to Figure 3 ) . Movie S3 : EB3 - GFP is not enriched at microtubule plus ends in ATP - depleted cells . Time stamps are in minutes : seconds . Bar , 10 \u03bc m . ( Related to Figure 4 ) . Movie S4 : ATP depletion attenuates , but does not completely eliminate the directed motions of mCherry - LAMP - 1 positive structures compared to control conditions . Corresponds to cells shown in Figure 4 . Time stamps are in minutes : seconds . Bar , 10 \u03bc m . ( Related to Figure 5 ) . References 1 . McMahon HT , Boucrot E . Molecular mechanism and physiological functions of clathrin - mediated endocytosis . Nat Rev Mol Cell Biol 2011 ; 12 : 517 \u2013 533 . 2 . Sandvig K , Pust S , Skotland T , van Deurs B . Clathrin - independent endocytosis : mechanisms and function . Curr Opin Cell Biol 2011 ; 23 : 413 \u2013 420 . 3 . Doherty GJ , McMahon HT . Mechanisms of endocytosis . Annu Rev Biochem 2009 ; 78 : 857 \u2013 902 . 4 . Howes MT , Mayor S , Parton RG . Molecules , mechanisms , and cellular roles of clathrin - independent endocytosis . Curr Opin Cell Biol 2010 ; 22 : 519 \u2013 527 . 5 . Mayor S , Pagano RE . Pathways of clathrin - independent endocytosis . Nat Rev Mol Cell Biol 2007 ; 8 : 603 \u2013 612 . 6 . Kumari S , Mg S , Mayor S . Endocytosis unplugged : multiple ways to enter the cell . Cell Res 2010 ; 20 : 256 \u2013 275 . 7 . Hansen CG , Nichols BJ . Molecular mechanisms of clathrin - independent endocytosis . J Cell Sci 2009 ; 122 : 1713 \u2013 1721 . 8 . Donaldson JG , Porat - Shliom N , Cohen LA . Clathrin - independent endocytosis : a unique platform for cell signaling and PM remodeling . Cell Signal 2009 ; 21 : 1 \u2013 6 . 9 . Mayor S , Parton RG , Donaldson JG . Clathrin - independent pathways of endocytosis . Cold Spring Harb Perspect Biol 2014 ; 6 : a016758 . 10 . Pang H , Le PU , Nabi IR . Ganglioside GM1 levels are a determinant of the extent of caveolae / raft - dependent endocytosis of cholera toxin to the Golgi apparatus . J Cell Sci 2004 ; 117 : 1421 \u2013 1430 . 11 . Parton RG . Ultrastructural localization of gangliosides ; GM1 is concentrated in caveolae . J Histochem Cytochem 1994 ; 42 : 155 \u2013 166 . 12 . Torgersen ML , Skretting G , van Deurs B , Sandvig K . Internalization of cholera toxin by different endocytic mechanisms . J Cell Sci 2001 ; 114 : 3737 \u2013 3747 . 13 . Massol RH , Larsen JE , Fujinaga Y , Lencer WI , Kirchhausen T . Cholera toxin toxicity does not require functional Arf6 - and dynamin - dependent endocytic pathways . Mol Biol Cell 2004 ; 15 : 3631 \u2013 3641 . 14 . Glebov OO , Bright NA , Nichols BJ . Flotillin - 1 de\ufb01nes a clathrin - independent endocytic pathway in mammalian cells . Nat Cell Biol 2006 ; 8 : 46 \u2013 54 . 15 . Lundmark R , Doherty GJ , Howes MT , Cortese K , Vallis Y , Parton RG , McMahon HT . The GTPase - activating protein GRAF1 regulates the CLIC / GEEC endocytic pathway . Curr Biol 2008 ; 18 : 1802 \u2013 1808 . 16 . Howes MT , Kirkham M , Riches J , Cortese K , Walser PJ , Simpson F , Hill MM , Jones A , Lundmark R , Lindsay MR , Hernandez - Deviez DJ , Hadzic G , McCluskey A , Bashir R , Liu L , et al . Clathrin - independent carriers form a high capacity endocytic sorting system at the leading edge of migrating cells . J Cell Biol 2010 ; 190 : 675 \u2013 691 . 17 . Kirkham M , Fujita A , Chadda R , Nixon SJ , Kurzchalia TV , Sharma DK , Pagano RE , Hancock JF , Mayor S , Parton RG . Ultrastructural identi\ufb01cation of uncoated caveolin - independent early endocytic vehicles . J Cell Biol 2005 ; 168 : 465 \u2013 476 . 18 . R\u00f6mer W , Berland L , Chambon V , Gaus K , Windschiegl B , Tenza D , Aly MR , Fraisier V , Florent JC , Perrais D , Lamaze C , Raposo G , Steinem C , Sens P , Bassereau P , et al . Shiga toxin induces tubular membrane invaginations for its uptake into cells . Nature 2007 ; 450 : 670 \u2013 675 . 19 . Chinnapen DJ , Chinnapen H , Saslowsky D , Lencer WI . Rafting with cholera toxin : endocytosis and traf\ufb01cking from plasma membrane to ER . FEMS Microbiol Lett 2007 ; 266 : 129 \u2013 137 . 20 . Badizadegan K , Wolf AA , Rodighiero C , Jobling M , Hirst TR , Holmes RK , Lencer WI . Floating cholera toxin into epithelial cells : functional association with caveolae - like detergent - insoluble membrane microdomains . Int J Med Microbiol 2000 ; 290 : 403 \u2013 408 . 21 . R\u00f6mer W , Pontani LL , Sorre B , Rentero C , Berland L , Chambon V , Lamaze C , Bassereau P , Sykes C , Gaus K , Johannes L . Actin dynamics drive membrane reorganization and scission in clathrin - independent endocytosis . Cell 2010 ; 140 : 540 \u2013 553 . 22 . Nichols BJ , Kenworthy AK , Polishchuk RS , Lodge R , Roberts TH , Hirschberg K , Phair RD , Lippincott - Schwartz J . Rapid cycling of lipid raft markers between the cell surface and Golgi complex . J Cell Biol 2001 ; 153 : 529 \u2013 541 . 23 . Nichols BJ . A distinct class of endosome mediates clathrin - independent endocytosis to the Golgi complex . Nat Cell Biol 2002 ; 4 : 374 \u2013 378 . 24 . Hansen CG , Bright NA , Howard G , Nichols BJ . SDPR induces membrane curvature and functions in the formation of caveolae . Nat Cell Biol 2009 ; 11 : 807 \u2013 814 . 25 . de Kreuk BJ , Nethe M , Fernandez - Borja M , Anthony EC , Hensbergen PJ , Deelder AM , Plomann M , Hordijk PL . The F - BAR domain protein PACSIN2 associates with Rac1 and regulates cell spreading and migration . J Cell Sci 2011 ; 124 : 2375 \u2013 2388 . 588 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Microtubules Tubulate the Plasma Membrane 26 . Pina DG , Johannes L . Cholera and Shiga toxin B - subunits : thermodynamic and structural considerations for function and biomedical applications . Toxicon 2005 ; 45 : 389 \u2013 393 . 27 . Wernick NL , Chinnapen DJ , Cho JA , Lencer WI . Cholera toxin : an intracellular journey into the cytosol by way of the endoplasmic reticulum . Toxins ( Basel ) 2010 ; 2 : 310 \u2013 325 . 28 . Wolf AA , Jobling MG , Saslowsky DE , Kern E , Drake KR , Kenworthy AK , Holmes RK , Lencer WI . Attenuated endocytosis and toxicity of a mutant cholera toxin with decreased ability to cluster GM1 . Infect Immun 2008 ; 76 : 1476 \u2013 1484 . 29 . Jobling MG , Yang Z , Kam WR , Lencer WI , Holmes RK . A single native ganglioside GM1 - binding site is suf\ufb01cient for cholera toxin to bind to cells and complete the intoxication pathway . MBio 2012 ; 3 : e00401 \u2013 e00412 . 30 . Ewers H , R\u00f6mer W , Smith AE , Bacia K , Dmitrieff S , Chai W , Mancini R , Kartenbeck J , Chambon V , Berland L , Oppenheim A , Schwarzmann G , Feizi T , Schwille P , Sens P , et al . GM1 structure determines SV40 - induced membrane invagination and infection . Nat Cell Biol 2010 ; 12 : 11 \u2013 18 ; sup pp 11 \u2013 12 . 31 . Johannes L , Mayor S . Induced domain formation in endocytic invagination , lipid sorting , and scission . Cell 2010 ; 142 : 507 \u2013 510 . 32 . Anitei M , Ho\ufb02ack B . Bridging membrane and cytoskeleton dynamics in the secretory and endocytic pathways . Nat Cell Biol 2012 ; 14 : 11 \u2013 19 . 33 . Stephens DJ . Functional coupling of microtubules to membranes - implications for membrane structure and dynamics . J Cell Sci 2012 ; 125 : 2795 \u2013 2804 . 34 . Leduc C , Campas O , Joanny JF , Prost J , Bassereau P . Mechanism of membrane nanotube formation by molecular motors . Biochim Biophys Acta 2010 ; 1798 : 1418 \u2013 1426 . 35 . Vaughan KT . Sur\ufb01ng , regulating and capturing : are all microtubule - tip - tracking proteins created equal ? Trends Cell Biol 2004 ; 14 : 491 \u2013 496 . 36 . Mimori - Kiyosue Y , Tsukita S . \u201cSearch - and - capture\u201d of microtubules through plus - end - binding proteins ( + TIPs ) . J Biochem 2003 ; 134 : 321 \u2013 326 . 37 . Vaughan PS , Miura P , Henderson M , Byrne B , Vaughan KT . A role for regulated binding of p150 ( Glued ) to microtubule plus ends in organelle transport . J Cell Biol 2002 ; 158 : 305 \u2013 319 . 38 . Lomakin AJ , Semenova I , Zaliapin I , Kraikivski P , Nadezhdina E , Slepchenko BM , Akhmanova A , Rodionov V . CLIP - 170 - dependent capture of membrane organelles by microtubules initiates minus - end directed transport . Dev Cell 2009 ; 17 : 323 \u2013 333 . 39 . Waterman - Storer CM , Salmon ED . Endoplasmic reticulum membrane tubules are distributed by microtubules in living cells using three distinct mechanisms . Curr Biol 1998 ; 8 : 798 \u2013 806 . 40 . Stepanova T , Slemmer J , Hoogenraad CC , Lansbergen G , Dortland B , De Zeeuw CI , Grosveld F , van Cappellen G , Akhmanova A , Galjart N . Visualization of microtubule growth in cultured neurons via the use of EB3 - GFP ( end - binding protein 3 - green \ufb02uorescent protein ) . J Neurosci 2003 ; 23 : 2655 \u2013 2664 . 41 . Jordan MA , Thrower D , Wilson L . Effects of vinblastine , podophyllotoxin and nocodazole on mitotic spindles . Implications for the role of microtubule dynamics in mitosis . J Cell Sci 1992 ; 102 : 401 \u2013 416 . 42 . Perez F , Diamantopoulos GS , Stalder R , Kreis TE . CLIP - 170 highlights growing microtubule ends in vivo . Cell 1999 ; 96 : 517 \u2013 527 . 43 . Hunt SD , Stephens DJ . The role of motor proteins in endosomal sorting . Biochem Soc Trans 2011 ; 39 : 1179 \u2013 1184 . 44 . Allan VJ . Cytoplasmic dynein . Biochem Soc Trans 2011 ; 39 : 1169 \u2013 1178 . 45 . Hirokawa N , Noda Y , Tanaka Y , Niwa S . Kinesin superfamily motor proteins and intracellular transport . Nat Rev Mol Cell Biol 2009 ; 10 : 682 \u2013 696 . 46 . Falcon - Perez JM , Nazarian R , Sabatti C , Dell\u2019Angelica EC . Distribution and dynamics of Lamp1 - containing endocytic organelles in \ufb01broblasts de\ufb01cient in BLOC - 3 . J Cell Sci 2005 ; 118 : 5243 \u2013 5255 . 47 . Toba S , Watanabe TM , Yamaguchi - Okimoto L , Toyoshima YY , Higuchi H . Overlapping hand - over - hand mechanism of single molecular motility of cytoplasmic dynein . Proc Natl Acad Sci U S A 2006 ; 103 : 5741 \u2013 5745 . 48 . Mallik R , Carter BC , Lex SA , King SJ , Gross SP . Cytoplasmic dynein functions as a gear in response to load . Nature 2004 ; 427 : 649 \u2013 652 . 49 . Kobayashi T , Murayama T . Cell cycle - dependent microtubule - based dynamic transport of cytoplasmic dynein in mammalian cells . PLoS One 2009 ; 4 : e7827 . 50 . Firestone AJ , Weinger JS , Maldonado M , Barlan K , Langston LD , O\u2019Donnell M , Gelfand VI , Kapoor TM , Chen JK . Small - molecule inhibitors of the AAA + ATPase motor cytoplasmic dynein . Nature 2012 ; 484 : 125 \u2013 129 . 51 . Schroer TA . Dynactin . Annu Rev Cell Dev Biol 2004 ; 20 : 759 \u2013 779 . 52 . Lencer WI , Constable C , Moe S , Rufo PA , Wolf A , Jobling MG , Ruston SP , Madara JL , Holmes RK , Hirst TR . Proteolytic activation of cholera toxin and Escherichia coli labile toxin by entry into host epithelial cells . Signal transduction by a protease - resistant toxin variant . J Biol Chem 1997 ; 272 : 15562 \u2013 15568 . 53 . Renard HF , Simunovic M , Lemiere J , Boucrot E , Garcia - Castillo MD , Arumugam S , Chambon V , Lamaze C , Wunder C , Kenworthy AK , Schmidt AA , McMahon HT , Sykes C , Bassereau P , Johannes L . Endophilin - A2 functions in membrane scission in clathrin - independent endocytosis . Nature 2015 ; 517 : 493 \u2013 496 . 54 . Galic M , Jeong S , Tsai FC , Joubert LM , Wu YI , Hahn KM , Cui Y , Meyer T . External push and internal pull forces recruit curvature - sensing N - BAR domain proteins to the plasma membrane . Nat Cell Biol 2012 ; 14 : 874 \u2013 881 . 55 . Meunier B , Quaranta M , Daviet L , Hatzoglou A , Leprince C . The membrane - tubulating potential of amphiphysin 2 / BIN1 is dependent on the microtubule - binding cytoplasmic linker protein 170 ( CLIP - 170 ) . Eur J Cell Biol 2009 ; 88 : 91 \u2013 102 . 56 . Wassmer T , Attar N , Harterink M , van Weering JR , Traer CJ , Oakley J , Goud B , Stephens DJ , Verkade P , Korswagen HC , Cullen PJ . The retromer coat complex coordinates endosomal sorting and dynein - mediated transport , with carrier recognition by the trans - Golgi network . Dev Cell 2009 ; 17 : 110 \u2013 122 . 57 . Traer CJ , Rutherford AC , Palmer KJ , Wassmer T , Oakley J , Attar N , Carlton JG , Kremerskothen J , Stephens DJ , Cullen PJ . SNX4 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 589 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e Day et al . coordinates endosomal sorting of TfnR with dynein - mediated transport into the endocytic recycling compartment . Nat Cell Biol 2007 ; 9 : 1370 \u2013 1380 . 58 . Hong Z , Yang Y , Zhang C , Niu Y , Li K , Zhao X , Liu JJ . The retromer component SNX6 interacts with dynactin p150 ( Glued ) and mediates endosome - to - TGN transport . Cell Res 2009 ; 19 : 1334 \u2013 1349 . 59 . Soppina V , Rai AK , Ramaiya AJ , Barak P , Mallik R . Tug - of - war between dissimilar teams of microtubule motors regulates transport and \ufb01ssion of endosomes . Proc Natl Acad Sci U S A 2009 ; 106 : 19381 \u2013 19386 . 60 . Verma P , Ostermeyer - Fay AG , Brown DA . Caveolin - 1 induces formation of membrane tubules that sense actomyosin tension and are inhibited by polymerase I and transcript release factor / cavin - 1 . Mol Biol Cell 2010 ; 21 : 2226 \u2013 2240 . 61 . McNally FJ . Mechanisms of spindle positioning . J Cell Biol 2013 ; 200 : 131 \u2013 140 . 62 . Laan L , Pavin N , Husson J , Romet - Lemonne G , van Duijn M , Lopez MP , Vale RD , Julicher F , Reck - Peterson SL , Dogterom M . Cortical dynein controls microtubule dynamics to generate pulling forces that position microtubule asters . Cell 2012 ; 148 : 502 \u2013 514 . 63 . Redemann S , Pecreaux J , Goehring NW , Khairy K , Stelzer EH , Hyman AA , Howard J . Membrane invaginations reveal cortical sites that pull on mitotic spindles in one - cell C . elegans embryos . PLoS One 2010 ; 5 : e12301 . 64 . van Deurs B , von Bulow F , Vilhardt F , Holm PK , Sandvig K . Destabilization of plasma membrane structure by prevention of actin polymerization . Microtubule - dependent tubulation of the plasma membrane . J Cell Sci 1996 ; 109 : 1655 \u2013 1665 . 65 . Chinnapen DJ , Hsieh WT , te Welscher YM , Saslowsky DE , Kaoutzani L , Brandsma E , D\u2019Auria L , Park H , Wagner JS , Drake KR , Kang M , Benjamin T , Ullman MD , Costello CE , Kenworthy AK , et al . Lipid sorting by ceramide structure from plasma membrane to ER for the cholera toxin receptor ganglioside GM1 . Dev Cell 2012 ; 23 : 573 \u2013 586 . 66 . Hsieh WT , Hsu CJ , Capraro BR , Wu T , Chen CM , Yang S , Baumgart T . Curvature sorting of peripheral proteins on solid - supported wavy membranes . Langmuir 2012 ; 28 : 12838 \u2013 12843 . 67 . Lauvrak SU , Walchli S , Iversen TG , Slagsvold HH , Torgersen ML , Spilsberg B , Sandvig K . Shiga toxin regulates its entry in a Syk - dependent manner . Mol Biol Cell 2006 ; 17 : 1096 \u2013 1109 . 68 . Hehnly H , Sheff D , Stamnes M . Shiga toxin facilitates its retrograde transport by modifying microtubule dynamics . Mol Biol Cell 2006 ; 17 : 4379 \u2013 4389 . 69 . Tetaud C , Falguieres T , Carlier K , Lecluse Y , Garibal J , Coulaud D , Busson P , Steffensen R , Clausen H , Johannes L , Wiels J . Two distinct Gb3 / CD77 signaling pathways leading to apoptosis are triggered by anti - Gb3 / CD77 mAb and verotoxin - 1 . J Biol Chem 2003 ; 278 : 45200 \u2013 45208 . 70 . Steffensen R , Carlier K , Wiels J , Levery SB , Stroud M , Cedergren B , Nilsson Sojka B , Bennett EP , Jersild C , Clausen H . Cloning and expression of the histo - blood group Pk UDP - galactose : Ga1beta - 4G1cbeta1 - cer alpha1 , 4 - galactosyltransferase . Molecular genetic basis of the p phenotype . J Biol Chem 2000 ; 275 : 16723 \u2013 16729 . 71 . Choy E , Chiu VK , Silletti J , Feoktistov M , Morimoto T , Michaelson D , Ivanov IE , Philips MR . Endomembrane traf\ufb01cking of Ras : the CAAX motif targets proteins to the ER and Golgi . Cell 1999 ; 98 : 69 \u2013 80 . 72 . Quintyne NJ , Gill SR , Eckley DM , Crego CL , Compton DA , Schroer TA . Dynactin is required for microtubule anchoring at centrosomes . J Cell Biol 1999 ; 147 : 321 \u2013 334 . 73 . Pralle A , Keller P , Florin EL , Simons K , Horber JK . Sphingolipid - cholesterol rafts diffuse as small entities in the plasma membrane of mammalian cells . J Cell Biol 2000 ; 148 : 997 \u2013 1008 . 74 . Kenworthy AK , Nichols BJ , Remmert CL , Hendrix GM , Kumar M , Zimmerberg J , Lippincott - Schwartz J . Dynamics of putative raft - associated proteins at the cell surface . J Cell Biol 2004 ; 165 : 735 \u2013 746 . 75 . Zacharias DA , Violin JD , Newton AC , Tsien RY . Partitioning of lipid - modi\ufb01ed monomeric GFPs into membrane microdomains of live cells . Science 2002 ; 296 : 913 \u2013 916 . 76 . Schindelin J , Arganda - Carreras I , Frise E , Kaynig V , Longair M , Pietzsch T , Preibisch S , Rueden C , Saalfeld S , Schmid B , Tinevez JY , White DJ , Hartenstein V , Eliceiri K , Tomancak P , et al . Fiji : an open - source platform for biological - image analysis . Nat Methods 2012 ; 9 : 676 \u2013 682 . 77 . Nehls S , Snapp EL , Cole NB , Zaal KJ , Kenworthy AK , Roberts TH , Ellenberg J , Presley JF , Siggia E , Lippincott - Schwartz J . Dynamics and retention of misfolded proteins in native ER membranes . Nat Cell Biol 2000 ; 2 : 288 \u2013 295 . 78 . Day CA , Kenworthy AK . Mechanisms underlying the con\ufb01ned diffusion of cholera toxin B - subunit in intact cell membranes . PLoS One 2012 ; 7 : e34923 . 79 . Umemura K , Kimura H . Hydrogen sul\ufb01de enhances reducing activity in neurons : neurotrophic role of H2S in the brain ? Antioxid Redox Signal 2007 ; 9 : 2035 \u2013 2041 . 80 . Macia E , Ehrlich M , Massol R , Boucrot E , Brunner C , Kirchhausen T . Dynasore , a cell - permeable inhibitor of dynamin . Dev Cell 2006 ; 10 : 839 \u2013 850 . 81 . Cole NB , Sciaky N , Marotta A , Song J , Lippincott - Schwartz J . Golgi dispersal during microtubule disruption : regeneration of Golgi stacks at peripheral endoplasmic reticulum exit sites . Mol Biol Cell 1996 ; 7 : 631 \u2013 650 . 82 . Goodwin JS , Drake KR , Rogers C , Wright L , Lippincott - Schwartz J , Philips MR , Kenworthy AK . Depalmitoylated Ras traf\ufb01cs to and from the Golgi complex via a nonvesicular pathway . J Cell Biol 2005 ; 170 : 261 \u2013 272 . 83 . Saslowsky DE , Cho JA , Chinnapen H , Massol RH , Chinnapen DJ , Wagner JS , De Luca HE , Kam W , Paw BH , Lencer WI . Intoxication of zebra\ufb01sh and mammalian cells by cholera toxin depends on the \ufb02otillin / reggie proteins but not Derlin - 1 or \u2212 2 . J Clin Invest 2010 ; 120 : 4399 \u2013 4409 . 84 . Jaqaman K , Loerke D , Mettlen M , Kuwata H , Grinstein S , Schmid SL , Danuser G . Robust single - particle tracking in live - cell time - lapse sequences . Nat Methods 2008 ; 5 : 695 \u2013 702 . 590 Traf\ufb01c 2015 ; 16 : 572 \u2013 590 16000854 , 2015 , 6 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1111 / t r a . 12269 by U n i v e r s it y O f W a s h i ng t o n , W il e y O n li n e L i b r a r y on [ 28 / 03 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti o n s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e g ov e r n e d by t h e a pp li ca b l e C r e a ti v e C o mm on s L i ce n s e",
    "debelly2023cell": "Article Cell protrusions and contractions generate long - range membrane tension propagation Graphical abstract Highlights d Forces engaging actin cortex generate rapid long - range membrane tension propagation d Forces applied to cell membrane alone fail to propagate membrane tension d Unifying mechanical model explains the requirements for membrane tension propagation d Membrane tension is consistent with a long - range integrator of cell physiology Authors Henry De Belly , Shannon Yan , Hudson Borja da Rocha , . . . , Herve\u00b4 Turlier , Carlos Bustamante , Orion D . Weiner Correspondence herve . turlier @ college - de - france . fr ( H . T . ) , carlosjbustamante2 @ gmail . com ( C . B . ) , orion . weiner @ ucsf . edu ( O . D . W . ) In brief Experiments and modeling reveal the requirements for rapid and robust membrane tension propagation . De Belly et al . , 2023 , Cell 186 , 3049 \u2013 3061 July 6 , 2023 \u00aa 2023 The Authors . Published by Elsevier Inc . https : / / doi . org / 10 . 1016 / j . cell . 2023 . 05 . 014 ll Article Cell protrusions and contractions generate long - range membrane tension propagation Henry De Belly , 1 , 2 , 10 Shannon Yan , 3 , 4 , 10 Hudson Borja da Rocha , 5 Sacha Ichbiah , 5 Jason P . Town , 1 , 2 Patrick J . Zager , 1 , 2 Dorothy C . Estrada , 1 , 2 Kirstin Meyer , 1 , 2 Herve\u00b4 Turlier , 5 , * Carlos Bustamante , 3 , 4 , 6 , 7 , 8 , 9 , * and Orion D . Weiner 1 , 2 , 11 , * 1 Cardiovascular Research Institute , University of California , San Francisco , San Francisco , CA , USA 2 Department of Biochemistry and Biophysics , University of California , San Francisco , San Francisco , CA , USA 3 Department of Molecular and Cell Biology , University of California , Berkeley , Berkeley , CA 94720 , USA 4 California Institute for Quantitative Biosciences , University of California , Berkeley , Berkeley , CA 94720 , USA 5 Center for Interdisciplinary Research in Biology ( CIRB ) , Colle ` ge de France , CNRS , Inserm , Universite\u00b4 PSL , Paris , France 6 Jason L . Choy Laboratory of Single - Molecule Biophysics , University of California , Berkeley , Berkeley , CA , USA 7 Department of Physics , University of California , Berkeley , Berkeley , CA , USA 8 Howard Hughes Medical Institute , University of California , Berkeley , Berkeley , CA , USA 9 Kavli Energy Nanoscience Institute , University of California , Berkeley , Berkeley , CA , USA 10 These authors contributed equally 11 Lead contact * Correspondence : herve . turlier @ college - de - france . fr ( H . T . ) , carlosjbustamante2 @ gmail . com ( C . B . ) , orion . weiner @ ucsf . edu ( O . D . W . ) https : / / doi . org / 10 . 1016 / j . cell . 2023 . 05 . 014 SUMMARY Membrane tension is thought to be a long - range integrator of cell physiology . Membrane tension has been proposed to enable cell polarity during migration through front - back coordination and long - range protrusion competition . These roles necessitate effective tension transmission across the cell . However , con\ufb02icting observations have left the \ufb01eld divided as to whether cell membranes support or resist tension propagation . This discrepancy likely originates from the use of exogenous forces that may not accurately mimic endoge - nous forces . We overcome this complication by leveraging optogenetics to directly control localized actin - based protrusions or actomyosin contractions while simultaneously monitoring the propagation of membrane tension using dual - trap optical tweezers . Surprisingly , actin - driven protrusions and actomyosin contractions both elicit rapid global membrane tension propagation , whereas forces applied to cell membranes alone do not . We present a simple unifying mechanical model in which mechanical forces that engage the actin cortex drive rapid , robust membrane tension propagation through long - range mem - brane \ufb02ows . INTRODUCTION For proper physiology , cells need a way to link short - range biochemical signaling events to long - range integration of cell - wide behaviors . Membrane tension is thought to serve as this global coordinator during cell migration . Membrane tension is the resistance of membrane to deformations . In cells , membrane tension is thought to be a combination of in - plane tension and adhesion between the membrane and underlying actin cytoskel - eton . 1 , 2 It has been proposed that membrane tension guides shape determination in motile cells by relaying actin - based pro - trusive forces at the front to the disassembly and contraction of the rear . 1 , 3 \u2013 14 Conversely , the retraction of the trailing edge ap - pears to modulate actin organization at the cell front through propagated membrane tension changes . 15 Long - range mem - brane tension propagation may similarly enable protrusions to communicate with one another for the winner - take - all competi - tion that establishes the axis of cell movement . 16 \u2013 19 Membrane tension is thought to serve as a central regulator in many other facets of cell and tissue physiologies , including cell spreading and membrane traf\ufb01cking , 11 , 20 \u2013 25 immune response , 26 , 27 cell fate , 28 , 29 cell division , 30 and organ homeostasis . 31 \u2013 33 To operate as a long - range integrator of cell shape and move - ment , membrane tension needs to propagate rapidly and ef\ufb01 - ciently across the cell . However , the actin cortex\u2019s attachment to the plasma membrane appears to inhibit membrane \ufb02ow and tension propagation when external forces are applied to the plasma membrane . 26 , 34 \u2013 36 It remains a source of signi\ufb01cant debate as to whether cell membranes support or resist long - range membrane tension propagation . This is a crucial point to resolve for understanding the role of membrane tension as a global integrator of cell shape and movement . Several factors could potentially underlie the discrepancies among these con\ufb02icting studies of membrane tension propaga - tion . For instance , tension propagation could be cell - type dependent , perhaps more ef\ufb01cient in migrating cells than in ll OPEN ACCESS Cell 186 , 3049 \u2013 3061 , July 6 , 2023 \u00aa 2023 The Authors . Published by Elsevier Inc . 3049 This is an open access article under the CC BY license ( http : / / creativecommons . org / licenses / by / 4 . 0 / ) . non - motile cells . 35 , 37 , 38 Alternatively , the origin of discrepancies could stem from the limitations of traditional tools for manipu - lating and analyzing membrane tension . For example , exoge - nously applied mechanical perturbations of the plasma membrane may elicit tension responses that are different from those elicited by the endogenously generated mechanical forces that are exerted during cell migration . To overcome these limita - tions , we implemented optogenetics to control localized actin - based cell protrusions or actomyosin contractions while simulta - neously monitoring membrane tension response at multiple locations around the cell using high - precision force measure - ments with dual - trap optical tweezers . We \ufb01nd that optogeneti - cally activated cell protrusions and actomyosin contractions both induce long - range membrane tension propagation within seconds . In contrast , perturbations affecting only the plasma membrane fail to elicit membrane tension propagation\u2014consis - tent with previous results . 34 , 35 , 37 We propose a simple unifying mechanical model in which the cortex resists membrane \ufb02ow when forces are applied to the plasma membrane alone . In contrast , when forces engage the cortex , the membrane and cortex act as an integrated system to ef\ufb01ciently transmit mem - brane tension throughout the cell . Our work demonstrates that membrane tension has the properties expected of a long - range integrator of cell physiology , critical for its role in regulating cell shape and movement . RESULTS Local cell protrusions elicit a rapid long - range increase in membrane tension To investigate membrane tension propagation upon endoge - nous force generation , we used an optogenetic approach ( Opto - PI3K ) 39 , 40 to activate localized actin - driven membrane protrusions in neutrophil - like HL - 60 cells ( Figures 1A , 1B , and S1A \u2013 S1E ; Video S1 ) and increase membrane tension at the pro - truding site . 9 , 16 , 41 The propagation of membrane tension can be A E F G B C D Figure 1 . Local cell protrusions elicit a sharp increase in membrane tension on the opposite side of the cell within seconds ( A ) Optogenetic control for light - induced activation of phosphoinositide 3 - kinase ( PI3K ) via localized recruitment of inter SH2 domain ( iSH2 ) , resulting in Rac GTPase activation that initiates actin - driven cell protrusions ( see STAR Methods ) . ( B ) Time - lapseconfocalimagesofaneutrophil - likeHL - 60cellexpressingopto - construct ( Opto - PI3K ) andmembranemarker ( CAAX - HaloTag ) , showinglocalized membrane protrusion upon light activation . ( C ) After light - activated protrusion on one side of the cell ( top of frame ) , changes in membrane tension on the opposite side ( bottom of frame ) are measured via a membrane tether held by an optical trap . ( Right ) Bright\ufb01eld image of a protruding cell during tether pulling assay . ( D ) After tether pulling measurements , the trapping laser is turned off , and the elastic recoil of the bead toward the cell is observed to con\ufb01rm the absence of cytoskeleton in the tether ( means \u00b1 SD ; n > 15 , N = 5 ) . ( E ) Representative time trace of trap force ( a direct readout of cell membrane tension change ) reveals robust and sharp increase in membrane tension over repeating cycles of light - activated protrusion on the opposite end of the cell ( as in C ) ; light : 90 s on ( shaded area ) . ( F ) Red : averaged time trace of trap force before ( steady state ) , during ( light ) , and after activating cell protrusion ( means \u00b1 SD ; n > 60 , N = 8 ) . Gray : as a control , averaged trace from cells treated with actin polymerization inhibitor ( 10 m M latrunculin B ) shows little membrane tension change upon optogenetic activation . ( G ) Averaged trap force before ( steady state ) and during activation . Box and whiskers : median and min to max ; p values from Wilcoxon paired Student\u2019s t test . Scale bars : 5 m m . See also Figure S1 and Video S1 . ll OPEN ACCESS 3050 Cell 186 , 3049 \u2013 3061 , July 6 , 2023 Article probed via a membrane tether pulled out on the opposite side of cell body using a bead ( coated with lectin to bind carbohydrate groups on the membrane ) and held by an optical trap ( a . k . a . trap - based tether pulling assay ; Figure 1C ; see STAR Methods ) . To verify that our optical trap experiments measure the forces exerted by the plasma membrane as opposed to potential actin polymerization within or along the membrane tether , 42 , 43 we ensured that the trapped beads linked to a membrane tether snap back to the cell within seconds upon the release of the op - tical trap at the end of our experimental measurements ( Fig - ure 1D ; Video S1 ; this was a standard control in our operation protocol for all sets of optical trap experiments ; see STAR Methods ) . In response to light - induced actin - driven protrusions , we observed a rapid long - range increase in membrane tension ( Figures 1E , 1F , and S1F \u2013 S1K ; Video S2 ) . The long - range rise in tension within (cid:1) 5 \u2013 15 s of light activation is in stark contrast to the conclusion arrived at in recent studies 34 that \u2018\u2018cell mem - branes resist \ufb02ow . \u2019\u2019 We also veri\ufb01ed that the observed increase in tension correlates with the local activation of the actin regu - lator , Rac GTPase , which is downstream of phosphoinositide 3 - kinase ( PI3K ) activation and precedes actin - driven protrusion ( Figures S1A \u2013 S1E ) . As an additional control , we treated the cells with the actin inhibitor latrunculin B and observed a lack of mem - brane tension increase after light activation ( Figures 1F , 1G , and S1L ) . These results demonstrate that actin - based protrusions elicit a rapid long - range propagation of membrane tension . Actin - driven protrusions stimulate global , unattenuated membrane tension propagation To examine the dynamics of membrane tension propagation in more detail , we performed a dual - tether pulling assay and simul - taneously monitored membrane tension on the side and back of the cell ( at 90 (cid:3) and 180 (cid:3) from the site of illumination , respectively ) throughout multiple cycles of light - induced protrusion ( Figures 2A \u2013 2C and S2 ; Video S2 ) . Interestingly , the two membrane tethers exhibit a near - simultaneous increase in A E B C D F Figure 2 . Actin - driven protrusions stimulate global , nearly undampened membrane tension propagation ( A ) Adual - tether pullingassaytosimultaneouslymonitormembrane tension onthefarend ( left , trap1at180 (cid:3) ) and onthesideofthecell ( top , trap2at90 (cid:3) ) during light - activated protrusion . ( B ) Representative time traces of dual trap forces over successive cycles of light - activated protrusion show coinciding tension increases on both membrane tethers adjacent to ( trap 2 ) and at the opposite cell surface from ( trap 1 ) protrusion ; light : 90 s on ( shaded area ) , 180 s off . ( C ) Correlation between trap forces at the two tether positions during activation ( blue ) remains robust from \ufb01rst activation cycle to the next ; for comparison , minimal correlation is seen between the two tethers before optogenetic activation ( gray ) . Dashed line : linear regression . ( D ) ( Left ) Time delay measured between tension rise on membrane tethers adjacent to ( trap 2 at 90 (cid:3) , blue ) and opposite from ( trap 1 at 180 (cid:3) , red ) cell protrusion . ( Right ) Inmostcells , thetrapsdetect membrane tension increaseonbothtetherswithin asecondorlessofoneanother , indicating arapidpropagationoftension across the cell . ( E ) Averaged traces of dual trap forces before , during ( light ) , and after activation ( means \u00b1 SD ; n > 25 , N = 4 ) . ( F ) Pearson correlation coef\ufb01cient between dual trap forces measured at steady state , during light activation , and recovery afterward ( 70 s post light ) . Error bar : means \u00b1 SD ; p values from Welch\u2019s unpaired Student\u2019s t test ( n > 10 , N > 4 ) . See also Figure S2 and Video S2 . ll OPEN ACCESS Cell 186 , 3049 \u2013 3061 , July 6 , 2023 3051 Article A D G J K B E H C F I Figure 3 . Membrane tension does not propagate upon direct mechanical pulling on the cell membrane ( A ) A dual - tether assay to detect tension propagation ( static tether , left ) while anearby force is exerted through the use of an optically trapped bead to pull on the membrane (cid:1) 2 - m m away ( moving tether , right ) . ( B ) Anexample timetrace oftrapforcefor dualmembranetension measurements , inwhichonemovingtrap ( T2 , gray ) dynamicallypullson thecellmembrane by continuouslypullingandextendingthemembranetether , whereastheothertrapcontrolsasecondstaticmembranetether ( T1 , black ) tomonitornearbychanges in membrane tension . The increase in the length of the extending tether from the cell body is plotted in gray along the right y axis . ( C ) Correlation plots of normalized trap forces between the moving and static tethers . Five representative measurements from different cells are shown ; dashed lines : linear regression . ( D \u2013 F ) Similarto ( A \u2013 C ) , butprobingtensioninblebs ( membranedetachedfromactincortexgeneratedbyusinglatrunculinBtreatmenttoweakentheactincortex ) ; here , a high correlation is observed between static and moving tethers . ( G \u2013 I ) Similar to ( A \u2013 C ) , but probing tension in cells where the actin cortex has been signi\ufb01cantly disassembled using a combination of latrunculin B treatment and osmotic shock ; a high correlation is observed between static and moving tethers even at a signi\ufb01cant distance from one another ( here , 90 (cid:3) , but in Figures S3H \u2013 S3J , 180 (cid:3) ) . ( J ) Pearson correlation coef\ufb01cient between dual trap forces measured before perturbations ( none ; light gray ) , upon light - activated protrusions ( purple ; Figure 2 ) , during cell membrane pulling ( pink ; A \u2013 C ) , during membrane pulling on a bleb ( light green ; D \u2013 F ) , and during cell membrane pulling in cells with heavily dis - assembled actin cortex ( dark green ; G \u2013 I ) . Error bar : means \u00b1 SD ; p values from Welch\u2019s unpaired Student\u2019s t test ( n > 15 , N > 3 ) . ( legend continued on next page ) ll OPEN ACCESS 3052 Cell 186 , 3049 \u2013 3061 , July 6 , 2023 Article tension , with a delay , on average , of 1 . 2 \u00b1 1 . 2 s between the two ( Figure 2D ) . Readouts on both tethers plateau toward similar ten - sion levels ( Figures 2B , 2E , and S2A \u2013 S2C ) . Furthermore , mem - brane tension measurements of the two tethers remain highly correlated during light - activated protrusion and during recovery ( Figure 2F ) . Our experiments indicate that endogenous actin - based protrusions generate a long - range increase in membrane tension , which is transmitted virtually unattenuated across the cell within seconds . The actin cortex resists membrane tension propagation when external forces are applied to the membrane alone The contradictory observations between this study and some previous studies 34 , 35 , 37 , 38 may originate from how a mechanical perturbation is applied to cell membranes . Here , we optogeneti - cally induce cellular membrane protrusion ( i . e . , endogenous actin driven ) , eliciting rapid global membrane tension propaga - tion . In this approach , the forces of actin polymerization are potentially applied to both the cortex and the plasma membrane . In contrast , previous studies concluding that membrane tension is locally constrained by the actin cytoskeleton 34 used a pair of membrane tethers to pull on the cell membrane ( i . e . , exogenous bead pulling ) , thereby applying forces to the plasma membrane alone . To test whether the membrane - tether - induced forces also fail to propagate in our cells , we repurposed our dual - tether assay to dynamically pull one tether by actively moving the \ufb01rst trap while measuring membrane tension on a nearby membrane tether held in place by the second trap ( i . e . , Figure 2A versus Fig - ure 3A ) . In line with analogous experiments performed in epithe - lial cells , we observe no propagation of membrane tension from the extending tether to the static one ( Figures 3A \u2013 3C , 3J , 3K , S3A , and S3D ; Video S2 ) \u2014even with the two tethers in close proximity ( < 2 m m apart ) . In contrast , when we performed the same dual - tether assay on cellular blebs ( membrane detached from actin cortex , achieved in latrunculin - treated cells ) , tension propagates almost instantly ( < 100 m s , i . e . , below the temporal resolution of the optical tweezers instrument ; Figures 3D \u2013 3F , 3J , 3K , S3B , S3C , S3E , and S3F ; Video S2 ) , in agreement with similar measurements in epithelial cells . 34 These bleb - based ex - periments can only test tension propagation at the size scale of cellular blebs ( < 4 m m ) . To test whether tension can propagate for longer distances , we performed tether experiments in cells treated with inhibitors that ef\ufb01ciently disassemble the actin cyto - skeleton . Because latrunculin does not suf\ufb01ce to depolymerize a population of latrunculin - resistant actin \ufb01laments , we used a combination of latrunculin treatment and osmotic shock , which has previously been shown to adequately depolymerize the actin cortex , 44 as we verify in our cells ( Figure S3G ) . These cortex - free cells exhibited rapid long - range propagation of membrane ten - sion , both for traps at 90 (cid:3) ( Figures 3G \u2013 3I ) and traps at opposite ends of the cell ( Figures S3H \u2013 S3J ; Video S2 ) . Both blebs and cortically depolymerized cells propagate tension when forces are applied to the plasma membrane alone , but cells with an intact cortex do not ( Figures 3J and 3K ) . Our observations that effective membrane tension propagation depends on how me - chanical perturbations are exerted within the context of the same cell type ( Figures 2 , 3 , S2 , and S3 ) suggest that existing disagreements in the \ufb01eld are at least partially methodological in nature . While mechanical perturbations via exogenous tether pulling fail to elicit membrane tension propagation ( consistent with the \u2018\u2018picket fence\u2019\u2019 model of cortex adhesion to the plasma membrane 34 , 38 ) , endogenous actin - based force generation ef\ufb01 - ciently promotes membrane tension propagation across the cell . Long - range tension propagation coincides with directed membrane and actin \ufb02ows toward the protrusion Next , we investigated the mechanism of tension propagation from the site of protrusion to the rest of the cell . We observed the enrichment of our plasma membrane probe in the opto - induced protrusions ( Figure 1B ; Video S3 ) on a similar timescale to that of cellular deformation and tension increase ( 10 \u2013 15 s ) following optogenetic protrusion generation ( Figures S1C \u2013 S1E ) . We hypothesized that membrane and cortical \ufb02ows could underlie the rapid propagation of membrane tension from the site of protrusion to the rest of the cell . To resolve the time - depen - dent \ufb02ow of the plasma membrane and actin cytoskeleton rela - tive to light - activated protrusions , we used \ufb02uorescent markers of the plasma membrane ( CAAX - HaloTag ) and actin cytoskel - eton ( Actin - HaloTag ) ; these markers were respectively examined in separate cells . During protrusion formation , the intensity of the plasma membrane probe is increased at the site of protrusion while decreasing elsewhere ( Figures 4A , 4B , and S4A \u2013 S4G ; Video S3 ) . Because the true width of the plasma membrane is likely to be constant during our experiments , these apparent shifts in intensity presumably represent the bunching and unfold - ing of sub - resolution plasma membrane folds . 45 Neutrophils have more than twice the amount of plasma membrane needed for their apparent cell size , and this excess is held in wrinkled plasma membrane reservoirs . 46 \u2013 48 The actin probe similarly accumulated at the site of protrusion and decreased on the side of the cell opposite from the protrusion ( Video S3 ) . To characterize the \ufb02ows of membrane and actin over time , we developed a novel \ufb02ow inference method based on kymographs to predict a \ufb02ow \ufb01eld that can explain the spatiotemporal redis - tribution of membrane ( or actin ) intensity ( Figure 4C ) . A model to rationalize our experimental observations is that the protrusion resulting from the actin polymerization pulls the actin cortex to - ward the protrusion front , which , in turn , drags the membrane around the cell at each point to which it is connected . In this case , it is reasonable to assume that the \ufb02ows resulting from the actin - driven protrusions are accompanied by dissipation generated by the friction between the membrane and its under - lying cortex . Under these conditions , the observed \ufb02ow reduces to a case of optimal transport 49 , 50 , which minimizes the dissipa - tion . Thus , it is possible to infer membrane and cortical velocity \ufb01elds from the experimental kymographs using the optimal ( K ) Relativeforcechanges ( yaxis ) formembranetensionmonitoredonthestatictetherasafunctionoftheextendingtetherlength ( xaxis ) uponcontinuouspulling . Inthecaseof blebsorcells withheavily disassembled actin cortex ( light and dark green ) , thetension onstatictether increases astheextending tether lengthens ; however , thereareno perceptible tension changes on thestatictethertensionfrom thecellbody ( pink , intactcortex ) even whentheother tetherhasextended by more than 60 m m ( n > 14 , N > 3 ) . Graphical data represent means \u00b1 SDs . See also Figure S3 and Video S2 . ll OPEN ACCESS Cell 186 , 3049 \u2013 3061 , July 6 , 2023 3053 Article transport theory ( see Methods S1 ; Figures S4H and S4I ) . We veri\ufb01ed that our inference method is able to recover velocity \ufb01elds from various simulated kymographs with high accuracy ( Figure S4J ; Methods S1 ) . Our analysis revealed the presence of a cell - wide \ufb02ow of both plasma membrane and actin cortex to - ward the protruding front during light - induced protrusion , and these \ufb02ows reverse direction during recovery ( Figures 4D , 4E , S4F , S4G , and S4K ; Video S3 ) . We used membrane photo - bleaching ( Figures 4F \u2013 4I and S4L \u2013 S4P ) and the tracking of microvilli movement ( Figure S4Q ; Video S3 ) to further validate plasma membrane \ufb02ows toward the protrusion . These directed membrane and cortex \ufb02ows provide a potential mechanism to mediate tension propagation upon cell protrusion . Actomyosin contractions also generate rapid long - range membrane tension propagation and membrane \ufb02ows Because actin - based protrusions elicit membrane and actin \ufb02ows ( Figure 4 ; Video S3 ) that mediate tension propagation ( Fig - ures 1 and 2 ; Video S2 ) , but forces applied to the membrane alone do not ( Figures 3 and S3 ; Video S2 ) , our data suggest that forces applied to the actin cortex could be central to A D E I F G H B C Figure 4 . Long - range tension propagation is accompanied by directed membrane and actin \ufb02ows toward the protrusion ( A ) Confocal images of opto - PI3K cells expressing membrane marker ( CAAX - HaloTag ) : before and during light - activated protrusion . Scale bars : 5 m m . ( B ) Kymographs of membrane \ufb02uorescence along the normalized cell circumference ( y axis ) show that over time ( x axis ) membrane accumulates toward the protruding cell front and is depleted from the back ( n > 50 , N = 6 ; Figure S4 ; see STAR Methods ) . ( C ) Flows of membrane and actin during protrusion are calculated assuming optimal transport ( see STAR Methods ) . ( D ) Membrane\ufb02ow\ufb01eldinferredusingoptimaltransportfromkymographintensitychangesovertime : shortlyafteractivationbegins ( t = 70s , darktealtraces ) , the magnitude of membrane \ufb02ow speed increases ( red dashed arrows ) , with positive speed for clockwise \ufb02ow along the cell upper half and negative speed for counter - clockwise \ufb02ow along the bottom half ( G ) , all moving toward the cell protruding front ( p ) . During recovery ( t = 170 s , light green traces ) , the direction of membrane \ufb02ow reverses ( blue dashed arrows ) . ( E ) Membrane \ufb02ow around the cell before , during , and after ( t = 30 , 70 , and 170 s ) right - side protrusion ; the \ufb02ow magnitude is denoted by the arrow size ( red : forward \ufb02ow , blue : backward ) . Membrane \ufb02ows toward the protrusion in the protruding phase and away from the protrusion during the recovery phase . ( F ) Alternative membrane diffusion assay in which we bleach the membrane marker CellMask across a wide section of the cell ( sparing a small section of the membrane maker ) , opto - activate a portion of the cell angled 90 (cid:3) from the unbleached area ( or use no light as control ) , and monitor the diffusion pattern of the unbleached area over time . ( G ) ( Top ) Example kymograph of unbiased diffusion in a control cell ( no activating light ) . ( Bottom ) Same as top but in a protruding cell , showing biased diffusion and bulk \ufb02ow of the unbleached membrane signal toward the protrusion . Heatmap similar as in ( B ) . ( H ) Sample\ufb01tsofindividualtimepointsofkymographdata ( pointscoloredbyrespectivetimepoints ) withagaussianequation ( thickcurves , coloredbyrespective time points ) . Shifts in the means of the gaussian \ufb01ts , quanti\ufb01ed bulk membrane \ufb02ow , are shown as vertical lines ( colored by respective timepoints ) . ( I ) Quanti\ufb01cation of mean shifts \ufb01t by linear regression to assay membrane \ufb02ow rate in control cells ( gray , no apparent \ufb02ow , u = 3 . 34 nm / s ) and protruding cells ( red , biased \ufb02ow toward side of protrusion , u = 35 . 51 nm / s ) ( N = 3 , n = 3 ) . See also Figure S4 and Video S3 . ll OPEN ACCESS 3054 Cell 186 , 3049 \u2013 3061 , July 6 , 2023 Article membrane tension propagation . As optogenetically activated protrusions exert forces on both the cortex and plasma mem - brane , we next sought to investigate the consequences of applying forces directly to the actin cortex . For this purpose , we leveraged an optogenetic approach to induce local actomy - osin contractility through the local recruitment of the Rho - acti - vating domain of LARG ( leukemia - associated Rho guanine nucleotide exchange factor ) ( Figures 5A \u2013 5C ) . 51 Focal Rho acti - vation elicited the local \ufb02attening of the cell ( Figure 5B ) , as ex - pected from local myosin activation . 52 , 53 Similar to light - acti - vated protrusions ( Figures 1 and 2 ) , light - activated actomyosin contractility also generated a long - range transmission of mem - brane tension ( Figure 5D and 5E ) that rapidly propagated virtually unattenuated across the cell ( Figures 5F \u2013 5H ) . As we observed for actin - based protrusions , the local generation of actomyosin contractility also generated \ufb02ows of both plasma membrane ( Figures 5I \u2013 5L ; Video S4 ) and actin cortex ( Figures S5E \u2013 S5H ; Video S4 ) toward the site of contractions . As an additional con - trol , we also used the speckle tracking of focal enrichments of the actin cortex to demonstrate cortical \ufb02ows toward the site of contractions ( Figure S5I ; Video S4 ) . These data suggest that forces applied to the actin cortex are suf\ufb01cient for ef\ufb01cient mem - brane \ufb02ows and membrane tension propagation in cells Mechanical forces engaging the actin cortex drive robust membrane tension propagation in cells To infer the critical requirements for cellular membrane tension propagation , we constructed a simple composite mechanical model in which an elastic plasma membrane is coupled to a viscous and contractile gel - like actomyosin cortex 52 via adhe - sive linkers ( Figure 6A ; see Methods S1 ) . The tension of a 2D membrane is overall of an entropic origin and corresponds to the unfolding of membrane \ufb02uctuations . In such entropic regime , membrane tension is proportional to the exponential of the area strain , as found experimentally 54 and predicted theoretically . 55 Our model assumes small strains , where this exponential be - haves approximately as an af\ufb01ne function and where the mem - brane can be considered as linearly elastic . For the simplicity of our model , we neglected the contribution of membrane reservoirs , as we do not envision that these domi - nate tension propagation . The presence of multiple membrane reservoirs that can unfold above a given tension threshold would simply limit the ability of tension to increase above this value . Our experimental data are consistent with reservoirs being accessed at the plateau phase ( maximum tension values ) of tension prop - agation . At early , pre - plateau phases of protrusion extension , membrane tension increases rapidly even for relatively small protrusions and then plateaus at a maximum even as the protru - sion continues to expand ( Figure S1F ) . Neutrophils have much larger plasma membrane reservoirs than other cells such as \ufb01 - broblasts ( Figures S6A and S6B 56 ) , making it unlikely that we are exhausting local reservoirs during early , pre - plateau phases of protrusion / contraction or during our tether pulling experi - ments . Intriguingly , both optogenetically induced protrusions and optogenetically induced contractions reach similar maximal membrane tension values , likely re\ufb02ecting the threshold of ac - cessing membrane tension buffers ( Figures 2 and 5 ) . Therefore , membrane tension propagation observed in the pre - plateau phases\u2014the focus of our study here\u2014is unlikely to be affected by the presence of folded membrane reservoirs as tension buffers , which manifest mostly in the plateau phase . In our model , the membrane displacement ( x i ) \u2014upon cortical \ufb02ows ( y i ) \u2014is determined by the overall friction imposed through the interconnecting layer of adhesive linkers ( e . g . , membrane - to - cortex attachment proteins [ MCAs ] , such as Ezrin ) . This friction m exerts a drag force on the cell membrane with a magnitude that is proportional to the relative tangential velocity between the cortex and membrane . Given a moderate membrane - cortex friction , this model adequately captures the known tension re - sponses upon different types of mechanical perturbations ( Figures 6B and 6C ) , including the absence of tension transmis - sion when only the membrane is pulled ( e . g . , exogenous tether pulling ) and rapid propagation of tension upon actin - based cell contraction / protrusion ( e . g . , endogenous force generation ) . Furthermore , the model suggests that perturbations engaging both membrane and cortex not only lead to tension propagation but also exhibit a robust tension transmission over a much wider range of membrane - cortex coupling conditions than perturba - tions engaging either component alone . To test how membrane tension propagation is affected by weakening MCA , we utilized NSC668394 , an inhibitor of Ezrin phosphorylation and Ezrin - actin binding . In accordance with the predictions of our model ( Figure 6C ) , this Ezrin inhibitor elicited only mild defects in protru - sion - mediated tension propagation ( Figures 6D , 6E , and S6C ) and elicited more signi\ufb01cant defects in contraction - mediated tension propagation ( Figures 6F , 6G , and S6C ) . Our modeling suggests that the key determinant of long - range membrane response is not the endogenous or exoge - nous application of force but rather whether the mechanical forces directly engage the actin cortex and whether the cortex is suf\ufb01ciently attached to the membrane ( i . e . , suf\ufb01cient friction / coupling ) to effectively transmit forces to produce membrane displacement upon cortex displacement . To test whether exog - enously applied forces can mediate membrane tension propa - gation , we implemented micropipette aspiration to apply me - chanical pulling on both the actin cortex and plasma membrane and monitored tension propagation using our dual - tether assay ( Figure 6H ; see STAR Methods ) . We detected a rapid , robust , and global increase in membrane ten - sion with little to no attenuation across the cell ( Figures 6I , 6J , and S6D \u2013 S6L ; Video S5 ) . Our unifying model indicates that the plasma membrane and actin cortex act as an integrated sys - tem for robust membrane tension propagation . DISCUSSION By combining optogenetics for local endogenous control of cell protrusion / contraction and optical trapping for direct membrane tension measurements in tether pulling assays , we demonstrate that local mechanical force generation such as through cellular protrusions and contractions elicits rapid long - range propaga - tion of membrane tension throughout the cell . In addition , our \ufb01ndings resolve the long - standing dispute as to whether the actin cortex facilitates or impedes tension propagation . When forces are applied to membranes alone ( e . g . , tether pull - ing ) , the actin cortex opposes membrane \ufb02ow and tension ll OPEN ACCESS Cell 186 , 3049 \u2013 3061 , July 6 , 2023 3055 Article A E I J K L F G H B C D Figure5 . Optogeneticallyinducedactomyosincontractionsgeneraterapidlong - rangemembranetensionpropagationandmembrane\ufb02ows ( A ) Optogenetic approach for light - induced activation of leukemia - associated Rho guanine nucleotide exchange factor ( LARG ) , resulting in Rho GTPase acti - vation to initiate actomyosin - driven cell contraction ( see STAR Methods ) . ( B ) Time - lapse confocal images of a neutrophil - like HL - 60 cell expressing opto - construct ( Opto - LARG ) and membrane marker ( CellMask ) , showing localized membrane contraction and cell \ufb02attening upon light activation . ( C ) Afterlight - activatedcontractionononesideofthecell ( top ) , changesinmembranetensionontheoppositeside ( bottom ) aremeasuredviaamembranetether held by an optical trap . ( D ) Averaged time trace of trap force before ( steady state ) , during ( light ) , and after activating cell contraction ( means \u00b1 SD ; n > 55 , N = 7 ) . ( E ) Averaged trap force before ( steady state ) and during activation . Box and whiskers : median and min to max ; p values from Wilcoxon paired Student\u2019s t test . ( F ) A dual - tetherpullingassayto simultaneouslymonitor membrane tension on thefar end ( left , trap1at180 (cid:3) ) and on theside ofthecell ( top , trap2at 90 (cid:3) ) during light - activated contraction . ( G ) Averagedtracesofdualtrapforcesbefore , during ( light ) , andafteractivationshowingcoincidingtensionincreasesonbothmembranetethersadjacentto ( trap 2 ) and at the opposite cell surface from ( trap 1 ) contraction ( means \u00b1 SD ; n = 25 , N = 4 ) . ( H ) Pearson correlation coef\ufb01cient between dual trap forces measured at steady state and during light activation . Error bar : means \u00b1 SD ; p values from Welch\u2019s unpaired Student\u2019s t test ( n > 20 , N > 4 ) . ( I ) Confocal images of opto - LARG cells stained with membrane marker ( CellMask ) before and during light - activated contraction . ( J ) Kymographs of membrane \ufb02uorescence along the normalized cell circumference ( y axis ) show that over time ( x axis ) membrane accumulates toward the contracting cell front and is depleted from the back ( n = 40 , N = 3 ; see STAR Methods ) . ( K ) Membrane \ufb02ow \ufb01eld inferred using optimal transport from kymograph intensity changes over time : shortly after activation begins ( t = 120 s , teal traces ) , the magnitude of membrane \ufb02ow speed increases ( red dashed arrows ) , with positive speed for clockwise \ufb02ow along the cell upper half and negative speed for counter - clockwise \ufb02ow along the bottom half , all moving toward the site of cell contraction ( p ) . During recovery ( t = 200 s , light green traces ) , the direction of membrane \ufb02ow reverses ( blue dashed arrows ) . ( L ) Membrane \ufb02ow around the cell before , during , and after ( t = 30 , 120 , and 240 s ) right - side contraction ; the \ufb02ow magnitude is denoted by the arrow size ( red : forward\ufb02ow , blue : backward ) . Membrane \ufb02ows towardthecontractioninthecontractingphaseand away fromthecontractionduringtherecoveryphase . Scale bars : 5 m m . See also Figure S5 and Video S4 . ll OPEN ACCESS 3056 Cell 186 , 3049 \u2013 3061 , July 6 , 2023 Article propagation . However , when forces engage the actin cortex un - derneath the membrane\u2014either upon optogenetically induced actin polymerization or actomyosin contraction or upon micropi - pette aspiration\u2014tension rapidly propagates nearly undamp - ened across the cell through the generation of actin - driven mem - brane \ufb02ows ( Figure 6K ) . A D H J K E F G B C I Figure 6 . Mechanical forces acting on the actin cortex drive rapid long - range membrane tension propagation in cells ( A ) A3 - tiercompositemodelformembranetensionpropagationincells : membranedisplacements ( x i ) asareadoutfortensionpropagationuponcortical\ufb02ows ( y i ) depend on the membrane elasticity ( k ) and the membrane - cortex friction m imposed through the adhesive linkers . ( B ) Modelpredictionsofmembranetensionresponseatmoderatemembrane - cortexfriction ( seeMethodsS1 ) : onlyactin - basedpullingleadstotensionincrease and propagation ( red rectangles ) ; external pulling on the membrane alone is inef\ufb01cient ( blue circles ) . ( C ) Predicted membrane tension transmission as a function of membrane - cortex friction ( x axis ) for different targets of force application : plasma membrane only ( blue ) and actin cortex only ( red ) . ( D ) Membrane tension measurements during light - induced protrusions in cells with decreased MCA by using 25 m M of Ezrin inhibitor NSC668394 . ( E ) Red : averaged time trace of trap force before ( steady state ) , during ( light ) and after activating cell protrusion in control cells ( same data as Figure 1F ) . Orange : averaged trace from cells with decreased MCA by using 25 m M of Ezrin inhibitor NSC668394 , showing slight defects in membrane tension propagation during light - activated protrusions ( means \u00b1 SD ; n > 25 , N = 3 ) . ( F and G ) Similar to ( D and E ) but using light - induced actomyosin contractions , in which decreases in MCA lead to severe defects in membrane tension prop - agation across the cell ( red : same data as Figure 5D ; means \u00b1 SD ; n > 25 , N = 4 ) . ( H ) A dual - tether assay to simultaneously monitor membrane tension on the far end ( bottom , trap 1 at 180 (cid:3) ) and on the side of the cell ( right , trap 2 at 90 (cid:3) ) during micropipette aspiration ( top ) , which mechanically pulls on both the membrane and underlying actin cortex ( see STAR Methods ) . ( I ) Representative time traces of dual trap forces over successive cycles of aspiration ( shaded area ) show coinciding tension increases and decreases on both membrane tethers , similar to that in Figure 2B . ( J ) Averaged trap forces measured before ( steady state ) and during aspiration . The robust increase in membrane tension upon aspiration of both membrane and cortex is consistent with our model prediction ( B ) . Box and whiskers : median and min to max ; p values from Wilcoxon paired Student\u2019s t test ( n > 25 , N = 5 ) . ( K ) Schematic of requirements for effective membrane tension propagation : in the presence of membrane - to - cortex attachments , force application to plasma membrane alonedoes not generate tension propagation , in agreement with the picket fence model . However , mechanical stimuli acting on actin cortex , such as contraction , lead to rapid , long - range membrane tension propagation in the presence of signi\ufb01cant membrane - to - cortex attachments . Perturbations affecting bothactin cortex and plasma membrane ( such asprotrusionsor micropipette aspiration ) leadtorobustlong - rangemembranetensionpropagation regardless of membrane - to - cortex attachment levels . See also Figure S6 , Methods S1 , and Video S5 . ll OPEN ACCESS Cell 186 , 3049 \u2013 3061 , July 6 , 2023 3057 Article It is noteworthy that the propagation of membrane tension af - ter cell protrusion / contraction is not only rapid but also unattenu - ated , an optimal behavior for coordinating processes at the scale of the entire cell . Our experiments and modeling suggest that one essential prerequisite for this ef\ufb01cient tension propagation is that the force transmits through the cortex . Accordingly , we propose that the cortical propagation of tension across the cell is supported by a continuous cortical network and that interac - tions between the cortex and cellular substrates other than the low - stiffness , highly compliant plasma membrane must be suf\ufb01 - ciently weak so as to minimize dissipative losses in tension propagation . Any discontinuities in the cortex or physical barriers that disrupt cortical \ufb02ow ( e . g . , the division between the apical versus basolateral portions of epithelia cells ) would be ex - pected to impede tension propagation . Consistent with this idea , we observe more robust membrane \ufb02ows toward the pro - trusion for the portions of the cell away from the substrate sur - face compared with the substrate - adhered ventral region of the cell ( Figures S6M \u2013 S6P ) . Actin - based protrusions and actomyosin contractions both mediate long - range membrane tension propagation and \ufb02ows of both actin and membrane toward the site of protrusion / contraction ( Figures 4 , 5 , S4 , and S5 ) . For actomyosin contrac - tion , the primary force is the myosin contractility that generates the actin \ufb02ows . In this case , the \ufb02ow of the plasma membrane and propagation of membrane tension depend on high MCA ( Figures 6C , 6F , and 6G ) . Compared with actomyosin contrac - tion , we have less of an understanding of why the cortex \ufb02ows toward the protrusion . We speculate that the newly polymerized actin at the leading edge generates a pushing force on the mem - brane while also generating a pulling force on the preexisting actin cortex . Future high - resolution electron microscopy images of protrusive and cortical actin could help reveal the relative or - ganization of these actin networks during motility . 15 , 57 Our modeling suggests that forces that engage both the cortex and plasma membrane could ensure robust membrane tension propagation over a wide range of membrane - to - cortex adhesion strengths ( Figures 6C \u2013 6E ) . During light - activated cell protrusions , forces from actin polymerization are exerted on both the plasma membrane and actin cortex , as can be observed by the \ufb02ow of membrane and cortex toward the site of protrusion ( Figure 4 ) . Membrane \ufb02ows enable membrane ten - sion propagation in regions of low MCA , cortical \ufb02ows permit membrane tension propagation in areas of high MCA , and forces applied to both networks should propagate tension in both set - tings . The ability of protrusions to engage both the plasma mem - brane and cortex may be particularly important for long - range tension propagation in motile cells with discontinuous MCAs . 58 We envision that actin polymerization at the leading edge , where the attachment between the cortex and plasma membrane is weak 58 , 59 , would extend the plasma membrane perpendicular to actin cortex and cause the plasma membrane to \ufb02ow toward the protrusion , whereas at the periphery of the cells ( where MCA goes back up ) , membrane tension would be propagated via pull - ing forces from the actin cortex . Our work indicates that membrane tension has the properties expected of a long - range integrator of cell physiology . Long - range propagation of membrane tension could mediate the competition among multiple protrusion sites for a \u2018\u2018winner - take - all\u2019\u2019 establishment of a dominant front 16 , 18 , 60 and could enable the front - back coordination that maintains cell shape and movement . 5 , 7 , 9 , 14 , 15 In contrast , the coordination of cellular processes that do not apply signi\ufb01cant forces to actin cortex may be more dependent on local membrane reservoirs ; this property could explain why \ufb01lopodia can coexist in adjacent re - gions of the cell without substantially affecting one another . 61 , 62 In future work , it will be critical to examine how cells modulate the dynamic range of membrane tension propagation based on the origin of the forces as well as the continuity and mechanical properties of the cortex . Limitations of the study Inthisstudy , weleveragemultiple modesofforce generation ( op - togenetic protrusion formation , optogenetic cell contractility , op - tical - trap - based tether pulling , and micropipette aspiration ) to probe the requirements for membrane tension propagation in cells . Our model system for this work is neutrophil - like HL - 60 cells . The mechanical model we propose explains our experi - mental results , correctly predicts the effects of MCA perturba - tions , and is consistent with both our experimental observations and those from other groups . Therefore , our general conclusions on membrane tension propagation are likely to translate to other cellular settings . However , it is likely that some of the quantitative featuresweobserveinourcells , inparticular , thenearlyunattenu - atedpropagation oftensionacrossthecellandtheprecisespeed of tension propagation , may differ in other cells where the cortex has different mechanical properties or where there is active me - chanosensory - based regulation of the membrane or cortex . Therefore , it will be important to extend our approach\u2014in partic - ular , the optogenetic engagement of endogenous membrane / cortex forces and direct measurement of membrane tension propagation\u2014to a broader diversity of cell types . STAR + METHODS Detailed methods are provided in the online version of this paper and include the following : d KEY RESOURCES TABLE d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND SUBJECT DETAILS d METHOD DETAILS B Transduction of HL - 60 cells B Microscopy hardware B Preparation of Opto - PI3K and Opto - LARG cells for confocal imaging B Preparation , settings , and operation procedures for membrane tethering pulling experiments on C - trap (cid:2) optical tweezers with confocal imaging B Optical trapping \u2013 setting and operations B Micropipette aspiration d QUANTIFICATION AND STATISTICAL ANALYSIS B Image and membrane tension analysis ll OPEN ACCESS 3058 Cell 186 , 3049 \u2013 3061 , July 6 , 2023 Article B Statistical analysis SUPPLEMENTAL INFORMATION Supplementalinformationcanbefoundonlineathttps : / / doi . org / 10 . 1016 / j . cell . 2023 . 05 . 014 . ACKNOWLEDGMENTS TheauthorsthankM . WuforkindlysharingplasmidsandreagentsandProf . S . X . Liuforcarefulreading andcommentsonourmanuscript . Theythankallpre - sent and past members of the Weiner , Turlier , and Bustamante labs for critical discussions . They also thank the support from Lumicks on C - trap application inlivecellstudies , speci\ufb01cally Drs . S . Leachman , N . Hadizadeh , H . Kelkar , and J . Janmaat for technical assistance in instrumentation and Drs . E . Lissek , W . Peutz , M . Johnson , and P . Wheeler for support with operational sustainability . Data for this study were acquired at the Center for Advanced Light Microscopy - Nikon Imaging Center at UCSF on instruments obtained using grantsfromtheUCSF ProgramforBreakthroughBiomedicalResearchfunded in part by the Sandler Foundation , the Strategic Advisory Committee , and the EVCP Of\ufb01ce Research Resource Program Institutional Matching Instrumenta - tion Award . This work was funded by The National Institutes of Health grant GM118167 ( O . D . W . ) ; the National Science Foundation / Biotechnology and Biological Sciences Research Council grant 2019598 ( O . D . W . ) ; the National Science Foundation Center for Cellular Construction grant DBI - 1548297 ( O . D . W . ) ; a Novo Nordisk Foundation grant for the Center for Geometrically Engineered Cellular Systems , NNF17OC0028176 ( O . D . W . ) ; the Labex Memo - Life , France under the program \u2018\u2018Investissements d\u2019Avenir\u2019\u2019 ANR - 10 - LABX - 54 ( H . B . d . R . ) ; the European Research Council ( ERC ) under the European Union\u2019s Horizon 2020 research and innovation programme , grant agreement no . 949267 ( H . B . d . R . and H . T . ) ; a QLife ( ANR - 17 - CONV - 0005 ) / QBio grant ( O . D . W . and H . T . ) ; EMBO ALTF 203 - 2021 ( H . D . B . ) and T32EB009383 ( P . J . Z . ) ; an American Heart Association Postdoctoral Fellowship ( K . M . ) ; the National Institutes of Health grant K99GM137074 ( S . Y . ) ; and the Howard Hughes Medical Institute ( C . B . ) . AUTHOR CONTRIBUTIONS Conceptualization , O . D . W . , H . D . B . , and S . Y . ; Methodology , H . D . B . , S . Y . , O . D . W . , H . T . , J . T . , S . I . , H . B . d . R . , P . J . Z . , D . C . E . , and K . M . ; Investigation , H . D . B . , S . Y . , S . I . , and H . B . d . R . ; Visualization , H . D . B . and S . Y . ; Funding acqui - sition , O . D . W . , H . T . , H . D . B . , H . B . d . R . , S . Y . , and C . B . ; Supervision , O . D . W . , C . B . , and H . T . ; Writing , H . D . B . , S . Y . , O . D . W . , C . B . , and H . T . DECLARATION OF INTERESTS The authors declare no competing interests . INCLUSION AND DIVERSITY We support inclusive , diverse , and equitable conduct of research . Received : October 17 , 2022 Revised : March 10 , 2023 Accepted : May 11 , 2023 Published : June 12 , 2023 REFERENCES 1 . Pontes , B . , Monzo , P . , Gole , L . , Le Roux , A . - L . , Kosmalska , A . J . , Tam , Z . Y . , Luo , W . , Kan , S . , Viasnoff , V . , Roca - Cusachs , P . , et al . ( 2017 ) . Mem - brane tension controls adhesion positioning at the leading edge of cells . J . Cell Biol . 216 , 2959 \u2013 2977 . https : / / doi . org / 10 . 1083 / jcb . 201611117 . 2 . Sitarska , E . , and Diz - Mun\u02dcoz , A . ( 2020 ) . Pay attention to membrane ten - sion : mechanobiology of the cell surface . Curr . Opin . Cell Biol . 66 , 11 \u2013 18 . https : / / doi . org / 10 . 1016 / j . ceb . 2020 . 04 . 001 . 3 . Batchelder , E . L . , Hollopeter , G . , Campillo , C . , Mezanges , X . , Jorgensen , E . M . , Nassoy , P . , Sens , P . , andPlastino , J . ( 2011 ) . Membranetension reg - ulates motility by controlling lamellipodium organization . Proc . Natl . Acad . Sci . USA 108 , 11429 \u2013 11434 . https : / / doi . org / 10 . 1073 / pnas . 1010481108 . 4 . Fogelson , B . , and Mogilner , A . ( 2014 ) . Computational estimates of mem - brane \ufb02ow and tension gradient in motile cells . PLoS One 9 , e84524 . https : / / doi . org / 10 . 1371 / journal . pone . 0084524 . 5 . Hetmanski , J . H . R . , deBelly , H . , Busnelli , I . , Waring , T . , Nair , R . V . , Sokleva , V . , Dobre , O . , Cameron , A . , Gauthier , N . , Lamaze , C . , et al . ( 2019 ) . Mem - brane tension orchestrates rear retraction in matrix - directed cell migra - tion . Dev . Cell 51 , 460 \u2013 475 . e10 . https : / / doi . org / 10 . 1016 / j . devcel . 2019 . 09 . 006 . 6 . Kapustina , M . , Elston , T . C . , and Jacobson , K . ( 2013 ) . Compression and dilation of the membrane - cortex layer generates rapid changes in cell shape . J . Cell Biol . 200 , 95 \u2013 108 . https : / / doi . org / 10 . 1083 / jcb . 201204157 . 7 . Keren , K . , Pincus , Z . , Allen , G . M . , Barnhart , E . L . , Marriott , G . , Mogilner , A . , andTheriot , J . A . ( 2008 ) . Mechanismofshapedeterminationinmotilecells . Nature 453 , 475 \u2013 480 . https : / / doi . org / 10 . 1038 / nature06952 . 8 . Lavi , I . , Goudarzi , M . , Raz , E . , Gov , N . S . , Voituriez , R . , and Sens , P . ( 2019 ) . Cellular blebs and membrane invaginations are coupled through mem - brane tension buffering . Biophys . J . 117 , 1485 \u2013 1495 . https : / / doi . org / 10 . 1016 / j . bpj . 2019 . 08 . 002 . 9 . Lieber , A . D . , Yehudai - Resheff , S . , Barnhart , E . L . , Theriot , J . A . , and Keren , K . ( 2013 ) . Membranetensioninrapidlymovingcellsisdeterminedbycyto - skeletal forces . Curr . Biol . 23 , 1409 \u2013 1417 . https : / / doi . org / 10 . 1016 / j . cub . 2013 . 05 . 063 . 10 . Ofer , N . , Mogilner , A . , andKeren , K . ( 2011 ) . Actindisassemblyclockdeter - mines shape and speed of lamellipodial fragments . Proc . Natl . Acad . Sci . USA 108 , 20394 \u2013 20399 . https : / / doi . org / 10 . 1073 / pnas . 1105333108 . 11 . Raucher , D . , and Sheetz , M . P . ( 2000 ) . Cell spreading and lamellipodial extension rate is regulated by membrane tension . J . Cell Biol . 148 , 127 \u2013 136 . https : / / doi . org / 10 . 1083 / jcb . 148 . 1 . 127 . 12 . Ren , C . , Yuan , Q . , Braun , M . , Zhang , X . , Petri , B . , Zhang , J . , Kim , D . , Guez - Haddad , J . , Xue , W . , Pan , W . , et al . ( 2019 ) . Leukocyte cytoskeleton polar - ization is initiated by plasma membrane curvature from cell attachment . Dev . Cell 49 , 206 \u2013 219 . e7 . https : / / doi . org / 10 . 1016 / j . devcel . 2019 . 02 . 023 . 13 . Schweitzer , Y . , Lieber , A . D . , Keren , K . , and Kozlov , M . M . ( 2014 ) . Theoret - ical analysis of membrane tension in moving cells . Biophys . J . 106 , 84 \u2013 92 . https : / / doi . org / 10 . 1016 / j . bpj . 2013 . 11 . 009 . 14 . Tsai , T . Y . - C . , Collins , S . R . , Chan , C . K . , Hadjitheodorou , A . , Lam , P . - Y . , Lou , S . S . , Yang , H . W . , Jorgensen , J . , Ellett , F . , Irimia , D . , et al . ( 2019 ) . Ef\ufb01 - cient front - rear coupling in neutrophil chemotaxis by dynamic myosin II localization . Dev . Cell 49 , 189 \u2013 205 . e6 . https : / / doi . org / 10 . 1016 / j . devcel . 2019 . 03 . 025 . 15 . Mueller , J . , Szep , G . , Nemethova , M . , deVries , I . , Lieber , A . D . , Winkler , C . , Kruse , K . , Small , J . V . , Schmeiser , C . , Keren , K . , et al . ( 2017 ) . Load adap - tation of lamellipodial actin networks . Cell 171 , 188 \u2013 200 . e16 . https : / / doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 . 16 . Houk , A . R . , Jilkine , A . , Mejean , C . O . , Boltyanskiy , R . , Dufresne , E . R . , An - genent , S . B . , Altschuler , S . J . , Wu , L . F . , and Weiner , O . D . ( 2012 ) . Mem - brane tension maintains cell polarity by con\ufb01ning signals to the leading edge during neutrophil migration . Cell 148 , 175 \u2013 188 . https : / / doi . org / 10 . 1016 / j . cell . 2011 . 10 . 050 . 17 . Zmurchok , C . , Collette , J . , Rajagopal , V . , and Holmes , W . R . ( 2020 ) . Mem - branetensioncanenhanceadaptationtomaintainpolarityofmigratingcells . Biophys . J . 119 , 1617 \u2013 1629 . https : / / doi . org / 10 . 1016 / j . bpj . 2020 . 08 . 035 . 18 . Diz - Mun\u02dcoz , A . , Thurley , K . , Chintamen , S . , Altschuler , S . J . , Wu , L . F . , Fletcher , D . A . , and Weiner , O . D . ( 2016 ) . Membrane tension acts through PLD2 and mTORC2 to limit actin network assembly during neutrophil migration . PLoS Biol . 14 , e1002474 . https : / / doi . org / 10 . 1371 / journal . pbio . 1002474 . 19 . Neilson , M . P . , Veltman , D . M . , Haastert , P . J . M . van , Webb , S . D . , Macken - zie , J . A . , and Insall , R . H . ( 2011 ) . Chemotaxis : a feedback - based ll OPEN ACCESS Cell 186 , 3049 \u2013 3061 , July 6 , 2023 3059 Article computationalmodelrobustly predictsmultiple aspectsofrealcellbehav - iour . PLoS Biol . 9 , e1000618 . https : / / doi . org / 10 . 1371 / journal . pbio . 1000618 . 20 . Akamatsu , M . , Vasan , R . , Serwas , D . , Ferrin , M . A . , Rangamani , P . , and Drubin , D . G . ( 2020 ) . Principles of self - organization and load adaptation by the actin cytoskeleton during clathrin - mediated endocytosis . Elife 9 , e49840 . https : / / doi . org / 10 . 7554 / eLife . 49840 . 21 . Appadurai , D . , Gay , L . , Moharir , A . , Lang , M . J . , Duncan , M . C . , Schmidt , O . , Teis , D . , Vu , T . N . , Silva , M . , Jorgensen , E . M . , and Babst , M . ( 2020 ) . Plasma membrane tension regulates eisosome structure and function . Mol . Biol . Cell 31 , 287 \u2013 303 . https : / / doi . org / 10 . 1091 / mbc . E19 - 04 - 0218 . 22 . Boulant , S . , Kural , C . , Zeeh , J . - C . , Ubelmann , F . , and Kirchhausen , T . ( 2011 ) . Actin dynamics counteract membrane tension during clathrin - mediated endocytosis . Nat . Cell Biol . 13 , 1124 \u2013 1131 . https : / / doi . org / 10 . 1038 / ncb2307 . 23 . Gauthier , N . C . , Fardin , M . A . , Roca - Cusachs , P . , and Sheetz , M . P . ( 2011 ) . Temporary increase in plasma membrane tension coordinates the activa - tion of exocytosis and contraction during cell spreading . Proc . Natl . Acad . Sci . USA 108 , 14467 \u2013 14472 . https : / / doi . org / 10 . 1073 / pnas . 1105845108 . 24 . Riggi , M . , Bourgoint , C . , Macchione , M . , Matile , S . , Loewith , R . , and Roux , A . ( 2019 ) . TORC2 controls endocytosis through plasma membrane ten - sion . J . CellBiol . 218 , 2265 \u2013 2276 . https : / / doi . org / 10 . 1083 / jcb . 201901096 . 25 . Thottacherry , J . J . , Kosmalska , A . J . , Kumar , A . , Vishen , A . S . , Elosegui - Ar - tola , A . , Pradhan , S . , Sharma , S . , Singh , P . P . , Guadamillas , M . C . , Chaudh - ary , N . , et al . ( 2018 ) . Mechanochemical feedback control of dynamin independent endocytosis modulates membrane tension in adherent cells . Nat . Commun . 9 , 4217 . https : / / doi . org / 10 . 1038 / s41467 - 018 - 06738 - 5 . 26 . Basu , R . , Whitlock , B . M . , Husson , J . , LeFloc\u2019h , A . , Jin , W . , Oyler - Yaniv , A . , Dotiwala , F . , Giannone , G . , Hivroz , C . , Biais , N . , et al . ( 2016 ) . Cytotoxic T Cells Use Mechanical Force to Potentiate Target Cell Killing . Cell 165 , 100 \u2013 110 . https : / / doi . org / 10 . 1016 / j . cell . 2016 . 01 . 021 . 27 . Masters , T . A . , Pontes , B . , Viasnoff , V . , Li , Y . , and Gauthier , N . C . ( 2013 ) . Plasma membrane tension orchestrates membrane traf\ufb01cking , cytoskel - etal remodeling , and biochemical signaling during phagocytosis . Proc . Natl . Acad . Sci . USA 110 , 11875 \u2013 11880 . https : / / doi . org / 10 . 1073 / pnas . 1301766110 . 28 . De Belly , H . D . , Stubb , A . , Yanagida , A . , Labouesse , C . , Jones , P . H . , Pal - uch , E . K . , andChalut , K . J . ( 2021 ) . MembranetensiongatesERK - mediated regulation of pluripotent cell fate . Cell Stem Cell 28 , 273 \u2013 284 . e6 . https : / / doi . org / 10 . 1016 / j . stem . 2020 . 10 . 018 . 29 . Bergert , M . , Lembo , S . , Sharma , S . , Russo , L . , Milovanovi (cid:1) c , D . , Gretars - son , K . H . , Bo\u00a8rmel , M . , Neveu , P . A . , Hackett , J . A . , Petsalaki , E . , and Diz - Mun\u02dcoz , A . ( 2021 ) . Cell surface mechanics gate embryonic stem cell differ - entiation . Cell Stem Cell 28 , 209 \u2013 216 . e4 . https : / / doi . org / 10 . 1016 / j . stem . 2020 . 10 . 017 . 30 . Gudipaty , S . A . , Lindblom , J . , Loftus , P . D . , Redd , M . J . , Edes , K . , Davey , C . F . , Krishnegowda , V . , andRosenblatt , J . ( 2017 ) . Mechanicalstretchtrig - gers rapid epithelial cell division through Piezo1 . Nature 543 , 118 \u2013 121 . https : / / doi . org / 10 . 1038 / nature21407 . 31 . Horsnell , H . L . , Tetley , R . J . , De Belly , H . , Makris , S . , Millward , L . J . , Benjamin , A . C . , Heeringa , L . A . , de Winde , C . M . , Paluch , E . K . , Mao , Y . , and Acton , S . E . ( 2022 ) . Lymph node homeostasis and adaptation to im - mune challenge resolved by \ufb01broblast network mechanics . Nat . Immunol . 23 , 1169 \u2013 1182 . https : / / doi . org / 10 . 1038 / s41590 - 022 - 01272 - 5 . 32 . Li , J . , Hou , B . , Tumova , S . , Muraki , K . , Bruns , A . , Ludlow , M . J . , Sedo , A . , Hyman , A . J . , McKeown , L . , Young , R . S . , etal . ( 2014 ) . Piezo1integrationof vascular architecture with physiological force . Nature 515 , 279 \u2013 282 . https : / / doi . org / 10 . 1038 / nature13701 . 33 . Yuge , S . , Nishiyama , K . , Arima , Y . , Hanada , Y . , Oguri - Nakamura , E . , Ha - nada , S . , Ishii , T . , Wakayama , Y . , Hasegawa , U . , Tsujita , K . , et al . ( 2022 ) . Mechanical loading of intraluminal pressure mediates wound angiogen - esis by regulating the TOCA family of F - BAR proteins . Nat . Commun . 13 , 2594 . https : / / doi . org / 10 . 1038 / s41467 - 022 - 30197 - 8 . 34 . Shi , Z . , Graber , Z . T . , Baumgart , T . , Stone , H . A . , and Cohen , A . E . ( 2018 ) . Cell membranes resist \ufb02ow . Cell 175 , 1769 \u2013 1779 . e13 . https : / / doi . org / 10 . 1016 / j . cell . 2018 . 09 . 054 . 35 . Gomis Perez , C . , Dudzinski , N . R . , Rouches , M . , Landajuela , A . , Machta , B . , Zenisek , D . , and Karatekin , E . ( 2022 ) . Rapid propagation of membrane tension at retinal bipolar neuron presynaptic terminals . Sci . Adv . 8 , eabl4411 . https : / / doi . org / 10 . 1126 / sciadv . abl4411 . 36 . Ellefsen , K . L . , Holt , J . R . , Chang , A . C . , Nourse , J . L . , Arulmoli , J . , Mekhd - jian , A . H . , Abuwarda , H . , Tombola , F . , Flanagan , L . A . , Dunn , A . R . , et al . ( 2019 ) . Myosin - II mediated traction forces evoke localized Piezo1 - depen - dent Ca2 + \ufb02ickers . Commun . Biol . 2 , 298 . https : / / doi . org / 10 . 1038 / s42003 - 019 - 0514 - 3 . 37 . Shi , Z . , Innes - Gold , S . , and Cohen , A . E . ( 2022 ) . Membrane tension prop - agation couples axon growth and collateral branching . Sci . Adv . 8 , eabo1297 . https : / / doi . org / 10 . 1126 / sciadv . abo1297 . 38 . Cohen , A . E . , and Shi , Z . ( 2020 ) . Do cell membranes \ufb02ow like honey or jiggle like JellO ? BioEssays 42 , e1900142 . https : / / doi . org / 10 . 1002 / bies . 201900142 . 39 . Saha , S . , Town , J . P . , and Weiner , O . D . ( 2023 ) . mTORC2 coordinates the leading and trailing edge cytoskeletal programs during neutrophil migra - tion . Mol . Biol . Cell 34 , ar35 . https : / / doi . org / 10 . 1091 / mbc . E22 - 05 - 0191 . 40 . Town , J . , and Weiner , O . ( 2023 ) . Rac negative feedback links local PIP3 rate - of - change to dynamic control of neutrophil guidance . Preprint at bio - Rxiv . https : / / doi . org / 10 . 1101 / 2022 . 12 . 30 . 521706 . 41 . Carlsson , A . E . ( 2018 ) . Membrane bending by actin polymerization . Curr . Opin . Cell Biol . 50 , 1 \u2013 7 . https : / / doi . org / 10 . 1016 / j . ceb . 2017 . 11 . 007 . 42 . Pontes , B . , Viana , N . B . , Salgado , L . T . , Farina , M . , Moura Neto , V . M . , and Nussenzveig , H . M . ( 2011 ) . Cell cytoskeleton and tether extraction . Bio - phys . J . 101 , 43 \u2013 52 . https : / / doi . org / 10 . 1016 / j . bpj . 2011 . 05 . 044 . 43 . Pontes , B . , Ayala , Y . , Fonseca , A . C . C . , Roma\u02dco , L . F . , Amaral , R . F . , Sal - gado , L . T . , Lima , F . R . , Farina , M . , Viana , N . B . , Moura - Neto , V . , and Nus - senzveig , H . M . ( 2013 ) . Membrane elastic properties and cell function . PLoS One 8 , e67708 . https : / / doi . org / 10 . 1371 / journal . pone . 0067708 . 44 . Roffay , C . , Molinard , G . , Kim , K . , Urbanska , M . , Andrade , V . , Barbarasa , V . , Nowak , P . , Mercier , V . , Garc\u0131\u00b4a - Calvo , J . , Matile , S . , et al . ( 2021 ) . Pas - sivecouplingofmembranetensionandcellvolumeduringactiveresponse of cells to osmosis . Proc . Natl . Acad . Sci . USA 118 . e2103228118 . https : / / doi . org / 10 . 1073 / pnas . 2103228118 . 45 . Hoffstein , S . T . , Friedman , R . S . , and Weissmann , G . ( 1982 ) . Degranulation , membrane addition , and shape change during chemotactic factor - induced aggregation of human neutrophils . J . Cell Biol . 95 , 234 \u2013 241 . https : / / doi . org / 10 . 1083 / jcb . 95 . 1 . 234 . 46 . Evans , E . , and Yeung , A . ( 1989 ) . Apparent viscosity and cortical tension of blood granulocytes determined by micropipet aspiration . Biophys . J . 56 , 151 \u2013 160 . 47 . Herant , M . , Heinrich , V . , and Dembo , M . ( 2005 ) . Mechanics of neutrophil phagocytosis : behavior of the cortical tension . J . Cell Sci . 118 , 1789 \u2013 1797 . https : / / doi . org / 10 . 1242 / jcs . 02275 . 48 . Ting - Beall , H . P . , Needham , D . , and Hochmuth , R . M . ( 1993 ) . Volume and osmotic properties of human neutrophils . Blood 81 , 2774 \u2013 2780 . 49 . Monge , G . ( 1781 ) . Me\u00b4moiresurlaThe\u00b4oriedesDe\u00b4blaisetdesRemblais ( De l\u2019Imprimerie Royale ) . 50 . Kantorovitch , L . ( 1958 ) . On the translocation of masses . Manag . Sci . 5 , 1 \u2013 4 . 51 . O\u2019Neill , P . R . , Castillo - Badillo , J . A . , Meshik , X . , Kalyanaraman , V . , Melgar - ejo , K . , and Gautam , N . ( 2018 ) . Membrane \ufb02ow drives an adhesion - inde - pendent amoeboid cell migration mode . Dev . Cell 46 , 9 \u2013 22 . e4 . https : / / doi . org / 10 . 1016 / j . devcel . 2018 . 05 . 029 . 52 . Borja da Rocha , H . , Bleyer , J . , and Turlier , H . ( 2022 ) . A viscous active shell theory of the cell cortex . J . Mech . Phys . Solids 164 , 104876 . https : / / doi . org / 10 . 1016 / j . jmps . 2022 . 104876 . ll OPEN ACCESS 3060 Cell 186 , 3049 \u2013 3061 , July 6 , 2023 Article 53 . Turlier , H . , Audoly , B . , Prost , J . , and Joanny , J . F . ( 2014 ) . Furrow constric - tion in animal cell cytokinesis . Biophys . J . 106 , 114 \u2013 123 . https : / / doi . org / 10 . 1016 / j . bpj . 2013 . 11 . 014 . 54 . Evans , E . , and Rawicz , W . ( 1990 ) . Entropy - driven tension and bending elasticity in condensed - \ufb02uid membranes . Phys . Rev . Lett . 64 , 2094 \u2013 2097 . https : / / doi . org / 10 . 1103 / PhysRevLett . 64 . 2094 . 55 . Fournier , J . B . , Ajdari , A . , and Peliti , L . ( 2001 ) . Effective - area elasticity and tensionofmicromanipulatedmembranes . Phys . Rev . Lett . 86 , 4970 \u2013 4973 . https : / / doi . org / 10 . 1103 / PhysRevLett . 86 . 4970 . 56 . Raucher , D . , and Sheetz , M . P . ( 1999 ) . Characteristics of a membrane reservoir buffering membrane tension . Biophys . J . 77 , 1992 \u2013 2002 . https : / / doi . org / 10 . 1016 / S0006 - 3495 ( 99 ) 77040 - 2 . 57 . Svitkina , T . M . ( 2020 ) . Actin cell cortex : structure and molecular organiza - tion . Trends Cell Biol . 30 , 556 \u2013 565 . https : / / doi . org / 10 . 1016 / j . tcb . 2020 . 03 . 005 . 58 . Bisaria , A . , Hayer , A . , Garbett , D . , Cohen , D . , and Meyer , T . ( 2020 ) . Mem - brane proximal F - actin restricts local membrane protrusions and directs cell migration . Science 368 , 1205 \u2013 1210 . https : / / doi . org / 10 . 1126 / science . aay7794 . 59 . Welf , E . S . , Miles , C . E . , Huh , J . , Sapoznik , E . , Chi , J . , Driscoll , M . K . , Isogai , T . , Noh , J . , Weems , A . D . , Pohlkamp , T . , et al . ( 2020 ) . Actin - membrane release initiates cell protrusions . Dev . Cell 55 , 723 \u2013 736 . e8 . https : / / doi . org / 10 . 1016 / j . devcel . 2020 . 11 . 024 . 60 . Diz - Mun\u02dcoz , A . , Fletcher , D . A . , and Weiner , O . D . ( 2013 ) . Use the force : membrane tension as an organizer of cell shape and motility . Trends Cell Biol . 23 , 47 \u2013 53 . https : / / doi . org / 10 . 1016 / j . tcb . 2012 . 09 . 006 . 61 . Bornschlo\u00a8gl , T . , and Bassereau , P . ( 2013 ) . The sense is in the \ufb01ngertips : the distal end controls \ufb01lopodial mechanics and dynamics in response to external stimuli . Commun . Integr . Biol . 6 , e27341 . https : / / doi . org / 10 . 4161 / cib . 27341 . 62 . Bornschlo\u00a8gl , T . , Romero , S . , Vestergaard , C . L . , Joanny , J . F . , Van Nhieu , G . T . , and Bassereau , P . ( 2013 ) . Filopodial retraction force is generated by cortical actin dynamics and controlled by reversible tethering at the tip . Proc . Natl . Acad . Sci . USA 110 , 18928 \u2013 18933 . https : / / doi . org / 10 . 1073 / pnas . 1316572110 . 63 . Schindelin , J . , Arganda - Carreras , I . , Frise , E . , Kaynig , V . , Longair , M . , Pietzsch , T . , Preibisch , S . , Rueden , C . , Saalfeld , S . , Schmid , B . , et al . ( 2012 ) . Fiji : an open - source platform for biological - image analysis . Nat . Methods 9 , 676 \u2013 682 . https : / / doi . org / 10 . 1038 / nmeth . 2019 . 64 . Sofroniew , N . , Lambert , T . , Evans , K . , Nunez - Iglesias , J . , Bokota , G . , Win - ston , P . , Pen\u02dca - Castellanos , G . , Yamauchi , K . , Bussonnier , M . , Doncila Pop , D . , etal . ( 2022 ) . Napari : amulti - dimensionalimageviewerforPython . Zenodo . https : / / doi . org / 10 . 5281 / zenodo . 7098045 . 65 . Fonseca , J . P . , Bonny , A . R . , Kumar , G . R . , Ng , A . H . , Town , J . , Wu , Q . C . , Aslankoohi , E . , Chen , S . Y . , Dods , G . , Harrigan , P . , et al . ( 2019 ) . A toolkit for rapid modular construction of biological circuits in mammalian cells . ACS Synth . Biol . 8 , 2593 \u2013 2606 . https : / / doi . org / 10 . 1021 / acssynbio . 9b00322 . 66 . van der Walt , S . , Scho\u00a8nberger , J . L . , Nunez - Iglesias , J . , Boulogne , F . , Warner , J . D . , Yager , N . , Gouillart , E . , and Yu , T . ; scikit - imageContributors ( 2014 ) . scikit - image : image processing in Python . PeerJ 2 , e453 . https : / / doi . org / 10 . 7717 / peerj . 453 . 67 . Virtanen , P . , Gommers , R . , Oliphant , T . E . , Haberland , M . , Reddy , T . , Cour - napeau , D . , Burovski , E . , Peterson , P . , Weckesser , W . , Bright , J . , et al . ( 2020 ) . SciPy 1 . 0 : fundamental algorithms for scienti\ufb01c computing in Py - thon . Nat . Methods 17 , 261 \u2013 272 . https : / / doi . org / 10 . 1038 / s41592 - 019 - 0686 - 2 . ll OPEN ACCESS Cell 186 , 3049 \u2013 3061 , July 6 , 2023 3061 Article STAR + METHODS KEY RESOURCES TABLE RESOURCE AVAILABILITY Lead contact Further information and requests for resources and reagents should be directed to and will be ful\ufb01lled by the lead contact , Orion Weiner ( orion . weiner @ ucsf . edu ) REAGENT or RESOURCE SOURCE IDENTIFIER Chemicals , peptides , and recombinant proteins RPMI 1640 supplemented with L - glutamine and 25 mM HEPES Corning 10 - 041 - CM Bovine Serum Albumin ( endotoxin - free , fatty acid free ) Sigma - Aldrich A8806 Heat - inactivated fetal bovine serum Gibco 16140071 DMEM Corning 10 - 017 - CV Bovine Calf Serum Sigma - Aldrich 12138C Latrunculin B Sigma - Aldrich 76343 - 94 - 7 NSC668394 Sigma - Aldrich 341216 Carboxyl latex bead Invitrogen C37278 Concanavalin A Sigma - Aldrich C2272 SPY650 - FastAct (cid:3) Cytoskeleton CY - SC505 CellMask (cid:3) Deep Red Thermo\ufb01sher C10046 Janelia Fluor 646 Janelia JF646X Lenti - X Concentrator Clontech 631231 TransIT - 293 Transfection Reagent Mirus Bio MIR2705 96 - well # 1 . 5 glass - bottom plates Azenta Life Sciences 4ti - 0223 u - Flux (cid:3) \ufb02ow cell ( 70 - mm chips ) Lumicks C1 Glass capillary tube King Precision Glass KG - 33 Deposited data Optimal transport and model GitHub Zenodo Github : https : / / github . com / VirtualEmbryo / membrane - cortex - tension ; Zenodo : https : / / doi . org / 10 . 5281 / zenodo . 7894202 Bleaching & gaussian \ufb01tting GitHub Zenodo Github : https : / / github . com / weinerlab / Inverse _ Photobleach _ Flow ; Zenodo : https : / / doi . org / 10 . 5281 / zenodo . 7894212 Experimental models : Cell lines HL - 60s Bourne lab N / A Opto - PI3K HL - 60s Weiner lab N / A Opto - LARG HL - 60s Weiner lab N / A 3T3 - Swiss Albino UCSF cell culture facility CCLZR083 HEK293T UCSF cell culture facility CCLZR076 Software and algorithms Fiji Schindelin et al 63 N / A Prism 9 Graphpad software , Inc N / A Adobe Illustrator Adobe N / A Excel Microsoft N / A Napari Sofroniew et al 64 N / A ll OPEN ACCESS e1 Cell 186 , 3049 \u2013 3061 . e1 \u2013 e5 , July 6 , 2023 Article Materials availability All unique / stable reagents generated in this study are available from the lead contact without restriction . Data and code availability d All data reported in this paper will be shared by the lead contact upon request . d All original code has been deposited on GitHub and Zenodo and is publicly available as of the date of publication . d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request . EXPERIMENTAL MODEL AND SUBJECT DETAILS HL - 60 cells are from the laboratory of Henry Bourne and were recently veri\ufb01ed via STR pro\ufb01ling in . 39 HL - 60s were cultured in R10 growth medium , which is RPMI 1640 supplemented with L - glutamine and 25 mM HEPES ( Corning ; Corning , NY ) and containing 10 % ( v / v ) heat - inactivated fetal bovine serum ( Gibco ; Waltham , MA ) . Cultures were kept at a density of 0 . 2 \u2013 1 . 0 million cells / mL at 37 (cid:3) C / 5 % CO 2 . HEK293T cells ( used to make lentivirus for transduction of HL - 60s ) are from UCSF cell culture facility ( CCLZR076 ) and were grown in DMEM ( Corning ; Corning , NY ) containing 10 % ( v / v ) heat - inactivated fetal bovine serum ( Gibco ; Waltham , MA ) and maintained at 37 (cid:3) C / 5 % CO 2 . All media were 0 . 22 - um \ufb01ltered . Opto - PI3K cells ( iLid - BFP - CAAX , iSH2 - GFP , Pak - PBD - mCherry ) were obtained from . 39 Plasmids used to generate Opto - LARG cells ( iLid - BFP - CAAX , DHPH - ARHGEF1 - GFP , AnillinRBD - mCherry ) , Opto - PI3K expressing CAAX - HaloTag , and Actin - HaloTag were assembled using a Golden - Gate - based modular cloning toolkit . 65 . 3T3 - Swiss Albino were obtained from UCSF cell culture facility ( CCLZR083 ) and were cultured in DMEM ( Corning ; Corning , NY ) supplemented with 10 % Bovine Calf Serum ( Sigma ; St . Louis , MO , 12138C ) and maintained at 37 (cid:3) C / 5 % CO 2 METHOD DETAILS Transduction of HL - 60 cells HEK293T cells were used to generate lentivirus and were seeded into 6 - well plates until approximately 80 % con\ufb02uent . For each transduction , 1 . 5 m g pHR vector ( containing the appropriate transgene ) , 0 . 167 m g vesicular stomatitis virus - G vector , and 1 . 2 m g cytomegalovirus 8 . 91 vector were prepared for transfection using TransIT - 293 Transfection Reagent ( Mirus Bio ; Madison , WI ) . Three days post transduction virus - containing supernatants were harvested and concentrated approximately 40 - fold using Lenti - X Concentrator ( Clontech ; Mountainview , CA ) . Concentrated viruses were frozen and stored at (cid:4) 80 (cid:3) C until needed . For all transduc - tions , thawed virus was mixed with approximately 0 . 3 million cells in growth media supplemented with polybrene ( 8 m g / mL ) and incubated overnight . Cells expressing desired transgenes were isolated using \ufb02uorescence - activated cell sorting ( FACS ) as appro - priate ( FACSAria2 ; BD Biosciences ; Franklin Lakes , NJ ) . Microscopy hardware Imaging depicted in Figures 1B ; 4A , 5B , S4A , S4D , S4F , S5E , S5I , and S6N and Videos S1 , S3 , and S4 were performed at 37 (cid:3) C on a Nikon Eclipse Ti inverted microscope equipped with a Borealis beam conditioning unit ( Andor ) , a CSU - W1 Yokogawa spinning disk ( Andor ; Belfast , Northern Ireland ) , a 100X PlanApo TIRF 1 . 49 numerical aperture ( NA ) objective ( Nikon ; Toyko , Japan ) , an iXon Ultra EMCCD camera ( Andor ) , and a laser merge module ( LMM5 , Spectral Applied Research ; Exton , PA ) equipped with 405 , 440 , 488 , and 561 - nm laser lines . All hardware was controlled using Micro - Manager ( UCSF ) . Optogenetic activation was performed using a LED ( 470 - nm ) via a custom DMD ( Andor Technology ) . Illumination intensities were varied by connecting the LEDs to the analog outputs of a digital - to - analogue converter and setting the LED voltages using serial commands via custom Python code . The microscope is equipped with two stacked dichroic turrets such that samples can be simultaneously illuminated with LEDs and imaged using a 488 - nm long - pass dichroic \ufb01lter ( Chroma Technology Corp . ) Preparation of Opto - PI3K and Opto - LARG cells for confocal imaging For experiments in which we monitored cells by confocal imaging , cells were seeded in a 96 - well # 1 . 5 glass - bottom plates ( Azenta Life Sciences ) in R + B imaging media , which is RPMI 1640 supplemented with L - glutamine and 25 mM HEPES ( Corning ; Corning , NY ) and containing 0 . 2 % Bovine Serum Albumin ( BSA , endotoxin - free , fatty acid free ; A8806 , Sigma ; St . Louis , MO ) . For optoge - netic activation , cells were illuminated using DMD ( see above ) at a chosen location ( using custom Python code ) in a circular pattern of varying size ( (cid:1) 2 microns radius for Opto - PIK , (cid:1) 1 micron radius for Opto - LARG ) for a duration of 90 seconds . For Figures S6M \u2013 S6P , we imaged the cells using a two - step Z - stack of the ventral side ( (cid:1) TIRF plane ) and mid - section of the cell . For plasma membrane and actin imaging using HaloTag ( Figures 1B , 4A , S4F , S4M , S4P , S4Q , and S6N ) , cells were stained with 100nM of JF646X for 10 min before being pelleted at 300 g for 3 min and resuspended in R + B imaging media ( RPMI + 0 . 2 % BSA ) . ll OPEN ACCESS Cell 186 , 3049 \u2013 3061 . e1 \u2013 e5 , July 6 , 2023 e2 Article For plasma membrane imaging using the membrane dye CellMask ( Figures 5B , 5I , and S4D ) , cells were \ufb01rst incubated with (cid:1) 2 - 5 m g / ml of CellMask (cid:3) Deep Red ( C10046 , Thermo\ufb01sher ) for 3 minutes at 37 (cid:3) C / 5 % CO 2 . Cells were then pelleted at 300 g for 3 min and resuspended in R + B imaging media ( RPMI + 0 . 2 % BSA ) . For actin imaging of Opto - LARG ( Figures S5E and S5I ) , cells were incubated with the actin dye SPY650 - FastAct (cid:3) ( CY - SC505 ) for 1h at 37 C / 5 % CO 2 . Cells were then pelleted at 300 g for 3 min and resuspended in R + B imaging media ( RPMI + 0 . 2 % BSA ) . Preparation , settings , and operation procedures for membrane tethering pulling experiments on C - trap (cid:2) optical tweezers with confocal imaging Cell preparation Opto - PI3K & Opto - LARG : 1 - 1 . 5 ml cells ( from culture at density of 0 . 6 \u2013 0 . 8 million cells / mL ) were stained ( with 0 . 5 m l of CellMask (cid:3) Deep Red or 100nM of JF646X ) , then pelleted down and resuspended in either R + B imaging medium ( RPMI + 0 . 2 % BSA ) or R10 medium ( all media 0 . 22 - um \ufb01ltered ) in the absence or presence of actin inhibitor ( 10 m M Latrunculin B ) for samples used in tether pulling assay . To heavily depolymerize the cortex ( Figures 3G \u2013 3K and S3G \u2013 S3J ) Cells were resuspended in a hypotonic media 60 % H2O and 40 % R10 ) containing 10 m M Latrunculin B . To decreases membrane - to - cortex attachment ( Figures 6D \u2013 6G and S6C ) , cells were resuspended in media ( R10 ) containing 25 m M of Ezrin inhibitor NSC668394 ( Sigma - Aldrich , 341216 ) . Bead preparation In a 1 . 7 ml Eppendorf tube , the following solutions were added : 9 m l of ultrapure water ( Corning , 46 - 000 - CM ) , 9 m l of carboxyl latex bead ( 4 % w / v , 2 m m ; Invitrogen , C37278 ) , and 2 m l of Concanavalin A ( 1 mg / ml ; Sigma - Aldrich , C2272 ) ; sample was vortexed at low speed at room temperature for 45 - 60 min ; 1 - 2 m l of this bead mixture stock was added into 1 ml of RPMI 1640 buffer ( 0 . 22 - um \ufb01ltered ) for samples used in tether pulling assay . Micro\ufb02uidics An u - Flux (cid:3) \ufb02ow cell ( 70 - mm chips ; Lumicks , C1 ) , installed on an heat - insulating PVC holder , was passivated with R + B imaging media ( 0 . 22 - um \ufb01ltered ) and pre - warmed at 35 - 37 (cid:3) C for 1 - 2 hours . A custom - made microchamber integrated with micropipettes ( descriptions on the assembly provided at the end ) was used in place of u - Flux (cid:3) \ufb02ow cell to apply aspiration in tether pulling assay performed on C - trap (cid:2) ( instrument operation procedures in the next section ) . During the assay , an air - pressured micro\ufb02ui - dics \ufb02ow system ( u - Flux (cid:3) , Lumicks ) , with pre - cleaned and proper dimensions of tubing connections , was used to deliver cell samples , bead solutions , and blank media / buffers ( for \ufb02ushing ) into the \ufb02ow cell or microchamber . Speci\ufb01cally , a tubing with large ID ( 1 / 32 inch ; Idex , 1520L ) was used to deliver cells at the lowest pressure setting ( 0 . 04 - 0 . 12 mbar , or sometimes just gravity \ufb02ow ) so as to minimize the shear force exerted to the cells during delivery . The delivery of beads and media was made with a narrower tubing ( ID 0 . 004 inch ; Upchurch Scienti\ufb01c , PM - 1148 - F ) . After \ufb02owing (cid:1) 200 - 500 m l ( suf\ufb01cient to displace dead volumes combined within the micro\ufb02uidics system ) of cells into the C - trap (cid:2) system pre - warmed at (cid:1) 36 (cid:3) C , incubated for 15 - 20 min so that the cells settle and stably attach to the bottom surface of the u - Flux (cid:3) \ufb02ow cell . Cell locations were then marked prior to the subsequent tether pulling experiments with optical traps . The cell samples were replenished every 1 . 5 - 2 hours , with abundant \ufb02ushing of R + B imaging medium in between ( which ensures the \ufb02ow cell surface remains properly passivated ) . Optical trapping \u2013 setting and operations A commercial dual - trap optical tweezers with 3 - color confocal imaging , aka C - trap (cid:2) , from Lumicks was used to perform the tether pulling assay with concurrent \ufb02uorescence imaging . The \ufb02ow cell , or microchamber , held paralleled to the table surface was aligned perpendicular between a water objective ( 60x , NA 1 . 2 ; Nikon , MRD07602 ) coming from the bottom and a matching condenser ( 60x , NA 1 . 4 , used with Type A immersion oil ; Leica ) coming from the top . The \ufb02ow cell was positioned in between the two such the IR laser beams ( 1064 nm ) focused down by the objective were formed inside the \ufb02ow cell (cid:1) 10 - 20 m m above the inner bottom surface ( with the \ufb02ow cell nano - stage set at the middle position ) . After the \ufb02ow cell , the condenser can adequately collect photons from the IR trapping beams and project on to position - sensitive detectors ( PSDs ) for accurate trap force measurement ( data acquisition at a rate of 78125 Hz and later down sampled to 10 Hz for analysis ) . The objective also di - rects \ufb02uorescence excitations in the visible wavelength range ( 488 , 532 , and 642 nm respectively for opto - tool , Rac / RhoA biosensor , and CellMask / HaloTag ) into the \ufb02ow cell ( or microchamber ) . The two set of light sources ( IR and visible ) were controlled by separate telescopes and mirror - steering systems upstream from the objective . The same objective collected the emission pho - tons from the imaging / optical trapping sample plane inside the \ufb02ow cell for \ufb02uorescence imaging ( bandpass \ufb01lters : 512 / 25 , 582 / 75 , and 680 / 42 ; camera pixel size : 100 nm ; frame rate depends on confocal scanning area size ) , whereas the condenser provided bright \ufb01eld imaging ( 850 nm LED light source ) recorded at 10 Hz . Both the objective and the condenser were pre - warmed to 35 - 37 (cid:3) C ( temperature control unit , Lumicks ) for at least 2 - 3 hours prior to cell experiments . The IR trapping power was typically set at 100 % trapping laser , 10 % overall power , and 50 - 50 split between trap 1 ( T1 ) and trap 2 ( T2 ) , which is about (cid:1) 175 mW per trap ( measured at the objective front ) and (cid:1) 0 . 2 pN / nm in trap stiffness for a 2 - m m bead ( bead corner frequency (cid:1) 2500 Hz ) . Low settings of excitation laser were suf\ufb01cient for \ufb02uorescence imaging ( typically (cid:1) 2 - 5 % of total power gives (cid:1) 0 . 02 - 0 . 04 m W measured at the objective front ) , minimizing the photo - toxicity to the cell during experiments . ll OPEN ACCESS e3 Cell 186 , 3049 \u2013 3061 . e1 \u2013 e5 , July 6 , 2023 Article At the beginning of each cell recording in the tether pulling assay , 2 - m m Concanavalin - coated beads were \ufb02owed into the micro - chamber ( e . g . , at 0 . 4 mbar via channel 5 in u - Flux (cid:3) C1 \ufb02ow cell ) and single beads were captured in either one or both traps ; we then moved the \ufb02ow cell stage to bring the beads to a cell location marked after incubation ( as described earlier ) . With beads in the vicinity of the cell , i . e . , in z - axis at the same confocal imaging plane for the cell ( (cid:1) 2 - 6 m m from the \ufb02ow cell bottom surface ) and (cid:1) 4 - 6 m m away from the cell body in the x - y plane , the trap stiffness was calibrated , and any residual force readout were zeroed before engaging the bead with the cell body to form membrane tethers . Region of interest ( ROI ) was cropped for bright \ufb01eld imaging ( typically an area of 35x45 m m ) , and continuous recording at 10 Hz was initiated . 1 . Tether pulling assay with light - activated cell protrusions : as seen from the bright \ufb01eld camera , we approached beads to position them in direct contact with the cell body ( even pressing a little , judging from the counter force acted on the bead in the trap ) , then we waited for several seconds before carefully ( slowly ) pulling out membrane tethers ( (cid:1) 4 - 10 m m in length ) at the desired con Figuration ( e . g . , two tethers right angle from each other ) . We monitored steady state tension for at least 1 min ( Figure S1G ) before local 488 - nm excitation ( ROI : 6x10 m m ) continuously for 90 sec on the opposite site of ( or right angle from ) the membrane tether . Upon localized 488 - nm illumination , the local recruitment of opto - controls ( iSH2 labelled with EGFP ) to trigger cell protrusions was also imaged simultaneously ( (cid:1) 1 - 1 . 3 sec / frame scanned ) . Post pro - trusion activation , we monitored cell membrane tension recovery for 180 sec and repeated activation cycles for as long as the tethers last ( see Video S2 ) . At desired time points , i . e . , before , during , or after 488 - nm light activation , the activated Rac was speci\ufb01cally imaged via 532 - nm illumination to visualize the distribution of the Rac biosensor ( Pak - PDB - mCherry ) inside the cell ( see Figures S1B \u2013 S1E and Video S1 ) . Similarly , the changes in cell membrane morphology were imaged over time with 642 - nm illumination ( for CellMask Deep Red or Halo - tag 660 if cells were stained earlier ) . 2 . Other experimental conditions in tether pulling assay , including controls : following the same bead engagement procedure described above , membrane tethers were pulled out from the cell body or from small patches of vesicle - like , outward budding membrane blebs that are detached from actin cortex upon Latrunculin treatment . Speci\ufb01cally , after the \ufb01rst membrane tether was formed , the second tether was pulled from a nearby location (cid:1) 2 m m away . The membrane tension was recorded in the same fashion as detailed earlier but for the following conditions : light activation on wild - type cells or drug - treated opto - PI3K cells ; in the absence of any light illumination , we moved one trap to extend the length of one tether on the cell body ( or bleb ) and monitor the tension response on the other ( see Video S2 ) ; or instead of 488 - nm illumination ( which triggers actin - driven cell protrusion ) , the cell was engaged with micropipette aspiration , which exerts mechanical pulling on both membrane and cortex , and the membrane tension was recorded over cycles of aspiration and relaxation ( see below ; Video S5 ) . For Figures S6A \u2013 S6B , we pulled tethers from either HL - 60s or 3T3 cells at a constant rate until eventual tether breakage ( following method from Raucher and Sheetz 56 ) . We then measured the bead - to - cell distance at the point of tether rupture as detected by our force measurements . At the end of each measurement , unless tethers had already broken on their own or by debris falling into the trap , trapping lasers were turned off to observe the tether and bead elastic recoil toward the cell as a control of the absence of cytoskeleton in the tethers ( Figures 1D , S3C , and S3J and Video S1 ) . Micropipette aspiration A custom - made microchamber was used to implement micropipette aspiration on the C - trap (cid:2) system . Speci\ufb01cally , a micropipette of 2 - 6 um tip diameter was prepared by gravity pulling a thin glass capillary tube ( ID 0 . 040 + / - 0 . 010 mm , OD 0 . 080 + / - 0 . 010 mm , length 150 mm ; King Precision Glass , KG - 33 ) that was threaded through a heated platinum coil ( (cid:1) 2 mm in diam . ; Pt wire is 0 . 3 mm in diam . , Alfa Aesar ) upon application of a desired voltage . The micropipette tip size generally correlates with the heating time required to pull the glass tube apart ; the faster heating , the more rapid the pull , giving micropipette tips in smaller diameters . The micropipette was then sandwiched between a 1 - mm glass cover slide ( 3\u2019\u2019x1\u2019\u2019x 1 mm ; VWR ) and a # 1 . 5 glass cover slip ( 24x60 mm ; VWR ) , held together with two pieces of melted Nesco\ufb01lm ( 100 um in thickness ; Karlan ) as sealant and spacer in the microchamber . 6 holes were drilled prior on the glass slide to provide inlets , which are connected to valves and uFlex (cid:3) pressurized syringe reservoirs ( for cell and bead samples delivery as well as buffer \ufb02ushing ) , and outlets towards the waste collection . The micropipette was con - nected to a separated micro\ufb02uidic pressurized system ( MFCS (cid:3) - EZ from Fluigent ; input : - 600 mbar , output : - 69 to 0 mbar ) powered by a small \ufb02oor pump ( KNF , model : N86KN . 18 , with manual regulator ) to provide aspiration control in the tether pulling assay on C - trap (cid:2) . The aspiration pressure zero point for each micropipette was carefully calibrated and set to have no outward nor inward \ufb02ow detectable to a laser trapped bead that was placed at the tip opening of the micropipette . During the experiments , cells were delivered into the microchamber at the same gentle \ufb02ow rate ( 0 . 04 - 0 . 12 mbar , or sometimes just gravity \ufb02ow ) and captured by the optical trap , which quickly brings the cell to the micropipette tip . A minute amount of suction was applied to keep the cell stably engage with the tip ( so it neither \ufb02oats away from the tip nor falls back into the optical trap ) but without any visible deformation of the cell morphology ( as seen in bright \ufb01eld camera ) . Then following the same bead calibration procedure and membrane tether pulling process as described earlier , consecutive rounds of aspiration and relaxation were performed on the cell for as long as the membrane tethers persist ( see Video S5 ) . ll OPEN ACCESS Cell 186 , 3049 \u2013 3061 . e1 \u2013 e5 , July 6 , 2023 e4 Article QUANTIFICATION AND STATISTICAL ANALYSIS Image and membrane tension analysis Fiji ( NIH ) , 63 Excel ( Microsoft ; Redmond , WA ) , custom Python code , and Prism ( Graphpad software , Inc ) were used for image and membrane tension analysis . Average trap force plots ( Figures 1F , 2E , 5D , 5G , 6E , 6G , S1J , S2A \u2013 S2B , and S5C ) were obtained by aligning trap force traces at time of light induction . Average linked trap force plots were made using Prism Graphpad ( software , Inc ) . In Figures 1G and 5E , average trace trap force was measured for 60 seconds before light induction ( steady state ) and for the duration of the light induction ( 90 seconds , Light ) . For Figure 6J , average trace trap force shown here for 30 seconds before aspiration ( steady state ) , for the duration of aspiration ( 15 - 30 seconds ) , and for intervening recovery periods . Pearson correlation coef\ufb01cients between T1 and T2 were calculated using Prism Graphpad software , Inc ) ( Figures 2F , 3J , 5H , and S6K ) . For Figures 2F and 3J , we used 30 seconds before activation for steady state , 30 seconds of light induction for opto - activated protrusion , and (cid:1) 10 - 30 seconds of active tether pulling on cell membrane ( tether length > 30 m m ) and on blebs . In Figure 2F , for \u2018 + Light\u2019 we used the full duration of light activation ( 90sec ) and for \u2018Recovery\u2019 70sec post light induction . In Figure S6K we used 15 - 30 sec pre - aspiration for steady state value and full duration of aspiration ( (cid:1) 15 - 30sec ) for aspiration . Delay between T2 and T1 during light induced protrusion ( Figures 2D , S5D , and S6J ) was calculated by measuring the time differ - ence between light induction and change in trap force slope for each trap . Of note , measuring time difference from light induction to plateau in force increase yields similar results ( i . e . , delay time between the two traps is still of (cid:1) 1 sec ) . For measurement of relative tether force over distance of moving tether ( Figure 3K ) , we normalized the trap force of static tether by its average when the extending tether was at distance < 1 m m ( namely , before any active pulling ) . For Figures S5E \u2013 S5F , we observed that FastAct ( see above ) intensity linearly increases during imaging . To make sure that these increases did not interfere with our quanti\ufb01cations , we used a set of (cid:1) 40 unstimulated cells , acquired in parallel of every Opto - LARG experiments using FastAct and corrected our measured \ufb02uorescence intensity to compensate for this passive intensity increase . For Figure S1F , in combination to trap forces measurements ( see above ) , cell diameter at the long axis was measured using a custom Fiji plugin and bright\ufb01eld imaging from optical trap setup as a proxy to roughly approximate cell shape changes during light - induced protrusions . For Figure S4Q , microvilli tracking was achieved by manually tracking microvilli over consecutive time frames using HaloTag - CAAX ( see above ) as membrane marker . For Figure S5I , Actin speckle tracking was achieved by using actin dye FastAct ( see above ) and by manually tracking distinct actin features ( e . g . , high intensity points ) over consecutive time frames . For tether length tracking ( Figures 1D , 3B , 3E , 3H , 3K , S3A \u2013 S3C , S3I , and S3J ) , we used a custom - made Fiji macro tracking the position of the bead overtime using bright\ufb01eld recordings during optical trapping experiments . The timestamps for turning the trap off were also recorded . In Figure 4G , kymographs were generated by segmenting the cell body through the HT - CAAX ( JFX - 549 ) channel and \ufb01nding the three - pixel wide boundary pixels that best capture the membrane of the CellMask channel . This segmented cell outline is unraveled and averaged over every three values to provide 1 x N arrays which are stacked to show the evolution of membrane signal over time . Image segmentation code utilized the Python package Scikit - Image . 66 In Figure 4H , f \u00f0 x \u00de = m , e (cid:4) \u00f0 x (cid:4) c \u00de 2 2 r 20 ! + o CellMask signal along the membrane is \ufb01tted with the parametric extension of the gaussian equation de\ufb01ned as : where x is the distance along the cell\u2019s perimeter in m m , m is the peak of the CellMask signal , r0 is the width of the CellMask signal , c position of the peak CellMask signal , and o is the offset of the CellMask signal from zero . The shift in the peak CellMask signal along the membrane was quanti\ufb01ed over time for both control and protruding cells in Figure 4I . Membrane \ufb02ow rates were calculated by taking the slope of the \ufb01tted linear regression lines and averaging the \ufb02ow rates within the control and protruding groups . Code for gaussian and linear regression \ufb01tting utilized the curve _ \ufb01t and linregress functions in the Python package Scipy . 67 Image analysis and gaussian \ufb01tting code in available on Github and Zenodo . Statistical analysis For all statistical analysis , PRISM 9 ( Graphpad software , Inc ) was used . Statistical details can be found in the legend of each \ufb01gure . N represents number of independent biological replicates . Pooled independent experiments are used in dot plots . Unless speci\ufb01ed otherwise , error bars always represent SD . ll OPEN ACCESS e5 Cell 186 , 3049 \u2013 3061 . e1 \u2013 e5 , July 6 , 2023 Article Supplemental \ufb01gures 0 20 40 60 0 10 20 30 40 50 T r ap f o r c e ( p N ) Time ( sec ) Light @ t = 0s A c t i v a t i ng li gh t ( C on f o c a l 488n m ) B r i gh tf i e l d + 2s + 10s + 15s + 30s iSH2 - EGFP 0 1 2 3 4 5 0 10000 20000 30000 40000 Distance ( \u03bc m ) I n t en s i t y ( a . u . ) 0s 2s 10s 30s 15s T 0 30s C D E 0 1 2 3 4 5 Distance ( \u03bcm ) A B Light Anchor Anchor Rac Local protrusion iSH2 iSH2 PI3K Pak - PBD B i nd s a c t i v e R a c PIP3 High Low H a l o T ag - C AAX R a c b i o s en s o r P a k PB D - m C he rr y Before activation After activation Other example post activation High Low Steady state membrane tension measurments Time ( sec ) WT cells + light ISH2 - GFP only cells ( no membrane anchor ) G H K L Light Light Light + 10\u03bcm LatB Light + 10\u03bcm LatB Pre - activation ( steady state ) Activation ( protruding ) Recovery Return to steady state Light Light Light Light Light eeeee Light ) ) Light 50 G F I J \u00d8 Light Risingforce Force plateauing ( legend on next page ) ll OPEN ACCESS Article Figure S1 . Optogenetic control of PI3K leads to local Rac activation , which triggers localized actin - driven cell protrusion and rapid mem - brane tension increase , related to Figure 1 ( A ) Membrane - anchored optogenetic control for light - induced activation of phosphoinositide 3 - kinase ( PI3K ) : upon localized 488 - nm excitation , the membrane anchor protein ( iLid - BFP - CAAX ) undergoes a conformational change , which results in the binding of inter SH2 domain ( iSH2 ) to the illuminated region . iSH2 proceedstorecruitPI3K , whoselipidproduct ( PIP 3 ) inducestheactivationofRacGTPase ( Rac ) . ActiveRacthentriggersactinpolymerizationleadingtolocalized membrane protrusion . By imaging the mCherry - labeled Rac biosensor ( Pak - PBD - mCherry ) , which recognizes and binds the active GTP - bound Rac , we can monitor Rac activation during light - induced protrusions ( see STAR Methods ) . ( B ) Time - lapse confocal images of HL - 60 cells expressing opto - construct ( Opto - PI3K ) , membrane marker ( CAAX - HaloTag , imaged on top ) , and Rac biosensor ( PAK - PBD - mCherry , imagedonbottom ) . Middleand right : localized recruitments ofactiveRaciscon\ufb01rmedatthesiteoflightactivationforcellprotrusion ( boxin black dashed line ) . Scale bar : 5 m m . ( C ) Time - lapse bright\ufb01eld ( top ) and confocal images ( bottom ) of an opto - PI3K cell during light activation . The speci\ufb01c recruitment of PI3K activator , ( iSH2 - EGFP ) to the illuminated area ( box in white dashed line ) is monitored upon 488 - nm excitation . Within 2 s ( between the \ufb01rst two frames ) , iSH2 has redistributed from the cytoplasm to the plasma membrane . Scale bar : 1 m m . ( D ) Fluorescence intensity line scans ( along the white dashed line in ( C ) show the enrichment of opto - construct ( iSH2 - GFP ) at the cell protruding site over time . ( E ) Kymograph of the above line scan ( white dashed line in C ) shows that after iSH2 is recruited to the membrane , the cell contour ( i . e . , its membrane ) rapidly expands outward . ( F ) In red , average time trace of cells before and during light - induced protrusion . In green , apparent cell diameter ( long axis ) over time as proxy of cell shape change and increases in apparent surface area during protrusion . Trap force and shape change are correlated during the initial phase of the protrusion ( rising phase ) but then are decoupled as the cell access its membrane reservoirs limiting further increases in membrane tension ( plateau phase ) even as the protrusion continues to extend ( means \u00b1 SD ; n = 15 , N = 5 ) . ( G ) Representative time trace of trap force measured from the tether pulling assay with a cell at steady state : membrane tension remains stable with low magnitude of stochastic \ufb02uctuations . ( H ) As a control , we light activate the wild - type ( WT ) cells , which lack opto - constructs , and use the same tether pulling assay described above to monitor membranetensionresponsebefore , during , andafter488 - nmillumination ( purpleshadedarea ) . Representativetimetraceoftrapforceforcellmembranetension recorded from WT cells with light activation . The activation light alone does not elicit any changes in cell morphology or membrane tension responses . ( I ) Inanothercontrol , welightactivatecellslackingthemembraneanchorproteinforopto - control ( iLid - BFP - CAAX ) andmonitortheirmembranetensionresponse upon 488 - nm illumination ( purple shaded area ) . No perceptible changes in cell morphology or membrane tension were observed . ( J ) Averaged time trace of trap force ( red ) for cell membrane tension recorded before ( steady state ) , during ( activation ) , and after ( recovery and return to steady state ) light - inducedprotrusionontheoppositesideofthecell ( seeFigure1C ) . Individualdatatraces areshowninlightgray ( samedataasinFigure1F , n > 60 , N = 8 ) . Cells at steady state exhibit stochastic \ufb02uctuations in membrane tension , similar to that shown earlier in ( G ) . Upon light activation ( purple shaded area ) , membrane tension rapidly increases and levels off to a plateau toward the end of activation ( total 90 s ) . The presence of a plateau potentially indicates that membranereservoirsunfoldtoprovideextramembrane , thusbufferingthetensionrise . Shortlyaftertheactivationlightisturnedoff , membranetensiongradually decreases to the steady - state level . ( K ) Two example time traces of trap force for membrane tension before , during , and after light - induced cell protrusion . ( L ) Same as ( K ) but recorded from cells treated with actin inhibitor ( 10 m M latrunculin B ) . We veri\ufb01ed that latrunculin B treatment neither impairs the opto - tool recruitment nor the subsequent Rac activation . This control shows that the increase in membrane tension measured at the opposite side of cell protrusion is dependent on the actin cytoskeleton . ll OPEN ACCESS Article Figure S2 . Membrane tension propagates within seconds across the cell during actin - driven protrusion , related to Figure 2 ( A ) Red and blue : averaged time traces of trap force for dual membrane tensionmeasurements before ( steady state ) , during ( light ) , and after ( recovery ) activating cell protrusion . A nearly coinciding tension increase is observed between the membrane tether adjacent to ( trap 2 , blue ) and opposite from ( trap 1 , red ) cell protrusion . Gray : as a control , averaged trace from cells treated with actin inhibitor ( 10 m M latrunculin B ) shows no membrane tension change upon activation ( means \u00b1 SD ; n > 15 , N > 4 ) . ( B ) Zoom - in on traces in ( A ) : increases in membrane tension emerge on both tethers within the \ufb01rst 5 \u2013 10 s of light activation . ( C ) Three example time traces of trap force for dual membrane tension measurements before , during , and after light - induced cell protrusion . At steady state , tensions from the two tethers show little correlation , but they becomehighly correlated upon light activation ( purple shaded area ) . During the recovery phase , we often observe a lag in time between the two tethers\u2019 tension drop , with the tether opposite from the protrusion site recovering more slowly ( red ) . ( D ) Three example time traces of trap force for dual membrane tension measurements with cells treated with actin inhibitor ( 10 m M latrunculin B ) before , during ( purple shaded area ) , and after light activation of cell protrusion . ll OPEN ACCESS Article ) m \u03bc ( ll e c - d a e b e c n a t s i D Moving trap away from cell Trap OFF elastic recoil of the tether T r a p f o r c e ( p N ) Time ( sec ) D i s t a n c e p u ll i n g t e t h e r ( \u03bc m ) Move trap 1 Move trap 2 T r a p 1 f o r c e ( n o r m ) Trap 1 force ( norm ) A G D E Move trap 1 Move trap 2 T r a p 1 f o r c e ( n o r m ) F S i R - A c t i n Trap 1 force ( norm ) Untreated Latrunculin B Latrunculin B + 60 % H20 * T r ap 2 pu lli ng T r a p f o r c e ( p N ) Time ( sec ) D i s t a n c e p u ll i n g t e t h e r ( \u03bc m ) Step - wise pulling * B T r ap 2 pu lli ng Bleb D i s t a n c e p u ll i n g t e t h e r ( \u03bc m ) T r a p f o r c e ( p N ) Time ( sec ) S t a t i c t r a p M o v i ng t r a p S t a t i c t r a p M o v i ng t r a p H I J R e l a t i v e d i s t an c e o f bead - c e ll Trap ON Trap OFF Elastic recoilof the tether Time ( sec ) C Time ( sec ) Pull on membrane in the absence of actin cortex ) m \u03bc ( ll e c - d a e b e c n a t s i D Moving trap away from cell Time ( sec ) Trap OFF elastic recoil of the tether Trap 1 static Trap 1 static Figure S3 . Mechanical perturbations affecting only the plasma membrane do not result in measurable membrane tension propagation in cells but do in blebs detached from actin cortex , related to Figure 3 ( A ) An example timetrace oftrap force for dual membrane tensionmeasurements , where onemovingtrap ( T2 , blue ) mechanically perturbs on thecellmembrane bycontinuouslypullingandextending themembrane tether , andtheothertrapremains static ( T1 , red ) tomonitorchangesandpropagationintensiontoanearby membranetether . Theincreaseinlengthoftheextendingtetherfromthecellbodyisplottedingrayalongtherightyaxis . \u2018\u2018 * \u2019\u2019annotateswhentheextendingtether broke . Note that a sudden tension release upon breakage of the extending tether ( blue , at t (cid:1) 50 s ) does not lead to changes in tension on the static tether ( red ) , whichisincloseproximitytotheextendingtether ( % 2 m m ) . Thisobservationshowsthatmechanicalperturbationsaffectingonlytheplasmamembraneincellsare locally constrained and inadequate to generate measurable tension propagation between the two tethers . ( B ) Similar operations as ( A ) but monitoring tension propagation between two membrane tethers on cellular blebs ( i . e . , a vesicle - like , small section of membrane detached from actin cortex upon latrunculin B treatment ) . The tension readouts between the extending and the static tethers on blebs appear highly correlated , ( legend continued on next page ) ll OPEN ACCESS Article unlike those on cell body in ( A ) . Speci\ufb01cally , during the \u2018\u2018step - wise pulling\u2019\u2019 to extend tether in trap 2 ( blue ) , the static tether held in trap 1 ( red ) exhibits immediate spikyrisesintension , mirroringthepatternintrap2 . Whenasmoothincreaseisexertedontheextendingtetherbytrap2 ( blue , att (cid:1) 13s ) , thetensionincreaseon statictether ( red ) accordinglybecomesgradual . Furthermore , thesuddendropintensionbacktoinitiallevelonthestatictether ( red , t (cid:1) 26s ) \u2014inresponsetothe sudden tether breakage ( * ) and thus tension release of the extending tether ( blue ) \u2014re\ufb02ects a direct tension transmission and rapid propagation ( see E ) within a membrane bleb detached from the constraining actin cortex . ( C ) Average time trace of relative distance between bead and cell in untreated cells and cells treated with 10 m M of the actin inhibitor latrunculin B . After tether pulling measurements , the trapping laser is turned off and the elastic recoil of the bead toward the cell is observed to con\ufb01rm the absence of cytoskeleton in the tether . Similar tether recoil is observed between untreated and latrunculin - treated cells ( means \u00b1 SD ; n > 13 , N > 3 ) . ( D ) Similar to ( A ) but we alternate which tether is pulling and which tether is static . Trap forces ( readout of membrane tension response ) from static tether is uncorrelated to that of moving tether ( i . e . , little to no change in tension on the static tether during pulling of the moving tether ) . ( E ) Similar to ( C ) but probing tension in blebs ( membrane detached from actin cortex ) ; here , a high correlation is observed between static and moving tethers . ( F ) Example zoom - in traces of dual trap forces ( raw data at 78 kHz ) showing the time difference between a sudden tension release upon breakage ( * ) of the extendingtether ( blue ) andthesubsequentreduction ( \u2726 ) intensiononthestatictether ( red ; tracesslightlyoffsetinyaxisforillustrationclarity ) . Typically , thistime delay observedis % 100 m s ( measured between thein\ufb02ection points , * and \u2726 , on each trace ) , which is right aroundthe temporal resolution of our opticaltrapping instrument ( limitedbythecornerfrequencyofa2 - m mbeadheldbyatrapwithstiffnessof (cid:1) 0 . 2pN / nm ) , indicatingthattheactualtimescaleoftensionpropagation on cellular blebs is likely too fast to be resolved in our experiments . ( G ) Representative confocal images of actin in cells using actin dye SiR - actin , comparing untreated cells as control with cells treated with either 10 m M of la - trunculin B or with a combination of 10 m M of latrunculin B and in hypotonic media ( + 60 % H20 ) . Scale bar : 10 m M . ( H ) Bright\ufb01eld image of dual - tether pulling from opposite sides of a cell treated with a combination of 10 m M of latrunculin B and hypotonic shock . ( I ) Representative force trace of a cell treated with a combination of 10 m M of latrunculin B and a hypotonic shock showing long - range membrane tension propagation in cells with heavily depolymerized cytoskeleton . ( J ) Two example timetraces ofdistancebetweenbead andcellincells treatedwith10 m Moftheactininhibitor latrunculin Bandwithanhypotonicosmoticshock toheavilydepolymerize theactincytoskeleton . Aftertetherpullingmeasurements , thetrappinglaseristurnedoffandtheelastic recoilofthebeadtowardthecell is observed to con\ufb01rm the absence of cytoskeleton in the tether . We observe similar tether recoil as with untreated and latrunculin - treated cells . ll OPEN ACCESS Article Pre - activation Protruding A F r on t B a ck B a ck A ppa r en t m e m b r ane t h i ck ne ss Thick Thin + Light B C E H I Cost Matrix Transport Plan Distance matrix Displacements a b c d e + Light Pre - activation High Low C e ll M a sk H ea t m ap Protruding Apparent membrane thickness Decrease in apparent membrane thickness Membrane unfolding / reservoirs being accessed D J \u22120 . 2 \u22120 . 1 0 . 0 0 . 1 0 . 2 Uniform Distributiont = 0 Flow Flow inference \u22120 . 2 \u22120 . 1 0 . 0 0 . 1 0 . 2 0 . 003 0 . 004 0 . 005 0 . 006 0 . 007 0 . 008 0 . 009 0 . 010 S pa t i a l d i m en s i on Time Time Time S pa t i a l d i m en s i on S pa t i a l d i m en s i on Simulated Velocity field Ground truth Velocity field OptimalTransport reconstruction Kymograph * K Bleaching H a l o T a g - C AA X UnbleachedCellMask Bleaching H a l o T a g - C AA X UnbleachedCellMask L i gh t i ndu c ed p r o t r u s i on Bleaching H a l o T o a g TT - C AA X g Time ( sec ) C e ll M a sk H a l o T ag - C AAX C e ll M a s k hea t m ap C e ll M a sk H a l o T ag - C AAX C e ll M a sk hea t m ap 0s 2 . 5s 5s L M N O Bleaching H a l o T a g - C AA X UnbleachedCellMask L i gh t i ndu c ed p r o t r u s i on Bleaching H a l o T o a g TT - C AA X P C e ll M a sk H a l o T ag - C AAX C e ll M a sk hea t m ap 0s 2 . 5s 5s 7 . 5s A c t i n - H a l o T ag H ea t m ap High Low Steady state ( pre - activation ) Protruding + Light Average membrane intensity ( CellMask ) R e l a t i v e m e m b r ane f l uo r e sc n c e i n t en s i t y High Low F r on t B a ck B a ck Time ( sec ) + Light F G R e l a t i v e a c t i n f l uo r e sc n c e i n t en s i t y High Low F r on t Time ( sec ) B a ck B a ck + Light Front Back Back 0 . 04 0 . 02 0 . 00 - 0 . 02 - 0 . 04 S peed v a l ue ( \u03bc m / s ) 0 Actin \u22120 . 2 \u22120 . 1 0 . 0 0 . 1 - 0 . 02 - 0 . 04 S pee Back0 acBa T = 30sec T = 70sec T = 170sec Protruding Recovery 0s 200s 100s 50s 175s 0s 10s 20s 30s 0s 30s 50s 70s Q ( legend on next page ) ll OPEN ACCESS Article Figure S4 . Long - range tension propagation coincides with directed membrane \ufb02ows toward the protrusion , related to Figure 4 ( A and B ) Apparent membrane thickness is measured based on the width of \ufb02uorescence intensity pro\ufb01le across the cell contour , e . g . , on the side of cell protrusion ( blackline ) . Atsteadystate ( pre - activation ) , thecellmembranecontourappearsrugged ( topimage ) andthickinwidth ( lightgreencurveinB ) , likelydue to the presence of membrane reservoirs . As the cell protrudes , the membrane intensity outside of the protruding region drops ( bottom image ) and becomes thinner in width ( purple curve in B ) . ( C ) Kymograph of averaged apparent membrane thickness along the normalized cell circumference ( y axis ) over time ( x axis ) : before , during , and after localized light - activated protrusion ( box in white dashed line ) . Apart from the protruding site , apparent membrane thickness reduces on average throughout the cell , likely re\ufb02ecting a decrease in membrane reservoirs and a redistribution of extra membranes toward the protrusion site . ( D ) Representative confocal images of an opto - PI3K cell stained with plasma membrane dye ( CellMask ) before light activation or during protrusion . ( E ) Kymograph of membrane \ufb02uorescence intensity ( from cells stained with CellMask ) along the normalized cell circumference ( y axis ) over time ( x axis ) : before , during , and after localized light - activated protrusion ( box in white dashed line ; n > 25 , N = 4 ) . ( F ) Confocal images of opto - PI3K cells expressing actin marker ( actin - HaloTag ) : before and during light - activated protrusion . ( G ) Kymographs of actin \ufb02uorescence along the normalized cell circumference ( y axis ) show that over time ( x axis ) actin accumulates toward the protruding cell front and is depleted from the back ( n > 30 , N = 6 ; see STAR Methods ) . ( H ) Left , evolutionofthetotalmembraneintensityacrossthecellcontour ( means \u00b1 SD ; n > 30 , N = 6 ) . Exceptforasmallintensitydecreaseduetothebleachingof the\ufb02uorophore , the membrane quantityis conserved . Right , evolution ofthe total actin intensity acrossthe cell contour ( means \u00b1 SD ; n > 50 , N = 6 ) . Bleaching of the \ufb02uorophore across time is visible . Actin intensity is conserved across time , with a higher standard deviation than the membrane intensity . ( I ) ( a \u2013 e ) An illustrative example of optimal transport between two discrete 1 - dimensional distributions , at time t ( blue ) and time t + 1 ( orange ) , which represent the amounts of membrane ( or actin ) along the membrane contour at two different time points . ( a ) Cost matrix C , in which C [ i , j ] indicates the value of the cost to displace an element from position i to the position j . Here , the cost function shown is the square of the curvilinear distance . ( b ) Transport Plan to go from the distribution at time t to the distribution at time t + 1 , minimizing the total cost of the displacement , computed from the cost matrix in ( a ) . ( c ) Distance matrix D , in whichD [ i , j ] indicatesthevalueofthedistancebetweenanelementattheposition i andanelementatthepositionj . Thedistancechosenisthecurvilineardistance . ( d ) The transport plan and the distance matrix allow to compute the mean displacement for every position between times t and t + 1 . ( e \u2013 g ) Matrices in the case of periodic boundary conditions , such as the circular contour of the cell . ( e ) Cost matrix with periodic boundary conditions . The cost function chosen is still the square of the curvilinear distance , but as the topology of the curve is periodic , the matrix is changed to re\ufb02ect this new topology . To keep track of the direction of the displacement , the distances can be positive or negative ( and subsequently the positive and negative speed shown in Figures 4D and 4E ) . A displacement in the clockwise direction ( increasing angle coordinate ) is positive , whereas a displacement in the counter - clockwise direction is negative . ( J ) Pipeline and example of \ufb02ow inference validation using computer simulated distributions ( see Methods S1 ) . Using Optimal Transports , \ufb02ows are inferred with minimal errors . ( K ) ( Top ) Actin \ufb02ow \ufb01eld inferred from kymograph intensity change over time using optimal transport . ( Bottom ) Actin \ufb02ow around the cell as inferred by optimal transportbefore , during , andafter ( t = 30 , 70 , and170s ) right - sideprotrusion ; the\ufb02owmagnitudeisdenotedbythearrowsize ( red : forward\ufb02ow , blue : backward ) . ( L ) Alternative membrane diffusion assay inwhichwe sequentially bleach the membrane marker CellMask across awide section of thecell , opto - activate thecell on the side of the unbleached area and monitor the diffusion pattern of the unbleached area over time . We use cells with no activating light as control . ( M ) ExampleconfocalimagesofthemembranemarkersHaloTag - CAAXandCellMaskinacellwithnoactivatinglight ( control , top ) andalight - inducedprotruding cell ( bottom ) . ( N ) Quanti\ufb01cation of shift centroid of signal intensity in control cells ( top , no apparent \ufb02ow ) and protruding cells ( bottom , biased \ufb02ow toward side of protrusion ) . ( O and P ) Similar to ( L and M ) but with an overlap between the unbleached and activation area . ( Q ) Two examples of microvilli tracking during light - induced cell protrusion . Tracked microvilli are circled in red and their trajectory is represented by lines of different colors . Scale bars : 5 m m . ll OPEN ACCESS Article Figure S5 . Optogenetically induced actomyosin contractions generate rapid long - range membrane tension propagation and actin \ufb02ows , related to Figure 5 ( A and B ) Representative time traces of trap force ( a direct readout of cell membrane tension change ) during light - induced actomyosin contraction . Revealing robust increase in membrane tension during light - activated contractions on the opposite end of the cell ; light : 90 s on ( shaded area ) . ( C ) Averaged time trace of trap force before ( steady state ) , during ( Light ) , and after activating cell contraction , measured at the side ( 90 (cid:3) ) of the contraction ( means \u00b1 SD ; n > 30 , N = 7 ) . ( D ) ( Left ) Time delay measured betweentension rise on membrane tethers adjacentto ( trap 2at 90 (cid:3) , blue ) and oppositefrom ( trap 1 at180 (cid:3) , red ) cellcontraction . ( Right ) Inmostcells , thetrapsdetect membrane tensionincreaseonbothtetherswithinasecondorless ofoneanother , indicatingarapid propagation oftension across the cell . Error bar : means \u00b1 SD . ( E ) Confocal images of opto - LARG cells stained with actin marker ( SPY650 - FastAct ) : before and during light - activated contraction . ( F ) Average kymograph of relative actin \ufb02uorescence intensity along the normalized cell circumference ( y axis ) show that over time ( x axis ) . Actin accumulates toward the contracting cell front ( n > 25 , N = 3 ; see STAR Methods ) . ( G ) Actin \ufb02ow \ufb01eld inferred using optimal transport from kymograph intensity changes over time : shortly after activation begins ( t = 120 s , teal traces ) , the magnitude of membrane \ufb02ow speed increases ( red dashed arrows ) , with positive speed for clockwise \ufb02ow along the cell upper half and negative speed for counter - clockwise \ufb02ow along the bottom half , all moving toward the cell contracting front ( p ) . During recovery ( t = 230 s , light yellow traces ) , the direction of membrane \ufb02ow reverses ( blue dashed arrows ) . ( H ) Actin \ufb02ow around the cell before , during , and after ( t = 30 , 80 , and 230 s ) right - side contraction ; the \ufb02ow magnitude is denoted by the arrow size ( red : forward \ufb02ow , blue : backward ) . Membrane \ufb02ows toward the contraction in the contracting phase and away from the protrusion during the recovery phase . ( I ) Twoexamplesofactinspeckletrackingduringlight - inducedcellcontraction . Trackedactinpatchesarecircledinredandtheirtrajectoryisrepresentedbylines of different colors . Scale bars : 5 m m . ll OPEN ACCESS Article ( legend on next page ) ll OPEN ACCESS Article Figure S6 . Mechanical perturbations applied on both membrane and cortex lead to rapid tension propagation across the cell , related to Figure 6 ( A ) Tether pulling assayin which tethers are pulled atconstantspeeduntil they break . Maximum tether length is used as aproxy for local membrane reservoirs 56 . ( B ) Maximum tether length comparison of 3T3s \ufb01broblasts versus HL - 60s cells . In red are cells for which the maximum pulling length was reached on our setup without tether breaking occurring , suggested high local membrane reservoir availability . Error bar : means \u00b1 SD ; n > 15 , N > 3 . ( C ) Average trap force of different opto - cells ( OptoPI3K - based protrusion induction and OptoLARG - based actomyosin contractility ) , before and after light in the absenceorpresenceoftheEzrininhibitorNSC668394 ( 25 m M ) . ThesedatashowthatloweringMCAonlyslightlyaffectsmembranetensionincreaseinprotruding cellsbutseverelyimpedesmembranetensionincreasesincontractingcells . Errorbar : means \u00b1 SD ; pvaluesfromWelch\u2019sunpairedStudent\u2019sttest ( n > 25 , N > 3 ) . ( D ) A dual - tether pulling assay to simultaneously monitor membrane tension on the far end ( bottom , trap 1 at 180 (cid:3) ) and on the side of the cell ( right , trap 2 at 90 (cid:3) ) during micropipette aspiration ( top , (cid:1) 4 \u2013 5 m m in tip diameter ) , which mechanically pulls on both the membrane and actin cortex underneath . ( E ) Representative time traces of dual trap forces over successive cycles of aspiration ( shaded area ) and relaxation ; the magnitude of aspiration progressively increased in the last two cycles ( + and + + ; the \ufb01rst three cycles were also shown in Figure 6I ) . The nearly superimposable tension rise and fall on the two membrane tethers show that membrane tension propagates rapidly across the cell upon mechanical perturbations exerted to both the cortex and membrane . Note that the pro\ufb01les of tension rise upon aspiration and of tension drop during relaxation resemble those observed with light - activated actin - driven protrusions ( Figure 2B ) . ( F ) Zoom - in on the \ufb01rst aspiration event shows that the trap force for membrane tension on the tether closer ( pink ) to the aspiration site started increasing slightly earlier and ended up slightly higher compared with that measured on the tether opposite from the aspiration ( purple ) . ( G ) An example trace of tether tension response monitored on the opposite side of micropipette aspiration ( trap 1 at 180 (cid:3) ) . Here , the recording lasted for six rounds of aspiration and relaxation . ( H ) Another example of dual - tether membrane tension measurement upon micropipette aspiration ; the tether in trap 2 broke ( * ) shortly after the aspiration stopped . ( I ) An example time trace of trap force for cell membrane tension exhibits robust responses over three aspiration cycles using a micropipette of slightly smaller diameter ( (cid:1) 2 m m ) . ( J ) ( Left ) Timedelay measuredbetweentensionriseonmembrane tethers adjacent to ( trap2at90 (cid:3) , pink ) and oppositefrom ( trap1at180 (cid:3) , purple ) cellaspiration using micropipettes . ( Right ) In most cells , the traps detect membrane tension increase on both tethers within a second or less of one another , indicating a rapid propagation of tension across the cell . ( K ) Pearson correlation coef\ufb01cient between dual trap forces measured before any perturbations ( steady state ) and during mechanical pulling upon micropipette aspiration . Error bar : means \u00b1 SD ; p values from Welch\u2019s unpaired Student\u2019s t test ( n > 15 , N > 3 ) . ( L ) Correlation plots of normalized trap forces between the two tethers during micropipette aspiration . Five representative measurements from different cells are shown ; dashed lines : linear regression . ( M ) Comparing membrane \ufb02ows of light - induced protrusions at the mid and ventral plane of the cell . ( N ) Confocal images ofacellmembrane ( visualizedusing CAAX - HaloTag ) before and duringprotrusion attwo different zplanes ( mid - section and ventralplane of the cell ) . Scale bar : 5 m m . ( O ) Average kymograph of relative membrane \ufb02uorescence intensity along the normalized cell circumference ( y axis ) at the ventral and mid - plane of the cell over time ( x axis ) showing a decreased membrane \ufb02ow at the ventral side of the cell , likely due to friction between the cell and the substrate ( n > 30 , N = 3 ) . ( P ) Normalized membrane \ufb02uorescence intensity across the blue dotted line in ( O ) . ll OPEN ACCESS Article",
    "day2015microtubule_supplement3": "Supplementary Figure 3 . ( Related to Figures 4 and 6 ) This figure provides additional evidence that intact microtubules and an intact dynactin complex are required for tubule elongation . It also shows the results of control experiments determining the concentration of ciliobrevin A required to inhibit delivery of CTxB to perinuclear compartments . ( A , B ) Pretreatment of cells with 16 . 6 (cid:0) M nocodazole prior to LatA ( A ) or Jasplakinolide treatment ( B ) blocks tubule formation . ( C , D ) Pretreatment with ciliobrevin A ( Cilio - A ) disrupts delivery of CTxB to perinuclear compartments in ATP replete cells in a dose - dependent manner . n = 36 - 108 cells ; error bars = SD . ( E , F ) Expression of GFP - p50 ( E ) , but not GFP alone ( F ) blocked the formation of long branched tubules in Dynasore - treated cells . Similar results were obtained for both wild type CTxB or monovalent CTx . Bars , 10 (cid:0) m .",
    "milewska2018entry": "Entry of Human Coronavirus NL63 into the Cell Aleksandra Milewska , a , b Paulina Nowak , a , b Katarzyna Owczarek , a , b Artur Szczepanski , a , b Miroslaw Zarebski , c Agnieszka Hoang , c Krzysztof Berniak , c Jacek Wojarski , d Slawomir Zeglen , d Zbigniew Baster , e Zenon Rajfur , e Krzysztof Pyrc a , b a Microbiology Department , Faculty of Biochemistry , Biophysics and Biotechnology , Jagiellonian University , Krakow , Poland b Laboratory of Virology , Malopolska Centre of Biotechnology , Jagiellonian University , Krakow , Poland c Department of Cell Biophysics , Faculty of Biochemistry , Biophysics and Biotechnology , Jagiellonian University , Krakow , Poland d Department of Cardiac Surgery and Transplantology , Silesian Center for Heart Diseases , Zabrze , Poland e Institute of Physics , Faculty of Physics , Astronomy and Applied Computer Sciences , Jagiellonian University , Krakow , Poland ABSTRACT The \ufb01rst steps of human coronavirus NL63 ( HCoV - NL63 ) infection were previously described . The virus binds to target cells by use of heparan sulfate pro - teoglycans and interacts with the ACE2 protein . Subsequent events , including virus internalization and traf\ufb01cking , remain to be elucidated . In this study , we mapped the process of HCoV - NL63 entry into the LLC - Mk2 cell line and ex vivo three - dimensional ( 3D ) tracheobronchial tissue . Using a variety of techniques , we have shown that HCoV - NL63 virions require endocytosis for successful entry into the LLC - MK2 cells , and interaction between the virus and the ACE2 molecule triggers recruitment of clathrin . Subsequent vesicle scission by dynamin results in virus internalization , and the newly formed vesicle passes the actin cortex , which requires active cytoskeleton rearrangement . Finally , acidi\ufb01cation of the endosomal microenvironment is required for successful fusion and release of the viral genome into the cytoplasm . For 3D tra - cheobronchial tissue cultures , we also observed that the virus enters the cell by clathrin - mediated endocytosis , but we obtained results suggesting that this pathway may be bypassed . IMPORTANCE Available data on coronavirus entry frequently originate from studies employing immortalized cell lines or undifferentiated cells . Here , using the most ad - vanced 3D tissue culture system mimicking the epithelium of conductive airways , we systematically mapped HCoV - NL63 entry into susceptible cells . The data obtained al - low for a better understanding of the infection process and may support develop - ment of novel treatment strategies . KEYWORDS HCoV - NL63 , clathrin , Coronaviridae , coronavirus , endocytosis , entry , infection , internalization H uman coronavirus NL63 ( HCoV - NL63 ) was discovered shortly after the emergence of the severe acute respiratory syndrome coronavirus ( SARS - CoV ) ( 1 ) . Extensive studies on the pathogen\u2019s biology and epidemiology revealed that it is prevalent worldwide , appearing with a seasonal distribution similar to that of other human coronaviruses . The clinical presentation may vary depending on the general health status of the patient . Usually , the virus causes a relatively mild respiratory tract disease , but fatal cases have been reported ( 2 \u2013 5 ) . Furthermore , broad studies on the association between infection and clinical symptoms reveal that HCoV - NL63 is associated with croup in young children ( 6 \u2013 9 ) . Phylogenetically , HCoV - NL63 clusters within the genus Alphacoronavirus , which Received 7 November 2017 Accepted 7 November 2017 Accepted manuscript posted online 15 November 2017 Citation Milewska A , Nowak P , Owczarek K , Szczepanski A , Zarebski M , Hoang A , Berniak K , Wojarski J , Zeglen S , Baster Z , Rajfur Z , Pyrc K . 2018 . Entry of human coronavirus NL63 into the cell . J Virol 92 : e01933 - 17 . https : / / doi . org / 10 . 1128 / JVI . 01933 - 17 . Editor Julie K . Pfeiffer , University of Texas Southwestern Medical Center Copyright \u00a9 2018 American Society for Microbiology . All Rights Reserved . Address correspondence to Krzysztof Pyrc , k . a . pyrc @ uj . edu . pl . A . M . and P . N . contributed equally to this article . VIRUS - CELL INTERACTIONS crossm February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 1 Journal of Virology on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m also includes another human pathogen , HCoV - 229E . Initially , these two viruses were considered distant relatives that diverged at some point due to an unknown reason . More recent research shows , however , that these two species most likely emerged in the human population during two separate zoonotic transmission events ( 10 \u2013 12 ) . From the perspective of genome structure , HCoV - NL63 is similar to other alphac - oronaviruses in that the 5 = - terminal two - thirds of the genome encodes a large poly - protein , which is cleaved to yield several nonstructural proteins . Five genes ( encoding S , ORF3 , E , M , and N ) are located at the 3 = terminus and encode structural proteins . The spike protein ( S ) is a class I fusion protein comprising a rod - like domain anchored to the virion via its C terminus and a globular head responsible for the interaction with cellular entry receptors ( 13 ) . It is generally assumed that alphacoronaviruses interact with and enter host cells using CD13 ( aminopeptidase N ) . However , HCoV - NL63 utilizes the ACE2 protein for this purpose , a characteristic shared only with SARS - CoV ( 14 , 15 ) . Virus tropism not only depends on the presence of a certain entry receptor but also may be modulated by other factors , e . g . , attachment receptors , protease availability , and the activity of pathways responsible for internalization and traf\ufb01cking of the virus particle ( 16 , 17 ) . While binding to their cognate entry receptor provides suf\ufb01cient stimulus for some viruses to initiate fusion between the viral and cellular membranes , most internalize via endocytosis ; acidi\ufb01cation and / or processing by cathepsins is then a prerequisite for fusion ( 13 ) . For a long time , endocytic entry of virions was classi\ufb01ed as clathrin dependent , clathrin independent , or clathrin and caveolin independent . During recent years , a number of other pathways were identi\ufb01ed , and this complex machinery has become better understood . The occurrence , abundance , and mechanistic details of these pathways appear to vary between cell types , tissues , and species . Most often , the selection of a speci\ufb01c endocytic route is linked to cargo - directed traf\ufb01cking and receptor - dependent traf\ufb01cking . Nevertheless , many receptors / cargoes allow \ufb02exibility due to their capacity to enter a cell via multiple pathways . The early stages of HCoV - NL63 infection have been described by us and others ( 18 \u2013 20 ) . Here , we made an effort to delineate events that occur early during HCoV - NL63 infection . First , the virus anchors to ciliated cells via heparan sulfate proteoglycans before interacting with the ACE2 entry receptor . Our results show that the virus - ACE2 interaction triggers recruitment of clathrin , followed by clathrin - mediated , dynamin - dependent endocytosis , which requires actin cortex remodeling . To ensure that our results were reliable , we used ex vivo cultured human airway epithelium ( HAE ) , which mimics the microenvironment at the infection site . RESULTS HCoV - NL63 enters the cell via endocytosis . We \ufb01rst determined whether entry of HCoV - NL63 requires endocytosis and acidi\ufb01cation of endosomes . For this , we studied the effect of ammonium chloride ( NH 4 Cl ) and ba\ufb01lomycin A , lysosomotropic agents that inhibit acidi\ufb01cation of endosomes ( 21 \u2013 23 ) , using two models of HCoV - NL63 infection : permissive LLC - Mk2 cells and HAE cultures . Cells were preincubated with NH 4 Cl ( 50 mM ) , ba\ufb01lomycin A ( 100 nM ) , or control dimethyl sulfoxide ( DMSO ) for 1 h at 37\u00b0C and subsequently incubated with the virus at a 50 % tissue culture infective dose ( TCID 50 ) of 100 / ml for LLC - Mk2 cells or 400 / ml for HAE fo r 2 h at32\u00b0C in the presence of the inhibitor . Subsequently , supernatants were removed , and cells were washed thrice with acidic buffer to inhibit the fusogenic activity of the virions retained on the surface ( 24 ) . Next , LLC - Mk2 cells were washed with 1 (cid:2) phosphate - buffered saline ( PBS ) ( pH 7 . 4 ) , overlaid with culture medium , and incubated at 32\u00b0C for 4 days . Supernatant samples were collected for virus replication analysis . Simultaneously , HAE cultures were washed with 1 (cid:2) PBS ( pH 7 . 4 ) and further maintained at an air - liquid interphase at 32\u00b0C for 5 days . During this time , HAE cultures were washed every 24 h with 1 (cid:2) PBS supplemented with a given inhibitor for 10 min at 32\u00b0C , and apical washes were collected for virus replication analysis . Subsequently , viral RNA was isolated and reverse Milewska et al . Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 2 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m transcribed ( RT ) , and the HCoV - NL63 yield was determined using a quantitative real - time PCR ( qPCR ) . Ba\ufb01lomycin A and NH 4 Cl inhibited HCoV - NL63 infection in LLC - Mk2 cells , proving that acidi\ufb01cation is a requirement for the virus infection in vitro . No inhibition was observed in HAE cultures ( Fig . 1A ) . No cytotoxic effect was observed in the presence of these inhibitors ( Fig . 1B ) . Next , we analyzed HCoV - NL63 colocalization with early endosome antigen 1 ( EEA1 ) , a hydrophilic protein localizing exclusively to early endosomes ( 25 ) . LLC - Mk2 cells were \ufb01xed after 10 , 20 , 30 , or 40 min postinoculation ( p . i . ) with gradient - puri\ufb01ed virus , stained with antibodies speci\ufb01c to HCoV - NL63 N protein and EEA1 , and analyzed under a confocal microscope . Measured colocalization , expressed as Manders\u2019 coef\ufb01cient , increases with time and reaches 0 . 68 at 40 min p . i . ( n (cid:3) 6 cells ) ( Fig . 1C ) . We validated the obtained results using the HAE model . Brie\ufb02y , HAE cultures were inoculated with gradient - puri\ufb01ed HCoV - NL63 and incubated at 32\u00b0C for 2 h . For this culture model , a longer incubation was required to observe virus attachment and entry , most likely due to the requirement to cross the mucus layer . Subsequently , cells were \ufb01xed and labeled with speci\ufb01c antibodies against HCoV - NL63 N protein and EEA1 . Colocalization of HCoV - NL63 virus particles with EEA1 protein was analyzed using a confocal microscope . Colocalization of virus and EEA1 was observed in inoculated cells ( Fig . 1D ) . Endocytosis of virus particles is induced by binding to the entry receptor . HCoV - NL63 virus employs the ACE2 protein for cellular entry , while heparan sulfate proteoglycans serve as attachment receptors ( 19 ) . Here , we analyzed the consequence FIG 1 Importance of endosomal entry for HCoV - NL63 infection . ( A ) Inhibition of HCoV - NL63 infection in LLC - Mk2 cells and HAE cultures by the lysosomotropic agents ammonium chloride ( NH 4 Cl ) ( 50 mM ) and ba\ufb01lomycin A ( Baf A ) ( 100 nM ) , as determined by RT - qPCR . Data on the y axis represent LRVs . The assay was performed in triplicate , and average values with standard errors are presented . P values of (cid:4) 0 . 05 were considered signi\ufb01cant and are denoted with an asterisk . ( B ) The cytotoxicity of the tested inhibitors was measured with an XTT assay . Data on the y axis represent viability of the treated cells compared to the untreated reference samples . The assay was performed in triplicate , and average values with standard errors are presented . ( C and D ) Confocal images showing colocalization of HCoV - NL63 virions with the early endosomal marker EEA1 on LLC - Mk2 cells ( C ) and HAE cultures ( D ) . Scale bars (cid:3) 5 (cid:2) m . Green , HCoV - NL63 ; red , EEA1 . Entry of HCoV - NL63 Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 3 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m of interaction between the virus particle and ACE2 . First , we inoculated naturally permissive LLC - Mk2 cells with HCoV - NL63 and incubated them for 40 min at 4\u00b0C to enable virus adhesion to a cell surface . Subsequently cells were \ufb01xed , the virus was labeled with speci\ufb01c antibodies , and its colocalization with ACE2 and clathrin was studied . As shown in Fig . 2A , HCoV - NL63 particles attach ef\ufb01ciently to the cell surface . However , only a proportion of virions colocalize with the ACE2 ( Manders\u2019 coef\ufb01cient (cid:3) 0 . 573 ; n (cid:3) 5 ) , suggesting that binding to the heparan sulfate precedes interaction with FIG 2 HCoV - NL63 binding to ACE2 triggers clathrin - mediated endocytosis . Precooled LLC - Mk2 cells were incubated with gradient - puri\ufb01ed HCoV - NL63 for 40 min at 4\u00b0C following 0 min ( A and B ) or 5 min ( C ) of incubation at 32\u00b0C . Colocalization of the virus ( green ) and ACE2 ( red ) was analyzed using confocal microscopy ( A ) . No colocalization with clathrin was observed after 0 min of incubation ( B ) . Triple colocal - ization of virus with ACE2 and clathrin ( blue ) is visible in panel C . Images on the right side are zoomed - in regions indicated by white rectangles on the left - side slides . A representative image is shown . Scale bars (cid:3) 10 (cid:2) m . Milewska et al . Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 4 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m the entry receptor . At that point , there is no colocalization of virus particles and clathrin - coated pits ( Manders\u2019 coef\ufb01cient (cid:3) 0 . 140 ; n (cid:3) 5 ) ( Fig . 2B ) . Next , we tested whether the virus binding to the adhesion or entry receptor triggers recruitment of common cellular proteins responsible for pit formation by incubating cells for 5 min at 32\u00b0C . Immunostaining showed that the virus particles bound to ACE2 start to colocalize with clathrin ( Manders\u2019 coef\ufb01cient (cid:3) 0 . 849 ; n (cid:3) 6 ) ( Fig . 2C ) , while there is no colocal - ization between non - ACE2 - bound virions and clathrin ( Manders\u2019 coef\ufb01cient (cid:3) 0 . 189 ; n (cid:3) 6 ) . HCoV - NL63 colocalizes with clathrin during entry . To determine whether colo - calization with clathrin following the ACE2 binding is relevant and the virus indeed enters the cell by use of clathrin - coated pits , we analyzed colocalization of intracellular virions with clathrin . Brie\ufb02y , LLC - Mk2 cells were incubated at 32\u00b0C for 5 to 20 min with gradient - puri\ufb01ed HCoV - NL63 , \ufb01xed , immunostained , and analyzed by confocal micros - copy . The results showed colocalization of virions entering the cell with clathrin ( Manders\u2019 coef\ufb01cient (cid:3) 0 . 584 ; n (cid:3) 7 ) ( Fig . 3A ) , whereas no colocalization with caveolin 1 was observed ( Manders\u2019 coef\ufb01cient (cid:3) 0 . 053 ; n (cid:3) 5 ) ( Fig . 3B ) . HCoV - NL63 colocaliza - tion with clathrin and caveolin was also studied in the HAE model . For this , cultures were incubated with gradient - puri\ufb01ed HCoV - NL63 at 32\u00b0C for 2 h ; the virus and the cellular proteins were immunostained and analyzed by confocal microscopy . HCoV - NL63 virions also colocalized with clathrin in this model , whereas no colocalization was observed for caveolin 1 ( Fig . 3 ) . Clathrin and dynamin are important for HCoV - NL63 entry . As we already knew that HCoV - NL63 virions migrate to clathrin - coated pits , in the subsequent step we aimed to determine whether the clathrin - mediated endocytosis is indeed important for the virus entry . For this reason , we blocked the pathway using Pitstop 2 { N - [ 5 - ( 4 - bromobenzylidene ) - 4 - oxo - 4 , 5 - dihydro - 1 , 3 - thiazol - 2 - yl ] naphthalene - 1 - sulfonamide } , a selective clathrin inhibitor targeting its amino - terminal domain , and tetradecyltrim - ethylammonium bromide ( MiTMAB ) , a dynamin I and II GTPase inhibitor . Activity of these compounds was veri\ufb01ed with the positive control , \ufb02uorescently labeled transfer - rin ( 26 , 27 ) . LLC - Mk2 cells were treated with Pitstop 2 , MiTMAB , or control DMSO for 30 min at 37\u00b0C following transferrin uptake for 45 min at 37\u00b0C . Confocal images showed that both inhibitors blocked transferrin endocytosis , as the protein was present only on the cell surface ( Fig . 4A to D ) . Subsequently , LLC - Mk2 cells were incubated with one of the inhibitors at 37\u00b0C for 30 min and inoculated with gradient - puri\ufb01ed HCoV - NL63 at 32\u00b0C for 5 min . Following immunostaining of the HCoV - NL63 N protein and actin , virus endocytosis was analyzed using confocal microscopy . The results showed that virus internalization was hampered in cells pretreated with clathrin and dynamin inhibitors compared to the DMSO - treated cells ( Fig . 4D to G ) . Simultaneously , a cytotoxicity test of the entry inhibitors was performed , which showed no toxic effect of the tested compounds on LLC - Mk2 cells FIG 3 HCoV - NL63 colocalizes with clathrin but not caveolin . ( A and B ) LLC - Mk2 cells were incubated with gradient - puri\ufb01ed HCoV - NL63 for 40 min at 4\u00b0C following 5 min ( A ) or 20 min ( B ) of incubation at 32\u00b0C . HAE cultures were incubated with gradient - puri\ufb01ed HCoV - NL63 for 40 min at 4\u00b0C following 120 min of incubation at 32\u00b0C . HCoV - NL63 colocalization with clathrin ( A ) or caveolin ( B ) was analyzed with confocal microscopy ( HCoV - NL63 , green ; clathrin and caveolin , red ; nuclei , blue ) . ( C ) Cells mock incubated and stained with isotypic antibodies were used as a control . Scale bars (cid:3) 5 (cid:2) m . Entry of HCoV - NL63 Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 5 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m ( Fig . 5 ) . In order to ensure that our observations were not biased , statistical analysis of virus entry was performed . For this , an algorithm was prepared for image analysis , a 3D representation of the cell was prepared , and the virus position in the cell was deter - mined ( Fig . 6 ) . FIG 4 Clathrin and dynamin inhibitors hamper internalization of HCoV - NL63 . ( A to C ) In order to verify the effectiveness of inhibitors , LLC - Mk2 cells were incubated with control DMSO ( A ) , 10 (cid:2) M Pitstop 2 ( B ) , or 10 (cid:2) M MiTMAB ( C ) for 30 min at 37\u00b0C and inoculated with Alexa Fluor 488 - labeled transferrin . Following incubation ( 45 min , 37\u00b0C ) , cells were \ufb01xed and stained for actin ( red ) . Transferrin entry was evaluated with confocal microscopy . ( E to G ) LLC - Mk2 cells were incubated with control DMSO ( E ) , 10 (cid:2) M Pitstop 2 ( F ) , or 10 (cid:2) M MiTMAB ( G ) for 30 min at 37\u00b0C . Cells were inoculated with puri\ufb01ed HCoV - NL63 and incubated at 32\u00b0C for 1 h . Subsequently , cells were \ufb01xed and immunostained for HCoV - NL63 particles ( green ) and actin ( red ) . ( D ) Mock - infected cells were used as a control . Scale bars (cid:3) 10 (cid:2) m . FIG 5 Cytotoxicity of Pitstop 2 and MiTMAB on LLC - Mk2 cells . The cytotoxicity of the endocytosis inhibitors was tested with an XTT assay . Cells were incubated with control DMSO , 10 (cid:2) M Pitstop 2 , or 10 (cid:2) M MiTMAB for 2 h at37\u00b0C Data on the y axis represent viability of the treated cells compared to the untreated reference samples . The assay was performed in triplicate , and average values with standard errors are presented . Milewska et al . Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 6 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m A similar experiment was conducted using HAE cultures . For this , cultures were incubated for 1 h at 37\u00b0C with the inhibitors described above following incubation with gradient - puri\ufb01ed HCoV - NL63 at 32\u00b0C for 2 h . A strong inhibition of virus internalization in cultures preincubated with clathrin or dynamin inhibitors compared to control cells was observed ( Fig . 7 ) . No cytotoxicity to HAE was observed for the tested inhibitors afte r 3 h ofincubation at 37\u00b0C ( Fig . 8 ) . Clathrin - mediated endocytosis is the main entry route for HCoV - NL63 . Even though certain cargo is usually internalized by a single route , frequently other pathways may be used as alternatives . We therefore aimed to test whether inhibition of clathrin - mediated entry with chemical inhibitors results in inhibition of virus replication . To address this , we incubated LLC - Mk2 cells with a given inhibitor at 37\u00b0C for 1 h and infected them with HCoV - NL63 ( TCID 50 (cid:3) 400 per ml ) for 2 h at32\u00b0C . Subsequently , media were removed and cells were washed thrice with acidic buffer following washing with 1 (cid:2) PBS ( pH 7 . 4 ) . Next , cells were overlaid with culture medium containing a given inhibitor and incubated at 32\u00b0C for 4 days . Cells were \ufb01xed and immunostained for HCoV - NL63 N protein to assess the number of infected cells . To assess the nonspeci\ufb01c effect of entry inhibitors , control cells were also treated with these at 4 h p . i . Clearly , in the presence of clathrin - mediated endocytosis inhibitors ( Pitstop 2 and MiTMAB ) , the number of HCoV - NL63 - infected cells was much lower than in the control . However , MiTMAB also inhibited virus replication at later stages of the infection ( Fig . 9 ) . To ensure FIG 6 Numerical image analysis : clathrin and dynamin inhibitors block HCoV - NL63 entry . ( B to D ) LLC - Mk2 cells were incubated with DMSO ( B ) , 10 (cid:2) M MiTMAB ( C ) , or 10 (cid:2) M Pitstop 2 ( D ) for 30 min at 37\u00b0C and subsequently inoculated with puri\ufb01ed HCoV - NL63 and incubated for 45 min at 32\u00b0C . Confocal images were digitalized , and the localization of each virus particle relative to the cellular membrane was assessed . ( A ) Number of internalized virus particles relative to number of virions on the cell surface ( y axis ) for cells treated with DMSO ( control ) , Pitstop 2 , or MiTMAB . In panels B , C , and D , raw data for cells treated with DMSO , Pitstop 2 , or MiTMAB , respectively , are presented . Histograms show the average number of virus particles ( y axis ) versus the distance from the cell surface ( x axis ) . Values of (cid:4) 0 on the x axis indicate that the virus is inside the cell , while for extracellular virions , the x value is (cid:3) 0 . Entry of HCoV - NL63 Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 7 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m that entry inhibitors affected HCoV - NL63 infection in LLC - Mk2 cells , we analyzed by RT - qPCR virus replication at 120 h p . i . in the presence of the tested compounds . The analysis showed an (cid:5) 2 - log decrease in virus progeny production in the presence of Pitstop 2 and MiTMAB compared to that in DMSO - treated cells and a slight increase of RNA copy levels in the presence of nystatin ( Fig . 10A ) . Importantly , no cytotoxic effect was observed for the tested inhibitors applied to LLC - Mk2 cells for 4 days at 32\u00b0C ( Fig . 10B ) . The in\ufb02uence of the tested inhibitors on HCoV - NL63 infection was also analyzed in HAE cultures . For this , cultures were preincubated with a given inhibitor ( Pitstop 2 , MiTMAB , nystatin , or control DMSO ) for 1 h at37\u00b0C and infected with HCoV - NL63 at a TCID 50 of 400 per ml for 2 h at32\u00b0C . Subsequently , noninternalized virions were inactivated by acid washing , and cultures were washed with 1 (cid:2) PBS and incubated with a given inhibitor for 10 min . After that time , supernatants were discarded and cultures were incubated for 5 days at 32\u00b0C . During this period , cultures were incubated with a given inhibitor for 10 min at 32\u00b0C every 24 h . Viral RNA from these samples was quanti\ufb01ed by RT - qPCR . Virus replication in HAE was not affected by any of the tested inhibitors ( Fig . 10A ) . TMPRSS2 is important during early stages of the infection . It was previously suggested that coronaviruses may bypass the endocytic entry route employing trans - membrane protease serine 2 ( TMPRSS2 ) , which primes the fusion protein and enables fusion of viral and cellular membranes on the cell surface ( 28 , 29 ) . We have tested FIG 7 Clathrin and dynamin inhibitors prevent HCoV - NL63 from entering the cell in the HAE model . HAE cultures were incubated with control DMSO ( A ) , 10 (cid:2) M Pitstop 2 ( B ) , or 10 (cid:2) M MiTMAB ( C ) fo r 1 h at37\u00b0C . Cells were then inoculated with puri\ufb01ed HCoV - NL63 and incubated at 32\u00b0C for 2 h . Subsequently , cells were \ufb01xed and immunostained for HCoV - NL63 particles ( green ) , actin ( red ) , and nuclei ( blue ) . Scale bars (cid:3) 5 (cid:2) m . FIG 8 Cytotoxicity of Pitstop 2 and MiTMAB on HAE cultures . The cytotoxicity of the endocytosis inhibitors was tested with an XTT assay . Cells were incubated with control DMSO , 10 (cid:2) M Pitstop 2 , or 10 (cid:2) M MiTMAB for 2 h at37\u00b0C . Data on the y axis represent the viability of the treated cells compared to the untreated reference samples . The assay was performed in triplicate , and average values with standard errors are presented . Milewska et al . Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 8 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m whether inhibition of TMPRSS2 with camostat affects the HCoV - NL63 infection . We observed that inhibition of TMPRSS2 hampers virus infection in HAE cultures , while it has no effect on virus replication in LLC - MK2 cells ( Fig . 11A ) . No inhibition of virus entry was observed in any of the models , as tracked with confocal microscopy visualizing the nucleoprotein ( Fig . 11B ) . As only single entry events per view were observed , several images for camostat - treated and control cells are presented . In total , 500 entry events into HAE cells were tracked , and no difference between the camostat - treated sample and the control sample was noted . HCoV - NL63 entry requires actin remodeling . We studied traf\ufb01cking of HCoV - NL63 inside the cell . As entry by endocytosis would probably require remodeling of the cytoskeleton , we evaluated virus internalization in the presence of cytochalasin D , jasplakinolide , or nocodazole . The \ufb01rst chemical inhibits actin polymerization , whereas the second binds F - actin and stabilizes actin \ufb01laments ( 30 , 31 ) . The last compound interferes with microtubule formation . The analysis showed that actin inhibitors pre - vented virus particles from penetrating the cell , with visible viral particle accumulation on actin cortex or unstructured actin deposits . The microtubule inhibitor did not affect virus entry ( Fig . 12 ) . No cytotoxicity was observed for the tested inhibitors ( Fig . 13 ) . DISCUSSION Previously , we and others described the \ufb01rst steps of the HCoV - NL63 infection process , showing that it begins with the virus binding to the cellular membrane via heparan sulfate proteoglycans , which then enable / facilitate interaction with the entry receptor , ACE2 ( 14 , 18 , 19 ) . Little is known about the subsequent virus internalization FIG 9 Clathrin and dynamin inhibitors limit the number of LLC - Mk2 infected cells . ( A to D ) LLC - Mk2 cells were incubated with control DMSO ( A ) , 5 (cid:2) g / ml nystatin ( B ) , 10 (cid:2) M MiTMAB ( C ) , or 10 (cid:2) M Pitstop 2 ( D ) fo r 1 h at37\u00b0C and inoculated with HCoV - NL63 ( TCID 50 (cid:3) 100 / ml ) . After 2 h ofincubation at 32\u00b0C , virions that were not internalized were inactivated with acidic buffer ( pH 3 ) , and cells were incubated for 4 days at 32\u00b0C in the presence of the tested inhibitors or control DMSO . ( E and F ) The identical procedure was applied to cells , but in these MiTMAB ( E ) and Pitstop 2 ( F ) were applied after the acid wash . Fixed cells were immunostained with anti - NL63 nucleocapsid protein ( green ) and nuclei ( blue ) , and confocal images were collected . Scale bar (cid:3) 200 (cid:2) m . Entry of HCoV - NL63 Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 9 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m and its traf\ufb01cking through the cytoplasm , and some published data are contradic - tory . For example , the role played by cathepsins and acidi\ufb01cation of the microen - vironment during transition of the HCoV - NL63 S protein to its fusogenic form remains unclear . We made an effort to systematically examine every step of the process . First , we tested whether the virus requires endocytosis for successful entry . To do this , we carried out experiments using chemical inhibitors of endosome acidi\ufb01cation ( ammonium chloride and ba\ufb01lomycin A ) . Both blocked virus infection in LLC - MK2 cells , suggesting a requirement for transport of virions to endosomes , which then undergo acidi\ufb01cation . However , such an approach may have several disadvantages . First , we examined the role of endosome acidi\ufb01cation based on virus replication ; thus , we cannot rule out interference with virus infection at later stages ( as shown for MiTMAB ) . Second , the speci\ufb01city and selectivity of chemical inhibitors are questionable . An indirect proof of the pH dependence of HCoV - NL63 entry may be provided by the fact that acidi\ufb01cation of the environment ( acid wash ) results in inactivation of the virus , suggesting a pH - directed structural switch in the S protein . To further con\ufb01rm our observations , we FIG 10 Clathrin and dynamin inhibitors hamper replication of HCoV - NL63 in LLC - MK2 cells . ( A ) HCoV - NL63 replication in LLC - Mk2 cells and HAE cultures in the presence of entry inhibitors or control DMSO was analyzed using RT - qPCR . Cultures were incubated with 10 (cid:2) M Pitstop 2 , 10 (cid:2) M MiTMAB , 5 (cid:2) g / ml nystatin , or DMSO for 1 h at37\u00b0C and inoculated with HCoV - NL63 ( TCID 50 (cid:3) 400 / ml ) . After 2 h of incubation at 32\u00b0C , virions that were not internalized were inactivated with acidic buffer ( pH 3 ) , and cells were incubated for 5 days at 32\u00b0C . The data are presented as log reduction value ( LRV ) compared to the control sample . The assay was performed in triplicate , and average values with standard errors are presented . P values of (cid:4) 0 . 05 were considered signi\ufb01cant and are denoted with an asterisk . ( B ) The cytotoxicity of the inhibitors was tested with an XTT assay . Cells were incubated with 10 (cid:2) M Pitstop 2 , 10 (cid:2) M MiTMAB , 5 (cid:2) g / ml nystatin , or DMSO for 5 days at 32\u00b0C . Data on the y axis represent viability of the treated cells compared to the untreated reference samples . The assay was performed in triplicate , and average values with standard errors are presented . Milewska et al . Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 10 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m developed a method of visualizing single virions as they entered the cell . Efforts to stain for virus surface proteins yielded poor results , most likely due to the lack of highly speci\ufb01c antibodies and posttranslational modi\ufb01cation of surface proteins , and the best results were obtained when antibodies speci\ufb01c to the N protein were used . Incubation of cells with puri\ufb01ed virions resulted in virus attachment , which was visualized by confocal microscopy and costaining for markers of the most commonly employed endocytic pathways and allowed us to study the colocaliza - tion . If signi\ufb01cant colocalization was detected , the results were con\ufb01rmed with chemical inhibitors . The results showed that HCoV - NL63 binding to ACE2 initiates recruitment of clathrin and subsequent formation of clathrin - coated pits ; no colocalization of the virus with other markers ( e . g . , caveolin ) was noted . Transferrin was used as a positive control for clathrin - mediated endocytosis ( 32 , 33 ) . Importantly , chemical inhibitors of clathrin completely blocked virus internalization , and the virus remained on the cell surface . Analysis of HAE cultures yielded identical results . The inhibitors of endocytosis also hampered virus infection on LLC - Mk2 cells , highlighting that this pathway is relevant and the lack of an equally effective alternative entry route in this culture model . Clathrin - mediated endocytosis requires a number of other proteins , such as dynamin , the GTPase responsible for scission of clathrin - coated vesicles from the cell surface ( 34 ) . Inhibiting dynamin also hampered virus internalization into LLC - MK2 cells and HAE cultures , con\ufb01rming our previous observations . However , in this case the MiTMAB compound blocked replication of HCoV - NL63 also during subsequent stages of the infection . It is noteworthy that we were not able to block virus infection of HAE cultures using inhibitors of endocytosis . This may be related to the fact that the cultures were exposed to inhibitors for a very short time during apical washes , which is not suf\ufb01cient to permanently block the infection . On the other hand , it is also possible that in HAE , HCoV - NL63 is able to enter the cell by an alternative route . Recent reports on other coronaviruses ( 28 , 29 , 35 ) suggested that these viruses may bypass the endocytic entry route using TMPRRS2 as the priming protease , enabling entry directly from the cell surface . Our experiments showed that inhibition of this protease indeed inhibited virus infection . Interestingly , it did not hamper virus internalization into the cell . Our data are consistent with the data presented by others ( 28 , 29 , 35 ) , yet we believe that there is a different mechanistic explanation for the observed phenomenon . We believe that FIG 11 TMPRSS2 is required for entry into HAE cells but does not enable virus - cell fusion on the cell surface . ( A ) HCoV - NL63 replication in LLC - Mk2 cells and HAE cultures in the presence of camostat or control DMSO was analyzed using RT - qPCR . Cultures were incubated with 100 (cid:2) M camostat or DMSO fo r 1 h at37\u00b0C and inoculated with HCoV - NL63 ( TCID 50 (cid:3) 400 / ml ) . After 2 h ofincubation at 32\u00b0C , virions that were not internalized were inactivated with acidic buffer ( pH 3 ) , and cells were incubated for 5 days at 32\u00b0C . The data are presented as log reduction value ( LRV ) compared to the control sample . The assay was performed in triplicate , and average values with standard errors are presented . P values of (cid:4) 0 . 05 were considered signi\ufb01cant and are denoted with asterisks . ( B ) HAE cultures were incubated with control DMSO or 100 (cid:2) M camostat for 1 h at37\u00b0C . Further , cells were inoculated with puri\ufb01ed HCoV - NL63 and incubated at 32\u00b0C for 2 h . Subsequently , cells were \ufb01xed and immunostained for HCoV - NL63 particles ( green ) , actin ( red ) , and nuclei ( blue ) . Scale bars (cid:3) 5 (cid:2) m . Entry of HCoV - NL63 Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 11 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m TMPRRS2 indeed is required for the virus - cell fusion , acting similarly to cathepsins , but it does not enable fusion on the cell surface , and the acidi\ufb01cation of the microenvi - ronment is required . Our \ufb01nal research question was about virus traf\ufb01cking . The endosome typically translocates through the depolymerizing actin cortex and is subsequently sorted at the endosomal hub and directed to different destinations . This sorting is highly dependent on the cargo . Using two chemical inhibitors ( jasplakinolide and cytochalasin B ) ( 31 , 36 ) , we showed that actin plays a vital role in virus entry . Stabilization of the actin cortex using jasplakinolide resulted in immobilization of the virus at the cell surface , similarly to the case for inhibition of actin polymerization using cytochalasin D . These two experiments suggest a scenario in which virus - carrying endosomes pass along the actin cortex , which actively unwinds and interacts with virions . In summary , we show that HCoV - NL63 enters the cell by clathrin - mediated endo - cytosis , but the pathway may be bypassed to some extent during infection ex vivo . HCoV - NL63 entry into the susceptible cell is summarized in Fig . 14 . FIG 12 Actin is important for HCoV - NL63 entry . LLC - MK2 cells were incubated with DMSO ( A ) , 10 (cid:2) M cytochalasin D ( B and E ) , 1 . 5 (cid:2) M jasplakinolide ( C and F ) , or 400 nM nocodazole ( D and G ) fo r 1 h at37\u00b0C and then inoculated with puri\ufb01ed HCoV - NL63 and incubated at 32\u00b0C for 1 h . Actin and virus localization was veri\ufb01ed with confocal microscopy ; \ufb01xed cells were immunostained for HCoV - NL63 particles ( green ) , actin ( red ) , and nuclei ( blue ) . Scale bars (cid:3) 10 (cid:2) m . Milewska et al . Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 12 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m MATERIALS AND METHODS Cell culture . LLC - Mk2 cells ( ATCC CCL - 7 ; Macaca mulatta kidney epithelial ) were maintained in minimal essential medium ( MEM ) ( two parts Hanks\u2019 MEM and one part Earle\u2019s MEM ; Thermo Fisher Scienti\ufb01c , Poland ) supplemented with 3 % heat - inactivated fetal bovine serum ( Thermo Fisher Scienti\ufb01c , Poland ) , penicillin ( 100 U / ml ) , streptomycin ( 100 (cid:2) g / ml ) , and cipro\ufb02oxacin ( 5 (cid:2) g / ml ) . Cells were cultured at 37\u00b0C under 5 % CO 2 . Ethics statement . Human tracheobronchial epithelial cells were obtained from airway specimens resected from adult patients undergoing surgery under Silesian Center for Heart Diseases - approved protocols . This study was approved by the Bioethical Committee of the Medical University of Silesia in Katowice , Poland ( approval no . KNW / 0022 / KB1 / 17 / 10 dated 16 February 2010 ) . Participants provided their written informed consent to participate in the study , as approved by the Bioethical Committee . HAE cultures . Primary human tracheobronchial epithelial cells were expanded on plastic to generate passage 1 cells and plated on permeable Transwell insert ( 6 . 5 - mm - diameter ) supports . Human airway epithelium ( HAE ) cultures were generated by provision of an air - liquid interface for 6 to 8 weeks to form well - differentiated , polarized cultures that resemble in vivo pseudostrati\ufb01ed mucociliary epithelium . Cultures were prepared and maintained as previously described ( 24 ) . Cell viability assay . LLC - Mk2 cells were cultured on 96 - well plates , and HAE cultures were prepared as described above . Cell viability assay was performed by using the 2 , 3 - bis - ( 2 - methoxy - 4 - nitro - 5 - sulfophenyl ) - 2H - tetrazolium - 5 - carboxanilide salt ( XTT ) cell viability assay ( Biological Industries , Israel ) according to the manufacturer\u2019s instructions . Brie\ufb02y , on the day of the assay , 100 (cid:2) l of the culture medium ( for LLC - Mk2 ) or 1 (cid:2) PBS ( for HAE ) with 30 (cid:2) l of the activated XTT solution was added to each well / culture insert . Following 2 h ofincubation at 37\u00b0C , the solution was transferred onto a 96 - well plate FIG 13 Cytotoxicity of the cytoskeleton - modifying compounds . The cytotoxicity of the endocytosis inhibitors was tested with an XTT assay . Cells were incubated with DMSO , 10 (cid:2) M cytochalasin D , 1 . 5 (cid:2) M jasplakinolide , or 400 nM nocodazole fo r 2 h at37\u00b0C . Data on the y axis represent viability of the treated cells compared to the untreated reference samples . The assay was performed in triplicate , and average values with standard errors are presented . FIG 14 Early events during HCoV - NL63 infection . Entry of HCoV - NL63 Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 13 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m and the signal was measured at (cid:4) (cid:3) 490 nm using a colorimeter ( Spectra Max 250 ; Molecular Devices ) . The results obtained were further normalized to the control sample , for which cell viability was set to 100 % . Virus preparation and titration . The HCoV - NL63 stock ( isolate Amsterdam 1 ) was generated by infecting monolayers of LLC - Mk2 cells . The virus - containing liquid was aliquoted and stored at (cid:6) 80\u00b0C . A control LLC - Mk2 cell lysate from mock - infected cells was prepared in the same manner . The virus yield was assessed by titration on fully con\ufb02uent LLC - Mk2 cells in 96 - well plates according to the method described by Reed and Muench ( 37 ) . Puri\ufb01cation of HCoV - NL63 . The virus stock was concentrated 25 - fold using centrifugal protein concentrators ( Amicon Ultra , 10 - kDa cutoff ; Merck , Poland ) and subsequently overlaid on 15 % iodixanol solution in 1 (cid:2) PBS ( OptiPrep medium ; Sigma - Aldrich , Poland ) . Following virus concentration using an iodixanol cushion ( centrifugation at 175 , 000 (cid:2) g for 3 h at4\u00b0C ) , it was overlaid on a 10 to 20 % iodixanol gradient in 1 (cid:2) PBS and centrifuged at 175 , 000 (cid:2) g for 18 h at 4\u00b0C . Fractions ( 1 ml ) collected from the gradient were analyzed by Western blotting using anti - HCoV - NL63 N IgGs ( 0 . 25 (cid:2) g / ml ; Ingenansa , Spain ) and a secondary antibody coupled with horseradish peroxidase ( 65 ng / ml ; Dako , Denmark ) . The virus - containing fractions were aliquoted and stored at (cid:6) 80\u00b0C . The control cell lysate ( mock ) was concentrated and prepared in the same manner as the virus stock . Inhibition of virus entry . LLC - Mk2 cells were seeded on coverslips in six - well plates ( TPP , Switzer - land ) and cultured for 2 days at 37\u00b0C . Subsequently , cells were incubated with a given inhibitor for 30 min at 37\u00b0C and later with 50 (cid:2) l of puri\ufb01ed HCoV - NL63 or mock sample for 1 h at 32\u00b0C . For the ex vivo experiment , HAE cultures were exposed to the tested inhibitor or control PBS for 1 h at37\u00b0C following inoculation with iodixanol - concentrated HCoV - NL63 or mock sample . Followin g 2 h ofincubation at 32\u00b0C , unbound virions were removed by washing with 1 (cid:2) PBS . Cells were then washed with 1 (cid:2) PBS and \ufb01xed with 4 % paraformaldehyde ( PFA ) . Transferrin and albumin were used as positive controls , as they were previously described to serve as a cargo in clathrin - and caveolin - dependent endocytosis , respectively ( 38 , 39 ) . LLC - Mk2 cells were seeded on coverslips in six - well plates ( TPP , Switzerland ) and cultured for 2 days at 37\u00b0C . Subsequently , cells were incubated with a given inhibitor for 30 min at 37\u00b0C following incubation with Alexa Fluor 488 - labeled transferrin ( 100 (cid:2) g / ml ; Molecular Probes ) , \ufb02uorescein isothiocyanate ( FITC ) - labeled albumin ( 500 (cid:2) g / ml ; Sigma - Aldrich , Poland ) , or control PBS for 45 min at 32\u00b0C . Cells were then washed with 1 (cid:2) PBS and \ufb01xed in 4 % PFA . Immunostaining and confocal imaging . Fixed cells were permeabilized with 0 . 1 % Triton X - 100 in 1 (cid:2) PBS and incubated overnight at 4\u00b0C in 1 (cid:2) PBS supplemented with 5 % bovine serum albumin ( BSA ) and 0 . 5 % Tween 20 . To visualize HCoV - NL63 particles , cells were incubated fo r 2 h atroom temperature with mouse anti - HCoV - NL63 N IgGs ( 0 . 25 (cid:2) g / ml ; Ingenansa , Spain ) , followed b y 1 h ofincubation with Alexa Fluor 488 - labeled goat anti - mouse IgG ( 2 . 5 (cid:2) g / ml ; Thermo Fisher Scienti\ufb01c , Poland ) . The following antibodies were used for endosomal markers : polyclonal goat anti - human clathrin HC coupled with tetramethylrhodamine ( 10 (cid:2) g / ml ; Santa Cruz Biotechnology ) , polyclonal rabbit anti - human early endo - some antigen 1 ( 2 (cid:2) g / ml ; Santa Cruz Biotechnology ) , polyclonal rabbit anti - human caveolin 1 ( 2 (cid:2) g / ml ; Sigma - Aldrich , Poland ) , and Alexa Fluor 633 - labeled goat anti - rabbit ( 2 . 5 (cid:2) g / ml ; Thermo Fisher Scienti\ufb01c , Poland ) . Actin \ufb01laments was stained using phalloidin coupled with Alexa Fluor 633 ( 0 . 2 U / ml ; Thermo Fisher Scienti\ufb01c , Poland ) . Nuclear DNA was stained with DAPI ( 4 = , 6 = - diamidino - 2 - phenylindole ) ( 0 . 1 (cid:2) g / ml ; Sigma - Aldrich , Poland ) . Immunostained cultures were mounted on glass slides in ProLong Gold antifade medium ( Thermo Fisher Scienti\ufb01c , Poland ) . Fluorescent images were acquired under a Leica TCS SP5 II confocal microscope ( Leica Microsystems GmbH , Mannheim , Germany ) and a Zeiss LSM 710 confocal microscope ( Carl Zeiss Microscopy GmbH ) . Images were acquired using Leica Application Suite Advanced Fluorescence LAS AF v . 2 . 2 . 1 ( Leica Microsystems CMS GmbH ) or ZEN 2012 SP1 software ( Carl Zeiss Microscopy GmbH ) deconvolved with Huygens Essential package version 4 . 4 ( Scienti\ufb01c Volume Imaging B . V . , The Netherlands ) and processed using ImageJ 1 . 47v ( National Institutes of Health , Be - thesda , MD , USA ) . Flow cytometry . LLC - Mk2 cells were seeded on 6 - well plates ( TPP ) and cultured for 2 days at 37\u00b0C with 5 % CO 2 . Cells in monolayer were incubated with each entry inhibitor fo r 1 h at37\u00b0C following infection with HCoV - NL63 at a TCID 50 of 100 / ml or inoculation of the mock sample . On day 4 p . i . , cells were washed with sterile PBS , \ufb01xed with 3 % PFA , permeabilized with 0 . 1 % Triton X - 100 in 1 (cid:2) PBS , and incubated for 1 h with 3 % BSA in 1 (cid:2) PBS with 0 . 1 % Tween 20 . To quantify HCoV - NL63 infection , \ufb01xed cells were scraped from the plastic and incubated fo r 2 h atroom temperature with mouse anti - HCoV - NL63 N IgG antibodies ( 1 (cid:2) g / ml ; Ingenansa ) , followed b y 1 h ofincubation with Alexa Fluor 488 - labeled goat anti - mouse antibody ( 2 . 5 (cid:2) g / ml ; Molecular Probes ) . Cells were then washed , resuspended in 1 (cid:2) PBS , and analyzed with a FACSCalibur instrument ( Becton Dickinson ) using Cell Quest software . Isolation of nucleic acids and reverse transcription . Viral nucleic acids were isolated from cell culture supernatants ( LLC - Mk2 cells ) or apical washes ( HAE cultures ) using the viral RNA / DNA isolation kit ( A & A Biotechnology , Poland ) according to the manufacturer\u2019s instructions . Reverse transcription was carried out with a high - capacity cDNA reverse transcription kit ( Thermo Fisher Scienti\ufb01c , Poland ) , according to the manufacturer\u2019s instructions . RT - qPCR . The HCoV - NL63 yield was determined by RT - qPCR ( 7500 Fast Real - Time PCR machine ; Life Technologies , Poland ) . Viral cDNA ( 2 . 5 (cid:2) l per sample ) was ampli\ufb01ed in a 10 - (cid:2) l reaction mixture containing 1 (cid:2) master mix ( RT Mix Probe ; A & A Biotechnology , Poland ) , a speci\ufb01c probe labeled with 6 - carboxy\ufb02uorescein ( FAM ) and 6 - carboxytetramethylrhodamine ( TAMRA ) ( 100 nM ) ( 5 = - ATG TTA TTC AGT GCT TTG GTC CTC GTG AT - 3 = ) , and primers ( 450 nM each ) ( sense , 5 = - CTG TGG AAA ACC TTT GGC ATC - 3 = ; antisense , 5 = - CTG TGG AAA ACC TTT GGC ATC - 3 = ) . Rox was used as the reference dye . The reaction Milewska et al . Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 14 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m conditions were as follows : 2 min at 50\u00b0C and 10 min at 92\u00b0C , followed by 40 cycles of 15 s at 92\u00b0C and 1 min at 60\u00b0C . In order to assess the copy number for the N gene , DNA standards were prepared . Brie\ufb02y , the N gene of HCoV - NL63 was ampli\ufb01ed and cloned into pTZ57R / T ( Thermo Fisher Scienti\ufb01c , Poland ) plasmid using the InsTAclone PCR cloning kit ( Thermo Fisher Scienti\ufb01c , Poland ) . Subsequently , DNA vectors were ampli\ufb01ed and linearized with EcoRI restriction enzyme . Linear nucleic acids were further puri\ufb01ed with the GeneJET PCR puri\ufb01cation kit ( Thermo Fisher Scienti\ufb01c , Poland ) according to the manufacturer\u2019s instructions , and its concentration was assessed using a spectrophotometer . The number of DNA copies per milliliter was assessed using Avogadro\u2019s constant and the molecular mass of RNA molecules . Samples were serially diluted and used as an input for real - time PCR . In this article , the data from quantitative PCR are presented as log removal values ( LRVs ) in order to enable comparison of results obtained from different assays . The LRV was calculated according to the formula LRV (cid:3) (cid:6) log ( c i / c 0 ) , where c i is the number of viral RNA copies per milliliter in the sample from the culture treated with a given polymer and c 0 is the number of viral RNA copies per milliliter in the control sample ( untreated cells ) . Image analysis . To evaluate the infection inhibition in the presence of various endocytosis inhibitors , image analysis was performed on 2 - mm by 2 - mm tile scan images . On each image , the number of nuclei ( expressed as a number of cells ) and the mean pixel intensity for the virus were calculated . For that , histograms of all images were adjusted to the minimum / maximum value , excluding signal from the virus derived from images with no infected cells . Results are presented as mean intensity of \ufb02uorescence per cell . Colocalization analyses were performed under ImageJ using the JACoP plugin ( 41 ) , where Manders\u2019 coef\ufb01cient was calculated for 3D images of more than 5 cells . Quantitative analysis of virus internalization in the presence of inhibitors was performed with the algorithm previously described by Berniak et al . with modi\ufb01cations ( 40 ) . The cell surface was estimated on each image slice manually using the polygon selection tool in ImageJ , and based on this information , the 3D cell surface was modeled . Coordinates of virus particles were determined using the 3D Object Counter ImageJ plugin . The relative localization and distance between the virus particle and cell surface were calculated . Results are presented as a ratio between virus particles inside a cell and the particles on the surface ( up to 1 . 5 (cid:2) m above ) . Statistical analysis . All the experiments were performed in triplicate , and the results are presented as mean (cid:7) standard deviation ( SD ) . To determine the signi\ufb01cance of the obtained results , a comparison between groups was conducted using the Student t test . P values of (cid:4) 0 . 05 were considered signi\ufb01cant . ACKNOWLEDGMENTS This work was supported by grants from the National Science Center ( UMO - 2012 / 07 / E / NZ6 / 01712 , UMO - 2012 / 07 / N / NZ6 / 02955 , and 2015 / 19 / N / NZ1 / 00323 to K . P . , A . M . , and K . B . , respectively ) and the Ministry of Science and Higher Education ( 0058 / DIA / 2015 / 44 to P . N . ) . K . P . acknowledges a networking contribution from the COST Action CM1407 \u201cChallenging Organic Syntheses Inspired by Nature\u2014from Natural Products Chemistry to Drug Discovery . \u201d The Faculty of Biochemistry , Biophysics and Biotechnol - ogy of Jagiellonian University is a bene\ufb01ciary of structural funds from the European Union ( grant no . POIG . 02 . 01 . 00 - 12 - 064 / 08 , \u201cMolecular Biotechnology for Health\u201d ) and is a partner of the Leading National Research Center supported by the Ministry of Science and Higher Education of the Republic of Poland . The funders had no role in study design , data collection and analysis , decision to publish , or preparation of the manuscript . REFERENCES 1 . van der Hoek L , Pyrc K , Jebbink MF , Vermeulen - Oost W , Berkhout RJ , Wolthers KC , Wertheim - van Dillen PM , Kaandorp J , Spaargaren J , Berk - hout B . 2004 . Identi\ufb01cation of a new human coronavirus . Nat Med 10 : 368 \u2013 373 . https : / / doi . org / 10 . 1038 / nm1024 . 2 . Konca C , Korukluoglu G , Tekin M , Almis H , Bucak I\u02d9 Uygun H , Altas AB , Bayrakdar F . 2017 . The \ufb01rst infant death associated with human Coro - navirus NL63 infection . Pediatr Infect Dis J 36 : 231 \u2013 233 . https : / / doi . org / 10 . 1097 / INF . 0000000000001390 . 3 . Mayer K , Nellessen C , Hahn - Ast C , Schumacher M , Pietzonka S , Eis - H\u00fcbinger AM , Drosten C , Brossart P , Wolf D . 2016 . Fatal outcome of human coronavirus NL63 infection despite successful viral elimination by IFN - alpha in a patient with newly diagnosed ALL . Eur J Haematol 97 : 208 \u2013 210 . https : / / doi . org / 10 . 1111 / ejh . 12744 . 4 . Cabe\u00e7a TK , Bellei N . 2012 . Human coronavirus NL - 63 infection in a Brazilian patient suspected of H1N1 2009 in\ufb02uenza infection : description of a fatal case . J Clin Virol 53 : 82 \u2013 84 . https : / / doi . org / 10 . 1016 / j . jcv . 2011 . 09 . 006 . 5 . Oosterhof L , Christensen CB , Sengel\u00f8v H . 2010 . Fatal lower respiratory tract disease with human coronavirus NL63 in an adult haematopoietic cell transplant recipient . Bone Marrow Transplant 45 : 1115 \u2013 1116 . https : / / doi . org / 10 . 1038 / bmt . 2009 . 292 . 6 . Fouchier RA , Hartwig NG , Bestebroer TM , Niemeyer B , de Jong JC , Simon JH , Osterhaus AD . 2004 . A previously undescribed coronavirus associ - ated with respiratory disease in humans . Proc Natl Acad Sci U S A 101 : 6212 \u2013 6216 . https : / / doi . org / 10 . 1073 / pnas . 0400762101 . 7 . Pyrc K , Berkhout B , van der Hoek L . 2007 . The novel human coronavi - ruses NL63 and HKU1 . J Virol 81 : 3051 \u2013 3057 . https : / / doi . org / 10 . 1128 / JVI . 01466 - 06 . 8 . van der Hoek L , Sure K , Ihorst G , Stang A , Pyrc K , Jebbink MF , Petersen G , Forster J , Berkhout B , Uberla K . 2005 . Croup is associated with the novel coronavirus NL63 . PLoS Med 2 : e240 . https : / / doi . org / 10 . 1371 / journal . pmed . 0020240 . 9 . van der Hoek L , Pyrc K , Berkhout B . 2006 . Human coronavirus NL63 , a new respiratory virus . FEMS Microbiol Rev 30 : 760 \u2013 773 . https : / / doi . org / 10 . 1111 / j . 1574 - 6976 . 2006 . 00032 . x . 10 . Pyrc K , Dijkman R , Deng L , Jebbink MF , Ross HA , Berkhout B , van der Entry of HCoV - NL63 Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 15 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m Hoek L . 2006 . Mosaic structure of human coronavirus NL63 , one thou - sand years of evolution . J Mol Biol 364 : 964 \u2013 973 . https : / / doi . org / 10 . 1016 / j . jmb . 2006 . 09 . 074 . 11 . Huynh J , Li S , Yount B , Smith A , Sturges L , Olsen JC , Nagel J , Johnson JB , Agnihothram S , Gates JE , Frieman MB , Baric RS , Donaldson EF . 2012 . Evidence supporting a zoonotic origin of human coronavirus strain NL63 . J Virol 86 : 12816 \u2013 12825 . https : / / doi . org / 10 . 1128 / JVI . 00906 - 12 . 12 . Tao Y , Shi M , Chommanard C , Queen K , Zhang J , Markotter W , Kuzmin IV , Holmes EC , Tong S . 2017 . Surveillance of bat coronaviruses in Kenya identi\ufb01es relatives of human coronaviruses NL63 and 229E and their recombination history . J Virol 91 : e01953 - 16 . https : / / doi . org / 10 . 1128 / JVI . 01953 - 16 . 13 . Fields BN , Knipe DM , Howley PM ( ed ) . 2013 . Fields virology , 6th ed . Lippincott Williams & Wilkins , Philadelphia , PA . 14 . Hofmann H , Pyrc K , van der Hoek L , Geier M , Berkhout B , P\u00f6hlmann S . 2005 . Human coronavirus NL63 employs the severe acute respiratory syndrome coronavirus receptor for cellular entry . Proc Natl Acad Sci U S A 102 : 7988 \u2013 7993 . https : / / doi . org / 10 . 1073 / pnas . 0409465102 . 15 . Li W , Moore MJ , Vasilieva N , Sui J , Wong SK , Berne MA , Somasundaran M , Sullivan JL , Luzuriaga K , Greenough TC , Choe H , Farzan M . 2003 . Angiotensin - converting enzyme 2 is a functional receptor for the SARS coronavirus . Nature 426 : 450 \u2013 454 . https : / / doi . org / 10 . 1038 / nature02145 . 16 . Wickramasinghe IN , de Vries RP , Gr\u00f6ne A , de Haan CA , Verheije MH . 2011 . Binding of avian coronavirus spike proteins to host factors re\ufb02ects virus tropism and pathogenicity . J Virol 85 : 8903 \u2013 8912 . https : / / doi . org / 10 . 1128 / JVI . 05112 - 11 . 17 . Promkuntod N , Wickramasinghe IN , de Vrieze G , Gr\u00f6ne A , Verheije MH . 2013 . Contributions of the S2 spike ectodomain to attachment and host range of infectious bronchitis virus . Virus Res 177 : 127 \u2013 137 . https : / / doi . org / 10 . 1016 / j . virusres . 2013 . 09 . 006 . 18 . P\u00f6hlmann S , Gramberg T , Wegele A , Pyrc K , van der Hoek L , Berkhout B , Hofmann H . 2006 . Interaction between the spike protein of human coronavirus NL63 and its cellular receptor ACE2 . Adv Exp Med Biol 581 : 281 \u2013 284 . https : / / doi . org / 10 . 1007 / 978 - 0 - 387 - 33012 - 9 _ 47 . 19 . Milewska A , Zarebski M , Nowak P , Stozek K , Potempa J , Pyrc K . 2014 . Human coronavirus NL63 utilizes heparan sulfate proteoglycans for attachment to target cells . J Virol 88 : 13221 \u2013 13230 . https : / / doi . org / 10 . 1128 / JVI . 02078 - 14 . 20 . Mathewson AC , Bishop A , Yao Y , Kemp F , Ren J , Chen H , Xu X , Berkhout B , van der Hoek L , Jones IM . 2008 . Interaction of severe acute respiratory syndrome - coronavirus and NL63 coronavirus spike proteins with angio - tensin converting enzyme - 2 . J Gen Virol 89 : 2741 \u2013 2745 . https : / / doi . org / 10 . 1099 / vir . 0 . 2008 / 003962 - 0 . 21 . Yoshimori T , Yamamoto A , Moriyama Y , Futai M , Tashiro Y . 1991 . Ba\ufb01 - lomycin A1 , a speci\ufb01c inhibitor of vacuolar - type H ( (cid:8) ) - ATPase , inhibits acidi\ufb01cation and protein degradation in lysosomes of cultured cells . J Biol Chem 266 : 17707 \u2013 17712 . 22 . Misinzo G , Delputte PL , Nauwynck HJ . 2008 . Inhibition of endosome - lysosome system acidi\ufb01cation enhances porcine circovirus 2 infection of porcine epithelial cells . J Virol 82 : 1128 \u2013 1135 . https : / / doi . org / 10 . 1128 / JVI . 01229 - 07 . 23 . Richards AL , Jackson WT . 2012 . Intracellular vesicle acidi\ufb01cation pro - motes maturation of infectious poliovirus particles . PLoS Pathog 8 : e1003046 . https : / / doi . org / 10 . 1371 / journal . ppat . 1003046 . 24 . Milewska A , Ciejka J , Kaminski K , Karewicz A , Bielska D , Zeglen S , Karolak W , Nowakowska M , Potempa J , Bosch BJ , Pyrc K , Szczubialka K . 2013 . Novel polymeric inhibitors of HCoV - NL63 . Antiviral Res 97 : 112 \u2013 121 . https : / / doi . org / 10 . 1016 / j . antiviral . 2012 . 11 . 006 . 25 . Mu FT , Callaghan JM , Steele - Mortimer O , Stenmark H , Parton RG , Camp - bell PL , McCluskey J , Yeo JP , Tock EP , Toh BH . 1995 . EEA1 , an early endosome - associated protein . EEA1 is a conserved alpha - helical periph - eral membrane protein \ufb02anked by cysteine \u201c\ufb01ngers\u201d and contains a calmodulin - binding IQ motif . J Biol Chem 270 : 13503 \u2013 13511 . 26 . Dutta D , Williamson CD , Cole NB , Donaldson JG . 2012 . Pitstop 2 is a potent inhibitor of clathrin - independent endocytosis . PLoS One 7 : e45799 . https : / / doi . org / 10 . 1371 / journal . pone . 0045799 . 27 . Joshi S , Perera S , Gilbert J , Smith CM , Mariana A , Gordon CP , Sakoff JA , McCluskey A , Robinson PJ , Braithwaite AW , Chircop M . 2010 . The dy - namin inhibitors MiTMAB and OcTMAB induce cytokinesis failure and inhibit cell proliferation in human cancer cells . Mol Cancer Ther 9 : 1995 \u2013 2006 . https : / / doi . org / 10 . 1158 / 1535 - 7163 . MCT - 10 - 0161 . 28 . Shirato K , Kanou K , Kawase M , Matsuyama S . 2017 . Clinical isolates of human coronavirus 229E bypass the endosome for cell entry . J Virol 91 : e01387 - 16 . https : / / doi . org / 10 . 1128 / JVI . 01387 - 16 . 29 . Reinke LM , Spiegel M , Plegge T , Hartleib A , Nehlmeier I , Gierer S , Hoffmann M , Hofmann - Winkler H , Winkler M , P\u00f6hlmann S . 2017 . Differ - ent residues in the SARS - CoV spike protein determine cleavage and activation by the host cell protease TMPRSS2 . PLoS One 12 : e0179177 . https : / / doi . org / 10 . 1371 / journal . pone . 0179177 . 30 . Casella JF , Flanagan MD , Lin S . 1981 . Cytochalasin D inhibits actin polymerization and induces depolymerization of actin \ufb01laments formed during platelet shape change . Nature 293 : 302 \u2013 305 . https : / / doi . org / 10 . 1038 / 293302a0 . 31 . Holzinger A . 2009 . Jasplakinolide : an actin - speci\ufb01c reagent that pro - motes actin polymerization . Methods Mol Biol 586 : 71 \u2013 87 . https : / / doi . org / 10 . 1007 / 978 - 1 - 60761 - 376 - 3 _ 4 . 32 . Hopkins CR , Miller K , Beardmore JM . 1985 . Receptor - mediated endocy - tosis of transferrin and epidermal growth factor receptors : a comparison of constitutive and ligand - induced uptake . J Cell Sci Suppl 3 : 173 \u2013 186 . https : / / doi . org / 10 . 1242 / jcs . 1985 . Supplement _ 3 . 17 . 33 . Warren RA , Green FA , Enns CA . 1997 . Saturation of the endocytic path - way for the transferrin receptor does not affect the endocytosis of the epidermal growth factor receptor . J Biol Chem 272 : 2116 \u2013 2121 . https : / / doi . org / 10 . 1074 / jbc . 272 . 4 . 2116 . 34 . Ferguson SM , De Camilli P . 2012 . Dynamin , a membrane - remodelling GTPase . Nat Rev Mol Cell Biol 13 : 75 \u2013 88 . https : / / doi . org / 10 . 1038 / nrm3266 . 35 . Park JE , Li K , Barlan A , Fehr AR , Perlman S , McCray PB , Gallagher T . 2016 . Proteolytic processing of Middle East respiratory syndrome coronavirus spikes expands virus tropism . Proc Natl Acad Sc i U S A113 : 12262 \u2013 12267 . https : / / doi . org / 10 . 1073 / pnas . 1608147113 . 36 . MacLean - Fletcher S , Pollard TD . 1980 . Mechanism of action of cytocha - lasin B on actin . Cell 20 : 329 \u2013 341 . https : / / doi . org / 10 . 1016 / 0092 - 8674 ( 80 ) 90619 - 4 . 37 . Reed L , Muench H . 1938 . A simple method of estimating \ufb01fty percent endpoints . Am J Epidemiol 27 : 493 \u2013 497 . https : / / doi . org / 10 . 1093 / oxford journals . aje . a118408 . 38 . Harding C , Heuser J , Stahl P . 1983 . Receptor - mediated endocytosis of transferrin and recycling of the transferrin receptor in rat reticulocytes . J Cell Biol 97 : 329 \u2013 339 . https : / / doi . org / 10 . 1083 / jcb . 97 . 2 . 329 . 39 . Schubert W , Frank PG , Razani B , Park DS , Chow CW , Lisanti MP . 2001 . Caveolae - de\ufb01cient endothelial cells show defects in the uptake and transport of albumin in vivo . J Biol Chem 276 : 48619 \u2013 48622 . https : / / doi . org / 10 . 1074 / jbc . C100613200 . 40 . Berniak K , Rybak P , Bernas T , Zare\u02dbbski M , Biela E , Zhao H , Darzynkiewicz Z , Dobrucki JW . 2013 . Relationship between DNA damage response , initiated by camptothecin or oxidative stress , and DNA replication , analyzed by quantitative 3D image analysis . Cytometry A 83 : 913 \u2013 924 . https : / / doi . org / 10 . 1002 / cyto . a . 22327 . 41 . Bolte S , Cordeli\u00e8res FP . 2006 . A guided tour into subcellular colocaliza - tion analysis in light microscopy . J Microsc 224 : 213 \u2013 232 . https : / / doi . org / 10 . 1111 / j . 1365 - 2818 . 2006 . 01706 . x . Milewska et al . Journal of Virology February 2018 Volume 92 Issue 3 e01933 - 17 jvi . asm . org 16 on J anua r y 23 , 2018 b y ABE - I PS h tt p : / / j v i . a s m . o r g / D o w n l oaded f r o m View publication stats",
    "jin2022branched": "ARTICLE Branched actin networks are organized for asymmetric force production during clathrin - mediated endocytosis in mammalian cells Meiyan Jin 1 , 6 , Cyna Shirazinejad 1 , 2 , 6 , Bowen Wang 3 , Amy Yan 1 , Johannes Sch\u00f6neberg 1 , 5 , Srigokul Upadhyayula 1 , 4 , Ke Xu 3 & David G . Drubin 1 \u2709 Actin assembly facilitates vesicle formation in several traf \ufb01 cking pathways , including clathrin - mediated endocytosis ( CME ) . Interestingly , actin does not assemble at all CME sites in mammalian cells . How actin networks are organized with respect to mammalian CME sites and how assembly forces are harnessed , are not fully understood . Here , branched actin network geometry at CME sites was analyzed using three different advanced imaging approaches . When endocytic dynamics of unperturbed CME sites are compared , sites with actin assembly show a distinct signature , a delay between completion of coat expansion and vesicle scission , indicating that actin assembly occurs preferentially at stalled CME sites . In addition , N - WASP and the Arp2 / 3 complex are recruited to one side of CME sites , where they are positioned to stimulate asymmetric actin assembly and force production . We pro - pose that actin assembles preferentially at stalled CME sites where it pulls vesicles into the cell asymmetrically , much as a bottle opener pulls off a bottle cap . https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 OPEN 1 Department of Molecular and Cell Biology , University of California , Berkeley , CA , USA . 2 Biophysics Graduate Group , University of California Berkeley , Berkeley , CA , USA . 3 Department of Chemistry , University of California , Berkeley , CA , USA . 4 Chan Zuckerberg Biohub , San Francisco , CA , USA . 5 Present address : Department of Pharmacology , and Department of Chemistry and Biochemistry , University of California , San Diego , CA , USA . 6 These authors contributed equally : Meiyan Jin , Cyna Shirazinejad . \u2709 email : drubin @ berkeley . edu NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications 1 1 2 3 4 5 6 7 89 0 ( ) : , ; F ormation of clathrin - coated vesicles requires forces to \ufb01 rst bend the membrane into a sphere or tube , and to then break the thin neck that connects the vesicle to the plasma membrane . These forces are generated through the combined actions of proteins that directly bend the membrane and actin \ufb01 lament assembly 1 \u2013 5 ( Supplementary Fig . 1a ) . Several studies have demonstrated that dependence of CME on actin assembly increases under elevated membrane tension 6 \u2013 10 . Interestingly , actin does not assemble at all CME sites in mammalian cells , suggesting highly localized differences in requirement for actin assembly , the nature of which is obscure 6 , 11 , 12 . A detailed understanding of how actin forces are harnessed to aid vesicle formation and scission depends on understanding which CME sites assemble actin , where \ufb01 lament assembly occurs around the endocytic membrane , and when . In yeast cells , where turgor pressure is particularly high , super - resolution data suggest that actin assembles symmetrically around CME sites and indicate that actin regulators including Las17 , which is yeast WASP , are present in a ring surrounding the base of the clathrin coat symmetrically 13 . On the other hand , studies on \ufb01 xed mammalian cells raised the possibility that actin assembly may at least in some cases be initiated asymmetrically at clathrin coats 14 , 15 . However , methods used for these studies prevented analysis of large numbers of sites , and suffered from possible loss of actin \ufb01 la - ments during unroo \ufb01 ng and extraction of the cells . Which CME sites assemble actin , and how actin networks are organized with respect to CME sites , has not been determined systematically in a large - scale , unbiased manner , particularly in live mammalian cells . This information is essential to understanding how and why actin assembly forces are harnessed for CME . Here , by combining \ufb01 xed and live - cell imaging of triple - gen - ome - edited , human induced pluripotent stem cells ( iPSCs ) , and newly developed machine - learning - based computational analysis tools , we report that N - WASP and the Arp2 / 3 complex localize at one side of the coat and neck of invaginating endocytic sites until vesicle scission . Most importantly , by comparing recruitment dynamics of proteins from three distinct endocytic modules for over one thousand unperturbed endocytic events , we found that branched actin assembly occurs predominantly at sites that have stalled between coat expansion and vesicle scission . We propose that these branched actin networks rescue stalled CME . Results Super - resolution imaging reveals asymmetric actin distribu - tion around endocytic sites . To investigate the physiological roles and spatiotemporal regulation of actin assembly at CME sites in mammalian cells , we applied genome - editing techniques to generate a human iPSC line ( hereafter referred to as ADA cells ) that co - expresses a TagRFP - T fusion of the mu subunit of the AP2 adaptor complex ( AP2M1 ) , a TagGFP2 fusion of dynamin2 ( DNM2 ) , and a HaloTag fusion of the ARPC3 subunit of the Arp2 / 3 complex as representatives of the CME coat , scission and actin modules , respectively 2 , 16 , 17 ( Supplementary Fig . 1a , b ) . Previous studies showed that endogenously tagged AP2M1 , DNM2 and ARPC3 can serve as reliable markers of these CME functional modules that avoid disruption of physiological spa - tiotemporal organization of the process as might be caused by overexpression of \ufb02 uorescently labeled proteins 12 , 18 \u2013 21 . We observed dynamic CME events on the basal plasma membrane of the genome - edited cells using Total Internal Re \ufb02 ection Fluores - cence ( TIRF ) microscopy ( Supplementary Fig . 1c and Supple - mentary Movie 1 , 2 ) . Consistent with previous studies , AP2 is recruited at early CME stages while DNM2 is recruited in two phases 11 , 16 , 20 , 22 . At the early stage of CME , a relatively small amount of DNM2 is recruited to CME sites . Shortly before the end of a CME event , the DNM2 recruitment rate increases rapidly with DNM2 levels reaching a peak concomitant with vesicle scission 16 , 20 , 23 ( Supplementary Fig . 1c ) . This later rapid - recruitment phase represents the assembly of the dynamin helix on the highly curved neck of the budding vesicle after the U to \u03a9 shape transition of the endocytic membrane 20 , 23 \u2013 27 . Super - resolution imaging of \ufb01 xed human skin melanoma SKMEL cells suggested that actin is arranged asymmetrically around CME sites 7 , consistent with indications from other previous studies 14 , 15 . To systematically analyze how actin networks are organized at CME sites in iPSCs , we \ufb01 rst performed two - color 3D Stochastic Optical Reconstruction Microscopy ( STORM ) imaging 28 on \ufb01 xed ADA cells , localizing either AF647 phalloidin - labeled actin \ufb01 laments 29 or HaloTag - fused ARPC3 at CME sites . Due to the dense phalloidin labeling of cortical actin \ufb01 laments under the plasma membrane , it was often challenging to unambiguously identify the CME - speci \ufb01 c actin structures in iPSCs . However , in regions with thinner cortical actin layers , we observed that actin was typically distributed asymmetrically around CME sites ( Fig . 1a \u2013 c ) . Antibody labeling of ARPC3 - Halotag in the ADA cells had the advantage of a less complex staining pattern . Besides being highly concentrated in lamellipodia , ARPC3 was associated with CME sites asymme - trically , like actin ( Fig . 1d \u2013 f ) . These data suggest that the Arp2 / 3 - mediated actin network is arranged asymmetrically around CME sites . Actin networks assembled at CME sites remain asymmetric through scission . We next used ADA cells to investigate actin assembly at CME sites in live cells , which has several advantages over studies in \ufb01 xed cells . During the \ufb01 xation and subsequent sample preparation , actin structures may not be faithfully pre - served . In addition , in live cells it is easier to identify the stage of CME , so the timing , geometry and dynamics of actin assembly can be related to the endocytic stage . More importantly , only by using live cells is it possible to trace a single CME event from start to \ufb01 nish , and to therefore identify those CME events wherein no detectable actin is ever assembled , so key parameters can be compared between events with and without associated actin assembly . By visualizing endogenously tagged AP2M1 to mark the coat and CME initiation , and DNM2 to mark the neck and scission , together with ARPC3 to speci \ufb01 cally label Arp2 / 3 - nucleated , branched actin \ufb01 laments ( Supplementary Fig . 1a ) , we were able to precisely study the spatial and temporal regulation of actin assembly during CME . Three - color labeling and analysis of the displacement between markers for the three modules allowed us to distinguish bona \ufb01 de asymmetric actin assembly from events that might artifactually appear asymmetric because the invagina - tions were elongated and tilted ( Fig . 2a ) . Using TIRF live - cell imaging , we observed ARPC3 - labeled branched actin networks at lamellipodia and at a subpopulation of CME sites ( Fig . 2b , c ) . Dynamic actin assembly and disassembly occurred at CME sites with different spatio - temporal characteristics , including discrete CME sites , clathrin plaques and at clathrin coat splitting sites , as previously reported 14 ( Fig . 2d and Supplementary Fig . 2a , b ) . In the analysis described below , we focus on the discrete CME events and not the more complex ones ( plaques and splitting events ) ( Supplementary Fig . 3a ) . Analysis of the discrete events with 1 s / frame temporal resolution revealed that ARPC3 is most robustly recruited during the late stages of CME shortly before scission 12 ( Fig . 2c , d ) . Interestingly , we observed clear spatial displacement between ARPC3 ( actin module ) and AP2 ( coat module ) as well as between ARPC3 and DNM2 ( neck ) before vesicle scission ( Fig . 2d ) . This observation supports the conclusion that ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 2 NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications asymmetric branched actin networks provide forces at endocytic sites through the time of scission . To investigate the spatiotemporal relationship between actin and dynamin with higher resolution , we used live - cell Airyscan2 confocal microscopy of live cells ( Fig . 3 and Supplementary Movie 3 ) . We observed clear displacement between ARPC3 and DNM2 signals at the time of vesicle scission at most CME sites ( Fig . 3b ) . Notably , ARPC3 remained on one side of DNM2 throughout the last second of CME until scission ( Fig . 3c , d ) . These observations of asymmetric branched actin network assembly at CME sites are consistent with the ones we made with TIRF live - cell imaging ( Fig . 2 ) . a clathrin top view actin c clathrin ARPC3 top view top view side view top view side view top view side view top view side view top view side view A c t i n C l a t h r i n M e r ge p l a s m a m e m b r ane cy t o s o l z top view side view top view side view top view side view top view side view top view side view A R P C 3 C l a t h r i n M e r ge p l a s m a m e m b r ane cy t o s o l z b d 0 20 40 60 80 100120140160180 0 5 10 15 20 Clathrin - actin Distance ( nm ) C oun t s N = 67 71 . 15 \u00b1 36 . 10 0 20 40 60 80 100 120 140 160 180 200 220 240 260 0 10 20 30 Clathrin - ARPC3 Distance ( nm ) C oun t s N = 161 107 . 26 \u00b1 49 . 66 e f 0 nm 1500 nm 0 nm 1500 nm NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 ARTICLE NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications 3 Fig . 1 Two - color , 3D stochastic optical reconstruction microscopy ( STORM ) shows that actin structures are off - centered with respect to clathrin coats . a , b Two - color 3D STORM image of bottom membrane of ADA cells immunolabeled with clathrin light chain antibody ( clathrin , CF - 680 , magenta ) and phalloidin ( actin , AF647 , rainbow ) . c A histogram of distances between centroids of clathrin and actin signals . Mean and standard deviation are reported on the graph . Data were analyzed using Prism 9 . Source data are provided in the Source Data \ufb01 le . d , e Two color 3D STORM image of the bottom membrane of ADA cells immunolabeled with clathrin light chain antibody ( clathrin , AF647 , magenta ) and HaloTag antibody ( ARPC3 - HaloTag , CF - 680 , rainbow ) . Dotted lines label lamellipodia . b , e The highlighted CME sites , which are labeled by white arrows in a , d , are rotated and shown in magni \ufb01 ed top and side view projections . Color bar shows the z position of ARPC3 - HaloTag . f A histogram of distances between centroids of clathrin and ARPC3 signals . Mean and standard deviation are reported on the graph . Data were analyzed using Prism 9 . Source data are provided in the Source Data \ufb01 le . Scale bars : a , d : 2 \u00b5 m , b , e : 100 nm . DNM2 - tagGFP2 AP2 - tagRFP - T ARPC3 - HaloTag Merge b DNM2 - tagGFP2 AP2 - tagRFP - T ARPC3 - HaloTag Merge 4 min c d ARPC3 - HaloTag AP2 - tagRFP - T ARPC3 - HaloTag DNM2 - tagGFP2 AP2 - tagRFP - T 1s DNM2 - tagGFP2 ARPC3 - HaloTag * * * * * Center of signal Coat module ( AP2 ) Scission module ( DNM2 ) Actin module ( ARPC3 ) Plasma membrane a Model 1 Model 2 Model 3 Fig . 2 Triple - genome - edited iPSCs reveal dynamic actin organization at CME sites . a Models of branched actin assembly at invaginating CME sites . Model 1 : Asymmetric actin assembly at CME sites results in separated actin - coat and actin - neck signals . Model 2 : Symmetric actin assembly at tilted CME sites results in separated actin - coat signals but overlapped actin - neck signals . Model 3 : Symmetric actin assembly at perpendicularly invaginating CME sites will result overlapped actin , coat and neck signals . b A representative single time frame image of a TIRF movie ( Supplementary Movie 2 ) of AP2M1 - tagRFP - T ( magenta ) , DNM2 - tagGFP2 ( green ) and JF635 ligand 51 - conjugated ARPC3 - HaloTag ( cyan ) in ADA cells . The highlighted region is boxed by a dashed line . Scale bar : 5 \u00b5 m . c A representative kymograph of AP2M1 - tagRFP - T ( magenta ) , DNM2 - tagGFP2 ( green ) and JF635 ligand - conjugated ARPC3 - HaloTag ( cyan ) at CME sites in ADA cells . Scale bar : 5 \u00b5 m . d Montage of a representative ARPC3 positive CME site in ADA cells . Individual channels and pair - wise merges are shown . * : Images from one frame before scission ( maximum DNM2 intensity ) are marked to show the displacement between the CME coat ( AP2 ) - ARPC3 and CME neck ( DNM2 ) - ARPC3 . Size of \ufb01 eld of view : 2 \u00b5 m \u00d7 2 \u00b5 m . Intervals : 1 s . ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 4 NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications To analyze the intrinsic recruitment order and timing for three endocytic proteins at CME sites quantitatively and systematically , we developed an automated , high - throughput method to analyze TIRF live - cell imaging data that avoids bias because it does not involve manual selection of CME sites ( see Materials and Methods ) . Brie \ufb02 y , AP2 tracks were identi \ufb01 ed using standard particle - tracking algorithms 30 . Novel \ufb01 ltering methods then extracted DNM2 - positive events marked by one or more DNM2 burst . The AP2 and DNM2 tracks were decomposed into dynamic features describing each events \u2019 position and brightness . These features were used for clustering via unsupervised machine learning , which enabled grouping of similarly - behaved tracks ( Supplementary Fig . 3a , b ) . DNM2 - positive events were re \ufb01 ned by a detection scheme that determined the number of DNM2 peaks using various characteristics of a single DNM2 - peak : the peak height , width , and minimum peak - to - peak distance ( Supplementary Fig . 3c ) . Events with a single DNM2 peak were analyzed as described below . The method detects low signals from endogenously tagged CME proteins , such as the low - level recruitment of DNM2 at the early CME stages , and accurately reveals the different CME stages ( Supplementary Fig . 3d ) . Next , the timing of actin network assembly at CME sites was determined using ARPC3 as a branched actin \ufb01 lament marker by analyzing over one thousand CME events . Although actin appearance early in CME has been reported 14 , determining the actin assembly timing can be challenging because it is dif \ufb01 cult to distinguish newly assembled branched actin at CME sites from the nearby cortical actin \ufb01 laments or actin \ufb01 laments associated with other vesicles or organelles . Also , whether actin functions during the early stage of CME has not yet been shown conclusively due to the potential side effects such as changes in membrane tension caused by actin inhibitors . Our endogenous ARPC3 tagging and large - scale computational analysis approach sidesteps these problems . We classi \ufb01 ed CME events into three groups : one group without ARPC3 appearance , one group with ARPC3 appearance early in CME , and one group with ARPC3 appearance late in CME ( Supplementary Fig . 4 ) . We observed that in most of the ARPC3 positive events a sharply increasing ARPC3 signal appears with similar timing to the rapid - recruitment phase of DNM2 concomitant with the U to \u03a9 membrane shape transition ( Fig . 4a and Supplementary Fig . 4b ) . This timing is consistent with the previously proposed role for actin in membrane invagination , as studies showed that actin inhibitors block the U to \u03a9 endocytic membrane shape transition 6 , 14 . In some cases we did detect ARPC3 signals at early CME stages ( Supplementary Fig . 4b ) . To test whether random overlap between nearby actin structures and CME sites might be responsible for the apparent early actin recruitment , we generated a randomized data set by pairing ARPC3 images with AP2 and DNM2 images from an unrelated movie ( Supplementary Fig . 4a ) . In this data set , we detected a signi \ufb01 cantly reduced fraction of ARPC3 positive events . However , early \u201c assembly \u201d of actin was observed in a similar fraction of CME events as in the real data set ( Supplementary Fig . 4b ) . Based on these observations we conclude that the presence of actin early in CME is very likely due to nearby actin structures overlapping with CME sites randomly . Our live - cell analysis allowed the timing of branched actin network assembly to be compared to the scission timing , and the spatial offset between the clathrin coat and the associated actin network to be determined . Super - resolution imaging of yeast CME sites suggested that actin and actin nucleators localize * * * DNM2 - tagGFP2 ARPC3 - HaloTag Merge 0 . 2s a c d 010020030040050060070080090010001100120013001400 0 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1 . 0 Distance ( nm ) I n t en s i t y ( a r b . un i t s ) DNM2 ARPC3 b 0 20 40 60 80100120140 160 180200220240260280300320 0 5 10 15 DNM2 - ARPC3 distance ( nm ) C oun t s N = 72 110 . 94 \u00b1 71 . 08 6000 2002500 200 DNM2 - tagGFP2 ARPC3 - HaloTag Merge Fig . 3 Airyscan live - cell imaging reveals asymmetric actin organization at CME sites . a A representative single time frame image of an Airyscan movie ( Supplementary Movie 3 ) of DNM2 - tagGFP2 ( red ) and JF635 ligand - conjugated ARPC3 - HaloTag ( cyan ) on the ventral plasma membrane surface of ADA cells . The highlighted region is boxed by a dashed line . Scale bar : 1 \u00b5 m . b Histogram of distance between the center of mass of DNM2 and ARPC3 at frame corresponding to scission . Mean and standard deviation are shown on the graph . Data were analyzed using Prism 9 . Source data are provided in the Source Data \ufb01 le . c Montage of average intensity projection of representative ARPC3 positive CME sites . Frame corresponding to scission was determined by maximum DNM2 intensity . 33 \u00d7 33 pixels square region centered at DNM2 maximum intensity pixel was cropped for six continuous frames ending in scission . Cropped frame series were rotated by 90 or 180 degrees as needed to align the ARPC3 signal to the left of the center of the image at the frame corresponding to scission . * : Averaged image of the frame corresponding to scission ( maximum DNM2 intensity ) is marked to show the displacement between the CME neck ( DNM2 ) and branched actin ( ARPC3 ) . Scale bar : 1 \u00b5 m . Intervals : 0 . 2 s . N = 72 . d The line scan function in ImageJ software was used to measure the \ufb02 uorescence intensity along a one pixel wide , 33 pixel long horizontal line drawn across the center of the averaged intensity image of the frame corresponding to scission * in c . The signals were normalized to the minimum signal along the line , and intensity was calculated as a ratio to the maximum signal along the line . Source data are provided in the Source Data \ufb01 le . NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 ARTICLE NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications 5 symmetrically in a ring around CME sites , and computational modeling suggested that an asymmetric actin arrangement would not provide suf \ufb01 cient force for the membrane invagination during yeast CME 13 . In contrast , in mammalian cells , which require less actin force production during CME , imaging of \ufb01 xed cells suggested that actin structures associate adjacent to apparent \ufb02 at clathrin coats 14 , 15 . However , these studies proposed that at the later CME stages the actin structures become larger and more symmetric to provide suf \ufb01 cient force for membrane deformation and scission 14 , 15 . Surprisingly , in our live cell studies designed to highlight sites of new actin assembly , we observed off - centered branched actin networks at CME sites throughout even the latest CME stages ( Fig . 2d and Fig . 3b \u2013 d ) . Furthermore , most ARPC3 - positive CME sites accomplish scission within 30 s of the initiation of ARPC3 recruitment ( Fig . 4b ) . The actin networks ( ARPC3 ) we observed were off center from the coat ( AP2 ) and neck ( DNM2 ) signals by approximately 150 nm at the time of vesicle scission ( Fig . 4c ) . Imaging \ufb02 uorescent beads using the same settings indicated that the displacement is not an artifact caused by misalignment between different imaging channels ( Supplementary Movie 4 and Supplementary Fig . 5a ) . By further analyzing \ufb02 uorescent bead images , we concluded that chromatic aberration contributes only a small portion of the AP2 - ARPC3 and AP2 - DNM2 separation we observed by TIRF imaging ( Supplementary Fig . 5b ) . This chromatic aberration might be the reason for slightly increased neck - actin separation we detected by TIRF microscopy ( Fig . 4c ) compared to Airyscan imaging ( Fig . 3b ) . Given the temporal separation between channel acquisition and the movement of AP2 spots , it was important to assess whether the spatial separation between channels can be attributed in part to an imaging artifact caused by puncta movement between subsequent channel acquisitions . When we measured the average movement of AP2 spots leading up to scission , we found that over 95 % of the events had AP2 - ARPC3 separations that exceed the frame - to - frame motility of AP2 ( Supplementary Fig 5c ) . Also , the puncta positional uncertainties indicated by the standard deviations determined when measuring the \ufb01 tted position of AP2 , range up to 40 nm , which is less than the determined displacements . Therefore , we utilized the AP2 - DNM2 separation , which is expected to be small , as the basis for comparison to the AP2 - ARPC3 and DNM2 - ARPC3 separations ( Fig . 4d ) . These results further support our conclusion that branched actin networks assemble asymmetrically at CME sites through the time of scission ( Fig . 2d ) . This observation is consistent with the observation that ring - shaped actin structures at clathrin coats were rarely observed in the high - resolution , live - cell imaging reported in a previous study 31 . In total , these live - cell data suggest that in mammalian cells , asymmetric actin network assembly can provide enough force to assist membrane deformation and scission during the late stages of CME . Asymmetric branched actin networks facilitate CME at stalled sites . To gain additional insights into the function of this asym - metric actin network assembly , we quantitatively compared kinetics of CME events with or without ARPC3 recruitment . We observed that about 30 % of CME events are completed in the absence of detectable actin assembly ( Supplementary Fig . 4b ) , which is consistent with the hypothesis that in mammalian cells a b c D i s t an c e ( \u00b5 m ) F r e quen cy den s i t y ARPC3 lifetime ( s ) 0 . 00 0 . 01 0 . 02 0 . 03 0 . 04 0 . 06 0 . 05 20 0 40 60 100 80 AP2 ARPC3DNM2 DNM2 - ARPC3 AP2 - ARPC3 0 100 200 300 400 500 0 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 Time ( s ) 0 - 10 - 20 10 20 - 30 AP2 ARPC3DNM2 Lifetime : 7 . 73 % 17 . 33 % 27 . 36 % 47 . 51 % 0 - 10 - 20 10 20 0 - 10 10 0 - 10 - 20 10 20 - 30 30 0 - 10 - 20 10 20 - 30 30 - 40 - 50 - 60 Time ( s ) 0 100 200 300 400 500 I n t en s i t y ( a r b . un i t s ) DNM2 - AP2 < 40s 40 - 60s 60 - 80s > 80s I n t en s i t y ( a r b . un i t s ) Fig . 4 Computational analysis of ARPC3 positive CME sites reveals asymmetric actin network assembly at the late stage of CME . a Averaged intensity vs time plots of cohorts of CME sites with late ARPC3 assembly in ADA cells . Events are grouped into cohorts by the lifetimes of AP2 and aligned to the frames showing the maximum DNM2 intensity ( time = 0 s ) . Percentage of the number of the CME sites in each cohort is shown next to the plot . b Histogram of ARPC3 - mediated actin network assembly duration . The assembly duration is measured from the \ufb01 rst frame of the ARPC3 signal to the presumed scission time ( the peak of DNM2 signal ) . Source data are provided in the Source Data \ufb01 le . c Averaged intensity ( solid lines ) and distance ( dashed lines ) vs time plots of ARPC3 positive CME sites in ADA cells . Events are aligned to the frames showing the maximum DNM2 intensity ( time = 0 s ) . Distance between centers of two signals are shown from - 10 s to 3 s when DNM2 and ARPC3 signals are relatively high . a \u2013 c N = 1 , 385 . a , c Error bar : \u00bc standard deviation . . ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 6 NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications actin assembly is required for CME only under certain conditions , such as relatively high membrane tension or speci \ufb01 c cargo internalization , which can vary regionally within cells 6 , 7 , 9 , 10 , 32 , 33 . Consistent with the possibility that unfavorable conditions such as increased membrane tension , might stall membrane defor - mation during CME 4 , 6 , 8 \u2013 10 , 34 \u2013 36 , CME lifetimes were markedly longer for ARPC3 positive events compared to the ARPC3 negative events ( Fig . 5a ) . In addition , when the AP2M1 intensity vs time pro \ufb01 les were compared between ARPC3 positive and negative CME sites , a plateau , which lasts for approximately 10 s , was observed for the ARPC3 positive events ( Fig . 5b ) . Based on these observations and previous experimental and computational modeling data 4 , 6 , 7 , we propose that this plateau in branched actin - positive CME events represents stalled membrane bending due to an unfavorable local membrane environment . We next tested the hypothesis that the asymmetric actin network might affect the lateral movements of endocytic coats on the plasma membrane . Interestingly , the ARPC3 positive CME sites showed signi \ufb01 cantly slower , but greater directional lateral movement before vesicle scission compared to the ARPC3 negative CME sites ( Fig . 5c , d ) . After scission both ARPC3 positive and negative vesicles showed fast , apparently random movements ( Fig . 5c , d ) . These data suggest that the asymmetric actin can stabilize the forming endocytic coat while pushing it in the plane of the plasma membrane with a lateral directional force . To test the function of actin network assembly on CME , we treated the cells with an Arp2 / 3 inhibitor , CK666 . We observed CME dynamics immediately after the treatment to minimize non - speci \ufb01 c side effects that might be caused by prolonged actin assembly disruption . Even with moderate inhibition of actin network assembly at CME sites , indicated by reduced ARPC3 intensity ( Supplementary Fig . 6a ) , we detected a small but signi \ufb01 cant increase in CME lifetimes ( Supplementary Fig . 6b ) . This result is consistent with results from previous studies in SKMEL cells 7 , 12 and supports the hypothesis that Arp2 / 3 - mediated actin assembly facilitates CME . a b d c F r equen cy den s i t y CME lifetime ( s ) AP2 ARPC3DNM2 ARPC3 - ARPC3 + Time ( s ) F l uo r e sc en c e i n t en s i t y ( a r b . un i t s ) 0 100 200 300 400 500 0 100 200 300 400 500 0 - 10 - 20 10 Average AP2 movement ( \u00b5 m ) C u m u l a t i v e f r equen cy 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 0 . 01 0 . 1 ARPC3 + before ARPC3 + after ARPC3 - after ARPC3 - before ARPC3 + before ARPC3 + after ARPC3 - after ARPC3 - before Straightness - index C u m u l a t i v e f r equen cy 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 ARPC3 - ARPC3 + 0 . 000 0 . 005 0 . 010 0 . 015 0 . 020 0 40 80 120 160 20 60 100 140 Fig . 5 Actin positive CME sites show distinct dynamics . a Histograms of ARPC3 negative ( blue ) and positive ( orange ) CME lifetimes . CME lifetime is measured from the \ufb01 rst frame of the AP2 signal to the presumed scission time ( the peak of DNM2 signal ) . ARPC3 positive CME events have longer lifetimes . p - value from two - sided Welch \u2019 s t - test : 2 . 11e - 76 . b Averaged intensity vs time plots of ARPC3 negative ( top ) and positive ( bottom ) CME sites in ADA cells . Events were aligned to the frames showing the maximum DNM2 intensity . Error bar : \u00bc standard deviation . c Lateral motility of ARPC3 negative ( blue ) and positive ( yellow ) CME sites before ( solid line ) and after ( dashed line ) vesicle scission . ARPC3 positive CME sites move slower than ARPC3 negative ones . p - value from two - sided Kolmogorov - Smirnov test : ARPC3 + before vs ARPC3 - before : 1 . 27e - 14 , ARPC3 + after vs ARPC3 - after : 5 . 18e - 08 . d Straightness - index of ARPC3 negative ( blue ) and positive ( yellow ) CME sites before ( solid line ) and after ( dashed line ) scission . The straightness - index is de \ufb01 ned by the ratio between the sum of frame - to - frame distances to the end - to - end distance of a single event \u2019 s trajectory , where a perfectly straight - lined trajectory would have an index of 1 . APRC3 positive CME sites move with a more straight trajectory . p - value from two - sided Kolmogorov \u2013 Smirnov test : ARPC3 + before vs ARPC3 - before : 1 . 07e - 11 , ARPC3 + after vs ARPC3 - after : 2 . 06e - 6 . a \u2013 d ARPC3 - : N = 840 , ARPC3 + : N = 1 , 385 . a , c , d Source data are provided in the Source Data \ufb01 le . NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 ARTICLE NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications 7 N - WASP is recruited asymmetrically to stalled CME sites . To further explore how asymmetric assembly of actin networks at CME sites is regulated , we endogenously tagged N - WASP , an actin nucleation promoting factor ( NPF ) that plays roles in CME , in AP2M1 - tagRFP - T / DNM2 - tagGFP2 genome - edited iPSCs ( hereafter referred to as ADW cells , Fig . 6a , Supple - mentary Fig . 7a and Supplementary Movie 5 ) . Quantitative imaging of budding yeasts demonstrated that initiation of productive actin assembly at CME sites requires the accumu - lation of yeast WASP and WIP ( WASP Interacting Protein ) to a certain level 37 . In our genome - edited iPSCs , we observed that N - WASP is recruited asymmetrically to CME sites mostly at the late stage of CME ( Fig . 6b , c and Supplementary Fig . 7b , c ) . Longer lifetimes and a plateau in the AP2 intensity vs time plot were observed speci \ufb01 cally for the N - WASP positive CME events ( Fig . 6d , e ) , similar to the ARPC3 positive events ( Fig . 5a , b ) . These data indicate that asymmetric NPF recruit - ment underlies the asymmetric architecture of branched actin networks at CME sites . Discussion Using unbiased analysis of thousands of CME sites in unper - turbed live cells , our study demonstrates that in mammalian cells clathrin coat assembly dynamics predict which sites will assemble actin , and show that at apparently stalled sites , actin assembles asymmetrically to facilitate successful vesicle formation . Based on the data presented here , we propose an updated model for actin assembly at mammalian CME sites in which , beyond global tension - dependent changes in the requirement for actin assembly , highly localized differences give rise to hetero - geneity even within the same patch of plasma membrane in the same cell ( Fig . 7 ) : ( 1 ) Where the local conditions are favorable for membrane deformation by coat proteins ( Fig . 7 upper scenario ) , the membrane can undergo \ufb02 at - U - \u03a9 shape transitions in a relatively short time without actin assembly . When the coat grows large enough to form a \u03a9 - shaped bud , suf \ufb01 cient dynamin can be recruited to perform scission , and there is little delay between coat expansion and scission ; ( 2 ) Where the local conditions are not favorable , presumably under high membrane tension or other AP2 - tagRFP - T HaloTag - N - WASP DNM2 - tagGFP2 Merge a b c M e r ge 4 min N - W ASP DN M 2 AP 2 S c a l ed i n t en s i t y ( a r b . un i t s ) D i s t an c e ( \u00b5 m ) Time ( s ) 0 - 10 - 20 10 0 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 00 0 . 25 0 . 50 0 . 75 1 . 00 d F r equen cy den s i t y CME lifetime ( s ) N - WASP - N - WASP + e AP2 N - WASP DNM2 N - WASP - N - WASP + F l uo r e sc en c e i n t en s i t y ( a r b . un i t s ) 0 100 200 300 0 100 200 300 Time ( s ) 0 - 10 - 20 10 AP2 N - WASP DNM2 DNM2 - N - WASP AP2 - N - WASP AP2 - DNM2 0 . 0000 0 . 0025 0 . 0050 0 . 0075 0 . 0100 0 . 0125 0 . 0150 50 0 100 150 200 - 30 20 1 . 25 Fig . 6 Asymmetric N - WASP recruitment to stalled CME sites . a A representative single time frame image of a TIRF movie ( Supplementary Movie 5 ) of AP2M1 - tagRFP - T ( magenta ) , DNM2 - tagGFP2 ( green ) and JF635 ligand - conjugated HaloTag - N - WASP ( cyan ) in ADW cells . The highlighted region is boxed by a dashed line . Scale bar : 5 \u00b5 m . b A representative kymograph of CME sites in ADW cells over a 4 min movie . Scale bar : 5 \u00b5 m . c Averaged intensity ( solid line ) and distance ( dashed line ) vs time plots of N - WASP positive CME sites in ADW cells . Events are aligned to the frames showing the maximum DNM2 intensity . Intensity is scaled to 1 at peaks for each channel . Error bar : \u00bc standard deviation . d N - WASP positive CME events have longer lifetimes . p - value from two - sided Welch \u2019 s t - test : 9 . 06e - 20 . e Intensity vs time plots of averaged N - WASP negative ( top ) and positive ( bottom ) CME sites in ADW cells . Events were aligned to the frames showing the maximum DNM2 intensity . Error bar : \u00bc standard deviation . Source data are provided in the Source Data \ufb01 le . c , d N - WASP negative CME sites : N = 299 , N - WASP positive CME sites : N = 1 , 199 . ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 8 NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications impediments , the coat protein - membrane interaction does not generate suf \ufb01 cient force to curve the membrane ( Fig . 7 lower scenario ) . Here , extra force generation from actin assembly is required 4 , 6 , 10 . Asymmetric N - WASP recruitment activates actin nucleation mostly at one side of the clathrin coat , generating an asymmetric force that pulls the membrane into the cell with a similar action to a bottle cap opener . We speculate that this asymmetrical force contributes to asymmetric membrane defor - mation at endocytic sites observed by high - speed atomic force microscopy 38 and may act with dynamin 39 to twist the clathrin pit to promote scission at the neck . CME events with associated actin assembly have longer lifetimes , likely due to a delay between the end of coat expansion and progress toward vesicle scission , requiring adaptive recruitment of actin regulators followed by actin network assembly and membrane remodeling . This result establishes that site - to - site heterogeneity in actin dependence and involvement can be observed without manipulating actin assembly . Our model provides further insights into the basis for incon - sistent effects of actin drugs on CME 6 , 14 , 18 , 40 \u2013 45 . Actin plays crucial roles in membrane shaping , cell adhesion , and membrane tension . Global disruption of actin dynamics is expected to dra - matically change membrane tension and the available pool of actin and associated proteins and therefore to have both direct and indirect effects on CME . Here , we focused on in - depth analysis of the unperturbed process and detected preference for actin assembly at stalled CME events . We used genome - edited ARPC3 - HaloTag protein expressed at endogenous level as a marker of branched actin networks . This \ufb02 uorescent protein allowed us to study the spatiotemporal dynamics of Arp2 / 3 - nucleated actin structures at CME sites speci \ufb01 cally ( Fig . 1a , d ) . A large population of unbranched corti - cal actin \ufb01 laments at CME sites has been observed by cryo - EM 46 and distinct effects of cortical actin - based structures on CME have been reported 36 . Previous studies on actin \u2019 s spatio - temporal dynamics and function at CME sites applied actin probes such as Lifeact , which label actin \ufb01 laments nucleated by different mechanisms , and which can perturb dynamic actin properties , possibly complicating conclusions reached in those studies 14 , 31 , 47 , 48 . WASP family proteins play a central role in regulation of Arp2 / 3 - mediated actin assembly both spatially and temporally during various cellular processes . We observed that N - WASP is recruited asymmetrically to stalled CME sites ( Fig . 6 ) . We pro - pose this asymmetric recruitment underlies the asymmetric actin architecture at CME sites . Quantitative live - cell imaging studies in yeast suggested concentration of yeast WASP at CME sites through multivalent protein interactions is important for robust , switch - like actin assembly at CME sites 37 . Future modeling together with quantitative in vivo and in vitro studies are needed to determine whether concentrating N - WASP into a small space , which is asymmetrically located relative to CME sites , is a more ef \ufb01 cient way to generate suf \ufb01 cient actin force for productive CME , compared to recruiting more N - WASP into a larger space that encircles the CME sites . Future computational modeling studies on how asymmetric actin network assembly provides forces and generates a torque for vesicle formation and mem - brane remodeling will deepen our understanding of actin \u2019 s functions in a host of actin - mediated processes . While this study provides evidence that actin assembles asymmetrically at mammalian CME sites , some questions remain open . For example , what is the molecular mechanism that couples recruitment of N - WASP and other actin assembly factors to the stalled CME sites ? Many factors , including membrane tension 6 , 9 , 10 , cell cycle 9 , cell adhesion 8 , 49 and cargo size and composition 32 , 33 appear to contribute to the actin requirement for CME . How do these different factors regulate the actin assembly machinery ? Answering these questions in the future will deepen our understanding of dynamic actin assembly regulation in membrane traf \ufb01 cking . Methods Cell culture . The WTC10 hiPSC line was obtained from the Bruce Conklin Lab at UCSF . hiPSCs were cultured on Matrigel ( hESC - Quali \ufb01 ed Matrix , Corning ) in StemFlex medium ( Thermo Fisher ) with Penicillin / Streptomycin in 37\u00b0C , 5 % CO 2 . Cultures were passaged with Gentle Cell Dissociation reagent ( StemCell Technologies , Cat # : 100 \u2013 0485 ) twice every week . Genome - editing . The AP2M1 gene was edited in WTC10 hiPSCs as previously described using TALENs targeting exon 7 of the AP2M1 gene 50 . Both alleles of AP2M1 were tagged with tagRFP - T . The Cas9 - crRNAtracrRNA complex electro - poration method was used sequentially to edit DNM2 and ARPC3 gene in AP2M1 - tagRFP - T genome - edited hiPSCs , as previously described 12 , 18 . The same method was used to edit the WASL gene in AP2M1 - tagRFP - T / DNM2 - tagGFP2 genome edited hiPSCs . S . pyogenes NLS - Cas9 was puri \ufb01 ed in the University of California Berkeley QB3 MacroLab . TracrRNA and crRNA that target CCTGCTCGAC - TAGGCCTCGA ( DNM2 ) , CCTGGACAGTGAAGGGAGCC ( ARPC3 ) and AGCTCATGGTTTCGCCGGCG ( WASL ) , were purchased from IDT . Gibson assembly ( New England Biolabs ) was used to construct donor plasmids containing DNM2 5 \u2032 homology - ggtaccagtggcggaagc - tagGFP2 - DNM2 3 \u2032 homology , ARPC3 5 \u2032 homology - ggatccggtaccagcgatccaccggtcgccacc - HaloTag - ARPC3 3 \u2032 homology , and WASL 5 \u2032 homology - HaloTag - agcgatccaccggtcgccaccggatcc - WASL 3 \u2032 homology sequences , respectively . Three days after electroporation ( Lonza , Cat # : VPH - 5012 ) Actin + CytosolTIRF field CytosolTIRF field Coat module ( AP2 ) Scission module ( DNM2 ) Actin module ( ARPC3 ) Plasma membrane Membrane tension Actin force NPF module ( N - WASP ) Actin - Fig . 7 An updated schematic model of actin - negative and actin - positive clathrin - coated pits in human cells . Actin assembly is induced at stalled CME sites , where asymmetric forces pull , bend and possibly twist the plasma membrane against membrane tension to drive membrane invagination and vesicle scission . NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 ARTICLE NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications 9 of the Cas9 - crRNA - tracrRNA complex and donor plasmid , the tagGFP2 or HaloTag positive cells were single cell sorted using a BD Bioscience In \ufb02 ux sorter ( BD Bioscience ) into Matrigel - coated 96 - well plates . Clones were con \ufb01 rmed by PCR and Sanger sequencing of the genomic DNA locus around the insertion site . Both alleles of DNM2 and ARPC3 were tagged with tagGFP2 and HaloTag , respectively , and one allele of WASL was tagged with HaloTag in the hiPSC lines used in this study . Western blotting . Cells were dissociated from the well using Gentle Cell Dis - sociation reagent ( StemCell Technologies , Cat # : 100 - 0485 ) . Total proteins were extracted by adding 1ml of cold 10 % TCA to the cell pellets , incubated on ice for 30min , and spun down by centrifuging at 4\u00b0C , 13400 \u00d7 g for 10min . Protein pellets were dissolved in loading buffer ( 50 mM HEPES , pH 7 . 4 , 150 mM NaCl , 1mM MgCl2 , 5 % BME , 5mM DTT and protease inhibitor ) and loaded onto an acrylamide gel for SDS - PAGE and transferred to nitrocellulose membranes for immunoblotting . Blots were incubated overnight at 4 \u00b0C with primary antibodies targeting Tag ( CGY ) FP ( 1 : 2000 dilution in 2 % milk , Evrogen , Cat # : AB121 ) , HaloTag ( 1 : 2000 dilution in 2 % milk or 1 : 1000 dilution in 0 . 5 % milk , Promega , Cat # : G9211 ) , GAPDH ( 1 : 100 , 000 dilution in 0 . 5 % milk , Proteintech , Cat # : 10494 - 1 - AP ) , respectively , and subsequently incubated in the dark at room temperature for 1 hr with secondary antibodies . TIRF live - cell imaging . Two days before imaging , hiPSCs were seeded onto Matrigel - coated 4 - well chambered cover glasses ( Cellvis ) . Halotag was labeled by JF635 - HaloTag ligand 51 . Cells were incubated in StemFlex medium with 100 mM JF635 - HaloTag for 45 min and the unbound ligands were washed away by three washes with 5min incubation in prewarmed StemFlex medium . Cells were imaged on a Nikon Ti - 2 inverted microscope \ufb01 tted with TIRF optics and a sCMOS camera ( Hamamatsu ) . Cells were maintained at 37\u00b0C with a stage top incubator ( OKO Lab ) in StemFlex medium with 10 mM HEPES . Images were acquired with Nikon Elements . Channels were acquired sequentially at a 1s interval and 300 ms expo - sure time over 4 min . For the CK666 treatment experiments , a 50 mM CK666 ( SML0006 , Sigma Aldrich ) stock solution was prepared in DMSO and kept at \u2212 80\u00b0C . 200 \u03bc M CK666 solution and 4 % DMSO ( v / v ) solution for a control were prepared fresh in StemFlex medium with 10mM HEPES and incubated at 37\u00b0C prior to use . Live - cell imaging was performed 1min after adding an equal volume of medium containing CK666 or DMSO into the imaging chamber to achieve a \ufb01 nal concentration of 100 \u03bc M CK666 or 2 % DMSO . TIRF image processing and analysis . Events ( i . e . , tracked diffraction - limited spots ) were extracted from cmeAnalysis 30 and processed in Python Jupyter Notebooks . AP2 - tagRFP - T was used as the \ufb01 ducial marker for clathrin - mediated endocytosis in cmeAnalysis tracking experiments . DNM2 was used as a secondary channel to mark vesicle scission and the termination of vesicle formation . ARPC3 - HaloTag was tracked separately from paired AP2 / DNM2 movies and linked to CCPs downstream of cmeAnalysis to allow for determining discrete ARPC3 nucleation and disassembly events . Four generalized processing steps were applied to identify clathrin - coated pits ( AP2 tracks ) with single DNM2 peaks : track feature abstraction , feature dimensionality reduction , event clustering , and DNM2 - peak detection . First , tracks that are de \ufb01 ned by \ufb01 tted positions and intensities for single events were generated using cmeAnalysis . Then , AP2 and DNM2 tracks were decomposed into dynamic features describing the dynamics of the position and brightness for each event . Each tracked event , which was once an arbitrary array of intensities and positions , was now a discrete vector of \ufb01 xed length . The mapping of each track to discrete features was done to generalize the dynamics of heterogeneous tracked events into a set of interpretable coordinates . Following feature abstraction , the output array is a 2 - dimensional matrix with N rows ( N tracked events ) and M columns ( M discrete features per track ) . These features were individually scaled to normal distributions to remove the variability in scale and dampen the effects of outliers . For instance , the \u2018 lifetime \u2019 feature ( AP2 lifetime ) ranged from a few seconds to several minutes on a scale of seconds , whereas the \u2018 DNM2 - peak fraction \u2019 feature ( where the DNM2 peak is located within one AP2 event ) ranges from 0 to 1 . Following feature re - scaling , these events , which each contain over thirty features , were projected to a lower - dimensional space via principal component analysis . The derived clusters were separated using a Gaussian Mixture Model and events were assigned to clusters based on their highest probability of identity to one cluster . DNM2 - positive events represented a distinct cluster of tracks that had detectable DNM2 throughout the event , were long lived , and were below the threshold of motility expected for transient , non - CME - derived clathrin - coated vesicle \u201c visitors \u201d 5 0 to the TIRF \ufb01 eld . To characterize a single - DNM2 peak , DNM2 - positive events were surveyed over a range of values set for the minimum DNM2 peak height , width , and peak - to - peak temporal distance . For each peak - de \ufb01 ning parameter combination , all DNM2 - positive events were categorized as having zero , one , or two or more peaks . After \ufb01 nding single - peaked events in a \ufb01 xed peak - parameter combination , the lifetime distribution of lifetimes for single peak events was \ufb01 t to the expected underlying distribution , a Rayleigh distribution 52 , where the best - \ufb01 tting parameter combination was chosen to identify single - peaked events . Single DNM2 - peaked events were kept as CME sites for the remainder of the analysis . All code associated with this analysis , generating Figs . 4 \u2013 6 , and a detailed step - by - step protocol , are available at https : / / github . com / DrubinBarnes / Jin _ Shirazinejad _ et _ al _ branched _ actin _ manuscript . Two - color 3D STORM imaging and analysis . 12mm round coverslips were sonicated in distilled water and sterilized for 20min in 70 % ethanol , air - dried and coated with Matrigel in 24 - well plates . Cells were seeded onto Matrigel - coated coverslips two days before \ufb01 xation . For clathrin and actin two - color imaging , cells were \ufb01 xed \ufb01 rst for 1min in 0 . 3 % ( v / v ) glutaraldehyde ( GA ) solution containing 0 . 25 % ( v / v ) Triton in cytoskeleton buffer ( CB : 10mM MES , 150 mM NaCl , 5mM EGTA , 5 mM Glucose , 5 mM MgCl 2 , 0 . 005 % NaN 3 , pH 6 . 1 ) and then immediately \ufb01 xed for 20min in 2 % ( v / v ) GA solution in CB . Both solutions were prepared fresh from a 10 % GA stock ( Electron Microscopy Science , Cat # : 16120 ) . After \ufb01 xation , samples were incubated twice for 5min in freshly prepared 0 . 1 % ( w / v ) NaBH4 in PBS . For clathrin and ARPC3 - HaloTag imaging , cells were \ufb01 xed for 20 min in 4 % ( v / v ) PFA ( Electron Microscopy Sciences , Cat # : 15710 ) in CB . Subsequently , both types of samples were washed 3 times for 10min in PBS . Samples were then blocked for 20min in blocking buffer [ 3 % ( w / v ) BSA and 0 . 1 % ( w / v ) Saponin in PBS ] . Clathrin light chain ( Invitrogen , Cat # : MA5 - 11860 , 1 : 200 dilution ) and Halotag ( Promega , Cat # : G9281 , 1 : 200 dilution ) antibodies were used in blocking solution . Primary antibody immunostaining was performed overnight at 4\u00b0C . On the next day , samples were washed three times in washing buffer ( 0 . 1\u00d7 blocking buffer in PBS ) for 10min . Samples were incubated with secondary antibody in blocking buffer for 30min at room temperature in the dark and were washed three times for 10min in washing buffer , and then three times for 10 min in PBS . Mouse secondary antibody ( Jackson ImmunoResearch , Code # : 715 - 005 - 151 ) conjugated with CF680 ( Biothium , Cat # : 92139 ) ( 1 : 50 ) was used to stain clathrin and actin samples . Commercial mouse secondary antibody - AF647 ( ThermoFisher , Cat # : A32787 ; 1 : 400 ) and rabbit secondary antibody ( Jackson ImmunoResearch , Code # : 711 - 005 - 152 ) conjugated with CF680 ( Biothium , Cat # : 92139 ) ( 1 : 50 ) were used to stain the clathrin and ARPC3 - HaloTag . Clathrin and actin samples were then stained with 0 . 5\u00b5M Phalloidin - AF647 ( Fisher Scienti \ufb01 c , Cat # : A22287 ) in PBS and kept at room temperature in the dark for 2 h . Samples were washed three times with PBS before STORM imaging . STORM imaging was performed as previously described on a homebuilt STORM setup 7 , 53 . Samples labeled by AF647 and CF680 were excited by an 647 nm laser . The emission of both AF647 and CF680 was then split into two light paths as two channels using a dichroic mirror ( Chroma , Cat # : T685lpxr ) , and each channel was projected onto one - half of an EMCCD camera ( Andor iXon Ultra 897 ) . Color assignment of each localization was based on its intensity in the two channels . A cylindrical lens was inserted into the transmitted channel to acquire 3D localization 28 . 3D position of each localization was determined from the ellipticity of each point spread function . The raw STORM data was processed according to previously described methods 28 , 54 and single - molecule localization and optical reconstruction were performed using the Insight3 software ( developed by Dr . Bo Huang at University of California , San Francisco and Dr . Xiaowei Zhuang at Harvard University ) . To quantify the distances between centroids of clathrin and actin or ARPC3 signals , we \ufb01 rst manually cropped regions that contain single CME sites associated with actin or ARPC3 structures using Insight3 software . Then we identi \ufb01 ed and saved the xy positions of single - molecule localizations in cropped regions as txt \ufb01 les . We next calculated the centroid of signals by averaging xy positions of each channel and calculated the distance between the centroids of two channels using MATLAB . Airyscan imaging and processing . Live cell sample preparation was performed as described in \u201c TIRF live - cell imaging \u201d . Cells were imaged on a Zeiss LSM 900 inverted microscope using an Airyscan 2 detector and Multiplex 4Y line scanning mode . Cells were maintained at 37\u00b0C , 5 % CO 2 in StemFlex medium . Images were acquired and processed using the ZEN 3 . 1 system . Channels were acquired sequentially for each line with 0 . 2s / frame intervals over 3 min . Images were pro - cessed with 3 . 7 deconvolution strength 2D Airyscan processing . Alignment between channels was corrected using \ufb02 uorescent bead images . Reporting summary . Further information on research design is available in the Nature Research Reporting Summary linked to this article . Data availability The data that support this study are available from the corresponding author upon reasonable request . Source data are provided with this paper . Code availability The Jupyter Notebooks used for live - cell imaging analysis can be found at https : / / doi . org / 10 . 5281 / zenodo . 6575159 . ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 10 NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications Received : 11 September 2021 ; Accepted : 8 June 2022 ; References 1 . Rottner , K . , Faix , J . , Bogdan , S . , Linder , S . & Kerkhoff , E . Actin assembly mechanisms at a glance . J . Cell Sci . 130 , 3427 \u2013 3435 ( 2017 ) . 2 . Lu , R . , Drubin , D . G . & Sun , Y . Clathrin - mediated endocytosis in budding yeast at a glance . J . Cell Sci . 129 , 1531 \u2013 1536 ( 2016 ) . 3 . Lacy , M . M . , Ma , R . , Ravindra , N . G . & Berro , J . Molecular mechanisms of force production in clathrin - mediated endocytosis . FEBS Lett . 592 , 3586 \u2013 3605 ( 2018 ) . 4 . Hassinger , J . E . , Oster , G . , Drubin , D . G . & Rangamani , P . Design principles for robust vesiculation in clathrin - mediated endocytosis . Proc . Natl Acad . Sci . USA 114 , E1118 \u2013 E1127 ( 2017 ) . 5 . Lanzetti , L . Actin in membrane traf \ufb01 cking . Current Opinion in Cell Biology 19 , 453 \u2013 458 ( 2007 ) . 6 . Boulant , S . , Kural , C . , Zeeh , J . C . , Ubelmann , F . & Kirchhausen , T . Actin dynamics counteract membrane tension during clathrin - mediated endocytosis . Nat . Cell Biol . 13 , 1124 \u2013 1132 ( 2011 ) . 7 . Kaplan , C . et al . Load adaptation of endocytic actin networks . Mol . Biol . Cell 33 , ( 2022 ) . 8 . Batchelder , E . M . & Yarar , D . Differential requirements for clathrin - dependent endocytosis at sites of cell - substrate adhesion . Mol . Biol . Cell 21 , 3070 \u2013 3079 ( 2010 ) . 9 . Kaur , S . , Fielding , A . B . , Gassner , G . , Carter , N . J . & Royle , S . J . An unmet actin requirement explains the mitotic inhibition of clathrin - mediated endocytosis . Elife 3 , e00829 ( 2014 ) . 10 . Ferguson , J . P . et al . Mechanoregulation of clathrin - mediated endocytosis . J . Cell Sci . 130 , 3631 \u2013 3636 ( 2017 ) . 11 . Grassart , A . et al . Actin and dynamin2 dynamics and interplay during clathrin - mediated endocytosis . J . Cell Biol . 205 , 721 \u2013 735 ( 2014 ) . 12 . Akamatsu , M . et al . Principles of self - organization and load adaptation by the actin cytoskeleton during clathrin - mediated endocytosis . Elife 9 , e49840 ( 2020 ) . 13 . Mund , M . et al . Systematic nanoscale analysis of endocytosis links ef \ufb01 cient vesicle formation to patterned actin nucleation . Cell 174 , 884 \u2013 896 ( 2018 ) . 14 . Yarar , D . , Waterman - Storer , C . M . & Schmid , S . L . A dynamic actin cytoskeleton functions at multiple stages of clathrin - mediated endocytosis . Mol . Biol . Cell 16 , 964 \u2013 975 ( 2005 ) . 15 . Collins , A . , Warrington , A . , Taylor , K . A . & Svitkina , T . Structural organization of the actin cytoskeleton at sites of clathrin - mediated endocytosis . Curr . Biol . 21 , 1167 \u2013 1175 ( 2011 ) . 16 . Taylor , M . J . , Perrais , D . & Merri \ufb01 eld , C . J . A high precision survey of the molecular dynamics of mammalian clathrin - mediated endocytosis . PLoS Biol . 9 , e1000604 ( 2011 ) . 17 . Kaksonen , M . , Toret , C . P . & Drubin , D . G . A modular design for the clathrin - and actin - mediated endocytosis machinery . Cell 123 , 305 \u2013 320 ( 2005 ) . 18 . Dambournet , D . et al . Genome - edited human stem cells expressing \ufb02 uorescently labeled endocytic markers allow quantitative analysis of clathrin - mediated endocytosis during differentiation . J . Cell Biol . 217 , 3301 \u2013 3311 ( 2018 ) . 19 . Doyon , J . B . et al . Rapid and ef \ufb01 cient clathrin - mediated endocytosis revealed in genome - edited mammalian cells . Nat . Cell Biol . 13 , 331 \u2013 337 ( 2011 ) . 20 . Cocucci , E . , Gaudin , R . & Kirchhausen , T . Dynamin recruitment and membrane scission at the neck of a clathrin - coated pit . Mol . Biol . Cell 25 , 3595 \u2013 3609 ( 2014 ) . 21 . Gibson , T . J . , Seiler , M . & Veitia , R . A . The transience of transient overexpression . Nat . Methods 10 , 715 \u2013 721 ( 2013 ) . 22 . Taylor , M . J . , Lampe , M . & Merri \ufb01 eld , C . J . A feedback loop between dynamin and actin recruitment during clathrin - mediated endocytosis . PLoS Biol . 10 , e1001302 ( 2012 ) . 23 . Merri \ufb01 eld , C . J . , Feldman , M . E . , Wan , L . & Almers , W . Imaging actin and dynamin recruitment during invagination of single clathrin - coated pits . Nat . Cell Biol . 4 , 691 \u2013 698 ( 2002 ) . 24 . Mar\u00eda Cabeza , J . , Acosta , J . & Al\u00e9s , E . Dynamics and Regulation of Endocytotic Fission Pores : Role of Calcium and Dynamin . Traf \ufb01 c 11 , 1579 \u2013 1590 ( 2010 ) . 25 . Ramachandran , R . & Schmid , S . L . Real - time detection reveals that effectors couple dynamin \u2019 s GTP - dependent conformational changes to the membrane . EMBO J . 27 , 27 \u2013 37 ( 2008 ) . 26 . Roux , A . et al . Membrane curvature controls dynamin polymerization . Proc . Natl Acad . Sci . USA 107 , 4141 \u2013 4146 ( 2010 ) . 27 . Iversen , T . G . , Skretting , G . , Van Deurs , B . & Sandvig , K . Clathrin - coated pits with long , dynamin - wrapped necks upon expression of a clathrin antisense RNA . Proc . Natl Acad . Sci . USA 100 , 5175 \u2013 5180 ( 2003 ) . 28 . Huang , B . , Wang , W . , Bates , M . & Zhuang , X . Three - dimensional super - resolution reconstruction microscopy . Science 319 , 810 \u2013 813 ( 2008 ) . 29 . Xu , K . , Babcock , H . P . & Zhuang , X . Dual - objective STORM reveals three - dimensional \ufb01 lament organization in the actin cytoskeleton . Nat . Methods 9 , 185 \u2013 188 ( 2012 ) . 30 . Aguet , F . , Antonescu , C . N . , Mettlen , M . , Schmid , S . L . & Danuser , G . Advances in analysis of low signal - to - noise images link dynamin and AP2 to the functions of an endocytic checkpoint . Dev . Cell 26 , 279 \u2013 291 ( 2013 ) . 31 . Li , D . et al . Extended - resolution structured illumination imaging of endocytic and cytoskeletal dynamics . Science 349 , aab3500 ( 2015 ) . 32 . Cureton , D . K . , Massol , R . H . , Saffarian , S . , Kirchhausen , T . L . & Whelan , S . P . J . Vesicular stomatitis virus enters cells through vesicles incompletely coated with clathrin that depend upon actin for internalization . PLoS Pathog . 5 , e1000394 ( 2009 ) . 33 . Piccinotti , S . , Kirchhausen , T . & Whelan , S . P . J . Uptake of rabies virus into epithelial cells by clathrin - mediated endocytosis depends upon actin . J . Virol . 87 , 11637 \u2013 11647 ( 2013 ) . 34 . Bucher , D . et al . Clathrin - Adaptor ratio and membrane tension regulate the \ufb02 at - To - curved transition of the clathrin coat during endocytosis . Nat . Commun . 9 , 1109 ( 2018 ) . 35 . Saleem , M . et al . A balance between membrane elasticity and polymerization energy sets the shape of spherical clathrin coats . Nat . Commun . 6 , 6249 ( 2015 ) . 36 . Liu , A . P . , Loerke , D . , Schmid , S . L . & Danuser , G . Global and local regulation of clathrin - coated pit dynamics detected on patterned substrates . Biophys . J . 97 , 1038 \u2013 1047 ( 2009 ) . 37 . Sun , Y . et al . Switch - like Arp2 / 3 activation upon WASP and WIP recruitment to an apparent threshold level by multivalent linker proteins in vivo . Elife 6 , e29140 ( 2017 ) . 38 . Yoshida , A . et al . Morphological changes of plasma membrane and protein assembly during clathrin - mediated endocytosis . PLoS Biol . 16 , e2004786 ( 2018 ) . 39 . Cheng , X . et al . Dynamin - dependent vesicle twist at the \ufb01 nal stage of clathrin - mediated endocytosis . Nat . Cell Biol . 23 , 859 \u2013 869 ( 2021 ) . 40 . Kaksonen , M . , Sun , Y . & Drubin , D . G . A pathway for association of receptors , adaptors , and actin during endocytic internalization . Cell 115 , 475 \u2013 487 ( 2003 ) . 41 . Aghamohammadzadeh , S . & Ayscough , K . R . Differential requirements for actin during yeast and mammalian endocytosis . Nat . Cell Biol . 11 , 1039 \u2013 1042 ( 2009 ) . 42 . Ferguson , S . et al . Coordinated actions of actin and BAR proteins upstream of dynamin at endocytic clathrin - coated pits . Dev . Cell 17 , 811 \u2013 822 ( 2009 ) . 43 . Durrbach , A . , Louvard , D . & Coudrier , E . Actin \ufb01 laments facilitate two steps of endocytosis . J . Cell Sci . 109 , 457 \u2013 465 ( 1996 ) . 44 . Lamaze , C . , Fujimoto , L . M . , Yin , H . L . & Schmid , S . L . The actin cytoskeleton is required for receptor - mediated endocytosis in mammalian cells . J . Biol . Chem . 272 , 20332 \u2013 20335 ( 1997 ) . 45 . Miya Fujimoto , L . , Roth , R . , Heuser , J . E . & Schmid , S . L . Actin assembly plays a variable , but not obligatory role in receptor - mediated endocytosis in mammalian cells . Traf \ufb01 c 1 , 161 \u2013 171 ( 2000 ) . 46 . Serwas , D . et al . Article Mechanistic insights into actin force generation during vesicle formation from cryo - electron tomography ll Article Mechanistic insights into actin force generation during vesicle formation from cryo - electron tomography . Dev . Cell 57 , 1132 \u2013 1145 ( 2022 ) . 47 . Xu , R . & Du , S . Overexpression of lifeact - GFP disrupts F - actin organization in cardiomyocytes and impairs cardiac function . Front . Cell Dev . Biol . 9 , 746181 ( 2021 ) . 48 . Flores , L . R . , Keeling , M . C . , Zhang , X . , Sliogeryte , K . & Gavara , N . Lifeact - GFP alters F - actin organization , cellular morphology and biophysical behaviour . Sci . Rep . 9 , 3241 ( 2019 ) . 49 . Grossier , J . P . , Xouri , G . , Goud , B . & Schauer , K . Cell adhesion de \ufb01 nes the topology of endocytosis and signaling . EMBO J . 33 , 35 \u2013 45 ( 2014 ) . 50 . Hong , S . H . , Cortesio , C . L . & Drubin , D . G . Machine - learning - based analysis in genome - edited cells reveals the ef \ufb01 ciency of clathrin - mediated endocytosis . Cell Rep . 12 , 2121 \u2013 2130 ( 2015 ) . 51 . Grimm , J . B . et al . A general method to \ufb01 ne - tune \ufb02 uorophores for live - cell and in vivo imaging . Nat . Methods 14 , 987 \u2013 994 ( 2017 ) . 52 . Loerke , D . et al . Cargo and dynamin regulate clathrin - coated pit maturation . PLoS Biol . 7 , 0628 \u2013 0639 ( 2009 ) . 53 . Wojcik , M . , Hauser , M . , Li , W . , Moon , S . & Xu , K . Graphene - enabled electron microscopy and correlated super - resolution microscopy of wet cells . Nat . Commun . 6 , 7384 ( 2015 ) . 54 . Rust , M . J . , Bates , M . & Zhuang , X . Sub - diffraction - limit imaging by stochastic optical reconstruction microscopy ( STORM ) . Nat . Methods 3 , 793 \u2013 795 ( 2006 ) . Acknowledgements M . J . was funded by American Heart Association Postdoctoral Fellowship ( 18POST34000029 ) . D . G . D . was funded by NIH MIRA grant R35GM118149 . K . X . is a NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 ARTICLE NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications 11 Chan Zuckerberg Biohub investigator and acknowledges support from NIH ( DP2GM132681 ) . S . U . is a Chan Zuckerberg Biohub Investigator and acknowledges support from the Philomathia Foundation , and Chan Zuckerberg Initiative Imaging Scientist program . We would like to thank Dr . Yidi Sun and Dr . Matthew Akamatsu for insightful comments on the manuscript ; the Conklin Lab at UCSF for providing the WTC10 human iPSC line ; the Lavis Lab at Janelia Research Campus for providing JF635 HaloTag ligand ; Dr . Sun Hae Hong for generating the AP2 - tagRFP - T iPSC line ; the UC Berkeley QB3 MacroLab for puri \ufb01 ed S . pyogenes NLS - Cas9 ; the Luo Lab at UC Berkeley for sharing their electroporator and the UC Berkeley Cancer Research Laboratory Flow Cytometry Facility for iPSC sorting . Author contributions M . J . , C . S . and D . G . D . conceived the study and experiments . M . J . and A . Y . generated the genome - edited cell lines . M . J . performed live cell data acquisition and sample preparation for super - resolution microscopy . B . W . and K . X . performed super - resolution microscopy and super - resolution data reconstruction . C . S . developed computational analysis tools and S . U . , J . S . , D . G . D . and M . J . supported the data analysis . M . J . , C . S . , and D . G . D . prepared the \ufb01 gures and wrote the manuscript with feedback from the other authors . Competing interests The authors declare no competing interests . Additional information Supplementary information The online version contains supplementary material available at https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 . Correspondence and requests for materials should be addressed to David G . Drubin . Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work . Peer reviewer reports are available . Reprints and permission information is available at http : / / www . nature . com / reprints Publisher \u2019 s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional af \ufb01 liations . Open Access This article is licensed under a Creative Commons Attribution 4 . 0 International License , which permits use , sharing , adaptation , distribution and reproduction in any medium or format , as long as you give appropriate credit to the original author ( s ) and the source , provide a link to the Creative Commons license , and indicate if changes were made . The images or other third party material in this article are included in the article \u2019 s Creative Commons license , unless indicated otherwise in a credit line to the material . If material is not included in the article \u2019 s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use , you will need to obtain permission directly from the copyright holder . To view a copy of this license , visit http : / / creativecommons . org / licenses / by / 4 . 0 / . \u00a9 The Author ( s ) 2022 ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 12 NATURE COMMUNICATIONS | ( 2022 ) 13 : 3578 | https : / / doi . org / 10 . 1038 / s41467 - 022 - 31207 - 5 | www . nature . com / naturecommunications",
    "day2015microtubule_supplement1": "Supplemental Figure 1 . ( Related to Figure 1 ) This figure shows a series of control experiments that characterize the properties of the tubular invaginations generated in response to ATP depletion . It also documents the efficacy of the ATP depletion treatment . ( A ) Alexa488 - STxB and Alexa488 - CTxB accumulate in tubular invaginations in ATP depleted COS - 7 cells . To facilitate STxB labeling , cells were transfected with Gb3 synthase . ( B , C ) As reported previously ( 78 ) , fluorescent CTxB labels both plasma membrane invaginations ( arrowheads ) and plasma membrane protrusions ( arrows ) in ATP depleted cells . B and C are single frames from a confocal z - stack . Dashes mark the position of the xz - sections shown below . ( D ) Alexa488 - STxB and Alexa555 - CTxB colocalize in invaginations in ATP depleted COS - 7 cells expressing Gb3 synthase . ( E ) Tubules prelabeled with Alexa555 - CTxB filled at approximately the same rate as newly added Alexa488 - CTxB accumulates at the plasma membrane , indicating that the tubules are open to the fluid phase in ATP depleted cells . Time stamps are in minutes : seconds . ( F - H ) ATP depletion inhibits uptake of dextran and internalization of transferrin ( Tfn ) ( Mean \u00b1 SD , N = 325 - 417 cells ) . * * * , p < 0 . 001 , Student\u2019s t - test . Bars , 10 \u00b5 m .",
    "sathe2018small": "ARTICLE Small GTPases and BAR domain proteins regulate branched actin polymerisation for clathrin and dynamin - independent endocytosis Mugdha Sathe 1 , Gayatri Muthukrishnan 1 , James Rae 2 , 3 , Andrea Disanza 4 , 5 , Mukund Thattai 1 , 6 , Giorgio Scita 4 , 5 , Robert G . Parton 2 , 3 & Satyajit Mayor 1 , 7 Using real - time TIRF microscopy imaging , we identify sites of clathrin and dynamin - independent CLIC / GEEC ( CG ) endocytic vesicle formation . This allows spatio - temporal localisation of known molecules affecting CG endocytosis ; GBF1 ( a GEF for ARF1 ) , ARF1 and CDC42 which appear sequentially over 60 s , preceding scission . In an RNAi screen for BAR domain proteins affecting CG endocytosis , IRSp53 and PICK1 , known interactors of CDC42 and ARF1 , respectively , were selected . Removal of IRSp53 , a negative curvature sensing protein , abolishes CG endocytosis . Furthermore , the identi \ufb01 cation of ARP2 / 3 complex at CG endocytic sites , maintained in an inactive state reveals a function for PICK1 , an ARP2 / 3 inhibitor . The spatio - temporal sequence of the arrival and disappearance of the molecules suggest a mechanism for a clathrin and dynamin - independent endocytic process . Coincident with the loss of PICK1 by GBF1 - activated ARF1 , CDC42 recruitment leads to the activation of IRSp53 and the ARP2 / 3 complex , resulting in a burst of F - actin polymerisation potentially powering scission . DOI : 10 . 1038 / s41467 - 018 - 03955 - w OPEN 1 National Centre for Biological Science ( TIFR ) , Bellary Road , Bangalore 560065 , India . 2 Institute for Molecular Bioscience , University of Queensland , Brisbane , QLD 4072 , Australia . 3 Centre for Microscopy and Microanalysis , University of Queensland , Brisbane , QLD 4072 , Australia . 4 IFOM , Fondazione Istituto FIRC di Oncologia Molecolare , Milan 20139 , Italy . 5 Department of Oncology and Hemato - Oncology , University of Milan , Milan 20122 , Italy . 6 Simons Centre for the Study of Living Machines , National Centre for Biological Sciences ( TIFR ) , Bellary Road , Bangalore 560065 , India . 7 Institute for Stem Cell Biology and Regenerative Medicine , Bellary Road , Bangalore 560065 , India . These authors contributed equally : Mugdha Sathe , Gayatri Muthukrishnan . Correspondence and requests for materials should be addressed to S . M . ( email : mayor @ ncbs . res . in ) NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications 1 1 2 3 4 5 6 7 89 0 ( ) : , ; There are amendments to this paper M ultiple endocytic pathways function in a eukaryotic cell 1 , 2 ; however , our understanding of the endocytic process is mainly derived from studies on clathrin - mediated endocytosis ( CME ) 3 \u2013 5 . Dynamin is responsible for vesicle scission in CME 6 , 7 and a host of clathrin - independent endocytic ( CIE ) pathways , such as the caveolar and fast endophilin - mediated endocytic pathway 8 \u2013 10 . On the other hand , among CIE pathways , the CLIC / GEEC [ clathrin and dynamin - independent carriers which form GPI - enriched endocytic com - partments ; CG ] pathway functions independent of both clathrin and dynamin in multiple cell types and contexts 11 \u2013 17 , while it is not present in others 18 . The actin polymerisation machinery has been implicated in the functioning of many CIE pathways at different stages 13 , 19 . Our focus , the CG pathway , is regulated by the small GTPases , ARF1 ( ADP - ribosylation factor 1 ) and CDC42 ( cell division control protein 42 ) 11 \u2013 16 . It is responsible for the uptake of many glycosylphosphotidylinositol ( GPI ) - anchored proteins , a major fraction of the \ufb02 uid phase , toxins such as Helicobacter pylori vacuolating toxin A 20 , cholera toxin 21 and viruses like adeno - associated virus 2 22 . The CLICs are formed in a polarised manner at the leading edge of migrating cells 23 and , the resulting GEECs subsequently fuse with the sorting endocytic vesicles via a Rab5 / phosphatidylinositol - 3 - kinase - dependent mechanism 24 . The CLICs / GEECs are high capacity endocytic carriers turning over the entire membrane surface in 12 min in \ufb01 broblasts , highlighting the role of CG pathway in regulating membrane homoeostasis 23 . Recent evidence suggests that this is required for generating a tubular vesicular endocytic network during cytokinesis 25 and serves to deliver ligands to their signalling receptors in a common endocytic compartment 26 . The molecular machinery to form a CG endocytic vesicle involves activating ARF1 at the plasma membrane by GBF1 ( Golgi - speci \ufb01 c brefeldin A resistance factor 1 ) 16 , a speci \ufb01 c ARF - GEF ( guanine nucleotide exchange factor ) . GTP \u2013 ARF1 recruits ARHGAP21 ( a RhoGAP for CDC42 ) , which removes CDC42 from the membrane 14 . Cholesterol removal , in addition , reduces the recruitment of ARF1 and CDC42 , along with accelerated cycling of CDC42 13 , 14 . Lastly , the CG pathway requires dynamic actin since both stabilisation and de - polymerisation of actin \ufb01 laments was found to affect endocytosis 13 . By visualising a forming CG endocytic vesicle , we wanted to understand the molecular mechanism responsible . We adapted a pH pulsing protocol that exploits the pH - sensitive properties of super ecliptic GFP ( SecGFP ) 27 , previously deployed to study CME 4 , 28 , 29 . We tagged the GPI - anchor with SecGFP to make model cargo SecGFP - GPI to assay , in real time , the sites of endocytic vesicle formation . We found that the CG endocytic vesicle formation was initiated by the accumulation of ARF1 / GBF1 followed by CDC42 and F - actin while dynamin and cla - thrin did not associate with forming endosomes . Hence , in the absence of a discernable coat 23 , alternate candidate proteins by generating / stabilising membrane curvature can assist in endocytic vesicle formation such as Bin / Amphiphysin / Rvs ( BAR ) domain - containing proteins ( BDPs ) 30 . Although several BDPs are involved in the CME pathway , only one has been identi \ufb01 ed to be associated with the CG endocytic pathway so far 31 . Using RNAi - screening , we identi \ufb01 ed two BDPs in particular , that affected CG endocytosis downstream of ARF1 and CDC42 . First , a CDC42 interaction partner and I - BAR protein , IRSp53 ( Insulin - responsive protein of mass 53 kDa ) was found to be necessary for CG endocytosis . Importantly , IRSp53 removal resulted in the disappearance of CLICs and loss of a GBF1 - dependent endocytic pathway . Second , an ARF1 interactor , PICK1 ( protein interacting with C kinase 1 ) emerged as a regulator of ARP2 / 3 activity in the early phases of CG vesicle formation . Lastly , ARP2 / 3 , an interaction partner of both IRSp53 and PICK1 , accumulated at the forming CG endocytic site and decreased CG endocytosis when inhibited . Together , the spatio - temporal dynamics of these proteins provided a mechanistic understanding of the forming CG endocytic vesicle . Results pH pulsing assay detects nascent CG endocytic sites . To monitor endocytic vesicle formation in real time , we employed the pH - sensitive \ufb02 uorescence of super ecliptic pHlourin GFP ( SecGFP ) 27 attached to a GPI - anchor ( SecGFP - GPI ) to differ - entiate cell surface - resident molecules from the newly internalised molecules . The \ufb02 uorescence of SecGFP is quenched reversibly when exposed to pH 5 . 5 4 , 27 \u2013 29 . SecGFP - GPI expressed in AGS ( human gastric cell line ) was endocytosed along with the CG cargo , \ufb02 uid phase ( 10 kDa dextran ) , but not with CME cargo , TfR , at both 37 \u00b0C and 30 \u00b0C ( Supplementary Fig . 1a \u2013 c and Supplementary Information ( S . I . ) ) as shown previously 15 , 32 . Endocytic events were identi \ufb01 ed by alternately exposing the cells to buffers equilibrated to pH 7 . 3 ( pH 7 ) or pH 5 . 5 ( pH 5 ) every 3 s at 30 \u00b0C ( Fig . 1a , Schematic and Supplementary Movies 1 \u2013 2 ) . SecGFP - GPI - containing endocytic events occurring during exposure to pH 7 remained \ufb02 uorescent due to their near neutral luminal pH right after formation . However , the buffer exchange from pH 7 to pH 5 quenched the \ufb02 uorescence of cell surface SecGFP - GPI . This enabled visualisation of the newly formed endocytic vesicle . The identi \ufb01 cation of the site of endocytic vesicle formation paved the way for the character - isation of the spatial and temporal dynamics of molecular players by the co - expression of a ( mCherry / TagRFPt / pRuby ) - tagged molecule of interest , \u2018 X - FP \u2019 ( Fig . 1a and Supplementary Movies 1 \u2013 2 ) . The dynamics of X - FP were extracted by looking at the history of the region , where vesicle formation was detected ( Fig . 1a montage ( bottom ) , see Methods and S . I . ) . To rule out the effect of pH 5 on the rate of endocytosis , we pre - treated the cells with either pH 5 or pH 7 buffer , followed by 5 - min pulse in pH 7 buffer and found no difference in CG endocytosis ( Supplemen - tary Fig . 1e ) . The pH pulsing movies were analysed using semi - automated scripts ( see Methods and S . I . ) . Brie \ufb02 y , the centroid of new spots appearing in the pH 5 channel provided a \ufb01 duciary marker for the time and location of the nascent endocytic vesicle ( Supplementary Fig . 1d , step 1 and S . I . ) . The relative enrichment of SecGFP - GPI and X - FP at the endocytic site was determined by normalising the average \ufb02 uorescence of the nascent endocytic spot to its local background annulus ( Supplementary Fig . 1a , d , step 2 \u2013 3 and see S . I . ) . The spots were then put through a series of automated and manual checks . The automated check ensured that the pH 5 intensity of the new spot had ( i ) signi \ufb01 cantly higher intensity than the background , ( ii ) persisted for at least 6 s and ( iii ) did not show an increase in intensity in the subsequent frame . Subsequently , a manual check was performed on the montages , ( i ) to remove any false positives that might have been missed by the automated check and ( ii ) to classify the new SecGFP - GPI spots into two groups based on whether X - FP co - detection was observed or not ( see S . I . ) . The data at the site of the spot were represented as the average fold change over the surrounding background , as a function of time ( Fig . 1b , solid traces ) , and compared to the average fold change of arbitrary regions within the cell ( \u2018 Random \u2019 ) ( Fig . 1b , dashed traces ) . The pro \ufb01 le obtained ( pooled from multiple cells ) represented a spatial and temporal pro \ufb01 le of the X - FP at SecGFP - GPI endocytic sites . Using the pH pulsing assay , we found that the rise in pH 7 SecGFP - GPI intensity occurred only ~ 3 s prior to vesicle ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w 2 NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications generation as opposed to nearly 40 s for CME , followed by monitoring SecGFP - TfR internalisation ( ref . 4 and Supple - mentary Fig . 2c ) . Furthermore , we observed a poor correlation ( r = 0 . 3 ) between pH 7 vs . pH 5 intensity per endosome , indicating that SecGFP - GPI is endocytosed without a major concentration in the endocytic vesicle ( Fig . 1c ) . In comparison to CME , it should be noted that we faced two main challenges , ( i ) lack of a cytoplasmic marker ( for e . g . , clathrin ) for the endocytic site , ( ii ) lack of strong concentration of the cargo prior to the endocytic vesicle pinching . Regardless , the protocol developed provided a reliable real - time assay for studying the spatio - temporal dynamics of the internalisation of GPI - anchored proteins ( see , S . I . ) . Cytoplasm a pH 7 . 0 pH 5 . 0 pH 7 . 0 pH 7 . 0 pH 5 . 0 secGPI - quenched Cytoplasm Extracellular milieu Cytoplasm Extracellular milieu Cytoplasm Extracellular milieu Extracellular milieu b c e d f \u2013 40 \u2013 30 \u2013 20 \u2013 10 0 10 20 \u2013 30 \u2013 25 \u2013 20 \u2013 15 \u2013 10 \u2013 5 0 Time ( s ) ARF1 Random \u2013 40 \u2013 30 \u2013 20 \u2013 10 0 10 20 1 . 0 1 . 2 1 . 4 1 . 6 \u2013 9 \u2013 6 \u2013 3 0 Time ( s ) CDC42 Random F o l d c hange ( a . u . ) \u2013 40 \u2013 30 \u2013 20 \u2013 10 0 10 20 \u2013 8 \u2013 6 \u2013 4 \u2013 2 0 Time ( s ) GBF1 Random Log10 ( p ) pH 5 pH 5 Random pH 7 Random pH 7 r = 0 . 3 secGPI - fluorescent N o r m a li s ed p H 7 i n t en s i t y ( a . u . ) Normalised pH 5 intensity ( a . u . ) Time ( s ) F o l d c hange ( a . u . ) 2 . 4 2 . 2 2 . 0 1 . 8 1 . 6 1 . 4 1 . 2 1 . 0 0 . 8 \u2013 42 \u2013 36 \u2013 30 \u2013 24 \u2013 18 \u2013 12 \u2013 6 0 6 12 18 50 100 150 200 250 120 100 80 60 40 20 0 1 . 0 1 . 2 1 . 4 1 . 6 1 . 0 1 . 2 1 . 4 1 . 6 pH 5 pH 7 ARF1 GBF1 pH 5 pH 7 pH 7 pH 5 CDC42 Time ( s ) \u2013 18 18 15 12 9 6 3 0 \u2013 3 \u2013 6 \u2013 9 \u2013 12 \u2013 15 Time ( s ) \u2013 18 18 15 12 9 6 3 0 \u2013 3 \u2013 6 \u2013 9 \u2013 12 \u2013 15 Time ( s ) \u2013 18 18 15 12 9 6 3 0 \u2013 3 \u2013 6 \u2013 9 \u2013 12 \u2013 15 pH 5 pH 7 X - FP pH 7 pH 5 X - FP \u2013 12 \u2013 9 \u2013 3 12 15 18 Time ( s ) 6 9 \u2013 18 3 0 \u2013 6 \u2013 15 NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w ARTICLE NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications 3 GBF1 , ARF1 and CDC42 are recruited to nascent CG endo - cytic sites . We visualised the temporal dynamics of co - expressed CDC42 at the site of formation of nascent SecGFP - GPI endocytic vesicles . TagRFPt - CDC42 recruitment was quanti \ufb01 ed as average fold accumulation of CDC42 relative to the local background in all the endocytic events recorded over a 40 s time window straddling the endocytic event . A signi \ufb01 cant change in the intensity of CDC42 began at around \u2212 9 s and peaked around 0 to + 3 s ( Supplementary Fig . 2a ) . Based on the presence or absence of a co - detected CDC42 spot ( within \u2212 18 to + 18 s time window ) at the endocytic site , we identi \ufb01 ed two populations via manual classi \ufb01 cation ( see S . I . for a detailed description ) . We called them CDC42 Coloc ( co - detection of CDC42 and SecGFP - GPI ) and CDC42 NoColc ( the remainder ) . We compared the fold accu - mulation of all the CDC42 spots ( CDC42 All ) , CDC42 Coloc and CDC42 NoColoc with the Random . While the CDC42 Coloc pro \ufb01 le was similar to that of CDC42 All , the CDC42 NoColoc pro \ufb01 le was comparable to the Random ( Supplementary Fig . 2b ) . Thus , the endocytic sites detected by our assay consisted of two populations wherein one fraction exhibited an accumulation of CDC42 while the second fraction failed to show a discernable accumulation . CDC42 Coloc corresponded to the 56 % ( of the CDC42 All ) SecGFP - GPI endocytic sites . As the removal of the events which did not coincide with the presence of CDC42 did not alter CDC42 recruitment pro \ufb01 le ( compare , Fig . 1d and Supplementary Fig . 2a ) , they were discarded from further ana - lysis . While the reasons for not detecting CDC42 at all endocytic events is a function of both the signal and noise in the data , it may also re \ufb02 ect a genuine lack of recruitment at some endocytic events ( see S . I . for a detailed explanation ) . Henceforth , for all X - FPs , we report pH pulsing traces that were classi \ufb01 ed as co - detected with the SecGFP - GPI ( see Methods , S . I . and Table 1 , for methods and statistical details and analysis ) . We next examined another CG pathway regulator , ARF1 14 , and found that mCherry - ARF1 was already accumulated at the forming CG endocytic sites at \u2212 36 s and peaked at 0 s ( Fig . 1e , Supplementary Fig . 6a and Table 1 ) . Predictably , mCherry - GBF1 , an ARF1 - GEF 16 , was also recruited to SecGFP - GPI spots ( Fig . 1f , Supplementary Fig . 6b and Table 1 ) . The temporal pro \ufb01 le of GBF1 was correlated to ARF1 ( r = 0 . 6 , Table 2 ) . When we extended the time window of observation for ARF1 and GBF1 further back in time ( nearly 60 s before scission ) , we concluded that the accumulation of ARF1 and GBF1 at CG endocytic sites initiates as early as \u2212 54 s ( Supplementary Fig . 2f \u2013 g ) . In absence of other molecules upstream of GBF1 and ARF1 , this pair currently serves as the earliest initiators of the CG endocytic pathway . Furthermore , when we assessed the ultrastructure of newly formed \ufb02 uid - \ufb01 lled endosomes by electron microscopy ( EM ) 23 in cells treated with the small - molecule inhibitor of GBF1 , LG186 33 , the number of CLICs and \ufb02 uid uptake was drastically reduced , whereas , CME - derived vesicles and uptake appeared relatively unaffected ( Supplementary Fig . 1e \u2013 f ) . In contrast to ARF1 , GBF1 and CDC42 , we did not observe a frequent recruitment of clathrin and dynamin to the CG endocytic sites . The recruitment pro \ufb01 le was comparable to random for > 65 % endocytic events for both mCherry - clathrin and dynamin ( Supplementary Fig . 2d \u2013 e and Table 1 ) . The remainder of the pro \ufb01 les exhibited high levels of clathrin and dynamin at SecGFP - GPI endocytic sites . Never - theless , both fraction showed a temporal trend similar to that observed for random ( Supplementary Fig . 2d \u2013 e , compare grey , blue and random traces ) . Conversely , mCherry - dynamin was recruited to SecGFP - TfR endocytic sites proximal to scission in 80 % of the cases ( Supplementary Fig . 2c ) in agreement with a previous study 4 . Despite being responsible for the timed removal of CDC42 from the plasma membrane 14 , ARF1 ( and GBF1 ) was recruited long before the recruitment of CDC42 . The pH pulsing analysis thus revealed a surprising facet of the recruitment of ARF1 ( and GBF1 ) , which suggested a CDC42 - independent function ( s ) for ARF1 in CG endocytosis . Taken together , these results establish a reliable real - time imaging assay to follow newly formed CG endocytic vesicles containing SecGFP - GPI , correlated with recruitment of its known regulators , ARF1 , GBF1 and CDC42 . Identi \ufb01 cation of BAR domain proteins in CG endocytosis . CLICs , visualised within 15 s of their formation , have a pleomorphic tubular appearance and lack a discernable protein coat when visualised by EM 21 , 23 . This prompted us to investigate Fig . 1 Identi \ufb01 cation of newly formed SecGFP - GPI endocytic vesicles using a pH pulsing assay . a Schematic ( top panel ) of the pH pulsing assay depicting the effect of pH on SecGFP - GPI \ufb02 uorescence during endocytic vesicle budding . Note the \ufb02 uorescence of SecGFP - GPI is retained at high pH ( top and bottom left panel ) or when exposed to low pH if sequestered in a newly formed endocytic vesicle ( bottom right panel ) , and quenched only when exposed to low pH ( top right panel ) . Sequential TIRF images of AGS cell co - expressing SecGFP - GPI and mCherry - ARF1 collected at pH 7 , pH 5 and in the RFP channels ( middle panel ) . Newly formed endocytic vesicles ( inset ) ( identi \ufb01 ed as in Supplementary Figure 1c ) are used to construct a single frame ( yellow rectangle ) of the montage depicted ( bottom panel ) . b Average of the normalised \ufb02 uorescence intensities of pH 5 and pH 7 traces at the site of newly formed SecGFP - GPI endocytic vesicles compared to their respective random traces constructed from 120 endocytic events ( pH 5 and pH 7 ) and 3428 random spots , derived from 17 cells pooled from four independent experiments . c The graph shows the fold enrichment of \ufb02 uorescence intensity over the local background of pH 5 vs . pH 7 at the time of formation of the endocytic vesicles ( data from 1b ) . d \u2013 f Graphs show the average normalised \ufb02 uorescence intensity vs . time traces for the recruitment of TagRFPt - CDC42 ( d ) , mCherry - ARF1 ( e ) and mCherry - GBF1 ( f ) to the forming SecGFP - GPI endocytic sites , and its corresponding random intensity trace ( n , Table 1 ) . The random traces were derived from randomly assigned spots of the same radius as the endocytic regions , as detailed in S . I . Endocytic distribution at each time point was compared to the random distribution by Mann \u2013 Whitney U test , and the log 10 ( p ) [ log 10 ( 0 . 05 ) is \u2212 1 . 3 and log 10 ( 0 . 001 ) is \u2212 2 . 5 ] is plotted below each trace ( d \u2013 f ) . Representative montages from the respective data sets are depicted below the graphs ( d \u2013 f ) . Arrowheads indicate the newly formed endocytic vesicle . Error bars , s . e . m . ( b , d \u2013 f ) . Scale bar , 1 . 5 \u00b5 m ( a , d \u2013 f ) Table 1 pH pulsing assay data set Molecule % Coloc # Spots SecGPI # Spots Random # Cell and Experiment CDC42 56 219 3428 17 and 6 ARF1 62 411 4952 12 and 3 GBF1 62 132 1917 6 and 2 IRSp53 60 309 4439 7 and 3 ARP3 48 170 3277 16 and 6 Lifeact 61 244 4277 7 and 3 PICK1 77 121 1968 22 and 10 N - WASP 27 256 1827 9 and 3 Clathrin 30 437 1801 8 and 4 Dynamin 34 130 1866 10 and 4 See section titled pH pulsing analysis ( Methods and S . I ) for details ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w 4 NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications the role of BAR domain proteins ( BDPs ) in CG endocytosis due to their capability to sense or stabilise membrane curvature . Additionally , BDPs contain domains that can bind lipids and / or regulate actin machinery , including RhoGTPases 30 . Thus , we performed a dsRNA screen in S2R + cells 15 , 16 to identify the BDPs involved in CG endocytosis using GBF1 ( garz ) and GFP dsRNA as positive and negative control , respectively . The screen yielded 10 \u2018 hits \u2019 which affected \ufb02 uid - phase uptake ( Fig . 2a , b and Supplementary Fig . 3 ) . Predictably , endophilin A required for dynamin - dependent CIE endocytosis of GPCRs and Shiga Toxin 9 , 10 was not selected , whereas , GRAF1 31 previously shown to be involved in CG endocytosis was identi \ufb01 ed as a \u2018 hit \u2019 . The other hits , sorting nexin 6 34 and centaurin \u03b2 1A 35 have been implicated in late stages of endosomal traf \ufb01 cking ; CIP4 36 and NWK 37 being dynamin interactors were not pursued in this study . We focused , instead , on two classes of BDPs , MIM / CG32082 ( I - BAR domain ) and PICK1 ( BAR domain ) , primarily due to their interactions with CDC42 and ARF1 , respectively . IRSp53 is necessary for CG endocytosis . IRSp53 , the mamma - lian orthologue of CG32082 38 , has been implicated in \ufb01 lopodia formation . IRSp53 has a multi - domain architecture consisting of I - BAR , CRIB , SH3 and PDZB domains . Using its SH3 domain , IRSp53 is known to interact with many actin regulatory proteins such as WASP - family verprolin - homologous protein 2 ( WAVE2 ) 39 \u2013 41 , Mena / VASP ( vasodilator - stimulated phosphoprotein ) 42 , 43 , Eps8 42 , 44 \u2013 46 , mDia 40 ( Fig . 3a ) . Furthermore , the recruitment of a b GFP garz PICK1 SNX6 CG32082 CG8176 FER ortholog SNX1 ENDO A AMPH GRAF CENT \u03b2 1A ARFAPTIN RHO92B ICA69 SYNDAPIN CIP4 NWK MIM Serial no . FlyBase ID Gene name ( D . melanogaster ) Mamalian ortholog ( DIOPT ) 1 . 2 Normalised endocytosis 1 . 0 0 . 8 0 . 6 0 . 4 0 . 2 0 . 0 1 FBgn0264560 Gartenzwerg ( garz ) GBF1 2 FBgn0032447 PICK1 ( Protein interacting with C kinase 1 ) PICK1 3 FBgn0032005 Sortin Nexin 6 ( SNX6 ) SNX6 , SNX32 , SNX5 4 FBgn0052082 CG32082 BAIAP2 / IRSp53 , BAIAP2L1 , BAIAP2L2 5 FBgn0037702 CG8176 FCHO2 , FCHO1 , 6 FBgn0000723 FER ortholog FER , FES 7 FBgn0031534 Sortin Nexin 1 ( SNX1 ) SNX2 , SNX1 , SNX33 , SNX18 , SNX30 8 FBgn0038659 Endophilin A ( Endo A ) SH3GL3 , SH3GL2 , SH3GL1 9 FBgn0027356 Amphiphysin ( Amph ) BIN1 , AMPH 10 FBgn0030685 GRAF ortholog ARHGAP26 11 FBgn0039056 Centaurin beta 1A ACAP2 , ACAP3 , ACAP1 12 FBgn0037884 Arfaptin ARFIP2 , ARFIP1 13 FBgn0038747 Rho GTPase activating protein at 92B ARHGAP17 , ARHGAP44 14 FBgn0037050 ICA69 ICA1L , ICA1 15 FBgn0053094 Syndapin PASCIN1 , PASCIN2 16 FBgn0035533 CIP4 ( Cdc42 - interacting protein 4 ) CIP4 17 FBgn0263456 Nervous wreck ( Nwk ) FCHSD1 , FCHSD2 18 FBgn0053558 Missing - in - metastasis ( MIM ) MTSS1L , MTSS1 3327 2791 985 737 779 884 848 736 704 450 539 671 606 844 834 572 794 747 707 Fig . 2 RNAi screen reveals BAR domain proteins involved in CG endocytosis . a List of Drosophila proteins in the PFAM database that contained one of the following BAR domains , PFAM IDs \u2014 PF06456 , PF09325 , PF06456 , PF00611 and PF08397 . The list was \ufb01 ltered to remove duplicates to give 18 genes . b The histogram shows normalised 5 - min \ufb02 uid - phase uptake in S2R + cells treated with 10 \u00b5 g of dsRNA for 4 days as indicated with dsRNA against GBF1 ( garz ) as positive , and GFP as negative controls . In a single experiment , mean uptake of one of GFP dsRNA coverslip was used to normalise the mean for rest of the coverslips . Data were pooled from three independent experiments and the cell numbers are indicated in the graph . The bars in green are signi \ufb01 cantly different from GFP dsRNA using two - sample t - test ( p value < 0 . 05 ) . Version 27 of the PFAM database was used to generate the list Table 2 Cross - correlation calculated for indicated traces Molecule pair Time interval ( s ) ( \u2212 36 s to \u2212 12 s ) Speci \ufb01 c interval GBF1 vs CDC42 0 . 5767 GBF1 vs LIFEACT n . s . GBF1 vs ARP3 n . s . GBF1 vs IRSp53 n . s . ARF1 vs GBF1 0 . 6323 ARF1 vs CDC42 0 . 564 ARF1 vs IRSp53 n . s . ARF1 vs ARP3 0 . 6347 ARF1 vs LIFEACT 0 . 7478 CDC42 vs IRSp53 0 . 65 a , b 0 . 5698 a ( \u2212 33 s to 0 s ) CDC42 vs LIFEACT 0 . 6087 0 . 54 ( \u2212 36 s to + 6 s ) CDC42 vs ARP3 n . s . ARP3 vs IRSp53 n . s . 0 . 7739 a ( \u2212 9 s to + 9 s ) ARP3 vs LIFEACT n . s . 0 . 6292 ( \u2212 33 s to 0 s ) ARF1 vs PICK1 n . s . 0 . 5698 ( \u2212 36 s to \u2212 12 s ) PICK1 vs IRSp53 \u2212 0 . 5353 Rest of the calculations are done with traces with spot radius = 3 pixels and background donut = 6 \u2013 8 pixels 1 pixel = 84nm n . s . not signi \ufb01 cant p value considered here is 0 . 05 calculated via t - statistic a Performed for traces with spot radius = 2 pixels and background donut = 6 \u2013 8 pixels b Performed with one frame shift NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w ARTICLE NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications 5 IRSp53 to the plasma membrane was compromised following ARF1 depletion 47 . Hence , IRSp53 was a good candidate to act as a signalling platform , linking CDC42 activation , membrane cur - vature and actin regulation for CG endocytosis . To address the function of IRSp53 we compared endocytosis between mouse embryonic \ufb01 broblasts ( MEFs ) generated from IRSp53 \u2212 / \u2212 mice ( IRSp53 \u2212 / \u2212 MEFs ) and IRSp53 \u2212 / \u2212 IRSp53WT addback MEFs ( IRSp53WT MEFs ) 42 . Loss of IRSp53 caused a signi \ufb01 cant reduction in \ufb02 uid - phase uptake , without affecting TfR internalisation ( Fig . 3b ) . We next addressed the nature of endocytosis in IRSp53 \u2212 / \u2212 MEFs and found that the \ufb02 uid - phase uptake in IRSp53 \u2212 / \u2212 MEFs remained refractory to * * n . s . d IRSp53 \u2013 / \u2013 LG186 PM PM PM 1 \u03bc m 1 \u03bc m 5 - min HRP pulse a b I - BAR CRIB / GBD SH3 WW Dimer inactive Dimer active PDZ - BD c IRSp53 WT IRSp53 \u2013 / \u2013 F L U I D TF TF R 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 1 . 4 * FluidTf / TfR n . s . IRSp53 \u2013 / \u2013 IRSp53 WT N o r m a li s ed endo cy t o s i s ( a . u . ) 1 . 6 n . s . IRSp53 WT LG186 ( 10 \u03bc M ) \u2013 \u2013 + + \u2013 + + \u2013 0 . 0 0 . 8 1 . 0 1 . 2 1 . 4 N o r m a li s ed endo cy t o s i s ( a . u . ) I R S p53 W T I R S p53 \u2013 / \u2013 DMSO LG186 WT PM CLICs EE CCV / Cav 0 2 4 6 8 10 12 14 16 WT MEFs IRSp53 \u2013 / \u2013 WT MEFs LG186 n . s . n . s . n . s . * * * N u m be r pe r f i e l d 319 273 305 235 418 325 403 327 Fig . 3 IRSp53 is involved in CG endocytosis . a Schematic depicting the domain organisation of IRSp53 . IRSp53 exists in an inactive dimer state , which upon binding to GTP - CDC42 is activated , allowing SH3 domain to bind to its effectors . b Histogram ( left ) shows 5 - min uptake of TfR and \ufb02 uid phase in IRSp53 WT cells normalised to IRSp53 \u2212 / \u2212 cells , along with representative images ( right ) . Data were pooled from two independent experiments and the number of cells indicated in the \ufb01 gure . c Histogram ( top ) shows normalised 5 - min \ufb02 uid uptake in IRSp53 \u2212 / \u2212 and IRSp53WT cells when treated with LG186 or vehicle ( DMSO ) along with representative images ( bottom ) . Data were pooled from two independent experiments and the number of cells indicated in the \ufb01 gure . d Histogram ( top ) shows an average number of endocytic structures quanti \ufb01 ed per \ufb01 eld from the electron microscope images ( bottom ) . Data pooled from three independent blocks . Untreated WT MEFs ( WT , top row ) , IRSp53 null MEFs ( IRSp53 \u2212 / \u2212 , bottom left ) or LG186 - treated WT MEFs ( LG186 , bottom right ) were incubated for 5 min at 37 \u00b0C with 10 mg / ml HRP as a \ufb02 uid - phase marker before processing for electron microscopy . Endocytic structures close to the plasma membrane ( PM ) are \ufb01 lled with the electron dense peroxidase precipitate . WT cells show a range of endocytic structures including vesicular structures ( double arrowheads ) and tubular / ring - shaped putative CLIC / GEECs ( large arrowheads ) but the IRSp53 \u2212 / \u2212 cells and LG186 - treated cells only show predominant labelling of vesicular pro \ufb01 les . p value < 0 . 05 ( * ) , 0 . 001 ( * * ) Mann \u2013 Whitney U test ( b \u2013 c ) and two - sample Student \u2019 s t test ( d ) . Error bars , s . d . ( b \u2013 d ) . Scale bar , 20 \u00b5 m ( b \u2013 c ) , 1 \u00b5 m ( d ) , respectively . Schematic ( a ) was adapted with permission from MBInfo ( www . mechanobio . info ) Mechanobiology Institute , National University of Singapore ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w 6 NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications LG186 - mediated GBF1 inhibition ( Fig . 3c ) . By contrast , GBF1 inhibition in IRSp53WT MEFs decreased \ufb02 uid - phase uptake to the levels observed in IRSp53 \u2212 / \u2212 MEFs ( Fig . 3c ) while endocytosed TfR remained unaffected ( Supplementary Fig . 4a ) . We con \ufb01 rmed the lack of the CG endocytic pathway in IRSp53 \u2212 / \u2212 MEFs by ultrastructural analysis of endocytic structures marked by the \ufb02 uid - phase marker , HRP using EM 14 , 23 . We observed a signi \ufb01 cant reduction of CLICs in IRSp53 \u2212 / \u2212 MEFs when compared with WT MEFs , while the number of clathrin and caveolae - derived structures was relatively unaffected . Similar to Supplementary Fig . 1e , f , the CLICs were reduced signi \ufb01 cantly upon LG186 treatment in WT MEFs as well ( Fig . 3d and Supplementary Fig . 4b ) . This led us to hypothesise that CG cargo would traf \ufb01 c via CME in the absence of IRSp53 . Therefore , we looked at the traf \ufb01 cking of GPI - AP ( GFP - GPI ) , \ufb02 uid phase and TfR ( CME ) in the absence of IRSp53 at high resolution . We \ufb01 rst counted the number of GFP - GPI and \ufb02 uid endosomes , and found them to be signi \ufb01 cantly lower in IRSp53 \u2212 / \u2212 MEFs relative to IRSp53WT MEFs ( Fig . 4a ) . Conversely , TfR endosomal number was unaffected ( Fig . 4a ) . We next , looked at co - localisation of GPI - AP / \ufb02 uid phase with TfR . A relatively higher fraction of GFP - GPI I - BAR CRIB / GBD SH3 + \u2013 / \u2013 + \u2013 / \u2013 + \u2013 / \u2013 + \u2013 / \u2013 + \u2013 / \u2013 a b c d 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 F r a c t i on o f F l d w i t h T f IRSp53 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 IRSp53 Cargo F r a c t i on o f G F P - G P I w i t h F l d / T f 300 600 900 1200 1500 # o f E ndo s o m e s / c e ll IRSp53 Cargo n n n GPI / Tf I R S p53 \u2013 / \u2013 I R S p53 W T GPI / Fluid Fluid Tf Merge I R S p53 W T I R S p53 \u2013 / \u2013 Fluid Transferrin N o r m a li s ed endo cy t o s i s ( a . u . ) 4KE I408P V522G I268N WT Fluid GFP - GPI Transferrin \u03b1 - GFP Tf I R S p53 \u2013 / \u2013 I R S p53 W T Fluid 0 . 0 0 . 2 0 . 8 1 . 0 1 . 2 1 . 4 1 . 6 * * * * * * * * * n . s . * * * WT I268N V522G 4KE I408P PDZ - BD I268N 4KE I408P V522G * * * * * n . s . 0 25 25 58 56 32 31 25 25 32 31 + \u2013 / \u2013 46 36 22 1 21 2 28 4 33 6 29 3 NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w ARTICLE NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications 7 ( Fig . 4b ) and \ufb02 uid phase ( Fig . 4c ) co - localised with co - pulsed Tf in IRSp53 \u2212 / \u2212 MEFs than IRSp53WT MEFs . Thus , removal of IRSp53 speci \ufb01 cally affected \ufb02 uid phase and GPI - AP endocytosis , while CME remained unaffected . Moreover , in cells lacking IRSp53 , GPI - AP and the residual \ufb02 uid phase is endocytosed via CME . We next analysed the contribution of different domains of IRSp53 on CG endocytosis by re - introducing into IRSp53 \u2212 / \u2212 MEFs , GFP - tagged IRSp53WT and a number of mutants of IRSp53 speci \ufb01 cally defective in various domains ( Fig . 4d , schematic ) . We found that GFP - IRSp53WT and GFP - IRSp53V522G rescued endocytosis while the rest of the mutants failed to do so ( Fig . 4d ) . In conclusion , these results indicated that IRSp53 is an essential and speci \ufb01 c regulator of CG endocytosis that requires functional I - BAR , CRIB and SH3 domains . IRSp53 is recruited to forming CG endocytic vesicles . The complete absence of CG endocytosis in IRSp53 \u2212 / \u2212 led us to hypothesise that IRSp53 has a direct role to play in CG vesicle formation . Hence , we used pH pulsing assay and examined the recruitment of mCherry - IRSp53 to the forming SecGFP - GPI endocytic sites . A majority of ( > 60 % ) endocytic events exhibited prominent recruitment of IRSp53 ( Fig . 5a , Supplementary Fig . 6c and Table 1 ) . Since the I - BAR domain of IRSp53 has been shown to sense / induce negative curvature in a membrane tension and protein concentration - dependent manner 48 , we looked at changes in its spatial distribution during the formation of endocytic vesicle using two types of masks \u2014 a spot and a ring mask ( Fig . 5a , schematic ) . Unlike the intensity pro \ufb01 les of CDC42 that did not exhibit any differential temporal patterns of recruitment between the two types of masks ( Fig . 5b , black vs . red trace ) , IRSp53 displayed a biphasic recruitment pattern ( Fig . 5a ) . In phase I ( \u2212 36 to \u2212 15 s ) , IRSp53 was \ufb01 rst recruited over a large area indicated by an increase in intensity in both spot and ring traces . Subsequently , in phase II ( \u2212 15 to 0 s ) , IRSp53 was con \ufb01 ned at the centre indicated by a concerted decrease in the ring mask intensity and increase in the spot mask intensity trace ( Fig . 5a , black vs . red trace ) . The increase of IRSp53 was more prominent in phase II toward the centre ( Fig . 5a , black vs . purple trace ) and correlated with the recruitment of CDC42 ( r = 0 . 6 ; Table 2 ) . To visualise the recruitment of GFP - IRSp53 at high spatial resolution , we co - expressed a GBP - APEX reagent ( GFP - binding protein soybean ascorbate peroxidase ) and processed for EM as described previously 49 ( Fig . 5c , Supplementary Fig . 4c and Supplementary Movies 3 \u2013 4 ) . GBP - APEX binds to GFP and converts 3 , 3 \u2032 - diamino - benzamidine into an osmiophilic polymer in the presence of H 2 O 2 49 . Images of 3D rendering from the electron densities revealed that IRSp53 associated with tubular structures characteristic of CLICs as described previously 23 and as expected , with \ufb01 lopodial structures 38 ( Fig . 5c , see Methods ) . In the 2D sections , IRSp53 was observed to accumulate as discrete patches at the plasma membrane ( PM ) ( Supplementary Fig . 4c i \u2013 ii , arrowheads ) and was frequently associated with tubulovesi - cular invaginations or tubular structures close to PM and \ufb01 lopodial tips ( Supplementary Fig . 4c iii , double arrowheads ) . We further validated recruitment of GFP - IRSp53 with an alternate technique , protein - retention expansion microscopy ( ProExM ) 50 , a derivative of expansion microscopy 51 . Expansion microscopy enables imaging of diffraction - limited structures by physically separating them using a swellable polymer cross - linked with the cell . This allows multi - colour super - resolution imaging of a sample with conventional regents and microscope . Thus , structures of around 250 nm will be scaled to 1 \u03bc m in a 4\u00d7 expanded sample ( see Methods ) , providing a potential apparent resolution of around 70 nm 51 with conventional imaging technology . We stained IRSp53 \u2212 / \u2212 GFP - IRSp53 MEFs for CD44 and IRSp53 , processed the cells ( Methods and ref . 50 ) , and imaged the samples using \u00d7100 objective in spinning disk microscope [ Supplementary Fig . 4e ( 1a \u2013 b ) and Supplementary Movie 5 ) ] . In accordance with our EM images , we observed enrichment of IRSp53 at the tips of the \ufb01 lopodia [ Supplementary Fig . 4e ( 2a \u2013 c ) and Supplementary Movie 5 ] . Additionally , we could identify several invaginations of CD44 [ Supplementary Fig . 4e ( 3 \u2013 8 , see arrowheads ) and Supplementary Movie 5 ] and found enrichment of IRSp53 . As the neck constriction is expected to develop [ Supplementary Fig . 4e ( 3 \u2013 4 and 6 , arrowheads ) and Supplementary Movie 5 ] , we could see progressive enrichment of IRSp53 around the region of the constriction , relative to other regions of the invagination . The pH pulsing trace showed a speci \ufb01 c pattern of enrichment of IRSp53 over time . This is consistent with two scenarios : ( 1 ) that IRSp53 enrichment occurs speci \ufb01 cally at the neck which constricts over time and , ( 2 ) that IRSp53 coats the entire tubule with higher enrichment around the necks which in turn constricts over time . Our interpretation of EM and proExM data support the latter scenario . Thus , the complete loss of CG endocytosis in IRSp53 null cells , and localisation of IRSp53 to forming CG endocytic vesicles suggests a role of IRSp53 in the vesicle scission . Branched actin nucleation is required for CG endocytosis . A functional CG endocytic pathway requires dynamic actin since inhibition of actin polymerisation ( Latrunculin A ) , or \ufb01 lament stabilisation ( Jasplakinolide ) , inhibited CG endocytosis 13 . CDC42 and IRSp53 are core components of a signalling axis that indir - ectly controls the location and activity of the ARP2 / 3 actin nucleation complex 39 , 52 , 53 . More pertinent , CK666 54 - mediated inhibition of ARP2 / 3 complex , impaired both \ufb02 uid - phase and TfR uptake in a dose - dependent manner . However , the extent of Fig . 4 CG pathway is abolished in the absence of IRSp53 . a The box plot shows the number of endosomes per cells ( left ) for endocytosed GFP - GPI ( \u03b1 - GFP Fab ) , \ufb02 uid phase and TfR in IRSp53 \u2212 / \u2212 and IRSp53 WT cells when pulsed for 2 min along with representative images ( right ) . Data were pooled from two independent experiments and the number of cells indicated below the graph . b Plot ( left ) shows quanti \ufb01 cation of the fraction of GFP - GPI endocytic vesicles containing \ufb02 uid phase or Tf . Images ( right ) show representative single confocal slices of a 2 - min pulse of \u03b1 - GFP Fab ( green ) / TMR - Dextran ( magenta ) and \u03b1 - GFP Fab ( green ) / A568 - Tf ( magenta ) in IRSp53 \u2212 / \u2212 ( top row ) and IRSp53WT ( bottom row ) cells . The inset depicts magni \ufb01 ed views of the indicated region ; single - channel images are in panel 4a . Data were pooled from two independent experiments and the number of cells is indicated below the graph . c Plot ( left ) showing quanti \ufb01 cation of the fraction of 1 - min \ufb02 uid - phase endocytic vesicles containing Tf . Images show representative single confocal slices of a 1 - min pulse of TMR - Dextran ( green ) and A647 - Tf ( magenta ) in IRSp53 \u2212 / \u2212 ( top row ) and IRSp53WT ( bottom row ) cells . Inset depicts magni \ufb01 ed views of the indicated region . Data were pooled from two independent experiments and the number of cells indicated below the graph . d Histogram ( left ) shows 5 - min uptake of \ufb02 uid phase in IRSp53 \u2212 / \u2212 MEFs transduced with virus expressing GFP - IRSp53 WT , GFP - IRSp53 4KE , GFP - IRSp53 I268N , GFP - IRSp53 I408P and GFP - IRSp53 V522G , normalised to that in IRSp53 \u2212 / \u2212 MEFs , along with representative images ( right ) . Data were pooled from two independent experiments and the number of cells indicated in \ufb01 gure except for IRSp53 \u2212 / \u2212 ( 381 ) . p value < 0 . 01 ( * ) and 0 . 001 ( * * ) by Mann \u2013 Whitney U test ( a \u2013 d ) . Scale bar , 20 \u00b5 m ( d ) , 5 \u00b5 m ( a \u2013 c ) , respectively . Error bars ( d ) represent s . d . ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w 8 NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications / / \u2013 40 / / / a c b Time ( s ) pH 7 pH 5 CDC42 pH 5 pH 7 IRSp53 Time ( s ) F o l d c hange ( a . u . ) IRSp53 Random Log10 ( p ) RAW THRESHOLDED 1 1a 2 3 Time ( s ) Time ( s ) \u2013 30 \u2013 20 \u2013 10 0 10 20 1 . 0 1 . 1 1 . 2 1 . 3 1 . 4 1 . 5 1 . 6 \u2013 40 \u2013 30 \u2013 20 \u2013 10 0 10 20 0 . 95 1 . 00 1 . 05 1 . 10 1 . 15 1 . 20 1 . 25 1 . 30 CDC42Random \u2013 6 \u2013 4 \u2013 2 0 18 15 12 9 6 3 0 \u2013 3 \u2013 6 \u2013 9 \u2013 12 \u2013 15 \u2013 18 18 15 12 9 6 3 0 \u2013 3 \u2013 6 \u2013 9 \u2013 12 \u2013 15 \u2013 18 Fig . 5 IRSp53 is recruited to forming CG endosomes . a Graphs show the average normalised \ufb02 uorescence intensity vs . time traces for the recruitment of three different regions [ circles , violet , r = 170 nm and black ( r = 250 nm ) and annulus , orange ( r = 250 \u2013 420 nm ) ] for the recruitment of IRSp53 - mCherry to the forming SecGFP - GPI endocytic sites and its corresponding random intensity trace ( n , Table 1 ) . b Graphs show the average normalised \ufb02 uorescence intensity vs . time traces for the recruitment of TagRFPt - CDC42 to the forming SecGFP - GPI endocytic sites and its corresponding random intensity trace to two different regions [ circle , black , r = 250 nm ; and annulus , orange ( r = 250 \u2013 420 nm ) ] . Error bars , ( a \u2013 b ) represent s . e . m . ( n , Table 1 ) . The random traces were derived from randomly assigned spots of the same radius as the endocytic regions , as detailed in S . I . Endocytic distribution at each time point was compared to the random distribution ( a ) by Mann \u2013 Whitney U test and the log 10 ( p ) is plotted below each trace [ log 10 ( 0 . 05 ) is \u2212 1 . 3 and log 10 ( 0 . 001 ) is \u2212 2 . 5 ] . Representative montages are depicted below the graphs ( a \u2013 b ) . Arrowheads indicate the newly formed endocytic vesicle . c Electron micrographs of AGS cells co - transfected - GFP - IRSp53 and GFP - binding protein coupled to Apex ( GBP - Apex ) . The DAB reaction was performed and the cells were processed for electron tomography . A single section of the original tomogram ( left ) and density - based thresholded of the same plane ( middle ) reveal electron dense structures containing IRSp53 at membrane surfaces . The whole of PM of the tomographic volume was rendered and different examples of enlarged tubular regions of interest show GFP - IRSp53 recruitment patterns ( right ) . Scale bar , 1 . 5 \u00b5 m ( a \u2013 b ) and 0 . 5 \u00b5 m ( c ) , respectively NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w ARTICLE NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications 9 inhibition of CG internalisation was markedly more prominent ( Fig . 6a ) . By contrast SMIFH2 55 - mediated inhibition of formins , failed to inhibit CG endocytosis ( Supplementary Fig . 5a ) . We next explored the spatio - temporal dynamics of F - actin and ARP2 / 3 complex using pRuby - lifeact 56 and mCherry - ARP3 , respectively , during the formation of the endocytic vesicle . ARP3 recruitment ( Fig . 6b , Supplementary Fig . 6d and Table 1 ) began earlier than \u2212 35 s and peaked at \u2212 6 s . This was unexpected since CDC42 , a key regulator of ARP2 / 3 57 was not recruited until \u2212 9 s . Instead , the ARP3 pro \ufb01 le was correlated with IRSp53 between \u2212 9 s and + 9 s ( r = 0 . 8 ; Table 2 ) . This indicated that , at least initially , the ARP2 / 3 complex might be recruited in a CDC42 - independent manner . On the other hand , F - actin accumulation began around \u2212 9 s and continued even after the scission event ( Fig . 6c , Supplementary Fig . 6e and Table 1 ) , highly correlated to the CDC42 pro \ufb01 le ( r = 0 . 6 ; Table 2 ) . Thus , F - actin pH 5 pH 7 Arp3 a b N o r m a li s ed endo cy t o s i s ( a . u . ) FluidTf / TfR CK ( \u03bc M ) 1 . 2 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 * * * * F l u i d T f T f R DMSO 50 \u03bc M CK 150 \u03bc M CK 300 \u03bc M CK c Time ( s ) \u2013 40 \u2013 30 \u2013 20 \u2013 10 0 10 20 1 . 0 1 . 1 1 . 2 1 . 3 1 . 4 1 . 5 1 . 6 F o l d c hange ( a . u . ) Time ( s ) Arp3Random 1 . 6 1 . 4 1 . 2 1 . 0 0 . 8 0 . 6 0 . 4 0 . 2 0 . 0 N o r m a li s ed f l u i d u p t a k e ( a . u . ) N - WASP \u0394 VCA N - WASPVCA N - WASPCA * * * * n . s . d \u2013 40 \u2013 30 \u2013 20 \u2013 10 0 10 20 1 . 0 1 . 1 1 . 2 1 . 3 1 . 4 \u2013 4 \u2013 3 \u2013 2 \u2013 10 F o l d c hange ( a . u . ) Time ( s ) N - WASP Coloc ( 27 % ) RandomN - WASP All Log 1 0 ( p ) Time ( s ) pH 5 pH 7 N - WASP e Time ( s ) pH 5 pH 7 Lifeact F o l d c hange ( a . u . ) \u2013 9 \u2013 6 \u2013 3 0 Log10 ( p ) 385 308 306 n 442 319 379 n 380 \u2013 40 \u2013 30 \u2013 20 \u2013 10 0 10 20 1 . 0 1 . 1 1 . 2 1 . 3 \u2013 3 \u2013 2 \u2013 10 Time ( s ) LifeactRandom Log10 ( p ) 300 150 50 0 18 15 12 9 6 3 0 \u2013 3 \u2013 6 \u2013 9 \u2013 12 \u2013 15 \u2013 18 18 15 12 9 6 3 0 \u2013 3 \u2013 6 \u2013 9 \u2013 12 \u2013 15 \u2013 18 18 15 12 9 6 3 0 \u2013 3 \u2013 6 \u2013 9 \u2013 12 \u2013 15 \u2013 18 ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w 10 NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications was generated coincident with the recruitment of CDC42 , a known regulator of actin polymerisation 57 . These observations suggested that ARP2 / 3 complex might be \ufb01 rst recruited in an inactive state , and then activated following the arrival of CDC42 . ARP2 / 3 is inhibited by PICK1 at nascent CG endocytic sites . To address how ARP2 / 3 complex was perhaps maintained at the forming endocytic pit in an inactive state , we analysed the role of PICK1 , another \u2018 hit \u2019 in the screen ( Fig . 2 ) . PICK1 , a highly con - served protein , possesses PDZ and BAR domain ( Fig . 7a ) that by intra - molecular interaction , maintains PICK1 in an auto - inhibited state . This auto - inhibited state is further stabilised upon GTP \u2013 ARF1 binding to the PDZ domain 58 . Additionally , activated PICK1 negatively regulates ARP2 / 3 - mediated actin polymerisa - tion 58 \u2013 60 . The ability of PICK1 to inhibit ARP2 / 3 complex is suppressed by GTP \u2013 ARF1 58 . To con \ufb01 rm a role of PICK1 in CG endocytosis in mammalian cells , we utilised a speci \ufb01 c small - molecule inhibitor of PICK1 , FSC231 61 . CG endocytosis ( \ufb02 uid phase ) was inhibited in a dose - dependent fashion ( Fig . 7b ) by this inhibitor . Additionally , in stable PICK1 knockdown lines , \ufb02 uid phase and folate receptor ( FR - GPI , another GPI - AP ) endosomal number were lower than that measured in scrambled shRNA stable lines , while TfR endocytosis remained unaffected ( Fig . 7c , d ) . Predictably , GFP - ARF1 and TagRFP - PICK1 co - localised in punctate spots at the TIRF plane in accordance with previous reports 58 ( Supplementary Fig . 5b ) . To test the effect of ARF1 activity on PICK1 recruitment we co - expressed TagRFP - PICK1 with either ARF1 - WT or dominant - negative ( ARF1 - DN ; T31N ) and active mutant ( ARF1 - DA ; Q71L ) and tracked the number and dynamics of PICK1 spots using TIRF microscopy ( see Methods ) . The residence time of PICK1 increased signi \ufb01 cantly in the presence of ARF1 - DN while it was reduced in the presence of ARF1 - DA ( Fig . 7e ) . ARF1 - DN and DA have been shown to decrease and increase endocytosis , respectively 14 . Thus , the local presence of ARF1 - GTP resulted in the removal of PICK1 from the membrane . Thus , we hypothesised that PICK1 was recruited to forming CG endosome at the early stage , rendering ARP2 / 3 inactive . To explore this possibility , we utilised the pH pulsing assay to visualise PICK1 at the CG endocytic site . We found that TagRFP - PICK1 was recruited to the forming CG endocytic sites ( Fig . 7f , Supplementary Fig . 6f and Table 1 ) in a pulsatile fashion . Maximum enrichment occurred at \u2212 12 s , with an eventual loss corresponding to the time of the rapid rise in ARF1 recruitment around \u2212 9 s ( Fig . 1e , r = 0 . 6 , Table 2 ) . Thus , the pH pulsing assay has led to the realisation that in CG endocytosis , the interplay of two BDPs , PICK1 and IRSp53 , regulate ARP2 / 3 complex . These BDPs interact with the ARP2 / 3 complex ( Supplementary Fig . 5b , upper panel and Supplementary Fig . 4d ) regulate its activity at the forming CG endocytic sites , in opposing fashion under the in \ufb02 uence of ARF1 and CDC42 . Discussion CG endocytosis was initially discovered as a route for the entry of toxins 62 , \ufb02 uid phase and GPI - anchored proteins when CME was perturbed 11 , 12 , 63 , raising some concerns regarding its physiological role in unperturbed cells 18 . Using a pH pulsing assay we show , here , that a majority of SecGFP - GPI - containing endocytic vesicles form due to a stereotypical and temporally orchestrated recruit - ment of the key molecular machinery namely , CDC42 , ARF1 and GBF1 , which , in turn , mediate the coordinated assembly of speci \ufb01 c BAR - containing , membrane deforming and actin regulatory pro - teins , IRSp53 and PICK1 . Notably , however , the vast majority of the endocytic vesicles are devoid of clathrin and dynamin . Thus , a dedicated complex protein machinery drives CG internalisation , similarly to that being observed in CME . The ability of the pH pulsing assay to provide a temporal pro \ufb01 le for the recruitment dynamics of the molecular players has considerably extended our understanding of CG endocytic vesicle formation . The GBF1 / ARF1 pair is the earliest module to be assembled and judging by their recruitment pro \ufb01 les , it takes around 1 min to assemble the molecular machinery for CG endocytosis . How this pair is concentrated at a forming CG endocytic site is an open question . The CG machinery includes CDC42 , ARP2 / 3 and F - actin along with BDPs , PICK1 and IRSp53 , and in the model ( Fig . 8 ) we propose a biphasic mechanism correlated with the ARF1 recruitment kinetics . In the \ufb01 rst phase , the accumula - tion is slow , accompanied by the presence of PICK1 and absence of CDC42 . In the second phase , beginning around \u2212 9 s ( before scission ) , the accumulation of ARF1 speeds up concomitant with the arrival of CDC42 and loss of PICK1 . How the kinetics of ARF1 recruitment is regulated is unclear since its GAP ( GTPase - activating protein ) is presently unknown , as is the GEF for CDC42 . Nevertheless , the presence of PICK1 provides an expla - nation behind ARP2 / 3 recruitment in an inactive state to the forming CG endocytic vesicles long before CDC42 . ARP2 / 3 is then induced to promote actin branching only upon the arrival of IRSp53 and CDC42 . The role of ARP2 / 3 in CG endocytosis is reminiscent of the endocytic process occurring in the budding yeast . In this system , the endocytic machinery strictly depends on Las17 , the yeast homologue of N - WASP but not so much on clathrin and dynamin 64 , 65 . There are however important differences . In CG endocytosis , the ARP2 / 3 complex appears to be activated inde - pendent of its canonical NPF , N - WASP , a CDC42 effector 57 . First , not only did N - WASP fail to recruit to forming CG endocytic sites ( Fig . 6d and Supplementary Fig . 6g ) , over - expression its dominant - negative mutants also failed to inhibit CG endocytosis ( Fig . 6e ) . By contrast , in CME , both ARP2 / 3 complex and N - WASP are recruited to budding CME vesicles , and in \ufb02 uence endocytosis in some cell types 28 . The unexpected recruitment pro \ufb01 le of ARP2 / 3 and the iden - ti \ufb01 cation of two BDPs , PICK1 and IRSp53 as upstream regulators Fig . 6 Arp2 / 3 - based actin machinery is required for CG endocytosis . a Histograms ( top ) show quanti \ufb01 cation of \ufb02 uid phase and TfR uptake in AGS cells treated with DMSO alone ( 0 \u00b5 M ) or the indicated concentrations of ARP2 / 3 inhibitor , CK666 , normalised to DMSO - treated controls , along with its representative images ( below ) . Data are pooled from two independent experiments and the number of cells shown indicated the graph . b \u2013 d Graphs show the average normalised \ufb02 uorescence intensity vs . time traces for the recruitment of mCherry - ARP3 ( b ) , pRuby - Lifeact ( c ) and mCherry - NWASP ( d ) to the forming SecGFP - GPI endocytic sites , and its corresponding random intensity trace ( n , Table 1 ) . The random traces were derived from randomly assigned spots of the same radius as the endocytic regions , as detailed in S . I . Endocytic distribution at each time point was compared to the random distribution by Mann \u2013 Whitney U test and the log 10 ( p ) [ log 10 ( 0 . 05 ) is \u2212 1 . 3 and log 10 ( 0 . 001 ) is \u2212 2 . 5 ] is plotted below each trace ( b \u2013 d ) . Representative montages are depicted below the graphs . Arrowheads indicate the newly formed endocytic vesicle . e Histogram ( left ) shows normalised 5 - min mean \ufb02 uid - phase uptake in AGS cells overexpressing pIRES - CA domain , GFP - VCA domain and GFP - N - WASP \u0394 VCA from N - WASP compared to un - transfected cells and representative images ( right ) . The transfected cells are outlined . Data were pooled from two independent experiments and the number of cells shown below the graph . Error bars represent s . e . m . ( b \u2013 d ) and s . d . ( a , e ) . p value < 0 . 01 ( * ) , and 0 . 001 ( * * ) by Mann \u2013 Whitney U test ( a , e ) . Scale bar , 1 . 5 \u00b5 m ( b \u2013 d ) , 20 \u00b5 m ( a , e ) NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w ARTICLE NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications 11 of ARP2 / 3 activity also suggests a reason for this biphasic mechanism . PICK1 operates in the early phases , and may func - tion as an inhibitor of ARP2 / 3 , consistent with the modest recruitment of F - actin in the presence of PICK1 observed here and as reported previously 59 , 60 . PICK1 recruitment occurs via its BAR domain since when mutated and overexpressed , it acts as a dominant negative for CG endocytosis ( Supplementary Fig . 5c ) . PICK1 recruitment is not only rapid , but also transient . We \ufb01 nd that the localisation of PICK1 to the membrane was negatively correlated to the activity of ARF1 , similar to that observed in neuronal cells 58 , wherein GTP \u2013 ARF1 interaction with PICK1 rendered PICK1 incapable of inhibiting ARP2 / 3 complex . This is followed by the second phase characterised by the simultaneous recruitment of CDC42 / IRSp53 effector complex enabling the activation of ARP2 / 3 and subsequent polymerisation of actin at the site of endocytosis . However , the NPF linking IRSp53 and ARP2 / 3 activation is not yet characterised and is the subject of investigation . There is no single unifying theme for vesicle scission in CIE and multiple modules may co - exist . Recently , endophilin A has been shown to facilitate tubule scission by a combination of scaffolding , dynamin recruitment and dynein - mediated elonga - tion of membrane tubules leading to an increase in friction 66 . In the CG pathway , IRSp53 emerges as a major player . This protein a b e d Time ( s ) pH 5 pH 7 PICK1 PDZ BAR Acidic Acidic 4 1220 110 152 362 380 \u2013 390 0 200 600 800 * * * Transferrin Fluid Cargo N u m be r o f endo s o m e s / c e ll PICK1 shRNA FR - GPI * * * n . s . P I G P Z P I C K 1 s h RN A Transferrin Fluid FR - GPI DMSO 50 \u03bc M 100 \u03bc M 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 1 . 4 * N o r m a li s ed endo cy t o s i s ( A . U . ) FluidTf / TfR * * * * * * D M S O 100 \u03bc M F S C 231 FLUID TF TFR P I G P Z P I C K 1 s h RN A PICK shRNA 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 N o r m a li s ed P I C K 1 l e v e l s ( a . u . ) * * \u2013 40 \u2013 30 \u2013 20 \u2013 10 0 10 20 1 . 0 1 . 2 1 . 4 1 . 6 F o l d c hange ( a . u . ) Time ( s ) PICK1 Random Log10 ( p ) f GFP ARF1 WT ARF1 DN HA ARF1 DA < R e s i den c e t i m e ( P I C K 1 ) , s > 186 82 68 98 n 63 64 77 32 17 20 13 * * * \u2013 4 \u2013 3 \u2013 2 \u2013 10 1 . 0 1 . 5 2 . 0 2 . 5 3 . 0 3 . 5 4 . 0 n 560 361 474 FSC231 n c \u2013 + \u2013 + \u2013 + 18 15 12 9 6 3 0 \u2013 3 \u2013 6 \u2013 9 \u2013 12 \u2013 15 \u2013 18 ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w 12 NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications may function by multiple mechanisms : it can couple negative curvature with membrane tension , it can scaffold membrane at moderate densities 48 , and it can regulate actin polymerisation as it does in \ufb01 lopodia formation 38 . A minimal model accounting for all these activities suggests that IRSp53 might be enriched at the vesicle neck , where it would regulate the actin machinery necessary to trigger CG vesicle scission . The spatio - temporal dynamics and ultrastructure analysis of IRSp53 recruitment at CG sites are consistent with such a model . However , the data does not permit an unequivocal picture ; better resolution and reagents are necessary to verify this speculation . Lastly , the complete and speci \ufb01 c loss of CG endocytosis ( but not CME ) in the absence of IRSp53 makes the requirement for IRSp53 necessary for CG endocytic process . In summary , we propose that CG endocytic vesicle formation begins with GBF1 / ARF1 concentrating at sites of endocytic pits . Following this , though ARP2 / 3 is recruited , it is held in an inactive state by PICK1 ( Fig . 8a ) . Meanwhile , IRSp53 is recruited ( potentially via its I - BAR domain ) and activated by CDC42 leading to ARP2 / 3 activation via unknown effector ( s ) ( Fig . 8b ) . The loss of IRSp53 and ARP2 / 3 from the membrane as the endocytic vesicle is pinched is consistent with their role in endosomal neck dynamics , providing a new candidate for mole - cular machinery of the pinching process in the absence of dynamin . It is conceivable that the IRSp53 provides a scaffold for a friction - based scission mechanism as recently suggested 66 , with actin polymerisation providing driving force for tubule elonga - tion . Alternatively , this force could arise from the involvement of a microtubule - based machinery as recently advocated in the internalisation of cholera toxin 67 . The assays developed here and the identi \ufb01 cation of a number of molecular players and their temporal recruitment pro \ufb01 le provides a path towards under - standing a molecular mechanism for the formation of a CG endocytic vesicle . Methods Cell culture , reagents , and plasmids . See Supplementary Information . pH pulsing assay . pH pulsing assay was adapted for Nikon TE 2000 TIRF microscope ( for the details of the microscope , see microscope section in S . I . ) from a similar assay as described previously 4 . Brie \ufb02 y , FR - AGS cells were plated on custom - designed coverslip bottom dishes and were transfected with SecGFP - GPI and X - FP constructs 12 \u2013 14 h before the assay . The dishes were then \ufb01 tted with a custom - designed holder to place inlet and outlet tubing . Tubing from a pH 7 . 4 ( HEPES ) and pH 5 . 5 ( MES ) buffers kept in a water bath at 38 \u00b0C went through a peristaltic \ufb02 ow controller ( Bioscience Tools ) into the cell chamber . The tempera - ture of the cells in the dish was maintained at 30\u00b0C by maintaining the buffers , the objective and the microscope ( using a chamber ) at appropriately high tempera - tures . Imaging at 30 \u00b0C slows the endocytic process to match the time resolution achieved by the \ufb02 ow setup ( 3 s ) . Images were captured using a script written in open source imaging software , Micromanager , to control the time of \ufb02 ow and imaging . Typically , buffers are exchanged every 3 s and three images are collected sequentially before the end of 3s in two channels , GFP and RFP , using camera exposure of 100ms . The chamber around the microscope was built with the help of NCBS Mechanical Workshop , AC department and Dr Manoj Matthew ( NCBS - CIFF ) . Endocytic assay . Mammalian endocytic assays : All population - based endocytic assays were performed as described 14 . For endocytic assays in mammalian cells , 2 - day plated cells on poly - D - Lysine - coated dishes were used . The assays were performed in the water bath maintained at 37\u00b0C . The media was removed and replaced with media containing \ufb02 uorescently labelled probes at appropriate dilu - tions for the required time ( TMR - Dextran was used at 1 mg / ml , \u03b1 - GFP was used at 20 \u00b5g / ml , Cy3 - Mov18 was used at 5\u00b5g / ml and Tf was used at 10\u00b5g / ml ) . The cells were then transferred to ice and washed with ice - cold medium 1 buffer ( 140mM NaCl , 20 mM HEPES , 1mM CaCl 2 , 1mM MgCl 2 , 5 mM KCl , pH 7 . 4 ) . The cells were then stripped for surface - bound Tf with ascorbate buffer ( 160mM sodium ascorbate , 40 mM ascorbic acid , 1mM MgCl 2 , 1mM CaCl 2 , pH 4 . 5 ) . In the case of GPI - AP ( GFP - GPI and FR - GPI ) uptake , the surface was removed using PI - PLC which cleaves GPI anchor 14 . Cells were treated with PI - PLC ( 50 \u00b5g / ml ) for 1h on ice . The cells were \ufb01 xed with 2 . 5 % paraformaldehyde and stained for surface TfR . BAR domain screen : RNAi screen for BDPs in Drosophila genome was done on S2R + cells stably expressing TfR 15 . Brie \ufb02 y , the cells were plated in a 12 - well plate ( 0 . 5 million cells / well ) for 1h . The media was then replaced with 600 \u00b5l of serum - free media supplemented with appropriate dsRNA at ( \ufb01 nal amount , 10 \u00b5g ) for 1h , post which 600 \u00b5l of serum containing media was added . After 4 days of depletion , the cells were assayed for endocytosis . On the 4th day , cells were deadhered from the well by manual pipetting and plated on coverslip bottom dishes . Cells were pulsed with TMR - dextran diluted in serum containing media for 5min . The cells were subsequently transferred to ice and washed with ice - cold medium 1 buffer ( supplemented with 1mg / ml BSA and glucose ) . The cells were then \ufb01 xed using 2 % paraformaldehyde ( 5 min on ice and 15 min at room temperature ) . dsRNA was prepared from the Drosophila Open Biosystems library v1 15 . HRP uptake and electron microscopy : WT and IRSp53 \u2212 / \u2212 MEFs were serum starved for 45min in the presence or absence of 10 \u03bc M LG186 compound . Cells were treated for 2 or 5 min at 37 \u00b0C with 10mg / ml HRP in serum - free medium , rapidly washed with complete medium on ice and subsequently \ufb01 xed with 2 . 5 % glutaraldehyde in PBS . Peroxidase development and further processing were performed as described previously 23 . Cells were sectioned parallel to the substratum and viewed unstained on a Jeol 1011 ( Tokyo , Japan ) transmission electron microscope equipped with a Morada Soft Imaging camera ( Olympus ) at two - fold binning . Quantitation was performed as follows : for 5 - min HRP uptake , images were captured at random across the monolayer by moving a de \ufb01 ned distance across the grid to avoid user bias . For 2 - min uptake , whole cells positive for HRP were imaged and montaged to generate a high - resolution image encompassing the entire cell . Five cells were imaged and montaged for each replicate . HRP - labelled elements per image or per cell pro \ufb01 le were classi \ufb01 ed using the following criteria ; vesicular elements including clathrin - coated vesicles and caveolae \u2014 circular pro \ufb01 les < 200nm in diameter ; early endosomes \u2013 circular and ring - shaped pro \ufb01 les , including multivesicular structures > 200nm in diameter ; CLIC / GEEC \u2013 other pro \ufb01 les including tubules and small ring - shaped structures < 200 nm in diameter . Ultrastructural localisation of IRSp53 . Thin sections : Localisation of GFP - tagged IRSp53 was performed as described previously 49 . Cells ( FR - AGS ) were seeded onto 30 mm tissue culture dishes ( TRP ) and transfected 24 h later using Lipofectamine Fig . 7 PICK1 is involved in CG endocytosis and is negatively regulated by ARF1 . a Schematic depicts domain organisation of PICK1 . b Histograms ( top ) show quanti \ufb01 cation of \ufb02 uid - phase and TfR uptake in AGS cells treated with DMSO alone ( 0 \u00b5 M ) or the indicated concentrations of PICK1 inhibitor , FSC231 , normalised to DMSO - treated controls , along with its representative images ( below ) . Data were pooled from two independent experiments with the cell numbers shown below the graph . c Box plot ( top ) shows the number of endosomes per cells for FR - GPI ( Cy3 - Mov18 ) , \ufb02 uid phase and TfR in scrambled ( PIGPZ ) and PICK1 shRNA - infected AGS cells when pulsed for 2 min along with representative images ( bottom ) . Data are pooled from two independent experiments and the number of cells indicated below the graph . d Histogram ( left ) shows normalised PICK1 levels measured by immunostaining in PICK1 shRNA - infected AGS cells along with representative images ( right ) . Data were pooled from two independent experiments with the cell numbers indicated in the \ufb01 gure except for PIGPZ ( 292 ) . e Box plot ( top ) shows the residence time of TagRFPt - PICK1 spots at the TIRF plane ( see Methods ) , averaged in an individual cell expressing either GFP , GFP - ARF1 WT , GFP - ARF1 DN or HA - ARF1 DA . The data are pooled from two independent experiments with cell number indicated below the graph . f The graph shows the average normalised \ufb02 uorescence intensity vs . time trace for the recruitment of TagRFPt - PICK1 to the forming SecGFP - GPI endocytic sites and its corresponding random intensity trace ( n , Table 1 ) . The random traces were derived from randomly assigned spots of the same radius as the endocytic regions , as detailed in S . I . Endocytic distribution at each time point was compared to the random distribution by Mann \u2013 Whitney U test and the log 10 ( p ) [ log 10 ( 0 . 05 ) is \u2212 1 . 3 and log 10 ( 0 . 001 ) is \u2212 2 . 5 ] is plotted below . Representative montage is depicted below . Arrowheads indicate the newly formed endocytic vesicle . Error bars represent s . e . m . ( f ) and s . d . ( b , d ) . p value < 0 . 01 ( * ) , 0 . 001 ( * * ) and 0 . 0001 ( * * * ) by Mann \u2013 Whitney U test ( b \u2013 e ) . Scale bar , 1 . 5 \u00b5 m ( f ) , 20 \u00b5 m ( b , d ) and 5 \u00b5 m ( c ) NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w ARTICLE NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications 13 3000 ( as per the manufacturer \u2019 s instructions ) . Cells were subsequently processed for electron microscopy 24 h after transfection . Brie \ufb02 y , AGS cells were double transfected with GFP - tagged IRSp53 and an expression vector encoding for a GFP - binding peptide conjugated to APEX2 ( Addgene plasmid # 67651 ) . Cells were washed in PBS , \ufb01 xed in 2 . 5 % glutaraldehyde in 0 . 1M sodium cacodylate buffer and subjected to the 3 , 3 \u2032 - diaminobenzidine ( DAB ; Sigma - Aldrich ) reaction in the presence of H 2 O 2 for 30 min . DAB reaction product was contrasted by 1 % osmium tetroxide for 2min . Cells were processed in situ and embedded in LX112 resin before sectioning parallel to the culture dish . Tomogram : Thick plastic sections ( 200nm ) were cut on an ultramicrotome ( UC6 , Leica ) and collected onto formvar - coated copper slot grids and lightly carbon coated . Dual - axis tilt series were acquired on a 120 kV TECNAI 12 ( FEI ) transmission electron microscope \ufb01 tted with a LC - 1100 4k x 4k lens coupled CCD camera ( Direct Electron ) and a tilt rotate holder ( Fischione ) utilising a tilt range of \u2212 60 to + 60\u00b0 . Microscope control and image acquisition was accomplished using the software SerialEM 68 . Tilt series were later reconstructed using weighted back projection and \ufb01 ducial markers ( 10 nm ) into a single volume with IMOD 69 . To examine areas with the greatest electron density in an unbiased manner , density - based thresholding was employed with the Isosurface render programme in IMOD as previously described 49 for APEX / DAB reaction product . The whole PM of the tomographic volume was rendered and tubular regions of interest were highlighted at greater magni \ufb01 cation . Protein - retention expansion microscopy . The ProExM protocol was adapted from previous reports 50 , 51 . IRSp53 \u2212 / \u2212 GFP - IRSp53 addback cells were grown on coverslips for 2 days . The cells were \ufb01 xed using 4 % PFA for 15min at room temperature ( RT ) . The surface CD44 was stained using \u03b1 - CD44 ( Rat ) following which the cells were permeablised using 0 . 05 % Tween20 for 15min . IRSp53 was dual stained using \u03b1 - GFP and IRSp53 antibodies both of which were generated in Rabbit . Following this , secondary antibodies against Rat ( Alexa - 568 ) and Rabbit ( Alexa - 488 ) were used . The cells then were treated with Acryloyl X - SE ( 10 mg / ml stock solution in DMSO , used 1 : 100 diluted in PBS ) for 12h at RT . The cells were washed with PBS 2\u00d7 15 min each before proceeding to gelation . For a 10 ml of monomer solution ( sod . acrylate ( \ufb01 nal concentration 8 . 6g / 100 ml ) , acrylamide ( \ufb01 nal concentration 2 . 5g / 100 ml ) , N , N \u2032 - methylenebisacrylamide ( \ufb01 nal concentra - tion 0 . 15 g / 100ml ) , NaCl ( \ufb01 nal concentration 11 . 7 g / 100ml ) were diluted in 1\u00d7 PBS . Monomer solution ( 48 \u00b5l ) was mixed with water ( 1 \u00b5l ) , TEMED ( 1 \u00b5l ) and APS ( 10 % , 1\u00b5l ) was added on to the cells . The cells were incubated at 37 \u00b0C for 30 min . The gel was incubated in the digestion buffer ( 50 mM Tris pH 8 . 0 , 1 mM EDTA , 0 . 5 % Triton X - 100 , 0 . 8M guanidine HCl , Proteinase K ( 1 : 100 , \ufb01 nal con - centration 8 units / ml , added before use ) for 6h at 37 \u00b0C . The gel was washed with double distilled water for 3 \u2013 5 times for 15min to achieve full 4\u00d7 expansion . The gel was placed on a coverslip and was imaged using \u00d7100 spinning disk microscope . Image analysis . In all cases , images were analysed with ImageJ and / or custom software written in MATLAB ( The Mathworks , Natick , Massachusetts , USA ) . The number of cells and repeats of the experiments are mentioned in the legends and \ufb01 gures . Statistical signi \ufb01 cance ( p ) was calculated by Mann \u2013 Whitney U test and two - sample Student \u2019 s t test , as reported in the legends . pH pulsing assay analysis : A semi - automated analysis was developed in MATLAB to identify newly formed endocytic vesicles in the pH pulsing assay and trace their intensity over time in pH 7 , pH 5 and RFP channels . The traces are an GTP F - Actin Arp2 / 3 complex ( Inactive ) Arp2 / 3 complex ( Active ) CDC42 Transferrin receptor Other GPI - APs Super ecliptic GFP ARF1 IRSp53 ( Active ) PICK1 ( Inactive ) IRSp53 ( Inactive ) PICK1 ( Active ) GTP GTP GTP GTP GTP GTP GTP GTP GTP GTP GTP GTP Extracellular milieu Cytoplasm GTP GTP GTP GTP GTP GTP GTP GTP a b c GTP KEY Fig . 8 Schematic depicting the proposed biphasic mechanism for CG endocytic vesicle formation . a Phase \u0399 : Characterised by the recruitment of ARF1 / GBF1 , PICK1 , ARP2 / 3 and IRSp53 but not the buildup of F - actin and CDC42 . Here , IRSp53 may be recruited by its I - BAR domain in the absence of GTP - CDC42 , keeping its SH3 domain in an intra - molecular inhibited state . PICK1 keeps ARP2 / 3 in an inactive state . b Phase \u0399\u0399 : Characterised by the recruitment of CDC42 and a sharp increase in ARF1 leading to the removal of PICK1 . This allows for the activation of ARP2 / 3 and buildup of F - actin . CDC42 binds to the CRIB domain of IRSp53 thereby activating it . The SH3 domain of IRSp53 can now bind to ARP2 / 3 activators and create F - actin . c Phase III : Characterised by endocytic vesicle formation , the presence of CDC42 , ARF1 / GBF1 and F - actin ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w 14 NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications average of many individual traces of all the endosomes pooled from different cells , which are compared with randomly placed spots within the cells . See S . I . for further details . Endocytic assay analysis : In all cases , images were analysed with ImageJ and custom software written in MATLAB ( The Mathworks , Natick , Massachusetts , USA ) . Each endocytic assay was performed with two technical duplicates . The number of repeats for each experiment is mentioned in its \ufb01 gure legend . For a given experiment , weighted mean for the technical duplicates was calculated as mentioned previously 14 . The total number of cells taken for analysis is mentioned in the legends and \ufb01 gures ( at least 40 \u2013 50 cells were taken from each technical duplicates ) . Subsequently , the cell - wise endocytosis distribution was normalised by weighted means of the control . This allowed data to be pooled from different days to be depicted as average and standard deviation as error bars . The statistical signi \ufb01 cance was calculated by Mann \u2013 Whitney U test . * represent p value < 0 . 05 . Co - localisation analysis : The analysis was performed by two methods . JaCoP ( Just another co - loc plugin ) . An ImageJ plugin which has multiple options by which co - localisation between two molecules can be measured 70 . The object - based overlap or Van Steensel cross - correlation function options were used . Brie \ufb02 y , images were thresholded manually and were given appropriate parameters for either of the two options . Spotter33 . It is a custom MATLAB script ( in - house ) as described previously 11 . The algorithm consists of applying a top - hat \ufb01 lter on the images , following which the structures are segmented . The threshold is manually determined for each channel . The segmented structures are then trimmed ( user determined ) until their are shape and size matched to the original structures . In a channel , each segmented particle mask \u2019 s centroid and pixel list is recorded , and the presence of a particle in that location is checked in the other channel . The number of pixels overlapped is normalied to the particle area . This value is averaged across all the particles for a given cell and reported . Residence time analysis : FR - AGS cells expressing the desired molecule tagged with a \ufb02 uorescent protein was imaged for the appropriate time 37 \u00b0C in TIRF . TagRFPt - PICK1 was imaged for 100 frames with 200 ms exposure and 500 ms interval and Nyquist criteria was satis \ufb01 ed . Spots were segmented and tracked using \u00b5 - track 71 . The residence time of TagRFPt - PICK1 spots per cell was calculated from the output of \u00b5 - track by custom MATLAB script . Brie \ufb02 y , the frame in which the spot was detected and the last frame a given spot of tracked is recorded . Spots that appeared in the \ufb01 rst and the last frame of the movie are discarded . Additionally , spots which appeared for only 1 frame are also discarded . The spots whose track end was the last frame was discared as well as the track may or may not have continued . Data availability . The authors declare that all data supporting the \ufb01 ndings of this study are available within the paper and its Supplementary Information Files or from the authors on reasonable request . Received : 23 August 2017 Accepted : 22 March 2018 References 1 . Johannes , L . , Parton , R . G . , Bassereau , P . & Mayor , S . Building endocytic pits without clathrin . Nat . Rev . Mol . Cell Biol . 42 , 1 \u2013 11 ( 2015 ) . 2 . Mayor , S . , Parton , R . G . & Donaldson , J . G . Clathrin - independent pathways of endocytosis . Cold Spring Harb . Perspect . Biol . 6 , a016758 ( 2014 ) . 3 . Kirchhausen , T . , Owen , D . & Harrison , S . C . Molecular structure , function , and dynamics of clathrin - mediated membrane traf \ufb01 c . Cold Spring Harb . Perspect . Biol . 6 , a016725 ( 2014 ) . 4 . Taylor , M . J . , Perrais , D . & Merri \ufb01 eld , C . J . A high precision survey of the molecular dynamics of mammalian clathrin - mediated endocytosis . PLoS Biol . 9 , e1000604 ( 2011 ) . 5 . Kaksonen , M . , Toret , C . P . & Drubin , D . G . Harnessing actin dynamics for clathrin - mediated endocytosis . Nat . Rev . Mol . Cell Biol . 7 , 404 \u2013 414 ( 2006 ) . 6 . Cocucci , E . , Gaudin , R . & Kirchhausen , T . Dynamin recruitment and membrane scission at the neck of a clathrin - coated pit . Mol . Biol . Cell 25 , 3595 \u2013 3609 ( 2014 ) . 7 . Shnyrova , A . V . et al . Geometric catalysis of membrane \ufb01 ssion driven by \ufb02 exible dynamin rings . Science 339 , 1433 \u2013 1436 ( 2013 ) . 8 . Parton , R . G . & Simons , K . The multiple faces of caveolae . Nat . Rev . Mol . Cell Biol . 8 , 185 \u2013 194 ( 2007 ) . 9 . Boucrot , E . et al . Endophilin marks and controls a clathrin - independent endocytic pathway . Nature 517 , 460 \u2013 465 ( 2015 ) . 10 . Renard , H . - F . et al . Endophilin - A2 functions in membrane scission in clathrin - independent endocytosis . Nature 517 , 493 \u2013 496 ( 2015 ) . 11 . Sabharanjak , S . , Sharma , P . , Parton , R . G . & Mayor , S . GPI - anchored proteins are delivered to recycling endosomes via a distinct cdc42 - regulated clathrin - independent pinocytic pathway . Dev . Cell 2 , 411 \u2013 423 ( 2002 ) . 12 . Guha , A . , Sriram , V . , Krishnan , K . S . & Mayor , S . Shibire mutations reveal distinct dynamin - independent and - dependent endocytic pathways in primary cultures of Drosophila hemocytes . J . Cell Sci . 116 , 3373 \u2013 3386 ( 2003 ) . 13 . Chadda , R . et al . Cholesterol - sensitive Cdc42 activation regulates actin polymerization for endocytosis via the GEEC pathway . Traf \ufb01 c 8 , 702 \u2013 717 ( 2007 ) . 14 . Kumari , S . & Mayor , S . ARF1 is directly involved in dynamin - independent endocytosis . Nat . Cell Biol . 10 , 3 0 \u2013 41 ( 2008 ) . 15 . Gupta , G . D . et al . Population distribution analyses reveal a hierarchy of molecular players underlying parallel endocytic pathways . PLoS ONE 9 , e100554 ( 2014 ) . 16 . Gupta , G . D . et al . Analysis of endocytic pathways in Drosophila cells reveals a conserved role for GBF1 in internalization via GEECs . PLoS ONE 4 , e6768 ( 2009 ) . 17 . Howes , M . T . , Mayor , S . & Parton , R . G . Molecules , mechanisms , and cellular roles of clathrin - independent endocytosis . Curr . Opin . Cell Biol . 22 , 519 \u2013 527 ( 2010 ) . 18 . Bitsikas , V . et al . Clathrin - independent pathways do not contribute signi \ufb01 cantly to endocytic \ufb02 ux . eLife 3 , e03970 ( 2014 ) . 19 . R\u00f6mer , W . et al . Actin dynamics drive membrane reorganization and scission in clathrin - independent endocytosis . Cell 140 , 540 \u2013 553 ( 2010 ) . 20 . Gauthier , N . C . et al . Helicobacter pylori VacA cytotoxin : a probe for a clathrin - independent and Cdc42 - dependent pinocytic pathway routed to late endosomes . Mol . Biol . Cell . 16 , 4852 \u2013 4866 ( 2005 ) . 21 . Kirkham , M . et al . Ultrastructural identi \ufb01 cation of uncoated caveolin - independent early endocytic vehicles . J . Cell Biol . 168 , 465 \u2013 476 ( 2005 ) . 22 . Nonnenmacher , M . & Weber , T . Adeno - associated virus 2 infection requires endocytosis through the CLIC / GEEC pathway . Cell Host Microbe 10 , 563 \u2013 576 ( 2011 ) . 23 . Howes , M . T . et al . Clathrin - independent carriers form a high capacity endocytic sorting system at the leading edge of migrating cells . J . Cell Biol . 190 , 675 \u2013 691 ( 2010 ) . 24 . Kalia , M . et al . Arf6 - independent GPI - anchored protein - enriched early endosomal compartments fuse with sorting endosomes via a Rab5 / phosphatidylinositol - 3 \u2019 - kinase - dependent machinery . Mol . Biol . Cell 17 , 3689 \u2013 3704 ( 2006 ) . 25 . Kettle , E . et al . A cholesterol - dependent endocytic mechanism generates midbody tubules during cytokinesis . Traf \ufb01 c 16 , 1174 \u2013 1192 ( 2015 ) . 26 . Hemalatha , A . , Prabhakara , C . & Mayor , S . Endocytosis of Wingless via a dynamin - independent pathway is necessary for signaling in Drosophila wing discs . Proc . Natl Acad . Sci . USA 113 , E6993 \u2013 E7002 ( 2016 ) . 27 . Miesenb\u00f6ck , G . , De Angelis , D . A . & Rothman , J . E . Visualizing secretion and synaptic transmission with pH - sensitive green \ufb02 uorescent proteins . Nature 394 , 192 \u2013 195 ( 1998 ) . 28 . Merri \ufb01 eld , C . J . , Perrais , D . & Zenisek , D . Coupling between clathrin - coated - pit invagination , cortactin recruitment , and membrane scission observed in live cells . Cell 121 , 593 \u2013 606 ( 2005 ) . 29 . Merri \ufb01 eld , C . J . , Feldman , M . E . , Wan , L . & Almers , W . Imaging actin and dynamin recruitment during invagination of single clathrin - coated pits . Nat . Cell Biol . 4 , 691 \u2013 698 ( 2002 ) . 30 . Qualmann , B . , Koch , D . & Kessels , M . M . Let \u2019 s go bananas : revisiting the endocytic BAR code . EMBO J . 30 , 3501 \u2013 3515 ( 2011 ) . 31 . Lundmark , R . et al . The GTPase - activating protein GRAF1 regulates the CLIC / GEEC endocytic pathway . Curr . Biol . 18 , 1802 \u2013 1808 ( 2008 ) . 32 . Kumari , S . Dynamin - independent endocytosis : molecular mechanisms and membrane dynamics . PhD Thesis , NCBS - TIFR , Manipal Univ . 65 , 148 , 150 \u2013 151 . ( 2008 ) 33 . Boal , F . et al . LG186 : an inhibitor of GBF1 function that causes golgi disassembly in human and canine cells . Traf \ufb01 c 11 , 1537 \u2013 1551 ( 2010 ) . 34 . Wassmer , T . et al . A loss - of - function screen reveals SNX5 and SNX6 as potential components of the mammalian retromer . J . Cell Sci . 120 , 45 \u2013 54 ( 2007 ) . 35 . Dai , J . et al . ACAP1 promotes endocytic recycling by recognizing recycling sorting signals . Dev . Cell 7 , 771 \u2013 776 ( 2004 ) . 36 . Hartig , S . M . et al . The F - BAR protein CIP4 promotes GLUT4 endocytosis through bidirectional interactions with N - WASp and dynamin - 2 . J . Cell Sci . 122 , 2283 \u2013 2291 ( 2009 ) . 37 . Rodal , A . A . , Motola - Barnes , R . N . & Littleton , J . T . Nervous Wreck and Cdc42 cooperate to regulate endocytic actin assembly during synaptic growth . J . Neurosci . 28 , 8316 \u2013 8325 ( 2008 ) . 38 . Scita , G . , Confalonieri , S . , Lappalainen , P . & Suetsugu , S . IRSp53 : crossing the road of membrane and actin dynamics in the formation of membrane protrusions . Trends Cell Biol . 18 , 52 \u2013 60 ( 2008 ) . 39 . Nakagawa , H . et al . IRSp53 is colocalised with WAVE2 at the tips of protruding lamellipodia and \ufb01 lopodia independently of Mena . J . Cell Sci . 116 , 2577 \u2013 2583 ( 2003 ) . 40 . Goh , W . I . et al . mDia1 and WAVE2 proteins interact directly with IRSp53 in \ufb01 lopodia and are involved in \ufb01 lopodium formation . J . Biol . Chem . 287 , 4702 \u2013 4714 ( 2012 ) . NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w ARTICLE NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications 15 41 . Miki , H . , Yamaguchi , H . , Suetsugu , S . & Takenawa , T . IRSp53 is an essential intermediate between Rac and WAVE in the regulation of membrane ruf \ufb02 ing . Nature 408 , 732 \u2013 735 ( 2000 ) . 42 . Disanza , A . et al . CDC42 switches IRSp53 from inhibition of actin growth to elongation by clustering of VASP . EMBO J . 32 , 2735 \u2013 2750 ( 2013 ) . 43 . Vaggi , F . et al . The Eps8 / IRSp53 / VASP network differentially controls actin capping and bundling in \ufb01 lopodia formation . PLoS Comput . Biol . 7 , e1002088 ( 2011 ) . 44 . Disanza , A . et al . Regulation of cell shape by Cdc42 is mediated by the synergic actin - bundling activity of the Eps8 \u2013 IRSp53 complex . Nat . Cell Biol . 8 , 1337 \u2013 1347 ( 2006 ) . 45 . Funato , Y . et al . IRSp53 / Eps8 complex is important for positive regulation of Rac and cancer cell motility / invasiveness . Cancer Res . 64 , 5237 \u2013 5244 ( 2004 ) . 46 . Kast , D . J . et al . Mechanism of IRSp53 inhibition and combinatorial activation by Cdc42 and downstream effectors . Nat . Struct . Mol . Biol . 21 , 413 \u2013 422 ( 2014 ) . 47 . Lewis - Saravalli , S . , Campbell , S . & Claing , A . ARF1 controls Rac1 signaling to regulate migration of MDA - MB - 231 invasive breast cancer cells . Cell Signal . 25 , 1813 \u2013 1819 ( 2013 ) . 48 . Pr\u00e9vost , C . et al . IRSp53 senses negative membrane curvature and phase separates along membrane tubules . Nat . Commun . 6 , 8529 ( 2015 ) . 49 . Ariotti , N . et al . Modular detection of GFP - labeled proteins for rapid screening by electron microscopy in cells and organisms . Dev . Cell 35 , 513 \u2013 525 ( 2015 ) . 50 . Tillberg , P . W . et al . Protein - retention expansion microscopy of cells and tissues labeled using standard \ufb02 uorescent proteins and antibodies . Nat . Biotechnol . 34 , 987 \u2013 992 ( 2016 ) . 51 . Chen , F . , Tillberg , P . W . & Boyden , E . S . Expansion microscopy . Science 347 , 543 \u2013 548 ( 2015 ) . 52 . Havrylenko , S . et al . WAVE binds Ena / VASP for enhanced Arp2 / 3 complex \u2013 based actin assembly . Mol . Biol . Cell 26 , 55 \u2013 65 ( 2015 ) . 53 . Lim , K . B . et al . The Cdc42 effector IRSp53 generates \ufb01 lopodia by coupling membrane protrusion with actin dynamics . J . Biol . Chem . 283 , 20454 \u2013 20472 ( 2008 ) . 54 . Nolen , B . J . et al . Characterization of two classes of small - molecule inhibitors of Arp2 / 3 complex . Nature 460 , 1031 \u2013 1034 ( 2009 ) . 55 . Rizvi , S . A . et al . Identi \ufb01 cation and characterization of a small molecule inhibitor of formin - mediated actin assembly . Chem . Biol . 16 , 1158 \u2013 1168 ( 2009 ) . 56 . Riedl , J . et al . Lifeact : a versatile marker to visualize F - actin . Nat . Methods 5 , 605 \u2013 607 ( 2008 ) . 57 . Rohatgi , R . et al . The interaction between N - WASP and the Arp2 / 3 complex links Cdc42 - dependent signals to actin assembly . Cell 97 , 221 \u2013 231 ( 1999 ) . 58 . Rocca , D . et al . The small GTPase Arf1 modulates Arp2 / 3 - mediated actin polymerization via PICK1 to regulate synaptic plasticity . Neuron 79 , 293 \u2013 307 ( 2013 ) . 59 . Rocca , D . L . , Martin , S . , Jenkins , E . L . & Hanley , J . G . Inhibition of Arp2 / 3 - mediated actin polymerization by PICK1 regulates neuronal morphology and AMPA receptor endocytosis . Nat . Cell Biol . 10 , 259 \u2013 271 ( 2008 ) . 60 . Nakamura , Y . et al . PICK1 inhibition of the Arp2 / 3 complex controls dendritic spine size and synaptic plasticity . EMBO J . 30 , 719 \u2013 730 ( 2011 ) . 61 . Thorsen , T . S . et al . Identi \ufb01 cation of a small - molecule inhibitor of the PICK1 PDZ domain that inhibits hippocampal LTP and LTD . Proc . Natl Acad . Sci . USA 107 , 413 \u2013 418 ( 2010 ) . 62 . Sandvig , K . , Torgersen , M . L . , Raa , H . A . & Van Deurs , B . Clathrin - independent endocytosis : from nonexisting to an extreme degree of complexity . Histochem . Cell Biol . 129 , 267 \u2013 276 ( 2008 ) . 63 . Jolla , L . et al . Clathrin - independent pinocytosis is induced in cells overexpressing a temperature - sensitive mutant of dynamin . J . Cell Biol . 131 , 69 \u2013 80 ( 1995 ) . 64 . Kaksonen , M . , Toret , C . P . & Drubin , D . G . A modular design for the clathrin - and actin - mediated endocytosis machinery . Cell 123 , 305 \u2013 320 ( 2005 ) . 65 . Kirkham , M . & Parton , R . G . Clathrin - independent endocytosis : new insights into caveolae and non - caveolar lipid raft carriers . Biochim . Biophys . Acta Mol . Cell Res . 1745 , 273 \u2013 286 ( 2005 ) . 66 . Simunovic , M . et al . Friction mediates scission of tubular membranes scaffolded by BAR proteins . Cell 170 , 172 \u2013 184 . e11 ( 2017 ) . 67 . Day , C . A . et al . Microtubule motors power plasma membrane tubulation in clathrin - independent endocytosis . Traf \ufb01 c 16 , 572 \u2013 590 ( 2015 ) . 68 . Mastronarde , D . N . Automated electron microscope tomography using robust prediction of specimen movements . J . Struct . Biol . 152 , 36 \u2013 51 ( 2005 ) . 69 . Kremer , J . R . , Mastronarde , D . N . & McIntosh , J . R . Computer visualization of three - dimensional image data using IMOD . J . Struct . Biol . 116 , 71 \u2013 76 ( 1996 ) . 70 . Bolte , S . & Cordeli\u00e8res , F . P . A guided tour into subcellular colocalization analysis in light microscopy . J . Microsc . 224 , 213 \u2013 232 ( 2006 ) . 71 . Jaqaman , K . et al . Robust single - particle tracking in live - cell time - lapse sequences . Nat . Methods 5 , 695 \u2013 702 ( 2008 ) . Acknowledgements We thank Ramya Purkanti , Kabir Hussain and Balaji Ramalingam for help with the analysis , Neeraj Sebastian for making TagRFPt - CDC42 , Rashmi Godbole for help with expansion microscopy , R . P . for her help with BAR domain database creation , Marcus Taylor ( UCSF / NCBS ) for mCherry - IRSp53 and SecGFP - TfR constructs , Gero Mie - senb \u04e7 ck ( University of Oxford ) for ecliptic - GPI , Paul Melan\u00e7on ( University of Alberta ) for ARF1 - mCherry , Catherine Jackson ( National Institutes of Health ) for GBF1 - mCherry , Roland Wedlich - Soeldner ( Universit\u00e4t M\u00fcnster ) for pRuby - lifeact , TagRFPt - PICK1 and GFP - PICK1 ( Harvey McMohan , MRC ) , Mike Way ( The Francis Crick Institute ) for GFP - NWASP \u0394 VCA and GFP - NWASP - VCA domain , J . Hanley ( Uni - versity of Bristol , UK ) for pIRES - EGFP - PICK KK - EE mutant and N - WASP CA domain and Manoj Matthew from CIFF ( Central Imaging and Flow cytometry Facility , NCBS ) for helping to set up the pH pulsing setup . Authors were supported by Wellcome Trust - DBT India Alliance Early Career Fellowship ( G . M . ) , NCBS - TIFR graduate student fel - lowship ( M . S . ) . J . C . Bose Fellowship and a Margadarshi Fellowship ( IA / M / 15 / 1 / 502018 ) from the Wellcome Trust - DBT Alliance ( S . M . ) , Wellcome Trust - DBT India Alliance Intermediate Fellowship and Simons Centre for Living Machines ( M . T . ) , and grants , DBT - CoE ( S . M ) , R . G . P was supported by the National Health and Medical Research Council ( NHMRC ) of Australia ( programme grant , APP1037320 and Senior Principal Research Fellowship , 569452 ) , and the Australian Research Council Centre of Excellence in Convergent Bio - Nanoscience and Technology ( CE140100036 ) , G . S . and A . D . were supported by Italian Association for Cancer Research Investigator Grant ( 10168 and 18621 to G . S . ) and European Research Council ( 268836 to G . S . ) . We acknowledge the Australian Microscopy & Microanalysis Research Facility at the Centre for Microscopy and Microanalysis at The University of Queensland . We acknowledge the CIFF ( Central Imaging and Flow cytometry Facility at National Centre for Biological Sciences , TIFR , India . We thank K . Joseph Matthew ( S . M . Laboratory ) for making the schematic in Fig . 1a and Fig . 8 . Author contributions M . S . and G . M . executed and analysed all \ufb02 uorescence microscopy experiments . M . S . and M . T . developed the analysis . R . G . P . and J . R . performed , analysed and interpreted all electron microscopy experiments . A . D . and G . S . helped with the IRSp53 knockout and IRSp53 mutant addback cell line construction . G . M . , M . S . and S . M . planned all experi - ments and wrote the manuscript with inputs from the remaining authors . Additional information Supplementary Information accompanies this paper at https : / / doi . org / 10 . 1038 / s41467 - 018 - 03955 - w . Competing interests : The authors declare no competing interests . Reprints and permission information is available online at http : / / npg . nature . com / reprintsandpermissions / Publisher ' s note : Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional af \ufb01 liations . Open Access This article is licensed under a Creative Commons Attribution 4 . 0 International License , which permits use , sharing , adaptation , distribution and reproduction in any medium or format , as long as you give appropriate credit to the original author ( s ) and the source , provide a link to the Creative Commons license , and indicate if changes were made . The images or other third party material in this article are included in the article \u2019 s Creative Commons license , unless indicated otherwise in a credit line to the material . If material is not included in the article \u2019 s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use , you will need to obtain permission directly from the copyright holder . To view a copy of this license , visit http : / / creativecommons . org / licenses / by / 4 . 0 / . \u00a9 The Author ( s ) 2018 ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / s41467 - 018 - 03955 - w 16 NATURE COMMUNICATIONS | ( 2018 ) 9 : 1835 | DOI : 10 . 1038 / s41467 - 018 - 03955 - w | www . nature . com / naturecommunications",
    "bibeau2023twist": "PNAS 2023 Vol . 120 No . 4 e2208536120 https : / / doi . org / 10 . 1073 / pnas . 2208536120 1 of 12 RESEARCH ARTICLE | Significance How actin filaments respond to mechanical loads is central to understanding cellular force generation and mechanosensing . While there is consensus on the actin filament bending stiffness , reported values of the filament torsional stiffness vary by almost 2 orders of magnitude . We used magnetic tweezers and hydrodynamic flow to determine how filaments respond to applied twisting and pulling loads . Twisting causes filaments to adopt a supercoil conformation . Pulling forces inhibit supercoil formation and fragment filaments . These observations explain how contractile forces generated by myosin motors accelerate filament severing by cofilin regulatory proteins in cells . Author affiliations : a Department of Molecular Biophysics and Biochemistry , Yale University , New Haven , CT 06520 Author contributions : J . P . B . , N . G . P . , and E . M . D . L . C . designed research ; J . P . B . , N . G . P . , and N . S . performed research ; J . P . B . , N . G . P . , S . G . , N . S . , C . V . S . , W . C . , and E . M . D . L . C . analyzed data ; and J . P . B . , N . G . P . , W . C . , and E . M . D . L . C . wrote the paper . The authors declare no competing interest . This article is a PNAS Direct Submission . D . V . is a guest editor invited by the Editorial Board . Copyright \u00a9 2023 the Author ( s ) . Published by PNAS . This article is distributed under Creative Commons Attribution - NonCommercial - NoDerivatives License 4 . 0 ( CC BY - NC - ND ) . 1 To whom correspondence may be addressed . Email : enrique . delacruz @ yale . edu . This article contains supporting information online at https : / / www . pnas . org / lookup / suppl / doi : 10 . 1073 / pnas . 2208536120 / - / DCSupplemental . Published January 19 , 2023 . BIOPHYSICS AND COMPUTATIONAL BIOLOGY Twist response of actin filaments Jeffrey P . Bibeau a , Nandan G . Pandit a , Shawn Gray a , Nooshin Shatery Nejad a , Charles V . Sindelar a , Wenxiang Cao a , and Enrique M . De La Cruz a , 1 Edited by Dimitrios Vavylonis , Lehigh University , Bethlehem , PA ; received June 1 , 2022 ; accepted December 16 , 2022 by Editorial Board Member Yale E . Goldman Actin cytoskeleton force generation , sensing , and adaptation are dictated by the bending and twisting mechanics of filaments . Here , we use magnetic tweezers and microfluid - ics to twist and pull individual actin filaments and evaluate their response to applied loads . Twisted filaments bend and dissipate torsional strain by adopting a supercoiled plectoneme . Pulling prevents plectoneme formation , which causes twisted filaments to sever . Analysis over a range of twisting and pulling forces and direct visualization of filament and single subunit twisting fluctuations yield an actin filament torsional persis - tence length of ~ 10 \u00b5m , similar to the bending persistence length . Filament severing by cofilin is driven by local twist strain at boundaries between bare and decorated segments and is accelerated by low pN pulling forces . This work explains how contractile forces generated by myosin motors accelerate filament severing by cofilin and establishes a role for filament twisting in the regulation of actin filament stability and assembly dynamics . actin | torsion | plectoneme | cofilin | severing Cells sense , respond , and adapt to internal and external forces ( 1 , 2 ) . The actin cytoskel - eton , a dynamic , branched , and cross - linked network of protein filaments ( 3 ) that behave as semiflexible polymers on cellular length scales ( 4 \u2013 6 ) , mediates many of these cellular responses . Cellular actin networks are pulled ( 7 , 8 ) , squeezed ( 9 , 10 ) , and twisted ( 11 , 12 ) during growth and remodeling and through interactions with contrac - tile and regulatory binding proteins ( 13 ) . These physical forces can stall network growth ( 14 ) , alter the filament structure ( 15 , 16 ) , modulate interactions among filaments ( 17 ) and with regulatory proteins ( 7 , 8 , 18 , 19 ) , and induce filament fragmentation ( 15 , 19 \u2013 23 ) , all of which influence network remodeling and mediate cellular \u201cmechano - sensing\u201d ( 24 , 25 ) . The capacity for actin networks to respond to force is dictated by the mechanical prop - erties of filaments . Relaxed ( i . e . , resting ) filaments are straight but helical with an intrinsic twist ( 26 ) . The forces required to twist and bend a filament scale with the filament mechan - ical properties ( 4 , 27 ) , specifically their bending and torsional stiffness , which are com - monly represented in terms of bending and twisting persistence lengths ( L B and L T ; we note these are effective persistence lengths because filaments are not homogeneous , iso - tropic materials ) . Filaments with larger persistence lengths are stiffer and require more force to deform than those with shorter persistence lengths . Similarly , stiff filaments store more elastic strain energy for any given deformation than more compliant ones . The elastic free energy ( i . e . , strain energy ) stored in the filament shape ( 16 ) can generate force and work when relaxing to the resting configuration . It can also fragment filaments ( 15 , 19 \u2013 23 , 28 ) and mediate interactions with binding partners ( 7 , 8 , 18 , 19 ) . Dissipation of elastic energy in bent filaments contributes to force generation at the leading edge of migrating cells ( 10 , 29 ) and during essential cellular processes such as endocytosis ( 9 ) . Twisted filaments are also strained . Such twisting has been implicated in symmetry break - ing ( 30 , 31 ) , network chirality ( 11 ) , and the buckling of actin networks in filopodia ( 12 ) . Quantitative knowledge of filament bending and twisting mechanics is therefore critical to reliably account for and model complex cellular behaviors . The bending mechanics of actin filaments have been extensively characterized . There is general agreement that filaments have a bending persistence length ( L B ) of ~ 10 \u00b5m ( 32 \u2013 34 ) , which can be modulated by regulatory proteins ( 35 , 36 ) and ligands ( 5 , 33 , 37 , 38 ) . A consensus on the filament twisting stiffness is lacking , with torsional persistence lengths reported from 0 . 5 to 20 \u00b5m ( 20 , 39 \u2013 44 ) . In addition , it is not known how filament twisting and bending are coupled ( 16 , 27 ) or how filaments respond to combinations of twisting , bending , and pulling forces , as experienced in cells . Here , we use a magnetic tweezers apparatus coupled with microfluidics to evaluate how single actin filaments respond to applied twisting and pulling loads . Our results and analyses provide multiple , independent determinations of the filament bending and twist - ing stiffness , demonstrate how bending and twisting are coupled , and show how this coupling is affected by pulling and filament fragmentation . These findings have D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . 2 of 12 https : / / doi . org / 10 . 1073 / pnas . 2208536120 pnas . org implications for actin cytoskeleton mechanosensing and network force generation and remodeling . Results Twisting and Pulling Actin Filaments . We developed an assay to twist actin filaments about their long axis with magnetic tweezers while simultaneously visualizing them by a TIRF microscope ( Fig . 1 ) . Short , Alexa 488 \u2013 labeled actin filament seeds were tethered to the surface of the microscope coverslip and elongated from the barbed end with purified , Alexa 647 \u2013 labeled actin monomers . These filaments were further elongated from their barbed ends with digoxigenin - conjugated actin monomers to which we attached a paramagnetic bead . The barbed - end \u2013 conjugated paramagnetic bead was twisted at a constant rate ( 0 . 31 rot s \u22121 ) with permanent magnets mounted on a stepper motor ( Fig . 1 B ) . Filament - attached beads and filaments rotated in phase with the permanent magnets ( Fig . 1 B \u2013 D and Movies S1 and S2 ) , indicating that no slipping occurs during manipulation . Filaments attached to two beads ( Fig . 1 B ) were used only to determine whether rotations were in phase with the permanent magnet ( Fig . 1 C ) ; filaments attached to a single bead were used in all subsequent experiments . Pulling forces exerted by the permanent magnet are negligible in our experimental setup ( SI Appendix , Fig . S1 ) , so they were applied with fluid flow using a microfluidic device . The force applied on the filament scales with the size of the bead ( d = 2 . 8 \u00b5m ) and the fluid velocity ( SI Appendix ) . This force is constant through - out the filament ( at a constant flow rate ) and is independent of the filament length in contrast to the much smaller and negligible forces exerted by flow on tethered filaments without conjugated beads ( 17 , 22 , 45 , 46 ) . Actin Filament Force \u2013 Extension Response . To establish the mechanical properties of actin filaments can be reliably determined with our experimental conditions , we measured the bending persistence length ( L B ) in the absence of twisting from the force \u2013 extension response of filaments undergoing thermally driven shape fluctuations ( Fig . 2 and Movie S3 ) . The pulling force required to straighten thermally bent filament scales with the bending stiffness ( L B ) . The filament end - to - end distance ( R ) , defined as the linear distance from the bead and surface attachment points , depends on the long - axis pulling force ( Fig . 2 ) . In the absence of fluid flow ( i . e . , pulling force ~ 0 ) filaments undergo thermally driven bending , so R is shorter than the filament contour length ( L ) , i . e . , R / L < 1 . At ~ 0 . 02 pN pulling force , the ratio of the filament end - to - end distance and the contour length ( R / L ) was ~ 0 . 92 ( Fig . 2 ) . The end - to - end distance approached the contour length R / L ~ 1 at Fig . 1 . Twisting actin filaments with magnetic tweezers . ( A ) Cartoon schematic of the experimental setup . Alexa 488 \u2013 labeled actin filament seeds ( green ) were attached to a Biotin - PEG - Silane surface through biotin ( yellow circles ) and neutravidin ( black diamonds ) interactions and elongated from the barbed ends with Alexa 647 \u2013 labeled actin ( red ) . Filaments were further elongated from their barbed ends with digoxigenin ( DIG ) - labeled actin ( purple ) . Paramagnetic beads ( 2 . 8 \u03bc m in diameter ; gray ) coated with DIG antibodies ( purple ) were attached to filaments at or near their barbed ends . Filament - attached beads can be rotated by a permanent magnet ( blue and red rectangles ) and pulled by buffer flow ( black arrow ) . Relative filament and bead sizes are not drawn to scale . Twisting clockwise or counterclockwise corresponds to under - or overtwisting , respectively . ( B ) Rotation of a phalloidin - decorated filament attached to two paramagnetic beads . ( C ) Cosine of the rotational angle of the second paramagnetic bead ( black trace ) and the magnet ( red trace ) indicates that the bead and magnet rotation are in phase . ( D ) Rotation of an Alexa 647 \u2013 labeled actin filament with visible attachment to paramagnetic bead indicates that the bead and filament rotation are in phase . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . PNAS 2023 Vol . 120 No . 4 e2208536120 https : / / doi . org / 10 . 1073 / pnas . 2208536120 3 of 12 > 3 pN pulling force ( Fig . 2 ) . The best fit of the force dependence of the filament end - to - end distance ( Fig . 2 ) to a worm - like chain model [ Eq . 1 ; ( 47 ) ] yields an actin filament bending persistence length ( L B ) of 10 . 7 ( \u00b11 . 0 ) \u00b5m ( Table 1 ) , consistent with previous wet lab and computational model determinations of bare filaments [ i . e . , without phalloidin or other binding partners ( 19 , 32 \u2013 35 , 42 , 48 \u2013 50 ) ] . The forces in these experiments are calculated using a bead - center distance from the surface ( d ) equal to the bead radius ( r ; d = r ) since the bead was at the surface of coverslip ( SI Appendix ) . Twisted Filaments Bend and Supercoil . Filaments undergo a series of shape transitions when continuously twisted with magnetic tweezers ( Movie S4 ) . In the absence of applied twist ( and with or without pulling loads ) , filaments bend randomly due to thermally driven forces . Applied twisting ( twist density < 0 . 8 rot \u00b5m \u22121 ) causes filaments to bend further in a nonrandom manner , provided long - axis pulling forces are weak ( \u22640 . 03 pN ; Fig . 3 A ) , and the value of R continues to shorten gradually with twisting until a critical twist density ( \u03c3 s ) is reached , at which point filaments form a looped segment . Additional twisting causes the linear , nonlooped filament segments to wind up and twist around each other , yielding an interwound , actin filament supercoil ( called a plectoneme ) like those observed with twisted DNA ( 51 , 52 ) . This transition is detected as a dramatic and abrupt reduction in R ( Fig . 3 B ) . Only a single loop ( i . e . , one plectoneme ) per filament was observed in our experiments . Plectoneme formation is reversible and relaxed ( i . e . , straight ) filaments can be recovered with untwisting ( Movie S5 ) . We note that rhodamine phalloidin \u2013 decorated actin filaments supercoil when subjected to high - intensity laser light ( 53 ) , presumably due to photo - induced torsional strain . Actin filaments are helical and can be described as having an intrinsic right - handed twist . Therefore , counterclockwise rotations increase the intrinsic twist ( \u201covertwisting\u201d ) , whereas clockwise rotations lower the intrinsic twist ( \u201cundertwisting\u201d ) . Binding of cofilin / ADF regulatory proteins , for example , also undertwists filaments ( 54 , 55 ) . Filament plectoneme formation depends on the pulling force but not on the twisting direction ( i . e . , over - ver - sus undertwisting ) , as indicated by the symmetry of the twist \u2013 extension response curves ( Fig . 3 B ) . The lack of a detectable asymmetry with over - and undertwisting is surprising given the intrinsic filament twist . However , this observed behavior likely arises from the fact that the deformations associated with plec - toneme formation are modest at the subunit level and not in the regime in which differences between the two directions could be detected ( 16 ) . Filaments adopt a plectoneme configuration at low twist densities ( ~ 0 . 2 rot \u00b5m \u22121 ) when the pulling force is low ( < 0 . 01 pN ) but require more twist ( ~ 0 . 8 rot \u00b5m \u22121 ) at higher pulling forces ( 0 . 03 pN ; Fig . 3 ) . Plectonemes did not form when the pulling force was 0 . 25 pN , even up to twist densities of ~ 1 rot \u00b5m \u22121 . The reversibility of plectoneme formation was also independent of the twisting direction . The plectoneme loop size also depends on the pulling force ( SI Appendix , Fig . S2 ) . The average loop radius was ~ 400 to 500 nm under 0 . 01 pN , whereas it was ~ 200 nm under 0 . 03 pN . The local filament curvatures at these radii are small compared with those predicted to significantly accelerate filament fragmentation ( 15 ) , consistent with these plectoneme loops being stable through - out the duration of our experiments . Filament Twist \u2013 Extension Response . The experimental data presented thus far demonstrate that twisted actin filaments bend and adopt supercoiled plectoneme structures when pulling forces are low ( < 0 . 25 pN ) . This response originates from the intrinsic filament bending and twisting mechanics and the coupling between these two deformations ( 16 , 52 , 56 ) . Fig . 2 . Actin filament force \u2013 extension response . ( A ) Representative fluorescent images of Alexa 647 \u2013 labeled actin filaments ( magenta ) conjugated to a paramagnetic bead ( cyan ) under fluid flow . No magnetic field is applied . ( B ) Force \u2013 extension curves for actin filaments of varying lengths ( colored points ) with the global best fit to Eq . 1 ( colored lines ) with pulling forces at d = r ( Methods ) . ( C ) Average force \u2013 extension curves , normalized to filament lengths , for 7 actin filaments and corresponding theory with the global best fit persistence length of 10 . 7 ( \u00b11 ) \u00b5m ( solid red line ) ( Eq . 1 ) . Theoretical force \u2013 extension curves ( Eq . 1 ) in descending order with L B = 100 , 5 , 1 , and 0 . 1 \u00b5m ( dashed red lines ) . Uncertainty bars indicate standard error of the mean ( SEM ) . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . 4 of 12 https : / / doi . org / 10 . 1073 / pnas . 2208536120 pnas . org Accordingly , the actin filament mechanical properties can be extracted from the data with appropriate theory , analysis , and modeling . A two - state model used to describe plectoneme formation in DNA ( 52 ) ( modified to include polymers with L ~ L B , such as actin filaments ; Methods ; Eq . 3 ) accounts for the twist \u2013 extension response and plectoneme formation of actin filaments over the range of pulling forces evaluated here ( Fig . 3 B ) . This model con - siders actin filament segments as semiflexible rods in either \u201cstraight\u201d or plectoneme states . Pulling forces favor the straight configuration . Twisting introduces strain energy , which is dissi - pated by bending and subsequent plectoneme formation . Prior to reaching a critical twist density ( \u03c3 s ) for plectoneme formation , twisted filaments bend to dissipate torsional strain , shortening R . Once a plectoneme has formed , applied twisting strain is dissi - pated through conversion of strained , straight state segments to more relaxed plectoneme configurations ( i . e . , straight segments shorten , while plectoneme segments elongate linearly with applied twisting loads ) . Since only the linear , nonlooped filament seg - ments contribute to the R value ( plectoneme segments do not ) , the R value decreases linearly with the applied twist until R = 0 , at which the entire filament is in a plectoneme configuration ( middle line in Eq . 3 ; Fig . 3 B ) . The applied twist at R = 0 is referred to as \u03c3 p . The best fit of the filament twist \u2013 extension data to this model ( Eq . 3 ; Fig . 3 B ) yields actin filament persistence lengths for bend - ing ( L B ) and twisting ( L T ) of 10 . 9 ( \u00b12 . 6 ) and 11 . 7 ( \u00b12 . 2 ) \u00b5m , respectively ( Table 1 ) . This value of the bending persistence length ( L B ) is comparable with the value of 10 . 7 ( \u00b11 . 0 ) \u00b5m obtained from the force \u2013 extension response ( Fig . 2 ) . Filament torsional persistence lengths an order of magnitude longer or shorter do not account well for the observed experimental data ( SI Appendix , Fig . S5 ) . We note that some twisted filaments \u201cwobbled\u201d slightly dur - ing twisting manipulations ( Movie S2 ) . To estimate the maxi - mum possible error introduced by bead wobbling in these cases , we analyzed the data assuming a larger force as expected if the bead moved away from the surface during rotation . Deviations far greater than observed for wobbling ( i . e . , one full bead height deviation , such that d = 2 r ; SI Appendix ) yield essentially iden - tical L T values ( L T = 11 . 7 \u00b1 2 . 2 \u00b5m versus L T = 11 . 4 \u00b1 2 . 1 \u00b5m for beads wobbling an entire bead height ) but ~ twofold lower L B values ( 10 . 9 \u00b1 2 . 6 \u00b5m versus 6 . 5 \u00b1 1 . 5 \u00b5m for beads wobbling an entire bead height ) . Thermally Driven , Filament Twist Fluctuations . We also determined the filament torsional stiffness by directly visualizing spontaneous , thermally driven , twist fluctuations ( Fig . 4 ) . A hollow , cylindrical magnet was positioned above the tethered filament to generate a magnetic field ( perpendicular to the surface of the sample chamber ) that held the filament orthogonally to the surface without constraining its rotation ( 57 ) ( Fig . 4 A and B and Movie S6 ) . The pulling force generated Table 1 . Actin filament bending and torsional persis - tence lengths Persistence Length ( \u00b5m ) Assay Bending ( L B ) 10 . 7 ( \u00b11 . 0 ) Force \u2013 extension , Fig . 2 10 . 9 ( \u00b12 . 6 ) Twist \u2013 extension , Fig . 3 , f at d = r Twisting ( L T ) 11 . 7 ( \u00b12 . 2 ) Twist \u2013 extension , Fig . 3 , f at d = r 12 . 9 ( \u00b12 . 4 ) Twist fluctuations , Fig . 4 8 . 2 ( \u00b10 . 2 ) Cryo - EM , refinement volume of 5 subunits , histogram fit , Fig . 5 5 . 5 ( \u00b10 . 2 ) Cryo - EM , refinement volume of 5 subunits , MLE * 4 . 4 ( \u00b12 . 7 ) Cryo - EM , refinement volume of 5 subunits , MCMC \u2020 5 . 8 ( \u00b10 . 3 ) Cryo - EM , refinement volume of 1 subunit , histogram fit 7 . 0 ( \u00b11 . 1 ) Cryo - EM , refinement volume of 1 subunit , MLE * 6 . 3 ( \u00b13 . 3 ) Cryo - EM , refinement volume of 1 subunit , MCMC \u2020 * Maximum likelihood estimation ( MLE , see Methods ) . \u2020 Markov chain Monte Carlo ( MCMC , see Methods ) . Fig . 3 . Actin filament twist \u2013 extension response and supercoiling . ( A ) Representative fluorescent images of actin filament twist \u2013 extension with 0 . 01 ( Top ) , 0 . 03 ( Middle ) , and 0 . 25 ( Bottom ) pN pulling force ( SI Appendix , Eq . S1 with d = r ) . ( Scale bar , 5 \u00b5m . ) ( B ) Twist \u2013 extension curves for actin filaments under 0 . 01 ( white circles ) , 0 . 03 ( gray circles ) , and 0 . 25 ( black circles ) pN pulling forces with the global best fit to Eq . 3 ( red lines ) . Solid and dashed red lines differentiate model before and after plectoneme formation , respectively . Rotations along the positive x - axis indicate filament overtwisting , and rotations along the negative x - axis indicate undertwisting . The complete dataset represents 50 filaments and n > 3 for each experimental condition . Uncertainty bars represent SEM . The asymmetric look of red dashed fitting lines for over - and undertwists under the same pulling force is due to the different fixed parame t ers ( 1 L ) in the fitting ( Methods ) and not because of differences in response to applied over - and undertwist . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . PNAS 2023 Vol . 120 No . 4 e2208536120 https : / / doi . org / 10 . 1073 / pnas . 2208536120 5 of 12 by the closely positioned hollow magnet maintains the filament relatively straight , thereby eliminating contributions from filament bending to the observed dynamics and allowing determination of the true filament torsional stiffness ( SI Appendix , Fig . S3 , ( 39 , 56 ) ) . Filament twisting fluctuations were monitored by tracking a smaller fluorescent marker bead attached to the paramag - netic bead at the filament barbed end ( Fig . 4 C and D and Movie S7 ) . The angular fluctuations of the marker bead reflect the cumulative rotational fluctuation of all subunits between the surface and paramagnetic bead attachment points ( L = 8 to 19 \u03bc m , ~ 365 subunits \u00b5m \u22121 ; SI Appendix , Fig . S6 ) . The rotation angle of the marker beads around the filament center axis fluctuates randomly ( Fig . 4 C and D ) . Time courses of the observed marker bead angle variance ( Fig . 4 E ) plateau at times > 1 , 000 s , indicating the entire accessible diffusive space of the filament - attached bead had been sampled ( 58 ) . The twist persis - tence length for each set of data in Fig . 4 D was determined from the angular variance calculated by directly averaging the set of data and the filament length according to Eq . 4 ( Methods ) . The averaged ( n = 5 ) value from separate experiments yields a filament torsional Fig . 4 . Direct visualization of actin filament twisting fluctuations . ( A ) Cartoon schematic of the experimental setup . A cylindrical magnet with a center hole was positioned 5 mm above the coverslip surface \u2013 anchored actin filament ( i . e . , in the z direction perpendicular to the surface ) . A DIG - coated marker bead was added to the paramagnetic bead to track rotations . No flow was applied during experiments . Filament length and bending are not to scale . ( B ) ( a ) Assay is set up by identifying a filament attached to a paramagnetic bead ( large dim bead ) with identifiable marker beads ( small bright bead ) under fluid flow . ( b ) Fluid flow is turned off . ( c ) Cylindrical magnet is lowered into position . ( d ) Filament is pulled out of the focal plane . ( e ) Focal plane is adjusted to observe the rotational fluctuations of both beads . Images were taken every 5 s . ( C ) Example images of the angular fluctuations of the filament visualized by the absolute angle of a line connecting the two beads to the x direction . ( D ) The mean - subtracted absolute angle ( SI Appendix , Eq . S25 ) of the marker beads over time . Black trace indicates the angular fluctuations from the filament tracked in ( C ) . Gray traces represent four other sample traces from different experiments performed at different times . Histogram represents the distribution of absolute angles from the black trace . The actin filament torsional persistence length of 12 . 9 ( \u00b12 . 4 ) \u03bc m is an average ( n = 5 ) of separate measurements , each was determined from the value of the variance at long times ( see Panel E ) and the filament length according to Eq . 4 . ( E ) Time - dependent variance of the traces in D . It demonstrates that the measurement time of whole filament angular fluctuation has to be long enough for the variance to reach equilibrium such that experiencing all possibilities . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . 6 of 12 https : / / doi . org / 10 . 1073 / pnas . 2208536120 pnas . org persistence length of 12 . 9 ( \u00b12 . 4 ) \u00b5m ( Table 1 ) comparable with the value of 11 . 7 ( \u00b12 . 2 ) \u00b5m determined from the filament twist \u2013 extension response and plectoneme formation ( Table 1 ) and with reported values of ~ 16 \u00b5m ( 20 ) and ~ 6 \u00b5m ( 39 ) . Single Subunit Twisting Fluctuations . We measured the filament twisting persistence length a third way , from the variance of twisting angles between subunits , as visualized by electron cryomicroscopy ( Fig . 5 ) . Alignment parameters output from the 3D structure refinement yield estimates of filament subunit orientations and hence the twisting angle between them ( 54 ) . Deviations of the observed intersubunit twist angle from the intrinsic ( average ) filament twist reflect thermally driven twist fluctuations . The distribution ( i . e . , width ) of these deviations scales with the filament torsional stiffness , such that filaments displaying a narrow distribution are less compliant in twisting than those with a broader distribution . A challenge with accurately calculating this angle distribution by cryo - EM is the low signal - to - noise ratio associated with the images . The noise introduces uncertainty in the angle measure - ments that can exceed the true intersubunit angle variance . Since the torsional stiffness and persistence length ( L T ) are determined from the distribution variance ( SI Appendix , Eq . S15 ) , and large uncertainties in individual intersubunit angle measurements ( SI Appendix , Eq . S17 ) yield a larger variance than the true vari - ance , neglecting contributions from these uncertainties causes filaments to appear more compliant than they actually are . We therefore developed an analysis method ( SI Appendix , Eqs . S17 \u2013 S21 ) that addresses the uncertainty in angle measure - ment to accurately measure L T from cryo - EM micrographs . Because each filament subunit is independently subject to ther - mally driven torsional angle fluctuations in a scale dictated by the torsional stiffness , the width of the true angular distributions \u03c3 n + 1 2 , measured across filament segments , increases linearly with the number of subunits according to \ud835\udf0e 2 obs , n + 1 = n \u0394 s L T + \ud835\udf0e 2 \ud835\udf00 ( SI Appendix , Eq . S21 ) . In contrast , the uncertainty in the esti - mated twist angle \ud835\udf0e 2 \ud835\udf00 remains constant . Our cryo - EM images confirmed this behavior ( Fig . 5 C ) , yield - ing estimates of the true intersubunit torsional variance and per - sistence length ( \u03c3 n + 1 2 and L T , respectively ; SI Appendix , Eq . S21 ) from the slope of the line relating the observed variance ( \ud835\udf0e 2 \ud835\udf00 ) to n ( Fig . 5 C ) . The intercepts of these lines reflect the contribution of noise to \ud835\udf0e 2 obs ( SI Appendix , Eq . S21 ) . The best linear fit of the n - dependent variance , obtained by Gaussian distribution fit to the angle \u03a6 histogram measured from a refinement volume of 5 subunits , to SI Appendix , Eq . S21 yields an L T of 8 . 2 ( \u00b10 . 2 ; \u00b1 indicates SDs of the fit ) comparable with the L T values of 11 . 7 ( \u00b12 . 2 ) and 12 . 9 ( \u00b12 . 4 ) \u00b5m determined by plectoneme formation ( Fig . 3 ) and thermally driven filament torsional fluctuations ( Fig . 4 ) , respectively . We include two additional analysis methods , directly averaging ( MLE ) and MCMC ( Methods ) , to inde - pendently determine n - dependent variance and thus L T to see if one is more robust than the others ( Table 1 ) . The differences among the three different methods are not significant . We also repeated the analysis to the angles measured from a refinement volume of 1 subunit ( Table 1 ) . These values are not significantly different from those from a refinement volume of 5 subunits . Twist - Induced Filament Fragmentation . Filaments often fragmented during twist \u2013 extension manipulations ( Fig . 3 ) . At 0 . 25 pN pulling force , filaments fragmented before forming a plectoneme , indicating that intersubunit bonds rupture if strain from applied twisting is not dissipated ( 15 , 19 ) . Most filaments ( 83 / 96 ) fragmented at the bead or surface attachment sites ( twist density = 1 . 2 ( \u00b10 . 6 ) rot \u00b5m \u22121 ; discussed below ) , suggesting that the weakest mechanical elements are filament attachment points . The remaining events ( 13 / 96 ) could be reliably discerned as fragmentation within the filament . This occurred at a twist density of ~ 1 . 1 ( \u00b10 . 5 ) rot \u00b5m \u22121 ( Movie S8 ) . At low pulling forces ( \u22640 . 03 pN ) , filaments adopt a plectoneme configuration , which dissipates the torsional strain from applied twisting . Accordingly , no fragmentation was observed , and all filaments formed a plectoneme under 0 . 01 pN pulling force . At a pulling force of 0 . 03 pN , some fragmentation events were observed . These occurred at the onset or during plectoneme for - mation [ twist density = 0 . 83 ( \u00b10 . 17 ) ; Fig . 6 and Movie S9 ] , with 9 / 17 filaments fragmenting exclusively within the filament . Cofilin Promotes Twist - Induced Filament Fragmentation . Filaments saturated with the actin regulatory protein , cofilin , referred to as cofilactin [ cofilin ] = 2 \u03bc M , which is saturating for this yeast isoform under our conditions ( 35 , 38 ) , did not form a plectoneme , even at low ( 0 . 03 pN ) pulling force , because they fragmented . Cofilactin filament fragmentation occurred at a lower twist density than fragmentation of bare filaments ( Fig . 6 and Movies S10 and S11 ) . The twist density dependence of the filament survival probability decay ( Fig . 6 ) indicated a midpoint of ~ 1 rot \u00b5m \u22121 for fragmentation of bare actin , while the midpoint for cofilactin filaments was significantly lower ( 0 . 64 ( \u00b10 . 28 ) and 0 . 43 ( \u00b10 . 12 ) rot \u00b5m \u22121 for overtwisting and undertwisting , respectively ; Fig . 6 and Movies S10 and S11 ) . The observed fragmentation originates from twisting strain rather than flow - mediated forces as neither untwisted bare nor cofilactin filaments fragmented on the timescales of these experiments ( SI Appendix , Fig . S4 and Movie S12 ) . Discussion Actin Filament Torsional Persistence Length is ~ 10 \u03bc m . Here , we have shown through twist \u2013 extension ( Fig . 3 ) , filament rotational fluctuations ( Fig . 4 ) , and single subunit fluctuations ( Fig . 5 ) that actin filaments have a torsional persistence length of ~ 10 \u03bc m , comparable with their bending persistence lengths ( Fig . 2 ) ( 19 , 32 \u2013 35 , 38 , 42 , 48 \u2013 50 , 59 ) . The reported torsional persistence length of actin filaments varies significantly from 0 . 5 to 20 \u00b5m ( 20 , 22 , 39 \u2013 44 ) . Our measured L T is consistent with reported values determined in optical traps ( 20 , 39 ) , fluorescent polarization microscopy ( 22 ) , and electron microscopy of filaments straightened with hydrodynamic flow ( 43 ) but differs from the shorter L T values determined with fluorescent polarization microscopy ( 41 ) , negative stain electron microscopy ( 44 ) , phosphorescence anisotropy ( 40 ) , and molecular dynamics simulations ( 42 ) . Our actin persistence lengths are about two orders of magnitude more rigid than DNA , which has an L T and L B of 100 and 50 nm , respectively ( 52 ) , although these values depend greatly on solution conditions ( e . g . , salt composition and concentration ) . The large uncertainties in the measured rotation angles of indi - vidual actin subunits could contribute to the short L T values deter - mined by electron microscopy ( 44 ) . Uncertainty in subunit rotation angles overestimates the angular fluctuations of adjacent filament subunits and yields an artificially short twisting persistence length . Using an analysis method as in this work ( Methods ) that accounts for these uncertainties in rotational angles yields a larger ( i . e . , stiffer ) torsional persistence length ( Fig . 5 ) . The discrepancy in the filament twisting persistence length values determined by fluorescence polar - ization and phosphorescence anisotropy ( 40 , 41 ) may be due to the independent movement of protein side chains or subunit domains to which the spectroscopic probe is conjugated . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . PNAS 2023 Vol . 120 No . 4 e2208536120 https : / / doi . org / 10 . 1073 / pnas . 2208536120 7 of 12 Actin Filaments Fragment at a Twist Density of ~ 1 to 2 deg sub \u22121 . In our twist \u2013 extension experiments with 0 . 25 pN pulling force , actin filaments fragmented at a twist density of 1 to 2 rot \u03bc m \u22121 ( Figs . 3 B and 7 A ) . A twist density of 1 rot \u03bc m \u22121 is equivalent to ~ 1 deg rotation per actin subunit , which corresponds to a 1 - deg change in relative twist between two laterally adjacent actin subunits and a 2 - deg change in twist between two longitudinally adjacent actin subunits . This twist density introduces only 0 . 65 and 1 . 3 k B T of strain energy ( SI Appendix , Eq . S4 ) at the lateral and longitudinal contacts of actin subunits ( Fig . 7 B ) , respectively , which is considerably less than the estimated bond energies associated with lateral ( 4 to 8 k B T ) and longitudinal ( 12 to 20 k B T ) filament contacts ( 26 ) . Why then do actin filaments fragment at such low twist den - sities ? For an actin filament to fragment , three intersubunit inter - faces\u2014two longitudinal and one lateral\u2014must rupture simultaneously ( 15 , 19 ) . The combined twist strain energy in these three bonds of a filament twisted to a density of ~ 1 deg rotation Fig . 5 . Torsional persistence length of actin filaments determined by electron cryomicroscopy . ( A ) Cartoon schematic of measured filament subunit twisting fluctuations . The top cartoon depicts an actin filament ( gray ) with the average , intrinsic twist ( \u0394\u03c6 1intrinsic ) between adjacent subunits i and i + 1 illustrated as a red curved arrow . The observed twist deviates from the intrinsic twist , either over or under , because of thermal fluctuations . The middle and bottom cartoons illustrate an undertwisted filament ( light gray ) overlaying a canonical filament with an intrinsic twist ( dark gray ) . Blue arrows illustrate the observed twist between subunits i and i + 1 ( Middle ) or i and i + 3 ( Bottom ) , which differs from the intrinsic twist by \u0394\u03c6 \u2019 n ( illustrated by black arrows ) . ( B ) Histograms of actin filament subunit twist fluctuations ( \u0394\u03c6 \u2019 n , in degrees ) estimated from cryo - EM alignment parameters ( reference volume of 5 subunits ) for n = 1 , 10 , and 50 subunits . Red lines represent fits to a normal distribution with mean zero and variance ( \ud835\udf0e 2 obs , n + 1 ) . ( C ) n dependence of the twist variance ( \ud835\udf0e 2 obs , n + 1 ) . The solid red lines represent the best fit to SI Appendix , Eq . S21 . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . 8 of 12 https : / / doi . org / 10 . 1073 / pnas . 2208536120 pnas . org per subunit is only ~ 3 . 2 k B T , more than an order of magnitude lower than the 44 k B T subunit \u22121 activation energy for fragmenta - tion ( 15 ) . Although the imposed twist strain energy does not directly overcome the activation energy for filament fragmenta - tion , it does accelerate the filament fragmentation rate constant ~ twofold ( calculated from exp ( E strain / k B T ) ; SI Appendix , Eq . S28 ) ( 15 ) . While this effect may seem small , it accounts for the rapid fragmentation of twisted filaments observed in our experiments when other factors contributing to severing are considered . Two additional factors that contribute to the rapid fragmen - tation of twisted filaments are the number of potential severing sites and the time duration of the applied twisting load . Filament severing occurs at subunit interfaces , so long filaments have more potential fragmentation sites than shorter ones . That is , the severing reaction is a microscopic process , but observed filament severing is a macroscopic process that scales with the filament length , a collective effect happening at individual sub - units . A typical filament in our experiments is > 10 \u00b5m in length ( > 3 , 700 subunits ) , which means the observed , macroscopic severing rate constant is the microscopic severing rate constant times 3 , 700 . In terms of fragmentation probability , if P is the microscopic fragmentation probability expressed in units subunit \u22121 , the macroscopic probability for fragmentation of a filament that comprised n subunits is given by 1\u2212 ( 1\u2212 P ) n ~ nP + 0 ( P 2 ) ( SI Appendix , Eq . S30 ) . The duration ( \u0394 t ) of applied twist deformation is a second crit - ical factor contributing to the observed fragmentation . Twist loads introduce strain energy and accelerate filament fragmentation according to the Arrhenius equation ( SI Appendix , Eq . S28 ) ( 15 ) from which the twofold acceleration is calculated above . However , the filament fragmentation probability ( P ) scales with the fragmen - tation rate constant ( k frag ) and the duration of the applied load ( \u0394 t ) according to P = 1\u2212 exp ( \u2212 k frag \u0394 t ) ( SI Appendix , Eq . S29 ) ( 15 ) . Therefore , fragmentation will not be significantly affected if the duration of the applied twist is short relative to the characteristic severing time ( \u0394 t \u226a 1 / k frag ) . We derived the filament survival probability as a function of time for a given rate of applied twisting strain ( SI Appendix , Eq . S31 ) . Survival curves , as a function of twist den - sity , were simulated according to SI Appendix , Eq . S31 under two different conditions : 1 ) slow rotation ( ~ 0 . 3 rot s \u22121 or 3 . 2 s for 1 rot ) of a 15 - \u03bc m filament , as carried out in our experiments , and 2 ) rapid rotation ( 2 . 2 rot s \u22121 ) of a short filament ( 0 . 1 \u03bc m ) , as previously modeled ( 15 ) ( Fig . 7 A ) . Long filaments with slow rotation break at ~ 2 rot \u03bc m \u22121 ( duration 96 s ) , whereas the short filaments with fast rotation break at 5 rot \u03bc m \u22121 ( duration 0 . 22 s ) . This behavior explains why filaments do not break under thermal twist fluctuation of ~ 1 deg sub \u22121 ( Fig . 7 A ) and suggests that filaments can undergo rather large structural changes with - out fragmenting , provided the duration of these shape changes are short . Fig . 6 . Twisted cofilactin filaments fragment more easily than twisted bare actin filaments . ( A ) Representative images of twist - induced fragmentation for undertwisted ( UT ) bare , overtwisted ( OT ) bare , undertwisted cofilin saturated , and overtwisted cofilin saturated filaments . ( Scale bar , 4 \u00b5m . ) ( B ) Survival analysis from the experiments in ( A ) at a pulling force of 0 . 03 pN . Log - rank test comparing UT bare to UT cofilin ( P < 0 . 0001 ) and OT bare and OT cofilin ( P = 0 . 0094 ) . Both log - rank tests and Gehan \u2013 Breslow \u2013 Wilcoxon tests yielded similar P values , which conclude that the observed twisting response of bare and cofilin - decorated filaments is statistically different . Fig . 7 . Modeling twist - induced fragmentation of actin filaments . ( A ) Experimental actin filament survival curves for undertwisted ( blue ) and overtwisted ( black ) bare actin filaments at 0 . 25 pN pulling force ( 2 \u00b5L min \u22121 flow rate ) and a twisting rate of \u03c9 = 0 . 3 rot s \u22121 . Data include instances where fragmentation occurs close to the bead or surface interfaces . For comparison , the plot includes simulations of filament survival curves as a function of twist density ( SI Appendix , Eq . S31 ) at the same twist rate of 0 . 3 rot s \u22121 with a length of L = 15 \u00b5m ( red trace ) , as that typical in our twist \u2013 extension experiments in this study , and a twist rate of \u03c9 = 2 . 2 rot s \u22121 with a length L = 0 . 1 \u00b5m ( gray trace ) . Inset image is an example of twist - induced fragmentation . Inset graph is the model - predicted filament torque ( SI Appendix , Eq . S9 ) . ( B ) Model - predicted twisting strain energy per subunit ( left y - axis , SI Appendix , Eq . S4 ) and the relative increase in fragmentation rate constant of strained relative to relaxed , native filaments ( right y - axis , SI Appendix , Eq . S28 ) . Dashed lines indicate the model - predicted twisting strain energy for the twist density imposed at boundaries of human cofilin clusters ( blue ) and by singly isolated bound human cofilin ( red ) . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . PNAS 2023 Vol . 120 No . 4 e2208536120 https : / / doi . org / 10 . 1073 / pnas . 2208536120 9 of 12 Local Twist Strain Drives Filament Severing by Cofilin . The actin filament severing protein , cofilin , binds between two adjacent longitudinal actin subunits and undertwists the filament ~ 4 . 3 deg sub \u22121 ( 54 , 55 ) . The filament twist changes abruptly , within ~ 1 to 2 subunits , at boundaries between bare and cofilin - decorated segments ( 54 , 55 , 60 ) . Filaments preferentially sever at these boundaries ( 45 , 61 , 62 ) within the bare actin side of the boundary ( 60 ) . If we assume that the ~ 4 . 3 - deg twist strain spreads evenly ( 54 ) among the two subunits at the boundary such that each experiences a twist change of ~ 2 . 1 deg , a twist density of ~ 2 . 1 deg sub \u22121 introduces strain energy of ~ 2 . 9 k B T sub \u22121 ( 1 2 L Ts \ud835\udf0e 2 , SI Appendix , Eq . S4 , Fig . 7 B ) . The strain energy of this magnitude is predicted to accelerate the bare actin intrinsic severing rate constant ~ 18 - fold ( exp ( 2 . 9 ) using SI Appendix , Eq . S28 ; Fig . 7 B ) . This value agrees with estimates of a boundary severing rate constant that is ~ 10 to 25 times faster than that of bare actin ( 32 , 35 , 38 ) . It has been reported that boundary severing rate constants vary among cofilin isoforms [ e . g . , Saccharomyces cerevisiae severs more rapidly than human cofilin ( 32 , 35 , 38 , 45 , 61 , 63 , 64 ) ] . Subtle differences in cofilin - induced twist could potentially account for large variability in severing , given the quadratic dependence of the severing rate constant on twist density ( Fig . 7 B ) . We note that the filament twist is constant and does not change within bare and cofilin - decorated segments and therefore does not introduce strain between subunits in those regions as it does at boundaries where twist discontinuities exist ( 62 ) . A singly bound [ i . e . , isolated ( 65 ) ] cofilin also changes the fil - ament twist , although less than at a boundary and only at the one subunit to which it directly binds ( 55 ) . Accordingly , a singly bound cofilin severs filaments but does so more slowly than boundaries ( 32 , 38 ) . Assuming the twist induced by isolated bound cofilin is ~ 1 to 2 deg , the local twist strain should accelerate fragmentation ~ 2 to 18 - fold ( Fig . 7 B ) , consistent with the reported ~ fivefold acceleration ( 32 ) . Cofilin Renders Twisted Actin Filaments Brittle . Cofilactin filaments ( i . e . , saturated with cofilin ) break at lower twist densities than bare actin filaments ( Fig . 6 ) . Several factors can potentially contribute to this mechanical instability . A twisted cofilactin filament could store more elastic strain energy than a bare filament for a given twist load , thereby resulting in more rapid fragmentation . While conceivable , cofilactin filaments are thought to be more compliant in bending ( 16 , 19 , 34 , 42 ) and twisting ( 40 , 42 ) than bare actin filaments so they have less strain energy for an identical shape deformation . It is also possible that cofilactin filaments are more fragile and fragment more easily than bare actin filaments . While this is also conceivable , cofilin bridging interactions stabilize cofi - lactin filaments and protect them from fragmentation ( 15 , 60 ) . These bridging interactions render fully decorated cofilactin filaments comparably stable to bare actin filaments ( 61 , 62 , 65 \u2013 68 ) . A third more likely possibility is that cofilin dissociation from actin filaments transiently introduces a boundary , which frag - ments more easily under twisting ( and bending ) loads ( 15 , 19 ) . Spontaneous cofilin dissociation can introduce this effect but it would be more prominent if twisting loads weakened cofilin binding and / or accelerated cofilin dissociation , as predicted from modeling studies ( 19 ) . Pulling Accelerates Cofilactin Filament Fragmentation . Surface tethering and cross - linking constrain filament bending and twisting , which prevents the dissipation of cofilin - induced torsional strain , thereby enhancing cofilin severing activity ( 15 , 19 , 22 , 67 ) . Long - axis pulling forces on actin filaments , of magnitude comparable with those exerted by myosin motors [ 3 to 5 pN per ATP hydrolyzed ( 69 ) ] , also dampen thermally driven ( Fig . 2 ) and twist - induced filament bending ( Fig . 3 ) and dramatically accelerate filament fragmentation . This behavior supports models in which contractile forces gen - erated by myosin motors rapidly sever twisted filaments such as those with bound cofilin ( 15 , 19 , 28 ) . Recent studies show that contractile forces produced by myosin motors \u201ccatalyze\u201d cofil - in - mediated actin filament disassembly and turnover in Aplysia neurons ( 70 ) and also contribute to actin filament turnover during contractile ring constriction in S . pombe ( 71 ) , demonstrating how combinations of pulling , bending , and twisting forces can dra - matically accelerate actin filament network fragmentation and turnover in cells . Twisting a Filament Bundle . With the torsional persistence length measured in this study , it is possible to make predictions about the behavior of bundled actin filaments . We can model a perfectly cross - linked actin filament bundle as a filament with different dimensions such that the radius of the bundle cross - section ( R ) is determined by the radius of a single actin filament ( r ) and the number ( n ) of filaments comprising the bundle . The area of the cross - section of the bundle is the sum of the cross - sectional areas of each filament forming the bundle ( i . e . , \u03c0 R 2 = n \u03c0 r 2 ) . Therefore , R 2 = nr 2 , and the bundle\u2019s torsional and bending persistence lengths become L T , bundle = G ( \u03c0 R 4 / 2 ) / k B T = n 2 G ( \u03c0 r 4 / 2 ) / k B T = n 2 L T and L B , bundle = E ( \u03c0 R 4 / 4 ) / k B T = n 2 E ( \u03c0 r 4 / 4 ) / k B T = n 2 L B , where G is the shear modulus , and E is the Young\u2019s modulus of an actin filament , respectively . Therefore , a filament bundle\u2019s torsional and bending persistence lengths and corresponding strain energies scale with n 2 . This suggests that twisted bundles are a result of very large applied torques as twisting a three - filament bundle requires nine times as much torque to twist compared with a single filament . The authors of this study ( 12 ) concluded that these torsional loads on actin bundles in filopodia are driven by myosin contractility , indicating the off - axis torques generated by myosin motors ( 72 \u2013 74 ) are sufficiently strong to twist filament bundles . Materials and Methods A brief description of the experimental materials and methods used is provided here , and for more details , see SI Appendix . Protein Purification . Actin was purified from rabbit skeletal muscle and labeled on surface lysines with NHS ester derivatives of Alexa 488 , Alexa 647 , biotin , or digoxigenin ( 17 ) . Alexa 488 phalloidin was purchased from Thermo Fisher ( catalog # A12379 ) . Ca 2 + - actin monomers ( 5 \u00b5M ) were converted to Mg 2 + - actin by addition of 50 \u00b5M MgCl 2 and 0 . 2 mM EGTA and equilibrated for 5 min on ice immediately before use ( 75 ) . Saccharomyces cerevisiae cofilin with a surface - en - gineered cysteine was purified and labeled with Alexa 488 ( 61 ) . Microscope Sample Preparation . Surface functionalization and passiva - tion of microscope coverslips with 2 to 5 % Biotin - labeled PEG - Silane slides were adapted from elsewhere ( 76 ) . Microfluidic chambers were assembled as described ( 17 , 77 ) . Superparamagnetic Dynabeads\u2122 M - 270 Epoxy ( 2 . 8 \u00b5m in diameter , Thermo Fisher catalog # 14301 ) were conjugated to antidigoxigenin antibody following the company - provided protocol . Samples were prepared and experiments carried out in KMI buffer ( 10 mM fluorescence - grade imidazole pH 7 . 0 , 50 mM KCl , 2 mM MgCl 2 , 0 . 2 mM ATP , and 2 mM DTT ) supplemented with 15 mM glucose , 0 . 02 mg mL \u22121 catalase , and 0 . 1 mg mL \u22121 glucose oxidase . Actin polymerization was done as previously D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . 10 of 12 https : / / doi . org / 10 . 1073 / pnas . 2208536120 pnas . org described ( 65 ) . Filament \u2013 bead conjugation in sample chambers is described in detail in SI Appendix . Microscopy . Imaging was conducted on a Till iMic digital total internal reflection fluorescence ( TIRF ) microscope equipped with a 100\u00d7 objective ( Olympus ) and an Andor iXon897 electron - multiplying charge - coupled device ( EMCCD ) camera ( 17 , 33 , 38 ) . Filament Force \u2013 Extension . In the force \u2013 extension experiments without twist , filament end \u2010 to \u2010 end length ( R ) was measured as a function of the ten - sile force ( f ) applied with buffer flow . Images were recorded at each unique buffer flow ( tensile force ) with an acquisition rate of 1 frame s \u22121 . The positions of the paramagnetic beads were tracked with the TrackMate plugin in ImageJ ( NIH , USA ) ( 78 ) . The value of R was determined as the direct straight - line distance between the two attachment points at the bead and surface . A fila - ment contour length ( L ) was measured as R at a high flow rate to make the filament straight . The measured force dependence of R ( i . e . , force \u2013 extension curve ) was fitted to the following equation describing the force \u2013 extension behavior of semiflexible polymers ( 6 , 47 ) : [ 1 ] R = L \u2212\u0394 L = L \u239b\u239c\u239c\u239d 1 \u2212 1 2 \ufffd k B T L B f \u239b\u239c\u239c\u239d coth \ufffd L \ufffd f L B k B T \ufffd \u2212 1 L \ufffd L B k B T f \u239e\u239f\u239f\u23a0\u239e\u239f\u239f\u23a0 = \u23a7\u23aa\u23aa\u23a8\u23aa\u23aa\u23a9 L \u2192 0 \u2192 L L \u2192 \u221e \u2192 L \u239b\u239c\u239c\u239d 1 \u2212 1 2 \ufffd k B T L B f \u239e\u239f\u239f\u23a0 , where L B is the filament bending persistence length , k B is Boltzmann con - stant , and T is the room temperature ( 296 K ) . \u0394 L is the deviation of R from L of a semiflexible polymer due to thermally driven random bending , and it does not scale with L linearly . For the polymers with very short L , \u0394 L ~ 0 , the deviation per L is negligible and R ~ L , whereas with very long L , the deviation of R from L reaches the maximum per unit length 1 2 \u221a k B T L B f . This nonlinear L dependence of \u0394 L is consistent with the conclusion by another study ( 79 ) , which claims a popular force \u2013 extension equation ( 80 ) with \u0394 L L = const is overly simplified . Using Origin software ( Originlab , Northampton , MA , USA ) , experimental rep - licates were globally fitted with L B as a global fitting parameter and L unique to each individual filament dataset ( Fig . 2 B ) . In addition to globally fitting individual filament force \u2013 extension curves , we compiled and fitted data for filaments of different lengths by averaging the nor - malized filament end \u2010 to - end length ( R / L ) , which is given as follows : [ 2 ] \ufffd R L \ufffd = 1 n n \ufffd i = 1 R i L i = 1 \u2212 1 2 \ufffd k B T L B f \u239b \u239c\u239c \u239c \u239d 1 + 2 n n \ufffd i = 1 e \u2212 2 L i \ufffd f LBkBT 1 \u2212 e \u2212 2 L i \ufffd f LBkBT \u2212 \ufffd L B k B T f 1 n n \ufffd i = 1 1 L i \u239e\u239f \u239f \u23a0 , \u223c 1 \u2212 1 2 \u221a k B T L B f + k B T 2 f ( 1 L ) , because the term 2 e \u2212 2 L i \u221a f LBkBT 1 \u2212 e \u2212 2 L i \u221a f LBkBT is < 0 . 08 and can be ignored since it is \u226a 1 and more than 1 order of magnitude smaller than \u221a L B k B T f 1 L i in our experimental force and filament length ranges . Filament Twist \u2013 Extension . Filament images in the twist \u2013 extension exper - iments were recorded with an acquisition rate of 1 frame s \u22121 while twisting filament at a constant rate of 0 . 31 rot s \u22121 and applying a given tensile force ( f ) with hydrodynamic flow . The position of the paramagnetic beads was tracked with the same procedure , and filament end \u2010 to \u2010 end distance ( R ) and contour length ( L ) were determined in the same manner as those in the preceding Force \u2013 Extension section . R / L is a function of twist density , \u03c3 ( unit : rot \u03bc m \u22121 ) , and L . Filament contour lengths in our experiments range from 7 to 20 \u03bc m . Normalized R / L data were averaged by binning according to the filament twist density ( rot \u03bc m \u22121 ) with a bin size of 0 . 025 rot \u00b5m \u22121 . To describe the twist \u2013 extension behavior of actin filaments , an analytical two - state model developed for DNA ( 52 ) was modified for actin filaments . In the model , it is assumed that a filament subunit exists in one of two states when a filament is twisted : plectonemic or nonplectonemic ( linear ) . A filament with a plectoneme consists of a mixture of plectonemic and nonplectonemic regions . The fraction of plectonemic regions increases linearly with twist density and does not contribute to R / L . Eq . 3 ( SI Appendix , Eq . S10 ) gives the force depend - ence of R / L contributed from only the fraction in the nonplectonemic state ( 52 ) ( SI Appendix ) as follows : [ 3 ] R L = \u23a7\u23aa\u23aa\u23aa\u23aa\u23aa\u23aa\u23aa\u23aa\u23a8\u23aa\u23aa\u23aa\u23aa\u23aa\u23aa\u23aa\u23aa\u23a9 1 \u2212 1 2 \ufffd k B T L B f + k B T 2 f \ufffd 1 L \ufffd \u2212 \ud835\udf14 02 \ud835\udf0e 2 \ufffd 4 L B L T \ufffd fL B k B T + 1 \ufffd 2 \ufffd k B TL B 3 f , \ufffd \ud835\udf0e \ufffd < \ufffd\ufffd \ud835\udf0e s \ufffd\ufffd \ufffd\ufffd\ufffd \ud835\udf0e p \ufffd\ufffd\ufffd \u2212 \ufffd \ud835\udf0e \ufffd \ufffd\ufffd\ufffd \ud835\udf0e p \ufffd\ufffd\ufffd \u2212 \ufffd\ufffd \ud835\udf0e s \ufffd\ufffd \u239b\u239c\u239c\u239d 1 \u2212 1 2 \ufffd k B T L B f + k B T 2 f \ufffd 1 L \ufffd \u2212 \ud835\udf14 02 \ud835\udf0e s 2 \ufffd 4 L B L T \ufffd fL B k B T + 1 \ufffd 2 \ufffd k B TL B 3 f \u239e\u239f\u239f\u239f\u23a0 , \ufffd\ufffd \ud835\udf0e s \ufffd\ufffd \u2264 \ufffd \ud835\udf0e \ufffd < \ufffd\ufffd \ud835\udf0e P \ufffd\ufffd 0 , \ufffd\ufffd\ufffd \ud835\udf0e p \ufffd\ufffd\ufffd \u2264 \ufffd \ud835\udf0e \ufffd . In the equation , \u03c3 s is a special applied twist density at which filaments form plectonemes , and \u03c3 p is applied twist density when the entire filament is plec - tonemic . The fraction of subunits in the nonplectonemic conformation is given by | | | \ud835\udf0e p | | | \u2212 | \ud835\udf0e | | | | \ud835\udf0e p | | | \u2212 | | \ud835\udf0e s | | . A custom routine was written with Origin software to globally fit the exper - imental data of R / L as a function of applied twist at different pulling forces to Eq . 3 ( Fig . 3 ) with L B and L T as shared unconstrained parameters . \u03c3 s for pulling force of 0 . 01 and 0 . 03 pN were fixed parameters from averaging experimentally observed values , but the other \u03c3 s for pulling force of 0 . 25 pN and value of \u03c3 p for all forces were also unconstrained during the fitting procedure but were not shared because they depend on and vary with the pulling force . The value of ( 1 L ) was constrained during fitting to the values calculated from the actual filament contour lengths measured under a high pulling force ( to straighten ) . Design of Freely Orbiting Magnetic Tweezers . The rotational fluctuations of actin filaments were measured with freely orbiting magnetic tweezers ( 57 ) . Five cylindrical magnets ( R422 - N52 , K & J Magnetics ) arranged in series were mounted on a linear XY micrometer stage ( XR25 , Thorlabs ) and centered directly above the microscope objective while visualizing with wide - field microscopy . The procedure for conjugating antidigoxigenin paramagnetic beads to digoxigenin marker beads for these experiments is given in detail in SI Appendix . Analysis of Filament Rotational Fluctuations . In whole filament twisting fluctu - ation experiments ( Fig . 4 A ) , we monitor how the angle \u03b8 ( t ) between the two filament D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . PNAS 2023 Vol . 120 No . 4 e2208536120 https : / / doi . org / 10 . 1073 / pnas . 2208536120 11 of 12 attachment points ( at the surface and on the bead , length L ) changes over time , and the variance ( \u03c3 \u03b8 2 ) of the angle fluctuation is given by Eq . 4 ( SI Appendix , Eq . S22 ) : [ 4 ] \ud835\udf0e 2 \ud835\udf03 = < ( \ud835\udf03 ( t ) \u2212 < \ud835\udf03 ( t ) > ) 2 > t \u2192 \u221e = L L T + \ud835\udf0e 2 \ud835\udf00 \u223c L L T . Here , \u03c3 \u03b5 2 is the variance of the uniform randomly distributed noise in a relative twist angle between two adjacent filament subunits , and we neglect it because it is much smaller than L / L T ( SI Appendix ) . Eq . 4 indicates that at times sufficiently long to sample the equilibrium distribution , indicated by plateau of the variance ( Fig . 4 E ) , the variance of the angle fluctuations at equilibrium is inversely proportional to L T , and scales with L linearly ( SI Appendix , Fig . S6 ) , which has been demonstrated for actin filaments by experimental observation ( 20 , 39 ) . Analysis of Filament Subunit Angular Fluctuations from cryo - EM Data . From the electron micrographs of actin filaments , which were classified as described previously ( 54 ) , we defined \u0394 s as the length of an actin subunit and \u03a6 as the rotational Euler angle around the filament centerline . Any rotation angle difference ( twist ) between two subunits spaced n subunits apart ( SI Appendix , Eq . S20 ) was calculated by the sum of all the observed relative twists between 2 adjacent subunits and it displays a Gaussian distribution with a variance that scales linearly with n ( SI Appendix , Eq . S21 ) . Fitting SI Appendix , Eq . S21 to the n dependence of the observed variance ( \ud835\udf0e 2 obs , n + 1 ) yields \u0394 s / L T from the slope of the best linear fit of the data and the noise variance ( \ud835\udf0e 2 \ud835\udf00 ) from the y intercept . The variance of the angle distribution was extracted from the data using three different analysis procedures . In one approach , the histogram of \u0394 \u03a6 I , obs , n + 1 for every value of n was independently fitted to a normal distribution , yielding \u03c3 2 i , obs , n + 1 . Then , \u03c3 2 i , obs , n + 1 as a function of n was fitted to SI Appendix , Eq . S21 . In addition , in maximum likelihood estimation ( MLE ) , each variance with spacing of n was directly calculated from the square of the SD as follows : [ 5 ] \ud835\udf0e 2 obs , n + 1 = \u27e8\ufffd \u0394\ufffd i , obs , n + 1 \u2212 \u27e8 \u0394 \ud835\udf03 obs , n + 1 \u27e9\ufffd 2 \u27e9 . The n dependence of \u03c3 2 i , obs , n + 1 was then fitted to SI Appendix , Eq . S21 . In the third method , we applied Bayesian inference using Markov chain Monte Carlo ( MCMC ; Metropolis \u2013 Hastings algorithm ) to sample the pos - terior probability distribution of the true variance ( 81 ) . The method was coded in language R ( www . r - project . org ) , and the most likely L T and \ud835\udf0e 2 \ud835\udf00 values were determined from the peaks of their probability distributions . These three methods of analyses were repeated for the rotation Euler angle \u03a6 determined from a refinement volume of 1 subunit instead of 5 with comparable results ( Table 1 ) ( 54 , 55 ) . Data , Materials , and Software Availability . All study data are included in the article and / or SI Appendix . ACKNOWLEDGMENTS . This research was supported by the NIH through R35 - GM136656 ( awarded to E . M . D . L . C . ) , J . P . B . was supported in part by the Department of Defense Army Research Office through a multidisciplinary uni - versity research initiative grant W911NF1410403 ( awarded to E . M . D . L . C . ) . S . G . was supported by R35GM136656 - S1 . C . V . S . was supported by R01 GM 110530 . 1 . P . A . Janmey , C . A . McCulloch , Cell mechanics : Integrating cell responses to mechanical stimuli . Annu . Rev . Biomed Eng . 9 , 1 \u2013 34 ( 2007 ) . 2 . I . Schoen , B . L . Pruitt , V . Vogel , The Yin - Yang of rigidity sensing : How forces and mechanical properties regulate the cellular response to materials . Annu . Rev . Mater Res . 43 , 589 \u2013 618 ( 2013 ) . 3 . T . D . Pollard , J . A . Cooper , Actin , a central player in cell shape and movement . Science 326 , 1208 \u2013 1212 ( 2009 ) . 4 . J . Howard , Mechanics of Motor Proteins and the Cytoskeleton ( Sinauer Associates , Publishers , Sunderland , Mass . , 2001 ) , p . xvi , 367 p . 5 . F . Gittes , B . Mickey , J . Nettleton , J . Howard , Flexural rigidity of microtubules and actin filaments measured from thermal fluctuations in shape . J . Cell Biol . 120 , 923 \u2013 934 ( 1993 ) . 6 . F . C . Mackintosh , J . Kas , P . A . Janmey , Elasticity of semiflexible biopolymer networks . Phys . Rev . Lett . 75 , 4425 \u2013 4428 ( 1995 ) . 7 . J . D . Winkelman , C . A . Anderson , C . Suarez , D . R . Kovar , M . L . Gardel , Evolutionarily diverse LIM domain - containing proteins bind stressed actin filaments through a conserved mechanism . Proc . Natl . Acad . Sci . U . S . A . 117 , 25532 \u2013 25542 ( 2020 ) . 8 . X . Y . Sun et al . , Mechanosensing through direct binding of tensed F - actin by LIM domains . Dev . Cell 55 , 468 \u2013 482 ( 2020 ) . 9 . M . Akamatsu et al . , Principles of self - organization and load adaptation by the actin cytoskeleton during clathrin - mediated endocytosis . Elife 9 , e49840 ( 2020 ) . 10 . E . S . Welf et al . , Actin - membrane release initiates cell protrusions . Dev . Cell 55 , 723 \u2013 736 ( 2020 ) . 11 . Y . H . Tee et al . , Cellular chirality arising from the self - organization of the actin cytoskeleton . Nat . Cell Biol . 17 , 445 \u2013 457 ( 2015 ) . 12 . N . Leijnse et al . , Filopodia rotate and coil by actively generating twist in their actin shaft . Nat . Commun . 13 , 1636 ( 2022 ) . 13 . N . A . Medeiros , D . T . Burnette , P . Forscher , Myosin II functions in actin - bundle turnover in neuronal growth cones . Nat . Cell Biol . 8 , 215 \u2013 226 ( 2006 ) . 14 . S . H . Parekh , O . Chaudhuri , J . A . Theriot , D . A . Fletcher , Loading history determines the velocity of actin - network growth . Nat . Cell Biol . 7 , 1219 \u2013 1223 ( 2005 ) . 15 . A . C . Schramm , G . M . Hocky , G . A . Voth , J . L . Martiel , E . M . De La Cruz , Plastic deformation and fragmentation of strained actin filaments . Biophys J . 117 , 453 \u2013 463 ( 2019 ) . 16 . E . M . De La Cruz , J . Roland , B . R . McCullough , L . Blanchoin , J . L . Martiel , Origin of twist - bend coupling in actin filaments . Biophys . J . 99 , 1852 \u2013 1860 ( 2010 ) . 17 . N . G . Pandit et al . , Force and phosphate release from Arp2 / 3 complex promote dissociation of actin filament branches . Proc . Natl . Acad . Sci . U . S . A . 117 , 13519 \u2013 13528 ( 2020 ) . 18 . V . I . Risca et al . , Actin filament curvature biases branching direction . Proc . Natl . Acad . Sci . U . S . A . 109 , 2913 \u2013 2918 ( 2012 ) . 19 . A . C . Schramm et al . , Actin filament strain promotes severing and cofilin dissociation . Biophys . J . 112 , 2624 \u2013 2633 ( 2017 ) . 20 . Y . Tsuda , H . Yasutake , A . Ishijima , T . Yanagida , Torsional rigidity of single actin filaments and actin - actin bond breaking force under torsion measured directly by in vitro micromanipulation . Proc . Natl . Acad . Sci . U . S . A . 93 , 12937 \u2013 12942 ( 1996 ) . 21 . M . P . Murrell , M . L . Gardel , F - actin buckling coordinates contractility and severing in a biomimetic actomyosin cortex . Proc . Natl . Acad . Sci . U . S . A . 109 , 20820 \u2013 20825 ( 2012 ) . 22 . H . Wioland , A . Jegou , G . Romet - Lemonne , Torsional stress generated by ADF / cofilin on cross - linked actin filaments boosts their severing . Proc . Natl . Acad . Sci . U . S . A . 116 , 2595 \u2013 2602 ( 2019 ) . 23 . Y . Arai et al . , Tying a molecular knot with optical tweezers . Nature 399 , 446 \u2013 448 ( 1999 ) . 24 . P . Bieling et al . , Force feedback controls motor activity and mechanical properties of self - assembling branched actin networks . Cell 164 , 115 \u2013 127 ( 2016 ) . 25 . J . Mueller et al . , Load adaptation of lamellipodial actin networks . Cell 171 , 188 \u2013 200 . e16 ( 2017 ) . 26 . A . Jegou , G . Romet - Lemonne , The many implications of actin filament helicity . Semin . Cell Dev . Biol . 102 , 65 \u2013 72 ( 2020 ) . 27 . L . D . Landau , E . M . Lifshitz , Theory of Elasticity ( Pergamon Press , Oxford , New York , 1986 ) . 28 . E . M . De La Cruz , J . L . Martiel , L . Blanchoin , Mechanical heterogeneity favors fragmentation of strained actin filaments . Biophys . J . 108 , 2270 \u2013 2281 ( 2015 ) . 29 . A . Mogilner , G . Oster , Force generation by actin polymerization II : The elastic ratchet and tethered filaments . Biophys . J . 84 , 1591 \u2013 1605 ( 2003 ) . 30 . A . Davison et al . , Formin is associated with left - right asymmetry in the pond snail and the frog . Curr . Biol . 26 , 654 \u2013 660 ( 2016 ) . 31 . S . R . Naganathan , S . Furthauer , M . Nishikawa , F . Julicher , S . W . Grill , Active torque generation by the actomyosin cell cortex drives left - right symmetry breaking . Elife 3 , e04165 ( 2014 ) . 32 . W . A . Elam et al . , Phosphomimetic S3D cofilin binds but only weakly severs actin filaments . J . Biol . Chem . 292 , 19565 \u2013 19579 ( 2017 ) . 33 . H . R . Kang et al . , Identification of cation - binding sites on actin that drive polymerization and modulate bending stiffness . Proc . Natl . Acad . Sci . U . S . A . 109 , 16923 \u2013 16927 ( 2012 ) . 34 . B . R . McCullough , L . Blanchoin , J . L . Martiel , E . M . De la Cruz , Cofilin increases the bending flexibility of actin filaments : Implications for severing and cell mechanics . J . Mol . Biol . 381 , 550 \u2013 558 ( 2008 ) . 35 . B . R . McCullough et al . , Cofilin - linked changes in actin filament flexibility promote severing . Biophys . J . 101 , 151 \u2013 159 ( 2011 ) . 36 . H . Kojima , A . Ishijima , T . Yanagida , Direct measurement of stiffness of single actin - filaments with and without tropomyosin by in - vitro nanomanipulation . Proc . Natl . Acad . Sci . U . S . A . 91 , 12962 \u2013 12966 ( 1994 ) . 37 . H . Isambert et al . , Flexibility of actin - filaments derived from thermal fluctuations \u2013 effect of bound nucleotide , phalloidin , and muscle regulatory proteins . J . Biol . Chem . 270 , 11437 \u2013 11444 ( 1995 ) . 38 . H . Kang et al . , Site - specific cation release drives actin filament severing by vertebrate cofilin . Proc . Natl . Acad . Sci . U . S . A . 111 , 17821 \u2013 17826 ( 2014 ) . 39 . R . Yasuda , H . Miyata , K . Kinosita Jr . , Direct measurement of the torsional rigidity of single actin filaments . J . Mol . Biol . 263 , 227 \u2013 236 ( 1996 ) . 40 . E . Prochniewicz , N . Janson , D . D . Thomas , E . M . De la Cruz , Cofilin increases the torsional flexibility and dynamics of actin filaments . J . Mol . Biol . 353 , 990 \u2013 1000 ( 2005 ) . 41 . J . N . Forkey , M . E . Quinlan , Y . E . Goldman , Measurement of single macromolecule orientation by total internal reflection fluorescence polarization microscopy . Biophys . J . 89 , 1261 \u2013 1271 ( 2005 ) . 42 . J . Fan et al . , Molecular origins of cofilin - linked changes in actin filament mechanics . J . Mol . Biol . 425 , 1225 \u2013 1240 ( 2013 ) . 43 . T . Fujii , A . H . Iwane , T . Yanagida , K . Namba , Direct visualization of secondary structures of F - actin by electron cryomicroscopy . Nature 467 , 724 \u2013 728 ( 2010 ) . 44 . E . H . Egelman , D . J . Derosier , Image - analysis shows that variations in actin crossover spacings are random , not compensatory . Biophys . J . 63 , 1299 \u2013 1305 ( 1992 ) . 45 . J . P . Bibeau , S . Gray , E . M . De La Cruz , Clusters of a few bound cofilins sever actin filaments . J . Mol . Biol . 433 , 166833 ( 2021 ) . 46 . N . Courtemanche , J . Y . Lee , T . D . Pollard , E . C . Greene , Tension modulates actin filament polymerization mediated by formin and profilin . Proc . Natl . Acad . Sci . U . S . A . 110 , 9752 \u2013 9757 ( 2013 ) . 47 . J . S . Palmer , M . C . Boyce , Constitutive modeling of the stress - strain behavior of F - actin filament networks . Acta Biomater . 4 , 597 \u2013 612 ( 2008 ) . 48 . J . S . Graham et al . , Multi - platform compatible software for analysis of polymer bending mechanics . PLoS One 9 , e94766 ( 2014 ) . 49 . H . Isambert , A . C . Maggs , Dynamics and rheology of actin solutions . Macromolecules 29 , 1036 \u2013 1040 ( 1996 ) . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 . 12 of 12 https : / / doi . org / 10 . 1073 / pnas . 2208536120 pnas . org 50 . M . J . Greenberg , C . L . Wang , W . Lehman , J . R . Moore , Modulation of actin mechanics by caldesmon and tropomyosin . Cell Motil . Cytoskeleton 65 , 156 \u2013 164 ( 2008 ) . 51 . M . T . van Loenhout , M . V . de Grunt , C . Dekker , Dynamics of DNA supercoils . Science 338 , 94 \u2013 97 ( 2012 ) . 52 . J . F . Marko , Torque and dynamics of linking number relaxation in stretched supercoiled DNA . Phys . Rev . E Stat . Nonlin . Soft Matter Phys . 76 , 021926 ( 2007 ) . 53 . T . Sanchez , I . M . Kulic , Z . Dogic , Circularization , photomechanical switching , and a supercoiling transition of actin filaments . Phys . Rev . Lett . 104 , 098103 ( 2010 ) . 54 . A . Huehn et al . , The actin filament twist changes abruptly at boundaries between bare and cofilin - decorated segments . J . Biol . Chem . 293 , 5377 \u2013 5383 ( 2018 ) . 55 . A . R . Huehn et al . , Structures of cofilin - induced structural changes reveal local and asymmetric perturbations of actin filaments . Proc . Natl . Acad . Sci . U . S . A . 117 , 1478 \u2013 1484 ( 2020 ) . 56 . J . D . Moroz , P . Nelson , Entropic elasticity of twist - storing polymers . Macromolecules 31 , 6333 \u2013 6347 ( 1998 ) . 57 . J . Lipfert , M . Wiggin , J . W . Kerssemakers , F . Pedaci , N . H . Dekker , Freely orbiting magnetic tweezers to directly monitor changes in the twist of nucleic acids . Nat . Commun . 2 , 439 ( 2011 ) . 58 . A . J . Hunt , J . Howard , Kinesin swivels to permit microtubule movement in any direction . Proc . Natl . Acad . Sci . U . S . A . 90 , 11653 \u2013 11657 ( 1993 ) . 59 . E . M . De La Cruz , M . L . Gardel , Actin mechanics and fragmentation . J . Biol . Chem . 290 , 17137 \u2013 17144 ( 2015 ) . 60 . G . M . Hocky , C . V . Sindelar , W . Cao , G . A . Voth , E . M . De La Cruz , Structural basis of fast - and slow - severing actin \u2013 cofilactin boundaries . J . Biol . Chem . 296 , 100337 ( 2021 ) . 61 . C . Suarez et al . , Cofilin tunes the nucleotide state of actin filaments and severs at bare and decorated segment boundaries ( vol 21 , pg 862 , 2011 ) . Curr . Biol . 21 , 862 \u2013 868 ( 2011 ) . 62 . E . M . De La Cruz , How cofilin severs an actin filament . Biophys . Rev . 1 , 51 \u2013 59 ( 2009 ) . 63 . H . Wioland et al . , ADF / cofilin accelerates actin dynamics by severing filaments and promoting their depolymerization at both ends . Curr . Biol . 27 , 1956 \u2013 1967 ( 2017 ) . 64 . K . Hayakawa , S . Sakakibara , M . Sokabe , H . Tatsumi , Single - molecule imaging and kinetic analysis of cooperative cofilin - actin filament interactions . Proc . Natl . Acad . Sci . U . S . A . 111 , 9810 \u2013 9815 ( 2014 ) . 65 . E . M . De La Cruz , Cofilin binding to muscle and non - muscle actin filaments : Isoform - dependent cooperative interactions . J . Mol . Biol . 346 , 557 \u2013 564 ( 2005 ) . 66 . E . Andrianantoandro , T . D . Pollard , Mechanism of actin filament turnover by severing and nucleation at different concentrations of ADF / cofilin . Mol . Cell 24 , 13 \u2013 23 ( 2006 ) . 67 . D . Pavlov , A . Muhlrad , J . Cooper , M . Wear , E . Reisler , Actin filament severing by cofilin . J . Mol . Biol . 365 , 1350 \u2013 1358 ( 2007 ) . 68 . W . A . Elam , H . Kang , E . M . De La Cruz , Biophysics of actin filament severing by cofilin . FEBS Lett . 587 , 1215 \u2013 1219 ( 2013 ) . 69 . E . M . De La Cruz , E . M . Ostap , Relating biochemistry and function in the myosin superfamily . Curr . Opin . Cell Biol . 16 , 61 \u2013 67 ( 2004 ) . 70 . X . F . Zhang et al . , Regulation of axon growth by myosin II - dependent mechanocatalysis of cofilin activity . J . Cell Biol . 218 , 2329 \u2013 2349 ( 2019 ) . 71 . M . Malla , T . D . Pollard , Q . Chen , Counting actin in contractile rings reveals novel contributions of cofilin and type II myosins to fission yeast cytokinesis . Mol . Biol . Cell 33 , mbcE21080376 ( 2021 ) . 72 . G . Lebreton et al . , Molecular to organismal chirality is induced by the conserved myosin 1D . Science 362 , 949 \u2013 952 ( 2018 ) . 73 . J . F . Beausang , H . W . Schroeder III , P . C . Nelson , Y . E . Goldman , Twirling of actin by myosins II and V observed via polarized TIRF in a modified gliding assay . Biophys . J . 95 , 5820 \u2013 5831 ( 2008 ) . 74 . S . Pyrpassopoulos , E . A . Feeser , J . N . Mazerik , M . J . Tyska , E . M . Ostap , Membrane - bound myo1c powers asymmetric motility of actin filaments . Curr . Biol . 22 , 1688 \u2013 1692 ( 2012 ) . 75 . E . M . De La Cruz , T . D . Pollard , Nucleotide - free actin : Stabilization by sucrose and nucleotide binding kinetics . Biochemistry 34 , 5452 \u2013 5461 ( 1995 ) . 76 . Y . Gidi , S . Bayram , C . J . Ablenas , A . S . Blum , G . Cosa , Efficient one - step PEG - silane passivation of glass surfaces for single - molecule fluorescence studies . ACS Appl Mater Interfaces 10 , 39505 \u2013 39511 ( 2018 ) . 77 . E . M . Johnson - Chavarria , U . Agrawal , M . Tanyeri , T . E . Kuhlman , C . M . Schroeder , Automated single cell microbioreactor for monitoring intracellular dynamics and cell growth in free solution . Lab . Chip 14 , 2688 \u2013 2697 ( 2014 ) . 78 . J . Y . Tinevez et al . , TrackMate : An open and extensible platform for single - particle tracking . Methods 115 , 80 \u2013 90 ( 2017 ) . 79 . Y . Seol , J . Li , P . C . Nelson , T . T . Perkins , M . D . Betterton , Elasticity of short DNA molecules : Theory and experiment for contour lengths of 0 . 6 - 7 microm . Biophys . J . 93 , 4360 \u2013 4373 ( 2007 ) . 80 . J . F . Marko , E . D . Siggia , Stretching DNA . Macromolecules 28 , 8759 \u2013 8770 ( 1995 ) . 81 . M . K . Cowles , Applied Bayesian statistics : With R and OpenBUGS Examples ( Springer , New York , 2013 ) , p . xiv , 232 pages . D o w n l o a d e d fr o m h tt p s : / / www . pn a s . o r g by \" UN I V E R S I T Y O F W A S H I NG T ON L I BR A R I E S , A RC S - S E R I A L S \" on M a r c h 13 , 2024 fr o m I P a dd r e ss 205 . 175 . 106 . 171 .",
    "lundmark2008gtpaseactivating": "Report The GTPase - Activating Protein GRAF1 Regulates the CLIC / GEEC Endocytic Pathway Richard Lundmark , 1 , 3 , 4 , * Gary J . Doherty , 1 , 3 Mark T . Howes , 2 Katia Cortese , 2 Yvonne Vallis , 1 Robert G . Parton , 2 and Harvey T . McMahon 1 , * 1 Medical Research Council Laboratory of Molecular Biology Hills Road Cambridge , CB2 0QH UK 2 Institute for Molecular Bioscience and Centre for Microscopy and Microanalysis University of Queensland Brisbane , Queensland 4072 Australia Summary Clathrin - independent endocytosis is an umbrella term for a variety of endocytic pathways that internalize numerous cargoes independently of the canonical coat protein Clathrin [ 1 , 2 ] . Electron - microscopy studies have de\ufb01ned the pleio - morphic CL athrin - I ndependent C arriers ( CLICs ) and G PI - E nriched E ndocytic C ompartments ( GEECs ) as related ma - jor players in such uptake [ 3 , 4 ] . This CLIC / GEEC pathway relies upon cellular signaling and activation through small G proteins , but mechanistic insight into the biogenesis of its tubular and tubulovesicular carriers is lacking . Here we show that the Rho - GAP - domain - containing protein GRAF1 marks , and is indispensable for , a major Clathrin - indepen - dent endocytic pathway . This pathway is characterized by its ability to internalize bacterial exotoxins , GPI - linked pro - teins , and extracellular \ufb02uid . We show that GRAF1 localizes to PtdIns ( 4 , 5 ) P2 - enriched , tubular , and punctate lipid struc - tures via N - terminal BAR and PH domains . These membrane carriers are relatively devoid of caveolin1 and \ufb02otillin1 but are associated with activity of the small G protein Cdc42 . This study provides the \ufb01rst speci\ufb01c noncargo marker for CLIC / GEEC endocytic membranes and demonstrates how GRAF1 can coordinate small G protein signaling and mem - brane remodeling to facilitate internalization of CLIC / GEEC pathway cargoes . Results and Discussion The protein G TPase R egulator A ssociated with F ocal Adhe - sion Kinase - 1 ( GRAF1 ) is predicted to comprise an N - terminal BAR domain , a PH domain , a RhoGAP domain , a proline - rich domain , and a C - terminal SH3 domain ( Figure 1A ) . GRAF1 ex - hibits GAP activity for the small G proteins RhoA and Cdc42 and has been shown to interact with the kinases FAK and PKN b [ 5 , 6 , 14 ] . The presence of a predicted BAR domain in GRAF1 suggests that it might function in membrane sculpting [ 7 ] . GRAF1 was found to be expressed in a variety of cell lines ( Figure S1A ) , and by immunocytochemistry we found that GRAF1 was predominantly localized to pleiomorphic tubular and punctate structures in HeLa and NIH 3T3 cells ( Figures 1B \u2013 1D and Figure S1B ) . Although GRAF1 can be found primar - ily on long tubules in some cells , other cells within the same population exhibit a predominantly punctate GRAF1 localiza - tion . GRAF1 - positive tubules are disrupted at low tempera - tures , and a 37 (cid:2) C \ufb01xation was required for their integrity to be preserved ( Figure S1C ) . Upon monitoring both GRAF1 lo - calization and endocytosis of either the plasma membrane marker DiI or the \ufb02uid - phase marker dextran by confocal mi - croscopy , we found that endocytic structures extensively co - localized with GRAF1 ( Figures 1B and 1C ) . Colocalization of GRAF1 was even evident when only very early endocytic struc - tures were examined , indicative of an early endocytic role for GRAF1 - positive membranes ( Figure 1D ) . We then examined the turnover of GRAF1 - positive membranes with time by over - expressing GFP - tagged full - length GRAF1 in HeLa cells and examining its localization by using four - dimensional spin - ning - disc confocal microscopy . We found GRAF1 - positive tu - bular structures to be spectacularly dynamic ( Figure 1E and Movie S1 ) . By electron microscopy , we found that GRAF1 - labeled tubules were around 40 nm in diameter in vivo ( Figure 1G ) . We also found that overexpressed GFP - tagged GRAF1 BAR + PH protein ( missing the GAP , proline - rich , and SH3 domains ) also labeled tubular membranes that could accumulate DiI ( Figure 1F , Movie S2 , and data not shown ) . However , these tubules were much more static than those la - beled by overexpressed full - length GRAF1 , suggesting that GRAF1 BAR + PH might act in a dominant - negative manner to stabilize early endocytic tubules ( see also below ) . To examine the speci\ufb01c properties of the predicted lipid - binding region of GRAF1 , we performed lipid cosedimentation assays with puri\ufb01ed GRAF1 BAR + PH protein ( Figure 2A ) . Us - ing liposomes of varying diameter , we found that GRAF1 BAR + PH bound better to smaller ( more highly curved ) lipo - somes , consistent with the presence of a membrane - curva - ture - sensing BAR domain [ 7 ] . Furthermore , we examined the effects of mutating key lysine residues [ 7 ] in the BAR domain to glutamates ( KK131 / 132EE ) and found that this mutant was now cytoplasmically distributed ( Figure S1D ) . Expression of the BAR domain alone in cells resulted in a predominantly cy - toplasmic and sometimes punctate localization ( Figure S1E ) , suggesting that both the BAR and PH domains are necessary for tubular localization of GRAF1 . We therefore tested whether , in addition to a curvature - sensing or - generating capability , the GRAF1 BAR + PH unit has phosphoinositide - binding speci\ufb01city . GRAF1 BAR + PH protein was subjected to lipid cosedimentation assays with 10 % phosphatidylserine - con - taining liposomes with varying phosphoinositide composition ( Figure 2B ) . Greatest binding was observed for PtdIns ( 4 , 5 ) P2 - enriched liposomes . PtdIns ( 4 , 5 ) P2 is plasma - membrane en - riched [ 8 ] , consistent with our observations that GRAF1 - asso - ciated traf\ufb01cking occurs from this site . GRAF1 BAR + PH protein was also capable of generating tubules in vitro from spherical liposomes as examined by electron microscopy * Correspondence : richard . lundmark @ medchem . umu . se ( R . L . ) , hmm @ mrc - lmb . cam . ac . uk ( H . T . M . ) 3 These authors contributed equally to this work 4 Present address : Department of Medical Biochemistry and Biophysics , Umea\u02da University , 901 87 Umea\u02da , Sweden Current Biology 18 , 1802 \u2013 1808 , November 25 , 2008 \u00aa 2008 Elsevier Ltd . Open access under CC BY license . DOI 10 . 1016 / j . cub . 2008 . 10 . 044 under negative - staining conditions ( Figure 2C ) . The diameter of these tubules was around 40 nm , consistent with the ob - served diameter of GRAF1 - positive tubules found in cells ( Figure 1G ) . Taken together , these data strongly suggest that the BAR and PH domains of GRAF1 function together to pro - duce and / or stabilize endocytic tubules in vivo . To identify interacting partners for GRAF1 , we immunopre - cipitated GRAF1 from rat brain cytosol . Interestingly , GRAF1 was found in a complex with the membrane scission protein Dynamin1 ( Figure 2D and S2A ) . We con\ufb01rmed binding to both Dynamin1 and Dynamin2 by using GST - tagged GRAF1 SH3 domain as bait in pull - down experiments against puri\ufb01ed Dynamin1 , brain cytosol , or HeLa cell lysate ( Figure 2E ; also Figures S2B and S2C ) . Monomeric GRAF1 SH3 domain was found to bind to a peptide from Dynamin1 proline - rich domain with a k d of 106 m M ( Figure S2D ) . Furthermore , treatment of HeLa cells with dynasore resulted in a profound reduction of tubular endocytosis in HeLa cells and a redistribution of GRAF1 to basally located puncta ( Figure S2E ) . These data , taken together with our observation that C - terminally trun - cated versions of GRAF1 ( lacking the SH3 domain ) are found on long static tubules , suggests that the complex between Dynamin and GRAF1 might function to regulate the scission and stability of these ordinarily pleiomorphic tubulovesicular structures . Clathrin polymers are rarely found on tubular membranes by electron microscopical techniques . Indeed , GRAF1 - positive tubules , and other GRAF1 - positive structures , were devoid of Clathrin and did not colocalize with internalized or internal - izing transferrin or the transferrin receptor , which is widely used as a marker for Clathrin - mediated endocytic events ( Fig - ures S3A \u2013 S3C ) . Alongside the canonical Clathrin - mediated en - docytic routes , other prevalent endocytic pathways coexist [ 1 , 2 ] . For example , the internalization of MHC class I proteins , many GPI - linked receptors , and bacterial exotoxins rely on traf\ufb01cking compartments with heretofore unde\ufb01ned coat com - ponents . The data presented above strongly suggest that GRAF1 might function as a coat within a prevalent Clathrin - independent endocytic pathway . To characterize further the nature of GRAF1 - positive tubules , we incubated GRAF1 - over - expressing HeLa cells with cholera toxin B subunit ( CTxB ) , a marker that is used for the study of Clathrin - independent en - docytic pathways but that also enters via Clathrin - mediated routes [ 9 ] . We found that internalized CTxB colocalized with Myc - GRAF1 - positive tubules in HeLa cells ( Figure 2F ) as well as with endogenous GRAF1 - positive structures after 5 min in - cubation at 37 (cid:2) C ( Figure 2G ) . In addition , we found that GRAF1 BAR + PH overexpression ( which leads to static tubules ; Figure 1F ) halved the number of cells internalizing CTxB with - out effecting transferrin uptake ( Figures S4A \u2013 S4C ) . We then followed CTxB and transferrin internalization in real time in NIH 3T3 cells ( almost all of which are capable of binding and internalizing CTxB ; cf . HeLa cells , which often lack its glyco - lipid receptor , GM1 ) . NIH 3T3 cells were transfected with constructs encoding either GFP - tagged GRAF1 or GRAF1 BAR + PH and incubated with CTxB and transferrin on ice be - fore stimulation of endocytosis by incubation of cells with pre - warmed ( 37 (cid:2) C ) media . Imaging commenced immediately ( Figure S4D and Movie S3 ) . Newly forming dynamic GRAF1 - positive tubules were shown to contain CTxB at extremely early stages of its internalization and were unassociated with internalizing / internalized transferrin . Furthermore , the more static , elongated GRAF1 BAR + PH - positive tubules maintained the presence of CTxB , again independently of transferrin , suggesting that here the toxin is trapped in structures that can - not undergo \ufb01ssion from the plasma membrane and progress further ( data not shown ) . Although GRAF1 - positive carriers resemble the previously characterized Arf6 - dependent structures that are responsible for uptake of MHC Class I proteins [ 10 ] , we found no evidence for overlap between these pathways . Neither GRAF1 - nor GRAF1 BAR + PH - positive tubules were found to contain endo - cytosed or recycled MHC class I proteins at 5 , 15 , or 45 min af - ter stimulation of their internalization when speci\ufb01c antibodies were used , and the amount of total MHC class I protein endo - cytosis was not altered upon overexpression of the dominant - negative GRAF1 BAR + PH construct ( Figures S6A and S6B and data not shown ) . In addition , the tubular localization of GRAF1 was independent of overexpression of wild - type Arf6 or dom - inant - active Arf6 ( Arf6Q67L ; data not shown ) . In contrast , Arf6 Q67L overexpression was found to completely block the inter - nalization of MHC class I molecules ( [ 11 ] and data not shown ) . Electron - microscopy studies , focusing on the uptake of CTxB and GPI - linked receptors , have suggested that a large portion of these proteins are internalized by Cl athrin - i ndepen - dent c arriers ( CLICs ) seemingly into G PI - AP - enriched e arly e n - dosomal c ompartments ( GEECs ) , together with \ufb02uid - phase markers such as dextran [ 4 ] . Our data have shown that GRAF1 - positive structures morphologically and functionally resemble the endocytic structures of this CLIC / GEEC endo - cytic pathway [ 4 ] . We therefore examined whether GRAF1 could regulate the internalization of a model GPI - linked protein known to undergo endocytosis into the CLIC / GEEC pathway . Consistent with this , we found by using immuno - electron microscopy that GRAF1 labeled GFP - GPI - positive tubules ( Figure 3A ) . We then examined the internalization of GFP - GPI in mouse embryonic \ufb01broblasts by binding anti - GFP anti - bodies and transferrin ( as a control ) to cells on ice and subse - quently moving these cells to 37 (cid:2) C for variable chase periods . We did this in both wild - type ( data not shown ) and caveolin1 - knockout mouse embryonic \ufb01broblasts ( to exclude caveolae - associated uptake ) that we transfected with either a Myc - tagged GRAF1 or a Myc - tagged GRAF1 BAR + PH construct and examined after cytosol washout ( Figure 3B ; note that washout treatment damages GRAF1 - positive membranes but is necessary for the objective quantitative analysis of co - localization of GRAF1 with internalized GFP - GPI because cy - toplasmic GRAF1 would otherwise result in overestimation of colocalization ) . GRAF1 was found to signi\ufb01cantly colocalize with internalized GFP - GPI ( 79 % colocalization ) but not trans - ferrin after a 2 min chase ( Figures 3B and 3C and data not shown ) . A reduced level of colocalization was seen after a 10 min chase ( 47 % ) , and even lower colocalization levels were observed after a 40 min chase ( when the protein is recy - cling back to the plasma membrane ; 23 % ) . These data are consistent with our previous observations , strongly suggest - ing a role for GRAF1 function in sculpting the highly curved membranes of the CLIC / GEEC endocytic pathway . In contrast , GRAF1 BAR + PH consistently colocalized with GFP - GPI after 2 min ( 60 % ) , 10 min ( 68 % ) , and 40 min ( 75 % ) chase periods , further supporting a role for this protein as a dominant - negative protein that traps early CLIC / GEEC carriers . To char - acterize this dominant - negative effect in greater detail , we examined the effect of GRAF1 BAR + PH protein on GFP - GPI in - ternalization ( Figures 3B and 3C ) . Compared with GFP - GPI in - ternalization in the presence of GRAF1 , the total amount of GFP - GPI internalization was profoundly reduced in the pres - ence of GRAF1 BAR + PH ( to around 10 % of control levels ) , Membrane Sculpting in Clathrin - Independent Endocytosis 1803 Figure 1 . GRAF1 Tubules Are Highly Dynamic and Mark a Prevalent Endocytic Pathway ( A ) Domain architecture of GRAF1 and the site of introduced BAR domain mutations ( * ) . ( B \u2013 D ) Micrographs showing that GRAF1 - positive tubules are derived from the plasma membrane as shown by colabeling with the membrane dye DiI after 5 min ( B ) or internalized dextran after either 5 min ( C ) or 1 min ( D ) of incubation . ( E ) Overlaid maximum projections of spinning - disc confocal micrographs of HeLa cells expressing GFP - tagged GRAF1 demonstrate that GRAF1 - positive tubules are completely turned over in 10 min . Z sections were performed continuously for 10 min in 0 . 5 m m steps . The initial maximum projection ( in green ) was then merged with the \ufb01nal maximum projection ( in red ) . See also Movie S1 . ( F ) Overlaid maximum projections as in ( a ) , but for a cell overexpressing GFP - tagged GRAF1 BAR + PH . See also Movie S2 . Current Biology Vol 18 No 22 1804 ( G ) Electron micrographs prepared as described and immunolabeled for GRAF1 ( 10 nm gold particles ) . Note the GRAF1 - positive tubular structures . Dis - tance between arrowhead tips = 40 nm . Scale bars represent 10 m m . Figure 2 . The BAR and PH Domains Localize GRAF1 to Highly Curved , PtdIns ( 4 , 5 ) P2 - Enriched Membranes , and the SH3 Domain Binds Dynamin ( A ) Coomassie - stained gel of liposome cosedimentationassay showing the preference of GST - tagged GRAF1 BAR + PH protein for binding to smaller - sized liposomes derived from total brain lipids ( the average diameters of liposomes are shown ) . Pellet ( P ) and supernatant ( S ) fractions were separated by ultra - centrifugation . The graph shows quanti\ufb01cations of total band intensities for each condition normalized to binding of the non - curvature - sensitive protein Dab2 ( as a way of controlling for total lipid in each experiment ) . The error bars show 95 % con\ufb01dence intervals ( calculated by t tests ) for each condition . ( B ) Liposome cosedimentation assay as performed in ( A ) but with 0 . 8 - m m - diameter liposomes enriched with varying phosphoinositides . ( C ) Electron micrographs of negatively stained liposomes incubated in the presence or absence of GST - tagged GRAF1 BAR + PH protein . Note the protein - dependent presence of tubular structures . The scale bar represents 200 nm . ( D ) Immunoprecipitation of GRAF1 from rat brain cytosol ( via Ab2 ) reveals a GRAF1 - Dynamin1 complex as identi\ufb01ed by mass spectrometry and con\ufb01rmed by immunoblot . ( E ) Coomassie - stained gels of pull - down experiments with puri\ufb01ed Dynamin and either bead - bound GST - tagged GRAF1 / Amphiphysin SH3 domain or GRAF1 BAR + PH protein . P = pellet fraction , S = supernatant fraction . ( F and G ) Epi\ufb02uorescence micrographs of HeLa cell overexpressing Myc - tagged GRAF1 and incubated with CTxB for 5 min before \ufb01xation and staining . ( G ) Epi\ufb02uorescencemicrographsofaHeLacellincubatedwithCTxBfor5minbefore\ufb01xationandstainingforendogenousGRAF1 . Scalebarsrepresent10 m m . Membrane Sculpting in Clathrin - Independent Endocytosis 1805 Figure 3 . GRAF1 Regulates the CLIC Endocytic Pathway ( A ) Representativeelectron micrographsof 65nm ultrathin cryosections of HeLa cellstransientlytransfectedwith GFP - GPI aloneorwithboth GFP - GPI and Myc - tagged GRAF1 FL . Cells were \ufb01xed in 2 % PFA with 0 . 2 % glutaraldehyde and labeled with anti - GFP and anti - Myc antibodies . Protein A 10 nm gold was used for revealing the GFP in the single labeling ( upper image ) . As shown , GFP - GPI was found in uncoated vesicles and tubules within the cell . Double la - beling of GRAF1 ( Protein A gold 10 nm ) and GFP - GPI ( Protein A gold 15 nm ) is shown in the bottom image , where colocalization of GRAF1 and GFP - GPI is seen in an intracellular tubule . Arrowheads point to 10 nm gold particles . Due to limitations of the Tokuyasu method , it is not possible to discriminate cell surface - connected tubules from intracellular tubules . ( B ) Caveolin1knockoutMEFswere cotransfected withGFP - GPI ( green ) andeitherMyc - tagged GRAF1 ( upper row ; red ) orGRAF1 BAR + PH ( lower row ; red ) . GFP antibodies ( blue ) were bound to cells for 30 min on ice prior to induction of internalization at 37 (cid:2) C and 5 % CO 2 for the indicated times . Surface labeling was removed , and cytosol extraction followed ( see Supplemental Experimental Procedures ) . Panels on the left side of each image show GRAF1 proteins ( top ) , internalized anti - GFP ( middle ) , and a merged , triple - labeled image ( bottom ) . Arrows indicate points of colocalization . ( C ) MicrographsofCaveolin1knockoutMEFsco - overexpressingGRAF1andGFP - GPI , afteranti - GFPinternalization , acidstripping , andcytosolextraction , were processed with Volocity 3 . 7 . 0 so that the percentage colocalization could be calculated . The histogram shows the number of anti - GFP pixels that co - localize with GRAF1 relative to all anti - GFP pixels . Five to seven images across three independent experiments were taken after 2 , 10 , and 40 min of inter - nalization . Error bars indicate the standard errors of the mean . Images were also processed in Adobe Photoshop CS2 so that the number of anti - GFPpixels relative to total pixels within the image could be calculated . The histogram represents standardized values of anti - GFP pixels . Error bars indicate the stan - dard errors of the mean . Current Biology Vol 18 No 22 1806 suggesting further that this protein traps cargoes in rather static early endocytic carriers that are incapable of undergoing \ufb01ssion from the plasma membrane . Importantly , the steady - state surface levels of GFP - GPI were indistinguishable be - tween cells expressing GRAF1 and those expressing GRAF1 BAR + PH ( Figure S6C ) . CTxB has been shown to be endocytosed via caveolin and \ufb02otillin - positive structures [ 12 ] . Although we did not \ufb01nd cav - eolin1 or \ufb02otillin1 in GRAF1 - positive structures at steady state by \ufb02uorescence microscopy ( Figures S5A and S5B ) , stabiliza - tion of early GRAF1 - positive carriers by GRAF1 BAR + PH over - expression caused \ufb02otillin1 , and to a lesser extent caveolin1 , to be found in such regions with CTxB ( Figures S5C and S5D ) , suggesting that these membrane regions can communi - cate with the CLIC / GEEC pathway . Uptake through the CLIC / GEEC endocytic pathway has pre - viously been shown to be Cdc42 dependent [ 3 ] . Consistent with this , we observed that overexpressed dominant - active Cdc42 protein colocalizes with GRAF1 in basally located puncta ( Figure 3D ) . The absence of Cdc42 in long tubular structures might suggest that the role of Cdc42 in this pathway is restricted spatially and temporarily to the earlier endocytic stages or that progression is blocked through the dominant - active construct , which would be expected to have pleiotropic effects . To determine whether GRAF1 was necessary for endocyto - sis , we depleted GRAF1 levels by using siRNA . This treatment was capable of reducing the amount of GRAF1 to background levels as assessed by immunocytochemistry and immunoblot - ting of HeLa cells and their lysates , respectively ( Figure 4A ; also Figure S6D ) . GRAF1 - depleted cells were assessed for their ability to endocytose dextran ( allowing assessment of total endocytic capacity ) , both by epi\ufb02uorescence microscopy and by a quantitative \ufb02uorimetric assay ( Figures 4B and 4C ) . GRAF1 depletion resulted in a 50 % \u2013 60 % reduction of total dextran endocytosis , similar to that observed upon depletion of AP2 , the major Clathrin adaptor at the plasma membrane ( Figure 4B ) , suggesting that GRAF1 - mediated endocytosis and Clathrin - mediated endocytosis account for roughly equal amounts of volume internalization in these cells . GRAF1 - depleted cells had no observable defects in their ability to en - docytose transferrin ( Figure 4D ) , demonstrating that GRAF1 is not necessary for Clathrin - mediated endocytosis . In this study we have shown that GRAF1 is found on tubular and vesicular membranes in vivo , and that these membranes de\ufb01ne a prevalent Clathrin - independent endocytic pathway . This endocytic pathway is capable of internalizing Cholera toxin , GPI - linked proteins , and large amounts of extracellular \ufb02uid , and it corresponds ( at least in part ) to the CLIC / GEEC en - docytic pathway . Activity of Arf1 , in addition to that of its GEF ( D ) Confocal micrographs of HeLa cells co - overexpressing GFP - tagged Cdc42 L61 ( dominant - active ) and Myc - tagged GRAF1 showing colocalization in discrete puncta and short tubules . Scale bars represent 10 m m . Figure 4 . GRAF1 Is Indispensable for Clathrin - Independent Fluid - Phase Uptake in Fibroblasts ( A ) Immunoblots on HeLa celllysates transfected withacontrolsiRNA oreither oftwosiRNAs directedagainstGRAF1 mRNA ( siRNAs aandb ) . Detectionof GRAF1 and tubulin ( loading control ) was performed with speci\ufb01c antibodies on lysates obtained 48 hr after transfection . ( B ) GRAF1 - depletedcellsshowamajorreductionin\ufb02uid - phaseendocytosisasshownbythedecreaseintheuptakeofFITC - labeleddextran ( controlsiRNA ( n = 5 ) , siRNAa ( n = 8 ) , or AP2 siRNA ( n = 3 ) ) . The error bars show the standard deviation of the mean . ( C and D ) HeLa cells depleted of GRAF1 were incubated with dextran ( C ) or transferrin ( D ) for 15 min before \ufb01xation and staining . Scale bars represent 10 m m . Membrane Sculpting in Clathrin - Independent Endocytosis 1807 ARHGAP10 , is necessary for uptake via the CLIC / GEEC path - way , and Arf1 activity modulation also interferes with Cdc42 dynamics , suggesting interplay between these proteins [ 13 ] . Indeed , GRAF1 is capable of downregulating the activity of Cdc42 [ 14 ] , which is known to be necessary for the CLIC / GEEC pathway [ 3 ] and which colocalizes with GRAF1 in vivo . The protein machinery responsible for the biogenesis and pro - cessing of CLIC / GEEC endocytic membranes is largely un - known . Although Dynamin certainly plays a role in CLIC / GEEC membrane processing ( it is necessary for delivery of CTxB to the Golgi apparatus [ 4 ] ) , its precise role in the CLIC / GEEC pathway is uncertain . Recent microinjection experi - ments with antibodies directed against speci\ufb01c isoforms of Dynamin , or with siRNAs directed against these , showed that Dynamin is involved in constitutive \ufb02uid - phase uptake and suggested that it might play an important role in high - vol - ume Clathrin - independent endocytic events [ 15 ] . We build on this work by showing that there exists a tight biochemical in - teraction between GRAF1 and Dynamin . Although we impli - cate Dynamin in this pathway , precisely determining whether it functions here in an analogous manner to Dynamin at Cla - thrin - coated pits or works via a noncanonical mechanism will require further work . Our experiments lead us to suggest that GRAF1 coordinates the highest volume Clathrin - independent endocytic pathway in HeLa cells . Although several Clathrin - independent endo - cytic pathways appear to coexist ( such pathways include cav - eolin1 - and \ufb02otillin1 - positive pathways , an Arf6 - associated tu - bular uptake pathway , and the CLIC / GEEC pathway ) , the precise contributions that each of these pathways makes to constitutive endocytosis , and the interrelationships between these pathways , are uncertain . Although we have found that GRAF1 - positive membranes contain little \ufb02otillin1 or caveolin1 at steady state , we have also shown that these membranes probably transiently communicate . We further show that GRAF1 - positive membranes are not associated with internal - izing or recycled MHC class I proteins . Taken together , our data link biochemical and cell biological observations to iden - tify the \ufb01rst important modulator of membrane curvature in a highly prevalent Clathrin - independent endocytic pathway . We believe that , in the context of our \ufb01ndings , study of Cla - thrin - independent endocytic pathways will reveal new layers of complexity underlying lipid and protein cargo selection , ap - propriate intracellular traf\ufb01cking of endocytic membranes , and the mechanisms of endocytic intermediate formation by mem - brane deformation and membrane - curvature stabilization . Supplemental Data Supplemental Data include Supplemental Experimental Procedures , six \ufb01g - ures , and three movies and are available with this article online at http : / / www . current - biology . com / supplemental / S0960 - 9822 ( 08 ) 01413 - 9 . Acknowledgments Many thanks to Sew Peak - Chew and Farida Begum for assistance with mass spectrometry analysis , Marijn Ford for providing puri\ufb01ed Dynamin , and Sven Carlsson and all members of the McMahon lab for help and sup - port . We also thank the Kempe and Magn Bergvallfoundations and Medical faculty Umea\u02da University for grants . Thanks to M . Fa\u00a8llman , B Nichols , J . W . G . Andersson and S . Mayor for providing reagents . R . L was supported by a postdoctoral fellowship from the Swedish Research Council . G . D was supported by a Trinity College Cambridge Internal Graduate Studentship and Research Scholarship . The main work was founded by Medical Re - search Council UK . Received : April 7 , 2008 Revised : October 4 , 2008 Accepted : October 8 , 2008 Published online : November 24 , 2008 References 1 . Mayor , S . , and Pagano , R . E . ( 2007 ) . Pathways of clathrin - independent endocytosis . Nat . Rev . Mol . Cell Biol . 8 , 603 \u2013 612 . 2 . Johannes , L . , andLamaze , C . ( 2002 ) . Clathrin - dependentornot : isitstill the question ? Traf\ufb01c 3 , 443 \u2013 451 . 3 . Sabharanjak , S . , Sharma , P . , Parton , R . G . , and Mayor , S . ( 2002 ) . GPI - anchored proteins are delivered to recycling endosomes via a distinct cdc42 - regulated , clathrin - independent pinocytic pathway . Dev . Cell 2 , 411 \u2013 423 . 4 . Kirkham , M . , Fujita , A . , Chadda , R . , Nixon , S . J . , Kurzchalia , T . V . , Sharma , D . K . , Pagano , R . E . , Hancock , J . F . , Mayor , S . , and Parton , R . G . ( 2005 ) . Ultrastructural identi\ufb01cation ofuncoatedcaveolin - indepen - dent early endocytic vehicles . J . Cell Biol . 168 , 465 \u2013 476 . 5 . Hildebrand , J . D . , Taylor , J . M . , andParsons , J . T . ( 1996 ) . AnSH3domain - containing GTPase - activating protein for Rho and Cdc42 associates with focal adhesion kinase . Mol . Cell . Biol . 16 , 3169 \u2013 3178 . 6 . Shibata , H . , Oishi , K . , Yamagiwa , A . , Matsumoto , M . , Mukai , H . , and Ono , Y . ( 2001 ) . PKNbeta interacts with the SH3 domains of Graf and a novel Graf related protein , Graf2 , which are GTPase activating pro - teins for Rho family . J . Biochem . ( Tokyo ) 130 , 23 \u2013 31 . 7 . Peter , B . J . , Kent , H . M . , Mills , I . G . , Vallis , Y . , Butler , P . J . , Evans , P . R . , and McMahon , H . T . ( 2004 ) . BAR domains as sensors of membrane curva - ture : the amphiphysin BAR structure . Science 303 , 495 \u2013 499 . 8 . Di Paolo , G . , and De Camilli , P . ( 2006 ) . Phosphoinositides in cell regula - tion and membrane dynamics . Nature 443 , 651 \u2013 657 . 9 . Torgersen , M . L . , Skretting , G . , van Deurs , B . , and Sandvig , K . ( 2001 ) . Internalization of cholera toxin by different endocytic mechanisms . J . Cell Sci . 114 , 3737 \u2013 3747 . 10 . Naslavsky , N . , Weigert , R . , andDonaldson , J . G . ( 2004 ) . Characterization of a nonclathrin endocytic pathway : membrane cargo and lipid require - ments . Mol . Biol . Cell 15 , 3542 \u2013 3552 . 11 . Caplan , S . , Naslavsky , N . , Hartnell , L . M . , Lodge , R . , Polishchuk , R . S . , Donaldson , J . G . , andBonifacino , J . S . ( 2002 ) . AtubularEHD1 - containing compartment involved in the recycling of major histocompatibility com - plex class I molecules to the plasma membrane . EMBO J . 21 , 2557 \u2013 2567 . 12 . Glebov , O . O . , Bright , N . A . , and Nichols , B . J . ( 2006 ) . Flotillin - 1 de\ufb01nes a clathrin - independent endocytic pathway in mammalian cells . Nat . Cell Biol . 8 , 46 \u2013 54 . 13 . Kumari , S . , and Mayor , S . ( 2008 ) . ARF1 is directly involved in dynamin - independent endocytosis . Nat . Cell Biol . 10 , 30 \u2013 41 . 14 . Taylor , J . M . , Macklem , M . M . , and Parsons , J . T . ( 1999 ) . Cytoskeletal changes induced by GRAF , the GTPase regulator associated with focal adhesion kinase , are mediated by Rho . J . Cell Sci . 112 , 231 \u2013 242 . 15 . Cao , H . , Chen , J . , Awoniyi , M . , Henley , J . R . , and McNiven , M . A . ( 2007 ) . Dynamin 2 mediates \ufb02uid - phase micropinocytosis in epithelial cells . J . Cell Sci . 120 , 4167 \u2013 4177 . Current Biology Vol 18 No 22 1808",
    "ezzatSpecificityAbstractionExamples2020": "H I C H A M E Z Z A T M A R I N E A G O G U (cid:1) E P A S C A L L E M A S S O N B E N O I T W E I L M A T H I E U C A S S O T T I Speci\ufb01city and Abstraction of Examples : Opposite Effects on Fixation for Creative Ideation ABSTRACT Fixation is one of the major obstacles that individuals face in creative idea generation contexts . Several studies have shown that individuals unintentionally tend to \ufb01xate to the examples they are shown in a cre - ative ideation task , even when instructed to avoid them . Most of these studies used examples formulated with high level of speci\ufb01city . However , no study has examined individuals\u2019 creative performance under an instruction to diverge from given examples , when these examples are formulated with a high level of abstraction . In the present study , we show that ( a ) instructing participants to avoid using common examples when formulated with a high level of speci\ufb01city increases \ufb01xation ; whereas ( b ) instructing participants to avoid such examples while using a more abstract level for stating these common examples \u2014 such as a cate - gorization of these examples \u2014 mitigates \ufb01xation and doubles the number of creative ideas generated . These \ufb01ndings give new insights on the key role of categorization in creative ideation contexts . Keywords : \ufb01xation effect , idea generation , examples , categorization . Creativity has been described as the capacity to generate ideas that are both novel and useful ( Amabile , 1996 ) , and constitutes one of the key cognitive skills that individuals use daily in various contexts . Accord - ing to divergent thinking scholars ( Guilford , 1959 ) , creative individuals usually exhibit high levels of idea - tional \ufb02uency , \ufb02exibility , and originality ( Runco & Acar , 2012 ) . However , the process of creative ideation is not always an easy task , and could be spontaneously and intuitively constrained by previously acquired knowledge , by individuals\u2019 own mental models or even by earlier ideas generated . Indeed , several studies have shown that these various factors could constrain one\u2019s cognitive ability to generate novel and creative ideas ( Agogu (cid:1) e , Kazakc (cid:1) i , et al . , 2014 ; Jansson & Smith , 1991 ; Smith , Ward & Schumacher , 1993 ) . Psycholo - gists have labeled this phenomenon as \ufb01xation ( Cassotti , Agogu (cid:1) e , Camarda , Houd (cid:1) e & Borst , 2016 ; Jansson & Smith , 1991 ) , aka the blind mental adherence to a set of ideas ( Cassotti , Camarda , Poirel , Houd (cid:1) e & Agogu (cid:1) e , 2016 ) . One classical task to illustrate \ufb01xation is called the \u201ctwo cords problem\u201d of Norman Maier ( Maier , 1931 ) . Two cords ( tied to the ceiling ) and pair of pliers were provided to participants , who were asked to tie together the free ends of these two cords . In this experiment , most participants were \ufb01xated on the knowledge they had of pliers , and were unable to look at it alternatively by simply tying the pliers to one of the cords to form a pendulum that will swing to reach the second one . In fact , the accumulated knowledge about the typical use of pliers was spontaneously activated in individuals mind , and prevented them from seeing alternative uses ( Maier , 1931 ) , and thus impending creative thinking . Such effect , characterized in this case in a problem - solving situation , has been demonstrated to occur as well in more openly framed prob - lems , where participants need to generate a lot of different ideas to a speci\ufb01c situation ( Agogu (cid:1) e , Poirel , Pineau , Houd (cid:1) e & Cassotti , 2014 ) . More speci\ufb01cally , multiple experimental psychology studies have showed that previously acquired knowl - edge in individuals\u2019 mind is most likely to act as a mental set promoting \ufb01xation ( Adamson , 1952 ; Sio , Kotovsky & Cagan , 2015 ) . Moreover , other studies related to the role of examples on creativity have shown indeed that exposure to examples could play either a positive or negative role to modulate \ufb01xation . Smith and colleagues demonstrated that designers unintentionally tend to conform to the features of the example The Journal of Creative Behavior , Vol . 54 , Iss . 1 , pp . 115 \u2013 122 \u00a9 2018 by the Creative Education Foundation , Inc . (cid:1) DOI : 10 . 1002 / jocb . 349 115 they were shown ( Smith et al . , 1993 ) . In three experiments using a creative ideation paradigm in which sub - jects had to imagine and sketch new exemplars of experimenter - de\ufb01ned categories , the authors showed that individuals surprisingly replicate the features of the example even when they are explicitly asked to propose ideas that are different from the given example . Deepening our understanding of the role - speci\ufb01c examples may play in creative thinking , Agogu (cid:1) e and colleagues showed that two types of examples could actually have opposite effects on \ufb01xation modulation ( Agogu (cid:1) e , Kazakc (cid:1) i , et al . , 2014 ) . In a creative ideation task where participants had to generate creative ideas to ensure that a hen\u2019s egg dropped from a height of ten meters does not break , the authors experimentally demonstrated that participants exposed to a common example ( within the \ufb01xation ) increased the number of solutions within the \ufb01xation ; whereas participants exposed to a creative example ( outside the \ufb01xation ) decreased the number of solutions within the \ufb01xation , and consequently increased originality of solutions . Focusing on the way instructions to be creative are delivered to participants , other studies have provided discrepant results regarding the role of warnings and constraints on \ufb01xation modulation . Chrysikou and Weisberg ( 2005 ) showed that the \ufb01xation effect can be reduced using instructions which warn subjects about the use of the \ufb02aws of an example ( Chrysikou & Weisberg , 2005 ) . On the contrary , Viswanathan and Linsey ( 2013 ) con\ufb01rmed Smith and colleagues\u2019 \ufb01ndings ( 1993 ) , and demonstrated that even when designers are presented with warnings about the undesirable example features along with the reasons for those warnings , \ufb01xation to those features is not mitigated ( Viswanathan & Linsey , 2013 ) . These discrepant results could let one think at \ufb01rst glance that there could be a certain relationship between the level of abstraction of the examples and the level of \ufb01xation mitigation . In fact , these \ufb01ndings may suggest that the more the con - straints and warnings on examples are abstract ( without clear indications about the features of the exam - ples ) , the more \ufb01xation would be mitigated . Similarly , one could also think that the more constraints on examples are speci\ufb01c ( with clear speci\ufb01cation of the features of the examples ) , the more \ufb01xation would be reinforced . In consistency with this role of speci\ufb01city and abstraction on \ufb01xation modulation in creative ideation , Baughman and Mumford ( 1995 ) demonstrated that when people are asked to combine exemplars from sep - arate categories to form a single inclusive category , which is considered an act of categorization , participants become more original ( Baughman & Mumford , 1995 ) . These results provided interesting indications about the role of abstraction for \ufb01xation mitigation and the generation of creative ideas . Moreover , Ward and col - leagues ( Ward , Patterson & Sifonis , 2004 ) con\ufb01rmed these \ufb01ndings on the role of abstraction for creativity . In their classical experiments where participants were asked to imagine life on other planets , the authors demonstrated that instructions encouraging participants to formulate the given task in more abstract ways led to more creativity than instructions encouraging participants to formulate the task in very speci\ufb01c ways . However , the in\ufb02uence of the level of abstraction on creativity has been only solely explored through the perspective of the task formulation . More focus could be placed on the relationship between the level of speci\ufb01city and abstraction , and the level of \ufb01xation mitigation by applying it to the use of examples . To assess this relationship between the level of abstraction and \ufb01xation mitigation \u2014 and in line with the abovementioned \ufb01ndings \u2014 the aim of the present study was to examine the effects of instructions warning about using either common examples formulated with a high level of speci\ufb01city , or more abstract levels of these same examples . To achieve this aim , participants were asked to solve a creative task ( i . e . , the hen\u2019s egg task ) and were provided with instructions warning about using common examples \u2014 either at a high level of speci\ufb01city or at a more abstract level \u2014 at the beginning of the task . We reasoned that if the in\ufb02uence of examples on creative ideation depends on the level of abstraction , then instructing individuals to avoid using common examples \u2014 at a high level of speci\ufb01city \u2014 should reinforce \ufb01xation ; whereas instructing individuals to avoid using common examples \u2014 by framing those examples at a more abstract level , such as a categoriza - tion of these examples \u2014 should mitigate \ufb01xation . METHOD PARTICIPANTS Seventy - \ufb01ve participants of an introductory course of innovation design were recruited for this study . To ensure that the content of the course had no in\ufb02uence on the performance of the participants , the experi - ment was made at the very beginning of the course . Participants were engineering students and professionals ( 91 % of the subjects were engineering students , while 9 % were professionals ) . Participants ( 69 % men ) were between 19 and 58 years old ( M = 25 . 6 years , SD = 6 . 9 ) . Participants were randomly assigned to one of the three experimental groups : a control group without constraints ( n = 25 , M = 24 . 84 years , SD = 5 . 51 , 18 116 Effects of Examples on Creative Ideation 21626057 , 2020 , 1 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / j o c b . 349 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 09 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e men ) , a group \u201cSpeci\ufb01c\u201d with constraints on examples ( n = 25 , M = 25 . 72 years , SD = 5 . 95 , 20 men ) and a group \u201cAbstract\u201d with constraints on categories of examples ( n = 25 , M = 26 . 36 , SD = 9 , 14 men ) . ANOVA and chi - squared analyses indicated that the mean ages ( F ( 2 , 72 ) < 1 ) and gender distributions ( v 2 = 3 . 51 , p = . 17 ) did not differ signi\ufb01cantly between the groups . Sample size was determined pre - hoc by running an a priori power analysis using G * Power 3 . 1 . 9 . 2 ( Faul , Erdfelder , Buchner & Lang , 2009 ) , revealing that a minimum of 66 participants would be needed to detect a medium effect size of 0 . 20 ( according to Cohen\u2019s effect size conventions ) , with a power ( 1 - b ) set at . 80 and an a set at . 05 . \u201d DESIGN AND PROCEDURE Each participant was given 10 minutes to perform individually a creativity task where the aim was to propose the maximum number of original solutions to \u201censure that a hen\u2019s egg dropped from a height of ten meters does not break\u201d . Individuals had to write down on a sheet of paper all the solutions they could come up with , and were prohibited to talk with each other . Not only the hen\u2019s egg task could be the appro - priate creativity task to select since it does not require speci\ufb01c knowledge and expertise from the partici - pants , but as we have mentioned earlier , previous studies of Agogu (cid:1) e and colleagues ( Agogu (cid:1) e , Kazakc (cid:1) i , et al . , 2014 ) used Concept - Knowledge theory ( Hatchuel & Weil , 2002 ) , to build a cartography of solutions of the hen\u2019s egg task , distinguishing between solutions labeled \ufb01xation ( common or less novel ideas based on the most accessible knowledge ) and solutions labeled expansion ( rare or more novel ideas based on less accessi - ble knowledge outside \ufb01xation ) . In fact , over the past 5 years , the authors demonstrated that 81 % of the solutions generated by participants to this task were \ufb01xated around three categories of solutions ( which are \u201cdamping the shock\u201d , \u201cprotecting the egg\u201d , and \u201cslowing the fall\u201d ) . These three categories were considered to be inside \ufb01xation . However , only 19 % of the solutions were outside \ufb01xation ( for instance : \u201cbefore and after the fall\u201d , \u201cwith a living device\u201d , \u201cusing intrinsic properties of the environment\u201d , etc . . . ) . This database comprises a total of 716 proposed solutions from 122 students ( from the Faculty of Psychology of Paris Descartes University ) who performed the hen\u2019s egg task in two previous studies ( Agogu (cid:1) e , Kazakc (cid:1) i , et al . , 2014 ; Agogu (cid:1) e , Poirel , et al . , 2014 ) , and is updated regularly with new solutions and categories of solutions if participants come up with new solutions that do not \ufb01t with the current categories . Table 1 shows a list of the range of categories of solutions of the hen\u2019s egg task . We used these works to delimitate the frontier between what is inside \ufb01xation ( simply labeled \u201c\ufb01xation\u201d according to these studies ) and outside \ufb01xation ( labeled \u201cexpansion\u201d according to these studies ) , and therefore identify the common examples ( inside \ufb01xa - tion ) for this creativity task . TABLE 1 . Categories of Solutions of the Egg\u2019s Task ( Agogu (cid:1) e , Kazakc (cid:1) i , et al . , 2014 ) Categories of Examples Example of Solutions Fixation / Expansion ( ideas that are not \ufb01xation ) Damping the shock Place a mattress at the reception Fixation Protecting the egg Pack the egg with bubble wrap Fixation Slowing the fall Hang the egg to a parachute Fixation Interrupting the fall Catch the egg with a net Expansion Acting before the fall Drop the egg at a height of 11 m Expansion Acting after the fall Replace the broken egg with an unbroken one Expansion Using a living device Train an eagle to take down the egg Expansion Modifying the properties of the egg Freezing the egg Expansion Using the natural properties of the egg Drop the egg on its most robust axis Expansion Using the properties of the environment Drop the egg at zero gravity Expansion 117 Journal of Creative Behavior 21626057 , 2020 , 1 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / j o c b . 349 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 09 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e Participants were randomly assigned to one of the three groups . All participants were provided with the typical instruction of the hen\u2019s egg task : \u201cYou are a designer and your manager gives you the following problem : Ensure that a hen\u2019s egg dropped from a height of 10 m does not break\u201d . The control group was provided with an additional guideline stating that : \u201cThe evaluation of your man - ager will be based on the number of original ideas you will propose\u201d . Participants of the group \u201cSpeci\ufb01c\u201d were provided with another additional guideline , imposing con - straints on speci\ufb01c examples inside \ufb01xation . The guideline stated that : \u201cThe evaluation of your manager will be based on the number of original ideas you will propose , knowing that your solutions must not use mat - tress , nor parachute , nor bubble wrap\u201d . These three speci\ufb01c examples were precisely chosen among others , due to the fact that they were the most generated examples in each of the three categories of \ufb01xation in an existing database of participants that performed this task in the past 5 years ( Agogu (cid:1) e , Kazakc (cid:1) i , et al . , 2014 ; Agogu (cid:1) e , Poirel , et al . , 2014 ) . Participants of the group \u201cAbstract\u201d were provided with another additional guideline , imposing con - straints on a more abstract level of these examples inside \ufb01xation , namely categories of these examples . The guideline stated that : \u201cThe evaluation of your manager will be based on the number of original ideas you will propose , knowing that your solutions must neither dampen the shock , nor slow the fall , nor protect the egg\u201d . These three abstract examples were considered the three categories of \ufb01xation of the present creative ideation task ( Agogu (cid:1) e , Le Masson , Dalmasso , Houd (cid:1) e & Cassotti , 2015 ; Agogu (cid:1) e , Kazakc (cid:1) i , et al . , 2014 ; Agogu (cid:1) e , Poirel , et al . , 2014 ; Cassotti , Agogu (cid:1) e , et al . , 2016 ; Cassotti , Camarda , et al . , 2016 ; Ezzat et al . , 2017 ) . DATA ANALYSIS The creative performance of the participants of the hen\u2019s egg task was quanti\ufb01ed by measuring the num - ber of solutions participants were given inside and outside \ufb01xation ( Agogu (cid:1) e et al . , 2015 ) . This could be done using an existing cartography of solutions of the hen\u2019s egg task , representing the distribution of solu - tions across different categories . To perform this measure , two trained raters assigned each of participants\u2019 solutions to one of the ten categories of solutions . The obtained inter - rater reliability score was 92 % . Only three categories among the ten were assigned inside \ufb01xation ( damping the shock , slowing the fall , and pro - tecting the egg ) . All solutions outside these three categories were assigned outside \ufb01xation ( modifying the properties of the egg , acting before / after the fall , or using a living device , etc . . . ) . Given that creativity requires both novelty / originality and usefulness / feasibility , we applied an external rating procedure to assess feasibility . More speci\ufb01cally , two independent raters were instructed to evaluate each idea on a \ufb01ve - point rating scale ranging from 1 ( \u201cnot feasible at all\u201d ) to 5 ( \u201chighly feasible\u201d ) . The raters displayed satisfactory intra - class correlation ( ICC = 0 . 90 ) . We also computed an objective measurement of the originality of the solutions by considering the fre - quency of the responses provided by all participants . For this score , the originality of a solution was de\ufb01ned as the normalized statistical infrequency of that particular solution . A mean originality score was calculated for each participant , which could range from 0 to 1 ( 0 represented lower originality and 1 represented higher originality ) . RESULTS In order to examine whether the numbers of proposed solutions ( ideational \ufb02uency ) inside \ufb01xation ( \ufb01x - ation ) and outside \ufb01xation ( expansion ) varied according to the experimental conditions , we conducted a repeated - measures analysis of variance ( ANOVA ) with the experimental conditions ( speci\ufb01c ; control and abstract ) as a between - subjects factor and the category of solution ( \ufb01xation vs . expansion ) as a within - sub - jects factor . We used the partial eta squared ( g 2 p ) and Cohen\u2019s d to assess the effect size . The ANOVA revealed a main effect of the category of solution ( F ( 1 , 72 ) = 7 . 83 , p = . 007 , g 2 p = . 10 ) indicating that the participants provided more solutions in the \ufb01xation path than in the expansion path . This analysis also showed a main effect of the experimental conditions ( F ( 2 , 72 ) = 4 . 13 , p = . 02 , g 2 p = . 10 ) . Moreover , there was a signi\ufb01cant experimental conditions 9 category of solution interaction ( F ( 2 , 72 ) = 48 . 06 , p < . 0001 , g 2 p = . 57 , see Figure 1a ) . Critically , the experimental conditions 9 category of solu - tion interaction was still signi\ufb01cant after controlling for the feasibility scores , F ( 2 , 71 ) = 32 . 70 , p < . 0001 , g 2 p = . 48 ) . Because main effects were modulated by the two - way interactions , we focused further analysis on the lat - ter . Planned comparisons , corrected with a Holm \u2013 Bonferroni procedure , revealed no signi\ufb01cant difference 118 Effects of Examples on Creative Ideation 21626057 , 2020 , 1 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / j o c b . 349 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 09 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e between the number of solutions in the expansion path in the control group ( M = 2 . 48 , SD = 1 . 78 ) and those in the group \u201cspeci\ufb01c\u201d ( M = 2 . 56 , SD = 1 . 74 , F ( 1 , 72 ) < 1 , d = . 04 ) . However , participants of the group \u201cabstract\u201d ( M = 5 . 16 , SD = 2 . 76 ) proposed much more solutions in the expansion path compared to the control group ( M = 2 . 48 , SD = 1 . 78 ; F ( 1 , 72 ) = 19 . 48 , p corr < . 0001 , d = 1 . 15 ) , and to the group \u201cspeci - \ufb01c\u201d ( M = 2 . 56 , SD = 1 . 74 , F ( 1 , 72 ) = 18 . 34 , p corr < . 0001 , d = 1 . 13 ) . Interestingly , the participants of the group \u201cabstract\u201d ( M = 1 . 36 , SD = 1 . 15 ) proposed fewer solutions in the \ufb01xation path than did those in the control group ( M = 5 . 24 , SD = 2 . 35 ; F ( 1 , 72 ) = 38 . 43 , p corr < . 0001 , d = 2 . 1 ) , as well as did the participants of the group \u201cspeci\ufb01c\u201d ( M = 6 . 52 , SD = 2 . 8 ; F ( 1 , 72 ) = 67 . 97 , p corr < . 0001 , d = 2 . 41 ) . Additionally , the group \u201cspeci\ufb01c\u201d ( M = 6 . 52 , SD = 2 . 8 ) proposed more solutions in the \ufb01xation path than the control group ( M = 5 . 24 , SD = 2 . 35 ; F ( 1 , 72 ) = 4 . 18 , p corr = . 04 , d = 0 . 5 ) . To examine whether originality scores varied according to the experimental conditions , we conducted a one - way analysis of variance ( ANOVA ) with the experimental conditions ( speci\ufb01c ; control and abstract ) as a between - subjects factor . We used the partial eta squared ( g 2 p ) and Cohen\u2019s d to assess the effect size . The ANOVA revealed a main effect of the experimental conditions ( F ( 2 , 72 ) = 3 . 18 , p = . 047 , g 2 p = . 08 , see Fig - ure 1b ) . Planned comparisons , revealed no signi\ufb01cant difference between the originality scores of the control group ( M = 0 . 32 , SD = . 16 ) and those of the group \u201cspeci\ufb01c\u201d ( M = . 31 , SD = . 18 , F ( 1 , 72 ) < 1 , d = . 06 ) . However , the solutions proposed by the group \u201cabstract\u201d ( M = . 42 , SD = . 16 ) were more original than those provided by the control group ( M = . 32 , SD = . 16 , F ( 1 , 72 ) = 4 . 26 , p = . 04 , d = . 62 ) and the group \u201cspeci\ufb01c\u201d ( M = . 31 , SD = . 18 , F ( 1 , 72 ) = 5 . 24 , p = . 025 , d = . 65 ) . DISCUSSION In the present study , we demonstrated that according to the level of the speci\ufb01city and abstraction of common examples presented prior to a creative idea generation task , we could obtain opposite effects on \ufb01xation mitigation . We showed that ( a ) instructing participants to avoid using common examples when for - mulated with a high level of speci\ufb01city increases \ufb01xation , and therefore constrains participants\u2019 capacity to generate creative ideas ; whereas ( b ) instructing participants to avoid such examples while using a more abstract level for stating these common examples \u2014 such as a categorization of these examples \u2014 mitigates \ufb01x - ation , and consequently increases the number of creative ideas generated \u2014 and in quite a signi\ufb01cant way . More precisely , statistical results show that constraints on common examples , formulated with high level of speci\ufb01city , lightly increase the number of solutions inside the \ufb01xation path , but however have no effect on the number of solutions outside the \ufb01xation path . However , statistical analysis shows that constraints on common examples , formulated with high level of abstraction \u2014 such as a categorization of these examples \u2014 approximately reduce the number of solutions inside the \ufb01xation path by more than one third , and FIGURE 1 . ( A ) Mean number of responses according to the experimental condition ( Speci\ufb01city / Control / Abstract ) and the type of solution ( Fixation / Expansion ) . ( B ) Mean originality scores according to the experimental condition ( Speci\ufb01city / Control / Abstract ) . 119 Journal of Creative Behavior 21626057 , 2020 , 1 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / j o c b . 349 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 09 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e surprisingly double the number of solutions outside the \ufb01xation path . Thus , the magnitude of the stimula - tion effect produced by explicit warning of not using certain categories is quite surprising and requires some discussion . Theoretically , these \ufb01ndings might suggest that when constraints on examples are formulated with high level of abstraction , individuals are forced to reason outside \ufb01xation , and therefore succeed to generate novel alternatives to the common and typical categories of solutions provided to them ( outside the \ufb01xation path ) . Whereas , when constraints on examples are formulated with high level of speci\ufb01city , individuals are more affected by the speci\ufb01c examples provided to them , and follow the path of least resistance , which consist of generating alternatives to the common and typical solutions ( inside the \ufb01xation path ) . These \ufb01ndings \ufb01rst of all con\ufb01rm the studies regarding the negative role of examples for creativity ( Agogu (cid:1) e , Kazakc (cid:1) i , et al . , 2014 ; Jansson & Smith , 1991 ; Smith et al . , 1993 ) , since the results of the group \u201cspeci\ufb01c\u201d showed that the introduction of a common example with high level of speci\ufb01city could highly increase \ufb01xation effect , and therefore constrains the capacity of individuals to generate creative ideas . Secondly , our \ufb01ndings add to the literature on new practical procedures to overcome \ufb01xation effects in creative ideation contexts ( Agogu (cid:1) e , Kazakc (cid:1) i , et al . , 2014 ; Linsey et al . , 2010 ; Moreno , Yang , Hern (cid:1) andez , Lin - sey & Wood , 2015 ; Zahner , Nickerson , Tversky , Corter & Ma , 2010 ) , through the appropriate use of con - straints and warnings . We demonstrate that two types of constraints could have opposite effects on \ufb01xation modulation , according to their level of speci\ufb01city and abstraction . The more constraints on inappropriate examples are abstract , the more \ufb01xation effect is mitigated . Similarly , the more the constraints on inappro - priate examples are speci\ufb01c , the more \ufb01xation effect is facilitated . Thirdly , our \ufb01ndings raised more questions than what they contribute to , especially regarding the crucial role of categorization in creative ideation contexts . First of all , the present study con\ufb01rms previous studies ( Baughman & Mumford , 1995 ; Nagai & Taura , 2009 ; Ward et al . , 2004 ) regarding the important role of abstraction for \ufb01xation mitigation . Furthermore , since we explicitly measure the number of solutions inside / outside the \ufb01xation path in the present experiments , using the statistical measurement of the variable \u201cex - pansivity\u201d ( Agogu (cid:1) e et al . , 2015 ) , our \ufb01ndings present further evidence regarding the key role that could be played by categorization for \ufb01xation modulation and creativity . Finally , from a purely managerial perspective , our \ufb01ndings are consistent with the view of the literature on expertise and categorization ( Chi , Feltovich & Glaser , 1981 ) , arguing that experts have the skills to both recognize and restructure problems ( Akin , 1990 ) , in a way in which it allow them seeing the problem from a broader and more abstract view than novices . Additionally , it gives new sights for understanding how leaders could brief their teams through initial instructions in creativity situations , in a way they could help their teams to avoid falling in the cognitive trap of \ufb01xation , and stimulate their creative performance ( Car - son & Carson , 1993 ; Chaffois , Gillier , Belkhouja & Roth , 2015 ; Ezzat et al . , 2017 ; Runco , Illies & Eisenman , 2005 ) . We show that it is not suf\ufb01cient and enough for leaders to simply impose constraints on unwanted ideas and solutions in creative projects for their teams , but leaders must have the ability to formulate these constraints in appropriate levels of abstraction , in a way that ensures \ufb01xation is majorly set aside , in order to open the way for creativity to \ufb02ourish . One possible limitation of our study was that we considered common or less novel ideas and solutions generated by participants , consisting of \u201cdamping the shock\u201d , \u201cprotecting the egg\u201d , or \u201cslowing the fall\u201d , as \ufb01xation categories . In fact , we did not measure or pretest in the present experiment whether the participants have used that techniques for similar purposes ( avoid breaking objects ) before . In line with this limitation , further experiments could consist of controlling this speci\ufb01c \ufb01xation issue using previous techniques tested on creative idea generation task ( Benedek et al . , 2014 ; Silvia , Nusbaum & Beaty , 2015 ) to assess whether individuals\u2019 responses were based on \u201cold\u201d ( responses retrieved from memory ) or \u201cnew\u201d ( responses gener - ated on the spot ) . CONCLUSION The present study demonstrates that common examples \u2014 if formulated in a high level of abstraction \u2014 can play a crucial role to help individuals overcome \ufb01xation effects occurring in creative idea generation sit - uations . Our results clearly suggest that the way the common examples are formulated prior to a creativity task \u2014 either with speci\ufb01city or abstraction \u2014 could have opposite effects on \ufb01xation . As a result , the present study provides new insights to the literature on the positive and negative role of examples in creative ideation . 120 Effects of Examples on Creative Ideation 21626057 , 2020 , 1 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / j o c b . 349 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 09 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e REFERENCES Adamson , R . E . ( 1952 ) . Functional \ufb01xedness as related to problem solving : A repetition of three experiments . Journal of Experimen - tal Psychology , 44 , 288 . Agogu (cid:1) e , M . , Kazakc (cid:1) i , A . , Hatchuel , A . , Masson , P . , Weil , B . , Poirel , N . , & Cassotti , M . ( 2014 ) . The impact of type of examples on originality : Explaining \ufb01xation and stimulation effects . The Journal of Creative Behavior , 48 , 1 \u2013 12 . Agogu (cid:1) e , M . , Le Masson , P . , Dalmasso , C . , Houd (cid:1) e , O . , & Cassotti , M . ( 2015 ) . Resisting classical solutions : The creative mind of industrial designers and engineers . Psychology of Aesthetics , Creativity , and the Arts , 9 , 313 . Agogu (cid:1) e , M . , Poirel , N . , Pineau , A . , Houd (cid:1) e , O . , & Cassotti , M . ( 2014 ) . The impact of age and training on creativity : A design - theory approach to study \ufb01xation effects . Thinking Skills and Creativity , 11 , 33 \u2013 41 . Akin , \u20ac O . ( 1990 ) . Necessary conditions for design expertise and creativity . Design Studies , 11 , 107 \u2013 113 . Amabile , T . ( 1996 ) . Creativity in context . Boulder , CO : Westview press . Baughman , W . A . , & Mumford , M . D . ( 1995 ) . Process - analytic models of creative capacities : Operations in\ufb02uencing the combina - tion - and - reorganization process . Creativity Research Journal , 8 , 37 \u2013 62 . Benedek , M . , Jauk , E . , Fink , A . , Koschutnig , K . , Reishofer , G . , Ebner , F . , & Neubauer , A . C . ( 2014 ) . To create or to recall ? Neural mechanisms underlying the generation of creative new ideas . NeuroImage , 88 , 125 \u2013 133 . Carson , P . P . , & Carson , K . D . ( 1993 ) . Managing creativity enhancement through goal - setting and feedback * . The Journal of Creative Behavior , 27 , 36 \u2013 45 . Cassotti , M . , Agogu (cid:1) e , M . , Camarda , A . , Houd (cid:1) e , O . , & Borst , G . ( 2016 ) . Inhibitory control as a core process of creative problem solving and idea generation from childhood to adulthood . New Directions for Child and Adolescent Development , 2016 , 61 \u2013 72 . Cassotti , M . , Camarda , A . , Poirel , N . , Houd (cid:1) e , O . , & Agogu (cid:1) e , M . ( 2016 ) . Fixation effect in creative ideas generation : Opposite impacts of example in children and adults . Thinking Skills and Creativity , 19 , 146 \u2013 152 . Chaffois , C . , Gillier , T . , Belkhouja , M . , & Roth , Y . ( 2015 ) . How task instructions impact the creativity of designers and ordinary partic - ipants in online idea generation . Paper presented at the 22nd innovation & product development management conference . Chi , M . T . , Feltovich , P . J . , & Glaser , R . ( 1981 ) . Categorization and representation of physics problems by experts and novices . Cog - nitive Science , 5 , 121 \u2013 152 . Chrysikou , E . G . , & Weisberg , R . W . ( 2005 ) . Following the wrong footsteps : Fixation effects of pictorial examples in a design prob - lem - solving task . Journal of Experimental Psychology : Learning , Memory , and Cognition , 31 , 1134 . Ezzat , H . , Camarda , A . , Cassotti , M . , Agogu (cid:1) e , M . , Houd (cid:1) e , O . , Weil , B . , & Le Masson , P . ( 2017 ) . How minimal executive feedback in\ufb02uences creative idea generation . PLoS ONE , 12 , e0180458 . Faul , F . , Erdfelder , E . , Buchner , A . , & Lang , A . G . ( 2009 ) . Statistical power analyses using G * Power 3 . 1 : Tests for correlation and regression analyses . Behavior Research Methods , 41 , 1149 \u2013 1160 . Guilford , J . P . ( 1959 ) . Traits of creativity . In H . H . Anderson ( Ed . ) , Creativity and its cultivation ( pp . 142 \u2013 161 ) . New York : Harper & Row . Hatchuel , A . , & Weil , B . ( 2002 ) . CK theory . Paper presented at the Proceedings of the Herbert Simon International Conference on \u00ab Design Sciences . Jansson , D . G . , & Smith , S . M . ( 1991 ) . Design \ufb01xation . Design Studies , 12 , 3 \u2013 11 . Linsey , J . S . , Tseng , I . , Fu , K . , Cagan , J . , Wood , K . L . , & Schunn , C . ( 2010 ) . A study of design \ufb01xation , its mitigation and perception in engineering design faculty . Journal of Mechanical Design , 132 , 041003 . Maier , N . R . ( 1931 ) . Reasoning and learning . Psychological Review , 38 , 332 . Moreno , D . P . , Yang , M . C . , Hern (cid:1) andez , A . A . , Linsey , J . S . , & Wood , K . L . ( 2015 ) . A step beyond to overcome design \ufb01xation : A design - by - analogy approach . In : Design computing and Cognition\u201914 ( pp . 607 \u2013 624 ) . Cham : Springer . Nagai , Y . , & Taura , T . ( 2009 ) . Design motifs : Abstraction driven creativity . Special Issue of Japanese Society for the Science of Design , 16 , 62 . Runco , M . A . , & Acar , S . ( 2012 ) . Divergent thinking as an indicator of creative potential . Creativity Research Journal , 24 , 66 \u2013 75 . Runco , M . A . , Illies , J . J . , & Eisenman , R . ( 2005 ) . Creativity , originality , and appropriateness : What do explicit instructions tell us about their relationships ? The Journal of Creative Behavior , 39 , 137 \u2013 148 . Silvia , P . J . , Nusbaum , E . C . , & Beaty , R . E . ( 2015 ) . Old or new ? Evaluating the old / new scoring method for divergent thinking tasks . The Journal of Creative Behavior , 51 , 216 \u2013 224 . Sio , U . N . , Kotovsky , K . , & Cagan , J . ( 2015 ) . Fixation or inspiration ? A meta - analytic review of the role of examples on design pro - cesses . Design Studies , 39 , 70 \u2013 99 . Smith , S . M . , Ward , T . B . , & Schumacher , J . S . ( 1993 ) . Constraining effects of examples in a creative generation task . Memory & Cog - nition , 21 , 837 \u2013 845 . Viswanathan , V . , & Linsey , J . ( 2013 ) . Mitigation of Design Fixation in Engineering Idea Generation : A Study on the Role of De\ufb01xa - tion Instructions . ICoRD\u201913 ( pp . 113 \u2013 124 ) : Springer . Ward , T . B . , Patterson , M . J . , & Sifonis , C . M . ( 2004 ) . The role of speci\ufb01city and abstraction in creative idea generation . Creativity Research Journal , 16 , 1 \u2013 9 . 121 Journal of Creative Behavior 21626057 , 2020 , 1 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / j o c b . 349 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 09 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e Zahner , D . , Nickerson , J . V . , Tversky , B . , Corter , J . E . , & Ma , J . ( 2010 ) . A \ufb01x for \ufb01xation ? Rerepresenting and abstracting as creative processes in the design of information systems . Arti\ufb01cial Intelligence for Engineering Design , Analysis and Manufacturing , 24 , 231 \u2013 244 . Hicham Ezzat , MINES ParisTech \u2013 PSL Research University Marine Agogu (cid:1) e , HEC Montr (cid:1) eal Pascal Le Masson , Benoit Weil , MINES ParisTech \u2013 PSL Research University Mathieu Cassotti , Paris Descartes University Correspondence concerning this article should be addressed to Hicham Ezzat , MINES ParisTech \u2013 PSL Research University , Center for Management Science , i3 UMR CNRS 9217 , France . E - mail : hicham . ezzat @ mines - paristech . fr ACKNOWLEDGMENT This research was \ufb01nanced by a grant from the French National Research Agency ( ANR - 13 - SOIN - 0004 - 02 ID (cid:1) e\ufb01xE ) . 122 Effects of Examples on Creative Ideation 21626057 , 2020 , 1 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / j o c b . 349 by U n i v e r s it y O f M a r y l a nd , W il e y O n li n e L i b r a r y on [ 09 / 05 / 2024 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e",
    "numasawa2020fluorescent": "German Edition : DOI : 10 . 1002 / ange . 201914826 Fluorescent Probes International Edition : DOI : 10 . 1002 / anie . 201914826 A Fluorescent Probe for Rapid , High - Contrast Visualization of Folate - Receptor - Expressing Tumors In Vivo Koji Numasawa , Kenjiro Hanaoka , * Naoko Saito , Yoshifumi Yamaguchi , Takayuki Ikeno , Honami Echizen , Masahiro Yasunaga , Toru Komatsu , Tasuku Ueno , Masayuki Miura , Tetsuo Nagano , and Yasuteru Urano * Abstract : Folate receptors ( FRs ) are membrane proteins involved in folic acid uptake , and the alpha isoform ( FR - a ) is overexpressed in ovarian and endometrial cancer cells . For fluorescence imaging of FRs in vivo , the near - infrared ( NIR ) region ( 650 \u2013 900 nm ) , in which tissue penetration is high and autofluorescence is low , is optimal , but existing NIR fluores - cent probes targeting FR - a show high non - specific tissue adsorption , and require prolonged washout to visualize tumors . We have designed and synthesized a new NIR fluorescent probe , FolateSiR - 1 , utilizing a Si - rhodamine fluorophore having a carboxy group at the benzene moiety , coupled to a folate ligand moiety through a negatively charged tripeptide linker . This probe exhibits very low background fluorescence and afforded a tumor - to - background ratio ( TBR ) of up to 83 in FR - expressing tumor - bearing mice within 30 min . Thus , FolateSiR - 1 has the potential to contribute to the research in the field of biology and the clinical medicine . Introduction Folic acid , which is required in one - carbon metabolic reactions and for the synthesis of nucleotide bases , emerged as a targeting ligand for imaging of cancer tissues in the 1990s . [ 1 , 2 ] It is internalized into cells through folate receptors ( FRs ) expressed on the cell surface . The alpha isoform of folate receptor ( FR - a ) is upregulated in about 40 % of human cancers , especially in malignant tissues such as ovarian cancer , whereas normal tissues , except for the kidney , do not accumulate folic acid or its conjugates . [ 3 \u2013 5 ] Folic acid binds to FR - a with high affinity ( K d & 10 @ 9 m ) even after conjuga - tion to imaging agents , and undergoes receptor - mediated endocytosis . [ 3 ] Consequently , various folate - linked drugs and imaging agents have been developed . [ 1 \u2013 3 ] Among available imaging modalities , fluorescence imaging provides real - time images with millimeter resolution , and has attracted interest for intraoperative fluorescence imaging of tumor tissues . [ 6 ] For example , 90 \u2013 95 % of epithelial ovarian cancers over - express FR - a , and a folate - linked fluorescent dye , folate - FITC , was recently employed for intraoperative tumor - specific fluorescence imaging in patients . [ 7 ] However , FITC emits green fluorescence ( around 520 nm ) , and is unsuitable for imaging deep tissues . For this purpose , dyes emitting in the near - infrared ( NIR ) region ( 650 \u2013 900 nm ) are most useful because tissue penetration is high and autofluorescence is low , resulting in a low background signal . [ 8 , 9 ] However , existing folate - linked NIR fluorescent dyes show nonspecific adsorp - tion on tissues , and require a washout period of up to 24 h to clearly image tumors in tumor - bearing mice ( Scheme 1 ) . [ 10 \u2013 12 ] Therefore , we aimed to develop a NIR fluorescent probe suitable for rapid , high - contrast detection of FR - expressing tumors in vivo , without the need for a washout procedure . Two approaches have so far been used to improve the signal - to - background ratio ( SBR ) of fluorescent probes . One approach is to develop activatable fluorescent probes whose fluorescence signal is emitted only in response to a specific feature of the local environment , such as pH . For example , a probe targeting human epidermal growth factor receptor type 2 ( HER2 ) employed a fluorescent dye - labeled antibody that is fluorescently activated in the low pH environment of lysosomes after cellular internalization . [ 13 ] This approach provides a high tumor - to - normal - tissue signal ratio . However , endocytotic transport of the probe to lysosomes takes up to 24 h . Yang et al . designed and synthesized an off / on - type fluorescent probe for detection of FRs by utilizing a pH - sensitive acyl hydrazone linker and a dark quencher , but failed to observe fluorescence enhancement after cellular internalization . [ 14 ] They also developed a folate conjugate whose fluorescence wavelength is changed by reduction of disulfide in the probe structure during receptor - mediated [ * ] Dr . K . Numasawa , Prof . K . Hanaoka , N . Saito , T . Ikeno , H . Echizen , Dr . T . Komatsu , Dr . T . Ueno , Prof . M . Miura , Prof . Y . Urano Graduate School of Pharmaceutical Sciences , The University of Tokyo 7 - 3 - 1 Hongo , Bunkyo - ku , Tokyo 113 - 0033 ( Japan ) E - mail : khanaoka @ mol . f . u - tokyo . ac . jp uranokun @ mol . f . u - tokyo . ac . jp Prof . Y . Yamaguchi Institute of Low Temperature Science , Hokkaido University Sapporo 060 - 0819 ( Japan ) Dr . M . Yasunaga Division of Developmental Therapeutics , Exploratory Oncology Research & Clinical Trial Center , National Cancer Center 6 - 5 - 1 Kashiwanoha , Kashiwa , Chiba 277 - 8577 ( Japan ) Prof . T . Nagano Drug Discovery Initiative , The University of Tokyo 7 - 3 - 1 Hongo , Bunkyo - ku , Tokyo 113 - 0033 ( Japan ) Prof . Y . Urano Graduate School of Medicine , The University of Tokyo 7 - 3 - 1 Hongo , Bunkyo - ku , Tokyo 113 - 0033 ( Japan ) andCREST ( Japan ) Agency for Medical Research and Development ( AMED ) 1 - 7 - 1 Otemachi , Chiyoda - ku , Tokyo 100 - 0004 ( Japan ) Supporting information and the ORCID identification number ( s ) for the author ( s ) of this article can be found under : https : / / doi . org / 10 . 1002 / anie . 201914826 . Angewandte Chemie Research Articles 6015 Angew . Chem . Int . Ed . 2020 , 59 , 6015 \u2013 6020 T 2020 Wiley - VCH Verlag GmbH & Co . KGaA , Weinheim endocytosis . [ 15 ] The latter probe was suitable for cellular applications , but the half - time of disulfide reduction was as long as 6 h after endocytosis . The second approach is to minimize nonspecific adsorption of the probe in vivo . Choi et al . reported that the zwitterionic heptamethine indocya - nine NIR fluorophore ZW800 - 1 exhibits low serum - protein binding , ultralow nonspecific tissue background , and rapid elimination from the body through renal filtration . [ 16 ] In tumor model applications , a tumor - to - background ratio ( defined as the contrast - to - background ratio in the tumor divided by the contrast - to - background ratio in nearby normal tissue ) of over 17 was achieved at 4 h after injection of ZW800 - 1 conjugated to cRGD . Herein , we build on this approach to design and synthesize an NIR fluorescent probe that we believe offers sufficiently high performance for practical in vivo imaging . Results and Discussion First , we needed a system to assess the occurrence of nonspecific adsorption of folate - linked NIR fluorescent dyes . For this purpose , we performed live - cell fluorescence imaging of FR - a - expressing and FR - a - non - expressing cultured cells , i . e . , KB cells ( FR + ) and OVCAR - 3 cells ( FR @ ) , with a commercially available NIR fluorescent probe , FolateR - Sense 680 ( PerkinElmer , USA ) . The expression of FRs on these cells was confirmed by immunostaining ( Supporting Information , Figure S1 ) . KB cells showed strong fluorescence at the cell membrane due to probe binding to FR - a on the cell surface , but we also observed some bright dots inside the cells , apparently due to nonspecific internalization of the probe ( Figure S2a ) . On the other hand , OVCAR - 3 cells showed many very bright dots within the cells ( Figure S2b ) . There - fore , we considered that this cellular imaging system was suitable to judge whether or not our newly synthesized folate - linked fluorescent dyes might show nonspecific adsorption when used for in vivo fluorescence imaging . In the molecular design of our folate - linked fluorescent dyes , we chose to conjugate the linker moiety to the folate glutamate moiety ( Figure 1a ) . The crystal structure of human FR - a complexed with folic acid has the folate pteroate moiety buried inside FR , whereas the glutamate moiety is exposed to the solvent and protrudes from the binding pocket entrance , so that it can be conjugated with a fluorescent dye without adversely affecting the binding to FR - a . [ 17 ] By this molecular design , we expected that fluorescent probes would show the high affinity for FR , i . e . , a K d value around 10 @ 9 m , because folate conjugates ( and folic acid ) normally bind to FR with this affinity . [ 3 ] We also employed a negatively charged peptide linker , Asp \u2013 Lys \u2013 Gly , in order to reduce the cell - membrane permeability of the folate - linked fluorescent dye , and we attached various xanthene fluorophores to the amino group of the lysine side chain in the linker ( Figure 1a and Figure S3 ) . We applied each compound to the live - cell fluorescence Scheme 1 . Imaging strategy of existing near - infrared fluorescent probes for in vivo fluorescence imaging of whole animals . Figure 1 . a ) Molecular design of fluorescent probes for detection of folate receptors . The structures of FolateSiR - 1 and FolateSiR - 2 are also shown . b ) Absorption and emission spectra of 1 m m FolateSiR - 1 in 100 m m sodium phosphate buffer at pH 7 . 4 , l ex = 652 nm . c ) Absorp - tion and emission spectra of 1 m m FolateSiR - 2 in 100 m m sodium phosphate buffer at pH 7 . 4 , l ex = 656 nm . Angewandte Chemie Research Articles 6016 www . angewandte . org T 2020 Wiley - VCH Verlag GmbH & Co . KGaA , Weinheim Angew . Chem . Int . Ed . 2020 , 59 , 6015 \u2013 6020 15213773 , 2020 , 15 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / a n i e . 201914826 by U n i v e r s it y O f W a s h i ng t on , W il e y O n l i n e L i b r a r y on [ 14 / 11 / 2023 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e imaging with KB cells and OVCAR - 3 cells . As shown in Figure S3 , basically fluorescein derivative - labeled folates such as Fluorescein Folate , 2 - Me DCTM Folate , and 2 - COOH DCTM Folate showed the selective detection of FR in living cells , while rhodamine derivative - labeled folate such as TAMRA Folate and Alexa488 Folate stained the cell membrane of KB cells , but they also showed nonspecific adsorption . Among them , we fortunately found a promising NIR fluorescent probe for detecting FR , FolateSiR - 1 , along with a control compound , FolateSiR - 2 , which has very similar chemical structure to FolateSiR - 1 ( Figure 1a ) . Both Folate - SiR - 1 and FolateSiR - 2 showed similar absorption and emis - sion spectra in the NIR region , and their fluorescence quantum yields were 7 . 6 % and 5 . 1 % , respectively ( Figure 1 ) . The relatively low fluorescence quantum yields were prob - ably due to dynamic quenching between the fluorophore and the electron - rich pteroate moiety , [ 18 ] but are sufficiently high for cellular and in vivo fluorescence imaging . When we applied FolateSiR - 1 to KB cells , the cell membrane was clearly stained ; there were no bright dots inside the cells , and the fluorescence disappeared in the presence of 1 m m folic acid as a competitor ( Figure 2a ) , indicating that nonspecific endocytosis of FolateSiR - 1 did not occur . FolateSiR - 1 may also possess the low cell - membrane permeability owing to the relatively large molecular weight and the negatively charged peptide linker . Further , when we applied FolateSiR - 1 to OVCAR - 3 cells , almost no fluorescence was observed , supporting the idea that there is little nonspecific adsorption of the dye ( Figure 2b ) . On the other hand , FolateSiR - 2 stained the cell membrane of KB cells , but also exhibited many bright dots inside the cells both in the presence and absence of 1 m m folic acid ( Figure 2a ) . It also afforded some bright dots inside OVCAR - 3 cells , supporting the existence of nonspecific adsorption ( Figure 2b ) . Although folate recep - tors can normally induce cell uptake of ligands through endocytosis , the endocytosis of FolateSiR - 1 and FolateSiR - 2 through folate receptors was not observed . In short , Folate - SiR - 1 showed a high S / N ratio with little nonspecific adsorption in cellular fluorescence imaging . Next , we applied these two fluorescent probes to ex vivo mouse embryos to visualize endogenously expressed folate - binding protein 1 ( Folbp1 ) , the mouse orthologue of human FR - a . Periconceptional folate supplementation significantly reduces the risk of neural tube defects , and Folbp1 is one of the membrane proteins that mediate cellular uptake of folate ; mice lacking Folbp1 are defective in early embryonic devel - opment . [ 19 ] Maternal anti - FR antibodies are also associated with neuronal tube defects in humans . [ 20 ] In situ hybridization revealed a distinct expression pattern of Folbp1 mRNA in the neural folds prior to the initiation of neural tube closure at the cervical region and the prosencephalic / mesencephalic boun - dary . [ 21 ] Folbp1 mRNA is mainly localized to the most dorsal regions of the neural folds , where fusion takes place , and as neural fold fusion proceeds , Folbp1 mRNA expression extends to the adjacent unfused neural folds ( Figure 3a ) . However , the expression pattern of Folbp1 protein has not been reported . We firstly applied FolateSiR - 1 to embryos at day 8 . 5 postcoitum . The living embryos were stained with 20 m m FolateSiR - 1 for 30 min at 37 88 C and fluorescence imaging was quickly performed . The neural tube closing region showed strong fluorescence compared with other regions , Figure 2 . a ) Bright - field ( left ) and fluorescence ( right ) images of KB cells incubated with 5 m m FolateSiR - 1 or FolateSiR - 2 in the presence or absence of 1 m m folic acid and 0 . 5 % DMSO as a cosolvent . White arrows indicate bright dots inside cells ; l ex = 650 nm , l em = 670 \u2013 750 nm . Scale bars = 20 m m . b ) Bright - field ( left ) and fluorescence ( right ) images of OVCAR - 3 cells incubated with 5 m m FolateSiR - 1 or FolateSiR - 2 and 0 . 5 % DMSO as a cosolvent . White arrows indicate bright dots inside cells , l ex = 650 nm , l em = 670 \u2013 750 nm . Scale bars = 20 m m . Figure 3 . a ) Schematic of folate receptor ( = Folbp 1 ) expression in mouse embryo ; regions reported to show folate receptor expression are shown in gray . b ) Fluorescence image of mouse embryo incubated with 20 m m FolateSiR - 1 . Locations stained with FolateSiR - 1 are indicat - ed by white arrowheads . c ) Fluorescence image of mouse embryo incubated with 20 m m FolateSiR - 2 . Angewandte Chemie Research Articles 6017 Angew . Chem . Int . Ed . 2020 , 59 , 6015 \u2013 6020 T 2020 Wiley - VCH Verlag GmbH & Co . KGaA , Weinheim www . angewandte . org 15213773 , 2020 , 15 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / a n i e . 201914826 by U n i v e r s it y O f W a s h i ng t on , W il e y O n l i n e L i b r a r y on [ 14 / 11 / 2023 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e while weak fluorescence was observed throughout the embryo ( Figure 3b ) . On the other hand , we observed strong , speckled fluorescence throughout the whole embryo when FolateSiR - 2 was loaded ( Figure 3c ) , suggesting the idea that FolateSiR - 2 binds non - specifically in the cells . We then performed a competition experiment of FolateSiR - 1 with 1 m m folic acid . The fluorescence images changed with time , probably because embryogenesis is rapid , so in this experi - ment we fixed the embryos with 4 % formaldehyde after staining with FolateSiR - 1 . We confirmed that FolateSiR - 1 was retained in the cell membrane of KB cells after the fixation process ( Figure S4a ) . Fluorescence imaging of the FolateSiR - 1 - stained , fixed embryos showed a similar pattern of strong fluorescence in the neural tube closing region to that shown in Figure 3b ( Figures S4b and S5 ) . This pattern disappeared in the presence of 1 m m folic acid ( Figure S4c ) . Thus , FolateSiR - 1 appears to stain Folbp1 - expressing regions in a folic acid - competitive manner . We further applied FolateSiR - 1 to in vivo fluorescence imaging of tumors in mouse models . FolateSiR - 1 was intra - venously injected through the tail vein into tumor - bearing mice prepared by subcutaneous inoculation of KB cells . FR - a - expressing tumors were clearly visualized with an extreme - ly low background ( TBR up to 83 ) just 30 min after injection of FolateSiR - 1 ( Figure 4a , b ) . The mice were sacrificed , and we confirmed that FolateSiR - 1 was strongly accumulated in the tumor ( fluorescence in the stomach may have been derived from the feed in this experiment ) ( Figure S6a ) . On the other hand , relatively strong fluorescence throughout the whole body was observed at the same time - point after injection of FolateSiR - 2 , resulting in a low TBR ( Figure 4c ) . Even at 6 h after injection of FolateSiR - 2 , the fluorescence was relatively strong throughout the whole body and the TBR was around 12 , which was much lower than that of FolateSiR - 1 ( Figure 4c , d ) , though tumor accumulation of FolateSiR - 2 was observed in excised tissues ( Figure S6b ) . We also prepared an FR ( @ ) - tumor model by subcutaneous injection of HT1080 cells . The absence of FR was confirmed by immunostaining ( Figure S1 ) . No fluorescence was observed when FolateSiR - 1 was injected into these mice ( Figure S7 ) . On the other hand , when we injected FolateSiR - 2 into these mice , fluorescence rapidly appeared throughout the whole body ( Figure S8 ) ; it subsequently decreased with time , but the tumor was not visualized . Moreover , when we injected 6 m m folic acid in saline into KB tumor - bearing mouse , followed by 100 m m FolateSiR - 1 in saline ( 100 m L ) , no fluorescence was observed except in the intestine , which might be due to the feed ( Figure S9 ) . These results indicate that FolateSiR - 1 binds specifically to FR - expressing tumors in these mice , affording high - contrast images unlike typical NIR fluorescent probes such as FolateSiR - 2 . To further investigate the potential applicability of FolateSiR - 1 for intraoperative tumor detection , we utilized a tissue microarray of patient - derived ovarian tumor tissues and normal tissues from regions adjacent to tumors ( Fig - ure S10 ) . FolateSiR - 1 successfully visualized ovarian tumor tissues , while exhibiting little binding to normal tissues ( Figure 5a and Figure S11 ) . Furthermore , the fluorescence image of the tissue microarray closely matched the immu - nostaining image ( Figure 5b and Figures S12 and S13 ) , confirming the ability of FolateSiR - 1 to visualize human tumors ex vivo . The selective staining of tumor tissues by FolateSiR - 1 was also observed in a specimen of tissue microarray , which contains both tumor tissues and non - tumor tissues ( vascular tissues and fibrous tissues ) ( Figure S13c ) . Conclusion In this study , we designed and synthesized a series of folate - fluorescent dye conjugates . We focused on the non - specific adsorption of typical NIR fluorescent dye such as cyanine dyes and folate conjugates to tissues inside the body when they are injected through the tail vein . We conjugated various xanthene - based fluorescent dyes to folic acid and applied them to the live - cell fluorescence imaging for selection of fluorescent probes showing low nonspecific adsorption . Among xanthene fluorescent dye - conjugated folates , we fortunately found that FolateSiR - 1 , consisting of Figure 4 . a ) Fluorescence images at different time points of KB tumor - bearing mouse injected with 100 m m FolateSiR - 1 in 100 m L saline containing 1 % DMSO as a cosolvent . Fluorescence and white light images were obtained before and 0 , 0 . 5 , 1 , 2 , 3 and 6 h after the probe injection ; l ex = 661 ( 641 \u2013 681 ) nm , l em = 700 \u2013 800 nm . T = tumor , M = muscle . Fluorescence intensity scale : gray scale 0 to 255 . b ) Time - dependent change of fluorescence intensity in tumor and non - tumor ( muscle ) areas of three mice , including the mouse in ( a ) . Error bar shows S . E . c ) Fluorescence images at different time points of KB tumor - bearing mouse injected with 100 m m FolateSiR - 2 in 100 m L saline containing 1 % DMSO as a cosolvent . Fluorescence and white - light images were obtained before and 0 , 0 . 5 , 1 , 2 , 3 and 6 h after the probe injection ; l ex = 661 ( 641 \u2013 681 ) nm , l em = 700 \u2013 800 nm . T = tumor , M = muscle . Fluorescence intensity scale : gray scale 0 to 255 . d ) Time - dependent change of fluorescence intensity in tumor and non - tumor ( muscle ) areas of three mice , including the mouse in ( c ) . Error bar shows S . E . Angewandte Chemie Research Articles 6018 www . angewandte . org T 2020 Wiley - VCH Verlag GmbH & Co . KGaA , Weinheim Angew . Chem . Int . Ed . 2020 , 59 , 6015 \u2013 6020 15213773 , 2020 , 15 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / a n i e . 201914826 by U n i v e r s it y O f W a s h i ng t on , W il e y O n l i n e L i b r a r y on [ 14 / 11 / 2023 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e a Si - rhodamine fluorophore having a carboxy group at the benzene moiety , coupled to a folate ligand moiety through a negatively charged tripeptide linker ( Figure 1a ) , achieved a very low level of nonspecific tissue binding . Indeed , this probe could rapidly provide high - contrast images of tumors in a mouse model in vivo without any washout procedure , due to the very low background fluorescence . We think the low background may be due to the combination of the intrinsic low level of nonspecific binding derived from the probe design and the preference of the 2 - COOH SiR650 fluoro - phore for the intramolecularly spirocyclized , nonfluorescent form in hydrophobic environments , such as those in plasma / inner membranes . [ 22 ] The spirocyclization converts the cat - ionic xanthene moiety to neutral form , whereas in contrast , the cationic charge of cyanines and Si - rhodamines favors nonspecific adsorption . Indeed , 2 - COOH SiR650 showed a large absorbance decrease in 100 m m sodium phosphate buffer ( pH 7 . 4 ) containing 10 % fetal bovine serum , which contains many proteins ( Figure S14 ) . This idea is also supported by the observation that when FolateSiR - 1 was dissolved in mouse serum , the absorbance was significantly decreased , presumably due to nonspecific binding to proteins of mouse serum and the formation of the intramolecularly spirocyclized , nonfluorescent form ( Figure S15 ) . This is also consistent with the finding of Weissleder et al . that a neutral 2 - COOH SiR650 derivative could follow the intracellular localization of a therapeutic drug in live cells . [ 23 \u2013 25 ] Notably , these ligand - 2 - COOH SiR650 conjugates show almost no fluorescence signal at non - targeted regions inside the cells , in accordance with the concept of using background - free \u201ctame\u201d fluorescent probes [ 26 ] to obtain clear fluorescence images in live cells . Our work extends the scope of this concept to in vivo fluorescence imaging of whole animals . Indeed , the effective blocking of nonspecific adsorption on non - targeted tissues enabled us to clearly visualize tumors within less than 30 min after probe injection . During the operation , at most 5 to 10 min may be allowed as an incubation time after probe injection , and FolateSiR - 1 has a potential to be used even during the operation . A similar approach might be applied to other tumor - related membrane proteins , such as carbonic anhydrase IX , which is overexpressed in hypoxic tumor tissues . [ 27 ] Moreover , the NIR probe OTL - 38 , which is based on the cyanine fluorophore , is currently in phase II clinical trial for intraoperative imaging of ovarian cancer . [ 28 \u2013 30 ] Our new probe FolateSiR - 1 might be superior to OTL - 38 in terms of high S / N ratio , and we believe it should be suitable for precise intraoperative fluorescence detection of tiny tumors in the millimeter size range , as well as for the efficient endoscopic detection of cancer cells in abdominal dropsy immediately after surgical operation . Recently , therapeutic antibodies have been applied to the treatment of various types of cancer . However , although it was confirmed to be safe , anti - FRs antibody did not meet the primary endpoints of progression - free survival or overall survival in a clinical trial . Therefore , anti - FRs antibody \u2013 drug conjugates ( ADCs ) have been de - veloped in order to enhance the efficacy . [ 31 , 32 ] However , poor penetration into solid tumors still poses a problem for ADCs as well as naked therapeutic antibodies . On the other hand , small - molecular drug conjugates are expected to overcome this issue . [ 33 ] The binding activity and specificity for FR of FolateSiR - 1 are thought to be high as well as anti - FRs antibodies , [ 3 ] and so FolateSiR - 1 might be available as a basic scaffold for the development of new small - molecular drug conjugates to treat FR - expressing ovarian cancer . Further - more , the theranostic features of such agents could be valuable to assess the pharmacokinetic profile within the tumor tissue . Acknowledgements This work was supported in part by JSPS KAKENHI Grant Numbers JP18H04609 and JP16H05099 to K . H . and JP26110005 to Y . Y . , SENTAN , JST to K . H , and Hoansha Foundation and Daiichi Sankyo Foundation of Life Science to K . H . This work was also supported by a grant JSPS Core - to - Core program , A . Advanced Research Networks and a Grant - in - Aid for Scientific Research on Innovative Areas \u201cSingularity Biology ( No . 8007 ) \u201d ( JP19H05414 to K . H . ) of The Ministry of Education , Culture , Sports , Science , and Technology , Japan . Conflict of interest The authors declare no conflict of interest . Keywords : cancer \u00b7 chromophores \u00b7 fluorescent probes \u00b7 imaging \u00b7 vitamins Figure 5 . a ) The ovarian cancer tissue microarray was incubated with 5 m m FolateSiR - 1 and 2 . 9 m m DAPI ( nuclear stain ) in phosphate - buffered saline containing 0 . 05 % Tween 20 ( PBST ) for 2 h . Then , the tissue microarray was washed with PBST three times and the fluores - cence image was obtained . The microarray contains 37 tumor tissues and 3 normal tissues . Fluorescence images of the 3 normal tissues and 3 representative tumor tissues are shown , scale bar = 2 mm . b ) Immunostaining of folate receptors in the tissue microarray . The immunostained 3 normal tissues and 3 tumor tissues correspond to those in ( a ) , scale bar = 2 mm . Angewandte Chemie Research Articles 6019 Angew . Chem . Int . Ed . 2020 , 59 , 6015 \u2013 6020 T 2020 Wiley - VCH Verlag GmbH & Co . KGaA , Weinheim www . angewandte . org 15213773 , 2020 , 15 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / a n i e . 201914826 by U n i v e r s it y O f W a s h i ng t on , W il e y O n l i n e L i b r a r y on [ 14 / 11 / 2023 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e How to cite : Angew . Chem . Int . Ed . 2020 , 59 , 6015 \u2013 6020 Angew . Chem . 2020 , 132 , 6071 \u2013 6076 [ 1 ] W . Xia , P . S . Low , J . Med . Chem . 2010 , 53 , 6811 \u2013 6824 . [ 2 ] Y . G . Assaraf , C . P . Leamon , J . A . Reddy , Drug Resist . Updates 2014 , 17 , 89 \u2013 95 . [ 3 ] P . S . Low , S . A . Kularatne , Curr . Opin . Chem . Biol . 2009 , 13 , 256 \u2013 262 . [ 4 ] K . R . Kalli , A . L . Oberg , G . L . Keeney , T . J . H . Christianson , P . S . Low , K . L . Knutson , L . C . Hartmann , Gynecol . Oncol . 2008 , 108 , 619 \u2013 626 . [ 5 ] S . Markert , S . Lassmann , B . Gabriel , M . Klar , M . Werner , G . Gitsch , F . Kratz , A . Hasenburg , Anticancer Res . 2008 , 28 , 3567 \u2013 3572 . [ 6 ] A . L . Vahrmeijer , M . Hutteman , J . R . van der Vorst , C . J . H . van de Velds , J . V . Frangioni , Nat . Rev . Clin . Oncol . 2013 , 10 , 507 \u2013 518 . [ 7 ] G . M . van Dam , G . Themelis , L . M . A . Crane , N . J . Harlaar , R . G . Pleijhuis , W . Kelder , A . Sarantopoulos , J . S . de Jong , H . J . G . Arts , A . G . J . van der Zee , J . Bart , P . S . Low , V . Ntziachristos , Nat . Med . 2011 , 17 , 1315 \u2013 1319 . [ 8 ] J . V . Frangioni , Curr . Opin . Chem . Biol . 2003 , 7 , 626 \u2013 634 . [ 9 ] R . Weissleder , V . Ntziachristos , Nat . Med . 2003 , 9 , 123 \u2013 128 . [ 10 ] W . K . Moon , Y . Lin , T . OQLoughlin , Y . Tang , D . E . Kim , R . Weissleder , C . H . Tung , Bioconjugate Chem . 2003 , 14 , 539 \u2013 545 . [ 11 ] C . H . Tung , Y . Lin , W . K . Moon , R . Weissleder , ChemBioChem 2002 , 3 , 784 \u2013 786 . [ 12 ] L . E . Kelderhouse , V . Chelvam , C . Wayua , S . Mahalingam , S . Poh , S . A . Kularatne , P . S . Low , Bioconjugate Chem . 2013 , 24 , 1075 \u2013 1080 . [ 13 ] Y . Urano , D . Asanuma , Y . Hama , Y . Koyama , T . Barrett , M . Kamiya , T . Nagano , T . Watanabe , A . Hasegawa , P . L . Choyke , H . Kobayashi , Nat . Med . 2009 , 15 , 104 \u2013 109 . [ 14 ] J . Yang , H . Chen , I . R . Vlahov , J . - X . Cheng , P . S . Low , J . Pharmacol . Exp . Ther . 2007 , 321 , 462 \u2013 468 . [ 15 ] J . Yang , H . Chen , I . R . Vlahov , J . - X . Cheng , P . S . Low , Proc . Natl . Acad . Sci . USA 2006 , 103 , 13872 \u2013 13877 . [ 16 ] H . S . Choi , S . L . Gibbs , J . H . Lee , S . H . Kim , Y . Ashitate , F . Liu , H . Hyun , G . L . Park , Y . Xie , S . Bae , M . Henary , J . V . Frangioni , Nat . Biotechnol . 2013 , 31 , 148 \u2013 153 . [ 17 ] C . Chen , J . Ke , X . E . Zhou , W . Yi , J . S . Brunzelle , J . Li , E . - L . Yong , H . E . Xu , K . Melcher , Nature 2013 , 500 , 486 \u2013 489 . [ 18 ] S . Gaw8da , G . Stochel , K . Szaci\u0142owski , Chem . Asian J . 2007 , 2 , 580 \u2013 590 . [ 19 ] J . A . Piedrahita , B . Oetama , G . D . Bennett , J . van Waes , B . A . Kamen , J . Richardson , S . W . Lacey , R . G . W . Anderson , R . H . Finnell , Nat . Genet . 1999 , 23 , 228 \u2013 232 . [ 20 ] V . Ramaekers , J . M . Sequeira , E . V . Quadros , Clin . Chem . Lab . Med . 2013 , 51 , 497 \u2013 511 . [ 21 ] H . Saitsu , M . Ishibashi , H . Nakano , K . Shiota , Dev . Dyn . 2003 , 226 , 112 \u2013 117 . [ 22 ] G . Lukinavic\u02c7ius , K . Umezawa , N . Olivier , A . Honigmann , G . Yang , T . Plass , V . Mueller , L . Reymond , I . R . Corr\u00dea , Jr . , Z . - G . Luo , C . Schultz , E . A . Lemke , P . Heppenstall , C . Eggeling , S . Manley , K . Johnsson , Nat . Chem . 2013 , 5 , 132 \u2013 139 . [ 23 ] M . A . Miller , E . Kim , M . F . Cuccarese , A . L . Plotkin , M . Prytyskach , R . H . Kohler , M . J . Pittet , R . Weissleder , Chem . Commun . 2018 , 54 , 42 \u2013 45 . [ 24 ] E . Kim , K . S . Yang , R . H . Kohler , J . M . Dubach , H . Mikula , R . Weissleder , Bioconjugate Chem . 2015 , 26 , 1513 \u2013 1518 . [ 25 ] E . Kim , K . S . Yang , R . J . Giedt , R . Weissleder , Chem . Commun . 2014 , 50 , 4504 \u2013 4507 . [ 26 ] S . H . Alamudi , R . Satapathy , J . Kim , D . Su , H . Ren , R . Das , L . Hu , E . Alvarado - Mart & nez , J . Y . Lee , C . Hoppmann , E . Pe\u00c7a - Cabrera , H . - H . Ha , H . - S . Park , L . Wang , Y . - T . Chang , Nat . Commun . 2016 , 7 , 11964 . [ 27 ] C . T . Supuran , Nat . Rev . Drug Discovery 2008 , 7 , 168 \u2013 181 . [ 28 ] C . E . S . Hoogstins , Q . R . J . G . Tummers , K . N . Gaarenstroom , C . D . de Kroon , J . B . M . Z . Trimbos , T . Bosse , V . T . H . B . M . Smit , J . Vuyk , C . J . H . van de Velde , A . F . Cohen , P . S . Low , J . Burggraaf , A . L . Vahrmeijer , Clin . Cancer Res . 2016 , 22 , 2929 \u2013 2938 . [ 29 ] E . de Jesus , J . J . Keating , S . A . Kularatne , J . Jiang , R . Judy , J . Predina , S . Nie , P . Low , S . Singhal , Int . J . Mol . Imaging 2015 , 469047 . [ 30 ] J . - Y . Winum , Expert Opin . Ther . Pat . 2016 , 26 , 1223 \u2013 1226 . [ 31 ] O . Ab , K . R . Whiteman , L . M . Bartle , X . Sun , R . Singh , D . Tavares , A . LaBelle , G . Payne , R . J . Lutz , J . Pinkas , V . S . Goldmacher , T . Chittenden , J . M . Lambert , Mol . Cancer Ther . 2015 , 14 , 1605 \u2013 1613 . [ 32 ] X . Cheng , J . Li , K . Tanaka , U . Majumder , A . Z . Milinichik , A . C . Verdi , C . J . Maddage , K . A . Rybinski , S . Fernando , D . Fernando , M . Kuc , K . Furuuchi , F . Fang , T . Uenaka , L . Grasso , E . F . Albone , Mol . Cancer Ther . 2018 , 17 , 2665 \u2013 2675 . [ 33 ] M . P . Deonarain , G . Yahioglu , I . Stamati , A . Pomowski , J . Clarke , B . M . Edwards , S . Diez - Posada , A . C . Stewart , Anti - bodies 2018 , 7 , 16 . Manuscript received : November 20 , 2019 Revised manuscript received : December 30 , 2019 Accepted manuscript online : January 26 , 2020 Version of record online : February 26 , 2020 Angewandte Chemie Research Articles 6020 www . angewandte . org T 2020 Wiley - VCH Verlag GmbH & Co . KGaA , Weinheim Angew . Chem . Int . Ed . 2020 , 59 , 6015 \u2013 6020 15213773 , 2020 , 15 , D o w n l o a d e d fr o m h tt p s : / / on li n e li b r a r y . w il e y . c o m / do i / 10 . 1002 / a n i e . 201914826 by U n i v e r s it y O f W a s h i ng t on , W il e y O n l i n e L i b r a r y on [ 14 / 11 / 2023 ] . S ee t h e T e r m s a nd C ond iti on s ( h tt p s : / / on li n e li b r a r y . w il e y . c o m / t e r m s - a nd - c ond iti on s ) on W il e y O n li n e L i b r a r y f o r r u l e s o f u s e ; OA a r ti c l e s a r e gov e r n e d by t h e a pp li ca b l e C r ea ti v e C o mm on s L i ce n s e",
    "yamada2016actin": "INTERNATIONAL JOURNAL OF ONCOLOGY 49 : 877 - 886 , 2016 Abstract . The endocytic protein dynamin participates in the formation of actin - based membrane protrusions such as podosomes , pseudopodia , and invadopodia , which facilitate cancer cell migration , invasion , and metastasis . However , the role of dynamin in the formation of actin - based membrane protrusions at the leading edge of cancer cells is unclear . In this study , we demonstrate that the ubiquitously expressed dynamin 2 isoform facilitates cell migration by stabilizing F - actin bundles in filopodia of the lung cancer cell line H1299 . Pharmacological inhibition of dynamin 2 decreased cell migration and filopodial formation . Furthermore , dynamin 2 and cortactin mostly colocalized along F - actin bundles in filopodia of serum - stimulated H1299 cells by immuno - fluorescent and immunoelectron microscopy . Knockdown of dynamin 2 or cortactin inhibited the formation of filopodia in serum - stimulated H1299 cells , concomitant with a loss of F - actin bundles . Expression of wild - type cortactin rescued the punctate - like localization of dynamin 2 and filopodial forma - tion . The incubation of dynamin 2 and cortactin with F - actin induced the formation of long and thick actin bundles , with these proteins colocalizing at F - actin bundles . A depolymer - ization assay revealed that dynamin 2 and cortactin increased the stability of F - actin bundles . These results indicate that dynamin 2 and cortactin participate in cell migration by stabilizing F - actin bundles in filopodia . Taken together , these findings suggest that dynamin might be a possible molecular target for anticancer therapy . Introduction Cancer cell migration , invasion , and metastasis are preceded by the formation of pseudopodia such as lamellipodia and filopodia . During these cellular processes , F - actin filaments remodel into a higher order structure and then assemble an intricate cytoskeletal network within cells ( 1 ) . These dynamic three - dimensional changes are mediated by several actin - bundling and crosslinking proteins , and are essential for supporting filopodia at the leading edge of migrating cells ( 2 ) . Dynamin plays an essential role in endocytosis , partici - pating in the membrane fission process ( 3 - 5 ) . Dynamin also functions in the formation of actin - rich structures , including lamellipodia and dorsal membrane ruffles ( 6 , 7 ) , invadopodia ( 8 ) , podosomes ( 9 ) , growth cones ( 10 - 12 ) , and phagocytic cups ( 13 , 14 ) . Three dynamin isoforms exist , namely , dynamin 1 , 2 , and 3 ( 5 ) . Dynamins are characterized by a GTPase domain at the N - terminus , a bundle signaling element , a stalk domain , a phosphoinositide - binding pleckstrin homology domain , and a proline and arginine - rich domain at the C - terminus ( PRD ) ( 15 , 16 ) . The PRD interacts with different proteins that contain the Src - homology - 3 ( SH3 ) domain . Of these GTPases , dynamin 2 is ubiquitously expressed . Cortactin , an F - actin - binding protein , was first identified as an Src substrate ( 17 ) . Cortactin also participates in cancer cell migration , invasion , and metastasis by regulating actin dynamics at the leading edge of migrating cells ( 18 ) . Cortactin is composed of an N - terminal acidic domain and a six - and - a - half tandem repeats domain , which directly binds to F - actin . Cortactin also contains an \u03b1 - helix , a proline - rich region , and an SH3 domain at the C - terminus , which interacts with the PRD of several binding partners ( 19 ) . Both dynamin and cortactin are implicated in the dynamics of cancer cells , including migration , invasion , and metastasis ( 18 ) . In addition , the pharmacological inhibition of dynamin by GTPase inhibitors suppresses specific cellular processes such as the lamellipodial formation and invasion of human osteocarcinoma cells ( 20 ) and the growth of human prostate adenocarcinoma cells ( 21 ) . Actin bundling by dynamin 2 and cortactin is implicated in cell migration by stabilizing filopodia in human non - small cell lung carcinoma cells HIROSHI YAMADA 1 , 3 , TETSUYA TAKEDA 1 , 3 , HIROYUKI MICHIUE 2 , TADASHI ABE 1 , 3 and KOHJI TAKEI 1 , 3 Departments of 1 Neuroscience and 2 Physiology , Graduate School of Medicine , Dentistry and Pharmaceutical Sciences , Okayama University ; 3 CREST , Japan Science and Technology Agency , Kita - ku , Okayama 700 - 8558 , Japan Received April 6 , 2016 ; Accepted May 25 , 2016 DOI : 10 . 3892 / ijo . 2016 . 3592 Correspondence to : Dr Hiroshi Yamada , Department of Neuro - science , Graduate School of Medicine , Dentistry and Pharmaceutical Sciences , Okayama University , 2 - 5 - 1 Shikata - cho , Kita - ku , Okayama 700 - 8558 , Japan E - mail : hiroyama @ md . okayama - u . ac . jp Key words : actin , cortactin , dynamin , filopodia , migration YAMADA et al : ACTIN BUNDLING BY DYNAMIN AND CORTACTIN IN CANCER CELL MIGRATION 878 A previous study reported that dynamin 2 binds to cortactin ( 7 , 12 ) . A disruption of this protein complex can affect the shape of cancer cells ( 7 ) , organization of the F - actin network within these cells ( 22 ) , and structure of growth cones ( 11 , 12 ) . However , the role of the dynamin 2 - cortactin complex in the dynamics of the actin cytoskeleton in cancer cells is unclear . In this study , we investigated whether dynamin 2 and cortactin regulate the F - actin bundle formation in filopodia in the human non - small cell lung carcinoma cell line H1299 . Materials and methods Antibodies and reagents . Rabbit polyclonal anti - dynamin 1 ( cat . no . PA1 - 660 ; Thermo Fisher Scientific , Waltham , MA , USA ) and anti - c - myc ( cat . no . C3956 ; Sigma - Aldrich , St . Louis , MO , USA ) antibodies , and a goat polyclonal anti - dynamin 2 ( cat . no . sc - 6400 ; Santa Cruz Biotechnology , Santa Cruz , CA , USA ) antibody , were purchased . In addition , mouse monoclonal anti - \u03b2 - actin ( cat . no . A5441 , Sigma - Aldrich ) , Dynasore ( cat . no . D7693 , Sigma - Aldrich ) , anti - c - myc ( cat . no . sc - 40 ; Santa Cruz Biotechnology ) , anti - green fluorescent protein ( GFP ; cat . no . sc - 9996 , Santa Cruz Biotechnology ) , and anti - cortactin ( cat . no . 05 - 180 ; EMD Millipore , Darmstadt , Germany ) antibodies were purchased . MitMAB and Dynole 34 - 2 were purchased from Abcam Biochemicals ( Bristol , UK ) . Alexa Fluor 488 - conjugated anti - rabbit IgG , rhodamine - conjugated anti - mouse IgG , and rhodamine or Alexa Fluor 488 - labeled phalloidin were obtained from Thermo Fisher Scientific . Purified rabbit skeletal \u03b1 - actinin was purchased from Cytoskeleton , Inc . ( Denver , CO , USA ) . Goat anti - mouse IgG - and goat anti - rabbit IgG - conjugated gold particles were purchased from British BioCell International ( Cardiff , UK ) . Cell culture . The human non - small cell lung carcinoma cell line H1299 ( Cat . no . ATCC CRL - 5803 ; American Type Culture Collection , Manassas , VA , USA ) was cultured in Dulbecco ' s modified Eagle ' s medium ( DMEM , Thermo Fisher Scientific ) supplemented with 10 % fetal bovine serum ( FBS ) at 37\u02daC in an atmosphere of 5 % CO 2 . Expression and purification of dynamin 2 and cortactin wild - types and mutants . GFP - tagged dynamin 2 cloned into pEGFP - N1 was a kind gift from Dr Mark McNiven ( Mayo Clinic , Rochester , MN , USA ) ( 6 ) . His - tagged dynamin 2 produced with the Bac - to - Bac baculovirus expression system ( Thermo Fisher Scientific ) was a kind gift from Dr Hiroshi Handa ( Tokyo Institute of Technology , Tokyo , Japan ) ( 23 ) . The dynamin solution was concentrated using a Centriplus YM50 ( Thermo Fisher Scientific ) and stored at - 80\u02daC . The protein suspension ( 2 - 5 mg / ml protein ) was thawed at 37\u02daC before use . The cDNAs encoding full - length rat cortactin and its mutants were prepared by polymerase chain reaction ampli - fication using specific primers ( 12 ) . Full - length cortactin or 1 - 450aa ( Cort \u2206 SH3 ) was subcloned into the plasmid pGEX - 6p vector as Bam HI - Eco RI fragments . GST - tagged cortactin W525K was generated by mutating pGEX - 6p - cortactin with the QuickChange site - directed mutagenesis kit ( Agilent Technologies , Santa Clara , CA , USA ) . For expression in cells , full - length cortactin or Cort \u2206 SH3 was subcloned into the pEF1 myc - His vector ( Thermo Fisher Scientific ) as Eco RI - Xba I fragments . The nucleotide sequences of the constructs were verified by DNA sequence analysis . The resulting plasmid was transformed into the bacterial BL21 ( DE3 ) pLysS strain for protein expression . The expression of GST - fusion proteins was induced by 0 . 1 mM isopropyl - 1 - thio - d - galactopyranoside at 37\u02daC for 3 - 6 h in LB medium supplemented with 100 \u00b5g / ml ampicillin to A 600 = 0 . 8 . The purification of GST - fusion proteins was performed as previously described ( 24 ) , and the cleavage of the GST with PreScission protease was performed according to the manufacturer ' s instructions . The protein was purified on a MonoQ column equilibrated in 20 mM Tris - HCl and 0 . 2 M NaCl , pH 7 . 7 . The eluted protein fraction ( 1 mg / ml protein ) was stored at - 80\u02daC . For the pull - down assay , the proteins were used without cleaving GST . siRNA - mediated interference . Pre - annealed siRNAs for human dynamin 2 and cortactin , and the negative control siRNA , were synthesized and purified ( Thermo Fisher Scientific ) . The sequences for the siRNAs for human dynamin 2 were as follows : 5 ' - GGAUAUUGAGGGCAAGAAGtt - 3 ' ( sense ) , 5 ' - CUUCUUGCCCUCAAUAUCCtt - 3 ' ( antisense ) for oligo 1 ; 5 ' - GCGAAUCGUCACCACUUACtt - 3 ' ( sense ) , 5 ' - GUAAGUG GUGACGAUUCGCtc - 3 ' ( antisense ) for oligo 2 ; and 5 ' - GGAC UUACGACGGGAGAUCtt - 3 ' ( sense ) , 5 ' - GAUCUCCCGU CGUAAGUCCtt - 3 ' ( antisense ) for oligo 3 . The sequences for the siRNAs for human cortactin were as follows : CCGAAUG GAUAAGUCAGUCtt - 3 ' ( sense ) , 5 ' - AGCUGACUUAUCCAU UCGGtc - 3 ' ( antisense ) for oligo 1 ; GGUUUCGGCGGCA AAUACGtt - 3 ' ( sense ) , CGUAUUUGCCGCCGAAACCtt - 3 ' ( antisense ) for oligo 2 ; and CGAAUAUCAGUCGAAACUUtt - 3 ' ( sense ) , AAGUUUCGACUGAUAUUCGtg - 3 ' ( antisense ) for oligo 3 . A scrambled siRNA with no significant sequence homology to all mouse , rat , or human gene sequences was used as the negative control . The day before transfection , the cells were plated in 6 - well plates ( 5x10 4 cells / well ) . One hundred picomoles of the duplex siRNAs was transfected into the cells using 4 \u00b5l of Lipofectamine 2000 ( Thermo Fisher Scientific ) . After 72 h , the cells were treated differently according to experimental design . In pilot experiments , we confirmed that all three transfections of siRNA for dynamin 2 and cortactin were effective . Filopodial formation . H1299 cells were serum - starved for 16 h . Thereafter , the cells were transfected with dynamin 2 siRNAs , cortactin siRNAs , or the control siRNA , followed by incubation with DMEM supplemented with 10 % FBS for 45 min . For the rescue experiments , cortactin was silenced in H1299 cells with oligo 3 , and the cells were cultured for 24 h . The cells ( 1x10 5 / coverslip ) were then transfected with rat wild - type cortactin or cortactin W525K ( 0 . 25 \u00b5g each ) cloned into the pIRES2 - AcGFP1 expression vector ( Clontech Laboratories , Santa Clara , CA , USA ) . Thereafter , the cells were stimulated with serum for 45 min , fixed , and stained with Alexa Fluor 488 or rhodamine - labeled phalloidin for visualization of filopodia . Wound healing assay . H1299 cells were cultured to confluence on glass - bottom dishes ( 35 mm diameter ; AGC Techno Glass Co . Ltd . , Tokyo , Japan ) in DMEM supplemented with 0 . 2 % FBS INTERNATIONAL JOURNAL OF ONCOLOGY 49 : 877 - 886 , 2016 879 for 8 h . Thereafter , the cell layer was wounded with a plastic pipette tip as previously described ( 25 ) . The cells were washed with DMEM supplemented with 0 . 2 % FBS and incubated for 8 h in the presence of Dynasore , Dynole 34 - 2 or MitMAB at the indicated concentrations . For the negative control , cells were incubated with 1 % dimethyl sulfoxide ( DMSO ) . The cells were visualized by Giemsa staining , followed by the acquisi - tion of phase contrast images from \u226520 randomly selected areas per dish . Areas filled with migrating cells were analyzed with ImageJ software ( National Institutes of Health , Bethesda , MD , USA ) . Formation of in vitro F - actin bundles . For the fluorescent detection of F - actin , non - muscle actin was polymerized in F - buffer ( 10 mM Tris - HCl , 0 . 2 mM DTT , 0 . 2 mM CaCl 2 , 2 mM MgCl 2 , 50 mM KCl , and 0 . 5 mM ATP , pH 7 . 5 ) for 1 h . Thereafter , 3 . 3 \u00b5M F - actin was incubated with 5 \u00b5M dynamin 1 or 2 and cortactin for 1 h , followed by an additional 30 min with 3 \u00b5M Alexa Fluor 488 - phalloidin . The samples were spread onto glass slides and mounted , and the F - actin bundles were observed under an epifluorescent microscope . For the immunolocalization of dynamin and cortactin , F - actin bundles were incubated with dynamin 1 or 2 and cortactin for 30 min with 3 \u00b5M phalloidin to stabilize the filaments , followed by centrifugation at 5 , 000 x g for 10 min . The pellet was resuspended with 50 \u00b5l of F - buffer and then immunostained in suspension for 30 min with 1 \u00b5l of primary antibody . The mixture was centrifuged at 5 , 000 x g , and the pellet was washed with F - buffer . The samples were incubated with secondary antibodies and washed as previously done for the primary antibody . All steps were performed at room temperature . The samples were spread onto glass slides and mounted . The samples were examined under a spinning disc confocal microscope system ( CSU10 , Yokogawa Electric Co . , Tokyo , Japan ) combined with an inverted microscope ( IX - 71 , Olympus Optical Co . , Ltd . , Tokyo , Japan ) and a CoolSNAP - HQ camera ( Roper Technologies , Sarasota , FL , USA ) . The confocal system was controlled by MetaMorph software ( Molecular Devices , Sunnyvale , CA , USA ) . Images were processed using Adobe Photoshop CS3 or Illustrator CS3 software . Immunoprecipitation assay . For the immunoprecipitation assay , H1299 cells were co - transfected with GFP - tagged dynamin 2 and either myc - tagged cortactin or cortactin \u2206 SH3 . The cells were lysed with 1 % NP - 40 , 100 mM KCl , 0 . 5 mM EDTA , 10 mM NaF , and 20 mM HEPES / KOH , pH 7 . 4 , and a protease inhibitor cocktail tablet ( Roche Diagnostics , Basel , Switzerland ) . The protein complexes were immunoprecipitated from 1 mg of cell extract using either 5 \u00b5g of the polyclonal anti - myc antibodies or preimmune IgG , and then visualized by western blotting with a monoclonal anti - GFP or anti - myc antibody . Immunostaining and fluorescent microscopy . H1299 cells were fixed with 4 % paraformaldehyde and stained by immu - nofluorescence as previously described ( 12 ) . Transmission electron microscopy . Specimens were embedded for immunoelectron microscopy as previously described ( 12 ) . In brief , H1299 cells were fixed with cytoskeleton buffer ( 10 mM 2 - ( N - morpholino ) ethanesulfonic acid , 150 mM NaCl , 5 mM EGTA , 5 mM MgCl 2 , and 5 mM glucose , pH 6 . 0 ) containing 10 \u00b5g / ml phalloidin , 0 . 1 % Triton X - 100 , and 3 % formaldehyde for 1 min . The cells were then fixed for an additional 30 min without Triton X - 100 , followed by washing with 5 \u00b5g / ml phal - loidin in phosphate - buffered saline ( PBS ) . After incubation in blocking solution ( 10 \u00b5g / ml phalloidin , 2 mg / ml BSA , and 100 mM glycine in PBS ) , the specimens were incubated with a primary antibody diluted in blocking solution , washed with 5 \u00b5g / ml phalloidin in PBS , incubated with goat anti - mouse or rabbit anti - goat IgG conjugated to 10 - nm gold particles , and then fixed with 2 . 5 % glutaraldehyde and 5 \u00b5g / ml phalloidin in PBS . The specimens were post - fixed with 1 % OsO 4 in 0 . 1 M sodium cacodylate buffer for 1 h , dehydrated , and embedded in EPON 812 for ultrathin sectioning . Cross - sections were visualized under a Hitachi H - 7100 transmission electron microscope . Determination of filopodial length . For the measurement of the filopodial length , H1299 cells were fixed and stained with rhodamine - or Alexa Fluor 488 - conjugated phalloidin . Membrane protrusions supported with F - actin bundles were defined as filopodia , and digital images were acquired at 400 - 1 , 000x magnifications . Up to five filopodia for each cell were randomly selected , and their lengths were measured with ImageJ software . Statistical analysis . Data were analyzed for statistical significance using KaleidaGraph software ( version 4 . 1 ) for the Macintosh ( Synergy Software Inc . , Essex Junction , VT , USA ) . Analysis of variance and Tukey ' s honest significant difference post hoc test were applied for more than two different groups , and Student ' s t - test was applied for two different groups . P - values of < 0 . 05 and 0 . 001 were considered as statistically significant . Results Inhibition of dynamin decreases the migration of the human non - small cell lung carcinoma cell line H1299 . To determine whether dynamin 2 is involved in cell migration , the effects of dynamin inhibition on cell migration were determined by a wound healing assay . Cell migration decreased after treatment of cells with Dynasore ( 26 ) , Dynole 34 - 2 ( 27 ) , and MitMAB ( 28 ) ( Fig . 1A ) . Dynasore ( 80 \u00b5M ) inhibited cell migration by ~ 80 % compared to that of control cells , whereas Dynole 34 - 2 and MitMAB inhibited cell migration by 20 - 40 % ( Fig . 1B ) . These results indicate that dynamin is important for the migra - tion of H1299 cells . Dynamin 2 colocalizes with cortactin along F - actin bundles in filopodia of H1299 cells . Fig . 1 shows that dynamin is involved in cell migration mediated by pseudopodia . Thus , we investi - gated whether dynamin 2 participates in filopodial formation in H1299 cells . Because cortactin functions with dynamin 1 in the bundling of F - actin , which is important for the stability of filopodia in human neuroblastoma cell line SH - SY5Y ( 12 ) , dynamin 2 and cortactin were immunostained in serum - stimu - lated H1299 cells . H1299 cells formed numerous filopodia YAMADA et al : ACTIN BUNDLING BY DYNAMIN AND CORTACTIN IN CANCER CELL MIGRATION 880 after serum stimulation ( Fig . 2A ) . Furthermore , dynamin 2 and cortactin colocalized to F - actin bundles in filopodia as bright dots ( Fig . 2A ) . The negative controls showed little immunoreactivity for dynamin 2 and cortactin ( Fig . 2B ) . In addition , immunoelectron microscopy revealed that both proteins localized to F - actin bundles in filopodia ( Fig . 2C ) . These results prompted us to examine the possible inter - action of dynamin 2 and cortactin by immunoprecipitation . Exogenously expressed dynamin 2 - GFP was co - precipitated with full - length cortactin - myc using a polyclonal anti - myc anti - bodies and H1299 cell lysates ( Fig . 2D , left ) . Cort \u2206 SH3 - myc , a dynamin 2 binding deficient mutant that lacks its SH3 domain , was unable to precipitate dynamin 2 ( Fig . 2D , right ) . Taken together , these results illustrate that these proteins interact at F - actin bundles in filopodia of H1299 cells . Dynamin 2 and cortactin are required for serum - induced filopodial formation in H1299 cells . To examine the role of dynamin 2 in filopodial formation , dynamin 2 was silenced in H1299 cells by RNAi . Compared with the control , knockdown of dynamin 2 in H1299 cells with specific siRNAs reduced its level by ~ 95 % as revealed by western blotting ( Fig . 3A ) . Compared with the length of filopodia in serum - stimulated control cells ( 10 . 2\u00b10 . 5 \u00b5m ) , dynamin 2 knockdown decreased filopodial extension in silenced cells ( 4 . 7\u00b10 . 6 \u00b5m ) ( Fig . 3B and C ) . In addition , dynasore inhibited filopodial extension ( 2 . 4\u00b10 . 08 \u00b5m ) . This effect was rescued after the inhibitor was removed ( 8 . 1\u00b12 . 4 \u00b5m ) ( Fig . 3D and E ) . We also examined the effects of cortactin knockdown by RNAi on filopodial formation . Compared with the control , knockdown of cortactin reduced its level by ~ 95 % as revealed by western blotting ( Fig . 4A ) . Compared with the length of filopodia in control cells ( 10 . 2\u00b10 . 39 \u00b5m ) , cortactin knockdown also decreased filopodial extension after serum - stimulation ( 5 . 6\u00b10 . 17 \u00b5m ) ( Fig . 4B and C ) . The inhibition of filopodial formation in cortactin - silenced cells was rescued by exogenous expression of wild - type cortactin ( 10 . 8\u00b10 . 54 \u00b5m ) but not by cortactin W525K , a binding - defective mutant of dynamin 2 ( 29 ) ( Fig . 4D and E ) . In addition , the punctate - like localization of dynamin 2 along F - actin bundles reappeared in wild - type cortactin expressing cells ( Fig . 4F , right ) . These results indi - cate that dynamin 2 and cortactin are required for filopodial formation . F - actin bundling by the dynamin 2 - cortactin complex stabi - lizes F - actin . The effects of dynamin 2 and cortactin on the formation of F - actin bundles were examined in vitro . In this experiment , preformed F - actin were incubated with or without cortactin and dynamin 2 in the presence of GTP . F - actin alone appeared as uniform filaments ( Fig . 5A , actin alone ) . The addition of dynamin 2 to F - actin filaments did not cause any visible change in their distribution ( Fig . 5A , + Dyn2 ) . However , F - actin incubated with wild - type cortactin , cortactin W525K or cortactin \u2206 SH3 often formed small clusters ( Fig . 5A , + Cort WT , + Cort W525K , or + Cort \u2206 SH3 ) , consistent with a previ - ously published report ( 12 ) . The presence of both dynamin 2 Figure 1 . Dynamin GTPase inhibitors inhibit the migration of H1299 cells . ( A ) Representative images acquired by light microscopy showing cell migration in a wound healing assay . Confluent H1299 cells were wounded and then incubated for 8 h in the presence or absence of dynamin GTPase inhibitors at the indicated concentrations . For the negative control ( control 0 h ) , cells were incubated with 1 % DMSO . Scale bar , 200 \u00b5m . ( B ) Morphometric analysis of the wound area filled by migrating cells after treatment with inhibitors at the indicated concentrations . The changes were normalized to the control . Results represent the means \u00b1 SEM of three independent experiments . INTERNATIONAL JOURNAL OF ONCOLOGY 49 : 877 - 886 , 2016 881 and wild - type cortactin resulted in the formation of long and thick F - actin bundles ( Fig . 5A , + Dyn2 + Cort WT ) , which were similar to those formed by dynamin 1 and cortactin ( Fig . 5A , + Dyn1 + Cort WT ) . On the other hand , the long and thick F - actin bundles were much less evident in the presence of dynamin 2 and cortactin W525K or \u2206 SH3 ( Fig . 5A , + Dyn2 + Cort W525K or + Dyn2 + Cort \u2206 SH3 ) . To localize dynamin 2 and cortactin to F - actin bundles , the preformed F - actin bundles were used for immunofluores - cent staining . Dynamin 2 and cortactin colocalized as bright dots along F - actin bundles ( Fig . 5B , left ) . The localiza - tion of dynamin 2 and cortactin was similar to that of the dynamin 1 - cortactin complex ( Fig . 5B , right ) ( 12 ) . Lastly , we examined whether actin bundling by the dynamin 2 - cortactin complex can affect F - actin stability . To address this , the depolymerization kinetics of preformed pyrene - labeled F - actin were examined after the solution was diluted 10 - fold with buffer . In the presence of dynamin 2 and cortactin , the rate of depolymerization by dilution decreased to a level comparable to that induced by \u03b1 - actinin , an actin - crosslinking protein , indicating that dynamin 2 and cortactin stabilize F - actin bundles ( Fig . 5C ) . These results indicate that the dynamin 2 - cortactin complex stabilizes F - actin bundles in filopodia prior to cell migration . Discussion The involvement of dynamin in the dynamics of cancer cells such as cell migration , invasion , and metastasis has been reported ( 18 ) . However , the precise role of dynamin in these Figure 2 . Dynamin 2 colocalizes with cortactin along F - actin bundles in filopodia of serum - stimulated H1299 cells . ( A ) Colocalization of dynamin 2 ( Dyn2 , left ) and cortactin ( Cort , middle ) by double - immunofluorescent staining in filopodia of serum - stimulated H1299 cells . Boxed areas correspond to enlarged images shown below . Dynamin 2 - and cortactin - positive puncta were present periodically along F - actin bundles in filopodia ( arrowheads ) . Scale bar , 5 \u00b5m ( upper panels ) , 1 . 6 \u00b5m ( lower panels ) . ( B ) In the negative controls , the primary antibodies were omitted for dynamin 2 ( left ) and cortactin ( right ) . Boxed areas correspond to enlarged images shown below . Bar , 10 \u00b5m ( top and bottom panels ) , 2 . 8 \u00b5m ( middle panels ) . ( C ) Representative images acquired by immunoelectron microscopy showing the localization of dynamin 2 ( top three panels ) and cortactin ( bottom three panels ) in filopodia of serum - stimulated H1299 cells . Immunoreactive dynamin 2 and cortactin were present along F - actin bundles ( arrowheads ) . Scale bar , 20 nm . ( D ) Immunoprecipitation ( IP ) results demonstrating an in vivo interaction between dynamin 2 and cortactin . H1299 cells were co - transfected with GFP - tagged dynamin 2 ( Dyn2 - GFP ) and either myc - tagged wild - type cortactin ( Cort WT - myc , left ) or cortactin \u2206 SH3 ( Cort \u2206 SH3 - myc , right ) . The protein complexes were immunoprecipitated using a polyclonal anti - myc antibody or preimmune IgG ( IgG ) , and then visualized by western blotting ( WB ) with monoclonal anti - GFP or anti - myc antibodies . Total cell lysates ( 4 , 10 and 20 \u00b5g ) were also analyzed ( input ) . YAMADA et al : ACTIN BUNDLING BY DYNAMIN AND CORTACTIN IN CANCER CELL MIGRATION 882 cellular processes is not entirely clear . We recently reported that actin bundling by the dynamin 1 - cortactin complex is crucial for neurite extension in developing neurons ( 12 ) . In this study , we examined the possibility that a similar F - actin - bundling mechanism is involved in the migration of H1299 cells , a human non - small cell lung carcinoma cell line . We showed that cortactin and dynamin 2 mostly colocal - ized along F - actin bundles in filopodia of serum - stimulated H1299 cells ( Fig . 2 ) . Pharmacological inhibition of dynamin 2 by Dynasore , Dynole 34 - 2 or MitMAB decreased cell migra - tion ( Fig . 1 ) and filopodial formation ( Fig . 3 ) . Furthermore , filopodia were shorter in dynamin 2 - and cortactin - depleted cells than in control cells ( Figs . 3 and 4 ) . In cortactin - silenced cells , the exogenous expression of wild - type cortactin rescued the punctate - like localization of dynamin 2 and filopodial formation ( Fig . 4 ) . Both dynamin 2 and cortactin bundled Figure 3 . Knockdown of dynamin 2 decreases filopodial formation in H1299 cells . ( A ) Western blotting showing knockdown of dynamin 2 ( Dyn2 ) expression by RNAi in H1299 cells . \u03b2 - actin served as the control . Three micrograms of cell lysate from each sample was analyzed by gel electrophoresis . ( B ) F - actin was visualized in H1299 cells by Alexa Fluor 488 - phalloidin staining after knockdown of dynamin 2 . Extensive filopodial formation was observed in cells after serum stimulation ( left ) . Filopodial formation was inhibited in dynamin 2 - silenced cells ( right ) . Boxed areas correspond to enlarged images shown below . Scale bar , 20 \u00b5m ( upper panels ) , 5 \u00b5m ( lower panels ) . ( C ) Filopodial length in H1299 cells cultured in the presence or absence of serum . The cells were visual - ized by fluorescent confocal microscopy , and filopodial length was measured as described in Materials and methods . ( D ) Inhibition of filopodial formation by dynasore in serum - stimulated H1299 cells . Serum - starved cells were incubated with 240 \u00b5M dynasore for 30 min , and then stimulated with 10 % FBS for 45 min in the presence of 240 \u00b5M dynasore ( middle ) . Thereafter , dynasore was removed , and the cells were incubated in serum - containing medium for 45 min ( right ) . For the negative control , cells were cultured in the presence of 3 % DMSO ( left ) . All steps were performed at 37\u02daC . ( E ) Analysis of filopodial formation in the H1299 cells shown in ( D ) . The cells were analyzed by fluorescent confocal microscopy , and filopodial length was measured . Results in ( C ) and ( E ) represent the means \u00b1 SEM from three independent experiments . INTERNATIONAL JOURNAL OF ONCOLOGY 49 : 877 - 886 , 2016 883 Figure 4 . Knockdown of cortactin decreases filopodial formation in H1299 cells . ( A ) Western blotting showing knockdown of cortactin expression by RNAi in H1299 cells . \u03b2 - actin was used as the control . Three micrograms of cell lysate from each sample was analyzed by gel electrophoresis . ( B ) F - actin was visualized in serum - stimulated H1299 cells by Alexa Fluor 488 - phalloidin staining . Boxed areas correspond to enlarged images shown below . Similar to results from dynamin 2 - depleted cells , filopodial formation decreased in cortactin - depleted cells ( right ) . Scale bar , 20 \u00b5m ( upper panels ) , 5 \u00b5m ( lower panels ) . ( C ) Analysis of filopodial formation in H1299 cells cultured with or without serum . The samples were analyzed by fluorescent confocal microscopy , and the filopodial length was measured . ( D ) Expression of wild - type cortactin rescues filopodial formation . Cortactin - depleted H1299 cells were transfected with rat wild - type cortactin ( left ) or cortactin W525K ( right ) cloned into the pIRES2 - AcGFP1 expression vector . Boxed areas correspond to enlarged images shown ( middle panels ) . Transfected cells were identified by GFP expression ( bottom panels ) . Scale bar , 20 \u00b5m ( top and bottom panels ) , 3 . 5 \u00b5m ( middle panels ) . ( E ) Analysis of filopodial formation in H1299 cells . The samples were analyzed by fluorescent confocal microscopy , and the filopodial length was measured . Results in ( C ) and ( E ) represent the means \u00b1 SEM from three independent experiments . ( F ) Rescue of the punctate - like localization of dynamin 2 along F - actin bundles in filopodia by re - expression of cortactin in cortactin - depleted cells . Cortactin - depleted H1299 cells ( left panels ) were transfected with rat wild - type cortactin or cortactin W525K cloned into the pIRES2 - AcGFP1 expression vector ( right panels ) . The cells were immunostained with an anti - dynamin 2 antibodies . Boxed areas correspond to enlarged images shown below . Transfected cells were identified by GFP expression ( right bottom panels ) . Scale bar , 5 \u00b5m ( top and right bottom panels ) , 2 . 3 \u00b5m ( left bottom and right middle panels ) . YAMADA et al : ACTIN BUNDLING BY DYNAMIN AND CORTACTIN IN CANCER CELL MIGRATION 884 F - actin , and these proteins increased F - actin stability ( Fig . 5 ) . These results indicate that dynamin 2 and cortactin partici - pate in cancer cell migration by stabilizing F - actin bundles in filopodia . Dynamin assembles at the neck of deeply invaginated endocytic pits ( 30 ) . Upon GTP hydrolysis , however , dynamin undergoes a conformational change , resulting in the fission of endocytic pits and release of endocytic vesicles ( 31 - 33 ) . In addition , dynamin 1 forms a ring - like complex with cortactin , which switches from an open to a closed state upon GTP hydrolysis . This change promotes the bundling of F - actin filaments ( 12 ) . The mechanism of actin bundling mediated by the dynamin 2 - cortactin complex is similar to that of the dynamin 1 - cortactin complex , because dynamin 2 and cortactin also facilitated the formation of long and thick F - actin bundles to which they colocalized ( Fig . 5 ) . This mechanochemical Figure 5 . Actin bundling by dynamin 2 and cortactin stabilizes F - actin bundles . ( A ) Long F - actin bundles were formed in the presence of dynamin 2 and cortactin ( lower left ) . Preformed F - actin ( 3 . 3 \u00b5M ) was incubated with or without the indicated proteins ( 5 \u00b5M each ) . F - actin was visualized with Alexa Fluor 488 - phalloidin . Scale bar , 30 \u00b5m . ( B ) Representative images acquired by fluorescent microscopy showing the localization of dynamin and cortactin along F - actin bundles . Actin bundles were formed in vitro by incubating dynamin 2 with wild - type cortactin ( left ) or dynamin 1 and wild - type cortactin ( right ) . Protein colocalization was performed by double - immunofluorescence . Scale bar , 2 \u00b5m . ( C ) Kinetics of F - actin disassembly induced in 10 - fold diluted preformed pyrene - labeled F - actin solution with buffer . F - actin bundles disassembled in the presence of dynamin 2 and cortactin , as well as in the presence of 5 \u00b5M \u03b1 - actinin . The rate of F - actin bundle disassembly was measured by pyrene - fluorescence . INTERNATIONAL JOURNAL OF ONCOLOGY 49 : 877 - 886 , 2016 885 property may be critical for the formation of F - actin bundles in filopodia of other cell types as well ( Fig . 6 ) . Additional studies are needed to determine the precise mechanism . Dynamin associates with tumorigenesis , particularly tumor cell migration and invasion . For example , increased dynamin 2 expression potentiates the migration and invasion of pancreatic ductal cancer cells ( 25 ) , and tyrosine phosphorylated dynamin 2 promotes the growth and invasiveness of glioblastomas ( 34 ) . Thus , the involvement of dynamin in the formation of F - actin bundles might promote cancer malignancy . In conclusion , we showed that dynamin 2 and cortactin participate in the formation of F - actin bundles , which stabi - lize filopodia in migrating cancer cells . Taken together , these results suggest that dynamin might be a potential molecular target for anticancer therapy . Acknowledgements The authors thank Yuki Masuoka , Dr Shun - AI Li , and Nana Okazaki for technical assistance . This study was supported in part by grants from the Ministry of Education , Science , Sports , and Culture of Japan ( grant no . 26670201 to H . Y . ; grant no . 15K1533007 to K . T . ) , the Astellas Foundation for Research on Metabolic Disorders ( to H . Y . ) , and the Japan Foundation for Applied Enzymology ( to H . Y . ) . References 1 . Arjonen A , Kaukonen R and Ivaska J : Filopodia and adhesion in cancer cell motility . Cell Adhes Migr 5 : 421 - 430 , 2011 . 2 . Ridley AJ : Life at the leading edge . Cell 145 : 1012 - 1022 , 2011 . 3 . Takei K , Slepnev VI , Haucke V and De Camilli P : Functional partnership between amphiphysin and dynamin in clathrin - mediated endocytosis . Nat Cell Biol 1 : 33 - 39 , 1999 . 4 . Mettlen M , Pucadyil T , Ramachandran R and Schmid SL : Dissecting dynamin ' s role in clathrin - mediated endocytosis . Biochem Soc Trans 37 : 1022 - 1026 , 2009 . 5 . Praefcke GJ and McMahon HT : The dynamin superfamily : Universal membrane tubulation and fission molecules ? Nat Rev Mol Cell Biol 5 : 133 - 147 , 2004 . 6 . Cao H , Garcia F and McNiven MA : Differential distribution of dynamin isoforms in mammalian cells . Mol Biol Cell 9 : 2595 - 2609 , 1998 . 7 . McNiven MA , Kim L , Krueger EW , Orth JD , Cao H and Wong TW : Regulated interactions between dynamin and the actin - binding protein cortactin modulate cell shape . J Cell Biol 151 : 187 - 198 , 2000 . 8 . Baldassarre M , Pompeo A , Beznoussenko G , Castaldi C , Cortellino S , McNiven MA , Luini A and Buccione R : Dynamin participates in focal extracellular matrix degradation by invasive cells . Mol Biol Cell 14 : 1074 - 1084 , 2003 . 9 . Ochoa GC , Slepnev VI , Neff L , Ringstad N , Takei K , Daniell L , Kim W , Cao H , McNiven M , Baron R , et al : A functional link between dynamin and the actin cytoskeleton at podosomes . J Cell Biol 150 : 377 - 389 , 2000 . 10 . Torre E , McNiven MA and Urrutia R : Dynamin 1 antisense oligonucleotide treatment prevents neurite formation in cultured hippocampal neurons . J Biol Chem 269 : 32411 - 32417 , 1994 . 11 . Kurklinsky S , Chen J and McNiven MA : Growth cone morphology and spreading are regulated by a dynamin - cortactin complex at point contacts in hippocampal neurons . J Neurochem 117 : 48 - 60 , 2011 . 12 . Yamada H , Abe T , Satoh A , Okazaki N , Tago S , Kobayashi K , Yoshida Y , Oda Y , Watanabe M , Tomizawa K , et al : Stabilization of actin bundles by a dynamin 1 / cortactin ring complex is necessary for growth cone filopodia . J Neurosci 33 : 4514 - 4526 , 2013 . 13 . Gold ES , Underhill DM , Morrissette NS , Guo J , McNiven MA and Aderem A : Dynamin 2 is required for phagocytosis in macrophages . J Exp Med 190 : 1849 - 1856 , 1999 . 14 . Otsuka A , Abe T , Watanabe M , Yagisawa H , Takei K and Yamada H : Dynamin 2 is required for actin assembly in phago - cytosis in Sertoli cells . Biochem Biophys Res Commun 378 : 478 - 482 , 2009 . 15 . Faelber K , Posor Y , Gao S , Held M , Roske Y , Schulze D , Haucke V , No\u00e9 F and Daumke O : Crystal structure of nucleotide - free dynamin . Nature 477 : 556 - 560 , 2011 . Figure 6 . Putative role of the dynamin 2 - cortactin complex in cancer cell migration . The dynamin 2 - cortactin complex bundles actin filaments , which stabilize filopodia . In addition , the dynamin 2 - cortactin complex participates in the formation of pseudopodia . The regulation of actin by dynamin 2 and cortactin may also be involved in cancer cell invasion and metastasis . YAMADA et al : ACTIN BUNDLING BY DYNAMIN AND CORTACTIN IN CANCER CELL MIGRATION 886 16 . Ford MG , Jenni S and Nunnari J : The crystal structure of dynamin . Nature 477 : 561 - 566 , 2011 . 17 . Wu H , Reynolds AB , Kanner SB , Vines RR and Parsons JT : Identification and characterization of a novel cytoskeleton - associated pp60src substrate . Mol Cell Biol 11 : 5113 - 5124 , 1991 . 18 . MacGrath SM and Koleske AJ : Cortactin in cell migration and cancer at a glance . J Cell Sci 125 : 1621 - 1626 , 2012 . 19 . Ammer AG and Weed SA : Cortactin branches out : Roles in regulating protrusive actin dynamics . Cell Motil Cytoskeleton 65 : 687 - 707 , 2008 . 20 . Yamada H , Abe T , Li SA , Masuoka Y , Isoda M , Watanabe M , Nasu Y , Kumon H , Asai A and Takei K : Dynasore , a dynamin inhibitor , suppresses lamellipodia formation and cancer cell invasion by destabilizing actin filaments . Biochem Biophys Res Commun 390 : 1142 - 1148 , 2009 . 21 . Yamada H , Abe T , Li SA , Tago S , Huang P , Watanabe M , Ikeda S , Ogo N , Asai A and Takei K : N ' - [ 4 - ( dipropylamino ) benzylidene ] - 2 - hydroxybenzohydrazide is a dynamin GTPase inhibitor that suppresses cancer cell migration and invasion by inhibiting actin polymerization . Biochem Biophys Res Commun 443 : 511 - 517 , 2014 . 22 . Mooren OL , Kotova TI , Moore AJ and Schafer DA : Dynamin2 GTPase and cortactin remodel actin filaments . J Biol Chem 284 : 23995 - 24005 , 2009 . 23 . Masaike Y , Takagi T , Hirota M , Yamada J , Ishihara S , Yung TM , Inoue T , Sawa C , Sagara H , Sakamoto S , et al : Identification of dynamin - 2 - mediated endocytosis as a new target of osteoporosis drugs , bisphosphonates . Mol Pharmacol 77 : 262 - 269 , 2010 . 24 . Slepnev VI , Ochoa GC , Butler MH and De Camilli P : Tandem arrangement of the clathrin and AP - 2 binding domains in amphi - physin 1 and disruption of clathrin coat function by amphiphysin fragments comprising these sites . J Biol Chem 275 : 17583 - 17589 , 2000 . 25 . Eppinga RD , Krueger EW , Weller SG , Zhang L , Cao H and McNiven MA : Increased expression of the large GTPase dynamin 2 potentiates metastatic migration and invasion of pancreatic ductal carcinoma . Oncogene 31 : 1228 - 1241 , 2012 . 26 . Macia E , Ehrlich M , Massol R , Boucrot E , Brunner C and Kirchhausen T : Dynasore , a cell - permeable inhibitor of dynamin . Dev Cell 10 : 839 - 850 , 2006 . 27 . Hill TA , Gordon CP , McGeachie AB , Venn - Brown B , Odell LR , Chau N , Quan A , Mariana A , Sakoff JA , Chircop M , et al : Inhibition of dynamin mediated endocytosis by the dynoles - - synthesis and functional activity of a family of indoles . J Med Chem 52 : 3762 - 3773 , 2009 . 28 . Quan A , McGeachie AB , Keating DJ , van Dam EM , Rusak J , Chau N , Malladi CS , Chen C , McCluskey A , Cousin MA , et al : Myristyl trimethyl ammonium bromide and octadecyl trimethyl ammonium bromide are surface - active small molecule dynamin inhibitors that block endocytosis mediated by dynamin I or dynamin II . Mol Pharmacol 72 : 1425 - 1439 , 2007 . 29 . Schafer DA , Weed SA , Binns D , Karginov AV , Parsons JT and Cooper JA : Dynamin 2 and cortactin regulate actin assembly and filament organization . Curr Biol 12 : 1852 - 1857 , 2002 . 30 . Takei K , McPherson PS , Schmid SL and De Camilli P : Tubular membrane invaginations coated by dynamin rings are induced by GTP - gamma S in nerve terminals . Nature 374 : 186 - 190 , 1995 . 31 . Sweitzer SM and Hinshaw JE : Dynamin undergoes a GTP - dependent conformational change causing vesiculation . Cell 93 : 1021 - 1029 , 1998 . 32 . Takei K , Haucke V , Slepnev V , Farsad K , Salazar M , Chen H and De Camilli P : Generation of coated intermediates of clathrin - mediated endocytosis on protein - free liposomes . Cell 94 : 131 - 141 , 1998 . 33 . Roux A , Uyhazi K , Frost A and De Camilli P : GTP - dependent twisting of dynamin implicates constriction and tension in membrane fission . Nature 441 : 528 - 531 , 2006 . 34 . Feng H , Liu KW , Guo P , Zhang P , Cheng T , McNiven MA , Johnson GR , Hu B and Cheng SY : Dynamin 2 mediates PDGFR \u03b1 - SHP - 2 - promoted glioblastoma growth and invasion . Oncogene 31 : 2691 - 2702 , 2012 .",
    "xu2024myosini": "1 Myosin - I Synergizes with Arp2 / 3 Complex to Enhance Pushing Forces 1 of Branched Actin Networks 2 3 Mengqi Xu 1 \u2020 , David M . Rutkowski 2 \u2020 , Grzegorz Rebowski 1 , Malgorzata Boczkowska 1 , Luther W . Pollard 1 * , 4 Roberto Dominguez 1 * , Dimitrios Vavylonis 2 * , E . Michael Ostap 1 * 5 1 Department of Physiology , Pennsylvania Muscle Institute , Perelman School of Medicine , University of 6 Pennsylvania , Philadelphia , PA 19104 . 2 Department of Physics , Lehigh University , Bethlehem , PA . 7 Email : ostap @ pennmedicine . upenn . edu , vavylonis @ lehigh . edu , 8 droberto @ pennmedicine . upenn . edu , Luther . Pollard @ pennmedicine . upenn . edu 9 Author Contributions : \u2020 These authors contributed equally . 10 Abstract ( 149 words ) 11 Myosin - Is colocalize with Arp2 / 3 complex - nucleated actin networks at sites of membrane protrusion and 12 invagination , but the mechanisms by which myosin - I motor activity coordinates with branched actin 13 assembly to generate force are unknown . We mimicked the interplay of these proteins using the \u201ccomet 14 tail\u201d bead motility assay , where branched actin networks are nucleated by Arp2 / 3 complex on the surface 15 of beads coated with myosin - I and the WCA domain of N - WASP . We observed that myosin - I increased 16 bead movement efficiency by thinning actin networks without affecting growth rates . Remarkably , myosin - 17 I triggered symmetry breaking and comet - tail formation in dense networks resistant to spontaneous 18 fracturing . Even with arrested actin assembly , myosin - I alone could break the network . Computational 19 modeling recapitulated these observations suggesting myosin - I acts as a repulsive force shaping the 20 network ' s architecture and boosting its force - generating capacity . We propose that myosin - I leverages its 21 power stroke to amplify the forces generated by Arp2 / 3 complex - nucleated actin networks . 22 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 2 Introduction 23 Actin assembly stimulated by Arp2 / 3 complex provides pushing forces for diverse cellular 24 processes ( 1 - 8 ) , including lamellipodial protrusion , endocytosis , phagocytosis , and cell adhesion . After 25 being activated by membrane - associated nucleation promoting factors ( NPFs ) , Arp2 / 3 complex nucleates 26 new actin branches from the sides of pre - existing mother filaments , generating branched , dendritic actin 27 networks that exert pushing forces against or deform the membrane . The network geometry , assembly 28 kinetics , and mechanical properties are dynamically adapted by a set of actin binding proteins ( e . g . , capping 29 proteins , nucleation promoting factors , profilin , cofilin , and crosslinkers ) that respond to mechanical loading 30 ( 9 - 19 ) . 31 Class - I myosins ( myosin - Is ) ( 20 , 21 ) frequently colocalize with Arp2 / 3 complex - nucleated branched 32 actin networks near membranes ( 22 - 26 ) . As an actin - activated ATPase , the myosin - I motor dynamically 33 detaches and attaches from actin filaments in an ATP - dependent manner while generating force through a 34 lever arm - mediated power stroke ( 27 ) . Myosin - Is are single - headed , membrane - anchored motors that bind 35 directly to cell membranes and function in membrane deformation , cell adhesion , and intracellular trafficking 36 ( 21 ) . Although multiple cellular studies have shown that the myosin - I motor activity cooperates with dynamic 37 actin assembly by Arp2 / 3 complex at the plasma membrane , the functional outcome of this interaction 38 remains unknown . 39 To investigate the functional interaction between Arp2 / 3 complex and myosin - I in actin assembly , 40 we developed a biomimetic system using a comet - tail bead motility assay ( 9 , 12 , 28 ) . This system 41 recapitulates the colocalization of NPFs and myosin - I observed on cellular membranes by employing 42 micron - sized beads coated with both NPF and myosin - I . We investigated the impact of myosin - I on Arp2 / 3 43 complex - mediated branched actin assembly at varying network densities achieved through different 44 capping protein ( CP ) concentrations ( 12 , 14 , 29 ) . 45 Our findings revealed that myosin - I alters actin assembly kinetics by reducing Arp2 / 3 complex - 46 mediated branching at NPF - coated surfaces , resulting in a sparser actin network that exhibits enhanced 47 elongation efficiency . This effect is attributed to the pushing force exerted by the myosin - I power stroke 48 directly on surrounding actin networks , propelling actin filaments away . Furthermore , the myosin - I power 49 stroke generates sufficient force to disrupt the network and trigger actin shell breakdown . Notably , a 50 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 3 computational model at the molecular level provides mechanistic insights , suggesting that myosin - I 51 promotes force generation during Arp2 / 3 complex - mediated actin polymerization by exerting a repulsive 52 force on the branched network via its power stroke . 53 54 Results 55 Comet - tail bead motility assay 56 The comet - tail bead motility assay was used to investigate the effect of myosin - I activity on Arp2 / 3 57 complex - mediated branched actin assembly ( 9 , 28 ) . Full - length , biotinylated Drosophila Myo1d was 58 attached to the bead surface via neutravidin ( myosin - bead ; see Methods ; Fig . 1A ) alongside a GST - tagged 59 WCA domain from human N - WASP . Drosophila Myo1d was chosen as it functions optimally at 20 - 22 \u00b0C 60 ( 30 , 31 ) . Bead - bound Myo1d activity was confirmed by processive bead movement along actin filaments 61 ( Fig . 1B ; Movie S1 ) . Control - beads were made identically to the myosin - beads , except that a biotinylated 62 far - red fluorescent dye ( CF640 , see Methods ) was bound to the neutravidin ( Fig . 1A ) . Actin assembly 63 around myosin - and control - beads was imaged simultaneously by epifluorescence microscopy ( Fig . 1C ) . 64 Upon mixing of assay components , surface - bound NPFs stimulate Arp2 / 3 complex - mediated 65 branched actin assembly . Polymerization at the bead surface displaces the branched actin arrays 66 outwards , building tension in the actin network , and ultimately breaking symmetry and forming polarized 67 comet tails ( 9 , 14 , 32 - 35 ) ( Fig . 1C ) . By varying CP concentration , we created networks of varying actin 68 density that ranged from tightly packed and symmetry - breaking resistant ( dense ) to loosely woven and 69 fracture - prone ( sparse ) , allowing us to investigate Myo1d ' s influence on symmetry breaking , comet tail 70 formation , and morphology . 71 The morphology of comet tails generated by myosin - beads differed from control - beads at all 72 network densities achieved by varying CP concentrations ( Fig . 1D ; Fig . S1 ) , and these differences were 73 more pronounced at higher Myo1d densities . We identified three distinct regimes : high , intermediate , and 74 low network densities , controlled by the CP concentration ( 200 nM , 40 nM - 100 nM , and 30 nM , respectively ) 75 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 4 ( Fig . 1D ; Fig . S1 ) , and describe each regime in greater detail using the 0 . 43 myosin / 1 NPF ratio in the 76 following sections . 77 78 Myosin - I prevents comet tail formation from sparse actin networks 79 Under high CP concentrations ( 200 nM ) , actin elongation was suppressed by capping , resulting in 80 actin arrays consisting of short branches . As a result , actin comet tails grown from control - beads were 81 irregularly organized with sparse networks ( Fig . 1D ; Fig . 2A ) , consistent with previous reports ( 12 , 14 , 36 ) . 82 Myosin - beads ( 0 . 43 : 1 myosin / NPF ) failed to generate a comet tail ( Fig . 2A ; Movie S2 ) , but instead formed 83 loose actin clouds which diffused away from the bead when flow was present in the chamber . The presence 84 of myosin compromised the network cohesion , making it more prone to disruption , and inhibiting comet tail 85 formation . 86 87 Myosin - I facilitates fracturing of dense actin networks 88 At a low CP concentration ( 25 nM ) , long actin filaments emanating from bead surfaces become 89 entangled , forming dense actin networks that are highly resistant to fracture by actin polymerization forces 90 ( 12 , 14 , 34 , 36 ) . Control - beads remained encapsulated within the shell , without symmetry breaking , for > 91 15 minutes after the initiation of the polymerization ( Fig . 1D ; Fig . 2B ) . Strikingly , myosin - beads grown under 92 the same condition broke symmetry within 10 min of mixing ( Fig . 2B ; Movie S3 ) . This result suggests that 93 myosin - I may enhance the force generation during actin assembly and / or may alter actin architecture that 94 promotes network fracture and symmetry breaking . 95 96 Myosin - I induces efficient comet tail growth 97 At intermediate CP concentrations ( 40 nM - 100 nM ) , myosin - beads generated comet tails with 98 sparser actin networks compared to control - beads ( Fig . 1D ; Fig . S1 ; Fig . 3A - B ; Movie . S4 ) . Following the 99 time course of tail elongation , we found that the sparser network architecture of myosin - beads arises from 100 a 0 . 76 - fold lower actin assembly rate ( p < 0 . 0001 , n = 11 , Fig . 3E - F ) as quantified by measuring the 101 rhodamine - actin fluorescence in the comet tail over time . Remarkably , despite the lower actin assembly 102 rate , myosin comet tails in the presence of 50 nM CP elongated ( 0 . 69 \u00b1 0 . 10 \u03bcm / min ) at the same speed 103 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 5 as the control - beads ( 0 . 66 \u00b1 0 . 20 \u03bcm / min ; Fig . 3C - D ) . If we define the growth efficiency of actin comet tail 104 as the comet tail length divided by the total amount of actin incorporated , the myosin - beads exhibited a 1 . 4 - 105 fold higher growth efficiency than the control - beads ( Fig . 3G ) . 106 We quantified the amount of a fluorescently labeled Arp2 / 3 complex ( SNAP - Arp2 / 3 complex ; see 107 Methods ) in the comet tail ( Fig . 4A - C ) . Myosin - beads showed significantly reduced levels of SNAP - Arp2 / 3 108 complex ( p < 0 . 0001 , n = 33 , Fig . 4C ) in the network . This reduced Arp2 / 3 complex level also resulted in 109 higher actin to Arp2 / 3 complex ratios on myosin - beads ( Fig . 4D ) , indicating a sparse network organization 110 with fewer branch points but longer filaments ( 15 ) . Despite the reduced amount of Arp2 / 3 complex in the 111 comet tails , we observed equivalent SNAP - Arp2 / 3 complex fluorescence on the control - and myosin - bead 112 surfaces ( Fig . 4E ) , suggesting that myosin - I does not affect the loading of Arp2 / 3 complex onto the bead - 113 bound NPF . Rather , Arp2 / 3 complex is recruited to NPF - coated beads and is ready for the arrival of a 114 mother filament and G - actin to initiate branched nucleation . Taken together , myosin - beads have less dense 115 network structure as a result of reduced Arp2 / 3 complex - stimulated actin branching . 116 We note that the myosin effect on growth efficiency depends on the CP concentration ( Fig . 1D ) . 117 Under low CP conditions ( < 50 nM ) , myosin beads grew even longer tails than control - beads with sparse 118 networks , suggesting an even higher growth efficiency than the 50 nM CP case as quantified above . High 119 CP concentrations ( > 50 nM ) resulted in highly diffuse networks whose cohesion was easily compromised 120 by myosin - I , causing actin dispersal , thus making their growth efficiency difficult to assess ( 12 , 14 , 17 ) . 121 122 The myosin power stroke is required to alter network architecture 123 We examined the role of myosin - I mechanochemical activity on actin assembly by removing ATP , 124 resulting in the population of a long - lived , actin - bound , rigor state ( i . e . rigor myosin ) ( 31 ) . To maintain normal 125 actin polymerization , G - actin was pre - treated with ATP and gel filtered to eliminate free nucleotide ( see 126 Methods ) . Rigor myosin inhibited the formation of monopolar comet tails , while control - beads generated 127 comet patterns as previously observed when free ATP was present ( Fig . 5A - C ) . At 50 nM CP , rigor myosin - 128 beads formed dense actin shells , which subsequently fractured , forming multiple short tails of high network 129 density ( Fig . 5A - C ; Fig . S3C ; Movie S5 ) . Restoring myosin motor activity by adding ATP reproduced the 130 sparse comet architecture with high growth efficiency as observed before ( Fig . 5B ; Fig . S3B ) . At the two 131 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 6 extreme CP conditions ( 15 nM and 200 nM ) , rigor myosin - beads neither enhanced symmetry breaking nor 132 shed actin away ( Fig . 5C ; Fig . S3D ) . Instead , actin shells were formed around the beads as a result of the 133 strong actin - binding characteristics of rigor myosin . In addition to probing the effect of rigor myosin , we 134 coupled biotinylated , Halo - tagged , actin - binding domain ( Halo - ABD ) of \u03b1 - actinin to the beads , which binds 135 actin filaments more dynamically than rigor myosin but does not undergo a power stroke ( 31 , 37 , 38 ) . 136 HaloABD - beads behaved similarly to rigor myosin ( Fig . 5C ; Fig . S3A ; Movie . S6 ) . 137 To further elucidate the role of the myosin mechanochemistry in modulating the actin network , we 138 uncoupled the myosin actin - activated ATPase activity from the power stroke by adding calcium . Calcium 139 weakens the affinity of the lever - arm stabilizing calmodulin light chains , resulting in an inhibited power 140 stroke but preserved ATPase activity ( 27 , 39 , 40 ) . We confirmed calcium inhibition using the in vitro gliding 141 assay , where the Myo1d - powered F - actin gliding ( speed 90 . 2 \u00b1 14 . 5 nm / s ) was completely halted in the 142 presence of 100 \u00b5M free calcium ( Fig . S3E - F ) . Under these conditions , myosin - beads generated slightly 143 shorter comet tails with network densities that were similar to control - beads ( Fig . 5D - F ) . We conclude that 144 myosin power stroke is required to alter the network architecture and induce a sparse actin organization 145 with higher growth efficiency . 146 147 The myosin power stroke alone can fracture the actin shell 148 We decoupled the force generated by myosin power stroke from the force generated by actin 149 polymerization using Arp2 / 3 complex inhibitor , CK - 666 , and the polymerization inhibitor , Latrunculin B 150 ( LatB ) ( 12 , 41 ) . The goal was to arrest actin assembly around the bead prior to the fracturing of the actin 151 shell that results in symmetry breaking . Polymerization inhibitors were added ~ 100 s after the initiation of 152 polymerization . Strikingly , most of the myosin - beads fractured the actin shell and ejected the bead 15 - 20 153 min after polymerization arrest . In contrast , control - beads remained enclosed in the actin shell without 154 observable changes during a 40 min time window ( Fig . 6A ; also see Fig . S4 and Movie S7 - 8 acquired in 155 the absence of phalloidin ) . 156 Shell fracture events were quantified by defining a shell - breaking angle , \ud835\udf03 , where \ud835\udf03 = 0 , indicates 157 no detected shell fracture ; 0 \u00b0 < \ud835\udf03 < 180 \u00b0 , indicates shell fracture ; and \ud835\udf03 \u2265 180 \u00b0 indicates the bead has 158 been ejected from the shell ( Fig . 6B ) . We found 80 % of the myosin - beads were ejected from the shell and 159 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 7 an additional 11 % had fractured shells , while the corresponding control - beads showed only 16 % shell 160 fracture ( Fig . 6C ) . Inhibition of myosin motile activity by adding 10 mM ADP ( 31 ) together with Latrunculin 161 B and CK - 666 resulted in substantial reduction in shell fracture events and eliminated bead ejection ( Fig . 162 6A , C ; Fig . S4C - D ) . Finally , we performed experiments with a higher CP concentration ( 100 nM ) that results 163 in control - beads of lower actin shell densities ( Fig . 6A ) and found only 2 % shell fracture ( Fig . 6C ) , which 164 further confirmed that the shell breaking was a direct result of myosin power stroke , and is not attributed to 165 differences in the shell network density . 166 167 Simulations show myosin - I forces produce sparser actin networks and aid bead propulsion 168 We developed a filament - level computational model with an overall system size comparable to our 169 experimental setup ( Fig . 7A , see Methods , Table S1 ) . We incorporated the myosin - I power stroke as a 170 repulsive force that pushes actin away from the bead surface , and explore whether this myosin - induced 171 pushing mechanism reproduces the experimentally observed comet tail patterns . In this model , we 172 represented semiflexible actin filaments as beads connected by springs , polymerizing at their barbed ends 173 and pushing against a spherical bead according to Brownian - ratchet - type force - elongation relationship . 174 Spontaneous filament nucleation and branching at 70\u00ba angles occurs close to the bead ; elongation stops 175 by capping when filaments reach a specified length . The effect of fluid and bead - filament friction was 176 combined into a single bead friction parameter . Excluded volume interactions prevent filament crossing , 177 resulting in tensile and compressive stresses developing within a shell of branched actin filaments that 178 nucleates uniformly around an initially bare bead . By allowing filaments to break or debranch above a 179 certain tensile force or branch angle threshold , we found that these networks can crack open , leading to 180 symmetry breaking and bead propulsion ( Fig . 7B ; Movie S9 ) as we observed in the experiments ( Fig . 1C ) . 181 To account for changes in CP concentration , we varied the average filament branch length in the 182 simulations . 183 The effect of the myosin - I power stroke was modeled as constant tangential pushing forces of 184 magnitude F myo acting along every actin filament segment close to the bead , with equal and opposite force 185 on the bead ( Fig . 7A ) . With myosin - I pushing force incorporated , simulations recapitulated many of the 186 experimental findings . For the short filament scenario ( high CP concentration ) , simulated myosin - beads 187 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 8 showed a significant delay or absence of symmetry breaking due to myosin pushing short filaments away 188 from the bead surface ( Fig . S5A , C ; Movie S10 ) . For the long filament scenario ( low CP concentration ) , 189 simulations reproduced the accumulation of a dense actin shell around the beads , which significantly 190 inhibited the symmetry breaking of the control - beads ( Fig . S5B , C ; Movie S11 ) . Myosin - beads by contrast 191 promoted the growth of an asymmetric dense cloud ( Fig . S5B , C ) , similar to the experimental observations 192 in early stages prior to bead ejection ( Fig . 2B ) . We note that simulations do not account for depletion of bulk 193 actin occurring during late stages of bead ejection in experiment ( Fig . 2B ) . We also note a faster 194 accumulation of actin in the simulations compared to the experiments ( Fig . S5B ) , which suggests additional 195 factors limiting network growth not considered in the model . For the intermediate filament length scenario 196 ( i . e . intermediate CP concentration ) , simulations showed a similar elongation speed for the myosin - bead 197 compared to the control - bead , albeit with a less dense actin network ( Fig . 7D - F ; Movie S12 ) , in agreement 198 with the experimental observations ( Fig . 3A - F ) . Further , simulation predicted that actin filaments in the 199 myosin comet tail experienced less stress ( Fig . 7C ) , possibly due to the sparse actin organization , which 200 reduces stresses arising from piling interconnected actin filaments on top of each other . 201 Notably , the ratio of actin to Arp2 / 3 complex density in simulated comet tails did not increase in the 202 presence of myosin ( Fig . 7F - G ; Fig . S6A ) , as observed in experiments ( Fig . 4D ) . This fixed ratio is a 203 consequence of our fixed - filament - length assumption to mimic a certain CP concentration . This assumption 204 validates when capping rate reduces equally as actin polymerization rate in response to opposing force , as 205 reported by Li et al ( 15 , 19 ) . However , the mismatch between our simulations and the experiments suggests 206 an inequivalent effect on actin polymerization and CP capping due to the pushing force of myosin . We also 207 tested whether the presence of myosin would promote actin debranching , which could also change the 208 actin to Arp2 / 3 ratio . We found that myosin forces do not enhance debranching in our simulations ( Fig . S6 ) . 209 Indeed , less debranching occurs since comet tail stresses are smaller in the presence of myosin ( Fig . 7C ) . 210 We next dissected the force contribution from different sources ( actin polymerization or myosin 211 power stroke ) that powered the bead propulsion in our simulations . Strikingly , we found that while in control - 212 beads the symmetry breaking and net comet propulsion were exclusively powered by actin polymerization , 213 in myosin - beads , the myosin pushing forces contributed significantly to bead propulsion after symmetry 214 breaking , and had a predominant effect at high F myo ( Fig . 7H ; Fig . S7 ) . Simulated myosin forces also partly 215 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 9 reorganized the network such that a larger fraction of actin filaments was polymerizing facing more 216 perpendicular to the bead surface at larger myosin forces ( Fig . 7I ) . When actin polymerization was arrested 217 before shell breaking , simulations replicated the experimental findings that myosin pushing forces alone 218 can eject the bead out of an actin shell ( Fig . 7J ; Movie S13 ) , provided the shell is not too dense or too 219 sparse ( Fig . S8 ) 220 Overall , our simulations with the myosin - I power stroke incorporated as an effective repulsive force 221 replicate most of our experimental observations and support the experimental finding that myosin - I 222 enhances force generation during branched actin assembly to promote shell breaking and enhance the 223 efficiency of bead propulsion . 224 225 Discussion 226 Myosin - I force generation and modulation of actin network density 227 The primary finding of this study is that Myo1d synergizes with Arp2 / 3 complex to enhance the 228 pushing forces of branched actin networks , and that this force enhancement requires the myosin power 229 stroke . Importantly , we also found that Myo1d alters the actin network composition grown from NPF beads , 230 producing a less dense actin network with decreased incorporation of Arp2 / 3 complex . 231 How does myosin - I modulate Arp2 / 3 complex incorporation ? We propose a model where myosin - 232 I motor activity pushes the actin filaments away from the bead surface , reducing the accessibility of actin 233 binding sites for NPF - activated Arp2 / 3 complex to bind and nucleate new branches ( Fig . 8 ) . Alternatively , 234 the mechanochemistry of myosin - I may differently affect the incorporation of actin monomer and CP to 235 filament ends in a way that promotes actin elongation while slowing capping ( 19 , 42 ) . Additional 236 experiments that measure the actin elongation rate and capping kinetics near the NPF - coated surface are 237 required to test this hypothesis . 238 We do not favor a model in which myosin - I sterically inhibits NPFs from binding and activating 239 Arp2 / 3 complex for three reasons . First , the concentration of SNAP - Arp2 / 3 complex bound on beads is 240 unchanged in the presence of myosin ( Fig . 4E ) . Second , networks grown in the presence of calcium where 241 the myosin power stroke was uncoupled from actin binding showed similar actin densities between myosin - 242 and control - beads ( Fig . 5 D , F ) , which ruled out the possibility that the sparse actin organization was due to 243 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 10 the competition between myosin - I and Arp2 / 3 complex for actin binding sites . Finally , calculations of 244 molecular occupancy on bead surface through protein quantification verified that the steric hindrance effect 245 of myosin - I is likely negligible ( See SI for detailed quantifications ) . 246 It remains possible that force generation by myosin - I debranches actin filaments , resulting in less 247 dense network with reduced Arp2 / 3 incorporation . Mechanical forces ranging from 0 pN to 2 pN dissociate 248 actin branches from their mother filaments together with Arp2 / 3 complex ( 43 ) . A myosin - I paralog ( Myo1b ) 249 that generates such forces has been shown to dissociate branches via motor activity ( 44 ) . Although our 250 simulations disfavor this hypothesis ( Fig . S6 ) , a role for myosin - I induced debranching should be explored 251 further in the cell . 252 Finally , it will be intriguing to explore these models further by performing experiments and 253 computational modeling to probe network growth under mechanical load in the presence of myosin - I , 254 especially given previous work that demonstrates the substantial loading effects on network architecture 255 and power generation ( 15 , 19 ) . 256 257 Myosin - I Cell Biology 258 Myosin - Is connect the actin cytoskeleton to cellular membranes where they contribute to plasma 259 membrane dynamics , organelle deformation , and shaping of actin network architecture . Although the 260 molecular details of myosin - I function have been difficult to determine , recent studies suggest some 261 paralogs are powerful motors that have active roles in shaping membranes . For example , myosin - Is in 262 budding yeast have mechanochemistry suitable for generating power working with polymerizing actin to 263 drive membrane invagination during endocytosis ( 23 , 24 , 45 ) . Our current study confirms the ability of 264 myosin - I to exert substantial pushing forces capable of fracturing actin networks . This power - generating 265 capacity likely translates across species , with vertebrate Myo1c regulating actin architecture in diverse 266 cellular regions ( 46 - 48 ) and Myo1e playing key roles in processes like endocytosis ( 49 , 50 ) , phagocytosis 267 ( 26 ) , and cancer invadosome formation ( 51 , 52 ) . Taken together , these findings point to some myosin - Is 268 as dynamic players , actively shaping cellular structures and processes through their unique ability to both 269 link and manipulate membranes and the actin cytoskeleton . 270 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 11 Not all paralogs are expected to modulated the actin networks as Myo1d . Notably , some myosin 271 paralogs have substantially slower kinetics which are better suited for a motor that functions as dynamic 272 tether , providing force - dependent linkages between actin and membranes ( 22 - 24 , 45 , 48 , 53 , 54 ) . Our 273 experiments performed at low ATP concentrations ( Fig . 5 A - C ) mimic the behavior of myosins with slow 274 motility rates , and show clearly that slow kinetics can inhibit network fracturing and comet tail growth . 275 Further studies to investigate how the intrinsic kinetic properties of different myosin - I paralogs influence 276 actin polymerization and membrane dynamics will further reveal the diverse role of myosin - I function ( 23 , 277 24 , 26 , 45 , 48 , 54 ) . 278 279 Summary 280 Overall , our study provides insights into how myosin - I molecules coordinate with Arp2 / 3 complex , 281 regulating the dynamics and architecture of the branched actin network and promoting the actin - based 282 motile force generation ( Fig . 8 ) . This work sheds light on synergy between myosin motor activity and actin 283 polymerization , underscoring their collective role in driving morphological changes at the cellular membrane 284 interfaces . Future work to determine how myosin - I affects actin network architecture and polymerization 285 forces will reveal the molecular roles of this important myosin family . 286 287 Materials and Methods 288 Protein Purification 289 Actin was purified from rabbit skeletal muscle acetone powder as previously described ( 55 ) . Monomeric G - 290 actin was purified by gel filtration on Sephacryl S - 300 in G - buffer ( 2 mM Tris \u2013 HCl [ pH 8 . 0 ] , 0 . 2 mM ATP , 291 0 . 1 mM CaCl 2 , 1 mM NaN 3 and 0 . 5 mM DTT ) and used within 2 - 3 weeks . Actin was labeled with NHS - 292 Rhodamine at random surface lysine residues ( 56 ) . Full - length Drosophila Myo1d , with C - terminal FLAG - 293 Avi tags , was expressed , purified as previously described ( 57 ) , and subsequently biotinylated at the C - 294 termini Avi tag sequence via BirA biotin - protein ligase ( Avidity ) ( 31 ) . GST - tagged WCA domain of human 295 WASP protein ( GST - WCA ) was purchased from Cytoskeleton ( Cat . # VCG03 - A ) and used without further 296 purification . Arp2 / 3 complex was isolated from bovine brain as previously described ( 58 ) . SNAP - tagged 297 Arp2 / 3 complex was constructed and purified as described ( 59 ) . SNAP - tagged Arp2 / 3 complex was labeled 298 with SNAP - Surface 488 ( Biolab , Cat . # S9124S ) using commercially provided protocol . Human CapZ was 299 expressed and purified as previously described ( 60 ) . Halo - ABD was constructed and purified as described 300 in ( 61 ) . 301 302 Bead Preparation 303 Carboxylate polystyrene beads ( Polybead , 2 . 0 \u03bcm diameter , CAT . # 18327 - 10 ) were purchased from 304 Polysciences . Beads were coated with NPF and neutravidin following a previous protocol ( 28 ) with 305 modifications . Briefly , 5 - 10 \u03bcL bead slurry was washed with X buffer ( 10mM HEPES [ pH7 . 5 ] , 100 mM KCl , 306 1 mM MgCl 2 , 100 \u03bcM CaCl 2 , 1 mM ATP ) , and then incubated with 50 - 100 \u03bcl 2 . 3 \u03bcM ( 0 . 1mg / ml ) GST - WCA 307 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 12 and various concentrations of Neutravidin ( ThermoFisher , Cat . # 31000 ) ( 8 . 3 \u03bcM ( 0 . 5mg / ml ) for 308 # myosin / NPF = 0 . 28 : 1 , 16 . 7 \u03bcM ( 1mg / ml ) for # myosin / NPF = 0 . 35 : 1 , 33 . 3 \u03bcM ( 2mg / ml ) for 309 # myosin / NPF = 0 . 43 : 1 , or 83 . 3 \u03bcM ( 5mg / ml ) for # myosin / NPF = 0 . 80 : 1 ) on a slow rotator at 4 \u00b0 C for 2hr . 310 Beads were then pelleted by spinning at 16000 g for 2 min at 4 \u00b0 C , to remove the unreacted reagents , and 311 then resuspended in 200 - 400 \u03bcL 10mg / mL BSA , incubating on ice for 30 min to block the free space left on 312 bead surface . Finally , beads were washed twice and stored in 50 - 100 \u03bcL 1mg / mL BSA in X buffer for up to 313 3 days . 314 315 The NPF and neutravidin - coated beads were next split in half , with one half coupled with biotinylated 316 Drosophila Myo1d ( myosin - bead ) or biotinylated Halo - ABD , and the other half coated with Biotin - CF640 317 ( Biotium , Cat . # 80032 ) fluorescence dye ( control - bead ) . The NPF and neutravidin - coated beads were first 318 washed with M buffer ( 20 mM HEPES [ pH7 . 5 ] , 100 mM KCl , 1 mM MgCl 2 , 1 mM EGTA , 2 mM ATP ) to 319 remove the free calcium in the storage X buffer , and were then incubated with 1 \u03bcM biotinylated Drosophila 320 Myo1d , 1 \u03bcM Halo - ABD , or 1 \u03bcM Biotin - CF640 fluorescence dye for 30 min on ice . Beads were then pelleted 321 under 16000xg for 2 min at 4 \u00b0 C , to remove the free unbound reagents , and washed with 1mg / mL BSA in 322 M buffer and used immediately for motility assay on the same day after preparation . 323 324 Bead motility assay / comet tail assay 325 Unless specified otherwise , a typical motility mixture contained 4 \u03bcM actin ( 5 % Rhodamine labeled ) , 200 326 nM Arp2 / 3 complex or SNAP - Arp2 / 3 complex , 6 . 5 - 200 nM CP , 2 \u03bcM Calmodulin , and 3 \u03bcL bead slurry 327 ( 1 . 5 \u03bcL of myosin - beads or HaloABD - beads , and 1 . 5 \u03bcL of control - beads ) , mixed in 20 mM HEPES [ pH7 . 5 ] , 328 100 mM KCl , 1 mM MgCl 2 , 1 mM EGTA , 1 mM MgATP , 40 mM DTT , 10 mg / mL BSA and 0 . 2 % 329 methylcellulose , contributing to a final volume of 50 \u03bcL . The activity of SNAP - Arp2 / 3 complex is slightly 330 lower than the unlabeled complex , so we changed the CP concentration ( 40 nM ) to achieve similar tail 331 lengths , actin densities and growth efficiencies as observed for the native complex ( Fig . 4 ) . For experiments 332 with calcium present , G - actin was pre - incubated with 200 \u00b5M EGTA and 50 \u00b5M MgCl 2 for 5 min to be 333 exchanged to Mg - G - actin before using . Calcium experiments included 1 . 1 mM CaCl 2 in place of Calmodulin 334 to disrupt the myosin power stroke . Beads were first mixed with all other reagents , excluding actin , with a 335 pipette , to ensure an even distribution . The reaction was then initiated by adding actin into the system , 336 mixed thoroughly , and denoted as time 0 . Slides and coverslips were wiped with 70 % ethanol and ddH 2 O , 337 followed by a plasma cleansing for 10 min . Upon mixing , a volume of 2 . 1 \u03bcL motility mixture was carefully 338 applied between a glass slide and a coverslip ( 22 mm \u00d7 22 mm ) forming a so called ' squeeze chamber ' 339 with a height of 4 . 3 \u03bcm . The squeeze chamber was then sealed with vacuum grease and imaged 340 immediately under the microscope . Time - lapse movies were acquired of microscope fields that included 341 both myosin - and control - beads . In most cases , the actin comet tails emerging from myosin - or control - 342 beads grew with a constant speed during the first ~ 10 minutes following symmetry breaking . As the 343 reagents in the polymerization mixture depleted , tail elongation slowed and eventually stopped ~ 30 min 344 after mixing . 345 346 347 Fluorescence imaging and Data analysis 348 Fluorescence microscopy imaging was performed via Leica DMIRB epifluorescence microscope ( 100x , oil - 349 immersive objective of numerical aperture 1 . 4 ) and Metamorph ( Molecular Devices ) imaging software . 350 Movies were recorded at 25 \u00b0 C and acquired every 10 s for 30 - 60 min . Exposure time was 200 ms for most 351 experiments , and 1s to image the SNAP - Surface 488 labeled Arp2 / 3 complex . Images were analyzed and 352 quantified using Fiji software . Actin comet tail length were measured manually at each frame using the 353 segmented line draw tool ( from the end of the tail to the center of the bead ) and converted to the microns . 354 Tail growth rate and fluorescent assembly rate were calculated by fitting the first 7 to 10 data points in the 355 time courses to get the initial slope of the growth and assembly . Growth efficiency was determined by 356 dividing the tail growth rate by the fluorescent assembly rate . Image brightness and contrast were carefully 357 adjusted using Fiji and Adobe Illustrator . Unless otherwise specified , the lookup table for each pair of 358 control - and myosin - bead ( or HaloABD - bead ) were kept the same . 359 360 Statistical analysis 361 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 13 The statistical significance was calculated using paired or unpaired two - tail Student\u2019s t test in Prism v9 . 0 . 362 Further details are described in figure legends . 363 364 Modeling methods 365 We simulated actin comet tail formation at the level of individual filaments , each represented as a series of 366 segments of length \ud835\udc59 ! , following earlier work ( 62 ) . The pointed ends of actin filament branches are assumed 367 to be connected to a point element of a mother filament , at the location of Arp2 / 3 complex . Such a filament 368 representation allows us to model the effect of myosin as a tangential force acting along filament segments 369 close to the nucleating bead . We thus generalize earlier filament models that did not explicitly account for 370 filament bending mechanics ( 35 , 63 - 65 ) , earlier mechanical models that did not monitor the whole network 371 of actin filaments ( 36 , 66 - 69 ) , or modeled the full process of symmetry breaking and propulsion ( 70 ) . We 372 do not explicitly consider diffusiophoretic contributions to bead motion ( 71 ) . ( Also see SI for further details ) . 373 Forces on actin filaments 374 The position \ud835\udc93 \ud835\udc8a of point element i of actin filaments / filament branches evolves according to ( 62 ) 375 \ud835\udf01 # \ud835\udc93 \ud835\udc8a # % = \ud835\udc6d & \u2019 ( ) & * + + \ud835\udc6d & , - * # + \ud835\udc6d & . * + / - + \ud835\udc6d & - 01 / 2 # - # + \ud835\udc6d & 345 . 376 Here \ud835\udf01 is an effective filament segment drag coefficient that allows the actin network to evolve through 377 approximate quasi - static mechanical equilibrium , while also providing numerical stability . The spring force 378 is \ud835\udc6d & \u2019 ( ) & * + = \ud835\udc58 . 1 % & * , \ud835\udc51 & ( & 78 ) \u2212 \ud835\udc59 ! / \ud835\udc510 & ( & 78 ) \u2212 \ud835\udc58 . 1 % & * , \ud835\udc51 ( & : 8 ) & \u2212 \ud835\udc59 ! / \ud835\udc510 ( & : 8 ) & where ( i - 1 ) , ( i + 1 ) are neighboring point 379 elements ( if they exist ) before and after i , \ud835\udc51 & ; is the separation distance between i and j , and \ud835\udc510 & ; is the unit 380 vector from i to j . The equilibrium length is \ud835\udc59 ! , except for ( i ) uncapped barbed ends that elongate on average 381 according to \ud835\udc59 ! ( \ud835\udf0f ) = \ud835\udc59 ! ! + \ud835\udeff \ud835\udc5f ! ( 5 / \ud835\udf0f after their initiation at \ud835\udf0f = 0 , in discrete steps of half - monomer size \ud835\udeff = 382 2 . 7 nm ( see below ) and ( ii ) a short branch segment joining the mother filament point element to the daughter 383 pointed end , which has length of \ud835\udc59 , ) . * 1 < . 384 The bending force is \ud835\udc6d & , - * # = \ud835\udf05 / \ud835\udc59 . = + [ \ud835\udf15 , \ud835\udc510 & ( & 78 ) \u2219 \ud835\udc510 ( & : 8 ) & / / \ud835\udf15\ud835\udc93 & + , \ud835\udc510 ( & : 8 ) & \u2219 \ud835\udc510 ( & : > ) ( & : 8 ) / / \ud835\udf15\ud835\udc93 & + , \ud835\udc510 ( & 78 ) ( & 7 > ) \u2219 385 \ud835\udc510 & ( & 78 ) / / \ud835\udf15\ud835\udc93 & ] , \ud835\udf05 = \ud835\udc58 ? \ud835\udc47\ud835\udc59 ( is the flexural rigidity , \ud835\udc59 ( is the persistence length of the actin filament , and \ud835\udc59 . = + is 386 the average length of the two filament segments composing the angle . For the straight angle that exists 387 among the first three beads in the connection between mother and daughter filaments \ud835\udc59 . = + = \ud835\udc59 ! for 388 numerical stability . 389 An angular potential keeps Arp2 / 3 complex branches at 70 \u00b0 . The angular force is \ud835\udc6d 3 . * + / - = 390 \ud835\udf16 . * + / - , \ud835\udc50\ud835\udc5c\ud835\udc60\ud835\udf03 & ; @ \u2212 \ud835\udc50\ud835\udc5c\ud835\udc60\ud835\udf03 ! / \ud835\udf15 , \ud835\udc51 0 ; @ \u2219 \ud835\udc51 0 & ; / / \ud835\udf15\ud835\udc93 3 where \ud835\udf16 . * + / - is a spring constant , i , j , k are the indices of the point 391 elements that make up the angle , and m is one of these indices . 392 The excluded volume force is due to repulsion between two actin filaments and between actin filaments 393 and the nucleating bead . The former prevents crossing of filaments and is exerted along the direction of 394 vector \ud835\udc85 AB that connects the two closest approach points on the filament segments \u03b1 and \u03b2 ( 72 ) . It is 395 modeled as a stiff spring force with a max range of \ud835\udc51 - 01 / 2 # - # , with \ud835\udc6d A - 01 / 2 # - # , . 1 % & * = \u2212\ud835\udc6d B - 01 / 2 # - # , . 1 % & * = 396 \ud835\udc58 - 01 / , \ud835\udc51 AB \u2212 \ud835\udc51 - 01 / 2 # - # / \ud835\udc510 AB , where \ud835\udc51 AB is the minimum distance between neighboring filament segments . 397 This force is distributed to the endpoints of filament segments \u03b1 and \u03b2 ( including element i ) according to a 398 level arm rule . Though the typical diameter of actin filaments is around 7 nm , we set \ud835\udc51 - 01 / 2 # - # to be 20 nm 399 in order to mimic the effect of thermally fluctuating filaments taking up a larger volume ( which is not included 400 in our simulations ) . Additionally , since the excluded volume is modeled as a stiff spring rather than a hard - 401 core potential , some level of overlap between segments is possible . At a separation of order the filament 402 diameter , 7 nm , the excluded volume force is 22 pN . Filament segments experience a radially - oriented 403 excluded volume force with the nucleating bead if they are closer to the nucleator than R ; this force can be 404 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 14 written as \ud835\udc6d & - 01 / 2 # - # , , - . # = \ud835\udc58 - 01 / ( \ud835\udc51 & D \u2212 \ud835\udc45 ) \ud835\udc510 & D where \ud835\udc51 & D is the closest approach distance between the 405 filament segment and the nucleator bead . 406 Myosin forces are given by \ud835\udc6d & 345 = \ud835\udc39 345 \ud835\udc510 & ( & : 8 ) , where i - 1 is the neighboring actin filament point element 407 along the pointed end direction . It acts on all filament elements i closer than \ud835\udc51 345 to the nucleator bead 408 surface . The myosin force acts tangentially along the segment towards the pointed end . We do not model 409 the individual binding , lever arm motion , and unbinding of myosin ; instead the magnitude of \ud835\udc39 345 410 approximates a time and ensemble average over many binding and unbinding cycles . 411 Nucleating bead motion 412 The nucleating bead evolves through time according to : 413 \ud835\udf01 * 21 / , - . # \ud835\udc51\ud835\udc93 * 21 / , - . # \ud835\udc51\ud835\udc61 = \u2212 G ( \ud835\udc6d & - 01 / 2 # - # , , - . # + \ud835\udc6d & 345\u2019 & * ) & 414 Where the sum is over all actin segments that contact the bead through excluded volume or myosin 415 interactions , according to the interaction distances defined above . We assume the relative motion between 416 actin network and nucleating bead is dominated by frictional forces between them . We used a value of 417 \ud835\udf01 * 21 / , - . # that was large enough to prevent the rapid ejection of bare beads out of a shell at the onset of 418 symmetry breaking , a phenomenon that is not seen in our experiments . We thus monitor the relative motion 419 of the bead and actin network in the limit of quasi - static mechanical equilibrium of the actin network , 420 including implicit transient attachment and detachment of actin filaments to the bead . This approximation 421 does not account for the varying concentration of actin near the bead or the absolute motion of actin comet 422 and bead in the lab frame , which in reality would be influenced by small forces between the bead or actin 423 and the glass slide . 424 Actin barbed end polymerization rate 425 Polymerization of the barbed end of uncapped filaments occurs with the last segment of the actin filament 426 lengthening in increments of half - monomer size \u03b4 at a rate given by the polymerization rate . If the end 427 segment reaches a length of \ud835\udc59 ! , a new segment is added with initial length \ud835\udc59 ! ! . Polymerization of the barbed 428 end is attenuated by compressive forces on the spring bond connecting the barbed end point element to 429 the rest of the filament . The polymerization rate is \ud835\udc5f ( 5 / ( \ud835\udc47 ) = \ud835\udc5f ! ( 5 / exp ( \u2212\ud835\udc47 \ud835\udeff / \ud835\udc58 ? \ud835\udc47 ) , where the free filament 430 elongation rate is \ud835\udc5f ! ( 5 / and \ud835\udc47 = | \ud835\udc6d & \u2019 ( ) & * + | when \ud835\udc6d & \u2019 ( ) & * + is the compressive tension on the barbed end point 431 element i ( otherwise \ud835\udc5f ( 5 / ( \ud835\udc47 ) = \ud835\udc5f ! ( 5 / ) ( 73 , 74 ) . Because each filament segment starts with an initial length of 432 \ud835\udc59 ! ! ( for numerical stability ) , the rate \ud835\udc5f ( 5 / ( 0 ) is effectively multiplied by a factor of 1 . 12 . 433 Barbed end capping 434 Filaments polymerize until they reach a final length specified for a given simulation , \ud835\udc3f F & / ( an integer multiple 435 of \ud835\udc59 ! ) , which we vary to simulate the effect of varying CP concentration . Here we didn\u2019t study the effects of 436 a varying filament length distribution . The assumption that \ud835\udc3f F & / is independent of the force is based on prior 437 experiments ( 19 ) and the fact that the filament length added by capping protein is close to \u03b4 . According to 438 this evidence , the polymerization to capping rate ratio , as well as the average filament length , remains 439 unchanged by force for a given CP concentration . 440 Branching 441 Branching occurs from actin filament point elements that are within \ud835\udc59 ! of the nucleating bead , at rate \ud835\udc5f ! , ) . * 1 < , 442 independent of the CP concentration . The rate was chosen to approximate the timescale of symmetry 443 breaking and comet speed at intermediate CP . The segment length connecting the mother filament point 444 element to the daughter filament pointed end point element is \ud835\udc59 ! \" # $ % & . Their orientation is chosen according 445 to a uniform angle distribution in the cone around the mother filament opening towards the barbed end . To 446 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 15 maintain a discretization of the network at segment length \ud835\udc59 ! , branches cannot branch from the barbed end 447 point element of the filament . Branches nucleating from the same mother filament have no torsional 448 restriction ( i . e . daughter filaments can rotate about the axis of the mother filament without restriction ) , but 449 implicit torsional restrictions are imposed due to the dense surrounding network limiting this motion . 450 De novo nucleation 451 Nucleation of new filaments of length \ud835\udc59 ! ! occurs at a rate of \ud835\udc5f ! # - * 5 = 5 . These small filaments are introduced 452 with a uniform spherical orientation with either their pointed or barbed end touching the nucleating bead 453 surface . This slow rate of filament introduction was tuned to allow for startup of the network on timescales 454 comparable to experiment , and to allow for the buildup of a thin actin cloud around the leading bead edge 455 during actin comet propulsion , as observed in experiments . At the beginning of the simulation , 200 of these 456 de novo filaments are added to provide enough initial filaments for symmetric growth of the shell . 457 Filament fragmentation and debranching 458 For simplicity , filament fragmentation was assumed to occur above a certain tensile force threshold \ud835\udc39 \u2019 \" # ( 459 ( instead of implementing a rate of severing on as a function of tension ) . 460 For filament segments the threshold was \ud835\udc39 \u2019 \" # ( = 50 pN . This is lower than the experimental fragmentation 461 force of phalloidin - labeled actin filaments , which is several hundred pNs ( 75 ) , but we found that the shell 462 would have difficulty breaking unless we had this lower force . Once a filament segment is fragmented , the 463 filament with the newly created barbed end is left uncapped and can grow to the final length specified for 464 that simulation . Debranching occurs for branches that have a branch bond with tension greater than 465 \ud835\udc39 # - , ) . * 1 < = 30 pN or if the angle deviates by more than \ud835\udee5\ud835\udf03 # - , ) . * 1 < = 25 \u00b0 off of the 70 \u00b0 equilibrium angle . 466 The value of \ud835\udee5\ud835\udf03 # - , ) . * 1 < was chosen to allow easy debranching when the branch is bent ( similar to 467 experiments where branches were bent and pulled by fluid forces of order pN , ( 43 ) ) . The value of \ud835\udc39 # - , ) . * 1 < 468 had to be sufficiently high such that the network does not easily fall apart . 469 470 Acknowledgments 471 We thank Daniel Safer , Faviolla A . Baez - Cruz and Richard Wike for their kind help with protein purifications 472 and technical assistance on the project . We thank everyone in Ostap Lab and Dominguez Lab for their 473 valuable inputs throughout the project duration . 474 475 Funding : This work was supported by NIH grants R37 GM057247 ( EMO ) and R01 GM073791 ( RD ) . DV 476 and DMR were supported by NIH grant R35GM136372 . Portions of this research were conducted on the 477 Rockfish ( Johns Hopkins ) cluster through allocations MCB180021 and BIO230116 from the Advanced 478 Cyberinfrastructure Coordination Ecosystem : Services & Support ( ACCESS ) program , which is supported 479 by NSF grants # 2138259 , # 2138286 , # 2138307 , # 2137603 , and # 2138296 . 480 481 Author contributions : 482 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 16 Conceptualization : M . X . , D . M . R . , L . W . P . , R . D . , D . V . , E . M . O . 483 Methodology : M . X . , D . M . R . , L . W . P . 484 Formal analysis : M . X . , D . M . R . 485 Investigation : M . X . , D . M . R . , L . W . P . 486 Resources : G . R . , M . B . 487 Visualization : M . X . , D . M . R . L . W . P . , D . V . , E . M . O . 488 Supervision : L . W . P . , R . D . , D . V . , E . M . O . 489 Writing\u2014original draft : M . X . , D . M . R . , D . V . , E . M . O . 490 Writing\u2014review & editing : M . X . , D . M . R . , L . W . P . , M . B . , R . D . , D . V . , E . M . O . 491 492 Competing interests : The authors declare that they have no competing interests . 493 494 Data and materials availability : All data needed to evaluate the conclusions in the paper are present in 495 the paper and / or the Supplementary Materials . Additional data related to this paper may be requested from 496 the authors . 497 498 References 499 500 1 . L . Blanchoin , R . Boujemaa - Paterski , C . Sykes , J . Plastino , Actin dynamics , architecture , and 501 mechanics in cell motility . Physiol Rev 94 , 235 - 263 ( 2014 ) . 502 2 . K . Rottner , J . Faix , S . Bogdan , S . Linder , E . Kerkhoff , Actin assembly mechanisms at a glance . 503 Journal of Cell Science 130 , 3427 - 3435 ( 2017 ) . 504 3 . T . Svitkina , The Actin Cytoskeleton and Actin - Based Motility . Cold Spring Harb Perspect Biol 10 , 505 ( 2018 ) . 506 4 . V . Papalazarou , L . M . Machesky , The cell pushes back : The Arp2 / 3 complex is a key orchestrator 507 of cellular responses to environmental forces . Curr Opin Cell Biol 68 , 37 - 44 ( 2021 ) . 508 5 . A . M . Gautreau , F . E . Fregoso , G . Simanov , R . Dominguez , Nucleation , stabilization , and 509 disassembly of branched actin networks . Trends Cell Biol 32 , 421 - 432 ( 2022 ) . 510 6 . M . Krendel , N . C . Gauthier , Building the phagocytic cup on an actin scaffold . Curr Opin Cell Biol 511 77 , 102112 ( 2022 ) . 512 7 . M . Jin et al . , Branched actin networks are organized for asymmetric force production during 513 clathrin - mediated endocytosis in mammalian cells . Nat Commun 13 , 3578 ( 2022 ) . 514 8 . D . N . Clarke , A . C . Martin , Actin - based force generation and cell adhesion in tissue 515 morphogenesis . Curr Biol 31 , R667 - R680 ( 2021 ) . 516 9 . T . P . Loisel , R . Boujemaa , D . Pantaloni , M . F . Carlier , Reconstitution of actin - based motility of 517 Listeria and Shigella using pure proteins . Nature 401 , 613 - 616 ( 1999 ) . 518 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 17 10 . F . Nakamura , E . Osborn , P . A . Janmey , T . P . Stossel , Comparison of filamin A - induced cross - 519 linking and Arp2 / 3 complex - mediated branching on the mechanics of actin filaments . J Biol Chem 520 277 , 9148 - 9154 ( 2002 ) . 521 11 . O . Chaudhuri , S . H . Parekh , D . A . Fletcher , Reversible stress softening of actin networks . Nature 522 445 , 295 - 298 ( 2007 ) . 523 12 . O . Akin , R . D . Mullins , Capping protein increases the rate of actin - based motility by promoting 524 filament nucleation by the Arp2 / 3 complex . Cell 133 , 841 - 851 ( 2008 ) . 525 13 . J . Stricker , T . Falzone , M . L . Gardel , Mechanics of the F - actin cytoskeleton . J Biomech 43 , 9 - 14 526 ( 2010 ) . 527 14 . A . Kawska et al . , How actin network dynamics control the onset of actin - based motility . Proc Natl 528 Acad Sci U S A 109 , 14440 - 14445 ( 2012 ) . 529 15 . P . Bieling et al . , Force Feedback Controls Motor Activity and Mechanical Properties of Self - 530 Assembling Branched Actin Networks . Cell 164 , 115 - 127 ( 2016 ) . 531 16 . J . Mueller et al . , Load Adaptation of Lamellipodial Actin Networks . Cell 171 , 188 - 200 e116 ( 2017 ) . 532 17 . R . D . Mullins , P . Bieling , D . A . Fletcher , From solution to surface to filament : actin flux into 533 branched networks . Biophys Rev 10 , 1537 - 1551 ( 2018 ) . 534 18 . J . Funk , F . Merino , M . Schaks , K . Rottner , S . Raunser , P . Bieling , A barbed end interference 535 mechanism reveals how capping protein promotes nucleation in branched actin networks . Nature 536 Communications 12 , ( 2021 ) . 537 19 . T . D . Li , P . Bieling , J . Weichsel , R . D . Mullins , D . A . Fletcher , The molecular mechanism of load 538 adaptation by branched actin networks . Elife 11 , ( 2022 ) . 539 20 . T . D . Pollard , E . D . Korn , Acanthamoeba Myosin . Journal of Biological Chemistry 248 , 4682 - 4690 540 ( 1973 ) . 541 21 . B . B . McIntosh , E . M . Ostap , Myosin - I molecular motors at a glance . J Cell Sci 129 , 2689 - 2695 542 ( 2016 ) . 543 22 . C . G . Almeida , A . Yamada , D . Tenza , D . Louvard , G . Raposo , E . Coudrier , Myosin 1b promotes 544 the formation of post - Golgi carriers by regulating actin assembly and membrane remodelling at 545 the trans - Golgi network . Nat Cell Biol 13 , 779 - 789 ( 2011 ) . 546 23 . R . T . A . Pedersen , D . G . Drubin , Type I myosins anchor actin assembly to the plasma membrane 547 during clathrin - mediated endocytosis . J Cell Biol 218 , 1138 - 1147 ( 2019 ) . 548 24 . H . E . Manenschijn , A . Picco , M . Mund , A . S . Rivier - Cordey , J . Ries , M . Kaksonen , Type - I 549 myosins promote actin polymerization to drive membrane bending in endocytosis . Elife 8 , ( 2019 ) . 550 25 . A . Capmany et al . , MYO1C stabilizes actin and facilitates the arrival of transport carriers at the 551 Golgi complex . J Cell Sci 132 , ( 2019 ) . 552 26 . S . R . Barger et al . , Membrane - cytoskeletal crosstalk mediated by myosin - I regulates adhesion 553 turnover during phagocytosis . Nat Commun 10 , 1249 ( 2019 ) . 554 27 . M . J . Greenberg , E . M . Ostap , Regulation and control of myosin - I by the motor and light chain - 555 binding domains . Trends Cell Biol 23 , 81 - 89 ( 2013 ) . 556 28 . R . Boujemaa - Paterski , R . Galland , C . Suarez , C . Guerin , M . Thery , L . Blanchoin , Directed actin 557 assembly and motility . Methods Enzymol 540 , 283 - 300 ( 2014 ) . 558 29 . V . Noireaux et al . , Growing an actin gel on spherical surfaces . Biophys J 78 , 1643 - 1654 ( 2000 ) . 559 30 . J . H . Lewis , T . Lin , D . E . Hokanson , E . M . Ostap , Temperature dependence of nucleotide 560 association and kinetic characterization of myo1b . Biochemistry 45 , 11589 - 11597 ( 2006 ) . 561 31 . F . A . Baez - Cruz , E . M . Ostap , Drosophila class - I myosins that can impact left - right asymmetry 562 have distinct ATPase kinetics . J Biol Chem 299 , 104961 ( 2023 ) . 563 32 . L . A . Cameron , M . J . Footer , A . van Oudenaarden , J . A . Theriot , Motility of ActA protein - coated 564 microspheres driven by actin polymerization . Proc Natl Acad Sci U S A 96 , 4908 - 4913 ( 1999 ) . 565 33 . A . Bernheim - Groswasser , S . Wiesner , R . M . Golsteyn , M . F . Carlier , C . Sykes , The dynamics of 566 actin - based motility depend on surface parameters . Nature 417 , 308 - 311 ( 2002 ) . 567 34 . J . van der Gucht , E . Paluch , J . Plastino , C . Sykes , Stress release drives symmetry breaking for 568 actin - based movement . Proc Natl Acad Sci U S A 102 , 7847 - 7852 ( 2005 ) . 569 35 . V . Achard et al . , A \" primer \" - based mechanism underlies branched actin filament network 570 formation and motility . Curr Biol 20 , 423 - 428 ( 2010 ) . 571 36 . M . J . Dayel , O . Akin , M . Landeryou , V . Risca , A . Mogilner , R . D . Mullins , In silico reconstitution of 572 actin - based symmetry breaking and motility . PLoS Biol 7 , e1000201 ( 2009 ) . 573 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 18 37 . D . H . Wachsstock , W . H . Schwarz , T . D . Pollard , Cross - linker dynamics determine the 574 mechanical properties of actin gels . Biophys J 66 , 801 - 809 ( 1994 ) . 575 38 . P . A . Kuhlman , J . Ellis , D . R . Critchley , C . R . Bagshaw , The kinetics of the interaction between 576 the actin - binding domain of alpha - actinin and F - actin . FEBS Lett 339 , 297 - 301 ( 1994 ) . 577 39 . J . H . Lewis , M . J . Greenberg , J . M . Laakso , H . Shuman , E . M . Ostap , Calcium regulation of 578 myosin - I tension sensing . Biophys J 102 , 2799 - 2807 ( 2012 ) . 579 40 . S . Manceva , T . Lin , H . Pham , J . H . Lewis , Y . E . Goldman , E . M . Ostap , Calcium regulation of 580 calmodulin binding to and dissociation from the myo1c regulatory domain . Biochemistry 46 , 581 11718 - 11726 ( 2007 ) . 582 41 . B . J . Nolen et al . , Characterization of two classes of small molecule inhibitors of Arp2 / 3 complex . 583 Nature 460 , 1031 - 1034 ( 2009 ) . 584 42 . P . Bieling et al . , WH2 and proline - rich domains of WASP - family proteins collaborate to accelerate 585 actin filament elongation . EMBO J 37 , 102 - 121 ( 2018 ) . 586 43 . N . G . Pandit et al . , Force and phosphate release from Arp2 / 3 complex promote dissociation of 587 actin filament branches . Proc Natl Acad Sci U S A 117 , 13519 - 13528 ( 2020 ) . 588 44 . J . Pernier et al . , Myosin 1b flattens and prunes branched actin filaments . J Cell Sci 133 , ( 2020 ) . 589 45 . R . T . A . Pedersen , A . Snoberger , S . Pyrpassopoulos , D . Safer , D . G . Drubin , E . M . Ostap , 590 Endocytic myosin - 1 is a force - insensitive , power - generating motor . J Cell Biol 222 , ( 2023 ) . 591 46 . F . S . Wang , C . W . Liu , T . J . Diefenbach , D . G . Jay , Modeling the role of myosin 1c in neuronal 592 growth cone turning . Biophys J 85 , 3319 - 3328 ( 2003 ) . 593 47 . J . L . Maravillas - Montero , P . G . Gillespie , G . Patino - Lopez , S . Shaw , L . Santos - Argumedo , Myosin 594 1c participates in B cell cytoskeleton rearrangements , is recruited to the immunologic synapse , 595 and contributes to antigen presentation . J Immunol 187 , 3053 - 3063 ( 2011 ) . 596 48 . A . M . Sokac , C . Schietroma , C . B . Gundersen , W . M . Bement , Myosin - 1c couples assembling 597 actin to membranes to drive compensatory endocytosis . Dev Cell 11 , 629 - 640 ( 2006 ) . 598 49 . J . Cheng , A . Grassart , D . G . Drubin , Myosin 1E coordinates actin assembly and cargo trafficking 599 during clathrin - mediated endocytosis . Mol Biol Cell 23 , 2891 - 2904 ( 2012 ) . 600 50 . M . Krendel , E . K . Osterweil , M . S . Mooseker , Myosin 1E interacts with synaptojanin - 1 and 601 dynamin and is involved in endocytosis . FEBS Lett 581 , 644 - 650 ( 2007 ) . 602 51 . J . L . Ouderkirk , M . Krendel , Myosin 1e is a component of the invadosome core that contributes to 603 regulation of invadosome dynamics . Exp Cell Res 322 , 265 - 276 ( 2014 ) . 604 52 . M . E . Garone , S . E . Chase , C . Zhang , M . Krendel , Myosin 1e deficiency affects migration of 4T1 605 breast cancer cells . Cytoskeleton ( Hoboken ) , ( 2023 ) . 606 53 . Y . Sun , A . C . Martin , D . G . Drubin , Endocytic internalization in budding yeast requires 607 coordinated actin nucleation and myosin motor activity . Dev Cell 11 , 33 - 46 ( 2006 ) . 608 54 . J . M . Laakso , J . H . Lewis , H . Shuman , E . M . Ostap , Myosin I can act as a molecular force sensor . 609 Science 321 , 133 - 136 ( 2008 ) . 610 55 . J . A . Spudich , S . Watt , The regulation of rabbit skeletal muscle contraction . I . Biochemical studies 611 of the interaction of the tropomyosin - troponin complex with actin and the proteolytic fragments of 612 myosin . J Biol Chem 246 , 4866 - 4871 ( 1971 ) . 613 56 . D . R . Kellogg , T . J . Mitchison , B . M . Alberts , Behaviour of microtubules and actin filaments in 614 living Drosophila embryos . Development 103 , 675 - 686 ( 1988 ) . 615 57 . G . Lebreton et al . , Molecular to organismal chirality is induced by the conserved myosin 1D . 616 Science 362 , 949 - 952 ( 2018 ) . 617 58 . M . Boczkowska , G . Rebowski , M . V . Petoukhov , D . B . Hayes , D . I . Svergun , R . Dominguez , X - ray 618 scattering study of activated Arp2 / 3 complex with bound actin - WCA . Structure 16 , 695 - 704 619 ( 2008 ) . 620 59 . A . Zimmet , T . Van Eeuwen , M . Boczkowska , G . Rebowski , K . Murakami , R . Dominguez , Cryo - 621 EM structure of NPF - bound human Arp2 / 3 complex and activation mechanism . Sci Adv 6 , 622 ( 2020 ) . 623 60 . J . N . Rao , Y . Madasu , R . Dominguez , Mechanism of actin filament pointed - end capping by 624 tropomodulin . Science 345 , 463 - 467 ( 2014 ) . 625 61 . M . J . Greenberg , H . Shuman , E . M . Ostap , Measuring the Kinetic and Mechanical Properties of 626 Non - processive Myosins Using Optical Tweezers . Methods Mol Biol 1486 , 483 - 509 ( 2017 ) . 627 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 19 62 . D . M . Rutkowski , D . Vavylonis , Discrete mechanical model of lamellipodial actin network 628 implements molecular clutch mechanism and generates arcs and microspikes . PLoS Comput Biol 629 17 , e1009506 ( 2021 ) . 630 63 . A . E . Carlsson , Growth of branched actin networks against obstacles . Biophys J 81 , 1907 - 1923 631 ( 2001 ) . 632 64 . J . B . Alberts , G . M . Odell , In silico reconstitution of Listeria propulsion exhibits nano - saltation . 633 PLoS Biol 2 , e412 ( 2004 ) . 634 65 . D . Holz , D . Vavylonis , Building a dendritic actin filament network branch by branch : models of 635 filament orientation pattern and force generation in lamellipodia . Biophys Rev 10 , 1577 - 1585 636 ( 2018 ) . 637 66 . K . Sekimoto , J . Prost , F . Julicher , H . Boukellal , A . Bernheim - Grosswasser , Role of tensile stress 638 in actin gels and a symmetry - breaking instability . Eur Phys J E Soft Matter 13 , 247 - 259 ( 2004 ) . 639 67 . K . John , P . Peyla , K . Kassner , J . Prost , C . Misbah , Nonlinear study of symmetry breaking in actin 640 gels : implications for cellular motility . Phys Rev Lett 100 , 068101 ( 2008 ) . 641 68 . F . Gerbal , P . Chaikin , Y . Rabin , J . Prost , An elastic analysis of Listeria monocytogenes 642 propulsion . Biophys J 79 , 2259 - 2275 ( 2000 ) . 643 69 . J . Zhu , A . Mogilner , Mesoscopic model of actin - based propulsion . PLoS Comput Biol 8 , 644 e1002764 ( 2012 ) . 645 70 . N . J . Burroughs , D . Marenduzzo , Nonequilibrium - driven motion in actin networks : comet tails and 646 moving beads . Phys Rev Lett 98 , 238302 ( 2007 ) . 647 71 . K . C . Lee , A . J . Liu , New Proposed Mechanism of Actin - Polymerization - Driven Motility . 648 Biophysical Journal 95 , 4529 - 4539 ( 2008 ) . 649 72 . T . Kim , W . Hwang , H . Lee , R . D . Kamm , Computational analysis of viscoelastic properties of 650 crosslinked actin networks . PLoS Comput Biol 5 , e1000439 ( 2009 ) . 651 73 . C . S . Peskin , G . M . Odell , G . F . Oster , Cellular motions and thermal fluctuations : the Brownian 652 ratchet . Biophys J 65 , 316 - 324 ( 1993 ) . 653 74 . A . Mogilner , G . Oster , Cell motility driven by actin polymerization . Biophys J 71 , 3030 - 3045 654 ( 1996 ) . 655 75 . Y . Tsuda , H . Yasutake , A . Ishijima , T . Yanagida , Torsional rigidity of single actin filaments and 656 actin - actin bond breaking force under torsion measured directly by in vitro micromanipulation . 657 Proc Natl Acad Sci U S A 93 , 12937 - 12942 ( 1996 ) . 658 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 20 Figures 659 660 661 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 21 Figure 1 . Alteration of comet tail morphology by myosin - I . ( A ) Schematic of ( left ) control - and ( right ) 662 myosin - beads coated with NPF and neutravidin , where neutravidin is further conjugated with ( control - bead ) 663 biotinylated CF640 fluorescent dye or ( myosin - bead ) biotinylated Drosophila Myo1d . ( B ) Time - lapse 664 sequence of a ( red ) myosin - bead walking on a ( green ) single - actin - filament track . Speed ~ 8 nm / s . Scale 665 bar 2 \u00b5m . ( C ) Time series of the growth of actin comet tail from symmetry breaking to generation of a 666 polarized comet tail from ( top ) control - bead and ( bottom ) myosin - bead . Scale bar 5 \u00b5m . ( D ) Phase map 667 showing representative examples of 2 \u00b5m - diameter beads coated with a range of myosin - densities in the 668 presence of 30 \u2013 200 nM CP , 4 \u00b5M ( 5 % Rhodamine labeled ) actin , and 200 nM Arp2 / 3 complex . For each 669 myosin density condition , the left column show control - beads that were acquired in the same imaging field 670 as the myosin - beads in the right column . The myosin and NPF densities were determined by SDS - PAGE 671 gel , where the NPF density is the same for group 0 . 28 : 1 , 0 . 35 : 1 and 0 . 43 : 1 , ~ 6000 um - 2 , and ~ 4000 um - 2 672 for group 0 . 80 : 1 . Images were acquired 20 - 35 min after mixing . Brightness and contrast were linearly set 673 to be the same for each control - and myosin - bead pair but different among panels at different conditions 674 for better visualization . Scale bar 5 \u00b5m . 675 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 22 676 677 Figure 2 . Disruption and fracturing of actin shells by myosin - I . ( A ) Time series of ( top ) control - beads 678 and ( bottom ) 0 . 43 : 1 myosin - beads acquired in the same imaging field in the presence of 200 nM CP 679 showing the inability of myosin - beads to form a comet tail . ( B ) Time series of ( top ) control - beads and 680 ( bottom ) 0 . 43 : 1 myosin - beads acquired in the same field in the presence of 25 nM CP showing fracturing 681 of the actin shell and comet tail growth from a myosin - bead but not the control - bead . Conditions : 4 \u00b5M actin 682 ( 5 % Rhodamine labeled ) , 200 nM Arp2 / 3 complex , 200 nM or 25 nM CP . Scale bar 5 \u00b5m . 683 A 180 s 380 s 580 s 780 s 980 s 1180 s 1380 s 116 s 216 s 316 s 416 s 516 s 616 s 716 s C on t r o l M y o s i n B C o n t r o l M y o s i n . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 23 684 685 Figure 3 . Myosin - I decreases the actin density of comet tails . ( A ) Time - series of ( top ) control - and 686 ( bottom ) 0 . 43 : 1 myosin - beads growing comet tails in the presence of 50 nM CP . The actin network growing 687 from the myosin - bead is less dense but has a similar tail length as the control . Conditions : 4 \u00b5M actin ( 5 % 688 Rhodamine labeled ) , 200 nM Arp2 / 3 complex and 50 nM CP . Scale bar 5 \u00b5m . ( B ) Comparison of the mean 689 fluorescence intensities ( Fluor . Int . / Area ) of actin comet tails grown from control - and myosin - beads , 690 captured 20 min after mixing . The solid lines connect experimental pairs ( N = 5 , n = 11 ) . ( C ) Actin comet tail 691 length as a function of time and ( D ) tail growth rates derived from the slopes of the time courses in ( C ) . ( E ) 692 Comet tail fluorescence as a function of time . Control - and myosin - bead experimental pairs are normalized 693 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 24 to the average fluorescence level of the control - beads from 1100 s \u2013 1300 s . ( F ) Rate of fluorescent actin 694 incorporation into comet tails derived from the slopes of the time courses in ( E ) . ( C , E ) Large points show 695 the averaged value at binned time interval ( every 100 sec ) . Traces are from individual beads with each 696 myosin - bead acquired with a control - bead in the same field of view ( N = 5 , n = 11 ) . Error bars are SD . ( G ) 697 Growth efficiency for control and myosin - beads . Efficiency is defined as comet tail length per unit actin 698 fluorescence intensity . Box plots ( B , D , F , G ) show median ( center line ) , interquartile range ( box ) and min - 699 max values ( whiskers ) . p values were calculated using two - tail paired t test . Each point represents a control - 700 and myosin - bead pair acquired in same field of view . See also Supplementary Fig . 701 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 25 702 703 Figure 4 . Comets grown from myosin - beads incorporate less Arp2 / 3 complex . ( A ) Representative 704 actin comet tails assembled from ( top ) control - and ( bottom ) myosin - beads showing ( magenta ) actin and 705 ( green ) SNAP - Arp2 / 3 complex . Images were acquired approximately 25 - 35 min after mixing . Brightness 706 and contrast were set differently for actin and SNAP - Arp2 / 3 complex panels for visualization . Conditions : 4 707 \u00b5M actin ( 5 % Rhodamine labeled ) , 200 nM Arp2 / 3 complex ( 80 % SNAP - Surface 488 labeled ) , 40 nM CP . 708 Scale bar is 5 \u00b5m . ( B ) Total actin fluorescence intensity , ( C ) total SNAP - Arp2 / 3 fluorescence intensity and 709 ( D ) actin to SNAP - Arp2 / 3 fluorescence ratio over the entire comet tail region . ( E ) Total SNAP - Arp2 / 3 710 fluorescence intensity on bead surface . Plot shows median ( center line ) , interquartile range ( box ) and min - 711 max values ( whiskers ) . p values were calculated using two - tail paired t test . Each point represents a pair 712 of control - and myosin - beads ( N = 2 , n = 33 ) 713 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 26 714 715 Figure 5 . The myosin power stroke is required for altering network architectures . ( A ) Time - series of 716 actin assembly around ( top ) control and ( bottom ) rigor myosin - beads at 50 nM CP . Rigor myosin heavily 717 delayed the growth of actin comet tails . ( B ) Representative images of actin comet tails grown under different 718 ATP concentrations ( top : control beads ; bottom : myosin - beads ) . Adding ATP back rescued comet tail 719 growth . Images were acquired 20 - 30 min after mixing . ( C ) Representative actin network patterns assembled 720 on control - , myosin - , rigor - myosin , and HaloABD - beads under three different CP concentrations , as 721 indicated . Images were acquired 20 \u2013 30 min after mixing . Brightness and contrast were set to the same 722 values for each panel . ( D ) Representative actin comet tails generated by ( top ) control - and ( bottom ) myosin - 723 beads in the absence and presence of 100 \u00b5M free Ca 2 + at 50 nM CP . Images were acquired approximately 724 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 27 15 - 20 min after mixing . ( E ) Length and ( F ) network density quantified by total fluorescence intensity per 725 area for control - and myosin - beads in the absence and presence of 100 \u00b5M free Ca 2 + ( N = 1 , n = 15 ) . Control - 726 and myosin - bead experimental pairs are normalized to the average fluorescence level of the control - beads . 727 Box plots ( E , F ) show median ( center line ) , interquartile range ( box ) and min - max values ( whiskers ) . p 728 values were calculated using two - tail paired t test . Conditions : 4uM actin ( 5 % Rhodamine labeled ) , 200 nM 729 Arp2 / 3 complex , 15 - 200 nM CP , 0 - 1 mM ATP , as indicated . Scale bar 5 \u00b5m . 730 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 28 731 732 Figure 6 . The myosin power stroke can fracture the actin shell . ( A ) Representative actin shells of 733 control - , myosin - beads assembled under 50 nM CP , arrested by adding 20\u03bcM ( 5 molar excess ) of 734 Latrunculin B and CK - 666 before symmetry breaking ; as well as myosin - beads assemble under the same 735 conditions but arrested with the addition of 10mM ADP to inhibit myosin power stroke . Myosin - beads 736 fractured and ejected from the actin shell , while control - and myosin - beads ( with 10mM ADP ) remained 737 enclosed in the shell . Control - bead assembled under 100 nM CP showing similar network density as the 738 myosin - beads , didn\u2019t show shell fracture or bead ejection . Image was captured approximately 40min after 739 arrest . ( B ) The extents of shell breaking was classified by shell - breaking angle \u03b8 : \u03b8 = 0 , no symmetry 740 breaking ; 0 < \u03b8 < 180 , shell fracture ; \u03b8 \u2265 180 , bead ejected . Scale bar is 5\u03bcm . ( C ) Percentage of populations 741 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 29 with different extents of shell breaking for control ( 50 nM CP ) ( n = 182 ) , myosin ( n = 158 ) , myosin ( with 10mM 742 ADP ) ( n = 89 ) , control ( 100 nM CP ) ( n = 65 ) , N = 2 . Conditions : 4\u03bcM actin ( 5 % Rhodamine labeled ) , 200 nM 743 Arp2 / 3 , 50 or 100 nM CP . 20\u03bcM ( 5 molar excess ) phalloidin and 2 mM ATP were also added to prevent 744 actin depolymerization and preserve myosin motor activity . Actin assembly was arrested 100s after mixing . 745 Scale bar is 5 \u00b5m . 746 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 30 747 748 Figure 7 . Filament level model of actin comet tail recapitulates experimental results . ( A ) Schematic 749 of model of actin polymerization around the nucleating bead . Model includes filament level nucleation , 750 branching , filament fragmentation , debranching , capping , and force exerted by implicit myosin . ( B ) 751 Timelapse of symmetry breaking event under intermediate capping condition ( branch length = 0 . 5 \u03bcm ) with 752 no myosin . Color scale indicates filament tension ( Red : tensile ; Blue : compressive ) . ( C ) Tension 753 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 31 distributions within actin comet tails formed at either control ( 0 . 0 pN ) or two myosin forces ( 0 . 2 pN or 0 . 4 754 pN ) under intermediate capping condition ( branch length = 0 . 5 \u03bcm ) . Image shows a cut through the center 755 of the comet tail . Color scale indicates filament tension ( Red : tensile ; Blue : compressive ) . ( D ) Simulated 756 timelapse of epifluorescence images under different myosin forces ( branch length = 0 . 5 \u03bcm ) . ( E ) Elongation 757 speed ( F ) actin intensity and ( G ) Arp2 / 3 complex intensity for simulated beads as a function of myosin force 758 ( branch length = 0 . 5 \u03bcm ) . ( H ) Forces acting on beads along the direction of bead propulsion due to actin 759 polymerization ( green ) and myosin pushing ( orange ) as a function of the myosin force ( branch length = 0 . 5 760 \u03bcm ) . ( I ) Orientation of filaments around the beads ( within 0 . 15 \u03bcm of the bead surface ) as a function of 761 myosin force ( branch length = 0 . 5 \u03bcm ) . ( J ) Simulated symmetry breaking after actin polymerization and 762 branching arrest . Actin was allowed to polymerize around the bead for 42 . 7 s ( 0 . 2 pN myosin force with 0 . 5 763 \u03bcm filament length ) to form a shell of intermediate thickness before halted . The time of arresting was set 764 as t = 0 s . 765 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 32 766 Figure 8 . Schematic of how myosin - I modulates actin network structure through its power - strok 767 Myosin - I Actin Arp2 / 3 Complex NPF - coated surface A B F . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 33 768 Supplementary Materials for 769 770 Myosin - I Synergizes with Arp2 / 3 Complex to Enhance Pushing 771 Forces of Branched Actin Networks 772 773 Mengqi Xu et al . 774 775 * Corresponding author . Email : ostap @ pennmedicine . upenn . edu , vavylonis @ lehigh . edu , 776 droberto @ pennmedicine . upenn . edu , Luther . Pollard @ pennmedicine . upenn . edu 777 778 779 780 781 782 This PDF file includes : 783 784 Supplementary Text 785 Figs . S1 to S8 786 Tables S1 787 Movies S1 to S13 788 789 Other Supplementary Materials for this manuscript include the following : 790 791 Movies S1 to S13 792 793 794 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 34 Supplementary Text 795 Myosin steric hindrance effect 796 We quantified the average spatial occupancy of individual components ( NPF or myosin - I ) bound 797 to bead surface for the first three myosin / NPF ratio groups ( 0 . 28 , 0 . 35 , 0 . 43 ) , all of which maintain 798 the same NPF densities ( ~ 6000 / um 2 ) . We found that individual molecular components occupied 799 similar areas in all three groups : 11 . 4x1 . 4 nm 2 / molecule for 7680 molecules in 0 . 28 : 1 800 ( myosin / NPF ) group , 11 . 1x11 . 1 nm 2 / molecule for 8100 molecules in 0 . 25 : 1 ( myosin / NPF ) group 801 and 10 . 7x10 . 7 nm 2 / molecule for 8580 molecules in 0 . 43 : 1 ( myosin / NPF ) group . Therefore , the 802 steric hindrance effect in each of the three groups should be nearly identical . In the 0 . 28 : 1 803 ( myosin / NPF ) group , both control - and myosin - beads displayed similar actin comet patterns , 804 which implies that the steric hindrance of myosin - I does not play a significant role in modifying the 805 Arp2 / 3 - medicated branching under the current conditions . 806 807 Bead gliding assay on actin filament tracks 808 F - actin was made by polymerizing 25 \u03bcM G - actin in MB buffer ( 10 mM MOPS [ pH7 . 0 ] , 25 mM 809 KCl , 1 mM EGTA , 1 mM MgCl 2 , 1 mM DTT ) with 2 mM ATP at room temperature for 1 hr , and 810 stabilized with AlexaFluor 488 Phalloidin ( ThermoFisher Cat . # A12379 ) . Myosin - beads were 811 prepared following the procedures outlined in Methods in the main text and labeled with TRITC 812 during the blocking process using 10 % TRITC - labeled BSA ( Sigma , Cat . # A2289 ) . Flow 813 chambers were made by sandwiching the plasma cleaned slides and coverslips ( 22 mm \u00d7 22 814 mm ) with two pieces of double - stick tape , forming a 5mm - wide channel that can hold a volume of 815 up to 10 \u03bcL . To immobilize single - actin - filament tracks on the coverslip , we coated the flow 816 chamber with 0 . 5 mg / mL NEM - myosin , incubated for 5 min , followed by two washes with 2 817 mg / mL casein ( 1 min for each ) to remove any unbound NEM - myosin and block the remaining free 818 spaces on the surface . Next , 5 - 10 nM AlexaFluor 488 Phalloidin stabilized F - actin was loaded 819 into the flow chamber to avoid any overlapping between the actin filaments , incubated for 2 min , 820 followed by a wash with 20uL MB buffer . Finally , 10 uL of the 50 uL bead motility mixture 821 containing 3uL TRITC - labeled myosin - bead slurry ( # myosin / NPF = 0 . 4 : 1 ) , 2 \u03bcM Calmodulin , 0 . 2 822 mg / mL glucose oxidase , 0 . 04 mg / mL catalase and 5 mg / mL glucose in MB buffer supplemented 823 with 5 mM Mg - ATP and 40 mM DTT , was loaded into the flow chamber , sealed with vacuum 824 grease and imaged immediately under the microscope . 825 826 Myosin inhibition experiments 827 828 Rigor myosin was prepared by excluding ATP from the motility mixture . Gel - filtered actin 829 monomers were incubated with 2 mM ATP for 15 minutes on ice , followed by passage through a 830 G25 Sephadex column to remove residual free nucleotides . The resulting ATP - treated actin 831 exhibited similar polymerization capability in ATP - free mixtures as to those containing 1 mM ATP . 832 Alternatively , myosin was inhibited by introducing 10 mM Mg - ADP to prolong the lifetime of the 833 myosin ADP - bound state . 834 835 Actin gliding assay 836 Coverslips were plasma - cleaned and coated with 0 . 5 % Nitrocellulose and used to make a flow 837 chamber . Flow chamber was first incubated with 0 . 1 mg / mL neutravidin for 2 min followed by 838 blocking with 2 mg / mL casein , twice , 1 min each . Then , 250 nM biotinylated Drosophila Myo1d 839 was loaded into the chamber , incubated for 2 min , and washed with 20uL MB buffer ( 10 mM 840 MOPS [ pH7 . 0 ] , 25 mM KCl , 1 mM EGTA , 1 mM MgCl 2 , 1 mM DTT ) . Next , 20 nM F - actin 841 stabilized with AlexaFluor 488 Phalloidin was loaded into the chamber , incubated for 2 min , 842 washed with 20uL MB buffer . Finally , the imaging mixture containing 8 . 4 \u03bcM Calmodulin , 1 mM 843 Mg - ATP , 40 mM DTT in 20 mM HEPES [ pH7 . 5 ] , 100 mM KCl , 1 mM MgCl 2 , 1 mM EGTA and 844 0 . 2 % methylcellulose was loaded into the chamber and observed immediately . For experiments 845 with calcium , 8 . 4 \u03bcM Calmodulin was replaced by 1 . 1 mM CaCl 2 to disrupt the myosin power - 846 stroke . 847 848 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 35 Actin arresting 849 Actin assembly was quenched by combining equal volume of motility mixture and arresting 850 solution , which contains 20 \u03bcM ( 5 molar excess ) of both Latrunculin B and CK - 666 ( together with 851 20 \u03bcM phalloidin for ' + Phalloidin ' experiments ) in 20 mM HEPES [ pH 7 . 5 ] , 100 mM KCl , 100 852 \u03bcg / ml BSA , 0 . 5 mM TCEP . Quenched motility mixture was either applied on a squeezer chamber 853 or loaded into a NEM - myosin coated flow chamber ( to make sure the observed shell - breaking 854 was not due to the squeezing effect between the slides and the coverslips ) and imaged under the 855 microscope immediately after sealing with vacuum grease . 856 857 Computational modeling 858 859 Reduction of de novo and branch nucleation rate due to crowding 860 If upon introduction to the system a de novo filament or new branch would strongly overlap with 861 other filaments or the nucleating bead , their insertion is rejected . Specifically , their introduction is 862 rejected if the closest distance between a newly added segment and an existing segment is less 863 than / 4 , or if the distance between the newly added branch and the nucleating bead is less than 864 ( \ud835\udc51 - 01 / 2 # - # / 2 + \ud835\udc45 ) / 4 . No explicit reattempts at introduction of a rejected filament are performed , 865 and the rate of branching , \ud835\udc5f ! , ) . * 1 < , and the rate of de novo introduction , \ud835\udc5f ! # - * 5 = 5 , are left 866 unchanged by a rejection event . 867 Simulation timescale 868 We wrote C + + code with force parallelization using OpenMP . For the case of Fig . 7B ( 0 . 5 \u03bcm 869 length filaments with no myosin force ) it takes 360 hours on 10 CPU cores to run the simulation to 870 260 s . The amount of time to run a simulated second increases as the simulation proceeds since 871 no filament segments ( beyond those that are severed ) are removed from the system . Simulations 872 with myosin or with shorter filaments take a shorter amount of time to run the same simulated 873 time since they have overall less filament segments at the same point in time . 874 Simulated fluorescence images 875 To generate simulated fluorescence images of actin comet tail formation we first rotate the 876 system so the net displacement of the nucleating bead occurs along the x - direction . For a given 877 frame , each actin or Arp2 / 3 complex spring bond centroid is placed in a 2d grid with gridlines 878 along the x and y axes and bin size 0 . 05 \u03bcm ( z - component is not considered ) . Actin filament 879 segments of length \ud835\udc59 ! contribute 37 actin monomers to the bin they are assigned , while shorter 880 filament segments contribute 37 \ud835\udc59 / \ud835\udc59 ! monomers , where \ud835\udc59 is the length of the segment . Arp2 / 3 881 complex bonds contribute a single Arp2 / 3 molecule to the bin they are placed in . A grayscale 882 image is constructed from the grid where each voxel in the image corresponds to a single bin and 883 the intensity corresponds to the sum of the count of monomers in each bin . Finally , a Gaussian 884 filter with standard deviation 0 . 1 \u03bcm is applied to the image to smooth the image to a comparable 885 level as in experiment . 886 Measurement of Forces 887 Forces in Figure 7H and Figure S7 are calculated by first rotating the system so the net 888 displacement of the nucleating bead occurs along the x - direction . Excluded volume forces acting 889 between actin segments and the nucleating bead are spatially divided into three regions based on 890 the closest approach point of the actin segment to the nucleating bead . Forces where the closest 891 approach point is greater than R / 4 behind the nucleating bead center in the x - direction are 892 considered back forces . Forces where the closest approach point is greater than R / 4 in front of 893 the nucleating bead center in the x - direction are considered front forces . All other forces are 894 labeled as side forces . Excluded volume forces are further subdivided into whether they are 895 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 36 caused by filament segments at their maximum length \ud835\udc59 ! ( actin excluded volume ) or filament 896 segments with length less than \ud835\udc59 ! ( actin polymerization ) . 897 Measurement of filament orientation with respect to nucleating bead 898 Filament angle distributions with respect to the nucleating bead , for filaments close to the 899 nucleating bead in Figure 7I are calculated by first rotating the system so the net displacement of 900 the nucleating bead occurs along the x - direction . Only filaments segments with length less than 901 \ud835\udc59 ! that are also in the back region ( as defined in Measurement of Forces ) and are also closer to 902 the nucleating bead surface than 0 . 15 \u03bcm are considered . For these actively growing filaments 903 we measure the angle between the filament displacement vector and the vector joining the 904 midpoint of the filament segment to the nucleating bead center . This angle is zero if the filament 905 is perpendicular to the nucleating bead surface and 90\u00b0 if the filament is lying flat on the bead 906 surface . We generate a probability distribution of these angles over approximately 70 s and when 907 the nucleating bead is moving at a relatively constant velocity . The probability of finding the angle 908 to be less than 70\u00b0 from these distributions is reported in Figure 7I . 909 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 37 910 911 Fig . S1 . Phase diagram of actin comet tails assembled under different CP concentrations 912 and myosin densities . Experimental points demonstrate control - and myosin - bead pairs 913 assembled under each specified conditions . Star \u2013 myosin - beads promote symmetry breaking ; 914 Triangle - myosin - beads have longer tails than control ; Circle \u2013 myosin - beads exhibit similar tail 915 length to control ; Square \u2013 myosin - beads have shorter tails than control ; Diamond \u2013 myosin - beads 916 shows no comet tail . For extremely low CP concentration ( e . g . 6 . 5 nM , Crossline ) , both myosin - 917 coated and control - beads displayed aster - like structure . No significant difference was observed 918 between the two types of beads . Conditions : 4 \u00b5M G - actin , 200 nM Arp2 / 3 complex , 0 - 200 nM of 919 CP as indicated in graph . 920 0 . 27 0 . 34 0 . 42 0 . 71 0 50 100 200 Myosin Density ( # myosin / NPF ) C P ( n M ) . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 38 921 Fig . S2 . Myosin - I decreases the actin density in comet tails ( supplement for Fig . 3 in the 922 main text ) . ( A ) Time - series of ( top ) control - and ( bottom ) 0 . 43 : 1 myosin - beads growing comet tails 923 in the presence of 50 nM CP . The actin network growing from the myosin - bead is less dense but 924 has a similar tail length as the control . ( B ) Actin comet tail length as a function of time ( C ) Comet 925 tail fluorescence as a function of time . The fluorescent intensity of control - and myosin - bead 926 experimental pairs are normalized to the average fluorescence level of the control - beads from 927 1100 - 1300 s . Large points show the averaged value at binned time interval ( every 100 sec ) . Traces 928 are from individual beads with each myosin - bead acquired with a control - bead in the same field of 929 view ( N = 5 , n = 11 ) . Error bars are SD . Conditions : 4 \u00b5M actin ( 5 % Rhodamine labeled ) , 200 nM 930 Arp2 / 3 complex and 50 nM CP . Scale bar 5 \u00b5m . 931 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 39 932 Fig . S3 . The myosin power - stroke is required for altering network architectures ( supplement 933 for Fig . 5 in the main text ) . ( A ) Time - series of actin assembly around ( top ) control - and ( bottom ) 934 HaloABD - bead at 50 nM CP . HaloABD of \ud835\udf36 - actinin heavily delayed the growth of actin comet tail 935 as rigor myosin . Scale bar 5 \u00b5m . ( B ) Quantification of comet tail length generated under different 936 ATP concentrations ( 0 - 1 mM ) for control - , myosin - and HaloABD - beads ( N = 2 , n > 15 ) . Error bars 937 are SD . p values were calculated using unpaired two - tail t test . ( C ) Network density quantified by 938 total fluorescence intensity per area for control - , myosin - , rigor - myosin and HaloABD - beads . 939 Normalized to the mean value of control - beads ( N = 2 , n > 11 ) . Dot plots show mean \u00b1 SD . p values 940 were calculated using unpaired two - tail t test . ( D ) Percentages of control - , myosin - , rigor - myosin 941 and HaloABD - beads that performed symmetry breaking or no symmetry breaking at 15nM CP 942 ( N = 3 , n > 23 ) . ( E ) Time collapsed images of actin filaments gliding ( with no Ca 2 + ) or swirling ( with 943 100 \u00b5M free Ca 2 + ) on a Myo1d - coated surface . Rainbow bar represents timescale over 61 frames 944 with 5 s frame interval . Scale bar 5 \u00b5m . ( F ) Quantification of actin gliding speed with and without 945 100 \u00b5M free Ca 2 + . ( G ) Total actin fluorescence intensity and ( H ) Growth efficiency for control - and 946 myosin - beads with and without the presence of 100 \u00b5M calcium ( N = 1 , n = 15 ) . Efficiency is defined 947 as comet tail length per unit actin fluorescence intensity . Box plots ( G , H ) show median ( center 948 line ) , interquartile range ( box ) and min - max values ( whiskers ) . p values were calculated using two - 949 tail paired t test . 950 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 40 951 Fig . S4 . Myosin power - stroke alone can break sparse actin shell ( w / o Phalloidin ) . ( A - B ) . 952 Time - lapse series of ( top ) control - and ( bottom ) myosin - bead that performed ( A ) shell fracture or 953 ( B ) bead ejection . Control - and myosin - beads were mixed with 4 \u03bcM actin ( 5 % Rhodamine labeled ) , 954 200 nM Arp2 / 3 , and 50 nM CP , incubated for 100s , and then actin assembly was arrested by adding 955 20\u03bcM ( 5 molar excess ) of Latrunculin B and CK - 666 . ( C ) Status of shell breaking approximately 956 50min after arrest . The extents of shell breaking was classified by shell - breaking angle \u03b8 . ( D ) 957 Percentage of different populations with different extents of shell breaking . Without Phalloidin : 958 control ( n = 141 ) , myosin ( n = 55 ) , myosin with 10mM ADP ( n = 125 ) ; With Phalloidin : control ( n = 182 ) , 959 myosin ( n = 158 ) , myosin with 10mM ADP ( n = 89 ) . Conditions : 4 \u03bcM actin ( 5 % Rhodamine labeled ) , 960 200 nM Arp2 / 3 complex , 50 nM or 100 nM CP was incubated for 100 s and arrested by adding 5 961 molar excess of Latrunculin B , CK - 666 and phalloidin ( for \u2018 + Phalloidin\u2018 experiments ) or 10mM ATP 962 ( for \u2018myosin with 10 mM ADP\u2019 experiments ) . Scale bar 5\u03bcm . 963 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 41 964 965 Figure S5 . Simulated comet tail structure and tensile / compression force distributions as a 966 function of branch length and myosin force F myo . ( A ) Simulated epifluorescence images as in 967 Figure 7D , but with filaments grow to a length of 0 . 3 \u03bcm before capping . The simulations can be 968 compared to high CP concentration experiments of Figure 2A where a comet tail fails to form for 969 myosin - beads . ( B ) Same as in panel A , but for filament length of 2 . 0 \u03bcm . The simulations mimic 970 the low CP concentration experiments of Figure 2B where myosin was found to promote symmetry 971 breaking . ( C ) State diagram for actin comet tails as a function of filament length and myosin force . 972 Images show a cut through the center of the comet tail . Color scale indicates filament tension ( red : 973 tensile ; blue : compressive ) . 974 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 42 975 976 Figure S6 . Simulated actin intensity , Arp2 / 3 intensity and Actin : Arp2 / 3 ratio across the entire 977 actin comet tail for intermediate length condition ( intermediate CP concentration ) under 978 different debranching thresholds . ( A ) Actin : Arp2 / 3 complex ratio ( calibrated to number of actin 979 monomers per Arp2 / 3 complex ) , average actin intensity , and average Arp2 / 3 complex intensity 980 from simulated epifluorescent images of comet tails formed at reference debranching conditions 981 with filament length 0 . 5 \u03bcm ( \u2248 187 subunits ) . Solid circles indicate averages calculated where the 982 Arp2 / 3 complex is removed upon debranching ; empty squares indicate averages calculated where 983 the Arp2 / 3 complex remains attached to the daughter filament upon debranching . ( B ) Same as in 984 panel A except under enhanced debranching condition ( debranching occurring at \u00b110\u00b0 away from 985 70\u00b0 equilibrium branch instead of \u00b125\u00b0 as in the reference case in A ) . The ratio varies with myosin 986 force when Arp2 / 3 complex dissociates after debranching , but in the opposite manner as compared 987 to the experimental results in Figure 4D . 988 B A Enhanced Debranching ( Debranching at \u00b110\u00b0 from equilibrium 70\u00b0 angle ) Reference Debranching ( Debranching at \u00b125\u00b0 from equilibrium 70\u00b0 angle ) Arp2 / 3 complex dissociates Arp2 / 3 complex remains Arp2 / 3 complex dissociates Arp2 / 3 complex remains . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 43 989 990 Figure S7 . Force acting on beads during comet tail elongation . Subdivision of forces acting on 991 beads along the direction of bead propulsion due to actin polymerization ( light green , gray , dark 992 green continuous ; calculated as excluded volume interactions between barbed end segment and 993 bead ) , actin excluded volume ( light green , gray , dark green dashed ; calculated as excluded volume 994 interactions on the bead from filament segments other than barbed ends ) , and myosin forces on 995 the bead ( orange line ) . The bead drag force that balances the sum of these forces is shown in 996 black . Actin forces are subdivided into whether they are located behind the bead ( light green ) , at 997 the side of the bead ( gray ) , or in front of the bead ( dark green ) . 998 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 44 999 1000 Figure S8 . Simulated expulsion of bead after halting actin polymerization / branching . 1001 Simulation snapshots at intermediate filament length ( 0 . 5 \u03bcm ) ( intermediate CP concentration ) 1002 show a cut through the bead center during the evolution of the actin network after halting actin 1003 polymerization and branching . Expulsion of the bead only occurs at intermediate cloud thickness 1004 ( Growth time = 42 . 7 s , middle row and Fig . 7J ) . Thinner shells fall apart ( Growth time = 31 . 5 s ) , 1005 while thicker shells cannot be broken by 0 . 2 pN myosin alone ( Growth time = 59 . 85 s ) . Network 1006 evolved in the presence of the same myosin force ( 0 . 2 pN ) . Histograms show filament tension 1007 distribution at t = 0 s ( time of growth arrest ) ; dashed lines indicate region containing 95 % of data , 1008 centered at zero . 1009 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 45 Table S1 . Table of simulation constants 1010 Parameter Description Value Justification R Radius of nucleating bead 0 . 75 \u03bcm Smaller but comparable to experiment , for numerical efficiency \u2206\ud835\udc61 Simulation timestep 3 . 5 \u00d7 10 : G s Allows for numerical stability of excluded volume interactions \ud835\udf01 Friction coefficient for actin segment 0 . 377 pN s \u03bcm - 1 Sufficiently high for numerical stability \ud835\udf01 * 21 / , - . # Friction coefficient for nucleating bead 2970 pN s \u03bcm - 1 Sufficiently high for bead to not eject bare from actin shell \ud835\udc59 ( Actin filament persistence length 17 . 0 \u03bcm \ud835\udc58 . 1 % & * Actin filament segment spring constant 1000 pN \u03bcm - 1 As large as allowed for numerical stability ( 62 ) \ud835\udc59 ! Actin segment spring equilibrium distance 0 . 1 \u03bcm Length of filament segment \ud835\udc59 ! ! Starting filament length 0 . 035 \u03bcm Numerical stability \ud835\udc59 , ) . * 1 < Bond length between mother point element and pointed end of daughter filament 0 . 02 \u03bcm Numerical stability \ud835\udc51 , ) . * 1 < Range of branching nucleation 0 . 1 \u03bcm Of order \ud835\udc59 ! \ud835\udf16 . * + / - Spring constant for filament branch 2 pN \u00b5m Maintains angle between 60\u00b0 - 80\u00b0 with thermal fluctuations ( 62 ) \ud835\udc51 - 01 / 2 # - # Excluded volume interaction range between filaments 0 . 02 \u03bcm Larger than 7 nm filament diameter for increased numerical stability \ud835\udc58 - 01 / 2 # - # Excluded volume interaction constant 1690 pN \u03bcm - 1 Prior computational model ( 72 ) \ud835\udc51 345 Range of myosin force 0 . 025 \u03bcm Comparable to size of myosin 1d \ud835\udc39 345 Magnitude of myosin force acting on actin filaments within range . 0 . 0 \u2013 0 . 4 pN Varied \ud835\udc5f ! ( 5 / Rate of polymerization of uncapped barbed ends 40 sub s - 1 4 \u03bcM actin solution . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 46 \ud835\udc5f ! , ) . * 1 < Rate of branching per filament point element 0 . 15 s - 1 Tuned to approximate timescale of symmetry breaking and comet speed ( intermediate CP ) \ud835\udc5f ! # - * 5 = 5 Rate of spontaneous introduction of new filaments 0 . 125 s - 1 Tuned to allow actin shell on timescales comparable to experiment \ud835\udc39 F ) . + Threshold force for tension fragmentation of filaments 50 pN Allows symmetry breaking \ud835\udc39 # - , ) . * 1 < Threshold force for tension debranching of branches 30 pN Of order \ud835\udc39 F ) . + \ud835\udee5\ud835\udf03 # - , ) . * 1 < \u03a4threshold angle deviation from 70\u00b0 for debranching 25\u00b0 Debranching by bending and pulling ( 72 ) 1011 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint 47 Movie S1 . Myosin - bead ( red ) processively moving along a ( green ) single - actin - filament track . 1012 1013 Movie S2 . Movie of control - beads and myosin - beads ( 0 . 43 : 1 # myosin / NPFs ) acquired in the 1014 same imaging field in the presence of 200 nM CP showing the inability of myosin - beads to form a 1015 comet tail . Scale bar 5 \u00b5m . 1016 1017 Movie S3 . Movie of control - beads and myosin - beads ( 0 . 43 : 1 # myosin / NPFs ) acquired in the 1018 same field in the presence of 25 nM CP showing fracturing of the actin shell and comet tail growth 1019 from a myosin - bead but not the control - bead . Scale bar 5 \u00b5m . 1020 1021 Movie S4 . Movie of control - beads and myosin - beads ( 0 . 43 : 1 # myosin / NPFs ) growing comet tails 1022 in the presence of 50 nM CP . The actin network growing from the myosin - bead is less dense but 1023 has a similar tail length as the control . Scale bar 5 \u00b5m . 1024 1025 Movie S5 . Movie of actin assembly around control - beads and rigor myosin - beads ( 0 . 43 : 1 1026 # myosin / NPFs ) at 50 nM CP . Rigor myosin heavily delayed the growth of actin comet tails . Scale 1027 bar 5 \u00b5m . 1028 1029 Movie S6 . Movie of actin assembly around control - and HaloABD - bead at 50 nM CP . HaloABD of 1030 \ud835\udefc - actinin heavily delayed the growth of actin comet tail as rigor myosin . Scale bar 5 \u00b5m . 1031 1032 Movie S7 . Movie of myosin - beads ( 0 . 43 : 1 # myosin / NPFs ) break actin shell apart acquired 11 min 1033 after actin polymerization was arrested by Latrunculin B and CK - 666 in the absence of phalloidin . 1034 Scale bar 5 \u00b5m . 1035 1036 Movie S8 . Movie of myosin - beads ( 0 . 43 : 1 # myosin / NPFs ) ejected out of actin shell acquired 24 1037 min after actin polymerization was arrested by Latrunculin B and CK - 666 n the absence of 1038 phalloidin . Scale bar 5 \u00b5m . 1039 1040 Movie S9 Simulated symmetry breaking and comet growth for intermediate length ( 0 . 5 \u03bcm ) 1041 filaments , without myosin forces ( Fig . 7B ) . Top video shows a cutaway ( filament point elements 1042 with position z > 0 are not shown ) while bottom video shows all filaments . Bead is semi - 1043 transparent . 1044 1045 Movie S10 . Movie of simulated actin comet tail formation for short filaments ( 0 . 3 \u03bcm , High 1046 CP ) , _ as a function of myosin force ( Fig . S5A ) . 1047 1048 Movie S11 . Movie of simulated actin comet tail formation for long filaments ( 2 . 0 \u03bcm , Low CP ) , as 1049 a function of myosin force ( Fig . S5B ) . 1050 1051 Movie S12 . Movie of simulated actin comet tail formation for intermediate filaments ( 0 . 5 \u03bcm , 1052 Intermediate CP ) , as a function of myosin force ( Fig . 7D ) . 1053 1054 Movie S13 . Movie of simulated symmetry breaking and bead ejection by myosin forces following 1055 arrest of actin polymerization and branching at 42 . 7 s ( Fig 7I ) . Myosin force is 0 . 2 pN throughout 1056 the simulation and filament length is 0 . 5 \u03bcm . 1057 . CC - BY - NC - ND 4 . 0 International license available under a ( which was not certified by peer review ) is the author / funder , who has granted bioRxiv a license to display the preprint in perpetuity . It is made The copyright holder for this preprint this version posted February 12 , 2024 . ; https : / / doi . org / 10 . 1101 / 2024 . 02 . 09 . 579714 doi : bioRxiv preprint",
    "kosmalska2015physical": "ARTICLE Received 12 Dec 2014 | Accepted 24 Apr 2015 | Published 15 Jun 2015 Physical principles of membrane remodelling during cell mechanoadaptation Anita Joanna Kosmalska 1 , 2 , Laura Casares 1 , 2 , Alberto Elosegui - Artola 1 , Joseph Jose Thottacherry 3 , Roberto Moreno - Vicente 4 , V\u0131\u00b4ctor Gonza\u00b4lez - Tarrago\u00b4 1 , 2 , Miguel A\u00b4ngel del Pozo 4 , Satyajit Mayor 3 , Marino Arroyo 5 , Daniel Navajas 1 , 2 , 6 , Xavier Trepat 1 , 2 , 7 , Nils C . Gauthier 8 & Pere Roca - Cusachs 1 , 2 Biological processes in any physiological environment involve changes in cell shape , which must be accommodated by their physical envelope\u2014the bilayer membrane . However , the fundamental biophysical principles by which the cell membrane allows for and responds to shape changes remain unclear . Here we show that the 3D remodelling of the membrane in response to a broad diversity of physiological perturbations can be explained by a purely mechanical process . This process is passive , local , almost instantaneous , before any active remodelling and generates different types of membrane invaginations that can repeatedly store and release large fractions of the cell membrane . We further demonstrate that the shape of those invaginations is determined by the minimum elastic and adhesive energy required to store both membrane area and liquid volume at the cell \u2013 substrate interface . Once formed , cells reabsorb the invaginations through an active process with duration of the order of minutes . DOI : 10 . 1038 / ncomms8292 OPEN 1 Institute for Bioengineering of Catalonia ( IBEC ) , Barcelona 08028 , Spain . 2 Department of Physiological Sciences I , University of Barcelona , Barcelona 08036 , Spain . 3 National Centre for Biological Sciences ( TIFR ) , Bangalore 560065 , India . 4 Centro Nacional de Investigaciones Cardiovasculares ( CNIC ) , Madrid 28029 , Spain . 5 LaCa ` N , Universitat Polite ` cnica de Catalunya - BarcelonaTech , Barcelona 08034 , Spain . 6 Ciber Enfermedades Respiratorias , Madrid 28029 , Spain . 7 Institucio\u00b4 Catalana de Recerca i Estudis Avanc \u00b8 ats ( ICREA ) , Barcelona 08010 , Spain . 8 Mechanobiology Institute , National University of Singapore , Singapore 117411 , Singapore . Correspondence and requests for materials should be addressed to N . C . G . ( email : mbinclg @ nus . edu . sg ) or to P . R . - C . ( email : rocacusachs @ ub . edu ) . NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications 1 & 2015 Macmillan Publishers Limited . All rights reserved . P hysiological processes in development , wound healing , breathing or any other scenario generally involve cell shape variations , which are constrained by the physical envelope of cells\u2014the plasma membrane . In any such process , the plasma membrane must adapt to often fast cell rearrangements , a requirement that is at odds with the very low membrane extensibility / compressibility given by its high stretching elastic modulus 1 , 2 . Other than simple extension and compression , the regulation of membrane area and shape therefore requires additional mechanisms , which could include active cell processes like endocytosis and exocytosis 3 \u2013 5 or the formation and \ufb02attening of membrane invaginations / evaginations , either at the micron scale as in membrane folds 6 , 7 , blebs 8 or vacuole - like dilations ( VLDs ) 9 or at the nanoscale as in caveolae 10 . However and despite extensive work on membrane mechanical interactions 11 \u2013 15 , there is no clear physical understanding of the manner in which the cell membrane responds to changes in area and shape while remaining highly con\ufb01ned by adjacent cells or substrates . Here we show that in response to changes in the area and volume of adherent cells , membrane remodelling occurs through a mechanical process that is passive , local , almost instantaneous and before any active response . This process generates invagina - tions with shapes that minimize the elastic and adhesive energy required to store both membrane area and liquid volume at the cell \u2013 substrate interface . Once formed , cells reabsorb the invagi - nations through an active process with duration of the order of minutes . Results Membrane response to changes in area and shape . To under - stand how cell membranes respond to changes in area and shape , we labelled the membrane of mouse embryonic \ufb01broblasts ( MEFs ) by transfection with a membrane \ufb02uorescent marker ( pEYFP - mem ) and seeded them on \ufb01bronectin - coated poly ( dimethylsiloxane ) ( PDMS ) membranes . We observed membrane dynamics after modifying two different cell shape parameters : cell volume ( regulated through changes in medium osmolarity ) and cell spreading area ( regulated through a custom - built biaxial stretch device , Supplementary Fig . 1 ) . After submitting cells to 6 % linear strain ( corresponding to a 12 % increase in surface ) , we noted that additional required area was obtained by \ufb02attening membrane ruf\ufb02es ( Fig . 1a and Supplementary Movie 1 ) . If stretch magnitude was increased , however , the membrane reservoir was depleted and the membrane teared within 3 min of constant stretch application ( Fig . 1a , e and Supplementary Movie 1 ) . In contrast , exposure to medium with a 50 % reduction in osmolarity for 3 min increased cell volume by 20 % , but only required an increase in plasma membrane area of 2 % ( Fig . 1c , d ) . By itself , this small increase in required area had a negligible effect on the membrane , as checked after stretching cells by 2 % ( Supplementary Fig . 2 ) . Accordingly , 50 % hypo - osmotic shock did not eliminate ruf\ufb02es ( Fig . 1b and Supplementary Movie 2 ) , and exposure to 100 % deionized water was required to eliminate membrane ruf\ufb02es through cell swelling , or to lyse the membrane ( Fig . 1b , f and Supplementary Movie 2 ) . After 3 min of stretch application , the release of stretch resulted in the accumulation of excess membrane in small membrane reservoirs of 0 . 5 \u2013 1 m m in diameter ( Fig . 2i \u2013 k ) , which extended from both the cell ventral and dorsal surfaces ( Fig . 2a , c and Supplementary Movie 3 ) . Through pEYFP - mem \ufb02uorescence quanti\ufb01cation , we calculated that these reservoirs stored approxi - mately 15 % of the total projected cell \u2013 substrate membrane area ( see methods ) , roughly matching the 12 % change in area associated with 6 % linear biaxial strain . Reservoirs were resorbed and eliminated by cells within B 2 min ( Fig . 2b ) , although their formation / resorption dynamics depended on temperature and stretch magnitude ( see Supplementary Note 1 and Supplementary Fig . 3 ) . Further , reservoirs appeared in open spaces devoid of actin \ufb01bres and focal adhesions , suggesting that membrane invaginations avoided cytoskeletal resistance ( Fig . 2d ) . Similarly , re - application of isotonic medium after 3 min of exposure to 50 % hypo - osmotic medium resulted in the formation of membrane invaginations in the cell ventral surface , which were however larger ( B 2 m m in diameter ) and with a spherical cap shape ( Fig . 2e , g and Supplementary Movie 3 ) . The invaginations were quanti\ufb01ed to store B 2 % of projected cell \u2013 substrate membrane area , thereby also matching the associated membrane requirement imposed by the osmotic shock ( Fig . 1d ) . Those invaginations were also eliminated by cells within B 3 min ( Fig . 2f ) , and their formation / resorption dynamics depended on temperature and magnitude of hypo - osmotic shock ( see Supplementary Note 1 and Supplementary Fig . 3 ) . The membrane structures were concentrated at the central part of the cell , with a less dense actin meshwork and less focal adhesions , and their formation had to displace actin \ufb01bres and disrupt adhesions ( Fig . 2h ) . This suggests that osmotically induced invaginations avoided sites of high cytoskeletal resistance like stretch - induced reservoirs , but due to their larger size also had to generate an opening through the cytoskeleton . The dynamic formation of those openings was con\ufb01rmed by time - lapse images of cells co - transfected with both membrane and actin markers ( Supplementary Fig . 4 ) . The formation of such structures ( termed VLDs ) upon increases in medium osmolarity has long been described in neurons and other cell types 9 , 16 , and has been hypothesized to constitute a mechanism to accommodate excess membrane area upon osmotic - induced cell shrinking . However , 3 - min incubation with 50 % hypo - osmotic medium only imposed a 2 % increase in membrane area ( Fig . 1d ) , which by itself had negligible effects on the membrane ( Supplementary Fig . 2 ) . Thus , other factors beyond regulation of membrane area may drive VLD formation . VLD formation is driven by water con\ufb01nement . Alternatively to being regulated by membrane area , VLDs formed after increasing osmolarity could be caused by water \ufb02ows exiting cells , which would be con\ufb01ned between cells and the substrate and thereby generate hydrostatic pressure . To test this , we seeded cells on polyacrylamide gels , through which water can \ufb02ow . In those conditions , VLDs did not form upon the change from hypo - to iso - osmotic medium ( Fig . 3a and Supplementary Movie 4 ) . This effect was due to the water - permeable properties of polyacrylamide and not by its lower stiffness , as VLDs clearly formed in softer but hydrophobic silicone elastomers ( Fig . 3a and Supplementary Movie 4 ) . Interestingly , application of stretch for 3 min and subsequent release in cells seeded on polyacrylamide gels resulted in the formation not only of reservoirs as expected , but also of VLDs ( Fig . 3b and Supplementary Movie 5 ) . This was due to the poroelastic properties of polyacrylamide gels 17 , by which gels gradually swelled when stretched for 3 min , and then gradually released water and shrank upon stretch release ( Supplementary Fig . 5 ) . Con\ufb01rming this , reservoirs but not VLDs formed on both soft silicone elastomers and polyacrylamide gels where swelling was prevented by submitting them to stretch only during a short pulse ( Fig . 3c ) . Thus , water pressure formed either through con\ufb01nement or poroelastic \ufb02ows was equivalently successful at generating VLDs . We then evaluated further the degree of water con\ufb01nement at the cell \u2013 substrate interface by submitting cells to a 3 - min 50 % hypo - osmotic shock , and then restoring iso - osmotic medium ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 2 NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications & 2015 Macmillan Publishers Limited . All rights reserved . labelled with membrane - impermeable red \ufb02uorescent dextran . We did this at a low temperature of 26 (cid:2) C , which slowed VLD formation and allowed us to distinguish the formation and resorption phases of VLD dynamics ( see Supplementary Note 1 ) . Whereas the medium surrounding cells immediately became \ufb02uorescent , the dextran - free water expelled by cells formed VLDs that had initially a very low \ufb02uorescence in the red channel ( Fig . 3d , e ) . This shows that medium in VLDs was indeed con\ufb01ned , and did not immediately mix with external medium . However , as time progressed VLDs gradually increased their \ufb02uorescence even after VLDs started decreasing in size ( Fig . 3e ) , demonstrating that water \ufb02ow and mixing was impaired but not eliminated . Thus , the water pressure driving VLD formation was not caused by a complete seal , but by a transient and dynamic con\ufb01nement generated by friction and \ufb02ow restriction at the cell \u2013 substrate interface . Consistently , cells submitted to a gradual rather than sudden osmolarity increase had suf\ufb01cient time to evacuate expelled water , preventing VLD formation ( Fig . 4b ) . Interestingly , gradual rather than abrupt de - stretch also reduced reservoir formation , leading instead to membrane accumulations at the cell edge ( Fig . 4a ) . This suggests that cells subjected to slow deformations can release membrane excess at locations where the membrane is not con\ufb01ned by a substrate , such as the cell edge . However , the increase in friction caused by fast de - stretch would prevent such long - scale rearrangements , forcing the membrane to release tension locally in reservoirs at the cell \u2013 substrate interface . Mechanism of membrane mechanical adaptation . In summary , reservoirs or VLDs were formed locally by mechanical stimuli imposing , respectively , a change in area ( through stretch ) or volume stored at the cell \u2013 substrate interface ( through osmotic shocks or poroelastic \ufb02ows , see Supplementary Note 2 ) . We then evaluated different potential mechanisms to explain how those mechanical stimuli led to the formation of membrane structures . First , reservoirs and VLDs could be mediated by caveolae \ufb02attening , reported to occur in response to both stretch and hypo - osmotic shocks 10 . However , neither reservoirs nor VLDs co - localized with caveolin during our experiments , and both types of membrane structures still formed and resorbed in caveolin 1 knockout cells ( Supplementary Fig . 6 ) . Further , reservoirs and VLDs formed and resorbed equally in caveolin 1 knockout cells reconstituted either with caveolin 1 - GFP or with an empty vector ( Supplementary Fig . 6 ) . Second , the membrane could adapt through any of the active ATP - dependent remodelling processes ( such as exo - or endocytosis ) regulating its area and shape 3 \u2013 5 . However , whereas ATP depletion inhibited reservoir and VLD resorption , it did not affect their formation ( Fig . 5a , e and Supplementary Movie 6 ) . ATP depletion also inhibited the dynamic reservoir rearrangements that occurred during their resorption by cells ( Supplementary Movie 6 ) , showing that reservoir resorption but not formation is mediated by an ATP - dependent process . Similarly , actin cytoskeleton depolymerization with cytochalasin D or a reduction in temperature slowed the resorption of both reservoirs and VLDs , but did not prevent their formation ( Supplementary Fig . 7 ) . In addition , both reservoirs and VLDs were consistently observed across different cell types from diverse species ( Supplementary Fig . 8 ) . Thus , whereas both reservoirs and VLDs resorbed through an active actin - and temperature - dependent response , they formed by a general % M e m b r ane t ea r i ng Strain ( % ) 100 80 40 6 8 10 Iso - osmotic medium Hypo - osmotic medium % Of osmolarity % M e m b r ane l ys i s 0 60 20 40 20 0 30 20 10 A r ea V o l u m e 100 80 40 0 60 20 % I n c r ea s e 60 80 6 % 10 % Strain 50 % 0 % 0 % t = 0 S t r a i n R e l a x ed t = 30 s t = 0 m i n H y po - o s m o t i c m ed i u m I s o - o s m o t i c m ed i u m t = 2 m i n Osmolarity Figure 1 | Membrane response to stretch and osmotic changes . ( a ) Cells transfected with pEYFP - mem before ( top panel ) and after ( middle and bottom panels ) applying different magnitudes of constant stretch . Yellow arrow indicates a membrane ruf\ufb02e \ufb02attened by stretch . ( b ) Cells transfected with pEYFP - mem before and after reducing medium osmolarity to either 50 or 0 % of original medium . Cells submitted to 0 % osmolarity ( de - ionized water ) for 3min sometimes rounded and \ufb02attened membrane ruf\ufb02es ( middle panel ) and sometimes underwent membrane lysis ( right panel ) . Yellow arrows indicate membrane ruf\ufb02es , which either remain or \ufb02atten after applying 50 or 0 % hypo - osmotic medium , respectively . ( c ) Confocal slice showing a cell before ( green ) and after ( red ) application of medium with 50 % osmolarity for 3min . ( d ) Corresponding quanti\ufb01cation of the increase in cell volume and required membrane area ( n \u00bc 5 cells ) . ( e ) % of cells showing membrane tearing after 3min of constant stretch application ( n \u00bc 70 cells ) . ( f ) % of cells showing membrane lysis after 3min of application of medium with different osmolarity ( n \u00bc 50 cells ) . Scale bars , 20 m m . Error bars are mean \u00b1 s . e . m . NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 ARTICLE NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications 3 & 2015 Macmillan Publishers Limited . All rights reserved . t = 3 min t = 6 min V L D f l uo r e sc en c e t = 0 min t ( min ) R e s e r v o i r f l uo r e sc en c e Iso - osmotic medium Hypo - osmotic medium t = 3 min t = 6 min t = 0 min Relaxed 6 % Strain pEYFP - mem t ( min ) 0 1 2 3 4 5 6 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 0 1 2 3 4 5 6 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 1 2 3 * * * VLDs 20 10 * * * Num per 100 \u00b5 m 2 pEYFP - mem Height ( \u00b5 m ) 3 2 1 Diameter ( \u00b5 m ) Reservoirs * * * pEYFP - mem Actin pEYFP - mem Pax Actin Paxillin pEYFP - mem Actin pEYFP - mem Pax Actin Paxillin Figure 2 | Cell membranes use different strategies to readapt to normal surface and volume . ( a ) pEYFP - mem - transfected cells before , during and after constant stretch application during 3min . ( b ) Quanti\ufb01cation of reservoir \ufb02uorescence after stretch release ( 1 : initial \ufb02uorescence , 0 : background ) . n \u00bc 100 reservoirs from 10 cells . ( c ) Confocal vertical slice from a pEYFP - mem - transfected cell before ( top ) and after ( bottom ) application of 6 % stretch for 3min . ( d ) Staining images of cells \ufb01xed immediately after stretch release showing the membrane ( pEYFP - mem transfection ) , paxillin and actin . Merged co - localization images are shown to the right . ( e ) pEYFP - mem - transfected cells before , during and after application of 50 % hypo - osmotic medium during 3min . ( f ) Quanti\ufb01cation of VLD \ufb02uorescence after re - application of iso - osmotic medium ( 1 : initial \ufb02uorescence , 0 : background ) . n \u00bc 100 VLDs from 10 cells . ( g ) Confocal images of a pEYFP - mem - transfected cell before ( top ) and after ( bottom ) application of 50 % hypo - osmotic medium for 3min . ( h ) Staining images of cells \ufb01xed immediately after re - application of iso - osmotic medium showing the membrane ( pEYFP - mem transfection ) , paxillin and actin . Merged co - localization images are shown to the right . ( i ) Quanti\ufb01cation of mean diameter of structures formed after stretch release ( reservoirs ) and re - application of iso - osmotic medium ( VLDs ) . n \u00bc 250 / 100 structures from 8 / 10 cells . ( j ) Quanti\ufb01cation of mean density of structures formed after stretch release ( reservoirs ) and re - application of iso - osmotic medium ( VLDs ) . n \u00bc 30 / 50 regions from 5 / 8 cells . ( k ) Quanti\ufb01cation of mean height of structures formed after stretch release ( reservoirs ) and re - application of iso - osmotic medium ( VLDs ) . n \u00bc 80 / 50 structures from 6 / 4 cells ( * * * P o 0 . 001 , two - tailed Student\u2019s t - test ) . We note that reservoir heights are close to the axial resolution of our confocal microscope ( 0 . 9 m m ) and thus represent upper estimates rather than accurate measurements . Scale bars are 5 m m in c , g and 20 m m in d , h . In all cases , zoomed insets ( 10 (cid:2) 6 m m 2 in c , g and 10 (cid:2) 10 m m 2 in d , h ) show a magni\ufb01cation of the area marked in the main image . Error bars are mean \u00b1 s . e . m . ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 4 NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications & 2015 Macmillan Publishers Limited . All rights reserved . passive mechanical process . Further con\ufb01rming the passive nature of reservoirs , their resorption in ATP - depleted cells could be induced by re - applying mechanical stretch ( Fig . 5a , b and Supplementary Movie 6 ) . Interestingly , whereas VLDs in ATP - depleted cells did not resorb , they gradually collapsed as water leaked from them ( Fig . 5e , f and Supplementary Movie 7 ) , leaving membrane accumulations similar to reservoirs . Those accumulations did not disappear upon re - application of hypo - osmotic medium , con\ufb01rming ( as observed in cells plated on polyacrylamide gels ) that osmotic changes per se do not directly regulate membrane invaginations . Thus , the two types of membrane invaginations could apparently convert to each other , pointing at a uni\ufb01ed framework of membrane mechanical adaptation . Given its passive nature , this framework could potentially mimic the behaviour of synthetic cell - free membrane systems . To explore this hypothesis , we adapted a theoretical approach ( see Supplementary Note 3 ) previously shown to reproduce the behaviour of passive synthetic bilayer membranes adhered to a deformable substrate 18 . In this approach , membrane invaginations are understood as structures that store excess membrane area or interstitial volume with the least energy penalty . The energy sources considered are the elastic strain energy required to stretch and bend the membrane , and the adhesion energy required to detach the membrane from the substrate . This adhesion energy includes nonspeci\ufb01c interactions ( as in the case of synthetic bilayers ) , but also speci\ufb01c bonds to the extracellular matrix mediated for instance by integrins . In the case of dorsal reservoirs substrate adhesion would not apply , but certain adhesion energy would still be required to detach the membrane from the underlying actin cortex . In this system , introducing liquid at the membrane \u2013 substrate interface results in the formation of VLD - like shallow spherical cap invaginations , which optimally store volume . In contrast , compressing the membrane results in the formation of tubular invaginations ( with much higher surface / volume ratio ) , which optimally store excess membrane area . As surface / volume requirements increase , the model provides a phase diagram with increasingly large shallow caps to store volume , increasingly long tubules to store surface , and spherical invaginations with a small connecting neck to store both ( Fig . 6a ) . If this framework applies to live cells , then the membrane reservoirs generated upon stretch release would in fact be short tubules , which should become longer for larger compressions . To test this , we seeded cells on pre - stretched membranes , and then further stretched the membrane . After 3 min , the total stretch ( between 12 and 22 % ) was released , compressing the membrane . By using this two - step approach , we prevented the membrane tearing generally observed upon high stretch ( Fig . 1 ) . As expected , releasing stretch above 12 % resulted 6 % S t r a i n pu l s e H y po - o s m o t i c m . ( 3 m i n ) 6 % S t r a i n ( 3 m i n ) R e l a x ed R e l a x ed I s o - o s m o t i c m ed i u m Dextran pEYFP - mem 0 1 2 3 4 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 Soft silicone elastomer Soft silicone elastomer Soft silicone elastomer PA gel PA gel PA gel p EY F P - m e m D e x t r an p EY F P - m e m D e x t r an p EY F P - m e m D e x t r an t ( min ) F l uo r e sc en c e t = 0 min t = 0 . 5 min t = 6 min t = 0 min t = 0 . 5 min t = 6 min 5 6 1 . 4 Figure 3 | VLD formation is driven by the con\ufb01nement of liquid \ufb02ows at the cell \u2013 substrate interface . Response of pEYFP - mem - transfected cells seeded on either poly - acrylamide ( PA ) gels or soft silicone elastomers to : ( a ) the application of 50 % hypo - osmotic medium for 3min , ( b ) the application of 6 % strain for 3min and ( c ) a fast 6 % strain pulse . Insets show zoomed views ( 10x10 m m 2 ) of membrane structures . Scale bars , 20 m m . No signi\ufb01cant differences were observed between any of the cases either in the diameter of reservoirs ( n \u00bc 150 reservoirs from 3 cells ) or in their density ( n \u00bc 30 cell regions from 3 cells ) . ( d ) Time sequence of VLD formation and resorption in pEYFP - mem - transfected cells exposed to dextran - labelled iso - osmotic media after 3min incubation with 50 % unlabelled hypo - osmotic media . ( e ) Zoomed insets ( 20 (cid:2) 20 m m 2 ) corresponding to red square in d showing the evolution of membrane and dextran \ufb02uorescence , and merged images . ( f ) Corresponding quanti\ufb01cation of pEYFP - mem and dextran relative \ufb02uorescence levels . NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 ARTICLE NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications 5 & 2015 Macmillan Publishers Limited . All rights reserved . in the formation of tubules , which became longer as stretch increased ( Fig . 6b and Supplementary Movie 8 ) . In some cases , tubules were observed to dynamically grow from reservoirs , showing that indeed reservoirs correspond to nascent membrane tubules ( Supplementary Movie 8 ) . Also as predicted by the model , VLD size , and therefore contained volume , increased with the magnitude of the hypo - osmotic shock ( Fig . 6c ) . We then explored more complex membrane deformation pathways within the phase diagram ( Fig . 7a \u2013 d ) . First , we generated VLDs in cells by decreasing and then restoring osmolarity , and then quickly re - applied hypo - osmotic medium . Cells quickly swelled and re - absorbed water in VLDs , leading to their immediate collapse ( Fig . 7e ) and con\ufb01rming that hydrostatic pressure is key to their formation and maintenance ( see Supplementary Note 2 ) . However , mere removal of hydrostatic pressure was not suf\ufb01cient to resorb membrane recruited upon VLD formation . As in ATP - depleted cells ( Fig . 5e - f ) , bright membrane accumulations ( \u2018collapsed\u2019 VLDs akin to reservoirs ) remained at VLD sites ( Fig . 7e ) . Subsequent application of stretch could then eliminate collapsed VLDs , closing the path in the phase diagram ( Fig . 7a ) . In contrast , full non - collapsed VLDs were maintained by hydrostatic pressure and only increased in diameter upon stretch ( Fig . 7f , i ) . Next , we submitted cells to both hypo - osmotic shock and stretch for 3 min and \ufb01rst restored iso - osmotic medium , leading to VLD formation as expected ( Fig . 7g ) . Upon stretch release , excess membrane did not form reservoirs or tubules but rather accumulated at the site of VLDs , as indicated by a sharp increase in pEYFP - mem \ufb02uorescence ( Fig . 7k ) . Confocal sections showed that VLDs became taller and more invaginated ( Fig . 7j ) , in agreement with the structures predicted by the model to store both volume and membrane area ( Fig . 7c ) . This latter \ufb01nding highlights that pre - existing membrane invaginations ( which are already bent and detached from the substrate ) act as seeds for further membrane storage . To con\ufb01rm this , we submitted cells to both stretch and hypo - osmotic medium and \ufb01rst released stretch , which resulted in reservoir formation ( Fig . 7h ) . When we then restored iso - osmotic medium , VLDs indeed formed at the \u2018seed\u2019 sites where reservoirs were previously located ( Fig . 7l ) . As VLD formation was now dictated by the denser reservoir network , VLDs appeared in higher number and smaller size than those formed either in the absence of stretch or before stretch release ( Fig . 7m ) . Finally , we evaluated the effect of modifying one of the key parameters of the system , adhesion energy . To this end , we treated cells with a blocking antibody against a 5 b 1 integrin , which we previously identi\ufb01ed to provide adhesion strength to \ufb01bronectin - coated substrates in the same cell type 19 . Decreasing adhesion strength reduced the density of stretch - mediated reservoirs ( Supplementary Fig . 9 ) . Similar to the case of slow stretch release ( Fig . 4 ) , this suggests that reduced friction and interaction at the membrane \u2013 substrate interface allowed cells to release excess membrane in more distant but less con\ufb01ned membrane regions . In contrast , VLDs in cells with reduced adhesion slightly increased in density , and markedly increased in diameter ( Supplementary Fig . 9 ) . This is consistent with model predictions , as a reduction in adhesion would make it energetically favourable to detach a larger membrane area to generate each VLD . Inhibition of a 5 b 1 also slowed VLD resorption , demonstrating that cells had an impaired ability to re - adhere detached membrane areas . In conclusion , the phase diagram provided by passive minimization of elastic and adhesive energies consistently predicted how the different perturbations generated membrane structures . The approximate dimensions of those structures , and their relative variations , were also correctly predicted after assuming parameter values consistent with experimental conditions ( see Supplementary Note 3 ) . Discussion Despite extensive work , the mechanisms of membrane adaptation to physical constraints have remained elusive . It was recently shown that membrane area can be stored or released upon mechanical stimulation through the assembly / disassembly of caveolae 10 . However , the estimated 0 . 3 % of membrane area contained in caveolae contrasts with membrane requirements of up to 10 % for instance in spreading cells 7 , suggesting that additional buffers are required . Such buffers can be provided in time scales from seconds to minutes by active exocytic / endocytic Iso - osmotic medium Hypo - osmotic medium Relaxed 6 % Strain 6 % Strain Hypo - osmotic medium Figure 4 | Effect of stimulus application rate on membrane structure formation . ( a ) Cell submitted to two successive steps of 6 % stretch for 3min , in which the \ufb01rst is released immediately and the second slowly ( 15s ) . ( b ) Cell submitted to two successive applications of 50 % hypo - osmotic media , in which iso - osmotic medium is restored \ufb01rst immediately and then slowly ( 1min ) . Scale bars , 20 m m . Insets show zoomed views ( 10 (cid:2) 10 m m 2 ) of membrane structures . ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 6 NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications & 2015 Macmillan Publishers Limited . All rights reserved . processes 3 \u2013 5 , which protect membrane integrity for instance in alveolar lung epithelial cells in response to both stretch 20 and osmotic changes 21 . Here we show that , before the onset of any such active process , membranes adapt almost instantaneously through a passive process minimizing membrane elastic and adhesion energies , akin to what is observed in synthetic lipid membranes 18 , 22 . This process leads to the nucleation and growth of reservoirs / tubules to accommodate membrane area fractions that can be above 10 % . The analogy between cell membranes and synthetic bilayers is surprising and has striking implications . First and addressing an unresolved issue 2 , our results suggest that if the perturbation is fast enough , membrane tension is released locally and not instantaneously transmitted across the cell ( Fig . 4 ) . The shape and size of membrane structures thus depends on both the magnitude and the dynamics of applied perturbations , which could lead for instance to the travelling membrane / cortex waves observed in cell - free surfaces 23 . Second and despite the complex molecular composition , cytoskeletal attachment and active behaviour of cell membranes 24 , we show that their mechanical adaptation can be successfully modelled by simply considering the two lipid layers . This is because the main mechanical parameter that constrains membrane deformation is the stretching modulus , which is determined by the membrane itself and not by the underlying actin cortex ( see Supplementary Note 3 ) . Finally , we note that reservoirs can store and release membrane area upon subsequent stretch cycles ( Fig . 5 ) , providing a regulatory mechanism potentially applicable to mechanical processes with time scales below the B 2 min required for active membrane resorption ( such as breathing , heart beating or muscle contraction ) . Our results also demonstrate that VLDs , which were broadly understood as membrane area containers 4 , 9 , 25 , are driven instead by hydrostatic pressure from water stored at the cell \u2013 substrate interface . Their role in processes such as cell adaptation to A T P dep l e t i on C on t r o l R e s e r v o i r f l uo r e sc en c e ATP depletion Control V L D f l uo r e sc en c e Iso - osmotic m . Hypo - osmotic medium t = 0 min t = 5 min t = 6 min t = 9 min A T P dep l e t i on C on t r o l t ( min ) t ( min ) 1 . 0 2 . 0 3 . 0 N u m be r pe r 100 \u00b5 m 2 0 . 2 0 . 6 1 . 0 D i a m e t e r ( (cid:2) m ) 5 10 15 20 t = 0 min t = 5 min t = 6 min t = 9 min Relaxed 6 % Strain 0 2 4 6 8 10 12 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 1 . 4 0 2 4 6 8 10 12 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 1 . 4 ATP depletion Control NS NS NS NS D i a m e t e r ( \u00b5 m ) N u m be r pe r 100 \u00b5 m 2 C on t r o l A T P C on t r o l A T P C on t r o l A T P C on t r o l A T P 1 . 0 2 . 0 3 . 0 Figure 5 | Membrane mechanical adaptation is a passive process followed by active recovery . ( a ) Examples of control and ATP - depleted pEYFP - mem - transfected cells before , during and after application of two 3 - min constant stretch pulses . ( b ) Quanti\ufb01cation of reservoir \ufb02uorescence after release of \ufb01rst stretch pulse and during application and release of the second pulse ( n \u00bc 50 / 70 reservoirs from 5 / 5 cells ) . The effect of ATP depletion was signi\ufb01cant ( P o 0 . 001 ) . ( c ) Quanti\ufb01cation of reservoir size in control and ATP - depleted cells ( n \u00bc 150 / 200 reservoirs from 3 / 4 cells ) . No signi\ufb01cant differences were observed , two - tailed Student\u2019s t - test . ( d ) Quanti\ufb01cation of reservoir density in control and ATP - depleted cells ( n \u00bc 50 / 50 regions from 5 / 5 cells ) . No signi\ufb01cant differences were observed . ( e ) Examples of control and ATP - depleted pEYFP - mem - transfected cells before , during and after application of 50 % hypo - osmotic medium in two 3 - min pulses . ( f ) Quanti\ufb01cation of VLD \ufb02uorescence after the \ufb01rst re - application of iso - osmotic medium and during application and release of the second pulse ( n \u00bc 35 / 40 VLDs from 3 / 3 cells ) . The effect of ATP depletion was signi\ufb01cant ( P o 0 . 001 , two - tailed Student\u2019s t - test ) . ( g ) Quanti\ufb01cation of VLD size in control and ATP - depleted cells ( n \u00bc 30 / 35 VLDs from 3 / 3 cells ) . No signi\ufb01cant differences were observed . ( h ) Quanti\ufb01cation of VLD density in control and ATP - depleted cells ( n \u00bc 20 / 25 regions from 3 / 3 cells ) . No signi\ufb01cant differences were observed . Scale bars , 20 m m . Insets show zoomed views ( 10 (cid:2) 10 m m 2 ) of membrane structures . NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 ARTICLE NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications 7 & 2015 Macmillan Publishers Limited . All rights reserved . shrinking should therefore be revisited , as cells in most physiological settings will be surrounded by permeable extracellular matrices , which would not constrain water \ufb02ow from cells . However , we show that VLDs can be equivalently formed by hydrostatic pressure arising from other sources ( such as poroelastic \ufb02ows , Fig . 3b ) , suggesting a signi\ufb01cant role in the cellular adaptation to any excess water pressure at the cell \u2013 matrix interface . The fact that increasing osmolarity leads to the immediate formation of VLDs rather than merely expelling water through the dorsal surface also con\ufb01rms that the cytoplasm exhibits limited water mobility and slow pressure redistribution 26 . Both reservoirs / tubules and VLDs form at sites of low cytoskeletal density and resorb through an actin - and ATP - dependent process . This active cell response likely involves actin polymerization to push the lamellipodium and re - stretch and \ufb02atten the cell membrane , endocytic processes to detach invaginated reservoirs or tubules , or vesiculation , that is , the detachment of membrane vesicles from tubules that has been observed both in live cells 27 and in passive bilayer systems 18 . However and regardless of the speci\ufb01c active mechanisms by which cells re - absorb membrane structures , the physical principles that drive their formation may themselves be harnessed by cells to respond to changes in cell shape arising in any instance of cell migration or deformation . Further , the biochemical activity of membrane curvature - sensitive molecules 2 , 28 , 29 could also be affected by the local curvature induced by tubules or VLDs , potentially initiating mechano - transduction cascades . Methods Cell culture and reagents . MEFs were previously described 19 , 30 and cultured in DMEM supplemented with 10 % fetal bovine serum ( FBS ) . One day before experiments , cells were transfected with the membrane - targeting plasmid pEYFP - mem ( Clontech ) or lifeact - ruby using the Neon transfection device according to the manufacturer\u2019s instructions ( Invitrogen ) . pEYFP - mem contains the N - terminal 20 amino acids of neuromodulin , which is palmytoylated post - translationally and targets the EYFP \ufb02uorophore to membranes . Cells were incubated with cytochalasin D for 30min to depolymerize the actin cytoskeleton ( 0 . 5 m M , Sigma ) and with 10mM deoxy - D - glucose plus 10mM NaN 3 ( Sigma ) to deplete ATP levels . Dextran experiments were carried out with 0 . 5mgml (cid:3) 1 of tetramethylrhodamine - labelled dextran ( 10 , 000 MW , Life technologies ) , and a 5 b 1 integrins were blocked with 10 m gml (cid:3) 1 a 5 b 1 antibody ( Merck Millipore , clone BMB5 ) . The role of caveolin 1 was analysed by using Caveolin1 Knock out MEFs reconstituted with Cav1 or IRES - GFP as a control 31 . Cav1 was cloned in the lentiviral vector pRR SIN 18 CMV IRES EGFP . To generate these stable cell lines , the infected cells were selected by sorting for GFP marker expression by \ufb02ow cytometry ( FACS ) . Cav1 expression was checked by western blot analysis and reverse transcription \u2013 quantitative PCR . Cells expressing levels of Cav1 - GFP comparable to endogenous Cav1 levels in wild - type MEFs were selected for experiments . Chinese Hamster Ovary cells were cultured in HF - 12 medium supplemented with 10 % FBS . Human keratinocytes ( HaCaT ) were cultured in DMEM supplemented with 10 % FBS . Human squamous carcinoma cells ( A431 ) were cultured in Earle\u2019s Balanced Salt Solution supplemented with 10 % FBS . Preparation of stretchable membranes . The stretchable PDMS membranes ( see schematic in Supplementary Fig . 1 ) were prepared by mixing the PDMS base and crosslinker at a 10 : 1 ratio , degassing for 1h , spinning the mixture on a 13 - cm sheet ( 500r . p . m . , 1min ) and curing at 65 (cid:2) C overnight . Once cured , PDMS membranes were peeled off and placed tightly between the rings of the stretching device ( Supplementary Fig . 1 ) . Membranes were then coated with 10 m gml (cid:3) 1 \ufb01bronectin ( Sigma ) overnight at 4 (cid:2) C , or attached to either polyacrylamide or soft silicone elastomers . To attach polyacrylamide gels to membranes 32 , gels were prepared by using a mixture of 10 % acrylamide and 0 . 3 % bis - acrylamide and polymerizing between two coverslips treated with repel - silane ( Young\u2019s modulus B 30kPa ) . Once polymerized , one coverslip was removed and the gel was pressed in contact with the PDMS membrane , which had previously been treated with 3 - aminopropyl triethoxysilane 10 % in ethanol for 1h at 65 (cid:2) C and with glutaraldehyde ( 1 , 5 % ) in PBS for 25m at room temperature . After overnight incubation at 37 (cid:2) C in a humid chamber for covalent binding , the other coverslip was removed and the gel was incubated with 10 m gml (cid:3) 1 \ufb01bronectin overnight at 4 (cid:2) C , resulting in membrane - attached gels ready for cell culture . Soft silicon elastomers ( CY 52 \u2013 276 , Dow Corning , with Young\u2019s modulus B 8kPa ( ref . 33 ) ) were prepared by mixing CyA and CyB components at a 1 : 1 ratio and curing at 80 (cid:2) C for 2h ( ref . 33 ) . The substrates were then attached to membranes following the same procedure as for polyacrylamide gels . In some experiments not involving stretch , PDMS membranes were cured directly on glass coverslips instead of placing them in the stretch system . Stretch and osmolarity experiments . Once PDMS membranes were either directly coated with \ufb01bronectin or covalently attached to \ufb01bronectin - coated polyacrylamide / soft silicone gels , cells were seeded on the membrane and allowed to spread in the incubator for 0 . 5h . Then , membranes were placed on the stretch system ( Supplementary Fig . 1 ) , consisting of a central loading post and an external ring . Vacuum was then applied through the space between the loading post and the external ring , thereby deforming and stretching the membrane . Cells spread on the membrane directly on top of the central loading post experienced an equibiaxial strain which depended on the vacuum pressure applied ( Supplementary Fig . 1 ) . The system was then mounted on the microscope stage . To modify osmolarity , cells were exposed to medium mixed with de - ionized water in which the concentrations of Ca 2 \u00fe and Mg 2 \u00fe had been corrected . 0 5 3 . 5 1 . 5 Surface storage V o l u m e s t o r age % D e s t r e t c h 6 12 t ( min ) 22 % Of destretch % Of osmolarity V L D d i a m e t e r ( \u00b5 m ) Leng t h ( \u00b5 m ) 7 6 5 4 3 2 1 5 10 15 20 25 1 2 3 4 80 40 20 60 0 Release Strain Figure 6 | Membrane mechanical adaptation is explained by minimization of the strain and adhesion energies required to generate surface and volume containers . ( a ) Phase diagram showing the predicted structures that require minimal energy to deform the membrane and detach it from the substrate in order to accommodate membrane surface area ( upon stretch release ) and liquid volume at the cell \u2013 substrate interface ( upon an increase in osmolarity ) 18 . Surface storage is achieved optimally with increasingly long tubules , whereas volume storage leads to the formation of spherical caps ( VLDs ) . When both volume and surface storage are required , spherical caps \u2018bud\u2019 and become more invaginated . ( b ) Left : time - course sequences of cell membrane regions showing the formation of either reservoirs or increasingly long tubules after releasing different stretch magnitudes . Right : Mean reservoir / tubule length ( black dots , experimental data , red line , theoretical prediction ) as a function of de - stretch magnitude ( for increasing stretch , n \u00bc 80 / 50 / 50 structures from 6 / 3 / 3 cells ) . ( c ) Left : images showing the formation of increasingly large VLDs after restoring iso - osmotic medium in cells previously exposed to different magnitudes of hypo - osmotic shocks for 3min . Right : mean VLD diameter ( black dots , experimental data , red line , theoretical prediction ) as a function of hypo - osmotic shock magnitude ( for increasing osmotic shock , 60 / 100 / 100 / 50 structures from 5 / 5 / 10 / 3 cells ) . Scale bars , 5 m m . Error bars are mean \u00b1 s . e . m . ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 8 NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications & 2015 Macmillan Publishers Limited . All rights reserved . Imaging . Cell images of \ufb02uorescently labelled cells were obtained using an upright microscope ( Nikon eclipse Ni - U ) with a water immersion objective ( (cid:2) 60 magni\ufb01cation , NA \u00bc 1 . 0 ) and an Orca Flash 4 . 0 camera ( Hamamatsu ) . To obtain three - dimensional stacks , an inverted microscope ( Nikon Eclipse Ti ) with a spinning disk confocal unit ( CSU - W1 , Yokogawa ) , a Zyla sCMOS camera ( Andor ) and a (cid:2) 60 objective was used . This objective was either of oil immersion osm . shock osm . shock + destretch t ( min ) 0 1 2 3 4 5 6 0 . 5 1 . 0 1 . 5 2 . 0 VLD diam . ( \u00b5 m ) 1 2 3 * * * * * * 2 4 6 N . VLD per 100 \u00b5 m 2 o s m . s ho ck o s m . s ho ck + D e s t r e t c h D e s t r e t c h + o s m . s ho ck * * * * * * F l uo r e sc en c e i n t en s i t y H y po - o s m o t i c m ed i u m 6 % S t r a i n R e l a x ed I s o - o s m o t i c m ed i u m H y po - o s m o t i c m ed i u m 6 % S t r a i n R e l a x ed I s o - o s m o t i c m ed i u m I m age # 3 I m age # 4 R e l a x ed I s o - o s m o t i c m ed i u m H y po - o s m o t i c m ed i u m R e l a x ed I s o - o s m o t i c m ed i u m H y po - o s m o t i c m ed i u m I m age # 4 T i m e 6 % S t r a i n 6 % S t r a i n I m age # 3 H y po t o i s o I s o t o h y po Stretch H y po t o i s o Stretch H y po t o i s o Destretch Destretch H y po t o i s o # 1 # 2 # 4 # 3 # 5 # 1 # 2 # 4 # 3 # 1 # 2 # 4 # 3 # 1 # 2 # 4 # 3 # 3 # 4 # 5 # 3 # 4 # 3 # 4 # 3 # 4 # 2 # 2 # 2 # 2 l Figure 7 | Membrane remodelling can be described through pathways along the surface / volume phase diagram . ( a \u2013 d ) Different pathways tested in the phase diagram by applying stretch and hypo - osmotic shocks ( red arrows , numbers refer to the corresponding image in the panel below ) . ( e \u2013 h ) Corresponding response of pEYFP - mem - transfected cells after applying stretch and osmotic shocks as indicated to follow the pathways . Time \ufb02ows from top to bottom . In all cases , the \ufb01rst application of stretch / hypo - osmotic shock ( second row of cells ) lasted 3min . Subsequent steps were carried out as quickly as possible to evaluate membrane response before cells had time to actively eliminate structures . ( i ) In cells submitted to hypo - osmotic shock , co - localization of membrane structures formed after \ufb01rst restoring iso - osmotic medium ( red ) and then applying stretch ( green ) . ( j ) Confocal vertical slices showing VLD shape before ( top ) and after ( bottom ) stretch release . Zoomed image to the right shows the superimposed shape prediction from the theoretical model in red . ( k ) Quanti\ufb01cation of VLD \ufb02uorescence for cells under hypo - osmotic medium after either restoring iso - osmotic medium ( blue symbols ) or restoring iso - osmotic medium and then releasing stretch application ( pink symbols , arrow indicates moment of stretch release ) . N \u00bc 100 / 50 structures from 10 / 5 cells . ( l ) In cells submitted to both hypo - osmotic shock and stretch , co - localization of membrane structures formed after \ufb01rst releasing stretch ( red ) and then restoring iso - osmotic medium ( green ) . ( m ) Quanti\ufb01cation of VLD diameter ( n \u00bc 100 / 50 / 70 structures from 10 / 3 / 3 cells , * * * P o 0 . 001 , analysis of variance ( ANOVA ) ) and density ( n \u00bc 50 / 30 / 30 regions from 8 / 3 / 3 cells , * * * P o 0 . 001 , ANOVA ) in cells submitted only to osmotic shocks or also to de - stretch ( stretch release ) before or after restoring iso - osmolarity ) . Scale bars are 5 m m in j and 20 m m elsewhere . Insets show zoomed views ( 10 (cid:2) 10 m m 2 ) of membrane structures . Error bars are mean \u00b1 s . e . m . NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 ARTICLE NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications 9 & 2015 Macmillan Publishers Limited . All rights reserved . ( NA \u00bc 1 . 42 ) or of water immersion ( NA \u00bc 1 . 0 ) for experiments involving stretch , as viscous oil droplets dragged the \ufb02exible PDMS membrane used for stretch and precluded proper focusing . Analysis of membrane structures . The evolution of membrane structures ( reservoirs and VLDs ) was analysed by measuring the time course of the pEYFP - mem \ufb02uorescence of each structure . To correct for photobleaching , this \ufb02uorescence was expressed as the fold - increase with respect to the background \ufb02uorescence of neighbouring cell regions without membrane structures , and normalized between 1 ( initial \ufb02uorescence after stretch release or osmolarity increase ) and 0 ( background cell \ufb02uorescence ) . The diameter and density of structures was also measured , and the height was obtained from confocal slices . We note , however , that reservoir heights are close to the axial resolution of our confocal microscope ( 0 . 9 m m ) and thus represent upper estimates rather than accurate measurements . The membrane fraction contained in membrane structures was estimated by comparing the average pEYFP - mem \ufb02uorescence of cell regions containing structures to the average \ufb02uorescence of structure - free zones within the same region . To ensure that we only considered the \ufb02uorescence of structures induced by stretch or osmotic shocks , the analysis was carried out in regions devoid of visible endomembrane structures before the application of stretch or osmotic shocks . In all time - lapse \ufb02uorescence images shown in \ufb01gures , contrast was adjusted in each image to correct for the effect of photobleaching and leave cell background at a uniform level . Supplementary Videos show the full time - lapse videos without this adjustment , thereby showing the effect of photobleaching . The videos also show a decrease in \ufb02uorescence upon application / release of stretch , caused by added photobleaching during the re - centering and re - focusing of cells after their movement . Cell volume and surface estimations . Changes in cell volume and required membrane surface were calculated from spinning disk confocal slices obtained in cells before and after a 3 - min 50 % hypo - osmotic treatment . Cells were \ufb01rst re - sliced in the XZ plane ( resulting in the images observed for instance in Fig . 2g ) , and membrane \ufb02uorescence images were binarized by thresholding . Cell volume was then calculated by adding the total number of pixels inside cells from all slices and multiplying by voxel size . To measure membrane surface , the cell perimeter in XZ - plane confocal slices was drawn manually . This avoided spurious increases in the estimated cell perimeter caused by the jagged cell edge resulting from binarization . We note that our area estimates do not correspond to the total membrane area , which may contain small folds not resolved by microscope images . Rather , our measurements estimate the increase in membrane area required to accommodate the global change in shape produced by cell swelling . Immunostaining . For \ufb02uorescence staining , cells were \ufb01xed with 4 % paraformaldehyde , permeabilized with 0 . 1 % Triton X - 100 and labelled \ufb01rst with primary antibodies ( 2h , room temperature ) , and then with Alexa - conjugated secondary antibodies ( Invitrogen , 2h , room temperature ) . Primary antibodies used were against paxillin ( 2 m gml (cid:3) 1 , clone 349 produced in mouse , ref . 610051 from BD Transduction Laboratories ) and Caveolin 1 ( 4 m gml (cid:3) 1 , CAV1 , polyclonal antibody produced in rabbit , ref . 610060 from BD Transduction Laboratories ) . Phalloidin - Tetramethylrhodamine B isothiocyanate ( Sigma ) was used instead of primary antibodies to label actin . Modelling . Theoretical modelling was carried out using a previously described approach 18 . Model assumptions , parameters and predictions are described in Supplementary Note 3 . Stastistical analysis . Statistical comparisons were carried out with two - tailed Student\u2019s t - tests when two cases were compared and with analysis of variance tests when more cases were analysed . All data are shown as mean \u00b1 s . e . m . References 1 . Hochmuth , R . M . , Mohandas , N . & Blackshear , Jr . P . L . Measurement of the elastic modulus for red cell membrane using a \ufb02uid mechanical technique . Biophys . J . 13 , 747 \u2013 762 ( 1973 ) . 2 . Diz - Munoz , A . , Fletcher , D . A . & Weiner , O . D . Use the force : membrane tension as an organizer of cell shape and motility . Trends Cell Biol . 23 , 47 \u2013 53 ( 2013 ) . 3 . Boulant , S . , Kural , C . , Zeeh , J . C . , Ubelmann , F . & Kirchhausen , T . Actin dynamics counteract membrane tension during clathrin - mediated endocytosis . Nat . Cell Biol . 13 , 1124 \u2013 1131 ( 2011 ) . 4 . Gauthier , N . C . , Masters , T . A . & Sheetz , M . P . Mechanical feedback between membrane tension and dynamics . Trends Cell Biol . 22 , 527 \u2013 535 ( 2012 ) . 5 . Gauthier , N . C . , Rossier , O . M . , Mathur , A . , Hone , J . C . & Sheetz , M . P . Plasma Membrane Area Increases with Spread Area by Exocytosis of a GPI - anchored Protein Compartment . Mol . Biol . Cell 20 , 3261 \u2013 3272 ( 2009 ) . 6 . Raucher , D . & Sheetz , M . P . Characteristics of a membrane reservoir buffering membrane tension . Biophys . J . 77 , 1992 \u2013 2002 ( 1999 ) . 7 . Gauthier , N . C . , Fardin , M . A . , Roca - Cusachs , P . & Sheetz , M . P . Temporary increase in plasma membrane tension coordinates the activation of exocytosis and contraction during cell spreading . Proc . Natl Acad . Sci . USA 108 , 14467 \u2013 14472 ( 2011 ) . 8 . Paluch , E . K . & Raz , E . The role and regulation of blebs in cell migration . Curr . Opin . Cell Biol . 25 , 582 \u2013 590 ( 2013 ) . 9 . Morris , C . E . & Homann , U . Cell surface area regulation and membrane tension . J . Membr . Biol . 179 , 79 \u2013 102 ( 2001 ) . 10 . Sinha , B . et al . Cells respond to mechanical stress by rapid disassembly of caveolae . Cell 144 , 402 \u2013 413 ( 2011 ) . 11 . Kozlov , M . M . et al . Mechanisms shaping cell membranes . Curr . Opin . Cell Biol . 29 , 53 \u2013 60 ( 2014 ) . 12 . Mogilner , A . & Keren , K . The shape of motile cells . Curr . Biol . 19 , R762 \u2013 R771 ( 2009 ) . 13 . Kabaso , D . , Shlomovitz , R . , Schloen , K . , Stradal , T . & Gov , N . S . Theoretical model for cellular shapes driven by protrusive and adhesive forces . PLoS Comput . Biol . 7 , e1001127 ( 2011 ) . 14 . Shaklee , P . M . et al . Bidirectional membrane tube dynamics driven by nonprocessive motors . Proc . Natl Acad . Sci . USA 105 , 7993 \u2013 7997 ( 2008 ) . 15 . Roux , A . The physics of membrane tubes : soft templates for studying cellular membranes . Soft Matter 9 , 6726 \u2013 6736 ( 2013 ) . 16 . Reuzeau , C . , Mills , L . R . , Harris , J . A . & Morris , C . E . Discrete and reversible vacuole - like dilations induced by osmomechanical perturbation of neurons . J . Membr . Biol . 145 , 33 \u2013 47 ( 1995 ) . 17 . Casares , L . et al . Hydraulic fracture during epithelial stretching . Nat . Mater . 14 , 343 \u2013 351 ( 2015 ) . 18 . Staykova , M . , Arroyo , M . , Rahimi , M . & Stone , H . A . Con\ufb01ned bilayers passively regulate shape and stress . Phys . Rev . Lett . 110 , 028101 ( 2013 ) . 19 . Roca - Cusachs , P . , Gauthier , N . C . , del Rio , A . & Sheetz , M . P . Clustering of a 5 b 1 integrins determines adhesion strength whereas a v b 3 and talin enable mechanotransduction . Proc . Natl Acad . Sci . USA 106 , 16245 \u2013 16250 ( 2009 ) . 20 . Vlahakis , N . E . , Schroeder , M . A . , Pagano , R . E . & Hubmayr , R . D . Deformation - induced lipid traf\ufb01cking in alveolar epithelial cells . Am . J . Physiol . Lung Cell Mol . Physiol . 280 , L938 \u2013 L946 ( 2001 ) . 21 . Wang , S . , Singh , R . D . , Godin , L . , Pagano , R . E . & Hubmayr , R . D . Endocytic response of type I alveolar epithelial cells to hypertonic stress . Am . J . Physiol . Lung Cell Mol . Physiol . 300 , L560 \u2013 L568 ( 2011 ) . 22 . Staykova , M . , Holmes , D . P . , Read , C . & Stone , H . A . Mechanics of surface area regulation in cells examined with con\ufb01ned lipid membranes . Proc Natl Acad . Sci . USA 108 , 9084 \u2013 9088 ( 2011 ) . 23 . Kapustina , M . , Elston , T . C . & Jacobson , K . Compression and dilation of the membrane - cortex layer generates rapid changes in cell shape . J . Cell Biol . 200 , 95 \u2013 108 ( 2013 ) . 24 . Rao , M . & Mayor , S . Active organization of membrane constituents in living cells . Curr . Opin . Cell Biol . 29 , 126 \u2013 132 ( 2014 ) . 25 . Dai , J . , Sheetz , M . P . , Wan , X . & Morris , C . E . Membrane tension in swelling and shrinking molluscan neurons . J . Neurosci . 18 , 6681 \u2013 6692 ( 1998 ) . 26 . Moeendarbary , E . et al . The cytoplasm of living cells behaves as a poroelastic material . Nat . Mater . 12 , 253 \u2013 261 ( 2013 ) . 27 . Romer , W . et al . Shiga toxin induces tubular membrane invaginations for its uptake into cells . Nature 450 , 670 \u2013 675 ( 2007 ) . 28 . Zhao , H . , Pykalainen , A . & Lappalainen , P . I - BAR domain proteins : linking actin and plasma membrane dynamics . Curr . Opin . Cell Biol . 23 , 14 \u2013 21 ( 2011 ) . 29 . Shen , H . , Pirruccello , M . & De Camilli , P . SnapShot : membrane curvature sensors and generators . Cell 150 , e1301 \u2013 1302 ( 2012 ) . 30 . Roca - Cusachs , P . et al . Integrin - dependent force transmission to the extracellular matrix by alpha - actinin triggers adhesion maturation . Proc . Natl Acad . Sci . USA 110 , E1361 \u2013 E1370 ( 2013 ) . 31 . Goetz , J . G . et al . Biomechanical remodeling of the microenvironment by stromal caveolin - 1 favors tumor invasion and metastasis . Cell 146 , 148 \u2013 163 ( 2011 ) . 32 . Elosegui - Artola , A . et al . Rigidity sensing and adaptation through regulation of integrin types . Nat . Mater . 13 , 631 \u2013 637 ( 2014 ) . 33 . Vedula , S . R . et al . Epithelial bridges maintain tissue integrity during collective cell migration . Nat . Mater . 13 , 87 \u2013 96 ( 2014 ) . Acknowledgements We acknowledge support from the Spanish Ministry for Economy and Competitiveness ( BFU2011 - 23111 , BFU2012 - 38146 , and FIS - PI11 - 00089 ) , a Career Integration Grant within the seventh European Community Framework Programme ( PCIG10 - GA - 2011 - 303848 ) , the European Research Council ( Grant Agreements 242993 , 240487 and 240487 ) , the Generalitat de Catalunya , Fundacio\u00b4 La Caixa , Fundacio\u00b4 la Marato\u00b4 de TV3 ( project 20133330 ) and the Mechanobiology Institute Singapore grant , initiative of the the National Research Foundation of Singapore ( to N . C . G ) . We thank F . Lolo and the members of the X . T . and P . R - C . laboratories for technical assistance and discussions . ARTICLE NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 10 NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications & 2015 Macmillan Publishers Limited . All rights reserved . Author contributions A . J . K . , M . A . , N . C . G . and P . R . - C . conceived the study , A . J . K . , S . M . , M . A . , X . T . , N . C . G . and P . R . - C . designed the experiments , A . J . K . and J . J . T . performed the experiments , A . E . - A , R . M . - V . and M . A . delP . contributed new reagents / analytical tools , A . J . K . , L . C . , V . G . - T . , D . N . and X . T developed the stretch device , M . A . carried out the theoretical modelling , and A . J . K . , M . A . , and P . R . - C . wrote the paper . Additional information Supplementary Information accompanies this paper at http : / / www . nature . com / naturecommunications Competing \ufb01nancial interests : The authors declare no competing \ufb01nancial interests . Reprints and permission information is available online at http : / / npg . nature . com / reprintsandpermissions / How to cite this article : Kosmalska , AJ . et al . Physical principles of membrane remodelling during cell mechanoadaptation . Nat . Commun . 6 : 7292 doi : 10 . 1038 / ncomms8292 ( 2015 ) . This work is licensed under a Creative Commons Attribution 4 . 0 International License . The images or other third party material in this article are included in the article\u2019s Creative Commons license , unless indicated otherwise in the credit line ; if the material is not included under the Creative Commons license , users will need to obtain permission from the license holder to reproduce the material . To view a copy of this license , visit http : / / creativecommons . org / licenses / by / 4 . 0 / NATURE COMMUNICATIONS | DOI : 10 . 1038 / ncomms8292 ARTICLE NATURE COMMUNICATIONS | 6 : 7292 | DOI : 10 . 1038 / ncomms8292 | www . nature . com / naturecommunications 11 & 2015 Macmillan Publishers Limited . All rights reserved .",
    "mueller2017load": "Article Load Adaptation of Lamellipodial Actin Networks Graphical Abstract Highlights d Lamellipodial actin density co - \ufb02uctuates with the size of the projected area d Lamellipodial actin density adapts to changes in membrane tension d Actin branch geometry prescribes adaptations in lamellipodial actin Authors Jan Mueller , Gregory Szep , Maria Nemethova , . . . , Kinneret Keren , Robert Hauschild , Michael Sixt Correspondence sixt @ ist . ac . at In Brief How do mechanical perturbations in\ufb02uence the density and the geometry of actin networks at the leading edge of migrating cells ? Mueller et al . , 2017 , Cell 171 , 1 \u2013 13 September 21 , 2017 \u00aa 2017 Elsevier Inc . http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 Article Load Adaptation of Lamellipodial Actin Networks Jan Mueller , 1 Gregory Szep , 1 Maria Nemethova , 1 Ingrid de Vries , 1 Arnon D . Lieber , 2 Christoph Winkler , 3 , 4 Karsten Kruse , 5 J . Victor Small , 6 Christian Schmeiser , 3 , 4 Kinneret Keren , 2 , 7 Robert Hauschild , 1 and Michael Sixt 1 , 8 , * 1 Institute of Science and Technology Austria ( IST Austria ) , am Campus 1 , 3400 Klosterneuburg , Austria 2 Department of Physics and Russell Berrie Nanotechnology Institute , Technion , Israel Institute of Technology , Haifa 32000 , Israel 3 RICAM , Austrian Academy of Sciences , Apostelgasse 23 , 1030 Vienna , Austria 4 Faculty of Mathematics , University of Vienna , Oskar - Morgenstern - Platz 1 , 1090 Vienna , Austria 5 NCCR Chemical Biology , Departments of Biochemistry and Theoretical Physics , University of Geneva , 30 , quai Ernest - Ansermet , 1211 Geneva , Switzerland 6 Institute of Molecular Biotechnology GmbH ( IMBA ) , Dr . Bohr - Gasse 3 , 1030 Vienna , Austria 7 Network Biology Research Laboratories , Technion , Israel Institute of Technology , Haifa 32000 , Israel 8 Lead Contact * Correspondence : sixt @ ist . ac . at http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 SUMMARY Actin \ufb01laments polymerizing against membranes power endocytosis , vesicular traf\ufb01c , and cell motility . In vitro reconstitution studies suggest that the structure and the dynamics of actin networks respond to mechanical forces . We demonstrate that lamellipodial actin of migrating cells responds to mechanical load when membrane tension is modulated . In a steady state , migrating cell \ufb01laments assume the canonical dendritic geometry , de\ufb01ned by Arp2 / 3 - generated 70 ! branch points . Increased tension triggers a dense network with a broadened range of angles , whereas decreased tension causes a shift to a sparse con\ufb01guration dominated by \ufb01la - ments growing perpendicularly to the plasma mem - brane . We show that these responses emerge from the geometry of branched actin : when load per \ufb01lament decreases , elongation speed increases and perpendicular \ufb01laments gradually outcompete others because they polymerize the shortest dis - tance to the membrane , where they are protected from capping . This network - intrinsic geometrical adaptation mechanism tunes protrusive force in response to mechanical load . INTRODUCTION From endocytic processes and vesicle traf\ufb01cking to cellular locomotion and intracellular pathogen movement , the actomy - osin cytoskeleton mediates most of the mechanical responses of eukaryotic cells ( Pollard and Cooper , 2009 ) . Actomyosin net - works have two principal mechanisms of force generation : ( 1 ) polymerization can expand the network by elongating \ufb01laments against a load , and ( 2 ) myosin motors can contract the network or transport cargo along \ufb01laments . Every actin \ufb01lament is born by a nucleation event . Nucleation is catalyzed by a number of molecular machines , the most prominent being formins and the Arp2 / 3 complex . While formins achieve de novo linear nucle - ation , the Arp2 / 3 complex binds to an existing \ufb01lament and branches off a new \ufb01lament at a 70 ! angle . Elongation , the addi - tion of new monomers to the growing ( barbed ) end of the \ufb01la - ment , is mainly mediated by formins or by proteins of the VASP family . Both nucleation and elongation are enhanced at the interface to membranes , where growing \ufb01laments are shielded from capping proteins , which otherwise terminate \ufb01la - ment elongation by sealing the \ufb01lament\u2019s barbed end ( Carlsson , 2010 ; Pollard , 2007 ) . Arp2 / 3 - dependent actin polymerization is a force - sensitive process . This has been demonstrated in vitro , in minimalist re - constituted systems , whereby increasing the mechanical load on a protruding reconstituted network led to higher network den - sity ( Bieling et al . , 2016 ; De\u00b4moulin et al . , 2014 ; Parekh et al . , 2005 ) . Thus , the network structure was able to adapt to the ambient mechanical conditions and to support force generation and mechanical resilience under varying loads . It has been shown that such adaptation is partially mediated by differential force sensitivities of nucleation , elongation , and capping , leading to enhanced branching under higher loads ( Bieling et al . , 2016 ) . However , these kinetic effects were only partially able to explain the changes in network density , and it was speculated that the remaining adaptation might be mediated by spatial rearrange - ments of the network . The mechanochemical response of protruding actin networks has not yet been studied in living cells . However , indirect evi - dence suggests that the leading edge of a motile cell is able to respond to load . When the lamellipodium , the \ufb02at protrusive front of a migrating cell , encounters an elastic obstacle ( e . g . , the ver - tical cantilever of an atomic force microscope ) , it was shown to substantially increase its pushing force , until collapsing when the load became excessive . For a migrating cell encountering a barrier , this means that its front pushes harder until the obstacle is pushed away or the leading edge stalls and the cell either retracts or circumnavigates the barrier ( Heinemann et al . , 2011 ; Prass et al . , 2006 ) . In \ufb01broblasts or epithelial cells , where lamellipodia typically undergo cycles of protrusion and retraction , it has been sug - gested that during each protrusive phase the polymerizing actin \ufb01laments experience a gradually growing load , because increase Cell 171 , 1 \u2013 13 , September 21 , 2017 \u00aa 2017 Elsevier Inc . 1 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 in projected cell area is accompanied by increase in membrane tension ( the cell stretches out the rather inextensible bag of plasma membrane ) ( Diz - Mun\u02dcoz et al . , 2013 ; Gauthier et al . , 2011 ; Raucher and Sheetz , 2000 ; Sens and Plastino , 2015 ) . This growing load on the pushing \ufb01laments is accompanied by accelerated incorporation of the actin nucleating Arp2 / 3 com - plex , which might , in turn , increase actin network density and boost force generation until the \ufb01laments stall and the leading edge retracts ( Ji et al . , 2008 ; Lee et al . , 2015 ; Ryan et al . , 2012 ) . While these response patterns might result from complex mechanosensitive signaling processes , theoretical consider - ations put forward the alternative idea that branched actin net - works respond to varying polymerization kinetics by changing their geometry ( Maly and Borisy , 2001 ; Schaus et al . , 2007 ; Weichsel and Schwarz , 2010 ) . Such responses would argue that load adaptation is an emergent geometrical property of branched actin . Here , we use the lamellipodium of \ufb01sh keratocytes to quantita - tively examine how protruding actin networks respond to varying forces . We use a combination of quantitative light and electron microscopy to describe geometrical changes of lamellipodial actin upon varying load regimes and employ stochastic modeling to elucidate how the network structure geometrically adapts to counter - forces . RESULTS Temporal Fluctuations in Actin Density Arise at the Leading Edge and Are Correlated with Projected Cell Area and Protrusion Speed To quantitatively explore how lamellipodial actin reacts to changes in load we employed \ufb01sh keratocytes migrating on planar surfaces ( Mogilner and Keren , 2009 ) . We established pro - tocols to derive keratocytes from adult zebra\ufb01sh , allowing ac - cess to stable transgenesis . On serum - coated coverslips , these cells migrate with exceptional speed , persistence , and morpho - logical stability ( Keren et al . , 2008 ) . We \ufb01rst chose a correlative approach and observed single migrating keratocytes expressing the actin reporter lifeact : GFP ( Riedl et al . , 2008 ) with high spatio - temporal resolution ( Figure 1A ; Movie S1 ) . This allowed us to monitor \ufb02uctuations in actin network dynamics and density together with protrusion speed and morphological parameters ( Figures 1A \u2013 1D , S1A , and S1B ) . The projected area of a kerato - cyte typically \ufb02uctuated \u00b1 5 % around a baseline , which was consistent with earlier studies ( Keren et al . , 2008 ) ( Figure S1A ) . These limited \ufb02uctuations supported the notion that the kerato - cyte\u2019s plasma membrane does not harbor major membrane res - ervoirs , which might be mobilized upon mechanical stretch ( Lieber et al . , 2013 ) . When we measured lifeact : GFP intensity within a \" 1 - m m broad zone behind the leading front , intensity \ufb02uctuations were more substantial than area changes and varied \u00b1 20 % ( Figures 1D and S1A ) . Strikingly , the normalized , temporal \ufb02uctuations in projected cell area were tightly correlated with life - act : GFP intensities as shown by cross - correlation peaks be - tween 0 . 6 \u2013 0 . 7 at time lag zero ( Figures 1D , 1E , and S1B ) . This correlation was also observed in actin : GFP - expressing kerato - cytes , but not when the plasma membrane was uniformly labeled ( Figure S2 ) . Correlation analysis of consecutive \" 1 - m m broad lamellipodial regions showed that the time delay between correlation peaks increased with the distance between the measured regions ( Fig - ures S1C \u2013 S1E ) . Dividing the respective lag times by the speed of each cell yields distances matching those between adjacent la - mellipodial regions ( Figure S1F ) . This analysis showed that , \ufb01rst , as previously demonstrated ( Theriot and Mitchison , 1991 ; Wil - son et al . , 2010 ) keratocytes exhibited minimal retrograde \ufb02ow of actin in relation to the substrate . Second , temporal \ufb02uctua - tions of actin density arise at the leading edge and propagate rearward . We next quanti\ufb01ed protrusion speed by optical \ufb02ow analysis at each pixel along the cell front . Protrusion speed was negatively correlated with both projected cell area and lifeact : GFP intensity with a negative cross - correlation peak between # 0 . 3 and # 0 . 4 at time lag zero ( Figures 1B \u2013 1E ) . Lifeact : GFP intensities were equally co - \ufb02uctuating with area changes when corrected for the \ufb02uctuations in protrusion speed ( Figure S1A ) . Taken together , our quantitative analysis demonstrates that keratocytes migrate slower and produce denser actin networks during intervals when their projected area is larger . Experimental Manipulations of Membrane Tension Reveal a Lamellipodial Response to Altered Load Among several interpretations of these correlative data we considered that a \ufb02uctuation in projected cell area might corre - spond to changes in membrane tension ( Raucher and Sheetz , 2000 ) . To dissect the causal relationship between actin network density and membrane tension , we experimentally increased membrane tension by aspirating the trailing edge of migrating keratocytes with a micropipette ( Houk et al . , 2012 ) ( Figure 2A ; Movie S2 ) . As membrane tension equilibrates almost instanta - neously over the whole cell , this manipulation allowed us to tune lateral tension at the lamellipodial tip ( Diz - Mun\u02dcoz et al . , 2013 ) . Experiments were performed by transiently applying four different vacuum levels ranging from # 10 to # 40 mbar . In all cases , we found that aspiration caused an increase in life - act : GFP signal concomitant with a moderate decrease in protru - sion speed ( Figures 2A \u2013 2C ) . Actin density and protrusion speed were dependent on the applied vacuum ( Figures 2D and 2E ) . Following transient aspiration , the band of denser actin network traveled backward ( in the cell frame of reference ) with the actin \ufb02ow ( Figures 2A and 2C ) . Under the given parameters the cells kept protruding during aspiration , demonstrating that the in - crease in tension was below the stall force of the lamellipodium . These results suggested that lamellipodial actin responds to an increase in membrane tension by increasing network density . To experimentally decrease membrane tension we took advantage of the fact that on adhesive substrates cells occa - sionally form tethered trailing edges , which leads to a stretched morphology ( Figure 2F ) . When these tethers spontaneously detach or when we cut them with a pulsed laser , tension is released and the tether snaps forward . Here , the projected cell area shrunk within a few seconds when the cell went from a stretched to a more compact con\ufb01guration . Such high recoil ve - locities indicate the rapid drop in tension , which had previously built up in the stretched cell . To directly test if shrinkage is accompanied by changes in membrane tension , membrane 2 Cell 171 , 1 \u2013 13 , September 21 , 2017 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 tethers were pulled from the plasma membrane of migrating ker - atocytes using an optical trap ( Lieber et al . , 2013 ) ( Figure S3A ) . While in steady - state migrating cells there was no consistent relation between projected area and tether force , abrupt retrac - tion events were accompanied by area shrinkage concurrent with a drop in tether force and thus membrane tension ( Fig - ure S3B ) . During abrupt retraction , the previously observed cor - relation between projected cell area and actin intensity at the leading edge manifested in a sudden drop in lamellipodial life - act : GFP intensity at the moment of area shrinkage ( Figures 2F and 2G ) . The resultant abrupt change in actin density was main - tained while it steadily traveled backward ( in the cell frame of reference ) with the actin \ufb02ow , further substantiating that altered polymerization originated at the leading edge of the cell ( Figures 2F and 2H ; Movie S2 ) . Cell shrinkage was also accompanied by a rapid but very transient increase in protrusion speed ( Figure 2G ) . These results suggested that lamellipodial actin responds to a decrease in membrane tension by decreasing network density . Figure 1 . Correlative Analysis of Actin , Projected Cell Area , and Protrusions in Migrating Zebra\ufb01sh Keratocytes ( A ) Confocal imaging of lifeact : GFP - expressing keratocyte moving on a glass coverslip . ( B ) Time frame from ( A ) showing the 1 . 09 - m m wide area of lifeact : GFP intensity measurements ( left ) , a binary mask used for quantifying the area ( middle ) , and a pseudo - colored Horn - Schunck optical \ufb02ow analysis image of the leading edge ( right ) . ( C ) Fluorescence intensity and leading edge velocity maps for the time lapse of the cell shown in ( A ) . ( D ) Temporal \ufb02uctuations of cell area , lifeact : GFP intensity , and protrusion speed averaged across the analyzed region shown in red in ( B ) . ( E ) Average of temporal cross - correlation functions of 21 migrating keratocytes . Mean and SEM are shown . See also Figures S1 and S2 and Movie S1 . Cell 171 , 1 \u2013 13 , September 21 , 2017 3 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 Figure 2 . Cytoskeletal Response to Manipulations of Membrane Tension ( A ) Lifeact : GFP imaging following micropipette aspiration . ( 1 ) Migrating keratocyte before contact with the micromanipulator , ( 2 ) aspiration of membrane , and ( 3 ) release of vacuum . ( B ) Temporal\ufb02uctuationsoflifeact : GFPintensityattheleadingedgede\ufb01nedasinFigure1B , area , andprotrusionspeed . Theareabetweenthedottedlinesshows where the cell was aspirated . ( C ) Kymograph along dashed yellow line in ( A ) . ( D ) Change in lifeact : GFP signal following aspiration with four different vacuum levels . ( E ) Change in cell edge protrusion for the same cells as in ( D ) . Mean and SEM are shown for 28 aspiration events in 13 individual cells in ( D ) and ( E ) . ( legend continued on next page ) 4 Cell 171 , 1 \u2013 13 , September 21 , 2017 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 As our mechanical manipulations of membrane tension might have caused additional changes in the cells , we chose an inde - pendent approach to challenge our hypothesis . Migrating cells were exposed to cycles of altered osmolarity ( Diz - Mun\u02dcoz et al . , 2013 ) , by supplementing the medium with sucrose versus pure water . Monitoring lifeact : GFP at the leading edge showed that increased osmolarity ( water ef\ufb02ux \u2013 cell shrinkage ) was accom - panied by a decrease in lifeact : GFP signal and this was reversed when water was added ( water in\ufb02ux \u2013 cell swelling ) ( Figures S3C and S3D ) . Together , our data show that lamellipodial actin responds to changes in membrane tension by altering its density : density in - creases following an increase in tension and decreases following a decrease in tension . Quantitative Analysis of Lamellipodial Ultrastructure during Steady - State Migration To gain quantitative insight into lamellipodial architecture at the single \ufb01lament level we employed three - dimensional ( 3D ) elec - tron microscopy of \ufb01xed and negatively stained keratocytes ( Vinzenz et al . , 2012 ) . This approach allows tracing single \ufb01la - ments in 3D space ( Winkler et al . , 2012 ) . In keratocyte lamellipo - dia , \ufb01laments are consistently oriented with their barbed ends toward the front ( Narita et al . , 2012 ; Sivaramakrishnan and Spu - dich , 2009 ; Svitkina et al . , 1997 ) , which is in line with the polar forward growth of actin at the leading cell front ( Figures S1C \u2013 S1F ) ( Keren , 2011 ; Lai et al . , 2008 ; Theriot and Mitchison , 1991 ) . In the digitalized tomograms we de\ufb01ned a barbed end as the end of a \ufb01lament proximal to the leading edge and a pointed end as the one distal to the leading edge . Any barbed end could either represent an actively growing or a capped \ufb01la - ment . A pointed end represents ( 1 ) a nucleation event generated by Arp2 / 3 ( a branch ) , ( 2 ) a nucleation event generated by an alternative nucleation factor , or ( 3 ) a debranched or severed \ufb01la - ment ( Ydenberg et al . , 2013 , 2011 ) . Throughout the lamellipo - dium of steady - state migrating keratocytes we found that the numbers of barbed and pointed ends were balanced ( Figures 3A \u2013 3D and S4A ) . Only at the leading front , barbed ends occurred in excess , which is expected due to ongoing polymer - ization at this site . Next , we analyzed the angles , which the \ufb01la - ments assumed relative to the leading membrane and found that the distribution of their orientations peaked at \u00b1 35 ! ( with 0 ! de\ufb01ned as perpendicular to the leading membrane ) ( Fig - ure S4B ) . This con\ufb01guration predominated in keratocytes as shown by a global order parameter de\ufb01ned as ( ( Filaments 0 ! \u2013 20 ! ) # ( Filaments 30 ! \u2013 50 ! ) ) / ( ( Filaments 0 ! \u2013 20 ! ) + ( Filaments 30 ! \u2013 50 ! ) ) as a readout for network orientation ( Weichsel and Schwarz , 2010 ) ( Figures 3E and 3F ) . Together , our tomography data provide a quantitative descrip - tion of lamellipodial actin as derived from \ufb01lament con\ufb01guration . We conclude that in a steady - state migrating keratocyte , nucle - ation , capping and elongation create an expanding network with a canonical branch geometry prescribed by the 70 ! branch structure of the Arp2 / 3 complex ( Svitkina et al . , 1997 ) . Network Con\ufb01guration following Load Increase We next used trailing edge aspiration to interrogate the ultra - structural changes in lamellipodial networks polymerizing against increased load . To this end we employed correlated \ufb02uorescence live - cell imaging and electron tomography on lifeact : GFP - expressing keratocytes migrating on electron micro - scopy grids . After 4 s of # 30mbar aspiration , cells were released , \ufb01xed and prepared for electron microscopy ( Figure 4A ; Movie S3 ) . Under the given conditions where density \ufb02uctuations arise from the leading edge , the temporal growth - history of lamellipo - dial actin is spatially encoded in the network : the pre - aspiration zone ( proximal to the nucleus ) represents steady - state growth , the aspiration zone the increase in load , and the post - aspiration zone at the leading edge the phase at which the load drops again due to the release of the vacuum . In the low - magni\ufb01cation electron micrograph , the transient in - crease in density during aspiration was clearly visible as a darker zone running parallel to the leading edge ( Figure 4A ) . This was consistent with the lifeact : GFP live cell imaging ( Figures 2A \u2013 2C ) . Electron tomography with automated tracking of actin \ufb01la - ments allowed us to trace the temporal evolution of the load response at the single \ufb01lament level . The pre - aspiration zone resembled a steady - state lamellipodium with a canonical 70 ! branched network . Within an 800nm broad aspiration zone ( cor - responding to a 4 - s aspiration time ) \ufb01lament density was mark - edly increased . At the beginning of the aspiration zone ( distal from the leading front , hence earlier in time ) a transient drop in the number of barbed ends and a gradual increase in pointed ends created a mismatch between \ufb01lament birth and death , along with the increase in \ufb01lament density ( Figures 4C and 4D ) . This increase in \ufb01lament density was also accompanied by an in - crease in the 0 ! \u2013 20 ! and 50 ! \u2013 70 ! fraction of \ufb01lament angles , whereas \ufb01laments at intermediate angles of 20 ! \u2013 50 ! increased only moderately ( Figures 4E and S4C ) . At the leading front , \ufb01la - ment density decreased to reach values comparable to levels before aspiration ( Figure 4D ) and \ufb01lament angles reverted to the canonical branch - pattern ( Figures 4E , S4C , and S4D ) . These data show that an increase in membrane tension causes an in - crease in network density and a change in geometry , with \ufb01la - ments growing at steeper angles toward the plasma membrane at higher membrane tensions . Network Con\ufb01guration following Load Decrease We followed the same correlative visualization strategy to analyze the network following a decrease in load . Similar to the laser cutting used in Figures 2F \u2013 2H rapid cell shrinkage could also be triggered by mechanically detaching one side of a cell with a micropipette ( Figure 5A ; Movie S4 ) . This maneuver al - lowed suf\ufb01cient experimental control to \ufb01x the cell seconds after ( F ) Lifeact : GFP signal following fast cell shrinkage . ( 1 ) Keratocyte exhibiting stretched morphology , while its trailing edge is tethered to the substrate . ( 2 ) The trailing edge is cut with a pulsed laser ( yellow arrowhead ) . ( G ) Temporal \ufb02uctuations of leading edge lifeact : GFP intensity , area , and protrusion speed . The rapid retraction is marked with a dotted line . ( H ) Kymograph along dashed yellow line in ( F ) . See also Figure S3 and Movie S2 . Cell 171 , 1 \u2013 13 , September 21 , 2017 5 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 ( legend on next page ) 6 Cell 171 , 1 \u2013 13 , September 21 , 2017 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 shrinkage . Analogous to the increase in actin density following aspiration , the decrease in actin density in the time window of rapid shrinkage was clearly visible as a transition zone in life - act : GFP signal as well as phalloidin staining ( Figure 5A ) . Both sig - nals showed a sharp decrease indicating the drop in \ufb01lament density . This was substantiated by low - magni\ufb01cation views of the same cell after extraction and negative stain , which showed a sharp decrease in electron density , overlapping with the \ufb02uo - rescent imaging ( Figures 5A and S6C ) . Quanti\ufb01cation of manu - ally as well as automatically traced \ufb01laments together with auto - mated image segmentation showed a drop in \ufb01lament numbers in the transition zone ( Figures 5C , 5D , S5 , and S6 ) . As expected , Figure 3 . Electron Tomography of Migrating Wild - Type Keratocytes ( A ) Overview electron micrograph of migrating keratocyte with acquired tomogram montage marked in red . ( B ) 5 . 5 - nm slice of a negatively stained tomogram of the actin network behind the leading edge . ( C ) Automated tracking results of the same region with \ufb01laments shown in green , barbed ends in red , and pointed ends in blue . ( D ) Normalized densities of \ufb01laments , barbed ends , and pointed ends in106 - nm - wide bins of four averaged tomogram montages . Graph shows mean and SEM . ( E ) Scheme showing the \ufb01lament angle bins used for calculating the global order parameter . ( F ) Histogram of combined \ufb01lament length growing at indicated angle toward the cell membrane ( black ) is shown together with a global order parameter ( blue ) in 212 - nm distance bins de\ufb01ned as ( ( Filaments 0 ! \u2013 20 ! ) # ( Filaments 30 ! \u2013 50 ! ) ) / ( ( Filaments 0 ! \u2013 20 ! ) + ( Filaments 30 ! \u2013 50 ! ) ) . See also Figures S4A and S4B . Figure 4 . Correlated Live Microscopy - Electron Tomography of Ultrastructural Changes in Networks with Increasing Filament Density ( A ) Amigrating keratocyteexpressinglifeact : GFPwas aspirated attherearwithamicropipette , \ufb01xed within z 3s , and prepared forelectron microscopy . Thecell shifted out of focus because of the manipulating procedure . A transient increase in actin density can be seen in the low - magni\ufb01cation electron micrograph on the right . ( B ) 5 . 5 - nm tomogram slice of the region marked by a red box in ( A ) . The rear of the cell is toward the left side of the picture , and the cell front is seen on the right . Regions of steady - state density , increased density , and decreased density are marked with black , blue , and red throughout the whole \ufb01gure . ( C ) Filament tracks of the lamellipodium shown in ( B ) , with actin \ufb01laments shown in green , barbed ends in red , and pointed ends in blue . ( D ) Filament numbers and densities of barbed and pointed ends in 106 - nm - wide spatial bins throughout the lamellipodium shown in ( B ) and ( C ) . ( E ) Histogram showing \ufb01lament angles to the cell edge in 212 - nm distance bins . See also Figures S4C and S4D and Movie S3 . Cell 171 , 1 \u2013 13 , September 21 , 2017 7 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 the pre - transition zone ( prior to mechanical manipulation ) showed the quantitative signature of the steady - state lamellipo - dium . The onset of the transition zone was marked by a transient increase of barbed ends , which preceded a drop in pointed ends ( Figure 5D ) . Here , the resultant gap between putative capping and nucleation events explained the concomitant drop in \ufb01la - ment density . The transition zone was followed by a leveling of barbed and pointed ends and a continuous recovery of the network toward the cell front . Notably , the transition zone was characterized by radically changed network geometry . When we quanti\ufb01ed and parame - terized \ufb01lament angles depending on their position in the network we found that at the onset of cell shrinkage the canon - ical \u00b1 35 ! dominated network abruptly changed into a con\ufb01gu - ration dominated by \ufb01laments growing perpendicularly ( 0 ! ) to the membrane . Toward the cell front the network again recov - ered gradually to canonical ( Figures 5E and S6D ) . The change in network geometry was due to selective elimination of \ufb01la - ments : in the transition zone the rate of \ufb01lament survival strictly depended on \ufb01lament orientation : \ufb01laments at a higher angle were preferentially eliminated , whereas low angle \ufb01laments had a higher rate of survival ( Figures S5E and S6D ) . These data show that a decrease in membrane tension causes a decrease in network density and a change in geometry , with more \ufb01laments growing perpendicularly to the plasma membrane . Figure 5 . Correlated Live Microscopy - Electron Tomography of Ultrastructural Changes in Networks with Decreasing Filament Density ( A ) Migrating keratocyte manipulated with a microneedle to induce a rapid decrease in projected cell area with an accompanying decrease in lifeact : GFP signal , \ufb01xed within z 3 s , and prepared for electron microscopy . A rapid decrease in the lifeact : GFP signal is preserved in the \ufb01xed lifeact : GFP sample and the low - magni\ufb01cation electron micrograph . ( B ) 5 . 5 - nm tomogram slice showing the region marked with a red box in ( A ) . The cell edge is seen on the right side , and the region of lower density is distin - guishable toward the middle of the micrograph . Region of steady - state network density and decreased density are marked in black and red throughout the whole \ufb01gure . ( C ) Filament tracks of the lamellipodium shown in ( B ) , with actin \ufb01laments shown in green , barbed ends in red , and pointed ends in blue . ( D ) Filament numbers and densities of barbed and pointed ends in 106 - nm - wide spatial bins throughout the lamellipodium shown in ( B ) and ( C ) . ( E ) Histogram showing \ufb01lament densities growing at the indicated angle from the membrane in 212 - nm distance bins . An order parameter ( blue ) in 212 nm is de\ufb01ned like in Figure 3 . See also Figures S5 and S6 and Movie S4 . 8 Cell 171 , 1 \u2013 13 , September 21 , 2017 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 A Stochastic 2D Model of Actin - Based Protrusion Recapitulates Light and Electron Microscopy Data To gain a quantitative understanding of the experimental data we built on earlier work by Maly and Borisy ( 2001 ) and Weichsel and Schwarz ( 2010 ) and developed a stochastic model of lamellipo - dial network growth . We formulated a two - dimensional ( 2D ) model taking into account \ufb01lament elongation and Arp2 / 3 - medi - ated ( 70 ! ) nucleation at the leading front . Elongation is termi - nated by capping whenever \ufb01laments detach from the mem - brane and thereby leave the zone in which they are protected from capping ( by elongation factors like VASP or formins ) . Fila - ments push the membrane forward according to a force - velocity relation based on thermal \ufb02uctuations as described in Dickinson ( 2009 ) and Mogilner and Oster ( 1996 , 2003 ) ( for details see STAR Methods and Tables S1 and S2 ) . Here , network geometry dic - tates that increased protrusion velocity ( as caused by decreased load ) rapidly decreases \ufb01lament density in an angle - dependent manner : A \ufb01lament growing at an angle 4 has to polymerize with \u00f0 1 = cos 4 \u00de the speed of the advancing membrane in order to keep contact with the membrane . Hence , with increasing pro - trusion speed \ufb01laments growing at higher angles will lag behind the leading edge and therefore detach from the membrane where they are protected from capping ( Figures 6A , 6B , S7D , and S7E ) . Consequently , higher angle \ufb01laments will be capped at higher rates and the family of low angle \ufb01laments ( and their accompanying \u00b1 70 ! side branches ) will outcompete the \u00b1 35 ! network . Slowing down the membrane has the opposite effect : More \ufb01laments at higher angles are able to catch up with the membrane , leading to a denser network consisting of \ufb01laments growing at a broad range of angles . The stochastic simulation is able to reproduce the correlative electron tomography \ufb01ndings of \ufb01lament orientation and density by using a sharp decrease from 300 pN as force input . 300 pN were taken as a realistic force reported for steady - state migrating \ufb01sh keratocytes ( Lieber et al . , 2013 ) . Together , our stochastic simulation faithfully repro - duced both kinetics and geometry of network parameters upon externally induced changes in load ( Figures 6C , 6D , S7A , and S7B ; Tables S1 and S2 ; Movie S5 ) . Next we wanted to test if the stochastic simulation can pro - duce the co - \ufb02uctuations of protrusion speed , actin density and projected area as we observed them in our light microscopy ex - periments in unperturbed migrating keratocytes ( Figure 1 ) . The amount of decrease in tension necessary to obtain the correct response at the \ufb01lament level in the correlated experiments was used as a scaling factor to relate membrane tension to pro - jected area . The experimental decrease and increase of mem - brane tension ( Figure 2 ) could be recapitulated by simulating representative protrusion events of 10 s with increasing or decreasing tension ( Figure S7C ) . By using the measured tempo - ral \ufb02uctuations in projected cell area ( Figure 1D ) with the ob - tained scaling factor as a proxy for tension we could recapitulate the positive correlation of membrane tension and actin density as well as the negative correlations of membrane tension , actin density , and protrusion speed ( Figures 7A and 7B ) . This showed that our \ufb01lament - level model is able to recapitu - late lamellipodial dynamics as we observed them at the cellular scale and using \ufb02uorescent markers . Our model predicts that keratocytes migrating on an adhesive substrate in 2D exhibit a canonically 70 ! branched network in their lamellipodia the vast majority of the time . This is consistent with the measurements on steady state migrating keratocytes . However , with a low fre - quency , periods of perpendicular dominated networks ( with an order parameter above 0 ) during phases of high protrusion speed are predicted ( Figure 7C ) . DISCUSSION We demonstrate that the lamellipodial actin network undergoes profound structural changes when \ufb01laments pushing against the membrane experience varying load . These changes happen at short time - scales and can be a sheer consequence of the geometricalandkineticpropertiesofArp2 / 3 - nucleatedbranched networks . The key - ingredients of the mechanism we propose are the force - velocity relation determining \ufb01lament elongation rates , the geometry of protruding networks , and the protection from capping near the membrane . Filaments polymerize faster when growing against decreased loads ( here , the load is due to mem - brane tension ) . With decreasing load the expanding network reaches a speed , where \ufb01laments growing perpendicularly to the membrane gradually \u2018\u2018outrun\u2019\u2019 \ufb01laments pushing at steeper angles because these have to travel a longer distance . Hence , these steeper - angle \ufb01laments are eliminated , because they lose contact with the membrane , where the polymerizing factors ( VASP and formins ) protect them from capping proteins . If the perpendicular \ufb01laments originate from \u2018\u2018noise\u2019\u2019 in the angular dis - tributions of a purely Arp2 / 3 - nucleated network or if they repre - sent a family nucleated by other factors like formins remains open , as \u2018\u2018non - branch\u2019\u2019 pointed ends in the tomograms might not only represent Arp2 / 3 - independent \ufb01laments but could also be Arp2 / 3 nucleated , but later severed or unbranched . Whereas this proposed mechanism will operate in any Arp2 / 3 - nucleated process like endocytosis , phagocytosis , vesicle traf\ufb01cking , intracellular pathogen transport , dendritic spine for - mation , etc . , its role in cell motility is most intuitive . Unlike in the idealized keratocyte system , cells migrating in physiological environments will almost never experience mechanically isotropic environments . First , a cell usually pushes against an in - terstitium with inhomogeneous viscoelastic features , and here it has been shown that lamellipodia can increase their force when counter - resistance increases locally ( Heinemann et al . , 2011 ; Prass et al . , 2006 ) . Second , the adhesiveness of the substrate is often variable ( e . g . , in any 3D \ufb01brillar environment ) . As the actin network slides back in areas of the cell , which are less coupled to the substrate , the load experienced by \ufb01laments polymerizing against the leading membrane will be reduced . It has been shown that such local inhomogeneities can be compensated by adaptations in polymerization speed , which \u2018\u2018\ufb01ll up the space\u2019\u2019 resulting from retrograde slippage of the network ( Barnhart et al . , 2011 ; Graziano and Weiner , 2014 ; Renkawitz et al . , 2009 ) . Thus , for a cell migrating through a complex 3D environment , the local adaptations we describe serve to keep the cell edge and its actin network coherent . This geometrical adaptation might provide another remarkable feature : wherever two adjacent regions of the network show differential \ufb02ow speed , the faster network will be less crosslinked to the neighboring , slower , regions , because it is less interconnected by branches . This might Cell 171 , 1 \u2013 13 , September 21 , 2017 9 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 Figure 6 . Stochastic Model Recapitulating Light and Electron Microscopy Results ( A ) The critical angle at which \ufb01laments are not able to keep up with the advancing membrane , 4 c , depends on the maximal polymerization rate , set here as 450nm / s , and theangleof the\ufb01lament towardthemembrane , 4 ( black line ) . A drop inexternal force leads toatransient increase inprotrusion speed leading toa decreasein 4 c andasubsequent , angle - dependentdecreasein\ufb01laments ( redline ) . Forceincreasecausesadecreaseinprotrusionspeed , leadingtoanincrease in 4 c and \ufb01lament density ( blue line ) . ( B ) Scheme of \ufb01laments growing at steady state at intermediate load ( middle , gray arrow ) . Following an increase in load ( top , blue arrow ) \ufb01lament density in - creasesandtheangledistributionbroadens , whereasatadecreaseinload ( bottom , redarrow ) \ufb01lamentsatahigherangle 4 i arepreferentiallycappedanddensity decreases . ( C ) Visualization ofmodelresultswith\ufb01laments showningreen , barbed endsinred , and pointedends inblue . Visualizationof modelresultsinaregime withforce increase ( top ) and decrease ( bottom ) , D F , at around 900 nm . Subsequent panels are aligned next to the respective force regimes . ( D ) Filamentnumbers , barbed , andpointedenddensityquanti\ufb01edforthethreeconditionslikeinFigures3 , 4 , and5 . Intheunperturbedsituationwith\ufb01xedexternal force , F , the network assumes a steady state with balanced \ufb01lament density , branching , and capping . After a decrease in F , capping increases , nucleation decreases , and the network is thinned out , whereas force increase causes a nucleation peak and an increase in \ufb01lament density . For all graphs , mean and SD for an average of 20 runs are shown . See also Figure S7 , Tables S1 and S2 , Movie S5 , and the STAR Methods . 10 Cell 171 , 1 \u2013 13 , September 21 , 2017 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 potentially allow the two networks to slide relative to each other and might explain how lamellipodia or actin cortices can harbor differentially \ufb02owing populations of actin as shown previously ( Ponti et al . , 2004 ; Vitriol et al . , 2015 ) . In our work we used manipulations of membrane tension as a proxy for altered load . For the protruding lamellipodium mem - brane tension likely plays the same mechanical role as an external load . However , lateral membrane tension cannot act locally as it equilibrates instantaneously across the cell surface ( Diz - Mun\u02dcoz et al . , 2013 ; Sens and Plastino , 2015 ) . Based on this potential signaling function it has been proposed that membrane tension acts as a global long - range regulator of cell polarity . Whenever lamellipodial actin stretches the leading membrane it suppresses polymerization at the sides and the trailing edge . Here , polymerization will stall under high mem - brane tension as it is not boosted by the biochemical enhancers driving the lamellipodium ( Houk et al . , 2012 ) . This mechanism to suppress competing protrusions has a pure mechanical compo - nent ( \ufb01laments stall under high load ) but is also complemented by a signaling mechanism , where membrane tension triggers mTORC2 signaling , which in turn dampens polymerization ( Diz - Mun\u02dcoz et al . , 2010 ) . Although this mechanism seems to coun - teract the geometrical adaptations we describe , it likely acts at much higher force regimes , where \ufb01laments ultimately fail to protrude the membrane . We think that the geometrical rules we describe here represent the most basic regulatory framework of Arp2 / 3 - mediated actin polymerization . They are a fundamental feature of the network structure , which demonstrates that mechanosensitivity is inherently built into the architecture of branched networks . Additional layers of regulation by chemoattractant - sensing and mechanosensitive signaling pathways will act on top of this basic Figure 7 . The Stochastic Model Recapitulates Cell Migration Parameters of Fish Keratocytes ( A ) Example of temporal \ufb02uctuations in lifeact : GFP intensity , area , and protrusion speed for the cell shown in Figure 1 compared to the migration parameters produced by a simulation run with the projected cell area as scaled tension input . ( B ) Average of temporal cross - correlation functions of the simulations obtained with the projected cell area of the 21 migrating keratocytes shown in Figure 1 . Temporal correlation peaks at time lag zero between lifeact : GFP intensity , projected cell area , and protrusion match the ones observed in vivo . Mean and SEM are shown . ( C ) Histogramoftheorderparametersobtainedforthe21steady - statemigratingkeratocytes . Thestochasticmodelpredictscellsmigratingwithapredominantly negative order parameter , with only occasional switches to positive order parameter values as cells undergo rapid shrinkage and a transient burst in protru - sion speed . See also Figure S7 . Cell 171 , 1 \u2013 13 , September 21 , 2017 11 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 mechanism and might add modulations e . g . , by altering the \ufb01la - ment distribution via different nucleators and elongators . STAR + METHODS Detailed methods are provided in the online version of this paper and include the following : d KEY RESOURCES TABLE d CONTACT FOR REAGENT AND RESOURCE SHARING d EXPERIMENTAL MODEL AND SUBJECT DETAILS B Zebra\ufb01sh d METHOD DETAILS B Keratocyte preparation B mRNA injection B Transfection B Membrane dye B Live cell imaging B Micromanipulators B Osmotic treatment B Tether pulling experiments and force measurements B Image analysis B Electron tomography B Actin \ufb01lament tracking and analysis B Correlated light and electron microscopy B Stochastic Simulation d QUANTIFICATION AND STATISTICAL ANALYSIS SUPPLEMENTAL INFORMATION Supplemental Information includes seven \ufb01gures , two tables , and \ufb01ve movies and can be found with this article online at http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 . AUTHOR CONTRIBUTIONS Conceptualization : J . M . and M . S . ; Methodology : J . M . , G . S . , A . D . L . , K . Keren , and R . H . ; Software : C . W . , K . Kruse , and C . S . ; Formal Analysis : J . M . , G . S . , and R . H . ; Investigation : J . M . , M . N . , I . d . V . , and A . D . L . ; Writing \u2013 Original Draft : J . M . and M . S . ; Writing \u2013 Review & Editing : all authors ; Visualization : J . M . and G . S . ; Project Administration : M . S . ; Funding Acquisition : M . S . ACKNOWLEDGMENTS Wethankthescienti\ufb01csupportfacilitiesofISTAustriaandBiocenterViennafor technical support and CP Heisenberg for \ufb01sh lines . This work was supported by the European Research Council ( ERC StG 281556 ) , a grant from the Aus - trian Science Foundation ( FWF ) ( to M . S . ) , and a grant from Vienna Science and Technology Fund ( WWTF ) ( No . LS13 - 029 to C . S . and M . S . ) . Research in the lab of K . Keren was supported by a grant from the United States - Israel Binational Science Foundation ( No . 2013275 with Alex Mogilner ) . Received : January 11 , 2017 Revised : May 21 , 2017 Accepted : July 31 , 2017 Published : August 31 , 2017 REFERENCES Barnhart , E . L . , Lee , K . C . , Keren , K . , Mogilner , A . , and Theriot , J . A . ( 2011 ) . An adhesion - dependent switch between mechanisms that determine motile cell shape . PLoS Biol . 9 , e1001059 . Behrndt , M . , Salbreux , G . , Campinho , P . , Hauschild , R . , Oswald , F . , Roensch , J . , Grill , S . W . , and Heisenberg , C . P . ( 2012 ) . Forces driving epithelial spreading in zebra\ufb01sh gastrulation . Science 338 , 257 \u2013 260 . Bieling , P . , Li , T . - D . , Weichsel , J . , McGorty , R . , Jreij , P . , Huang , B . , Fletcher , D . A . , and Mullins , R . D . ( 2016 ) . Force feedback controls motor activity and me - chanical properties of self - assembling branched actin networks . Cell 164 , 115 \u2013 127 . Carlsson , A . E . ( 2010 ) . Actin dynamics : from nanoscale to microscale . Annu . Rev . Biophys . 39 , 91 \u2013 110 . De\u00b4moulin , D . , Carlier , M . - F . , Bibette , J . , and Baudry , J . ( 2014 ) . Power trans - duction of actin \ufb01laments ratcheting in vitro against a load . Proc . Natl . Acad . Sci . USA 111 , 17845 \u2013 17850 . Dickinson , R . B . ( 2009 ) . Models for actin polymerization motors . J . Math . Biol . 58 , 81 \u2013 103 . Diz - Mun\u02dcoz , A . , Krieg , M . , Bergert , M . , Ibarlucea - Benitez , I . , Muller , D . J . , Pal - uch , E . , and Heisenberg , C . - P . ( 2010 ) . Control of directed cell migration in vivo by membrane - to - cortex attachment . PLoS Biol . 8 , e1000544 . Diz - Mun\u02dcoz , A . , Fletcher , D . A . , and Weiner , O . D . ( 2013 ) . Use the force : mem - brane tension as an organizer of cell shape and motility . Trends Cell Biol . 23 , 47 \u2013 53 . Gauthier , N . C . , Fardin , M . A . , Roca - Cusachs , P . , and Sheetz , M . P . ( 2011 ) . Temporary increase in plasma membrane tension coordinates the activation of exocytosis and contraction during cell spreading . Proc . Natl . Acad . Sci . USA 108 , 14467 \u2013 14472 . Graziano , B . R . , and Weiner , O . D . ( 2014 ) . Self - organization of protrusions and polarity during eukaryotic chemotaxis . Curr . Opin . Cell Biol . 30 , 60 \u2013 67 . Heinemann , F . , Doschke , H . , andRadmacher , M . ( 2011 ) . Keratocytelamellipo - dial protrusion is characterized by a concave force - velocity relation . Biophys . J . 100 , 1420 \u2013 1427 . Houk , A . R . , Jilkine , A . , Mejean , C . O . , Boltyanskiy , R . , Dufresne , E . R . , Ange - nent , S . B . , Altschuler , S . J . , Wu , L . F . , and Weiner , O . D . ( 2012 ) . Membrane ten - sion maintains cell polarity by con\ufb01ning signals to the leading edge during neutrophil migration . Cell 148 , 175 \u2013 188 . Ji , L . , Lim , J . , andDanuser , G . ( 2008 ) . Fluctuationsofintracellularforcesduring cell protrusion . Nat . Cell Biol . 10 , 1393 \u2013 1400 . Keren , K . ( 2011 ) . Cell motility : the integrating role of the plasma membrane . Eur . Biophys . J . 40 , 1013 \u2013 1027 . Keren , K . , Pincus , Z . , Allen , G . M . , Barnhart , E . L . , Marriott , G . , Mogilner , A . , and Theriot , J . A . ( 2008 ) . Mechanism of shape determination in motile cells . Nature 453 , 475 \u2013 480 . Kremer , J . R . , Mastronarde , D . N . , and McIntosh , J . R . ( 1996 ) . Computer visualization of three - dimensional image data using IMOD . J . Struct . Biol . 116 , 71 \u2013 76 . Kreshuk , A . , Straehle , C . N . , Sommer , C . , Koethe , U . , Cantoni , M . , Knott , G . , andHamprecht , F . A . ( 2011 ) . Automateddetectionandsegmentationofsynap - tic contacts in nearly isotropic serial electron microscopy images . PLoS ONE 6 , e24899 . Lai , F . P . L . , Szczodrak , M . , Block , J . , Faix , J . , Breitsprecher , D . , Mannherz , H . G . , Stradal , T . E . B . , Dunn , G . A . , Small , J . V . , and Rottner , K . ( 2008 ) . Arp2 / 3 complex interactions and actin network turnover in lamellipodia . EMBO J . 27 , 982 \u2013 992 . Lee , C . T . , Hoopes , M . F . , Diehl , J . , Gilliland , W . , Huxel , G . , Leaver , E . V . , McCann , K . , Umbanhowar , J . , and Mogilner , A . ( 2001 ) . Non - local concepts and models in biology . J . Theor . Biol . 210 , 201 \u2013 219 . Lee , K . , Elliott , H . L . , Oak , Y . , Zee , C . - T . , Groisman , A . , Tytell , J . D . , and Dan - user , G . ( 2015 ) . Functional hierarchy of redundant actin assembly factors re - vealed by \ufb01ne - grained registration of intrinsic image \ufb02uctuations . Cell Syst . 1 , 37 \u2013 50 . Lieber , A . D . , Yehudai - Resheff , S . , Barnhart , E . L . , Theriot , J . A . , and Keren , K . ( 2013 ) . Membranetensioninrapidlymovingcellsisdeterminedbycytoskeletal forces . Curr . Biol . 23 , 1409 \u2013 1417 . 12 Cell 171 , 1 \u2013 13 , September 21 , 2017 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 Lou , S . S . , Diz - Mun\u02dcoz , A . , Weiner , O . D . , Fletcher , D . A . , andTheriot , J . A . ( 2015 ) . Myosin light chain kinase regulates cell polarization independently of mem - brane tension or Rho kinase . J . Cell Biol . 209 , 275 \u2013 288 . Ma\u0131\u02c6tre , J . L . , Berthoumieux , H . , Krens , S . F . G . , Salbreux , G . , Ju\u00a8licher , F . , Pal - uch , E . , and Heisenberg , C . P . ( 2012 ) . Adhesion functions in cell sorting by me - chanically coupling the cortices of adhering cells . Science 338 , 253 \u2013 256 . Maly , I . V . , and Borisy , G . G . ( 2001 ) . Self - organization of a propulsive actin network as an evolutionary process . Proc . Natl . Acad . Sci . USA 98 , 11324 \u2013 11329 . Manhart , A . , Oelz , D . , Schmeiser , C . , and Sfakianakis , N . ( 2015 ) . An extended \ufb01lament based lamellipodium model produces various moving cell shapes in the presence of chemotactic signals . J . Theor . Biol . 382 , 244 \u2013 258 . Mastronarde , D . N . ( 2005 ) . Automated electron microscope tomography using robust prediction of specimen movements . J . Struct . Biol . 152 , 36 \u2013 51 . Mogilner , A . , and Keren , K . ( 2009 ) . The shape of motile cells . Curr . Biol . 19 , R762 \u2013 R771 . Mogilner , A . , and Oster , G . ( 1996 ) . Cell motility driven by actin polymerization . Biophys . J . 71 , 3030 \u2013 3045 . Mogilner , A . , and Oster , G . ( 2003 ) . Force generation by actin polymerization II : the elastic ratchet and tethered \ufb01laments . Biophys . J . 84 , 1591 \u2013 1605 . Narita , A . , Mueller , J . , Urban , E . , Vinzenz , M . , Small , J . V . , andMae\u00b4da , Y . ( 2012 ) . Direct determinaftion of actin polarity in the cell . J . Mol . Biol . 419 , 359 \u2013 368 . Parekh , S . H . , Chaudhuri , O . , Theriot , J . A . , and Fletcher , D . A . ( 2005 ) . Loading history determines the velocity of actin - network growth . Nat . Cell Biol . 7 , 1219 \u2013 1223 . Pollard , T . D . ( 2007 ) . Regulation of actin \ufb01lament assembly by Arp2 / 3 complex and formins . Annu . Rev . Biophys . Biomol . Struct . 36 , 451 \u2013 477 . Pollard , T . D . , and Cooper , J . A . ( 2009 ) . Actin , a central player in cell shape and movement . Science 326 , 1208 \u2013 1212 . Ponti , A . , Machacek , M . , Gupton , S . L . , Waterman - Storer , C . M . , and Danuser , G . ( 2004 ) . Two distinct actin networks drive the protrusion of migrating cells . Science 305 , 1782 \u2013 1786 . Prass , M . , Jacobson , K . , Mogilner , A . , and Radmacher , M . ( 2006 ) . Direct mea - surement of the lamellipodial protrusive force in a migrating cell . J . Cell Biol . 174 , 767 \u2013 772 . Raucher , D . , and Sheetz , M . P . ( 2000 ) . Cell spreading and lamellipodial exten - sion rate is regulated by membrane tension . J . Cell Biol . 148 , 127 \u2013 136 . Renkawitz , J . , Schumann , K . , Weber , M . , LAmmermann , T . , P\ufb02icke , H . , Piel , M . , Polleux , J . , Spatz , J . P . , and Sixt , M . ( 2009 ) . Adaptive force transmission in amoeboid cell migration . Nat . Cell Biol . 11 , 1438 \u2013 1443 . Riedl , J . , Crevenna , A . H . , Kessenbrock , K . , Yu , J . H . , Neukirchen , D . , Bista , M . , Bradke , F . , Jenne , D . , Holak , T . A . , Werb , Z . , et al . ( 2008 ) . Lifeact : a versatile marker to visualize F - actin . Nat . Methods 5 , 605 \u2013 607 . Ryan , G . L . , Petroccia , H . M . , Watanabe , N . , and Vavylonis , D . ( 2012 ) . Excitable actin dynamics in lamellipodial protrusion and retraction . Biophys . J . 102 , 1493 \u2013 1502 . Schaus , T . E . , Taylor , E . W . , and Borisy , G . G . ( 2007 ) . Self - organization of actin \ufb01lament orientation inthe dendritic - nucleation / array - treadmilling model . Proc . Natl . Acad . Sci . USA 104 , 7086 \u2013 7091 . Schindelin , J . , Arganda - Carreras , I . , Frise , E . , Kaynig , V . , Longair , M . , Pietzsch , T . , Preibisch , S . , Rueden , C . , Saalfeld , S . , Schmid , B . , et al . ( 2012 ) . Fiji : an open - source platform for biological - image analysis . Nat . Methods . 9 , 676 \u2013 682 . Sens , P . , and Plastino , J . ( 2015 ) . Membrane tension and cytoskeleton organi - zation in cell motility . J . Phys . Condens . Matter 27 , 273103 . Sivaramakrishnan , S . , and Spudich , J . A . ( 2009 ) . Coupled myosin VI motors facilitate unidirectional movement on an F - actin network . J . Cell Biol . 187 , 53 \u2013 60 . Small , J . V . , Herzog , M . , andAnderson , K . ( 1995 ) . Actin\ufb01lament organization in the \ufb01sh keratocyte lamellipodium . J . Cell Biol . 129 , 1275 \u2013 1286 . Svitkina , T . M . , Verkhovsky , A . B . , McQuade , K . M . , and Borisy , G . G . ( 1997 ) . Analysis of the actin - myosin II system in \ufb01sh epidermal keratocytes : mecha - nism of cell body translocation . J . Cell Biol . 139 , 397 \u2013 415 . Theriot , J . A . , and Mitchison , T . J . ( 1991 ) . Actin micro\ufb01lament dynamics inloco - moting cells . Nature 352 , 126 \u2013 131 . Vinzenz , M . , Nemethova , M . , Schur , F . , Mueller , J . , Narita , A . , Urban , E . , Win - kler , C . , Schmeiser , C . , Koestler , S . A . , Rottner , K . , etal . ( 2012 ) . Actinbranching in the initiation and maintenance of lamellipodia . J . Cell Sci . 125 , 2775 \u2013 2785 . Vitriol , E . A . , McMillen , L . M . , Kapustina , M . , Gomez , S . M . , Vavylonis , D . , and Zheng , J . Q . ( 2015 ) . Two functionally distinct sources of actin monomers sup - ply the leading edge of lamellipodia . Cell Rep . 11 , 433 \u2013 445 . Weichsel , J . , and Schwarz , U . S . ( 2010 ) . Two competing orientation patterns explain experimentally observed anomalies in growing actin networks . Proc . Natl . Acad . Sci . USA 107 , 6304 \u2013 6309 . Wilson , C . A . , Tsuchida , M . A . , Allen , G . M . , Barnhart , E . L . , Applegate , K . T . , Yam , P . T . , Ji , L . , Keren , K . , Danuser , G . , andTheriot , J . A . ( 2010 ) . MyosinIIcon - tributes to cell - scale actin network treadmilling through network disassembly . Nature 465 , 373 \u2013 377 . Winkler , C . , Vinzenz , M . , Small , J . V . , and Schmeiser , C . ( 2012 ) . Actin \ufb01lament tracking in electron tomograms of negatively stained lamellipodia using the localized radon transform . J . Struct . Biol . 178 , 19 \u2013 28 . Ydenberg , C . A . , Smith , B . A . , Breitsprecher , D . , Gelles , J . , and Goode , B . L . ( 2011 ) . Cease - \ufb01re at the leading edge : new perspectives on actin \ufb01lament branching , debranching , and cross - linking . Cytoskeleton 68 , 596 \u2013 602 . Ydenberg , C . A . , Padrick , S . B . , Sweeney , M . O . , Gandhi , M . , Sokolova , O . , and Goode , B . L . ( 2013 ) . GMF severs actin - Arp2 / 3 complex branch junctions by a co\ufb01lin - like mechanism . Curr . Biol . 23 , 1037 \u2013 1045 . Cell 171 , 1 \u2013 13 , September 21 , 2017 13 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 STAR + METHODS KEY RESOURCES TABLE CONTACT FOR REAGENT AND RESOURCE SHARING Further information and requests for resources and reagents should be directed to and will be ful\ufb01lled by the Lead Contact , Michael Sixt ( sixt @ ist . ac . at ) . EXPERIMENTAL MODEL AND SUBJECT DETAILS Zebra\ufb01sh Wild - type zebra\ufb01sh lines AB and TL and the transgenic line Tg ( actb1 : lifeact - GFP ) ( Behrndt et al . , 2012 ) were housed and bred in the IST Austria \ufb01sh facility . Zebra\ufb01sh were maintained at 28 ! C and embryos collected according to standard \ufb01sh laboratory protocol . Animals were sacri\ufb01ced for experiments at 6 months to about 1 . 5 years of age regardless of sex . REAGENT or RESOURCE SOURCE IDENTIFIER Chemicals , Peptides , and Recombinant Proteins Dulbecco\u2019s Modi\ufb01ed Eagle\u2019s Medium GIBCO Cat # 10566 - 016 L - 15 Leibovitz medium Sigma Cat # L1518 PCR puri\ufb01cation kit Qiagen Cat # 28106 SP6 mMessage mMachine kit Invitrogen Cat # AM1340 Fugene 6 Roche Cat # 11814443001 Cell Mask Orange Molecular Probes Cat # C10045 PLL ( 20 ) - g [ 3 . 5 ] - PEG ( 2 ) / PEG ( 3 . 4 ) - RGD Surface Solutions N / A PLL - PEG - RGD Surface Solutions N / A Experimental Models : Organisms / Strains Zebra\ufb01sh line : Tg ( actb1 : lifeact - GFP ) Gift from Heisenberg Laboratory ( IST Austria ) Behrndt et al . , 2012 N / A Zebra\ufb01sh line : wt AB Gift from Heisenberg Laboratory ( IST Austria ) N / A Zebra\ufb01sh line : wt TL Gift from Heisenberg Laboratory ( IST Austria ) N / A Recombinant DNA pEGFP - actin Clontech Cat # 6084 - 1 zf actin : GFP This paper N / A Software and Algorithms MATLAB 2015b The MathWorks https : / / ch . mathworks . com / products / matlab Fiji 1 . 50d Schindelin et al . , 2012 https : / / \ufb01ji . sc / Ilastik 1 . 1 . 5 The ilastik Team http : / / ilastik . org / index . html SerialEM 3 . x Mastronarde , 2005 http : / / bio3d . colorado . edu / ftp / SerialEM / IMOD 4 . 7 Kremer et al . , 1996 http : / / bio3d . colorado . edu / imod / Python 2 . 7 Python Software Foundation https : / / www . python . org / download / releases / 2 . 7 / Prism 5 . 0b GraphPad https : / / www . graphpad . com / scienti\ufb01c - software / prism / LabView National Institutes http : / / www . ni . com / en - us / shop . html NIS - Elements AR 3 . 2 Nikon Instruments https : / / www . nikoninstruments . com / en _ EU / Products / Software / NIS - Elements - Advanced - Research Stochastical Simulation \u2013 Code repository Authors of this study https : / / github . com / gszep / lamellipodium Other 200mesh HF15 Gold \ufb01nder grids Maxtaform Cat # G245A 200mesh hexagonal gold grids Agar Scienti\ufb01c Cat # AGG2450A e1 Cell 171 , 1 \u2013 13 . e1 \u2013 e6 , September 21 , 2017 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 Primary keratocyte cultures for these experiments were prepared from the Central American cichlid Hypsophrys Nicaraguensis as described previously ( Lieber et al . , 2013 ) . METHOD DETAILS Keratocyte preparation Keratocytes were prepared from adult zebra\ufb01sh scales ( plucked from sacri\ufb01ced animals ) and washed three times with Dulbecco\u2019s Modi\ufb01ed Eagle\u2019s Medium ( DMEM ) ( GIBCO ) . Scales were incubated in START medium ( Small et al . , 1995 ) at room temperature for one or two days to allow keratocytes to migrate off in a monolayer . The monolayer was then washed three times in PBS , incubated for 40min with Running Buffer ( Small et al . , 1995 ) with 1mM EGTA and the scales removed . The remaining adherent cells were washed three times with PBS , trypsinized for 2min with 0 . 25 % Trypsin - EDTA ( GIBCO ) at room temperature , resuspended in the same volume of trypsin inhibitor ( Sigma ) and transferred to a coverslip or electron microscopy grid . After allowing the cells to adhere for 40min the medium was changed to START medium , the cells were incubated for another 40min before live cell imaging or \ufb01xation for electron microscopy . mRNA injection To produce synthetic mRNA , zf actin : GFP was linearized by double strand cutting with NotI , puri\ufb01ed with a PCR puri\ufb01cation kit ( Qiagen ) and mRNA was synthesized using the SP6 mMessage mMachine Kit ( Ambion ) . 50pg of mRNA were injected in freshly har - vested embryos . Zebra\ufb01sh embryonic keratocytes were prepared from these embryos 1 day post fertilization as described elsewhere ( Lou et al . , 2015 ) . Transfection The plasmid used for transfection was pEGFP - actin ( Clontech ) . Transfection was performed by mixing 100 m l OPTIMEM , 6 m l Fugene 6 ( Roche ) and 1 m g plasmid DNA for 30min and adding it to the cells in L - 15 Leibovitz medium ( Sigma ) for 5h . Then the medium was changed to START medium again and the transfected cells were re - plated onto coverslips as described above . Membrane dye As a membrane dye for negative controls CellMask Orange plasma membrane stain ( Molecular probes ) was used according to the manufacturer\u2019s protocol for live , adherent cell cultures with the following modi\ufb01cations : Keratocytes seeded on glass coverslips were incubated at room temperature for 15min , washed \ufb01ve times with PBS and imaged immediately in START medium . Live cell imaging All live cell imaging was performed at room temperature in START medium . Confocal microscopy was performed with an inverted microscope ( Zeiss ) , equipped with a Spinning disk system ( Yokogawa X1 , iXon897 , Andor ) , a C - Apochromat 63x / 1 . 2 Water Objective ( Zeiss ) , a motorized stage and 488nm and 561nm lasers . While imaging with this microscope , laser ablations of the trailing edge were performed with a 355nm laser cutter . For wide\ufb01eld imaging an inverted wide\ufb01eld microscope ( Nikon ) , equipped with 60x / 1 . 4 and 100x / 1 . 4 Apochromat Oil Objectives , a motorized stage and a light source with \ufb02exible excitation bands ( Lumencor ) was used . Micro - pipette aspiration experiments were performed on an upright confocal microscope ( SP5 , Leica ) , equipped with HCX 63x / 1 . 4 Apo - chromat Oil Objective ( Leica ) , a motorized stage and a 488nm laser line . The time interval was 0 . 2 s - 1 . 0 s . Micromanipulators Micropipettes were produced from 1 . 0mm diameter glass capillaries as described before ( Ma\u0131\u02c6tre et al . , 2012 ) . For cell manipulations negative pressures from 10mbar to 40mbar were used for aspirating the cell membrane and 0 - 2mbar of positive pressure to release the cell again . For correlated live and electron microscopy cell membrane was aspirated with 30mbar for about 5 s and extracted and \ufb01xed about 3 s after release from the micropipette . The same micromanipulator system was used to induce rapid cell shrinking by carefully detaching part of the cell from the underlying substrate . In this case , the cell was \ufb01xed on the grid immediately afterward and prepared for electron microscopy as described below . Osmotic treatment For experiments with changing osmotic pressure , lifeact : GFP expressing keratocytes were subjected to repeated additions of 50 m l 0 . 2M sucrose in START medium and 50 m l double distilled water to a starting volume of 300 m l START medium . Tether pulling experiments and force measurements Primary keratocyte cultures for these experiments were prepared from the Central American cichlid Hypsophrys Nicaraguensis as described previously ( Lieber et al . , 2013 ) . One day old cultures were replated and cultured at room temperature in Leibovitz\u2019s L - 15 media ( GIBCO BRL , Grand Island , NY ) , supplemented with 14 . 2 mM HEPES pH 7 . 4 , 10 % Fetal Bovine Serum ( Invitrogen , Grand Island , NY ) , and 1 % antibiotic - antimycotic ( GIBCO BRL ) . Tether force measurements were carried out as in ( Lieber et al . , 2013 ) with a laser tweezers system ( PALM microtweezers , Carl Zeiss MicroImaging GmbH , Jena , Germany ) using a 63 3 1 . 2 NA water immersion Cell 171 , 1 \u2013 13 . e1 \u2013 e6 , September 21 , 2017 e2 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 objective and a motorized stage ( Ludl Electronic Products , Hawthorne , NY ) on an inverted microscope ( Axiovert 200M , Zeiss , Jena , Germany ) . Trapping was done with a 3W 1064nm Nd : YAG laser focused to a diffraction limited spot , and imaging by bright \ufb01eld was done simultaneously . Tether force measurements on cells were done by attaching a Concanavalin A coated bead to the rear part of the cell body and pulling it away by moving the stage . The measured tether force was corrected for the contribution of the dynamic friction due to stage movement , when relevant , as in ( Lieber et al . , 2013 ) . A membrane tether was held continuously during sponta - neous instances of abrupt rear retraction , providing a read out of the changes in the tether force that accompany the large changes in the projected cell area during the retraction event . Membrane tension values T were calculated from the tether force F T using , T = \u00f0 F 2 T = 8 p 2 B \u00de , where B = 0 : 14 pN , m m is the measured bending modulus of the membrane in keratocytes ( Lieber et al . , 2013 ) . Image analysis Time - lapse images of migrating keratocytes were analyzed in Fiji by \ufb01rst de\ufb01ning the projected cell area by applying a Gaussian blur to the raw image stack and then threshold it to obtain a stack of binary images . The same binary image was used to de\ufb01ne a 1 . 09 m m wide region along the leading edge as the lamellipodial region of interest . An optical \ufb02ow analysis based on the Horn - Schunck method was then used to calculate the optical \ufb02ow between the binary mask frames . Both the original image stack and the optical \ufb02ow data were stabilized using a custom - made image - stabilizing MATLAB algorithm . The resulting stabilized stacks were further analyzed by skeletonizing the 1 . 09 m m wide lamellipodium region for each stack slice and plotting both the gray value of the original image and the amount of shift determined by the optical \ufb02ow analysis along the whole length of the leading edge for every time step . A custom - made MATLAB program was used for this step . The resulting 2D heatmaps of lifeact : GFP , actin : GFP or membrane dye intensity and protrusion were averaged along the length of the leading edge and plotted against time and the whole images were correlated to each other and the projected area . For cross correlation analysis the \ufb01rst derivative of the lifeact : GFP , actin : GFP or membrane dye intensity and protrusion heatmaps as well as the projected cell area were taken and the cross covariance function of MATLAB was used normalized so that the auto - covariance at zero lag equaled 1 . For analysis of signal \ufb02ow within the lamellipo - dium 1 . 09 m m wide bands were eroded from the leading edge in consecutive steps and the resulting bands were used to quantify the lifeact : GFP intensity in adjacent parts further removed from the leading edge . The resulting heatmaps were correlated to each other using the same procedure as described above and the resulting peaks were multiplied by the speed of the respective cell to obtain the distance lag . The calculated distance was plotted against the known distance between the eroded bands . Electron tomography Cells were grown on 200mesh HF15 Gold \ufb01nder grids ( Maxtaform ) and 200mesh hexagonal gold grids ( Agar Scienti\ufb01c ) coated with 2 % Formvar and a 2nm carbon layer and incubated with 0 . 5mg / ml PLL ( 20 ) - g [ 3 . 5 ] - PEG ( 2 ) / PEG ( 3 . 4 ) - RGD ( Surface Solutions ) for 50min . Extraction and \ufb01xation was performed using 0 . 25 % glutaraldehyde and 0 . 5 % Triton X - 100 in Cytoskeleton Buffer ( 10mM MES buffer , 150mM NaCl , 5mM EGTA , 5mM glucose , 5mM MgCl2 , pH 6 . 1 ) for 1min at room temperature . The grids were then trans - ferred to 2 % glutaraldehyde in Cytoskeleton buffer for 10min and kept in another dish with 2 % glutaraldehyde in Cytoskeleton buffer at 4 ! C until staining for electron microscopy . For staining , grids were carefully blotted with \ufb01lter paper from the bottom and imme - diately stained with 70 m l 4 % sodium silicotungstate supplemented with BSA - gold \ufb01ducials from a gold stock as described before ( Vinzenz et al . , 2012 ) . Double axis tilt series with typical tilt angles from # 65 ! to + 65 ! and 1 ! increments following the Saxton scheme were acquired on an FEI Tecnai G20 transmission electron microscope operated at 200kV equipped with a Eagle 4k HS CCD camera ( Gatan ) . The automated acquisition of double axes tilt series was driven by SerialEM 3 . x ( Mastronarde , 2005 ) . The defocus was set to # 5 m m for all acquisitions and the primary on - screen magni\ufb01cation was 25 , 000x . For montage tomograms montages of 2x2 or 3x2 images were acquired for each tilt angle . Back - projection of the tilt series was performed using the IMOD package 4 . 7 with gold particle \ufb01ducials as markers and high - pass \ufb01ltering of the \ufb01nal reconstruction ( Kremer et al . , 1996 ) . A typical tomogram comprised 50 - 70 stacks of 0 . 92nm and an area of 1 . 9 m m x 1 . 9 m m . The montage tomograms had dimensions of 3 . 1 m m x 3 . 1 m m and 4 . 0 m m x 3 . 1 m m , respectively . Actin \ufb01lament tracking and analysis Filaments were quanti\ufb01ed in three different ways : Manual tracking of the region shown in Figure S6B using 3dmod software . Auto - mated tracking using a MATLAB - based tracking algorithm ( Winkler et al . , 2012 ) . Computerized segmentation of actin \ufb01laments by the pixel classi\ufb01cation function of ilastik 1 . 1 was used to quantify actin \ufb01lament density in the correlated tomograms ( Summarized in Fig - ures S6A \u2013 S6C ) ( Kreshuk et al . , 2011 ) . The manually and automatically tracked \ufb01laments were further analyzed in the following way . In the tomogram montages the re - corded \ufb01laments were divided in 106 . 15nm wide distance bins spanning from the leading edge up to 2 . 8 m m toward the back of the lamellipodium ( shown in Figures S4D and S6E ) . For the single tomograms of actin \ufb01lament density steps two regions of 300nm each before and after the step were used . In automated tomograms \ufb01laments shorter than 55nm were excluded from analysis . Filament numbers indicate the number of individual \ufb01laments crossing a plane in the middle of the respective analysis region . Pointed and barbed ends were de\ufb01ned as ends of \ufb01laments furthest and closest from the leading edge , which was clearly visible in every e3 Cell 171 , 1 \u2013 13 . e1 \u2013 e6 , September 21 , 2017 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 case , respectively . Filament angles were de\ufb01ned as the angle subtended between a \ufb01lament and the leading edge in the plane of the \ufb02at lamellipodium , with 0 ! meaning a \ufb01lament growing perpendicularly toward the leading edge . Angles were weighted by the total length of the \ufb01laments growing in that direction . The density in the different angle bins was taken as number of \ufb01laments growing at that angle normalized by the size of the angle bin . As a global measure of \ufb01lament architecture an order parameter was de\ufb01ned as : \u00f0 Filaments # 10 ! to + 10 ! \u00de # \u00f0 Filaments # 40 ! to # 20 ! + Filaments + 20 ! to + 40 ! \u00de = 2 \u00f0 Filaments # 10 ! to + 10 ! \u00de + \u00f0 Filaments # 40 ! to # 20 ! + Filaments + 20 ! to + 40 ! \u00de = 2 ( Weichsel and Schwarz , 2010 ) . Correlated light and electron microscopy For correlated light and electron microscopy 200mesh HF15 Gold Finder grids were coated with 0 . 5mg / ml PLL - PEG - RGD ( Surface Solutions ) for 50min . Then keratocytes were allowed to adhere for z 1h , imaged directly on the grid with an inverted confocal spinning disk microscope and either aspirated for about 4 s and then released or partly detached from the grid using micromanipulators as described above . Immediately after release from the micropipette or induction of a rapid shrinking event the cells were \ufb01xed by taking off the medium and adding the extraction and \ufb01xation buffers . To \ufb01nd the correlated cells back , the grids were \ufb01rst completely map - ped by using the navigator map function of SerialEM on an FEI Tecnai G20 transmission electron microscope . Once the relevant grid square was found , high magni\ufb01cation montages and tomograms of the transition zones were acquired . Stochastic Simulation Stochastic 2D models of actin - based lamellipodial protrusion have been proposed in the past to show the self - organization of orien - tation patterns ( Maly and Borisy , 2001 ; Schaus et al . , 2007 ; Weichsel and Schwarz , 2010 ) . These models considered network steady state analysis and simulations with \ufb01xed velocity . We build on the published models to include transient responses of leading edge velocity and network organization due to changing external forces exerted on the \ufb01laments . This simulation was implemented in py - thon 2 . 7 and has been made publicly available at https : / / github . com / gszep / lamellipodium . We formulate a 2D model , where each actin \ufb01lament is rigid and immobile . The \ufb01rst assumption can be justi\ufb01ed by electron tomo - grams , where mostly straight \ufb01laments are observed . Immobility of \ufb01laments has to be understood relative to each other , which still permits a coordinated retrograde \ufb02ow . Thus the state of the network close to the leading edge at time t can be described by the set F \u00f0 t \u00de with N \u00f0 t \u00de number of growing barbed ends , each having 2D position vectors x \u00f0 t \u00de = \u00f0 x \u00f0 t \u00de ; y \u00f0 t \u00de\u00de and constant orientations u . F \u00f0 t \u00de = n \u00f0 x \u00f0 t \u00de ; u \u00de 1 ; . ; \u00f0 x \u00f0 t \u00de ; u \u00de N \u00f0 t \u00de o ( 1 ) u encodes the direction and magnitude j u j = d of elongation with the addition of one actin monomer . The plasma membrane is modeled as a \ufb02at edge parallel to the y - axis at position p \u00f0 t \u00de along the x - axis . With all \ufb01laments only pushing from one side , the resul - tant protrusion velocity v \u00f0 t \u00de is positive in the x - direction . Periodic boundary conditions are imposed in the interval 0 % y \u00f0 t \u00de < L along the y - axis . We shall describe elongation , branching , and capping events at a barbed end at position x \u2013 dropping time variable for conve - nience \u2013 at time t as Poisson processes with rates l \u00f0 x ; t \u00de ; b \u00f0 x ; t \u00de and k \u00f0 x ; t \u00de respectively . This means that in a suf\ufb01ciently small time step from t to t + D t , the \ufb01lament will experience one of these events given by rate u \u00f0 x ; t \u00de with probability u \u00f0 x ; t \u00de D t . We introduce a useful notation n \u00f0 x ; t j u \u00de which is a Boolean variable that tells us whether an event with rate u \u00f0 x ; t \u00de happened or not at a barbed end at position x and time t . Elongation happens in the strip of width w behind the leading edge position p \u00f0 t \u00de with the rate l 0 and zero otherwise . This assump - tion is motivated by the observation that most growth happens at the leading edge . l \u00f0 x ; t \u00de = ! l 0 p \u00f0 t \u00de # w < x \u00f0 t \u00de < p \u00f0 t \u00de 0 elsewhere ( 2 ) An elongation event for a particular barbed end means x \u00f0 t + D t \u00de = x \u00f0 t \u00de + u . There are N \u00f0 t \u00de barbed ends , meaning we have to check events at each one of them . Branching occurs at the leading edge with a rate that depends on the linear density of nucleation factors a \u00f0 t \u00de such as Arp2 / 3 at the leading edge ( Weichsel and Schwarz , 2010 ) . Let a 0 be the rate of activation that leads to branching . b \u00f0 x ; t \u00de = ! a 0 a \u00f0 t \u00de p \u00f0 t \u00de # w < x \u00f0 t \u00de < p \u00f0 t \u00de 0 elsewhere Consider the \ufb01nite availability of nucleation factors and activation being the rate - limiting step in branching new \ufb01laments . Let D \u00f0 t \u00de be the linear density of barbed ends along the leading edge , which increases with branching rate a 0 a \u00f0 t \u00de and decreases with possibly time dependent capping rate k \u00f0 t \u00de : In addition , we know that there is a \ufb01nite positive rate of arrival b 0 of nucleation factors at the lead - ing edge coming from a reservoir . Finally , nucleation factors are depleted when branching occurs , with a rate a 0 D \u00f0 t \u00de depending of barbed end concentration . Cell 171 , 1 \u2013 13 . e1 \u2013 e6 , September 21 , 2017 e4 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 v t D \u00f0 t \u00de = a 0 a \u00f0 t \u00de D \u00f0 t \u00de # k \u00f0 t \u00de D \u00f0 t \u00de v t a \u00f0 t \u00de = b 0 # a 0 a \u00f0 t \u00de D \u00f0 t \u00de Assuming that the arrival of nucleation factors at the leading edge is faster than branching or capping events , we can employ the quasi - steady state assumption ( Manhart et al . , 2015 ) , setting v t a \u00f0 t \u00de = 0 . Rearranging for the concentration a \u00f0 t \u00de ; we can \ufb01nally write : b \u00f0 x ; t \u00de = 8 > < > : b 0 D \u00f0 t \u00de p \u00f0 t \u00de # w < x \u00f0 t \u00de < p \u00f0 t \u00de 0 elsewhere ( 3 ) Branching add a new \ufb01lament \u00f0 x 0 \u00f0 t + D t \u00de ; u 0 \u00de to the frontier at the position of the barbed end x 0 \u00f0 t + D t \u00de = x \u00f0 t \u00de of the mother \ufb01lament . The orientation is rotated by a random branching angle q sampled from a distribution . This can be written as a rotation matrix R \u00f0 q \u00de acting on the mother \ufb01lament orientation u 0 = R \u00f0 q \u00de u . The branching angle distribution B \u00f0 q j m ; s \u00de is the sum of two Gaussians with respective means at \u00b1 m and standard deviations s taken from electron microscopy data . The daughter \ufb01lament is required to be oriented toward the leading edge , meaning that the random choice might have to be repeated , until this requirement is ful\ufb01lled . This model neglects that there is a preferred distance between branches , due to the helical structure of actin \ufb01laments ( Vinzenz et al . , 2012 ) . Capping irreversibly removes a barbed end from the frontier . We assume there is a low capping rate k 0 within width w behind the leading edge , and a high rate k 1 further behind , leading to \ufb01laments being immediately capped outside this zone . k \u00f0 x ; t \u00de = 8 < : k 0 p \u00f0 t \u00de # w < x \u00f0 t \u00de < p \u00f0 t \u00de k 1 x \u00f0 t \u00de < p \u00f0 t \u00de # w 0 elsewhere ( 4 ) Actin \ufb01laments polymerizing against the plasma membrane exert a force that in turn sets the protrusion velocity . It has to balance the restricting force due to membrane tension . We assume an equal distribution of the total force among the pushing \ufb01laments and the absence of retrograde \ufb02ow as observed in keratocytes migrating on serum - coated glass coverslips . We use a force - velocity relation based on thermal \ufb02uctuations of the membrane and / or the \ufb01laments , as is appropriate for the Brownian ratchet model ( Lee et al . , 2001 ; Mogilner and Oster , 1996 ) of polymerization , but also for other established models like the tethered ratchet or end - tracking motors ( Dickinson , 2009 ) . This leads to the following rule for updating the position of the leading edge from small time steps : p \u00f0 t + D t \u00de = p \u00f0 t \u00de + v \u00f0 t \u00de D t v \u00f0 t \u00de = l 0 d e # d F \u00f0 t \u00de k B T D \u00f0 t \u00de ( 5 ) Here the maximum possible speed is equal to the maximal polymerization speed l 0 d of unobstructed \ufb01laments . By F \u00f0 t \u00de we denote the membrane force per leading edge length , D \u00f0 t \u00de the linear density of barbed ends and k B T is the thermal energy . The experiments with a rapid decrease of membrane tension and subsequent return to a steady protruding state are approximated by a stepwise decrease D F in force density at time t for a duration D t , followed by a stepwise increase back to original force den - sity F 0 . F \u00f0 t \u00de = ! F 0 # D F t < t < t + D t F 0 elsewhere ( 6 ) The initial barbed end positions are sampled from a uniform distribution within the polymerization zone . The orientations are sampled from a uniform distribution along a half circle of radius d . The density D \u00f0 t \u00de is updated before each time step and stored for the sampling of branching events and for updating the velocity of the leading edge . Then we iterate through all binding sites in the frontier , checking \ufb01rst for capping events , then branching , and \ufb01nally elongation . Finally the position of the membrane is updated , along with the elapsed time . The process is repeated until \ufb01nal simulation time T 0 . First the simulation is run for the above parameters except we set D F = 0 . The initialization is uniform . Beyond a time T 0 \" 10s a steady state in both \ufb01lament density and angular distribution with respect to the membrane is reached . Using the \ufb01nal state of a pre - vious run as an initial state for the next run , we simulate 20 runs each of time T 0 = 10 s . By recording all the history of barbed end move - ment during a run , we collect statistics at the end of each run . This way we end up with 20 data points per statistic , of which we return the mean and standard deviation to produce wild - type regime shown in Figures 6B \u2013 6D ( middle ) , Figures S7A \u2013 7C ( middle ) and Movie S5 ( \ufb01rst part ) . e5 Cell 171 , 1 \u2013 13 . e1 \u2013 e6 , September 21 , 2017 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 This unperturbed wild - type steady state with orientation peaks at \u00b1 35 ! is used as an initial condition for every run of the perturbed simulation . Setting D F = \u00b1 270pN m m # 1 and executing another 20 runs , collecting statistics at the end of each run obtains the results shown in Figures 6B \u2013 6D ( top and bottom ) , Figures S7A \u2013 7C ( top and bottom ) and Movie S5 ( second and third part ) . To analyze if our stochastic model would be able to reproduce the correlations shown in Figure 1 , we used the measured changes in projected cell area as input . Using the correlated cells shown in Figures 5 and S3 with known decreases in projected cell area and changes in network parameters , we scaled the changes in projected cell area by normalizing to the maximum value , subtracting a constant value and setting the mean to 300pN . The mean force of the \ufb01rst 2 s was used as constant tension for 5 s to let the simulation reach a steady state , before the force changes scaled from the projected area changes were used as input . The \ufb02uctuations in pro - jected area from the 21 migrating keratocytes used for Figure 1 were then used as force input and the resulting output parameters averaged and cross - correlated to generate Figure 7 . QUANTIFICATION AND STATISTICAL ANALYSIS MATLAB ( MathWorks ) and Prism ( GraphPad ) were used for statistical analysis . For live - cell imaging 21 and 7 individual cells were used for analysis of lifeact : GFP and membrane dye \ufb02uctuations , respectively . For membrane aspiration experiments 28 aspiration events in 13 individual cells were performed . For tomography experiments 4 wild - type tomography montages were used as well as 7 correlated shrinking events in 3 individual cells and one correlated aspiration event . Statistical parameters for individual exper - iments can be found within the \ufb01gure legends . A paired t test was used for changes in median angle at the density step ( Figure S5D ) . Pearson correlation was used for correlation analysis of measured distances from the leading edge to calculated distances ( Fig - ure S1F ) and for angle to density decrease analysis ( Figure S5E ) . A p value of < 0 . 05 was considered statistically signi\ufb01cant . Cell 171 , 1 \u2013 13 . e1 \u2013 e6 , September 21 , 2017 e6 Please cite this article in press as : Mueller et al . , Load Adaptation of Lamellipodial Actin Networks , Cell ( 2017 ) , http : / / dx . doi . org / 10 . 1016 / j . cell . 2017 . 07 . 051 Supplemental Figures Figure S1 . Details of Quanti\ufb01cation of Live Cell Imaging Parameters , Related to Figure 1 ( A ) Temporal \ufb02uctuations of lifeact : GFP signal , projected cell area and protrusion speed of the cell shown in Figure 1 are shown normalized from the maximum value to zero . The lifeact : GFP signal multiplied by the instantaneous speed is shown normalized from the maximum to zero as well . ( legend continued on next page ) ( B ) Temporal crosscorrelation functions of Area \u2013 lifeact : GFP intensity , Area - protrusion speed and lifeact : GFP intensity \u2013 Protrusion speed for 21 individual cells are shown in black . Averaged cross - correlation functions from a spline \ufb01t to the combined single - cell cross - correlation functions are shown in purple ( Area \u2013 lifeact : GFP intensity ) , green ( Area \u2013 Protrusion speed ) and yellow ( lifeact : GFP intensity \u2013 Protrusion speed ) . ( C ) Confocal images of a migrating keratocyte expressing lifeact : GFP were analyzed using consecutive 1 . 09 m m wide regions spanning the lamellipodium . ( D ) Resulting lifeact : GFP intensity maps are plotted against time . ( E ) Temporal cross correlation functions of region \u20181\u2019 at the very front of the cell to consecutive regions are plotted as a function of time lag . ( F ) Time lag between peaks of the curves shown in H multiplied by cell velocity and resulting distances are plotted against distance between analyzed regions . Measured mean and s . d . of twelve cells are shown together with a linear regression ( Pearson constant = 0 . 9997 , p = 0 . 0003 ) . ( legend on next page ) Figure S2 . Live Cell Imaging Controls with a Membrane Marker and actin : GFP , Related to Figure 1 ( A ) Example images of migrating keratocytes incubated with the membrane dye CellMask . ( B and C ) Temporal \ufb02uctuations of the resulting intensity maps are shown normalized as inFigure 1D ( B ) and as in Figure S1A ( C ) . Neither an ampli\ufb01cation of area \ufb02uctuations nor a correlation could be observed . ( D ) Resulting temporal cross correlation functions for eight individual cells are shown in black and averaged cross correlation in dark blue . The averaged cross correlation as mean with s . e . m . is shown in dark blue on the right . ( E ) Zebra\ufb01sh keratocytes transfected with actin : GFP were analyzed in the same way as shown before for lifeact : GFP . ( F ) Temporal cross correlation analysis area \u2013 GFP signal intensity showing peaks at time lag zero for transfected cells expressing actin : GFP and lifeact : GFP . Additionally cells generated from actin : GFP microinjected Zebra\ufb01sh embryos were analyzed and showed apositive cross correlation coef\ufb01cient at time lag zero . Plot shows averaged cross correlation and s . e . m . for seven cells for both actin : GFP transfection and mRNA injection . FigureS3 . MembraneTensionMeasurementbyTetherPullingandMembraneTensionManipulationbyChangingOsmoticPressure , Related to Figure 2 ( A ) Amigratingkeratocytewasimagedwithbright\ufb01eldmicroscopyandsimultaneouslyacoatedbeadcontrolledbyalasertweezerwasusedtopullamembrane tether . ( B ) The measured tether force decrease correlated with rapid decreases in projected cell area . ( C ) A lifeact : GFP expressing keratocyte was imaged by confocal microscopy and subjected to rapid changes in osmotic pressure of the medium by alternating additions of 0 . 2M sucrose and pure water . ( D ) lifeact : GFP signal is negatively correlated with the osmolarity of the medium . When the pressure increases upon addition of 0 . 2M sucrose , lifeact : GFP signal drops , which is reversible upon restoring a lower osmotic pressure by addition of pure water . Figure S4 . Actin Network Parameters for Individual Tomogram Montages of Keratocyte Lamellipodia , Related to Figures 3 and 4 ( A ) Filament density , barbed and pointed ends for individual cells shown in Figure 3 . Data for individual lamellipodia are shown in black together with average . ( B ) Filamentdensitiesgrowingattheindicatedanglefromthecellmembranein212nmdistancebinsofthesteadystatelamellipodiashowninFigure3 . Meanand s . e . m . are shown . ( C ) Filamentdensitiesgrowing attheindicated anglefromthecellmembranein212nmdistancebinsofthecorrelatedtomogramshowninFigure4 . Thedensities are shown as in Figure 4 beginning from toward the rear of the cell on the left until the leading edge on the right . ( D ) 5 . 5nm slice of a negatively stained electron tomogram montage showing the region marked by a red box in Figure 4A ( left ) and corresponding automatically tracked \ufb01laments shown color - coded according to their angle from the leading edge ( right ) . The spatial bins used for quanti\ufb01cation of network parameters in Figures 4 and S4C are also shown on the right . Figure S5 . Changes in Filament Parameters at a Decrease in Filament Density , Related to Figure 5 ( A and B ) Network structure analysis of seven lamellipodia tomograms in three correlated cells revealed consistent features at the point , where the actin density decreased due to the manipulation by the micropipette . Overview electron micrographs and 5 . 5nm tomogram slices are shown for all regions . ( C ) Averages of \ufb01lament density , barbed and pointed ends and median angle from the cell edge showed the consistent changes at the density step . ( D ) Mediananglesofallanalyzedregionsareplottedfor300nmspacebinsbeforeandafterthedecreaseinactin\ufb01lamentdensity ( right ) . ( Pairedttest , p = 0 . 0249 ) . ( E ) Normalized ratio of \ufb01laments after density step to \ufb01laments before density step is plotted against the \ufb01lament angle from the cell edge together with a linear regression curve . ( Pearson constant , r = ! 0 . 936 ) Mean and s . e . m . are shown on all graphs . ( legend on next page ) Figure S6 . Quanti\ufb01cation of Correlated Live Microscopy - Electron Tomography , Related to Figure 5 ( A ) A 5 . 5nm slice of the tomogram montage used for Figure 5 ( left ) is shown together with automated tracking results ( middle and right ) . ( B ) Close - upofregioncontainingthedropin\ufb01lamentdensityandchangeinactinarchitecturewith\ufb01lamentsshowningreen , barbedendsinredandpointedends in blue . Manually tracked \ufb01laments , barbed and pointed ends of a region including the density step are shown . ( C ) Filament density quanti\ufb01ed by manual and automated tracking , ilastik image analysis and the lifeact : GFP signal of the correlated cell shown in Figure 5 . ( D ) Filamentdensitiesin212nmdistancebinsgrowingatdifferentanglestowardthemembrane . ThedensitiesareshownasinFigure5beginningfromtowardthe rear of the cell on the left until the leading edge on the right . ( E ) 5 . 5nmsliceofnegativelystainedelectrontomogrammontageshowingregionmarkedwithredboxinFigure5A ( left ) andcorresponding automaticallytracked \ufb01laments shown color - coded according to their angle from the leading edge ( right ) . Additionally , the spatial bins used for quanti\ufb01cation of network parameters in Figures 5 and S6D are displayed on the left . The cell edge is seen on the right side . Figure S7 . Details of Stochastic Simulation , Related to Figures 6 and 7 The graphs in A \u2013 C are aligned as in Figure 6 . ( A ) Thetemporalchangesin\ufb01lamentsinthreedifferentanglebinsareshown . Thenumberof\ufb01lamentsintheindicatedanglebinsisconstantatgrowthatconstant externalforce . Intheforcedecreasescenario\ufb01lamentsgrowingathigheranglesarecappedpreferentiallyandthenetworkthinsout . Whenforceincreaseisused as an input , \ufb01laments in all angle bins increase . ( B ) Histograms of \ufb01lament density ( black ) and order parameter as de\ufb01ned in Figure 3 ( blue ) . Without perturbation the network architecture is dominated by \u00b1 35 \" peaks and the order parameter is consistently negative . Histograms of \ufb01lament density before ( black ) and after ( red ) decrease in F and order parameter ( blue ) showbiasedelimination of\ufb01laments growing athigher angle . This leads toachange from the \u00b1 35 \" architecture toonedominated by straight , 0 \" , \ufb01laments and a transiently positive order parameter . Filaments of all angles increase upon increase in force , which is also re\ufb02ected in a negative order parameter . ( C ) The migration parameters force ( the input parameter ) , actin density and protrusion speed \ufb02uctuate around steady state values during the unperturbed simulation . The force decrease causes an increase in protrusion speed and a decrease in actin density , with the reciprocal situation observed for the force increase . ( D ) Capping , elongation and branching rates used for the stochastic model . ( E ) Underlying lamellipodial feedback loop : Force , F , is set as an external parameter and decreases velocity , v , which is in turn increased by increased \ufb01lament density , D , as suggested by the elastic ratchet model of actin polymerization . Actin branching is modeled as a zeroth order process and therefore actin density decreases branching rate , b . The parts of the feedback loop linking velocity , v , to the \ufb01lament density , D , shown in this study are marked in red : Velocity , v , increasescappingrate , k , inanangle - dependentmannerandleadstoadecreasein\ufb01lamentdensity , D . Forallgraphsmeanands . d . foranaverageof20runsare shown , for details see STAR Methods .",
    "zwettler2020molecular": "ARTICLE Molecular resolution imaging by post - labeling expansion single - molecule localization microscopy ( Ex - SMLM ) Fabian U . Zwettler 1 , Sebastian Reinhard 1 , Davide Gambarotto 2 , Toby D . M . Bell 3 , Virginie Hamel 2 \u2709 , Paul Guichard 2 \u2709 & Markus Sauer 1 \u2709 Expansion microscopy ( ExM ) enables super - resolution \ufb02 uorescence imaging of physically expanded biological samples with conventional microscopes . By combining ExM with single - molecule localization microscopy ( SMLM ) it is potentially possible to approach the resolution of electron microscopy . However , current attempts to combine both methods remained challenging because of protein and \ufb02 uorophore loss during digestion or denaturation , gela - tion , and the incompatibility of expanded polyelectrolyte hydrogels with photoswitching buffers . Here we show that re - embedding of expanded hydrogels enables d STORM imaging of expanded samples and demonstrate that post - labeling ExM resolves the current limitations of super - resolution microscopy . Using microtubules as a reference structure and centrioles , we demonstrate that post - labeling Ex - SMLM preserves ultrastructural details , improves the labeling ef \ufb01 ciency and reduces the positional error arising from linking \ufb02 uorophores into the gel thus paving the way for super - resolution imaging of immunolabeled endogenous proteins with true molecular resolution . https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 OPEN 1 Department of Biotechnology and Biophysics , Biocenter , University of W\u00fcrzburg , Am Hubland , 97074 W\u00fcrzburg , Germany . 2 Department of Cell Biology , Sciences III , University of Geneva , Geneva , Switzerland . 3 School of Chemistry , Monash University , Clayton , VIC 3800 , Australia . \u2709 email : virginie . hamel @ unige . ch ; paul . guichard @ unige . ch ; m . sauer @ uni - wuerzburg . de NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications 1 1 2 3 4 5 6 7 89 0 ( ) : , ; B y linking a \ufb02 uorophore or a protein of interest into a dense , cross - linked network of a swellable polyelectrolyte hydrogel , biological specimen can be physically expanded allowing for magni \ufb01 ed imaging with subdiffraction - resolution on conventional microscopes . Since its introduction in 2015 1 , expansion microscopy ( ExM ) has shown impressive results including the magni \ufb01 ed visualization of pre - or postexpansion labeled proteins and RNAs with \ufb02 uorescent proteins ( FPs ) , antibodies , and oligonucleotides , respectively , in isolated organelles , cells , pathogens , tissues , and human clinical specimen 2 \u2013 4 . In addition , various protocols have been developed to anchor proteins or RNA into charged poly - acrylamide hydrogels 5 \u2013 8 . Using 2 . 5 % ( w / w ) acrylamide and 8 . 55 % sodium acrylate with 0 . 15 % ( w / w ) of the cross - linker N - N \u2032 - methylenebisacrylamide accomplishes a ~ 4 . 5x linear expansion of biological specimens . Decreasing the cross - linker concentration usually permits higher gel expansion factors of up to 10x but also increases proportionally the linkage error de \ufb01 ned by the af \ufb01 nity reagent , linker and \ufb02 uorophore and leads to greater gel instability 9 . It is also possible to expand samples in series enabling gel expansion factors of 20x and higher with a demonstration of 53x expansion of microtubules 10 . However , \ufb02 uorescence imaging of such greatly enlarged samples is complicated by the dilution of \ufb02 uorescent labels and dramatic increase in the physical separation between the \ufb02 uorophore and its target due to the linkage error . Nevertheless , ExM with lower expansion factors enables confocal diffraction - limited \ufb02 uorescence imaging with spatial resolutions comparable to that of super - resolution microscopy methods 11 , 12 . To further enhance the resolution , ExM has been combined with structured illumination microscopy ( SIM ) 13 , 14 and stimu - lated emission depletion ( STED ) microscopy 2 , 15 . By careful optimization of the expansion protocol U - ExM demonstrated that even ultrastructural details of multiprotein complexes such as centrioles can be truthfully preserved 2 . Combining ExM with SMLM methods ( Ex - SMLM ) can then potentially further improve the spatial resolution to enable true molecular resolution and bridge the gap to the electron microscopy regime . However , despite these apparent advantages , attempts to combine ExM with SMLM have remained rare and unoptimized due to several challenges 5 , 16 . There are two major determinants that control the resolution of SMLM , the localization precision and the localiza - tion density 11 , 12 . The localization precision remains unaltered by sample expansion and therefore allows achieving an improved resolution depending on the expansion factor . The localization density is arguably the more important determinant for SMLM on expanded samples . According to information theory , the required density of \ufb02 uorescent probes has to be suf \ufb01 ciently high to satisfy the Nyquist \u2013 Shannon sampling theorem 17 . At its most basic level , the theorem states that the mean distance between neighboring localized \ufb02 uorophores ( the sampling interval ) must be at least twice as \ufb01 ne as the desired resolution . In real samples , however , the relationship between localization density and reso - lution is far more complex 18 . Empirically , it seems that for a given resolution the distance between neighboring localizations should be signi \ufb01 cantly less than that indicated by a naive application of the Nyquist limit 19 . These considerations illustrate the challenges Ex - SMLM is confronted with . First , the \ufb02 uorophore density is considerably diluted in expanded samples 9 , 10 , which often results in unclear views of biological structures and complicates SMLM data interpretation . For example , a 4x expansion in three dimensions effectively lowers the labeling density 64 - fold . Second , addition of a thiol - containing phosphate - buffered saline ( PBS ) photoswitch - ing buffer as required for d STORM 20 , 21 to a swellable polyelec - trolyte hydrogel with hydrophilic ionic side groups results in substantial shrinking of the gel in the worst case down to its initial size . Finally , ExM protocols use free - radical polymerization to form polymers . However , free radicals also have the potential to react with the \ufb02 uorophores which can irreversibly destroy them 1 , 3 , 5 . Therefore , the \ufb02 uorophore density will be further diluted in ExM protocols that use pre - expansion labeling and consequently reduce the structural resolution . The extent of irreversible \ufb02 uorophore destruction during gelation varies across \ufb02 uorophores . Unfortunately , the best suited dyes for d STORM , the carbocyanine dyes Cy5 and Alexa Fluor 647 19 , 20 , are almost completely destroyed during gelation 1 , 3 , 5 Here , post - expansion labeling approaches ( post - labeling ExM ) offer acceptable solutions 2 , 7 , 8 , though they require preservation of protein epi - topes during expansion . An additional bene \ufb01 t of post - labeling ExM is improved epitope accessibility for antibodies and a reduction of the linkage error proportional to the expansion factor compared to pre - labeling ExM 2 . For instance , after 4x expansion , the immunolabeling linkage error of 17 . 5 nm de \ufb01 ned by the primary and secondary antibodies 22 , would reduce to 4 . 4 nm , which is the size of a tubulin monomer 23 . Combining SMLM with post - labeling ExM reduces the linkage error by the expansion factor and could thus enable \ufb02 uorescence imaging with molecular resolution . Here , we set out to develop post - labeling Ex - SMLM with organic \ufb02 uor - ophores with minimal linkage error . Results Re - embedding of expanded samples enables Ex - SMLM in photoswitching buffer . A major problem of Ex - SMLM is the shrinking of the expanded hydrogels in photoswitching buffer due to ionic interactions between ions of the buffer and the ionic side groups of the gel . Therefore , we re - embedded expanded charged hydrogels in an uncharged polyacrylamide gel as recently introduced for ExM of RNA 6 . We started using pre - labeling ExM in combination with standard immunostaining using unmodi \ufb01 ed primary and \ufb02 uorophore labeled secondary antibodies to realize Ex - SMLM ( Supplementary Fig . 1 ) . We used microtubules as reference structure to investigate the expansion factor , spatial resolution , structural distortions , and the labeling density . Microtubules are assembled from \u03b1 \u00df tubulin hetero - dimers , which stack head - to - tail into polar proto \ufb01 laments with a periodicity of 8 nm , with ~ 13 proto \ufb01 laments associating laterally in parallel to form a hollow , polar cylinder ( Fig . 1a ) 23 , 24 . As previously measured by transmission electron microscopy ( TEM ) , microtubules are hollow tubes with an outer diameter of 25 nm and 60 nm , respectively , after immunostaining with pri - mary and secondary antibodies 22 . This results in a linkage error de \ufb01 ned by the size of the primary and secondary antibody of 17 . 5 nm ( Fig . 1a ) . We used the proExM protocol , in which proteins are directly anchored to the hydrogel using the succinimidyl ester of 6 - ( ( acryloyl ) amino ) hexanoic acid ( AcX ) 5 . To minimize \ufb02 uorophore loss during gelation in pre - labeling ExM methods , we used the rhodamine derivative Alexa Fluor 532 , which retains ~ 50 % of its pre - gelation brightness after expansion 1 , 3 , 5 . To prevent shrinking of the hydrogel upon addition of photoswitching buffer , expanded hydrogels were re - embedded in acrylamide for serial staining of the expanded specimen 6 . Hydrogels were incubated twice in 10 % AA , 0 . 15 % bis - acrylamide , 0 . 05 % APS , 0 . 05 % TEMED in 5 mM Tris ( pH 8 . 9 ) for 30 min each and subsequently transferred onto coverslips functionalized with acrydite via glass silanization to minimize lateral drift of expanded samples ( Supplementary Fig . 1 ) . After polymerization of the re - embedding gel , hydrogels were immersed in photoswitching buffer containing 100 mM mercaptoethylamine ( MEA ) in PBS . The expansion factor was determined by comparing the post - expansion and post re - embedding \ufb02 uorescence images with ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 2 NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications pre - expansion \ufb02 uorescence images . The results showed a low distortion introduced by the re - embedding process and a reduction in gel size of ~ 20 % from ~ 3 . 9x before to ~ 3 . 1x after re - embedding ( Supplementary Figs . 2 and 3 ) . A caveat of imaging expanded samples is that super - resolution imaging methods , and in particular SMLM , are most effective when used on thin samples located within a few micrometers above the coverslip surface . However , expanded specimen can be easily located several tens of micrometers above the coverslip . In addition , expanded specimens are transparent because they consist largely of water . Hence , the use of oil - immersion objectives and total internal re \ufb02 ection \ufb02 uorescence ( TIRF ) microscopy as used in most SMLM applications to achieve a higher signal - to - background ratio is in this case not the best choice . Therefore , we decided to use a water - immersion objective and epi - illumination in all SMLM experi - ments . The corresponding d STORM images of pre - labeled expanded microtubules showed homogeneously labeled \ufb01 laments with some labeling gaps re \ufb02 ecting \ufb02 uorophore and protein loss during polymerization and enzymatic digestion , respectively ( Fig . 1b , c ) . In addition , we imaged unexpanded microtubules by d STORM under identical experimental conditions ( Fig . 1f , g ) . To examine the achieved spatial resolution , cross - sectional pro \ufb01 les of selected microtubule areas are often consulted 21 . If the two - dimensional ( 2D ) projection of the \ufb02 uorescence intensity distribution measured from microtubule \ufb01 laments show a bimodal distribution , the peak - to - peak distance can then be \ufb01 tted with a sum of two Gaussians and used as an estimate of the spatial resolution . To ensure an objective evaluation and comparison of the spatial resolution achieved , we developed \u201c Line Pro \ufb01 ler \u201d , an automated image processing software . Line Pro \ufb01 ler automatically evaluates potential regions of interest along \ufb01 lamentous structures to generate cross - sectional pro \ufb01 les that can be \ufb01 t by a sum of two Gaussians to determine the peak - to - peak distance between the sidewalls of the \ufb01 lamentous structure ( Supplementary Fig . 4 ) . In order to compare the experimentally measured peak - to - peak distances of different expansion protocols , we simulated trans - verse pro \ufb01 les of unexpanded and expanded microtubules using a cylindrical distribution function to describe the hollow annular structure of microtubules ( Fig . 2a and Supplementary Fig . 5 ) similar to the approach used recently for iterative expansion 10 . The resulting peak - to - peak distances were used to determine the molecular expansion factor of expanded immunolabeled microtubules considering the in \ufb02 uence of the label size on the broadening of the microtubule diameter ( Supplementary Table 1 and Supplementary Note 1 ) . \u03b1 - tubulin \u03b2 - tubulin 8 nm sab sab pab pab a b c proExM AI532 137 . 1 \u00b1 10 . 1 nm pre post 36 . 2 \u00b1 5 . 4 nm Unexpanded AI532 1 . 0 0 . 5 F r equen cy [ a . u ] 0 . 0 1 . 0 0 . 5 I n t en s i t y [ a . u . ] 0 . 0 \u2013 400 \u2013 200 \u2013 200 400 0 Distance [ nm ] 120 130 140 150 160 Peak - to - peak distance [ nm ] 1 . 0 0 . 5 F r equen cy [ a . u ] 0 . 0 1 . 0 0 . 5 I n t en s i t y [ a . u . ] 0 . 0 \u2013 200 200 0 Distance [ nm ] 20 25 30 35 40 45 5055 Peak - to - peak distance [ nm ] g f d e h i Fig . 1 Re - embedding enables Ex - d STORM . a Model of microtubules with an outer diameter of 25 nm stained with conventional primary ( pab ) and \ufb02 uorescently labeled secondary IgG antibodies ( sab ) results in a total diameter of 60 nm with a linkage error ( de \ufb01 ned by the size of the primary and secondary antibody ) of 17 . 5 nm 22 . b d STORM image of pre - labeled proExM expanded and re - embedded Cos - 7 cells stained with primary antibodies against \u03b1 - tubulin and secondary Alexa Fluor 532 conjugated antibodies ( Al532 ) . The small logo in the upper left corner symbolizes that microtubules have been immunolabeled before expansion ( pre - labeled ) . c Zoom in on highlighted region in ( b ) . d Averaged cross - sectional pro \ufb01 le of nine microtubule segments with a total length of 29 . 1 \u00b5 m ( segment lengths range from 2 . 1 - 4 . 5 \u00b5 m ) measured in two cells from 1 expanded sample . e Histogram of peak - to - peak distances with normalized normal distribution curve ( red ) determined by bi - Gaussian \ufb01 tting of the data analyzed in ( c ) with an average distance of 137 . 1 \u00b1 10 . 1 nm ( mean \u00b1 sd ) . The data were obtained from n = 9 microtubule segments in 2 cells from 1 expanded sample . f Unexpanded Cos - 7 cells labeled with an anti \u03b1 - tubulin primary antibody and Alexa Fluor 532 ( Al532 ) conjugated IgG secondary antibodies . The small logo in the upper left corner symbolizes that microtubules have been immunolabeled and not expanded . g Zoom in of the white boxed region in ( f ) . h Average intensity pro \ufb01 le of 35 microtubule segments with a length between 1 . 1 and 5 . 8 \u00b5 m ( mean = 2 . 0 \u00b5 m ) and a total length of 69 . 6 \u00b5 m analyzed in 12 d STORM images . i Histogram of peak - to - peak distances with normalized normal distribution curve ( red ) determined by bi - Gaussian \ufb01 tting of cross - sectional pro \ufb01 les of the analyzed microtubule segments in ( h ) with a mean peak - to - peak distance of 36 . 2 \u00b1 5 . 4 nm ( mean \u00b1 sd ) . The data were obtained from n = 35 microtubule segments in 12 cells and 3 independent experiments . Scale bars , 2 \u00b5 m ( b , f ) , 500 nm ( c , g ) . NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 ARTICLE NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications 3 Pre - labeling Ex - SMLM . d STORM images of unexpanded and expanded microtubules clearly showed a bimodal signal distribution along the \ufb01 laments , similar to that of previous super - resolution microscopy studies ( Fig . 1c , d and 1g , h ) 21 , 25 . When the cross - sectional pro \ufb01 le of unexpanded microtubules was \ufb01 t with a sum of two Gaussians , the peak - to - peak distance between the sidewalls showed a mean value of 36 . 2 \u00b1 5 . 4 nm ( mean \u00b1 s . d . ) analyzed over several microtubule \ufb01 lament segments ( Fig . 1i ) . This value is expected for the projection of a 25 nm inner diameter cylinder that has been broadened by primary and secondary antibodies on both sides by 17 . 5 nm 22 ( Fig . 1a ) and corresponds well to the simulated peak - to - peak distance of 32 . 0 nm for unexpanded microtubules ( Fig . 2a and Supplementary Note 1 ) . The mean peak - to - peak dis - tance of proExM treated and expanded microtubules was deter - mined to 137 . 1 \u00b1 10 . 1 nm ( mean \u00b1 s . d . ) ( Fig . 1e ) . This value corresponds to an expansion factor of 3 . 1x determined from a b c g pre post pre post Antibodies unexpanded ( 32 . 0 nm ) Antibodies pre - labeling ( proExM / GA ) ( 143 . 0 nm ) DNA Cy5 unexpanded ( 41 . 5 nm ) DNA Cy5 expanded ( 202 . 0 nm ) DNA AI532 expanded ( 226 . 5 nm ) 1 . 0 0 . 5 0 . 0 \u2013 150 \u2013 100 \u2013 50 0 Distance [ nm ] I n t en s i t y [ a . u . ] 50 100 150 f j k 133 . 8 \u00b1 13 . 2 nm 1 . 0 0 . 5 F r equen cy [ a . u ] 0 . 0 1 . 0 0 . 5 I n t en s i t y [ a . u . ] 0 . 0 \u2013 400 \u2013 200 \u2013 200 400 0 Distance [ nm ] 120 110 100 130140150 170 160 Peak - to - peak distance [ nm ] d e 1 . 0 0 . 5 F r equen cy [ a . u ] 0 . 0 1 . 0 0 . 5 I n t en s i t y [ a . u . ] 0 . 0 \u2013 400 \u2013 200 \u2013 200 400 0 Distance [ nm ] 195 180 165 210 225 240 Peak - to - peak distance [ nm ] 133 . 8 \u00b1 13 . 2 nm 1 . 0 0 . 5 F r equen cy [ a . u ] 0 . 0 1 . 0 0 . 5 I n t e n s i t y [ a . u . ] 0 . 0 \u2013 200 200 0 Distance [ nm ] Peak - to - peak distance [ nm ] h i Antibody - AI532 ( GA ) Unexpanded DNA - Cy5 DNA - Cy5 DNA - AI532 30 35 40 45 50 55 o n 201 . 0 \u00b1 12 . 9 nm p q 1 . 0 0 . 5 F r equen cy [ a . u ] 0 . 0 1 . 0 0 . 5 I n t en s i t y [ a . u . ] 0 . 0 \u2013 400 \u2013 200 \u2013 200 400 0 Distance [ nm ] 210 195 180 225 240 255 270 Peak - to - peak distance [ nm ] 226 . 7 \u00b1 15 . 3 nm l m Fig . 2 Pre - labeling Ex - d STORM . a Simulated intensity pro \ufb01 les using a cylindrical distribution function to describe unexpanded or 3 . 2x expanded immunostained microtubules ( labeled with IgG antibodies or DNA modi \ufb01 ed IgG antibodies pre - expansion ) and resulting peak - to - peak distances of the cross - sectional pro \ufb01 les . b d STORM image of expanded and re - embedded \u03b1 - and \u03b2 - tubulin pre - labeled with secondary Alexa Fluor 532 IgG antibodies ( Al532 ) using the MA - NHS / GA method 6 , i . e . antibodies are cross - linked with glutaraldehyde ( GA ) into the hydrogel ( Antibody - Al532 ( GA ) ) . c Zoom in of white boxed region in ( b ) . d Averaged cross - sectional pro \ufb01 le of 8 microtubule segments with a length between 1 . 5 \u2013 6 . 4 \u00b5 m and 28 . 6 \u00b5 m in total measured in 4 expanded cells . e Histogram of peak - to - peak distance distribution with normalized normal curve ( red ) of microtubule segments analyzed in ( d ) at n = 8 microtubule segments in 4 cells from 1 expansion experiment with a mean distance of 133 . 8 \u00b1 13 . 2 nm ( mean \u00b1 sd ) . f Unexpanded d STORM image of ssDNA - Cy5 secondary antibody hybridized with Cy5 bearing oligonucleotides pre - expansion ( DNA - Cy5 protocol ) . g Magni \ufb01 ed view of white boxed region in ( f ) . h Average cross - sectional pro \ufb01 le of 7 microtubule segments with a length between 1 . 0 \u2013 1 . 8 \u00b5 m and 8 . 7 \u00b5 m in total . i Histogram of peak - to - peak distances with normalized normal distribution curve ( red ) of the data analyzed in ( h ) along n = 7 microtubule segments in 2 cells from 1 experiment with a mean distance of 43 . 9 \u00b1 3 . 7 nm ( mean \u00b1 sd ) . j Expanded d STORM image of microtubules labeled with \u03b1 - tubulin and dsDNA ( DNA - Al532 ) conjugated secondary antibodies exhibiting a methacryloyl group to crosslink the DNA with \ufb02 uorophores pre - expansion into the hydrogel ( original ExM trifunctional label concept ) 1 . k Zoom - in of white boxed region in ( j ) . l Average intensity pro \ufb01 le of 26 microtubule segments with a length of 2 . 4 \u2013 10 . 7 \u00b5 m and 118 . 6 \u00b5 m in total . m Histogram of peak - to - peak distances with normalized normal distribution curve ( red ) determined from n = 26 microtubule segments in 4 cells from 1 expanded sample showing a mean distance of 226 . 7\u00b1 15 . 3 nm ( mean \u00b1 sd ) . n d STORM image of \u03b1 - and \u03b2 - tubulin expanded according to the DNA - Cy5 protocol strategy with labels at Cy5 - bearing oligonucleotides introduced post - re - embedding . o Zoom in of white boxed region in ( n ) . p Average intensity pro \ufb01 le of 15 microtubule segments with a length between 1 . 6 \u2013 25 . 1 \u00b5 m and a total length of 126 . 0 \u00b5 m in 1 expanded sample . q Histogram of peak - to - peak distances with normalized normal distribution curve ( red ) determined by \ufb01 tting the cross - sectional pro \ufb01 les analyzed in ( p ) along n = 22 microtubule segments in 4 cells from 1 expanded sample showing a mean distance of 201 . 0 \u00b1 12 . 9 nm ( mean \u00b1 sd ) . The small logos in the upper left corner symbolize the labeling method , e . g . pre - and post - immunolabeled with or without DNA - linker , respectively . Scale bars , 2 \u00b5 m ( b , f , j , n ) , 500 nm ( c , g , k , o ) . ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 4 NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications simulation of expanded microtubules pre - labeled with primary and secondary IgG antibodies ( Supplementary Table 1 ) . We next used the post - stain linker - group functionalization method ( MA - NHS / GA method ) 7 as an alternative pre - labeling Ex - SMLM method . Here , the entire sample was functionalized with polymer - linking groups after conventional immunostaining . The resulting d STORM images showed a peak - to - peak distance between the microtubule sidewalls of 133 . 8 \u00b1 13 . 2 nm ( mean \u00b1 s . d . ) ( Fig . 2b \u2013 e ) corresponding to a simulated expansion factor of 3 . 0x ( Supplementary Table 1 ) . The determined peak - to - peak distance is in good agreement with the peak - to - peak distance determined from proExM expanded microtubules ( Fig . 1b \u2013 e ) . Variations in the measured peak - to - peak distances can be well explained by varying initial expansion factors of hydrogels which are typically in the range of ~ 4 . 0 \u2013 4 . 5x for the used ExM gel composition . Considering a ~ 20 % reduction in gel size caused by re - embedding of the hydrogel , an ultimate expansion factor of ~ 3 . 1 \u2013 3 . 6x can be expected which \ufb01 ts well with the determined molecular expansion factors . Next , we tested the original ExM protocol with trifunctional DNA - modi \ufb01 ed secondary antibodies 1 , which can be labeled with dye - functionalized complementary oligonucleotides that contain an acrydite linker modi \ufb01 cation . Alternatively , anti - bodies can be modi \ufb01 ed with a single stranded DNA that is incorporated directly into the hydrogel . Antisense dye - labeled oligonucleotides can then be hybridized after re - embedding of the hydrogel , which enables the use of \ufb02 uorophores that would not survive the radical polymerization process . Since the linkage error is mainly determined by the IgG antibodies and the 40 bases long DNA strand ( Supplementary Table 2 ) both methods still belong to the pre - labeling Ex - SMLM method ( Supplementary Fig . 1 ) . First , we tested the approach on unexpanded microtubules and obtained peak - to - peak distances of 43 . 9 \u00b1 3 . 7 nm ( mean \u00b1 s . d . ) ( Fig . 2f \u2013 i ) and 37 . 0 \u00b1 4 . 8 nm ( mean \u00b1 s . d . ) ( Supplementary Fig . 6 ) for labeling with Cy5 - and Alexa Fluor 532 - modi \ufb01 ed oligonu - cleotides , respectively . These values are in good agreement with the theoretically expected value of 41 . 5 nm for immunolabeling with 42 bases long trifunctional oligonucleotide - modi \ufb01 ed sec - ondary antibodies ( Fig . 2a ) . Due to the additional modi \ufb01 cation of the secondary antibodies , the peak - to - peak distances should be a few nanometers larger than the value measured for standard immunolabeled microtubules of 36 . 2 \u00b1 5 . 4 nm ( mean \u00b1 s . d . ) ( Fig . 1f \u2013 i ) . If the oligonucleotide - modi \ufb01 ed secondary antibodies are labeled with complementary Alexa Fluor 532 - modi \ufb01 ed oligonucleotides prior to expansion , we measured a peak - to - peak distance of 226 . 7 \u00b1 15 . 3 nm ( mean \u00b1 s . d . ) from d STORM images ( Fig . 2j \u2013 m ) . Since Cy5 does not survive gelation 1 , 3 , 5 , we performed an additional experiment labeling the oligonucleotide - modi \ufb01 ed secondary anti - bodies after expansion with complementary Cy5 - modi \ufb01 ed oligo - nucleotides , performed d STORM imaging in photoswitching buffer and determined a slightly shorter peak - to - peak distance of 201 . 0 \u00b1 9 . 3 nm ( mean \u00b1 s . d . ) ( Fig . 2n \u2013 q ) . Both values are in excellent agreement with the theoretical peak - to - peak distance of 226 . 5 nm and 202 nm , respectively ( Fig . 2a ) , simulated for 3 . 2x expanded microtubules taking into account the length of the 42 base pair trifunctional oligonucleotide , the position of \ufb02 uoro - phores within the DNA strand and its spatial orientation ( Supplementary Fig . 7 and Supplementary Note 1 ) . The slightly shorter peak - to - peak distance measured in the Cy5 - experiment where the dye - labeled complementary strand was hybridized after expansion can be explained most likely by coiling of the single - stranded trifunctional oligonucleotide during gelation ( Supple - mentary Fig . 7 ) . These results indicate that Ex - SMLM can resolve linker differences of 42 DNA base pairs ( corresponding to ~ 14 . 3 nm ) and , interestingly conformational differences between single and double - stranded DNA linkers . Noteworthy is that the total size of an expanded sample is not only determined by the biomolecule of interest , e . g . microtubules , but also by the \ufb02 uorescent probe , e . g . primary and secondary antibodies , used to label the biomolecule of interest . Unexpanded , microtubules labeled with primary and secondary IgG antibodies exhibit a diameter of ~ 60 nm with a linkage error ( de \ufb01 ned by the size of the primary and secondary antibody ) of 17 . 5 nm 22 . For example , for 3 . 3x expansion this translates into a microtubule diameter of 3 . 3 \u00d7 25 nm = 82 . 5 nm whereas the diameter of the immunolabeled microtubule is substantially broadened to ~ 198 nm because of the linkage error of 3 . 3 \u00d7 17 . 5 nm = 57 . 75 nm introduced by the primary and secondary antibody that has to be added to both sides of the microtubule ( Supplementary Fig . 5 ) . In other words , even though SMLM achieves high localization precisions 12 , 13 , a linkage error of > 50 nm undoes much , or even all , of the gain in resolution . Pre - versus post - labeling Ex - SMLM . In order to reduce the linkage error , we next explored post - labeling Ex - SMLM . It has been shown that the \ufb02 uorescence signals from some genetically encoded FPs as well as conventional \ufb02 uorescently labeled sec - ondary antibodies and streptavidin that are directly anchored to the gel are at least partially preserved by proExM even when subjected to the strong nonspeci \ufb01 c proteolytic digestion used in the original ExM protocol 1 , 5 . Therefore , we anticipated that pro - tein epitopes might survive the proExM protocol 26 . To compare the labeling density of pre - and post - labeling Ex - SMLM we immunostained microtubules with Alexa Fluor 532 before and additionally after expansion . Intriguingly , combining pre - with post - labeling resulted in a substantial shortening of the average peak - to - peak distance of the sidewalls of microtubules to 79 . 5 \u00b1 6 . 6 nm ( mean \u00b1 s . d . ) determined from d STORM images ( Fig . 3 ) . We speculated that the protease digestion step may destroy a large fraction of the pre - labeled antibody complexes but to our surprise , the majority of tubulin epitopes survives this critical step . Toge - ther with the increased accessibility of tubulin epitopes for pri - mary antibodies and primary antibody epitopes for secondary antibodies after expansion this results in peak - to - peak distances undistinguishable from solely post - labeled microtubules . To examine more quantitatively epitope survival and increased epitope accessibilities , we simulated the cross - sectional pro \ufb01 les expected for pre - and post - labeled microtubules . Here we assumed a ~ 10 - fold signal dilution ( 3 . 2 2 ) for the 2D projection of the \ufb02 uorescence signals of 3 . 2x expanded pre - labeled antibodies ( Fig . 3e and Supplementary Fig . 5 ) . Hence , the cross - sectional microtubule pro \ufb01 les show the superposition of the pro \ufb01 le calculated for the 3 . 2x expansion of 25 nm diameter microtubules post - immunolabeled and 60 nm diameter micro - tubules pre - immunolabeled . The resulting superposition pro \ufb01 le shows a peak - to - peak distance of 79 . 5 nm ( Fig . 3e ) emphasizing the advantage of post - labeling Ex - SMLM . Post - labeling Ex - SMLM using the proExM protocol 5 provides an improved labeling ef \ufb01 ciency and a reduced linkage error . In fact , the immunolabeling linkage error of ~ 58 nm for pre - labeling reduces to ~ 5 nm for post - labeling considering a 3 . 2x expansion factor and thus improves the effective achievable resolution ( Supplementary Fig . 8 ) . Therefore , d STORM images of Alexa Fluor 532 labeled microtubules clearly revealed the hollow cylinder of micro - tubules ( Fig . 3c ) using a water - immersion objective and epi - illumination , similar to recently published results obtained by DNA - PAINT TIRF microscopy and experimental point spread function \ufb01 tting 27 . The average distance between the sidewalls of NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 ARTICLE NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications 5 the xz - projection of a 6 . 5 \u00b5m long microtubule \ufb01 lament was determined to 81 . 2 nm ( Fig . 3g \u2013 h ) highlighting the high spatial resolution of pre - labeling 3D Ex - d STORM . Post - labeling Ex - SMLM of centrioles . Motivated by the results , we set out to explore the molecular organization of centriole organelles by Ex - SMLM . We used isolated Chlamydomona s cen - trioles , which have a characteristic 9 - fold microtubule triplet - based symmetry , forming a polarized cylinder ~ 500 nm long and ~ 220 nm wide 28 ( Supplementary Fig . 9 ) . Recently 2 , U - ExM has been developed as an extension of ExM that allows for near - native expansion of organelles and multiprotein complexes and visuali - zation of preserved ultrastructures by optical microscopy . When combined with STED microscopy , details of the ultrastructural organization of isolated centrioles such as the 9 - fold symmetry and centriolar chirality could be visualized 2 . Advantageously , U - ExM uses post - labeling to improve the epitope accessibility after expansion . Here , we used U - ExM treated centrioles in combination with post - labeling with Alexa Fluor 647 secondary antibodies to enable d STORM imaging , which has previously be impossible due to shrinking of expanded hydrogels in photoswitching buffer . Therefore , samples were re - embedded and transferred onto coverslips functionalized with acrydite via glass silanization to minimize lateral drift . This allowed us to perform post - labeling 3D Ex - d STORM on ~ 3 . 4x expanded centrioles ( Fig . 4a \u2013 d and Supplementary Fig . 10 ) . Alternatively , we used the spontaneously blinking Si - rhodamine dye HMSiR 29 that enables SMLM in the absence of photoswitching buffer and does thus not require re - embedding . Using double - deionized water , we achieved a molecular expansion factor of ~ 4x ( Fig . 4d \u2013 f and Supplementary Fig . 9 ) . Unfortunately , since the pH of double - deionized water is below 7 . 0 , HMSiR does not exhibit optimal blinking characteristics 29 . Addition of PBS buffer , pH 7 . 4 improved the blinking characteristics of HMSiR but reduced the expansion factor to ~ 2x , which limits the spatial resolution of the SMLM experiments ( Fig . 4d , g ) . In contrast to 3D d STORM images of unexpanded centrioles ( Fig . 4h ) post - labeling 3D Ex - SMLM clearly visualized the centriole as a bundle of nine microtubule triplets . SMLM images of expanded isolated Chlamydomonas centrioles showed the 9 - fold symmetry of the procentrioles ( Fig . 4b , f ) with tubulin diameters of ~ 220 nm in agreement with previous studies 2 , 30 . Even in side views of centrioles imaged by 3D Ex - d STORM the neighboring microtubule triplets are clearly separated ( Fig . 4c ) . Furthermore , 3D Ex - d STORM allowed us to resolve ring - like sub - structures of centrioles indicating hollow cylinders of microtubule triplets ( Supplementary Fig . 11 ) . According to these results , re - embedding of the sample and d STORM in photoswitching buffer provides currently the best Ex - SMLM performance . Since microtubule triplets separated by 15 \u2013 20 nm 30 are very well resolved in the expanded images post - labeling Ex - SMLM exhibits a spatial resolution that is way below 15 \u2013 20 nm reaching the structural resolution required to resolve the molecular architecture of centrioles . Discussion Electron microscopy has been the only viable method able to reveal the ultrastructure of organelles and molecular complexes for decades because of the diffraction limit of optical microscopy . Super - resolution microscopy offers up to ~ 10x higher resolution than conventional diffraction - limited microscopy 11 , 12 . Improved super - resolution microscopy methods can now localize single emitters with a precision of a few nanometers 31 \u2013 33 , but limitations in labeling ef \ufb01 ciency and linkage error have thus far prevented the translation of high localization precision into sub - 10 - nm spatial resolution . Therefore , the spatial resolution provided by all these inventive methods is currently still too low a b e f g h d c pre 630 0 V iV i i ii ii iV / V iii z [ nm ] y x z x z x post 79 . 5 \u00b1 6 . 6 nm 81 . 2 nm 1 . 0 0 . 5 F r equen cy [ a . u ] 0 . 0 I n t en s i t y [ a . u . ] 1 . 0 0 . 5 0 . 0 \u2013 200 \u2013 100 100 200 0 Distance [ nm ] 1 . 0 0 . 5 I n t en s i t y [ a . u . ] 0 . 0 \u2013 150 \u2013 100 \u2013 50 100 50 150 0 Distance [ nm ] 70 75 80 85 90 95100 65 60Peak - to - peak distance [ nm ] Fig . 3 3D post - labeling Ex - d STORM . SMLM image of re - embedded and post - expansion labeled microtubules . a 3D d STORM image of re - embedded Cos - 7 cells expanded according to the Protein - Retention protocol ( proExM ) 4 pre - labeled with anti \u03b1 - and \u03b2 - tubulin antibodies and additionally post - labeled with anti \u03b1 - tubulin . The secondary antibodies were labeled with Alexa Fluor 532 ( proExM Al532 ) . The small logo in the upper left corner symbolizes the labeling method , e . g . pre - and post - immunolabeling with Al532 secondary antibodies . b Magni \ufb01 ed view of highlighted region in ( a ) . c xz - side view cross - sections ( white lines ) ( i ) and ( ii ) shown in ( b ) revealing the hollow structure of microtubules . d Magni \ufb01 ed view of highlighted region ( white box ) in ( b ) . Since post - labeling dominates the signal , the method is termed proExM Al532 ( post - labeled ) . e Averaged cross - sectional pro \ufb01 le ( blue ) of 11 analyzed microtubule segments along a total of 28 . 2 \u00b5 m \ufb01 lament ( 2 . 1 \u2013 5 . 5 \u00b5 m segments ) of one expanded sample . The simulated cross - sectional pro \ufb01 le for 3 . 2x proExM expanded pre - and post - labeled microtubule assuming a pre - to post - labeling ratio of 0 . 1 is shown in red . f Histogram of peak - to - peak distances with normalized normal curve ( red ) of \ufb01 tted pro \ufb01 les analyzed in ( e ) with an average distance of 79 . 5 \u00b1 6 . 6 nm ( mean \u00b1 sd ) analyzed along n = 11 microtubule segments in 2 cells from 1 expanded sample . g Image projection of the xz - axes averaged along two microtubule \ufb01 laments ( iv ) and ( v ) shown in ( b ) ( red dotted lines ) using the \u201c z projection analysis \u201d of the software \u201c Line Pro \ufb01 ler \u201d . h Cross - sectional pro \ufb01 le ( blue dots ) of the xz - projection shown in ( g ) . Using a bi - Gaussian \ufb01 t ( red ) the peak - to - peak distance is determined to 81 . 2 nm . Scale bars , 10 \u00b5 m ( a ) , 5 \u00b5 m ( b ) , 1 \u00b5 m ( c ) , 500 nm ( d ) , 100 nm ( g ) . ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 6 NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications to unravel the composition and molecular architecture of protein complexes or dense protein networks . Expansion microscopy ( ExM ) represents an alternative approach to bypass the diffrac - tion barrier . By linking a protein of interest into a cross - linked network of a swellable polyelectrolyte hydrogel , biological speci - mens can be physically expanded allowing for sub - diffraction resolution imaging on conventional microscopes 1 \u2013 10 . However , even in combination with super - resolution microscopy techni - ques , spatial resolutions below ~ 20 nm have so far proven to be very dif \ufb01 cult to achieve by ExM 16 . Here , we have shown that re - embedding of expanded hydrogels enables the use of standard photoswitching buffers and d STORM imaging of ~ 3 . 2x expanded samples . Our results demonstrate that post - labeling ExM using the proExM protocol 5 or U - ExM 2 provides solutions for the two major limiting problems of improved super - resolution micro - scopy , the labeling ef \ufb01 ciency and linkage error . First , as shown for microtubules , expansion of the sample increases the epitope accessibility and thus the labeling ef \ufb01 ciency . Comparison experiments demonstrated that post - labeling outperforms pre - labeling several times in this regard ( Fig . 3 ) . Second , post - labeling ExM reduces the linkage error proportionally to the expansion factor . Hence , post - immunolabeling of 3 . 2x expanded micro - tubules reduces the linkage error from 17 . 5 nm 22 to ~ 5 nm ( Fig . 3 ) . Since the linkage error also in \ufb02 uences the localization accuracy and thus the effective achievable resolution ( Supple - mentary Figs . 8 and 9 ) 34 , 35 our \ufb01 ndings are highly relevant . Very recently 36 , 37 , trifunctional linkers have been introduced that are inert to polymerization , digestion and denaturation , and enable direct covalent linking of target molecules and functional groups to the hydrogel . Therefore , trifunctional linkers can retain a high number of labels and \ufb02 uorescence molecules available for post - expansion imaging . However , since the target molecules are labeled with primary and secondary antibodies or enzymatic tags ( e . g . SNAP - tags ) functionalized with the trifunctional anchor before expansion , linkage errors remain . The improved labeling ef \ufb01 ciency of post - labeling Ex - d STORM in combination with small ( 1 . 5 \u00d7 2 . 5 nm ) camelid antibodies ( \u201c nanobodies \u201d ) 38 , 39 and 10 \u2212 20x expansion factors 9 , 10 can thus pave the way for true molecular resolution imaging of endogenous proteins with 1 \u2013 5 nm spatial resolution . On the other hand , at such a small length scale , distortions of the structure may occur . To realize more homogeneous gel network structures , a new gelation method based on a highly homogeneous expansion microscopy polymer composed of tetrahedron - like monomers has been introduced 40 . The new tetra - gel polymer chemistry may introduce fewer spatial errors than earlier versions , and enable molecular resolution post - labeling Ex - d STORM with reduced distortion . Nevertheless , already ~ 3x Ex - SMLM can resolve small linker length and con - formational differences between labeling approaches as shown here for oligonucleotide - functionalized secondary antibodies ( Fig . 2 ) . In addition , we have shown that post - labeling 3D Ex - d STORM exhibits excellent structure preservation and already 0 . 8 25 \u2013 75 % IQR 1 . 5IQR Mean Median ( IQR - Interquartile range ) p < 0 . 02 D i a m e t e r [ \u03bc m ] 0 . 7 0 . 6 0 . 5 0 . 4 0 . 3 AI647MEA HMSiRPBS HMSiRWater a b c d e f g h Fig . 4 Ex - SMLM of U - ExM expanded centrioles . a - c 3D d STORM image of U - ExM expanded and re - embedded Chlamydomonas centrioles stained post re - embedding with anti \u03b1 - tubulin primary antibody and Alexa Fluor 647 conjugated secondary antibodies measured in MEA buffer . b Zoom - in on highlighted region in ( a ) revealing the 9 - fold symmetry of the shown procentriole . c Side view of two mature centrioles with clearly separated triplets . The inlet shows the cross - sectional pro \ufb01 le along the centriole ( white box ) showing \ufb01 ve distinct peaks of microtubule triples ( marked with arrow heads ) . d Comparison of the diameters determined from expanded centrioles measured using different protocols ( re - embedded and labeled with Alexa Fluor 647 , and imaged in MEA photoswitching buffer , labeled with HMSiR 647 and imaged in double - deionized water or in pH optimized PBS ( 1x ) buffer with pH 7 . 4 ) . Mean values are 657 \u00b1 90 nm ( mean \u00b1 sd ) for Alexa Fluor 647 in MEA buffer ( n = 12 centrioles ) , 428 \u00b1 74 nm ( mean \u00b1 sd ) for HMSiR in PBS ( n = 7 centrioles ) , and 750 \u00b1 34 nm ( mean \u00b1 sd ) for HMSiR 647 in water ( n = 8 centrioles ) . Data from n = 2 independent experiments for each condition . Divided by the previously analyzed diameter of \u03b1 - tubulin labeled centriole expansion factors translates into ~ 3 . 4x , ~ 2 . 2x , and ~ 3 . 9x for expanded centrioles labeled with Alexa Fluor 647 in MEA buffer , HMSiR in PBS ( 1x ) , and HMSiR 647 in water , respectively . Statistical signi \ufb01 cance was assessed by one - way ANOVA : the mean values of the diameters are signi \ufb01 cantly different with p < 0 . 02 ( F = 3 . 80 ) ( * P \u2264 0 . 05 , * * P \u2264 0 . 01 ) . e \u2013 g 2D d STORM image of U - ExM expanded centrioles labeled with HMSiR 647 imaged in water ( e \u2013 f ) or PBS ( 1x ) ( g ) f Zoom - in on highlighted region in ( e ) . h 3D d STORM image of unexpanded isolated Chlamydomonas centrioles immunostained with antibodies against glutamylated tubulin and Alexa Fluor 647 conjugated secondary antibodies . Scale bars , 1 \u00b5 m ( a , e ) , 500 nm ( b , f , g ) , 1 . 5 \u00b5 m ( c ) , 250 nm ( h ) . NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 ARTICLE NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications 7 3 . 4x expansion using standard protocols can provide a suf \ufb01 cient structural resolution to resolve details of the molecular archi - tecture of centrioles ( Fig . 4 and Supplementary Fig . 11 ) . Methods Reagents . Acetic acid ( A6283 , Sigma ) , Acrylamide ( AA , 40 % , A4058 , Sigma ) , Acryloyl - X , SE , 6 - ( ( acryloyl ) amino ) hexanoic Acid , Succinimidyl Ester ( A20770 , Thermo Fisher ) , Agarose ( A9539 , Sigma ) , Ammonium persulfate ( APS , A3678 , Sigma ) , Bind Silane ( GE17 - 13330 - 01 , Sigma ) , Bovine Serum Albumin ( BSA , A2153 , Sigma ) , Cysteamine hydrochloride ( MEA , 6500 , Sigma ) , Dextran sulfate ( D8906 , Sigma ) , DMEM / HAM \u2019 s F12 with L - glutamine ( Sigma , D8062 ) , Ethanol ( absolute , \u2265 99 . 8 % , 32205 , Sigma ) , Ethylenediaminetetraacetic acid ( E1644 , Sigma ) , Ethylene glycol - bis ( 2 - aminoethylether ) - N , N , N \u2032 , N \u2032 - tetraacetic acid ( EGTA , 03777 , Sigma ) , 10 % FBS ( Sigma , F7524 ) , Formaldehyde ( FA , 36 . 5 - 38 % , F8775 , Sigma ) , Glutaraldehyde ( GA , 25 % , G5882 , Sigma ) , Guanidine hydrochloride ( 50933 , Sigma ) , 2 - ( N - morpholino ) ethanesulfonic acid ( MES , M3671 , Sigma ) , N , N \u2032 - methylen - bisacrylamide ( BIS , 2 % , M1533 , Sigma ) , N , N , N \u2032 , N \u2032 - Tetramethylethylenediamine ( TEMED , T7024 , Sigma ) , Poly - D - lysine hydropromide ( P6407 , Sigma ) , Polyoxyethylene ( 20 ) sorbitan monolaurate solution ( Tween - 20 , 10 % , 93774 , Sigma ) , Potassium hydroxide ( P5958 , Sigma ) . Proteinase K ( P4850 , Sigma ) , Saline sodium citrate buffer ( SSC , 20x , 15557 , Thermo Fisher ) , Sodium acrylate ( SA , 97 - 99 % , 408220 , Sigma ) , Sodium chloride ( NaCl , S7653 , Sigma ) , Sodium dodecyl sulfate ( SDS , L3771 , Sigma ) , streptomycin ( 0 . 1 mg / ml ) ( Sigma , R8758 ) , Tris base ( T6066 , Sigma ) , Triton X - 100 Surfact - Amps Detergent Solution ( 10 % ( w / v ) , 28313 , Thermo Fisher ) , Yeast tRNA ( AM7119 , Thermo Fisher ) . Antibodies and labeling reagents . Rabbit anti \u03b1 - tubulin antibody ( ab18251 , abcam ) , Mouse anti \u03b2 - tubulin antibody ( T8328 , Sigma ) , Mouse anti poly - glutamylated tubulin , mAb ( GT335 ) ( Adipogen , AG - 20B - 0020 ) , Alexa Fluor 647 F ( ab \u2018 ) 2 of goat anti rabbit IgG ( A - 21246 , Thermo Fisher ) , Alexa Fluor 647 F ( ab \u2018 ) 2 of goat anti mouse IgG ( A - 21235 , Thermo Fisher , Al532 - Goat anti Rabbit IgG ( H + L ) ( A - 11009 , Thermo Fisher ) , Al532 - Goat anti Rabbit IgG ( H + L ) ( A - 11002 , Thermo Fisher ) , HMSiR 647 ( A208 - 01 , MoBiTec ) conjugated to goat anti rabbit IgG F ( ab \u2018 ) 2 ( SAB3700946 , Sigma ) , TetraSpeck Microspheres ( 0 . 1\u00b5m , T27279 , Thermo Fisher ) . Chlamydomonas reinhardtii centriole isolation . Centrioles were isolated from the cell wall - less Chlamydomonas reinhardtii strain CW15 by centrifugation at 600 g for 10min in 50 ml conical tubes 41 . Isolated centrioles were thawed on ice and diluted with cold K - Pipes 10mM pH 7 . 2 . Centrioles were then loaded in a 15ml Corex tube with a homemade adaptor and concentrator , and spun onto a 12 mm Poly - D - lysine coated coverslip through centrifugation at 10 , 000 g for 10 min with a JS - 13 . 1 swinging bucket rotor ( Beckman ) at 4\u00b0C . Coverslips were then processed for immunostaining and expansion microscopy . Cell culture of mammalian cells . COS - 7 monkey kidney cells ( purchased from CLS Cell Line Servie GmbH ) were cultured at 37\u00b0C and 5 % CO 2 in DMEM / HAM \u2019 s F12 medium with L - glutamine containing FBS ( 10 % ) and peni - cillin ( 100 U / ml ) and streptomycin ( 0 . 1 mg / ml ) . 20 \u2013 30 , 000 cells per well were seeded on round 18 mm high precision cover glasses ( No 1 . 5 ) in 12 - well culture plates ( Techno Plastic Products , 92012 ) and grown for 24 h prior to \ufb01 xation . Sample preparation . For \ufb01 xation , all solutions were pre - warmed to 37\u00b0C and \ufb01 xation was conducted on a heating plate set to 37 \u00b0C . Right before \ufb01 xation samples were rinsed once with pre - warmed Cytoskeleton buffer ( CB - buffer , 10 mM MES , 150 mM NaCl , 5mM EGTA , 5mM glucose and 5mM MgCl2 , pH 6 . 1 ) . Cells were then \ufb01 xed and permeabilized simultaneously incubating a primary \ufb01 xative solution of 0 . 3 % glutaraldehyde and 0 . 25 % Triton X - 100 in CB - buffer for 90s followed by a second \ufb01 xation using 2 % glutaraldehyde in CB - buffer for 10 min . Fixation was stopped by a 7min reduction step with freshly prepared 0 . 5 % NaBH 4 in PBS . Specimen were then washed three times with PBS ( 1x ) for 5 min each and treated differently depending on subsequent expansion method described below . Unless otherwise stated all incubations were carried out at room temperature in the following protocols . Immunostaining was either performed pre - gelation ( referred to as pre - labeling ) , post - expansion ( post - labeling ) or post - re - embedding ( post re - embedding labeling ) . Sequences and modi \ufb01 cations of DNA labels are listed in Supplementary Table 2 . A list of primary and secondary antibodies used for immunostaining in the corresponding Figures is provided in Supplementary Table 3 and Supplementary Table 4 with details about the expansion protocol used . Immunostaining of unexpanded Cos - 7 cells . Cells were placed in blocking buffer ( 5 % BSA in PBS ) for 1 h and then incubated for 1h with anti - alpha tubulin primary antibody solution ( ab1825 , diluted 1 : 500 , \ufb01 nal concentration c end = 2\u00b5g / ml ) diluted in blocking buffer . Samples were washed thrice in PBS ( 1x ) for 5 min each and incubated with secondary Alexa Fluor 532 IgG antibody solution in blocking buffer ( A - 11002 , diluted 1 : 200 , c end = 10 \u00b5g / ml ) for 1 h followed by three washes in PBS ( 1x ) for 10min each . ExM protocol using DNA trifunctional labels ( ExM protocol ) . After blocking with 5 % BSA in PBS for 1 h , cells were incubated with anti - alpha tubulin primary antibody ( ab1825 , diluted 1 : 500 , c end = 2 \u00b5g / ml ) in blocking buffer ( 5 % BSA in PBS ) for 1h , followed by three washes in PBS ( 1x ) for 5 min each and incubation of \u201c Antibody B \u201d DNA - labeled secondary antibodies ( 10 \u00b5g / ml ) in hybridization buffer ( 2x saline sodium citrate ( SSC ) , 10 % dextran sulfate , 1mg / ml yeast tRNA , 5 % BSA ) for 3h . Antisense DNA B1 - Alexa Fluor 532 and DNA B2 - Alexa Fluor 532 oligos were hybridized simultaneously at a total DNA concentration of 1 . 0 ng / \u00b5l for 3h in hybridization buffer . Then samples were washed three times with PBS ( 1x ) for 10min each . Gelation was performed on the lid of a 4 - well cell culture plate put on ice and covered with para \ufb01 lm that served as a \ufb02 at hydrophobic gelation surface . 18 mm cover glasses with cells facing down were placed on top of 90\u00b5l pre - chilled ExM monomer solution ( 8 . 625 % ( w / w ) SA , 20 % ( w / w ) AA , 0 . 15 % ( w / w ) BIS , 2M NaCl in PBS ) supplemented with 0 . 2 % APS and 0 . 2 % TEMED . Samples were then carefully transferred to a humidi \ufb01 ed chamber and incubated for 1 . 5 h at 37 \u00b0C for chemical crosslinking of acrylic monomers and trifunctional labels . After gelation samples were treated with 8 U Proteinase K in digestion buffer ( 50 mM Tris ( pH 8 . 0 ) , 0 . 5 % TritonX - 100 , 0 . 8M guanidine hydrochloride , 1 mM EDTA ) and then expanded in double - deionized water . Water was exchanged several times until the maximum expansion factor of the hydrogel was reached . The expansion factor was determined by measuring the diameter of the gel using a calipser . When the expansion factor did not change within three water exchanges this factor was assumed as maximum expansion of the hydrogel . Protein Retention protocol ( proExM protocol ) . Blocking and immunostaining were performed as described under \u201c Immunostaining of unexpanded Cos - 7 cells \u201d incubating anti - \u03b1 - tubulin antibody ( ab1825 , diluted 1 : 500 , c end = 2\u00b5g / ml ) and anti - \u00df - tubulin ( T8328 , diluted 1 : 200 , c end = 10 \u00b5g / ml ) simultaneously in blocking buffer as primary antibodies and Alexa Fluor 532 IgG antibodies ( A - 11002 and A - 1109 , each diluted 1 : 200 to c end = 10\u00b5g / ml ) diluted in blocking buffer as secondary antibodies . For copolymerization of amine groups into the hydrogel , cells were treated with the amine reactive agent Acryloyl X - SE ( 0 . 1mg / ml ) in PBS . The agent was freshly prepared from desiccated stock aliquots kept at \u2212 20 \u00b0C , incubated overnight in a humidi \ufb01 ed chamber , and subsequently washed twice for 15min each in PBS ( 1x ) . Hydrogel formation , Proteinase K digestion and expansion in water were performed as described under \u201c ExM protocol \u201d . After re - embedding of expanded hydrogels as described in section \u201c Bind - silane treatment and re - embedding \u201d , samples were labelled with \u03b1 - tubulin primary antibody solution ( ab1825 , diluted 1 : 500 , c end = 2 \u00b5g / ml ) in 2 % BSA for 3h at 37\u00b0C and then washed twice with 0 . 01 % Tween in PBS for 20min each and twice with PBS ( 1x ) for 10min each . Secondary antibodies were incubated for 3h at 37 \u00b0C and washed twice with 0 . 01 % Tween in PBS for 30 min each and twice with PBS ( 1x ) for 30 min followed by a washing step over night in PBS ( 1x ) . ExM protocol with glutaraldehyde linker ( ExM - GA protocol ) . Blocking and immunostaining were performed as described under \u201c Immunostaining of unex - panded Cos - 7 cells \u201d using \u03b1 - ( ab1825 , diluted 1 : 500 , c end = 2\u00b5g / ml ) and \u00df - tubulin ( T8328 , diluted 1 : 200 , c end = 10 \u00b5g / ml ) antibodies as primary antibodies and a mixture of Alexa Fluor 532 IgG secondary antibodies ( A - 11002 and A - 1109 , each diluted 1 : 200 to c end = 10\u00b5g / ml ) in blocking buffer . After washing with PBS ( 1x ) , cells were incubated with 0 . 25 % GA in PBS for 10min and washed thrice in PBS ( 1x ) for 5min each before proceeding with gelation of the samples . Gelation , digestion , and expansion was performed as described under \u201c ExM protocol \u201d . DNA label with Cy5 ( DNA - Cy5 protocol ) . Blocking and immunostaining were performed as described under \u201c ExM protocol \u201d with a mixture of primary \u03b1 - and \u00df - tubulin antibodies ( ab1825 diluted 1 : 500 with 2\u00b5g / ml and T8328 diluted 1 : 200 with 10\u00b5g / ml ) and DNA conjugated secondary antibodies \u201c Antibody B Cy5 \u201d or \u201c Antibody C Cy5 \u201d in hybridization buffer that were then directly incorporated into the hydrogel . Hydrogel formation , proteinase K digestion and expansion were performed as described under \u201c ExM protocol \u201d . After re - embedding on 24 - mm silanized round coverslips samples were incubated over night with Cy5 antisense oligos with a DNA concentration of 0 . 5ng / \u00b5l for each oligo in hybridization buffer . Ultrastructure expansion microscopy ( U - ExM ) . Twelve millimeters cover glasses with isolated 2 centrioles were placed in a solution containing 0 . 7 % FA , 1 % AA diluted in PBS ( 1x ) . Next , 35 \u00b5l of pre - chilled U - ExM monomer solution ( 19 % ( w / w ) SA , 10 % ( w / w ) AA , 0 . 1 % ( w / w ) BIS ) supplemented with 0 . 5 % APS and 0 . 5 % TEMED in PBS for 1min on a para \ufb01 lm coated plate put on ice . Gelation proceeded for 1h at 37\u00b0C in a humidi \ufb01 ed chamber . Samples were placed in denaturation buffer ( 200mM SDS , 200mM NaCl in 50mM Tris ( pH 9 . 0 ) ) for 15 min and then gels were carefully removed from the cover glasses and transferred to 1 . 5ml cen - trifuge tubes \ufb01 lled with denaturation buffer . Hydrogels were then incubated for 30 min at 95\u00b0C and then expanded in double deionized water until the maximum expansion of the gels were reached . After re - embedding on Bind - silane treated 24 - mm cover glasses , centrioles were labelled with anti alpha - tubulin primary antibodies ( ab1825 , c end = 2 \u00b5g / ml ) diluted 1 : 500 in 2 % BSA in PBS for 3 h at 37 \u00b0C , washed twice with 0 . 01 % Tween in PBS for 20 min each and twice with PBS ( 1x ) for 10min each . Next , secondary Alexa Fluor 647 F ( ab \u2018 ) 2 antibodies ( A - 21246 , ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 8 NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications 1 : 200 , c end = 10\u00b5g / ml ) diluted in 2 % BSA were incubated 3h at 37\u00b0C followed by two washing steps in 0 . 01 % Tween in PBS for 30 min each and two washes with PBS ( 1x ) for 30min . Before imaging gels were washed once more overnight in PBS ( 1x ) . For imaging of unexpanded centrioles the primary antibody anti poly - glutamylated tubulin ( Adipogen , 1 : 500 ) was diluted in 5 % BSA in PBS and incu - bated for 1 h at room temperature , washed thrice in PBS for 5 min each , followed by incubation with secondary Alexa Fluor 647F ( ab \u2018 ) 2 antibodies ( A - 21246 , 1 : 200 , c end = 10\u00b5g / ml ) diluted in 2 % BSA for 1h . The samples were then washed twice in 0 . 01 % Tween in PBS and once in PBS for 10min each . Re - embedding of expanded hydrogels ( Re - embedding protocol ) . To avoid shrinking caused by d STORM photoswitching buffer and to prevent drifting of the hydrogel during image acquistion an uncharged acrylamide gel was crosslinked throughout the hydrogel while chemically binding it on Bind - silane treated cover glasses . Round 24 - mm cover glasses ( high precision ) were sonicated successively in double - deionized water , absolute ethanol and 5M potassium hydroxide for 20 min each and washed with deionized water between every sonication step and \ufb01 nally oven dried at 100\u00b0C . 200 \u00b5l of Bind - silane working solution ( 5 \u00b5l Bind - Silane in 8 ml absolute ethanol , 200 \u00b5l glacial acetic acid , 1 . 8ml double deionized water ) were distributed evenly on cleaned 24 - mm cover glasses and left for around 1h until the solution was fully evaporated . Cover glasses were then rinsed with doubly deio - nized water and air - dried . Glasses were prepared shortly before use . For re - embedding expanded hydrogels were placed in 6 - well cell culture plates and each sample was covered with 3ml of freshly prepared Re - embedding solution ( 10 % acrylamide , 0 . 15 % bis - acrylamide , 0 . 05 % APS , 0 . 05 % TEMED in 5 mM Tris ( pH 8 . 9 ) ) . Samples were incubated on a platform shaker twice with freshly prepared solution for 30min each . Shaking of the Re - embedding solution is crucial in this step as it brings oxygen into the solution that prevents it from gelling to early . The stirring speed should be adjusted so that the liquid is in motion but the gels are not damaged . After the second incubation , samples were transferred on silanized coverglasses while carefully removing excess solution from the hydrogels using laboratory wipes . Another coverglass that was not silanized was placed on top of the hydrogels during the following steps . The whole setup was transferred to a humidi \ufb01 ed container equipped with gas injection holes . To accelerate gelation oxygen was extracted from the container by purging the chamber with nitrogen for 15min . The samples were then incubated at 37\u00b0C for 2h . After polymerization of the re - embedding gel samples were washed at least thrice for 30 min in double deionized water . Coverglasses on top of the hydrogel come off themselves during washing or can be detached carefully after the \ufb01 rst washing steps . Re - embedded gels were then placed in imaging buffer or staining buffer depending on the sub - sequent protocol . Microscopes . Single - molecule localization microscopy ( SMLM ) image acquisition was performed on a custom - built setup with an inverted Zeiss Axio Observer Z1 ( Carl Zeiss Microscopy ) microscope equipped with a De \ufb01 nite Focus autofocusing system . For excitation of different \ufb02 uorescent molecules the setup provides three iBeam smart diode lasers with 405 nm ( 100mW output power ) , 488nm ( 200 mW output power ) and 640nm ( 150mW output power ) and a DPSS ( diode pumped solid state ) 532nm laser ( gem532 , Laserquantum ) . Lasers were \ufb01 ltered with laser clean - up \ufb01 lters according to the speci \ufb01 c wavelength and focused on the back focal plane of the objective to achieve a wide \ufb01 eld illumination . To match the aqueous refractive index of expanded samples awater - immersion objective ( LD C - Apochromat 63x / 1 . 15 W Corr M27 , Carl Zeiss Microscopy ) is implemented in the microscope . The excitation light passes a quad - band dichroic beam splitter ( Di01 - R405 / 488 / 532 / 635 - 25\u00d736 , BrightLine ) combined with a quad - band rejection \ufb01 lter ( ZET405 / 488 / 532 / 642m , Chroma ) . For recording the emission of excited \ufb02 uor - ophores the setup is equipped with two Andor Ixon Ultra 897 EMCCD ( electron - multiplying charge - coupled device ) cameras at the side port of the microscope . The software Andor Solis ( Version 4 . 28 . 30014 ) was used to control the EMCCD cameras . The \ufb02 uorescence light is parallelized through a 160 mm achromatic lens ( Thorlabs ) and can be spectrally separated by a 630 DCXR ( Chroma ) dichroic beam splitter . In this con \ufb01 guration , two different emission wavelengths can be focussed on two cameras arranged perpendicular to each other . For all d STORM measurement in this work the beam splitter was removed and the emission light was directed to one camera . Suitable emission light \ufb01 lters were placed in front of the camera depending on the detected \ufb02 uorescent wavelength . For 3D imaging , an additional achromatic cylindrical lens ( f = 250 mm , Thorlabs ) was placed in the detection path close to the imaging plane before the relay system . Rescanning confocal imaging ( RCM ) was performed on a Nikon TiE inverted microscope equipped with an RCM unit ( Confocal . nl ) that is based on the image scanning principle 42 . The setup was operated by the microscope software NIS - Elements ( version 4 . 6 ) . Mounting and SMLM image conditions . Re - embedded hydrogels immobilized on 24 - mm cover glasses were immersed in photoswitching buffer consisting of 100 mM cysteamine hydrochloride ( MEA ) in PBS with optimized pH ( adjusted with KOH ) depending on the utilized \ufb02 uorescent dye . For Alexa Fluor 647 and Cy5 \ufb02 uorophores , the pH of the imaging buffer was adjusted to pH 7 . 7 and to pH 7 . 9 when using Alexa Fluor 532 , respectively . The buffer was prepared freshly before use . The hydrogel was incubated in photoswitching buffer twice for 20 min each before imaging . U - ExM treated samples labeled with the spontaneously blinking Si - rhodamine dye HMSiR were immobilized on Poly - L - lysine ( 0 . 1 % ) coated 24 - mm high - precision cover glasses and additionally embedded in 1 % ( w / v ) Agarose . As imaging buffer , double deionized water or pH adjusted PBS buffer ( 1x , pH 7 . 4 ) was used . For unexpanded d STORM imaging samples were placed in 100 mM MEA in PBS adjusted to pH 7 . 5 ( with KOH ) for DNA - Cy5 and Streptavidin - Alexa Fluor 647 and pH 7 . 9 ( with KOH ) when using Alexa Fluor 532 . 3D d STORM calibration . To obtain 3D calibration curves , \ufb02 uorescent beads were mixed in U - ExM or ExM monomer solution for 3D image acquisitions of U - ExM or ExM samples , respectively . Therefor \ufb02 uorescent marker stock supension ( 0 . 1 \u00b5m , ~ 1 . 8 \u00d7 10 11 particles / mL , TetraSpeck Microspheres , Thermo Fisher ) was vortexed for ~ 1 min and then diluted 1 : 50 in the corresponding monomer solution . After adding TEMED and APS in the appropriate concentrations , the bead - gel solution was vortexed again for ~ 20 s , polymerized , and expanded as described under the respective expansion protocol ( omitting the digestion or denaturation step ) . The expanded gels were then transferred on poly - L - lysine ( 0 . 1 % ) coated coverslips and additionally embedded in 1 % ( w / v ) Agarose . 4 \u00b5m z - stacks of several \ufb02 uorescent markers dispersed in the hydrogel ~ 50 \u2013 400 \u00b5m above the coverslip were recorded and used to generate 3D calibration curves as described below . The software Micro - manager 1 . 4 was used for image acquisition and to control the piezo driven stage . Image processing . For 2D and 3D d STORM image reconstruction super - resolution images were analyzed , post - processed and visualized using the analysis platform SMAP ( Superresolution Microscopy Analysis Platform ) with the GPU based 3D \ufb01 tter \ufb01 t3Dcspline 27 and the ImageJ plugin ThunderSTORM 43 . The respective integrated calibration tools were used for generating 3D astigmatism calibration curves . Localizations were further corrected for drift using the cross - correlation method , \ufb01 ltered for molecules with poor precision and grouped to one localization when molecules appeared in several consecutive images . Expansion factor determination . Centriole diameters of U - ExM expanded sam - ples were determined by averaging peak - to - peak distances of two cross - sectional pro \ufb01 les that were drawn through the center of the ninefold - symmetrical \u03b1 - tubulin signal using the line pro \ufb01 le tool of Fiji 44 . Peaks were then determined by using the peak \ufb01 nder minitool implemented in the analyse software Origin ( OriginLab , Northampton , MA ) . To determine the expansion factor post - expansion and post - re - embedding , Cos - 7 cells were labeled with a - tubulin and \u00df - tubululin and expanded according to the \u201c proExM protocol \u201d . An additional post - expansion immunostaining for \u03b1 - tubulin was performed using the same primary and sec - ondary antibodies . RCM images of the same cells were acquired before gelation , after expansion and after re - embedding in different imaging buffers . Images were then registered via rigid ( similarity ) and non - rigid registration ( B - spline ) using the open source , command - line program elastix 6 . The transform parameters of the similarity transformation of pre - and post - expansion RCM images were used to determine the initial expansion factor of the sample . RCM images acquired post - re - embedding in PBS ( 1x ) and cysteamine hydrochloride as well as a d STORM image in photoswitching buffer of the same area were registered in the same way using elastix to determine the expansion factor after re - embedding in different imaging buffers . Furthermore , a deformation vector \ufb01 eld of pre - expansion and post re - embedding RCM images was created using elastix and transformix 6 . Elastix and transformix code were executed in Wolfram Mathematica 11 . 2 . Analysis of microtubule transversal pro \ufb01 les . To analyze and compare the dif - ferent expansion protocols we developed a home written software that detects \ufb01 ber like structures and automatically determines the transversal pro \ufb01 le along these structures in reconstructed SMLM images . In detail the SMLM images are \ufb01 rst convolved with a Gaussian blur compensating for noise discontinuity or holes . A thresholding algorithm 45 then converts the image from grayscale to binary . Using Lees algorithm 46 the expanded lines are reduced to one pixel width . The pixel coordinates from all still connected lines are then retrieved and tested for continuity . Points of discontinuity are used as breakpoints and all following coordinates are connected to a new line . Lines , shorter than the minimum required length are discarded . An univariate spline of degree 3 ( c - spline ) is \ufb01 tted to each line . Note that shape and gradient of the line depend on the smoothing parameter . The result is a table containing the spline coordinates and the local derivatives . Perpendicular to the derivative a line pro \ufb01 le is extracted from the original image at each coordinate point . The averaged pro \ufb01 les for each spline are \ufb01 tted with the following functions ( Eqs . ( 1 \u2013 5 ) ) : Gaussian : y \u00bc h e (cid:2)\u00f0 x (cid:2) c \u00de 2 2 w 2 \u00fe b \u00f0 1 \u00de ( where h is the intensity , c the center , b the offset , and w the variance of the distribution . Optimal for single pro \ufb01 les ) . Bi (cid:2) Gaussian : y \u00bc h 1 e (cid:2)\u00f0 x (cid:2) c 1 \u00de 2 2 w 2 1 \u00fe h 2 e (cid:2)\u00f0 x (cid:2) c 2 \u00de 2 2 w 2 2 \u00fe b \u00f0 2 \u00de NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 ARTICLE NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications 9 ( optimal for pro \ufb01 les containing a dip ) . Tri (cid:2) Gaussian : y \u00bc h 1 e (cid:2)\u00f0 x (cid:2) c 1 \u00de 2 2 w 21 \u00fe h 2 e (cid:2)\u00f0 x (cid:2) c 2 \u00de 2 2 w 22 \u00fe h 3 e (cid:2)\u00f0 x (cid:2) c 3 \u00de 2 2 w 23 \u00fe b \u00f0 3 \u00de ( optimal for pro \ufb01 les exhibiting a dip and high background signal ) . Cylinder : y \u00bc h \ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03 r 22 (cid:2) \u00f0 x (cid:2) c \u00de 2 q (cid:2) \ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03 r 21 (cid:2) \u00f0 x (cid:2) c \u00de 2 q (cid:3) (cid:4) ; if x k k < r 1 h \ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03\ufb03 r 22 (cid:2) \u00f0 x (cid:2) c \u00de 2 q(cid:3) (cid:4) ; if x k k \u2265 r 1 ; x k k < r 2 0 ; else 8 > > > > > < > > > > > : \u00f0 4 \u00de ( y describes the theoretical intensity pro \ufb01 le of microtubules where r 1 and r 2 denote the inner and outer cylinder radius . The quality of the \ufb01 t strongly depends on the initial estimation of the parameters , due to the nonlinearity of the cylinder function . ) Multi (cid:2) Cylinder : y \u00bc cyl \u00f0 i 1 ; c ; 25 e x = 2 (cid:2) 2 a ; 25 e x = 2 (cid:2) a \u00de \u00fe cyl \u00f0 i 2 ; c ; 42 : 5 e x = 2 ; 42 : 5 e x = 2 \u00fe a \u00de \u00fe cyl \u00f0 i 3 ; c ; 25 e x = 2 \u00fe a ; 25 e x = 2 \u00fe 2 a \u00de \u00fe b \u00f0 5 \u00de ( includes the theoretical dimensions of microtubules leaving less degrees of freedom . Might result in a better \ufb01 t ) . Note that the \ufb01 t intensity ( h ) gives a good estimation for the relative labeling density . Using the splines \ufb01 tted to the maximum intensity projection we constructed xz - pro \ufb01 le projections of microtubules , by taking line pro \ufb01 les in each z - stack of the 3D image . Averaging the aligned line pro \ufb01 les in a layer yields the intensity values for the corresponding row of the xz - projection . Reporting summary . Further information on research design is available in the Nature Research Reporting Summary linked to this article . Data availability All data that support the \ufb01 ndings described in this study are available within the manuscript , the related supplementary information or deposited at https : / / doi . org / 10 . 6084 / m9 . \ufb01 gshare . 12415787 . v1 . Additional information is available from the corresponding authors upon reasonable request . Code availability The automated image processing software Line Pro \ufb01 ler is available at https : / / line - pro \ufb01 ler . readthedocs . io / en / latest / . Received : 14 April 2020 ; Accepted : 5 June 2020 ; References 1 . Chen , F . , Tillberg , P . W . & Boyden , E . S . Expansion microscopy . Science 347 , 543 \u2013 548 ( 2015 ) . 2 . Gambarotto , D . et al . Imaging cellular ultrastructures using expansion microscopy ( U - ExM ) . Nat . Methods 16 , 71 \u2013 74 ( 2019 ) . 3 . Wassie , A . T . , Zhao , Y . & Boyden , E . S . Expansion microscopy : principles and uses in biological research . Nat . Methods 16 , 33 \u2013 41 ( 2019 ) . 4 . Kunz , T . C . , G\u00f6tz , R . , Sauer , M . & Rudel , T . Detection of Chlamydia developmental forms and secreted effectors by expansion microscopy . Front . Cell Infect . Microbiol . 9 , 276 ( 2019 ) . 5 . Tillberg , P . W . et al . Protein - retention expansion microscopy of cells and tissues labeled using standard \ufb02 uorescent proteins and antibodies . Nat . Biotechnol . 34 , 987 \u2013 992 ( 2016 ) . 6 . Chen , F . et al . Nanoscale imaging of RNA with expansion microscopy . Nat . Methods 13 , 679 \u2013 684 ( 2016 ) . 7 . Chozinski , T . J . et al . Expansion microscopy with conventional antibodies and \ufb02 uorescent proteins . Nat . Methods 13 , 485 \u2013 488 ( 2016 ) . 8 . Ku , T . et al . Multiplexed and scalable super - resolution imaging of three - dimensional protein localization in size - adjustable tissues . Nat . Biotechnol . 34 , 973 \u2013 981 ( 2016 ) . 9 . Truckenbrodt , S . , Sommer , C . , Rizzoli , S . O . & Danzl , J . G . A practical guide to optimization in X10 expansion microscopy . Nat . Protoc . 14 , 832 \u2013 863 ( 2019 ) . 10 . Chang , J . B . et al . Iterative expansion microscopy . Nat . Methods 14 , 593 \u2013 599 ( 2017 ) . 11 . Sauer , M . & Heilemann , M . Single - molecule localization microscopy in eukaryotes . Chem . Rev . 117 , 7478 \u2013 7509 ( 2017 ) . 12 . Schermelleh , L . et al . Super - resolution microscopy demysti \ufb01 ed . Nat . Cell Biol . 21 , 72 \u2013 84 ( 2019 ) . 13 . Halpern , A . R . , Alas , G . C . M . , Chozinski , T . J . , Paredez , A . R . & Vaughan , J . C . Hybrid structured illumination expansion microscopy reveals microbial cytoskeleton organization . ACS Nano 11 , 12677 \u2013 12686 ( 2017 ) . 14 . Wang , Y . et al . Combined expansion microscopy with structured illumination microscopy for analyzing protein complexes . Nat . Protoc . 13 , 1869 \u2013 1895 ( 2018 ) . 15 . Gao , M . et al . Expansion stimulated emission depletion microscopy ( ExSTED ) . ACS Nano 12 , 4178 \u2013 4185 ( 2018 ) . 16 . Xu , H . et al . Molecular organization of mammalian meiotic chromosome axis revealed by expansion STORM microscopy . Proc . Natl Acad . Sci . USA 116 , 18423 \u2013 18428 ( 2019 ) . 17 . Shannon , C . E . Communication in the presence of noise . Proc . IEEE Inst . Electr . Electron Eng . 37 , 10 \u2013 21 ( 1949 ) . 18 . Baddeley , D . & Bewersdorf , J . Biological insight from super - resolution microscopy : what we can learn from localization - based images . Annu . Rev . Biochem 87 , 965 \u2013 989 ( 2018 ) . 19 . Legant , W . R . et al . High - density three - dimensional localization microscopy across large volumes . Nat . Methods 13 , 359 \u2013 365 ( 2016 ) . 20 . Heilemann , M . et al . Subdiffraction - resolution \ufb02 uorescence imaging with conventional \ufb02 uorescent probes . Angew . Chem . Int . Ed . 47 , 6172 \u2013 6176 ( 2008 ) . 21 . Dempsey , G . T . , Vaughan , J . C . , Chen , K . H . , Bates , M . & Zhuang , X . Evaluation of \ufb02 uorophores for optimal performance in localization - based super - resolution imaging . Nat . Methods 8 , 1027 \u2013 1036 ( 2011 ) . 22 . Weber , K . , Rathke , P . C . & Osborn , M . Cytoplasmic microtubular images in glutaraldehyde - \ufb01 xed tissue culture cells by electron microscopy and by immuno \ufb02 uorescence microscopy . Proc . Natl Acad . Sci . USA 75 , 1820 \u2013 1824 ( 1978 ) . 23 . Zhang , R . , Alushin , G . M . , Brown , A . & Nogales , E . Mechanistic origin of microtubule dynamic instability and its modulation by EB proteins . Cell 162 , 849 \u2013 859 ( 2015 ) . 24 . Zwettler , F . et al . Synaptonemal Complex Line Pro \ufb01 ler . 10 . 5281 / zenodo . 2643214 https : / / doi . org / 10 . 5281 / zenodo . 2643214 ( 2019 ) . 25 . Vaughan , J . C . , Jia , S . & Zhuang , X . Ultrabright photoactivatable \ufb02 uorophores created by reductive caging . Nat . Methods 9 , 1181 \u2013 1184 ( 2012 ) . 26 . Cahoon , C . K . et al . Superresolution expansion microscopy reveals the three - dimensional organization of the Drosophila synaptonemal complex . Proc . Natl Acad . Sci . USA 114 , E6857 \u2013 E6866 ( 2017 ) . 27 . Li , Y . et al . Real - time 3D single - molecule localization using experimental point spread functions . Nat . Methods 15 , 367 \u2013 369 ( 2018 ) . 28 . Hamel , V . et al . Identi \ufb01 cation of chlamydomonas central core centriolar proteins reveals a role for human WDR90 in ciliogenesis . Curr . Biol . 27 , 2486 \u2013 2498 ( 2017 ) . 29 . Uno , S . N . et al . A spontaneously blinking \ufb02 uorophore based on intramolecular spirocyclization for live - cell super - resolution imaging . Nat . Chem . 6 , 681 \u2013 689 ( 2014 ) . 30 . Pigino , G . et al . Cryoelectron tomography of radial spokes in cilia and \ufb02 agella . J . Cell Biol . 195 , 673 \u2013 687 ( 2011 ) . 31 . Gu , L . et al . Molecular resolution imaging by repetitive optical selective exposure . Nat . Methods 16 , 1114 \u2013 1118 ( 2019 ) . 32 . Cnossen , J . et al . Localization microscopy at doubled precision with patterned illumination . Nat . Methods 17 , 59 \u2013 63 ( 2020 ) . 33 . Gwosch , K . C . et al . MINFLUX nanoscopy delivers 3D multicolor nanometer resolution in cells . Nat . Methods 17 , 217 \u2013 224 ( 2020 ) . 34 . Deschout , D . et al . Precisely and accurately localizing single emitters in \ufb02 uorescence microscopy . Nat . Methods 11 , 253 \u2013 266 ( 2014 ) . Ehemals 30 . 35 . Virnat , D . et al . A peptide tag - speci \ufb01 c nanobody enables high quality labeling for dSTORM imaging . Nat . Commun . 9 , 930 ( 2018 ) . 36 . Wen , G . et al . Evaluation of direct grafting strategies via trivalent anchoring for enabling linpid membrane and cytoskeleton staining in expansion microscopy . ACS Nano https : / / doi . org / 10 . 1021 / acsnano . 9b09259 ( 2020 ) . 37 . Shi , X . et al . Label - retention expansion microscopy . Preprint at https : / / doi . org / 10 . 1101 / 687954 ( 2019 ) . 38 . Ries , J . , Kaplan , C . , Platonova , E . , Eghlidi , H . & Ewers , H . A simple , versatile method for GFP - based super - resolution microscopy via nanobodies . Nat . Methods 9 , 582 \u2013 584 ( 2012 ) . 39 . Mikhaykova , M . et al . Resolving bundled microtubules using anti - tubulin nanobodies . Nat . Commun . 6 , 7933 ( 2015 ) . 40 . Gao , R . et al . A highly homogeneous expansion microscopy polymer composed of tetrahedron - like monomers . Preprint at https : / / doi . org / 10 . 1101 / 814111 ( 2019 ) . 41 . Klena , N . et al . Isolation and \ufb02 uorescence imaging for single - particle reconstruction of Chlamydomonas centrioles . J . Vis . Exp . 2018 , e58109 ( 2018 ) . 42 . Luca , G . M . Rde et al . Con \ufb01 gurations of the re - scan confocal microscope ( RCM ) for biomedical applications . J . Microsc . 266 , 166 \u2013 177 ( 2017 ) . 43 . Ovesn\u00fd , M . , K \u0159 \u00ed \u017e ek , P . , Borkovec , J . , \u0160 vindrych , Z . & Hagen , G . M . ThunderSTORM : a comprehensive ImageJ plug - in for PALM and STORM data analysis and super - resolution imaging . Bioinformatics 30 , 2389 \u2013 2390 ( 2014 ) . ARTICLE NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 10 NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications 44 . Schindelin , J . et al . Fiji : an open - source platform for biological - image analysis . Nat . Methods 9 , 676 \u2013 682 ( 2012 ) . 45 . Otsu , N . A threshold selection method from grey level histograms . IEEE Trans . Syst . Man Cybernetics 9 , 62 \u2013 66 ( 1979 ) . 46 . Lee , T . C . , Kashyap , R . L . & Chu , C . N . Building skeleton models via 3 - D medial surface / axis thinning algorithms . Computer Vis . Graph . Image Process . 56 , 462 \u2013 478 ( 1994 ) . Acknowledgements The authors thank P . Gessner and L . Behringer - Pliess for assistance in immunocy - tochemistry and cell culture preparation . This work was supported by the German Research Foundation ( DFG , TRR 166 ReceptorLight , project A04 ) and the European Regional Development Fund ( EFRE project \u201c Center for Personalized Molecular Immu - notherapy \u201d ) . This work is supported by the European Research Council ( ERC ; StG 715289 ACCENT to P . G . ) and the Swiss National Science Foundation ( SNSF ) PP00P3 _ 187198 to P . G . Author contributions F . U . Z . , S . R . , D . G . , T . D . M . B . , V . H . , P . G . , and M . S . conceived and designed the project . M . S , V . H . , and P . G supervised the project . F . U . Z . performed all Ex - SMLM experiments . S . R . developed Line Pro \ufb01 ler and analyzed the data together with F . U . Z . , D . G . provided the centriole samples . All authors wrote and revised the \ufb01 nal manuscript . Competing interests The authors declare no competing interests . Additional information Supplementary information is available for this paper at https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 . Correspondence and requests for materials should be addressed to V . H . , P . G . or M . S . Peer review information Nature Communications thanks the anonymous reviewer ( s ) for their contribution to the peer review of this work . Peer reviewer reports are available . Reprints and permission information is available at http : / / www . nature . com / reprints Publisher \u2019 s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional af \ufb01 liations . Open Access This article is licensed under a Creative Commons Attribution 4 . 0 International License , which permits use , sharing , adaptation , distribution and reproduction in any medium or format , as long as you give appropriate credit to the original author ( s ) and the source , provide a link to the Creative Commons license , and indicate if changes were made . The images or other third party material in this article are included in the article \u2019 s Creative Commons license , unless indicated otherwise in a credit line to the material . If material is not included in the article \u2019 s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use , you will need to obtain permission directly from the copyright holder . To view a copy of this license , visit http : / / creativecommons . org / licenses / by / 4 . 0 / . \u00a9 The Author ( s ) 2020 NATURE COMMUNICATIONS | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 ARTICLE NATURE COMMUNICATIONS | ( 2020 ) 11 : 3388 | https : / / doi . org / 10 . 1038 / s41467 - 020 - 17086 - 8 | www . nature . com / naturecommunications 11",
    "linseyStudyDesignFixation2010": "J . S . Linsey Mechanical Engineering Department , Texas A & M University , 224 Engin . Physics , 3123 TAMU College Station , TX 77843 e - mail : jlinsey @ tamu . edu I . Tseng e - mail : iht @ andrew . cmu . edu K . Fu e - mail : katherine . fu @ gmail . com J . Cagan e - mail : cagan @ tamu . edu Mechanical Engineering Department , Carnegie Mellon , Scaife Hall 214 , 5000 Forbes Avenue , Pittsburgh , PA 15213 K . L . Wood Mechanical Engineering Department , The University of Texas at Austin e - mail : wood @ utexas . edu C . Schunn Department of Psychology , University of Pittsburg , LRDC 821 , 3939 O\u2019Hara St . , Pittsburgh , PA 15260 e - mail : schunn @ pitt . edu A Study of Design Fixation , Its Mitigation and Perception in Engineering Design Faculty The bridge between engineering design and cognitive science research is critical to understand the effectiveness of design methods as implemented by human designers . The study reported in this paper evaluates the effects of design \ufb01xation in a group of engi - neering design faculty , and also provides evidence for approaches to overcome design \ufb01xation . Three conditions are compared , a control , a \ufb01xation group whom were provided with an example solution , and a de\ufb01xation group whom were also given materials to mitigate their design \ufb01xation . Measures include indicators of design \ufb01xation and partici - pant perceptions . The study demonstrates that the engineering design faculty show sta - tistically signi\ufb01cant evidence of design \ufb01xation , but only partially perceive its effects . This study also indicates that design \ufb01xation can be mitigated . The group of participants in this study , due to their background in engineering design research and experience with student design teams , was expected to have more accurate perceptions or awareness of design \ufb01xation than the typical participant . Understanding the incongruities between participant perceptions and quantitative design outcomes are particularly of interest to researchers of design methods . For this study , clear evidence exists that designers , even those that study and teach design on a regular basis , do not know when they are being in\ufb02uenced or \ufb01xated by misleading or poor information . (cid:1) DOI : 10 . 1115 / 1 . 4001110 (cid:2) Keywords : design \ufb01xation , analogy , conceptual design 1 Introduction Engineering design was rigorously studied , arguably since Si - mon\u2019s Sciences of the Arti\ufb01cial was published in 1968 (cid:1) 1 (cid:2) . The \ufb01eld has explored formal and heuristic approaches to design re - \ufb01nement , manufacturing , generation , and computation (cid:1) 2 \u2013 4 (cid:2) . A critical part of engineering design is how designers think about the problem , how they reason about problem - relevant information , and how they are able to generate novel problem solutions . The study of such questions falls in the \ufb01eld of cognitive - based inno - vation and requires methods and knowledge from the \ufb01eld of cog - nitive psychology , integrated with process knowledge and partici - pants from the \ufb01eld of engineering design . The \ufb01eld of researchers who have actively pursued cognitive - based engineering is small but diverse , examples falling into the following \ufb01ve nonexhaus - tive categories . (cid:127) Representation\u2014the means to reason about and search for a solution to a design problem (cid:1) 5 \u2013 13 (cid:2) ; (cid:127) Fixation\u2014barriers to solution based on real and perceived constraints (cid:1) 14 \u2013 17 (cid:2) ; (cid:127) Analogy\u2014the mapping of knowledge from one domain to another supported by abstract representations (cid:1) 9 , 18 \u2013 29 (cid:2) ; (cid:127) Computational models\u2014models for cognitive simulation and generative tools (cid:1) 30 \u2013 32 (cid:2) ; (cid:127) Teams\u2014effective negotiation strategies , compatible repre - sentations , effective communication , trust among team members (cid:1) 7 , 33 \u2013 42 (cid:2) . Results in this \ufb01eld have begun to bear fruit , and the time is ripe for rigorous research efforts in this area . As such , in January 2008 , an NSF sponsored workshop was held in Knoxville , TN , as part of the CMMI Grantees Meeting . The workshop , entitled \u201cDiscussion on Individual and Team - Based Innovation , \u201d brought approximately 50 educators and re - searchers from the \ufb01eld of engineering design together for a day to learn about the current work and discuss potential directions for new research in the area of cognitive - based engineering design . As part of the workshop , participants took part in a formal cogni - tive study on the role of \ufb01xation and the use of analogies to overcome \ufb01xation . The experiment was formally developed , pi - loted , approved by the lead institution\u2019s Internal Review Board , and then run during the workshop . This paper presents the results of this study . One goal of the study was to allow participants to experience a formal and rigorous cognitive experiment . Since most of the participants had only engineering backgrounds , it was unlikely that many had participated in such studies . Another goal of the study was to advance the state of the \ufb01eld of cognitive - based engineering design by learning (cid:3) 1 (cid:4) if engineer - ing educators experience design \ufb01xation during a design problem solving exercise , (cid:3) 2 (cid:4) how design \ufb01xation can be overcome , adding to the current knowledge base of the \ufb01eld , and (cid:3) 3 (cid:4) whether the Contributed by the Design Education Committee of ASME for publication in the J OURNAL OF M ECHANICAL D ESIGN . Manuscript received June 19 , 2009 ; \ufb01nal manuscript received January 24 , 2010 ; published online April 13 , 2010 . Assoc . Editor : Janis Terpenny . Journal of Mechanical Design APRIL 2010 , Vol . 132 / 041003 - 1 Copyright \u00a9 2010 by ASME Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm participants accurately perceive the effects of provided examples and materials to mitigate design \ufb01xation . On one hand , the study group is interesting because they have experience in design , as demonstrated by survey results showing considerable industrial projects and patents , but on the other hand , they think about the process of design through teaching courses in design and through research in the broader \ufb01eld of design . It is interesting to see whether this cohort is susceptible to aspects of \ufb01xation that have been prevalent in other studies of design students or practitioners . What follows is a discussion of the current state of the art of cognitive - based engineering relevant to this work , a presentation of the experiment , results , discussion , and insights from the study . 2 Background / Previous Work A number of studies showed that design \ufb01xation effects can occur when example solutions are introduced to participants (cid:1) 15 , 43 \u2013 47 (cid:2) . Jansson and Smith (cid:1) 15 (cid:2) were the \ufb01rst to apply an experimental approach to study engineering design \ufb01xation . They found that showing example solutions can reduce the range of design solutions generated by a designer , and that aspects of the example solution , including aspects that were shown to violate goals of the problem statement , can \ufb01nd their way into the design - ers\u2019 solutions . A number of later experiments by others used the same and similar design problems to further investigate the issue of design \ufb01xation (cid:1) 47 , 48 (cid:2) . Purcell and Gero (cid:1) 47 (cid:2) suggested that the susceptibility of a designer to \ufb01xation may depend on the discipline of the designer , and that design \ufb01xation is more likely if the example problem embodies principles that are in line with the knowledge base of that discipline . These studies , as a whole , dem - onstrate that introducing examples can cause design \ufb01xation , re - sulting in less creativity during ideation . 2 . 1 Possible Approaches to Overcoming Design Fixation . Some approaches to reducing design \ufb01xation have been identi\ufb01ed . Using the same \ufb01xating examples as Jansson and Smith , Chrysikou and Weisberg (cid:1) 48 (cid:2) found that including de\ufb01xation in - structions can negate the \ufb01xating effects of the examples . Another possible approach to breaking or mitigating design \ufb01xation , be - yond de\ufb01xation instructions , is to assist the designer in \ufb01nding a new way to frame the problem , which may lead to new and im - proved solutions . The power of analogical inspiration , as part of problem framing , is supported by empirical evidence , as well as by examples of professional designers using analogies to solve problems (cid:1) 9 , 49 \u2013 51 (cid:2) . Within the literature , a number of approaches for enhancing analogical retrieval and use have been noted . Some of these de - pend on the expertise of the participants , and some are more gen - eral \ufb01ndings . Casakin and Goldschmidt (cid:1) 49 (cid:2) found that visual analogies can improve design problem solving for both novice and expert architects . Ball et al . (cid:1) 52 (cid:2) found experts use more analogies than novices do , so experience seems to increase re - trieval frequency . Expertise also enhances the ability to retrieve high - level principles derived from sets of analogies (cid:3) schema - driven (cid:4) . Novices tended to use more case - driven analogies (cid:3) analo - gies where a speci\ufb01c concrete example was used to develop a new solution (cid:1) 29 (cid:2)(cid:4) rather than schema - driven analogies (cid:3) more general design solution derived from a number of examples (cid:1) 53 , 54 (cid:2)(cid:4) . This difference can be explained because novices have more dif\ufb01culty retrieving relevant information when needed and have more dif\ufb01 - culty mapping concepts from disparate domains due to a lack of experience (cid:1) 55 (cid:2) . Tseng et al . (cid:1) 17 (cid:2) found that the effectiveness of analogical in - spiration in design was dependent on the timing of when the in - spiring information is given , as well as how apparently similar the information is to the problem being solved . More speci\ufb01cally , information that shares similar keywords or domains can be ap - plied to problem solving , even if the information is given before the designer has begun work on the problem , while information that is relevant but does not share similarity of keywords or do - mains only affects problem solving when the designer has already begun work on the problem . Additionally , Dahl and Moreau (cid:1) 56 (cid:2) demonstrated that subjects exposed to within - domain examples employed fewer far - domain analogies in generating solutions , and that the originality of the solutions produced was increased when subjects were encouraged to use analogies extensively and given no example solutions . Marsh et al . (cid:1) 57 (cid:2) found that within - domain examples caused sub - jects to be biased toward generating solutions that had similar features to those found in the examples . These \ufb01ndings led to the expectation for our experiment that a within - domain example so - lution given to participants prior to ideation would cause design \ufb01xation , as exhibited through fewer solutions generated and the appearance of features from the example in the solutions gener - ated . 2 . 1 . 1 Breaking or Mitigating Fixation : Formal Design - by - Analogy Methods . Analogy is a likely candidate for alleviating design \ufb01xation . A few formal methods have been developed to support design - by - analogy . These include Synectics (cid:1) 58 (cid:2) , French\u2019s work on inspiration from nature (cid:1) 59 , 60 (cid:2) , Biomimetic concept gen - eration (cid:1) 26 , 27 (cid:2) , the WordTree Design - by - Analogy method (cid:1) 20 , 61 (cid:2) , and analogous design through the use of the function and \ufb02ow basis (cid:1) 24 (cid:2) . Synectics is a group idea generation method that uses four types of analogies to solve problems : personal (cid:3) be the problem (cid:4) , direct (cid:3) functional or natural (cid:4) , symbolic , and fantasy (cid:1) 58 (cid:2) . Synectics gives little guidance to designers about how to \ufb01nd successful analogies . Other methods also base analogies on the natural world . French (cid:1) 59 , 60 (cid:2) highlights powerful examples that nature provides for design . Biomimetic concept generation pro - vides a systematic tool to index biological phenomena (cid:1) 26 , 62 (cid:2) . In biomimetic concept generation , the functional requirements of the problem and the keywords are \ufb01rst derived . The keywords are then referenced and relevant entries can be found . Like biomi - metic concept generation , the WordTree method is also based on keywords . The WordTree method takes the key functions or cus - tomer needs of a design problem , and systematically rerepresents them through the intuitive knowledge of the designers and through the WordNet database , particularly with synonyms known as hypernyms and troponyms . Analogous concepts can be also identi\ufb01ed by creating abstracted functional models of concepts and comparing the similarities between their functionality . Analo - gous and nonobvious products can be explored using the func - tional and \ufb02ow basis (cid:1) 24 (cid:2) . Other database supported computation tools for design - by - analogy have been recently developed . Examples of such tools are the work by Kurtoglu and Campbell (cid:1) 63 (cid:2) , Chui and Shu (cid:1) 26 (cid:2) , and Chakrabarti et al . (cid:1) 64 (cid:2) . Each has created an automated tool to provide inspiration to designers as part of the idea generation process . Based on the function or behavior of a device , analogies from nature or other devices are provided as potential sources of inspiration to the designer . 2 . 2 The Perception of Being Fixated . One reason why de - sign \ufb01xation is dif\ufb01cult to overcome is that designers are often not conscious of the fact that they are \ufb01xated . Ward and co - workers (cid:1) 45 , 65 (cid:2) found that the examples were not constraining the subjects consciously by causing them to believe that they should produce solutions similar to the given examples , but rather subconsciously constraining them ; when participants were asked to avoid produc - ing solutions that were similar to the examples , the similarity between the examples and generated solutions did not signi\ufb01 - cantly decrease when compared with participants\u2019 solutions who were not told to avoid solutions similar to the given examples . In general , participants did not have control over their use of the knowledge gained from the examples . These results suggest that designers are unaware that they are being in\ufb02uenced by example solutions or previously generated solutions (cid:1) 44 , 45 , 65 (cid:2) . Consistent 041003 - 2 / Vol . 132 , APRIL 2010 Transactions of the ASME Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm with this \ufb01nding , Linsey et al . (cid:1) 19 (cid:2) also observed that engineers were unaware that they were implementing prior examples to which they had been exposed . 2 . 3 Summary and Breaking New Ground . It is clear from this literature review that experiments have been deployed for studying \ufb01xation and the underlying causes within designers . We build on these studies and the associated results in this paper . Mitigating strategies for alleviating \ufb01xation have been studied , but much fertile ground has yet to be explored , especially in the do - main of engineering and for the use of analogies . It is also clear , to the best of our knowledge , that academics from engineering de - sign research have not been a focal group as part of \ufb01xation stud - ies . Our cognitive - based study here addresses these limitations , and constructs a framework on the foundation of the previous research . 3 Research Questions Design \ufb01xation is a common problem for both inexperienced and experienced designers . In this study , we seek to answer three fundamental research questions : (cid:3) 1 (cid:4) Do engineering academicians , both design researchers and educators , experience design \ufb01xation ? (cid:3) 2 (cid:4) How can design \ufb01xation be overcome or mitigated ? (cid:3) 3 (cid:4) Do the participants accurately perceive the effects of provided example solutions and of the materials to mitigate design \ufb01xation ? These three research questions , our associated hypotheses , and our mo - tivation for answering these questions are discussed in the follow - ing sections . In our study , three experimental conditions are implemented : a control , a \ufb01xation condition in which a negative example solution is presented , and a de\ufb01xation condition in which the negative example solution is given along with a list of possible solution directions to consider ; these conditions are referred to in presenting our hypotheses and de\ufb01ned in Sec . 4 . In addition , all participants \ufb01lled out a survey prior to the workshop that obtained demographic information and perceptions about the design pro - cess . 3 . 1 Evidence of Design Fixation . For this study , we explore the effects of \ufb01xation on experienced academic engineering de - signers . The group of participants for this experiment has a unique background that makes them interesting to study . All of the par - ticipants attended a workshop on the cognitive aspects of engi - neering design and developing cognitive experiments in engineer - ing design . They have clear interest in design and cognition . In addition , most of this group has experience teaching design and most are researchers in design . Therefore , this group is aware of design methods , they have spent time thinking about many of the issues related to design , particularly the \u201cfuzzy - front end , \u201d and they are likely to be aware of some of the dif\ufb01culties that design - ers have during idea generation . They are also aware of methods such as design - by - analogy and some of the short - comings of tra - ditional group brainstorming . Overcoming design \ufb01xation is a dif - \ufb01cult task . Yet , because of this group\u2019s background in design theory and methods , with their knowledge and skills , might they be able to more effectively overcome design \ufb01xation ? We there - fore seek to answer the following research question and contem - plate the following hypothesis : Research Question 1 : Do academic engineering design educa - tors show evidence of design \ufb01xation ? Hypothesis 1 : Academic engineering design educators will show evidence of design \ufb01xation . They will produce fewer total ideas when provided with example solutions and repeat ideas from provided examples as compared with the control group . 3 . 2 Overcoming / Mitigating Design Fixation . Prior research has shown that it is possible to reduce design \ufb01xation by instruct - ing participants to not focus on the negative aspects of the design and describing those short - comings (cid:1) 48 (cid:2) . This is clearly one ap - proach for reducing \ufb01xation , but based on anecdotal commentary in the design literature , it is likely that there are other approaches to mitigating \ufb01xation . Many product design books describe the bene\ufb01ts of functions , analogies , categories , and back - of - the - envelope calculations in the design process (cid:1) 66 \u2013 69 (cid:2) . In addition , analogy is noted as a tool for innovative design and a proli\ufb01cally implemented strategy by designers (cid:1) 9 , 19 , 70 (cid:2) . These observations lead to the following research question and hypothesis : Research Question 2 : What can engineers do to reduce their \ufb01xation on particular design solutions ? Can analogies , functions , categories of energy sources , and back - of - the - envelope calcula - tions assist in overcoming design \ufb01xation ? Hypothesis 2 : Design \ufb01xation can be reduced . The de\ufb01xation group will produce more ideas and repeat fewer ideas from the provided example solution than the \ufb01xation group . The de\ufb01xation group will implement more analogies than the other two condi - tions . 3 . 3 Participant Perceptions . Participants\u2019 perceptions fre - quently are not consistent with quantitative outcomes of their per - formance (cid:1) 37 , 44 , 45 (cid:2) . Unfortunately , perceptions are easily obtain - able and may be the basis that an individual or a company uses to choose to implement a particular method . For example , one of the reasons for group brainstorming\u2019s popularity , in spite of numerous studies contradicting its purported effectiveness , is that individu - als feel more productive during group brainstorming than when generating ideas alone (cid:1) 37 (cid:2) . In contrast to the participant\u2019s percep - tions of productivity , numerous studies quantitatively demonstrate a reduction in the number of ideas per person when comparing brainstorming in a group to individual brainstorming (cid:3) see Ref . (cid:1) 71 (cid:2) for a review (cid:4) . The basic group brainstorming method must be adapted to produce the quantity of results greater than the sum of the individuals (cid:1) 64 (cid:2) . The group of participants in this study has experience with design methods and is at least somewhat familiar with their short - comings . In addition , the majority of this group has taught design classes and observed their students\u2019 performance . Therefore , it is likely that the participants in this study will be much more aware of the effects of a provided example solution and additional de - \ufb01xation materials on their performance than participants who do not study design . In contrast , the prior literature indicates that participants are likely to inaccurately perceive the effects of an introduced example solution and the associated de\ufb01xation materi - als , if they are introduced . Therefore , we seek to answer the fol - lowing research question and test the related hypothesis : Research Question 3 : How well do participant perceptions of design results correspond to quantitative assessment of the results ? Hypothesis 3 : Participants will inaccurately perceive the effects of a provided example solution and associated de\ufb01xation materi - als . Results from survey questions collecting the participants\u2019 per - ceptions will be inconsistent with the quantitative metrics . 4 Experimental Method The experiment evaluates the effects of \ufb01xation on experienced academic engineering designers . To answer the research questions and hypotheses , we implement three experimental conditions : a control , a \ufb01xation , and a de\ufb01xation condition . All participants are given the same experimental procedure and documentation media . Participants in the \ufb01xation condition are provided with an ex - ample solution . Participants in the de\ufb01xation condition are also provided with the same example solution , but also with additional materials to potentially break or mitigate the design \ufb01xation (cid:3) de - tailed below in Sec . 4 . 4 (cid:4) . All participants are told that the goal of the experiment is to generate as many solutions to the design problem as possible , where a prize will be given to participants with the greatest number of solutions . This prize is an incentive for participants to devote serious effort to the design activity . All conditions end with a short post - experiment survey , which mea - sures prior exposure to the design problem , perceptions of partici - pants\u2019 performance and perceived in\ufb02uence of a provided example Journal of Mechanical Design APRIL 2010 , Vol . 132 / 041003 - 3 Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm solution . In the case of the de\ufb01xation condition , participations are asked questions regarding the de\ufb01xation materials and their per - ceptions of these materials . 4 . 1 Description of the Design Problem . All participants are provided with the same design problem . The design problem is to design a device to quickly shell peanuts for use in places like Haiti and West African countries , and is based on a real - world problem posted on ThinkCycle (cid:1) 72 (cid:2) . Participants are told that no electrical energy sources are available and are given customer needs (cid:3) Fig . 1 (cid:4) . This problem is chosen because it is a real world problem that is appropriate for an engineer , it has intrinsic incentive for solu - tions given its need - based nature , and the problem has a diverse set of available solutions . This problem has also been used in previous research on idea generation (cid:1) 35 , 73 , 74 (cid:2) . It is very unlikely that any of the participants would have extensive prior experience in solving this problem , yet shelling a peanut is a task that all of the participants have likely experienced . 4 . 2 Control Group . The control group is given the design problem as stated in Fig . 1 . They are not provided with an ex - ample solution or alternative representation of the problem . 4 . 3 Experimental Fixation Group . The \ufb01xation group is given the design problem as stated in Fig . 1 and an additional negative solution example (cid:3) Fig . 2 (cid:4) . They are not given an alterna - tive representation of the problem . The example solution uses a gasoline powered press to crush the shell , and does not separate the nut from the shell . The example solution focuses exclusively on a mechanical concept that crushes the shell and uses external fuel energy . This solution is dif\ufb01cult to control in terms of dam - aging the peanut , complex , and costly to manufacture for the West African environment . The participants all have graduate degrees in engineering , so these short - comings should be obvious to them . In addition , these particular process solutions are many of the common solutions found by participants in a prior experiment (cid:1) 35 , 73 (cid:2) . Common solutions to design problems were shown to create greater \ufb01xation (cid:3) fewer total solutions (cid:4) than unusual solu - tions (cid:1) 14 , 75 (cid:2) . 4 . 4 Experimental De\ufb01xation Group . The de\ufb01xation group is presented with the design problem , as in Fig . 1 , and also alterna - tive representations of the problem (cid:3) Fig . 3 (cid:4) . The alternative rep - resentations provide a brief functional description , useful analo - gies , a list of available energy sources , and a quick back - of - the - envelope calculation result . Some of the analogies were identi\ufb01ed using the WordTree Design - by - Analogy method with the key word of \u201cremove\u201d and \u201cshell\u201d to \ufb01nd the associated hypernyms and troponyms from WordNet (cid:1) 20 (cid:2) . 4 . 5 Participants . Fifty engineering academics expressed in - terest in attending the NSF sponsored workshop : \u201cDiscussion on Individual and Team - Based Innovation . \u201d 38 from this group \ufb01lled out the online presurvey for the workshop , and 34 actually at - tended the workshop . These thirty - four participants are randomly assigned to one of three conditions prior to the workshop , equally distributing the senior (cid:3) associate and full professors (cid:4) and junior faculty (cid:3) assistant professors (cid:4) . The study serves to demonstrate to the workshop participants an example cognitive study in engineer - ing design , while at the same time , providing useful experimental data . Based on the preworkshop survey results , which are only partially presented here , participants are faculty members (cid:3) 85 % (cid:4) , plus a few research scientists and graduate students (cid:3) 12 % (cid:4) , and federal government employees (cid:3) 3 % (cid:4) . There were no participants from industry . Almost half the participants are assistant professors (cid:3) 45 % (cid:4) ; 12 % are associate professors and 27 % are full professors . Most participants have mechanical engineering backgrounds . Most have at least 1 year of industrial experience (cid:3) 64 % (cid:4) and have consulted with industry at least once (cid:3) 79 % (cid:4) . There is also a high representation of women relative to the \ufb01eld of engineering (cid:3) 33 % females (cid:4) . A number of preregistered intended participants did not attend the beginning of the workshop so three participants were switched to different groups to compensate . Unintentionally , they were switched from the de\ufb01xation condition to a different condition , and they therefore had brie\ufb02y seen the de\ufb01xation materials . These three participants were removed from the data set . Fig . 1 Design problem description Fig . 2 Example solution provided to the participants in the \ufb01xation group 041003 - 4 / Vol . 132 , APRIL 2010 Transactions of the ASME Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm 5 Metrics To understand the effects of design \ufb01xation and evaluate the research questions , a set of measures are employed . To quantify the degree of \ufb01xation , \ufb01ve metrics are implemented : (cid:3) 1 (cid:4) number of ideas , (cid:3) 2 (cid:4) number of times features from the example solution appear in generated concepts , (cid:3) 3 (cid:4) percentage of features from the example solution that appear at least once in participant solutions , (cid:3) 4 (cid:4) number of energy domains , and (cid:3) 5 (cid:4) percentage of the solutions that employ a gas engine . To evaluate the effects of providing de\ufb01xation materials , along at least one dimension , the number of analogies is also measured . To provide inter - rater reliability , one of the authors evaluated all of the data for each metric , while a second rater evaluated two from each condition or at least 18 % of the data . 5 . 1 Quantity of Ideas . Building from the procedure devel - oped by Shah et al . (cid:1) 76 (cid:2) , a set of procedural rules are de\ufb01ned for what constitutes a single idea , see Ref . (cid:1) 77 (cid:2) for more details . Our basic de\ufb01nition for an idea is something that solves one or more functions of the design , as de\ufb01ned by the functional basis (cid:3) a clearly de\ufb01ned and tested language for expressing design func - tions (cid:1) 78 , 79 (cid:2)(cid:4) . The total number of unique (cid:3) nonredundant , nonre - peated (cid:4) ideas is calculated for each person . Pearson\u2019s correlation coef\ufb01cient (cid:1) 80 (cid:2) was 0 . 97 , indicating that the measure is highly reliable . A high degree of Pearson\u2019s correlation indicates that if one evaluator had a higher score , then the other evaluator also tended to give a higher score . 5 . 2 Repeated Example Solution Features and Percentage of Features Used . Figure 4 illustrates the example solution pro - vided to the participants and all of the design ideas contained within it (cid:3) number of ideas (cid:4) , categorized by function . The number of times each participant employs one of the design features from the example solution is counted . This procedure results in two metrics , which indicate the degree of \ufb01xation : the total number of times a feature is repeated and the total percentage of features from the example that is implemented at least once . One of the authors evaluated all of the data , while a second rater measured two from each condition or 18 % of the data . In half of the cases , Fig . 3 De\ufb01xation materials Fig . 4 Example solution provided to participants Journal of Mechanical Design APRIL 2010 , Vol . 132 / 041003 - 5 Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm the two raters had identical scores for the number of repeated features , and their Pearson\u2019s correlation coef\ufb01cient (cid:1) 80 (cid:2) was 0 . 97 , indicating that the measure is highly reliable . Cohen\u2019s Kappa (cid:1) 80 (cid:2) is not used since the metric of interest is the relative number of repeated features across conditions , not whether or not the partici - pants use a particular feature or the absolute number of repeated features in this setting . To calculate the percentage of the features that each participant reused from the example , the number of fea - tures used at least once is divided by the total number of ideas within the example solution (cid:3) eight ideas , Fig . 4 (cid:4) . 5 . 3 Energy Domains and Percentage of Solutions Employ - ing a Gas Engine . In addition to recording the quantity of ideas , the number of energy sources used by each participant is analyzed (cid:3) Table 1 (cid:4) . These energy sources are categorized into 16 energy categories (cid:3) wind , solar , water streams , captured rain water at a height , water (cid:3) other (cid:4) , human , animal , nuclear , electrical outlet , \ufb01re , gas engine , engine (cid:3) other (cid:4) , fuel cell , \ufb02uid density difference , chemical , and genetic (cid:4) . The original tally included 18 categories , but it was found , due to the universality of gravity that the two gravity driven categories are dif\ufb01cult to measure reliably between raters , and were thus removed . The total of all energy sources used by each participant is recorded . Since the goal is to deter - mine the breath of energy sources spanned , a participant receives the same score , regardless of whether they use an energy source once or multiple times . Since the de\ufb01xation materials provided a list of energy sources to directly break \ufb01xation on the gas engine , the percentage of solutions using a gas engine is also calculated . 5 . 4 Analogies - Breaking Fixation . After 45 min of ideation , analogies were identi\ufb01ed by the participants by going back through their solutions , and circling or labeling any analogies used with a red permanent marker . The total number of unique analo - gies identi\ufb01ed by each participant is measured . Some participants circled components of their designs , some drew arrows with tex - tual descriptions , and others described the analogies they used while also circling them . Initially , all of this information is tabu - lated . If a participant circled something and wrote no textual de - scription , this analogy is counted for that participant toward the total number of analogies they used . If a participant wrote a de - scription , the analogy is counted for that participant toward their total number of analogies . If the participant both circles and writes text , it is determined if the two pertained to separate analogies . One participant stated that they believed all of their ideas were analogies since all their ideas came from things they had seen before . Although this statement that most of an engineers\u2019 design ideas come from prior exposure is true , this is not a working use of the concept of analogy , where a speci\ufb01c idea is borrowed or mapped onto the current problem ; thus , this participant\u2019s data are not included in the analogy analysis . Due to the fact that each participant\u2019s de\ufb01nition of \u201canalogy\u201d is unique , the analogies identi\ufb01ed by the participant are counted as valid analogies . This reduces the amount of biases introduced by the researchers , but there is still subjectivity due to differing par - ticipants\u2019 de\ufb01nitions . However , the participants\u2019 de\ufb01nitions of analogy , as provided in the presurvey , are surprisingly consistent across the participants . Prior studies have used indication of analogies such as \u201cdevice X is like device Y\u201d to identify analogies in concurrent think - aloud protocols or recorded team conversa - tions (cid:1) 9 , 23 , 52 (cid:2) . This study did not include concurrent think - aloud descriptions as the participants were working , nor was there any other indication of which ideas were based on analogies , there - fore , the participants\u2019 identi\ufb01cation of analogy had to be used . Since participants did not provide details on the analogies , it was not possible to further classify analogies along common dimen - sions such as analogical distance . In tabulating the analogies metric , as with all other metrics , an inter - rater agreement analysis is performed to ensure objectivity and consistency . In tallying the total number of analogies remov - ing any repeats , the raters achieved 99 % agreement and a Pear - son\u2019s Correlation of 0 . 99 . These measures indicate that the metric is highly reliable , and there is strong consistency between the two evaluators . 6 Results : Design Fixation A key outcome of this study is on understanding design \ufb01xa - tion , participants\u2019 perception of it and how to break or mitigate \ufb01xation when it occurs . Figures 5 and 6 illustrate examples of typical participant results with high and low degrees of \ufb01xation (cid:3) samples sizes are in Table 2 (cid:4) . Four measures are implemented to assess each participant\u2019s degree of \ufb01xation . From these measures , a participants\u2019 \ufb01xation may be ascertained and the hypotheses tested . The number of nonredundant ideas varies across the three con - ditions (cid:3) Fig . 7 (cid:4) . An ANOVA (cid:3) analysis of variance (cid:4) shows a statis - tically signi\ufb01cant effect across the \ufb01xation conditions (cid:3) F = 3 . 7 , p Table 1 Energy source categories Wind Anything powered by naturally occurring wind . Includes using a wind generator to generate electricity , a windmill to directly turn a mechanism , and naturally allowing the wind to blow peanut shells away . Does not include a fan or other wind source powered by another energy source . Solar Anything powered by the sun . Includes both solar panels to generate electricity and using the sun directly to heat water or roast the peanuts . Water streams Uses naturally existing water streams such as a water wheel in a river to power the system , either by generating electricity or directly powering the mechanical system . Captured rain water at a height Capturing rain water at a height to accomplish much the same as the previous category . Water (cid:3) other (cid:4) Can include the soaking of peanuts to soften them , and many other uses . Human Includes manual shelling , turning a hand crank , and any other human power . Animal Includes using animals to motivate a mechanical system , training or genetically manipulating animals to shell peanuts . Nuclear Nuclear power station to provide energy . Electrical outlet Any solution where the concept involves plugging into a wall socket for electricity . Fire Anything that uses a \ufb01re to generate power , excludes internal combustion or steam engines , but includes using a \ufb01re to roast peanuts or burn off shells . Gas engine Gasoline powered internal combustion engine . Engine (cid:3) other (cid:4) Any other type of fuel engine . Fuel cell Hydrogen or other type fuel cell . Fluid density difference Concepts based on whether something \ufb02oats or sinks , often used to sort peanuts and their shells . Chemical Any chemical process to generate power that is different than seen above , or to burn or dissolve the peanut shell off . Genetic Genetically altering the peanut itself . Does not include genetic manipulations on animals (cid:3) which should be categorized as animal (cid:4) . 041003 - 6 / Vol . 132 , APRIL 2010 Transactions of the ASME Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm (cid:1) 0 . 04 (cid:4) . 1 A t - test shows that the control group produces more ideas than the \ufb01xation group (cid:3) t = 2 . 94 , p (cid:1) 0 . 02 (cid:4) . The other pair - wise comparisons are not statistically signi\ufb01cant . The variation in the number of nonredundant ideas indicates that the example did cause design \ufb01xation , resulting in fewer ideas being generated . The trend in this data is that the de\ufb01xation group produces more ideas on average than the \ufb01xation group , indicating that the additional materials assist in reducing their \ufb01xation . 6 . 1 Number of Example Solution Features Used . Another indicator of the degree of \ufb01xation is the number of times that the participants reuse features from the provided example solution . This metric differs across the three conditions and ranges from 1 to 43 repeated features (cid:3) Fig . 8 (cid:4) . The control group did not see the example solution , but they still may think of the same features that are present in the example solution . These data do not satisfy the assumptions for a standard ANOVA since Shapiro \u2013 Wilk\u2019s test of normality shows the data are not normally distributed and Lev - ene\u2019s test for equality of variances shows that the variances are not homogenous ; therefore , a Kruskal \u2013 Wallis ANOVA is imple - mented instead . ANOVA can be used when there are only small departures from normality , but if there are also unequal variances across the groups , a different approach is required . A Kruskal \u2013 Wallis ANOVA is analogous to a standard ANOVA , but is the nonparametric statistical equivalent and evaluates the relative ranks of the data points . Implementing a Kruskal \u2013 Wallis ANOVA , there is a statistically signi\ufb01cant difference across the three con - ditions (cid:3) H = 7 . 3 , df = 2 , p = 0 . 03 , N = 31 (cid:4) . 1 The data are not normally distributed but ANOVA is robust for departures from normality . The remaining assumptions for ANOVA are met . To con\ufb01rm the ANOVA results , a Kruskal \u2013 Wallis ANOVA is also completed on the data (cid:3) H = 5 . 7 , p (cid:1) 0 . 06 (cid:4) . Table 2 Sample size for each condition Group Sample Size Control 9 Fixation 12 De\ufb01xation 10 Fig . 5 A set of solutions showing a low degree of \ufb01xation on the provided example solution Fig . 6 A set of solutions showing a high degree of \ufb01xation on the provided example solution Fig . 7 The \ufb01xation group produced fewer ideas , on average , than either the control group or the de\ufb01xation group . Each er - ror bar is \u00b11 standard error . Journal of Mechanical Design APRIL 2010 , Vol . 132 / 041003 - 7 Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm The \u201cnumber of features\u201d data also indicate that the example solution caused \ufb01xation , as the features from the example (cid:3) the \ufb01xation condition (cid:4) are reused more often than for the control . This data indicate that the additional materials are effective in mitigat - ing or reducing the design \ufb01xations , since the de\ufb01xation condition reuses fewer features from the example solution than the \ufb01xation condition . In the case of the control group , participants generated features in their concepts corresponding to the example solution . This result is to be expected since some of the features , as shown in Fig . 4 , are perfectly acceptable solutions , whereas others , such as the gas - press are not . The control group , however , did not overly use the solutions of the provided example for the \ufb01xation condition , whereas the \ufb01xation group did offer solutions powered by gasoline or similar fuel - based systems . 6 . 2 Percentage of Features Used . The three conditions also caused the participants to implement different percentages of the features from the example solution (cid:3) Fig . 9 (cid:4) , another indicator of \ufb01xation . The number of times that the participants implemented one of the features from the example solution was counted . Again , these data do not satisfy the assumptions for a standard ANOVA . The Shapiro \u2013 Wilk\u2019s test of normality shows that the data is not normally distributed , and Levene\u2019s test for equality of variances shows that the variances are not homogenous ; therefore , a Kruskal \u2013 Wallis ANOVA is completed instead . There is a statisti - cally signi\ufb01cant difference across the three conditions (cid:3) H = 7 . 3 , df = 2 , p = 0 . 03 , N = 31 (cid:4) . A Wilcoxon\u2019s Rank - sum test shows that the \ufb01xation group incorporated more of the features from the ex - ample in their solutions than the de\ufb01xation group (cid:3) W s = 85 . 5 , n 1 = 10 , n 2 = 12 , p = 0 . 08 (cid:4) and the control (cid:3) W s = 67 . 5 , n 1 = 9 , n 2 = 12 , p = 0 . 02 (cid:4) . Again the \u201cpercentage of features\u201d measure shows that the ex - ample solution causes designer \ufb01xation . The de\ufb01xation condition is different from the \ufb01xation condition , based on statistical signi\ufb01 - cance , again showing that the de\ufb01xation materials are assisting the designers in overcoming the \ufb01xation induced by the presented example . 7 Results : Energy Sources Fixation In addition to the number of solutions , the energy source used in the design can be another indicator of design \ufb01xation . The de\ufb01xation condition contained a categorical list of energy sources , Fig . 3 . Participants in the \ufb01xation condition were given an example that is powered by a gas engine . It is expected that this example would \ufb01xate individuals on using a gas engine . Participants in this condition were given information in addition to the gas engine powered example that is intended to aid in breaking the induced \ufb01xation . Individuals in the control condition were given no ex - ample on which to \ufb01xate . Both predicted effects are observed in the results (cid:3) Fig . 10 (cid:4) . Again , these data do not meet the assump - tions for a standard ANOVA , as the data are not normally distrib - uted and the variances are not homogenous . The Kruskal \u2013 Wallis ANOVA compares the data based on the relative rank of the re - sults , but this approach is not accurate when there are a large number of equal outcomes as there are with this data set . For this data set , one - way ANOVA via randomization is implemented (cid:1) 81 , 82 (cid:2) . This approach is analogous to the other approaches , but does not make any assumptions about the distribution or the rank - ing of the data . Based on the graphical results shown in Fig . 10 , the \ufb01xation group is clearly different and distinct from the other two groups . One might expect the other two conditions not to use gas engines at all , but in this study , gas engines were occasionally used . Using a one - way ANOVA via randomization , there is statistically signi\ufb01 - cant difference across the groups (cid:3) p = 0 . 05 (cid:4) (cid:1) 83 , 84 (cid:2) . The \ufb01xation group produced a larger percentage of gas powered designs than the control group , indicating that the example solution caused \ufb01xation . The \ufb01xation group also showed a strong trend for pro - ducing a larger percentage of gas powered designs than the de\ufb01x - ation group (cid:3) t = 1 . 97 , p (cid:1) 0 . 08 (cid:4) , suggesting that the de\ufb01xation in - formation is effective in breaking or mitigating the induced \ufb01xation . Similar to the other measures of design \ufb01xation , the re - sults show that \ufb01xation is occurring , and the de\ufb01xation materials are having a statistically signi\ufb01cant impact . The total number of energy sources used in all stages of peanut shelling differed across the three conditions (cid:3) Fig . 11 (cid:4) . Again , these data do not satisfy the assumptions for a standard ANOVA (cid:3) data are not normally distributed and the variances are not homog - enous (cid:4) so a Kruskal \u2013 Wallis ANOVA is implemented . There is not a statistically signi\ufb01cant difference across the three conditions Fig . 8 The \ufb01xation group repeated , on average , features from the example solution more often than the other two groups . Each error bar is \u00b11 standard error . Fig . 9 The \ufb01xation group used , on average , a higher percent - age of the features from the example solution in their concepts . Each error bar is \u00b11 standard error . Fig . 10 The de\ufb01xation group produced more designs powered by a gas engine than individuals in the other conditions . The error bars are \u00b11 standard error . 041003 - 8 / Vol . 132 , APRIL 2010 Transactions of the ASME Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm (cid:3) H = 3 . 28 , df = 2 , p = 0 . 194 , N = 31 (cid:4) . However , the \ufb01xation condi - tion produced , on average , fewer energy sources than the de\ufb01x - ation condition (cid:3) Wilcoxon\u2019s Rank - sum test , W s = 112 . 5 , n 1 = 10 , n 2 = 12 , p = 0 . 09 (cid:4) , suggesting that the de\ufb01xation materials are ef - fective in breaking the induced \ufb01xation . The other pair - wise com - parisons are not statistically signi\ufb01cant . Consistent with other results from the study , these results sug - gest that the de\ufb01xation materials are assisting in mitigating or reducing design \ufb01xation . The additional categories of available energy sources are assisting the designers in identifying solutions . 8 General Fixation Results Discussion The various measures related to \ufb01xation clearly illustrate that the example solution causes design \ufb01xation . This result is shown by the lower number of ideas generated , by a higher number of features from the example being used in the solutions , and by fewer energy categories being implemented in the participants\u2019 concepts . This \ufb01xation is of particular interest since these partici - pants are not novice designers . All participants have the required domain knowledge to identify short - comings in the presented ex - ample solution . Design \ufb01xation is experienced by the engineering design faculty . Providing participants with analogies and rerepresentations of the problem through categories did assist in reducing their \ufb01xation on the example solution , but it did not completely eliminate it . Participants in the control group still outperformed both the \ufb01xa - tion and the de\ufb01xation group in total number of concepts , whereas the de\ufb01xation group employed a greater diversity of energy - based solutions . 9 Results : Analogies Many designers employ analogies to inspire solutions to a given problem . The focal metric to consider in this study when examining analogies is simply the quantity identi\ufb01ed by the par - ticipants across the three conditions . The analogies were identi\ufb01ed by the participants after the 45 min of ideation by revisiting their solutions , circling and labeling any analogies used with a red marker . The number of reported analogies employed did not vary sta - tistically across the conditions (cid:3) Fig . 12 (cid:4) . Again , the data are non - normally distributed with unequal variances , requiring nonpara - metric tests . A Kruskal \u2013 Wallis ANOVA shows there is a not statistically signi\ufb01cant difference across the three conditions (cid:3) H = 1 . 4 , df = 2 , p = 0 . 5 , N = 30 (cid:4) . The de\ufb01xation group did , however , use slightly more analogies , on average , implementing many of the analogies provided in the de\ufb01xation materials in a variety of ways . 9 . 1 Analogies Results Discussion . The three conditions implemented in this study produced distinct levels of analogy use . The \ufb01xation group and the control have a similar level of analogy use . However , the de\ufb01xation group , given both the example solu - tion and the extra information , used slightly more analogies , on average , than that given just the example solution . This could be a result of the \ufb01xation caused by the example solution , or further indication that more supplemental information is better for a broader search of the design space . Many of the analogies used by the group given extra information were directly inspired by the key word list supplied in the materials . Future work , with more participants per condition , should investigate whether the ob - served trend replicates . An important obstacle to note is the subjectivity of identifying analogies . In order to remove the experimenters\u2019 interpretations and biases , a red marker was given to each participant after the ideation session , with which they were to circle and label any analogies used . This removed subjectivity from the experiment - ers\u2019 perspectives , however , not from the participant\u2019s perspective . Many participants showed signs of a broader understanding of the de\ufb01nition of analogy , often times circling mechanisms or pro - cesses that were taken directly from industry , for example , press . Since the press was being used in an identical way to that which is used in practice , it is not legitimately considered an analogy for the purposes of this study , but more an application of a technol - ogy . Despite this discrepancy , all analogies identi\ufb01ed by partici - pants were counted as analogies because omitting these would create too much uncertainty in the analysis . 10 Results from Post - Experiment Survey A post - experiment survey measures a variety of items including the following : participants\u2019 opinions about the design problem ; perceptions about the effect of the example solution and the addi - tional material ; and if they had exposure to the design problem and its solutions prior to the experiment . These measures serve to provide further insights and validity to the experimental results . The participants\u2019 overall opinions of the design problem and activity indicate that , on average , they found it interesting and seriously committed themselves to the task . The participants , on a semantic difference scale , felt that they worked hard on the activ - ity (cid:3) mean (cid:3) SD (cid:4) : 1 . 8 (cid:3) 0 . 90 (cid:4) 1 = worked hard , 5 = minimal effort (cid:4) and found the activity to be somewhat motivating but not inspiring (cid:3) mean (cid:3) SD (cid:4) : 2 . 2 (cid:3) 0 . 86 (cid:4) (cid:3) 1 = motivating and 5 = demotivating (cid:4) and 2 . 6 (cid:3) 0 . 78 (cid:4) (cid:3) 1 = inspiring and 5 = frustrating (cid:4)(cid:4) . None of these results show a mean shifted substantially to the right of the scales mid - point . For the validity of this study , it is important that the partici - pants were motivated and put in a substantial effort since in a more realistic design setting , it is expected that engineers are gen - erally well motivated to solve a given design problem . Since the presented design problem is an actual existing issue , it is possible that the participants may have had prior exposure to Fig . 11 The de\ufb01xation group used , on average , more energy sources in total than participants in the other two groups . Each error bar is \u00b11 standard error . Fig . 12 All groups used analogies in developing solutions to the design problem . The de\ufb01xation group , on average , em - ployed slightly more analogies during ideation . The error bars are \u00b11 standard error . Journal of Mechanical Design APRIL 2010 , Vol . 132 / 041003 - 9 Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm the problem and to the solutions . A total of seven participants , evenly distributed across the conditions , had prior exposure to the design problem , with four of them also having exposure to solu - tions to the design problem . These four participants believed their prior exposure had an insigni\ufb01cant to only some in\ufb02uence on their results (cid:3) the four responses were 1\u2014insigni\ufb01cant , 2\u2014minor , and 1\u2014some in\ufb02uence (cid:4) . Due to the low level of prior exposure and that the participants were evenly distributed across the conditions , we do not believe that the prior exposure affects the results of this experiment . The key outcome of the survey is the participants\u2019 perception about their performance . The \ufb01xation and the de\ufb01xation groups were asked , using a Likert scale question , if they felt the provided example solution had in\ufb02uenced them , and then if it had posi - tively or negatively in\ufb02uenced them (cid:3) Figs . 13 \u2013 15 , error bars are 1 standard error (cid:4) . Both groups felt that the provided example solu - tion had in\ufb02uenced them . The participants are recognizing the fact that they are being in\ufb02uenced by the provided design example . The \ufb01xation group tended to believe that the effect of the ex - ample solution is positive whereas the de\ufb01xation group is unsure of the in\ufb02uence . The differences between the groups are not sta - tistically signi\ufb01cant , although this lack of statistical signi\ufb01cance may re\ufb02ect the moderate group sizes of this study . The partici - pants\u2019 perceptions are in contrast to the quantitative \ufb01xation re - sults that indicate that the example is having a strong negative effect on the \ufb01xation group , meaning that the designer may not be aware of the negative in\ufb02uence . In addition , the participants\u2019 perceptions of the effects of the de\ufb01xation materials were also measured (cid:3) Table 3 (cid:4) . In this case , participants\u2019 correctly believe that the additional information is bene\ufb01ting them with a mean of only agree to somewhat agree (cid:3) 2 . 7 (cid:4) and a fairly high standard deviation (cid:3) 1 . 33 (cid:4) . This high stan - dard deviation may indicate the some of the participants are more accurate in their perceptions . The quantitative results indicate that there is a very strong positive effect in overcoming the \ufb01xation due to the provided de\ufb01xation materials . While the participant perceptions are generally accurate , in this case , they do not strongly match the quantitative results . This indicates that partici - pant perceptions are not an accurate tool for evaluating design methods . 11 Addressing the Research Questions Research Question 1 : Do academic engineering design educa - tors show evidence of design \ufb01xation ? Academic engineering design educators do show evidence of design \ufb01xation . The \ufb01xation group produced fewer ideas , reused more of the features from the example solution , and implemented fewer categories of energy sources than the control group . Design \ufb01xation is evidenced by the presence of a considerable number of solution elements that are clearly not appropriate for the context of the design problem . This group of participants has a high de - gree of knowledge and can clearly recognize the short - comings of the presented design . Qualitatively compared with undergraduate students in a previous study (cid:1) 73 (cid:2) , the design educators produced a larger number of highly novel solutions not identi\ufb01ed by the stu - dents , providing some validation of their status as design experts . It should be noted , however , that the students did produce a range of novel ideas for the design problem . Question 2 : What can engineers do to mitigate their \ufb01xation on design solutions ? Can analogies , functions , categories of energy sources , and back - of - the - envelope calculations assist in overcom - ing design \ufb01xation ? Some of the strategies that may mitigate design \ufb01xation are analogies , a functional decomposition , rerepresentation of the problem , categories of solutions (cid:3) such as energy sources (cid:4) , and Table 3 Participant perceptions of the effect of the de\ufb01xation materials Survey question De\ufb01xation group mean (cid:3) SD (cid:4) The provided additional information (cid:3) functions , analogies , calculation , energy domains (cid:4) bene\ufb01ted me . 2 . 7 = agree / somewhat agree (cid:3) 1 . 33 (cid:4) The provided additional information (cid:3) functions , analogies , calculation , energy domains (cid:4) hindered me . 5 . 1 = somewhat disagree (cid:3) 1 . 45 (cid:4) Scale : 1 = strongly agree , 2 = agree , 3 = somewhat agree , 4 = neutral , 5 = somewhat disagree , 6 = disagree , 7 = strongly disagree Fig . 13 Participants believe they were in\ufb02uenced by the pro - vided example solution Fig . 14 Participants in the \ufb01xation and the de\ufb01xation group felt the example solution had a positive in\ufb02uence or at least were unsure that the in\ufb02uence was positive or not Fig . 15 Participants were undecided if the example solution negatively in\ufb02uenced them 041003 - 10 / Vol . 132 , APRIL 2010 Transactions of the ASME Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm back - of - the envelope calculations . Results from this study clearly indicate that design \ufb01xation can be mitigated or reduced through these means . The de\ufb01xation group did produce more ideas , on average , than the \ufb01xation group . In addition , they repeated fewer features from the example solution and implemented a greater number of different energy categories than the \ufb01xation group . The results do not indicate which materials are most effective for de - \ufb01xation , but that this set , as a whole , is effective . Future studies will need to consider particular categories of de\ufb01xation materials and strategies . Unlike previous studies on design \ufb01xation , this study directed participants toward the use of analogy to break design \ufb01xation . Participants in past studies likely implemented analogies since analogy is a common and effective design strategy , but our study is much more literal about analogy use . Our study provides further but somewhat indirect support for the importance and impact of analogical reasoning in the design process . Additional research is needed to fully understand the type of information that eliminates design \ufb01xation , including the numer - ous representations currently existing in the design literature , and how these materials may be generated for novel design problems . Design methods currently exist for functional decomposition (cid:1) 63 , 66 , 79 (cid:2) and for analogies (cid:1) 26 , 62 , 64 , 74 , 85 (cid:2) . As more ap - proaches to reducing design \ufb01xation are discovered , new design methods will need to be developed to assist designers . Research Question 3 : How well do participant perceptions of design results correspond to quantitative assessment of the re - sults ? Consistent with other studies on idea generation (cid:1) 37 (cid:2) , partici - pants\u2019 perceptions of effectiveness during idea generation do not always match the quantitative outcome . Participants in the \ufb01xation group inaccurately believe , in general , that the example solution has a positive in\ufb02uence on their idea generation process , or they are unsure as to this in\ufb02uence . It is clear from the \ufb01xation results that the example reduces the number of ideas generated , but this is not perceived by the participants . The de\ufb01xation group felt that the example solution in\ufb02uenced them , but are not sure if it was positive or negative . The de\ufb01xation group also correctly perceived that they are assisted by the addi - tional information that is provided , but they do not feel strongly about this . By contrast , the solution data demonstrate that the de\ufb01xation group is strongly supported by the additional materials . These participant perception results strongly warn against their use as an accurate measurement of ideation effectiveness . 12 Conclusions This study evaluates design \ufb01xation , the use of mitigation strat - egies , and the perception of design \ufb01xation in a group of mostly engineering design faculty . Results show that design \ufb01xation is a dif\ufb01culty encountered even by this group , indicating the strength and importance of this effect in the design process . The partici - pants\u2019 perceptions of the effects of the provided \ufb01xation example solution and de\ufb01xation materials are generally not accurate , ex - cept with respect to de\ufb01xation materials . This result is not ex - pected for a group of individuals who study design . This incon - gruity in perception presents a unique obstacle to engineering design methods research since one of the simplest measures to obtain is the users\u2019 perceptions of the method\u2019s effectiveness . Based on this study and past work , the participants\u2019 evaluation of a method are frequently inconsistent with the quantitative mea - sures and not suitable for accurate evaluation or assessment . This study compared three groups of participants : (cid:3) 1 (cid:4) a control group , which only received the design problem , (cid:3) 2 (cid:4) a \ufb01xation group , which also was provided a negative example solution , and (cid:3) 3 (cid:4) a de\ufb01xation group , which , in addition to the negative solution , also received a set of materials to mitigate or reduce \ufb01xation . The example solution caused design \ufb01xation , as demonstrated by a reduction in the number of ideas generated , a greater number of design features from the example being reused and fewer catego - ries of energy sources considered . Consistent with prior studies , design \ufb01xation can be mitigated . The unique \ufb01xation mitigation materials , which included functions , energy sources , and analo - gies , increased the number of ideas generated . It also reduced the frequency of design solutions that were highly similar to the ex - ample , and increased the number of energy categories spanned . Fixation is a commonplace during the idea generation process and warrants much further investigation . Solutions presented or the individuals\u2019 own ideas can cause \ufb01xation , thus limiting the ideas that are considered . The situations that tend to increase de - sign \ufb01xation need to be identi\ufb01ed , and more approaches for miti - gating or reducing \ufb01xation should be created . Acknowledgment The authors would like to thank all participants of the 2008 NSF Workshop : Discussion on Individual and Team - Based Inno - vation and NSF for supporting the workshop and this paper under Grant No . CMMI0740529 . The authors speci\ufb01cally thank Judy Vance for proposing the workshop . Support is also acknowledged from University of Texas at Austin College of Engineering , the Cullen Trust Endowed Professorship in Engineering No . 1 , with Grant Nos . NSF SBE0738058 and NSF BCS0717957 . A version of this paper and some of the data was previously presented at the ICED Conference : Linsey , J . , Tseng , I . , Fu , K . , Cagan , J . , and Wood , K . , 2009 , \u201cReducing and Perceiving Design Fixation : Ini - tial Results from a NSF - Sponsored Workshop , \u201d Proceedings of the 2009 International Conference on Engineering Design , Stan - ford , CA . References (cid:1) 1 (cid:2) Simon , H . A . , 1968 , The Sciences of the Arti\ufb01cial , MIT , Cambridge , MA . (cid:1) 2 (cid:2) Antonsson , E . K . , and Cagan , J . , 2001 , Formal Engineering Design Synthesis , Cambridge University Press , Cambridge , England . (cid:1) 3 (cid:2) Chakrabarti , A . , 2002 , Engineering Design Synthesis : Understanding , Ap - proaches and Tools , Springer - Verlag , London . (cid:1) 4 (cid:2) Stone , R . B . , Wood , K . L . , and Crawford , R . H . , 2000 , \u201cA Heuristic Method to Identify Modules From a Functional Description of a Product , \u201d Des . Stud . , 21 , pp . 5 \u2013 31 . (cid:1) 5 (cid:2) Gero , J . S . , 1990 , \u201cDesign Prototypes : A Knowledge Representation Schema for Design , \u201d AI Mag . , 11 (cid:3) 4 (cid:4) , pp . 26 \u2013 36 . (cid:1) 6 (cid:2) Shah , J . J . , Smith , S . M . , and Vargas - Hernandez , N . , 2003 , \u201cMetrics for Mea - suring Ideation Effectiveness , \u201d Des . Stud . , 24 (cid:3) 2 (cid:4) , pp . 111 \u2013 134 . (cid:1) 7 (cid:2) Shah , J . J . , Vargas - Hern\u00e1ndez , N . , Summers , J . S . , and Kulkarni , S . , 2001 , \u201cCollaborative Sketching (cid:3) C - Sketch (cid:4) \u2014An Idea Generation Technique for En - gineering Design , \u201d J . Creat . Behav . , 35 (cid:3) 3 (cid:4) , pp . 168 \u2013 198 . (cid:1) 8 (cid:2) Yang , M . C . , and Epstein , D . J . , 2005 , \u201cA Study of Prototypes , Design Activ - ity , and Design Outcomes , \u201d Des . Stud . , 26 , pp . 649 \u2013 669 . (cid:1) 9 (cid:2) Christensen , B . T . , and Schunn , C . , 2007 , \u201cThe Relationship of Analogical Distance to Analogical Function and Pre - Inventive Structures : The Case of Engineering Design , \u201d Mem . Cognit . , 35 (cid:3) 1 (cid:4) , pp . 29 \u2013 38 . (cid:1) 10 (cid:2) Hirschi , N . W . , and Frey , D . D . , 2002 , \u201cCognition and Complexity : An Ex - periment on the Effect of Coupling in Parameter Design , \u201d Res . Eng . Des . , 12 , pp . 123 \u2013 131 . (cid:1) 11 (cid:2) Gero , J . S . , and Kannengiesser , U . , 2004 , \u201cThe Situated Function - Behavior - Structure Framework , \u201d Des . Stud . , 25 (cid:3) 4 (cid:4) , pp . 373 \u2013 391 . (cid:1) 12 (cid:2) Yang , M . C . , and Charm , J . G . , 2007 , \u201cAn Analysis of Sketching Skill and Its Role in Early Stage Engineering Design , \u201d ASME J . Mech . Des . , 129 , pp . 476 \u2013 482 . (cid:1) 13 (cid:2) Schunn , C . , Lovell , M . , Wang , Y . U . , and Yang , A . , 2008 , \u201cMeasuring Inno - vative Apples and Oranges : Towards More Robust and Ef\ufb01cient Measures of Product Innovation , \u201d Proceedings of the Design Creativity Conference . (cid:1) 14 (cid:2) Perttula , M . , and Sipila , P . , 2007 , \u201cThe Idea Exposure Paradigm in Design Idea Generation , \u201d J . Eng . Design , 18 (cid:3) 1 (cid:4) , pp . 93 \u2013 102 . (cid:1) 15 (cid:2) Jansson , D . , and Smith , S . , 1991 , \u201cDesign Fixation , \u201d Des . Stud . , 12 (cid:3) 1 (cid:4) , pp . 3 \u2013 11 . (cid:1) 16 (cid:2) Moss , J . , Kotovsky , K . , and Cagan , J . , 2007 , \u201cThe In\ufb02uence of Open Goals on the Acquisition of Problem - Relevant Information , \u201d J . Exp . Psychol . Learn . Mem . Cogn . , 33 (cid:3) 5 (cid:4) , pp . 876 \u2013 891 . (cid:1) 17 (cid:2) Tseng , I . , Moss , J . , Cagan , J . , and Kotovsky , K . , 2008 , \u201cThe Role of Timing and Analogical Similarity in the Stimulation of Idea Generation in Design , \u201d Des . Stud . , 29 (cid:3) 3 (cid:4) , pp . 203 \u2013 221 . (cid:1) 18 (cid:2) Linsey , J . , Laux , J . , Clauss , E . F . , Wood , K . , and Markman , A . , 2007 , \u201cIncreas - ing Innovation : A Trilogy of Experiments Towards a Design - by - Analogy Method , \u201d Proceedings of the ASME Design Theory and Methodology Confer - ence . (cid:1) 19 (cid:2) Linsey , J . , Wood , K . , and Markman , A . , 2008 , \u201cModality and Representation in Analogy , \u201d Artif . Intell . Eng . Des . Anal . Manuf . , 22 (cid:3) 02 (cid:4) , pp . 85 \u2013 100 . (cid:1) 20 (cid:2) Linsey , J . , Wood , K . , and Markman , A . , 2008 , \u201cIncreasing Innovation : Presen - Journal of Mechanical Design APRIL 2010 , Vol . 132 / 041003 - 11 Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm tation and Evaluation of the Wordtree Design - by - Analogy Method , \u201d Proceed - ings of the ASME IDETC Design Theory and Methodology Conference . (cid:1) 21 (cid:2) Schunn , C . D . , and Dunbar , K . , 1996 , \u201cPrimary , Analogy , and Awareness in Complex Reasoning , \u201d Mem . Cognit . , 24 (cid:3) 3 (cid:4) , pp . 271 \u2013 284 . (cid:1) 22 (cid:2) Dunbar , K . , 1997 , \u201cCreative Thought : An Investigation of Conceptual Struc - tures and Processes , \u201d How Scientists Think : On - Line Creativity and Concep - tual Change in Science , American Psychological Association , Washington , DC . (cid:1) 23 (cid:2) Eckert , C . M . , Stacey , M . , and Earl , C . , 2005 , \u201cReferences to Past Designs , \u201d Studying Designers \u201805 , Sydney , Australia , J . S . Gero and N . Bonnardel , eds . Key Centre of Design Computing and Cognition , pp . 3 \u2013 21 . (cid:1) 24 (cid:2) Mcadams , D . , and Wood , K . , 2002 , \u201cA Quantitative Similarity Metric for Design by Analogy , \u201d ASME J . Mech . Des . , 124 (cid:3) 2 (cid:4) , pp . 173 \u2013 182 . (cid:1) 25 (cid:2) Nagel , R . L . , Midha , P . A . , Tinsley , A . , Stone , R . B . , Mcadams , D . , and Shu , L . , 2008 , \u201cExploring the Use of Functional Models in Biomimetic Conceptual Design , \u201d J . Mech . Des . , 130 (cid:3) 12 (cid:4) , p . 121102 . (cid:1) 26 (cid:2) Chiu , I . , and Shu , L . H . , 2007 , \u201cBiomimetic Design Through Natural Lan - guage Analysis to Facilitate Cross - Domain Information Retrieval , \u201d Artif . In - tell . Eng . Des . Anal . Manuf . , 21 (cid:3) 1 (cid:4) , pp . 45 \u2013 59 . (cid:1) 27 (cid:2) Hacco , E . , and Shu , L . H . , 2002 , \u201cBiomimetic Concept Generation Applied to Design for Remanufacture , \u201d Proceedings of the ASME IDETC , Montreal , Quebec , Canada . (cid:1) 28 (cid:2) Hey , J . , Linsey , J . , Agogino , A . , and Wood , K . , 2008 , \u201cAnalogies and Meta - phors in Creative Design , \u201d Int . J . Eng . Educ . , 24 (cid:3) 2 (cid:4) , pp . 283 \u2013 294 . (cid:1) 29 (cid:2) Goel , A . K . , 1997 , \u201cDesign , Analogy , and Creativity , \u201d IEEE Expert , 12 (cid:3) 3 (cid:4) , pp . 62 \u2013 70 . (cid:1) 30 (cid:2) Campbell , M . , Cagan , J . , and Kotovsky , K . , 2003 , \u201cTheA - DesignApproach to Managing Automated Design Synthesis , \u201d Res . Eng . Des . , 14 (cid:3) 1 (cid:4) , pp . 12 \u2013 14 . (cid:1) 31 (cid:2) Olson , J . , and Cagan , J . , 2004 , \u201cInter - Agent Ties in Computational Con\ufb01gu - ration Design , \u201d Artif . Intell . Eng . Des . Anal . Manuf . , 18 (cid:3) 2 (cid:4) , pp . 135 \u2013 152 . (cid:1) 32 (cid:2) Davies , J . , Goel , A . , and Nersessian , N . , 2009 , \u201cA Computational Model of Visual Analogies in Design , \u201d Cognitive Systems Research , Special Issue on Analogies - Integrating Cognitive Abilities , 10 , pp . 204 \u2013 215 . (cid:1) 33 (cid:2) Leung , P . , Ishii , K . , Abell , J . , and Benson , J . , 2008 , \u201cDistributed System De - velopment Risk Analysis , \u201d J . Mech . Des . , 130 (cid:3) 5 (cid:4) , p . 051403 . (cid:1) 34 (cid:2) Paulus , P . B . , and Nijstad , B . A . , 2003 , Group Creativity : Innovation through Collaboration , Oxford University Press , New York . (cid:1) 35 (cid:2) Linsey , J . , Clauss , E . F . , Kurtoglu , T . , Murphy , J . T . , Wood , K . L . , and Mark - man , A . B . , \u201cAn Experimental Study of Group Idea Generation Techniques , \u201d Understanding the Roles of Idea Representation and Viewing Methods . (cid:1) 36 (cid:2) Yang , M . C . , and Jin , Y . , 2008 , \u201cAn Examination of Team Effectiveness in Distributed and Co - Located Engineering Teams , \u201d Int . J . Eng . Educ . , 24 (cid:3) 2 (cid:4) , pp . 400 \u2013 408 . (cid:1) 37 (cid:2) Paulus , P . B . , and Yang , H . C . , 2000 , \u201cIdea Generation in Groups : A Basis for Creativity in Organizations , \u201d Org . Behav . Hum . Decis . Process , 82 (cid:3) 1 (cid:4) , pp . 76 \u2013 87 . (cid:1) 38 (cid:2) Dong , A . , and Agogino , A . M . , 1997 , \u201cText Analysis for Constructing Design Representations , \u201d Artif . Intell . Eng . , 11 (cid:3) 2 (cid:4) , pp . 65 \u2013 75 . (cid:1) 39 (cid:2) Cronin , M . A . , and Weingart , L . R . , 2007 , \u201cRepresentational Gaps , Informa - tion Processing , and Con\ufb02ict in Functionally Diverse Teams , \u201d Acad . Manage . Rev . , 32 (cid:3) 3 (cid:4) , pp . 761 \u2013 773 . (cid:1) 40 (cid:2) Weingart , L . R . , Cronin , M . A . , Houser , C . S . , Cagan , J . , and Vogel , C . M . , 2005 , \u201cFunctional Diversity and Con\ufb02ict in Cross - Functional Product Devel - opment Teams : Considering Representational Gaps and Task Characteristics , \u201d Understanding Teams , IAP , Greenwich , CT . (cid:1) 41 (cid:2) Hanna , L . , and Cagan , J . , 2009 , \u201cEvolutionary Multi - Agent Systems : An Adaptive and Dynamic Approach to Optimization , \u201d J . Mech . Des . , 131 (cid:3) 1 (cid:4) , p . 011010 . (cid:1) 42 (cid:2) Dong , A . , Hill , A . W . , and Agogino , A . M . , 2004 , \u201cA Document Analysis Method for Characterizing Design Team Performance , \u201d J . Mech . Des . , 126 (cid:3) 3 (cid:4) , pp . 378 \u2013 385 . (cid:1) 43 (cid:2) Smith , S . M . , and Blankenship , S . E . , 1991 , \u201cIncubation and the Persistence of Fixation in Problem Solving , \u201d Am . J . Psychol . , 104 , pp . 61 \u2013 87 . (cid:1) 44 (cid:2) Ward , T . B . , 1994 , \u201cStructured Imagination : The Role of Category Structure in Exemplar Generation , \u201d Appl . Cognit . Psychol . , 27 (cid:3) 1 (cid:4) , pp . 1 \u2013 40 . (cid:1) 45 (cid:2) Marsh , R . L . , Ward , T . B . , and Landau , J . D . , 1999 , \u201cThe Inadvertent Use of Prior Knowledge in a Generative Cognitive Task , \u201d Mem . Cognit . , 27 (cid:3) 1 (cid:4) , pp . 94 \u2013 105 . (cid:1) 46 (cid:2) Jaarsveld , S . , and Leeuwen , C . , 2005 , \u201cSketches From a Design Process : Cre - ative Cognition Inferred From Intermediate Products , \u201d Cogn . Sci . , 29 , pp . 79 \u2013 101 . (cid:1) 47 (cid:2) Purcell , A . T . , and Gero , J . S . , 1996 , \u201cDesign and Other Types of Fixation , \u201d Des . Stud . , 17 (cid:3) 4 (cid:4) , pp . 363 \u2013 383 . (cid:1) 48 (cid:2) Chrysikou , E . G . , and Weisberg , R . W . , 2005 , \u201cFollowing the Wrong Foot - steps : Fixation Effects of Pictorial Examples in a Design Problem - Solving Task , \u201d J . Exp . Psychol . Learn . Mem . Cogn . , 31 (cid:3) 5 (cid:4) , pp . 1134 \u2013 1148 . (cid:1) 49 (cid:2) Casakin , H . , and Goldschmidt , G . , 1999 , \u201cExpertise and the Use of Visual Analogy : Implications for Design Education , \u201d Des . Stud . , 20 (cid:3) 2 (cid:4) , pp . 153 \u2013 175 . (cid:1) 50 (cid:2) Leclercq , P . , and Heylighen , A . , 2002 , Arti\ufb01cial Intelligence in Design \u201802 , 5 , 8 Analogies Per Hour . (cid:1) 51 (cid:2) Chiu , I . , and Shu , L . H . , 2007 , \u201cUsing Language as Related Stimuli for Con - cept Generation , \u201d Artif . Intell . Eng . Des . Anal . Manuf . , 21 (cid:3) 2 (cid:4) , pp . 103 \u2013 121 . (cid:1) 52 (cid:2) Ball , L . J . , Ormerod , T . C . , and Morley , N . J . , 2004 , \u201cSpontaneousAnalogising in Engineering Design : AComparativeAnalysis of Experts and Novices , \u201d Des . Stud . , 25 (cid:3) 5 (cid:4) , pp . 495 \u2013 508 . (cid:1) 53 (cid:2) Gick , M . L . , and Holyoak , K . J . , 1983 . (cid:1) 54 (cid:2) Gick , M . L . , and Patterson , K . , 1992 , \u201cDo Contrasting Examples Facilitate Schema Induction and Analogical Transfer ? , \u201d Can . J . Psychol . , 46 , pp . 539 \u2013 550 . (cid:1) 55 (cid:2) Kolodner , J . L . , 1997 , \u201cEducational Implications of Analogy : A View from Case - Based Reasoning , \u201d Am . Psychol . , 52 (cid:3) 1 (cid:4) , pp . 57 \u2013 66 . (cid:1) 56 (cid:2) Dahl , D . W . , and Moreau , P . , 2002 , \u201cThe In\ufb02uence and Value of Analogical Thinking During New Product Ideation , \u201d J . Mark . Res . , 39 (cid:3) 1 (cid:4) , pp . 47 \u2013 60 . (cid:1) 57 (cid:2) Marsh , R . L . , Landau , J . D . , and Hicks , J . L . , 1996 , \u201cHow Examples May (cid:3) and May Not (cid:4) Constrain Creativity , \u201d Mem . Cognit . , 24 (cid:3) 5 (cid:4) , pp . 669 \u2013 680 . (cid:1) 58 (cid:2) Gordon , W . J . J . , 1961 , Synectics : The Development of Creative Capacity , Harper , New York . (cid:1) 59 (cid:2) French , M . , 1996 , Conceptual Design , Springer - Verlag , London . (cid:1) 60 (cid:2) French , M . , 1988 , Invention and Evolution : Design in Nature and Engineering , Cambridge University Press , Cambridge , England . (cid:1) 61 (cid:2) Linsey , J . , Markman , A . , and Wood , K . , 2008 , \u201cWordtrees : A Method for Design - by - Analogy , \u201d Proceedings of the 2008 American Society for Engineer - ing Education Annual Conference , Pittsburg , PA . (cid:1) 62 (cid:2) Shu , L . H . , Hansen , H . N . , Gegeckaite , A . , Moon , J . , and Chan , C . , 2006 , \u201cCase Study in Biomimetic Design : Handling and Assembly of Microparts , \u201d Proceedings of theASME IDETC , International Design and Engineering Tech - nical Conferences . (cid:1) 63 (cid:2) Kurtoglu , T . , and Campbell , M . , 2008 , \u201cAutomated Synthesis of Electrome - chanical Design Con\ufb01gurations From Empirical Analysis of Function to Form Mapping , \u201d J . Eng . Des . , 18 (cid:3) 6 (cid:4) , pp . 83 \u2013 104 . (cid:1) 64 (cid:2) Chakrabarti , A . , Sarkar , P . , Leelavathamma , B . , and Nataraju , B . S . , 2005 , \u201cA Functional Representation for Aiding Biomimetic and Arti\ufb01cial Inspiration of New Ideas , \u201d Artif . Intell . Eng . Des . Anal . Manuf . , 19 (cid:3) 2 (cid:4) , pp . 113 \u2013 132 . (cid:1) 65 (cid:2) Smith , S . M . , Ward , T . B . , and Schumacher , J . S . , 1993 , \u201cConstraining Effects of Examples in a Creative Generation Task , \u201d Mem . Cognit . , 21 (cid:3) 6 (cid:4) , pp . 837 \u2013 845 . (cid:1) 66 (cid:2) Pahl , G . , and Beitz , W . , 1996 , Engineering Design\u2014A Systematic Approach , Springer , New York . (cid:1) 67 (cid:2) Otto , K . , and Wood , K . , 2001 , Product Design : Techniques in Reverse Engi - neering and New Product Development , Prentice - Hall , Upper Saddle River , NJ . (cid:1) 68 (cid:2) Ullman , D . , 1997 , The Mechanical Design Process , McGraw - Hill , New York . (cid:1) 69 (cid:2) Ulrich , K . , and Eppinger , S . , 2004 , Product Design and Development , McGraw - Hill , New York . (cid:1) 70 (cid:2) Ahmed , S . , and Christensen , B . T . , 2008 , \u201cUse of Analogies by Novice and Experienced Design Engineers , \u201d Proceedings of the ASME IDETC 2008 . (cid:1) 71 (cid:2) Mullen , B . , Johnson , C . , and Salas , E . , 1991 , \u201cProductivity Loss in Brain - storming Groups : A Meta - Analytic Integration , \u201d J . Appl . Soc . Psychol . , 12 (cid:3) 1 (cid:4) , pp . 3 \u2013 23 . (cid:1) 72 (cid:2) 2004 , Thinkspace : Peanut Sheller , March 2004 , http : / / www . thinkcycle . org / tc - space / tspace ? tspace _ id (cid:2) 41963 (cid:1) 73 (cid:2) Linsey , J . , Green , M . G . , Murphy , J . T . , Wood , K . L . , and Markman , A . B . , 2005 , \u201cCollaborating to Success : An Experimental Study of Group Idea Gen - eration Techniques , \u201d Proceedings of the ASME Design Theory and Methodol - ogy Conference . (cid:1) 74 (cid:2) Linsey , J . , Markman , A . , and Wood , K . , 2008 , \u201cWordtrees : A Method for Design - by - Analogy , \u201d Proceedings of the 2008 American Society for Engineer - ing Education Annual Conference . (cid:1) 75 (cid:2) Dugosh , K . L . , and Paulus , P . B . , 2005 , \u201cCognitive and Social Comparison Processes in Brainstorming , \u201d J . Exp . Soc . Psychol . , 41 , pp . 313 \u2013 320 . (cid:1) 76 (cid:2) Shah , J . J . , Kulkarni , S . V . , and Vargas - Hern\u00e1ndez , N . , 2000 , \u201cEvaluation of Idea Generation Methods for Conceptual Design : Effectiveness Metrics and Design of Experiments , \u201d ASME J . Mech . Des . , 122 , pp . 377 \u2013 384 . (cid:1) 77 (cid:2) Linsey , J . S . , Green , M . G . , Murphy , J . T . , Wood , K . L . , and Markman , A . B . , 2005 , \u201cCollaborating to Success : An Experimental Study of Group Idea Gen - eration Techniques , \u201d Proceedings of the ASME Design Theory and Methodol - ogy Conference . (cid:1) 78 (cid:2) Stone , R . , and Wood , K . , 2000 , \u201cDevelopment of a Functional Basis for De - sign , \u201d J . Mech . Des . , 122 (cid:3) 4 (cid:4) , pp . 359 \u2013 370 . (cid:1) 79 (cid:2) Hirtz , J . , Stone , R . B . , and Mcadams , D . A . , 2002 , \u201cA Functional Basis for Engineering Design : Reconciling and Evolving Previous Efforts , \u201d Res . Eng . Des . , 13 , pp . 65 \u2013 82 . (cid:1) 80 (cid:2) Clark - Carter , D . , 1997 , Doing Quantitative Psychological Research : From De - sign to Report , Psychology Press , United Kingdom . (cid:1) 81 (cid:2) Howell , D . C . , 2002 , Statistical Methods for Psychology , Duxbury , Paci\ufb01c Grove , CA . (cid:1) 82 (cid:2) Howell , D . C . , 2007 , Resampling Statistics : Randomization and the Bootstrap , 3 / 2009 , http : / / www . uvm . edu / ~ dhowell / StatPages / Resampling / Resampling . html (cid:1) 83 (cid:2) Howell , D . C . , 2000 , Resampling Procedures (cid:3) Software (cid:4) , 3 / 2009 , http : / / www . uvm . edu / ~ dhowell / StatPages / Resampling / Resampling . html (cid:1) 84 (cid:2) Good , P . I . , 2005 , Resampling Methods : A Practical Guide to Data Analysis , Birkhauser , Boston , MA . (cid:1) 85 (cid:2) Chakrabarti , A . , Sarkar , P . , Leelavathamma , B . , and Nataraju , B . S . , 2005 , \u201cA Behavioural Model for Representing Biological and Arti\ufb01cial Systems for In - spiring Novel Designs , \u201d Proceedings of the International Conference on Engi - neering Design . 041003 - 12 / Vol . 132 , APRIL 2010 Transactions of the ASME Downloaded 26 Mar 2012 to 130 . 49 . 139 . 178 . Redistribution subject to ASME license or copyright ; see http : / / www . asme . org / terms / Terms _ Use . cfm",
    "johnston2018novel": "* Forcorrespondence : goode @ brandeis . edu \u2020 These authors contributed equally to this work Competing interests : The authors declare that no competing interests exist . Funding : See page 24 Received : 21 August 2018 Accepted : 22 October 2018 Published : 23 October 2018 Reviewing editor : Anna Akhmanova , Utrecht University , Netherlands Copyright Johnston et al . This article is distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use and redistribution provided that the original author and source are credited . A novel mode of capping protein - regulation by twinfilin Adam B Johnston 1\u2020 , Denise M Hilton 1\u2020 , Patrick McConnell 2 , Britney Johnson 3 , Meghan T Harris 1 , Avital Simone 1 , Gaya K Amarasinghe 3 , John A Cooper 2 , Bruce L Goode 1 * 1 Department of Biology , Rosenstiel Basic Medical Science Research Center , Brandeis University , Waltham , United States ; 2 Department of Biochemistry and Molecular Biophysics , Washington University , St Louis , United states ; 3 Department of Pathology and Immunology , Washington University , St Louis , United States Abstract Cellular actin assembly is controlled at the barbed ends of actin filaments , where capping protein ( CP ) limits polymerization . Twinfilin is a conserved in vivo binding partner of CP , yet the significance of this interaction has remained a mystery . Here , we discover that the C - terminal tail of Twinfilin harbors a CP - interacting ( CPI ) motif , identifying it as a novel CPI - motif protein . Twinfilin and the CPI - motif protein CARMIL have overlapping binding sites on CP . Further , Twinfilin binds competitively with CARMIL to CP , protecting CP from barbed - end displacement by CARMIL . Twinfilin also accelerates dissociation of the CP inhibitor V - 1 , restoring CP to an active capping state . Knockdowns of Twinfilin and CP each cause similar defects in cell morphology , and elevated Twinfilin expression rescues defects caused by CARMIL hyperactivity . Together , these observations define Twinfilin as the first \u2018pro - capping\u2019 ligand of CP and lead us to propose important revisions to our understanding of the CP regulatory cycle . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 001 Introduction Assembly of cellular actin structures with distinct architectural and dynamic properties requires the convergence and coordination of numerous actin assembly , stabilization , and disassembly mecha - nisms . Although our understanding of the functions and mechanisms of individual actin - binding pro - teins has grown tremendously , there is a need to consider more deeply how seemingly disparate and sometimes competing factors work together in vivo and take on new mechanistic roles within more complex mixtures . One particularly enigmatic example is the interaction of Twinfilin with Cap - ping Protein ( CP ) . These two conserved proteins directly interact with high - affinity , and yet have seemingly opposite effects on the barbed ends of actin filaments . Twinfilin is one of five proteins in the Actin Depolymerization Factor - Homology ( ADF - H ) domain family , of which ADF / Cofilin is the founding member ( Poukkula et al . , 2011 ) . Twinfilin is unique among the members of this family in containing two ADF - H domains , which are joined by a small linker region and followed by a short C - terminal tail . Initial biochemical studies categorized Twinfilin as an actin monomer sequestering factor because of its high affinity for ADP - bound G - actin and abil - ity to inhibit subunit addition to either end of the filament ( Goode et al . , 1998 ; Vartiainen et al . , 2000 ; Wahlstro\u00a8m et al . , 2001 ) . However , mouse Twinfilin was later shown to interact directly with the barbed ends of actin filaments ( Helfer et al . , 2006 ; Paavilainen et al . , 2007 ) , and more recently yeast Twinfilin was shown to accelerate depolymerization at actin filament ends ( Johnston et al . , 2015 ) . Alone , yeast Twinfilin enhanced barbed end depolymerization by 3 - fold through a processive filament end - attachment mechanism . Further , in conjunction with Srv2 / CAP ( cyclase - associated pro - tein ) , yeast Twinfilin increased the rate of pointed - end depolymerization by over 15 - fold Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 1 of 28 RESEARCH ARTICLE ( Johnston et al . , 2015 ) . More recently , it was shown that mouse Twinfilin isoforms accelerate barbed end depolymerization , similar to yeast Twinfilin , but do not induce robust pointed end depo - lymerization in conjunction with Srv2 / CAP ( Hilton et al . , 2018 ) . Collectively , these studies highlight the biological significance of Twinfilin . The conserved barbed - end effects of Twinfilin are particularly interesting given that both yeast and mammalian Twinfilins bind to CP ( Falck et al . , 2004 ; Palmgren et al . , 2001 ) . Further , a barbed - end regulatory role for Twinfilin is suggested by its localization to the tips of stereocilia and filopo - dia , and to the barbed ends of Drosophila actin bristles ( Peng et al . , 2009 ; Rzadzinska et al . , 2009 ; Wahlstro\u00a8m et al . , 2001 ) . In addition , Twinfilin localizes to endocytic actin patches in yeast , and to lamellipodia and cell - cell junctions in animal cells ( Goode et al . , 1998 ; Vartiainen et al . , 2000 ) . Twinfilin\u2019s localization to cortical actin patches in yeast is dependent on its interaction with CP ( Palmgren et al . , 2001 ) . In both yeast and mammals , this interaction is mediated by conserved sequences in the C - terminal tail region of Twinfilin ( Falck et al . , 2004 ) . Despite the high affinity of the Twinfilin - CP interaction ( K d ~ 10 nM for the yeast homologs [ Poukkula et al . , 2011 ] ) , studies have revealed no significant effects of Twinfilin on the barbed end capping activity of CP in vitro , and reciprocally , no obvious effect of CP on Twinfilin interactions with ADP - actin monomers ( Falck et al . , 2004 ) . Thus , the functional significance of the Twinfilin - CP interaction has remained highly enigmatic . CP is an obligate heterodimer , consisting of alpha and beta subunits , and binds stably to the barbed ends of actin filaments to block subunit addition and loss . CP is ubiquitous and highly eLife digest Plant and animal cells are supported by skeleton - like structures that can grow and shrink beneath the cell membrane , pushing and pulling on the edges of the cell . This scaffolding network \u2013 known as the cytoskeleton \u2013 contains long strands , or filaments , made from many identical copies of a protein called actin . The shape of the actin proteins allows them to slot together , end - to - end , and allows the strands to grow and shrink on - demand . When the strands are the correct length , the cell caps the growing ends with a protein known as Capping Protein . This helps to stabilize the cell\u2019s skeleton , preventing the strands from getting any longer , or any shorter . Proteins that interfere with the activity of Capping Protein allow the actin strands to grow or shrink . Some , like a protein called V - 1 , attach to Capping Protein and get in the way so that it cannot sit on the ends of the actin strands . Others , like CARMIL , bind to Capping Protein and change its shape , making it more likely to fall off the strands . So far , no one had found a partner that helps Capping Protein limit the growth of the actin cytoskeleton . A protein called Twinfilin often appears alongside Capping Protein , but the two proteins seemed to have no influence on each other , and had what appeared to be different roles . Whilst Capping Protein blocks growth and stabilizes actin strands , Twinfilin speeds up their disassembly at their ends . But Johnston , Hilton et al . now reveal that the two proteins actually work together . Twinfilin helps Capping Protein resist the effects of CARMIL and V - 1 , and Capping Protein puts Twinfilin at the end of the strand . Thus , when Capping Protein is finally removed by CARMIL , Twinfilin carries on with disassembling the actin strands . The tail of the Twinfilin protein looks like part of the CARMIL protein , suggesting that they might interact with Capping Protein in the same way . Attaching a fluorescent tag to the Twinfilin tail revealed that the two proteins compete to attach to the same part of the Capping Protein . When mouse cells produced extra Twinfilin , it blocked the effects of CARMIL , helping to grow the actin strands . V - 1 attaches to Capping Protein in a different place , but Twinfilin was also able to interfere with its activity . When Twinfilin attached to the CARMIL binding site , it did not directly block V - 1 binding , but it made the protein more likely to fall off . Understanding how the actin cytoskeleton moves is a key question in cell biology , but it also has applications in medicine . Twinfilin plays a role in the spread of certain blood cancer cells , and in the formation of elaborate structures in the inner ear that help us hear . Understanding how Twinfilin and Capping Protein interact could open paths to new therapies for a range of medical conditions . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 002 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 2 of 28 Research article Cell Biology conserved across eukaryotes , and has universal roles in controlling the assembly of actin networks that drive cell morphogenesis and cell motility ( Cooper and Sept , 2008 ; Hart and Cooper , 1999 ; Mejillano et al . , 2004 ; Schafer et al . , 1994 ; Schafer et al . , 1995 ) . In vitro , CP binds to the barbed ends of actin filaments with sub - nanomolar affinity , and dissociates from barbed ends very slowly ( half - life of ~ 30 min ) ( Schafer et al . , 1996 ) . Given the relatively high abundance of CP in the cytosol ( 1 \u2013 3 m M ) and the strength of its interactions with barbed ends ( Cooper and Sept , 2008 ) , it is not surprising that cells have evolved a number of regulatory mechanisms to spatiotemporally restrict CP activity . Cellular protein inhibitors of CP broadly fall into two classes : steric inhibitors and allosteric inhibi - tors . Steric inhibitors , which include V - 1 / myotrophin , bind to CP in a manner that physically obstructs its association with barbed ends ( Bhattacharya et al . , 2006 ; Kim et al . , 2007 ; Schafer et al . , 1996 ) . V - 1 is a highly abundant 13 kDa protein that binds CP with a K d ~ 40 nM and sterically blocks its abil - ity to bind barbed ends ( Bhattacharya et al . , 2006 ; Taoka et al . , 2003 ) . Notably , however , V - 1 does not catalyze dissociation of CP from barbed ends ( Bhattacharya et al . , 2006 ) . In contrast , allo - steric inhibitors induce conformational changes in CP that catalyze its dissociation from barbed ends ( \u2018uncapping\u2019 or \u2018displacing\u2019 CP ) , and also decrease but do not abolish its ability to bind barbed ends . The major class of allosteric inhibitors is the capping protein interaction ( CPI ) motif family of pro - teins ( Edwards et al . , 2014 ) . The founding and best characterized member of the CPI family is CAR - MIL ( Capping Protein , ARP2 / 3 and Myosin I linker ) , which is conserved across metazoans ( Stark et al . , 2017 ) . CARMIL catalyzes CP dissociation from barbed ends , reducing CP\u2019s affinity for barbed ends by ~ 100 fold , transforming it into a transient capper ( Fujiwara et al . , 2014 ; Stark et al . , 2017 ; Uruno et al . , 2006 ; Yang et al . , 2005 ) . CARMIL localizes to the leading - edge plasma membrane , where it promotes cell migration through direct interactions with CP ( Fujiwara et al . , 2014 ; Liang et al . , 2009 ; Stark et al . , 2017 ; Yang et al . , 2005 ) . Other proteins with CPI motifs include CD2AP , CKIP - 1 , CapZIP , CIN85 , and WASHCAP ( FAM21 ) ; their roles in regu - lating CP are less well understood . CPI - motif proteins share a common mode of interaction with CP , but are otherwise unrelated to each other ( Edwards et al . , 2014 ; Hernandez - Valladares et al . , 2010 ) . To date , binding partners of CP that antagonize its inhibitors , and thus function as \u2018pro - cap - ping\u2019 factors , have not been reported . Here , we uncover a novel role for Twinfilin in protecting CP from the negative regulatory effects of V - 1 and CARMIL , and thus promoting actin filament capping . These and other data lead us to propose important revisions to current models for the CP regulatory cycle . Results CP inhibits mTwf1 - mediated depolymerization by capping barbed ends Because CP binding proteins have been studied predominantly in mammalian systems , we focused our investigation on mouse rather than yeast CP and Twinfilin . Mutagenesis on the yeast Twinfilin tail previously identified a mutant , twf1 - 11 , that targets a cluster of positively charged residues ( R328A , K329A , R330A , R331A ) necessary for binding CP ( Falck et al . , 2004 ) . While truncations of the C - ter - minal tail in mouse Twinfilin ( mTwf1 ) also disrupt CP binding , the residues involved have not yet been defined . We therefore first sought to generate a specific mutant in mTwf1 that disrupts the interaction , analogous to yeast twf1 - 11 . An alignment of the three mouse and three human Twinfilin isoforms , along with the single Twinfilin genes expressed in S . cerevisiae and D . melanogaster ( Figure 1A ) , revealed a region that includes two of the basic residues mutated in the yeast twf1 - 11 mutant . We mutated these two residues in mTwf1 , changing them to alanines , to produce mTwf1 - 11 ( K332A , R333A ) . To quantify binding of mTwf1 to CP , we performed fluorescence anisotropy assays using a mTwf1 tail peptide ( 317 - 350 ) labeled at its N - terminus with HiLyte488 . The mTwf1 tail pep - tide displayed high affinity , concentration - dependent binding to CP a 1 b 2 , a major non - muscle iso - form of CP in mammalian cells ( Figure 1B ) . Moreover , full - length mTwf1 protein ( unlabeled ) competed with the labeled mTwf1 tail for CP binding , whereas full - length mTwf1 - 11 ( unlabeled ) did not ( Figure 1C ) . Thus , the mTwf1 - 11 mutant effectively uncouples mTwf1 binding to CP . Using mTwf1 - 11 , we addressed how CP binding affects Twinfilin\u2019s actin depolymerization activi - ties in total internal reflection fluorescence ( TIRF ) microscopy assays , by directly observing Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 3 of 28 Research article Cell Biology depolymerization at actin filament barbed ends in real time . In agreement with previous observa - tions using yeast and mouse Twinfilin ( Hilton et al . , 2018 ; Johnston et al . , 2015 ) , 1 m M mTwf1 accelerated barbed end depolymerization by 2 \u2013 3 fold compared to control reactions ( Figure 1D ) , and the addition of CP blocked this effect ( Figure 1E ) . Further , mTwf1 - 11 exhibited a similar rate ( Figure 1D ) , indicating that this mutant has wild type depolymerization activity , and thus separates Twinfilin\u2019s ability to bind CP from its ability to promote barbed - end depolymerization . Interestingly , the addition of CP was still able to block barbed - end depolymerization by mTwf1 - 11 ( Figure 1E ) . These observations suggest that CP sterically blocks mTwf1 access to barbed ends , independent of its direct interaction with mTwf1 . However , this left open the question of whether CP binding to mTwf1 might alter its mechanism of depolymerization independent of blocking the barbed end . To Figure 1 . Barbed end capping by Capping Protein inhibits Twinfilin1 - mediated depolymerization . ( A ) Mouse Twinfilin - 1 ( mTwf1 ) domain organization : ADF - H , actin depolymerization factor homology domain ; L , linker ; T , tail . Sequence alignment of tail regions of Twinfilin isoforms from different species with boxed region highlighting conservation of residues critical for binding to Capping Protein ( CP ) . mTwf1 - 11 carries a mutation in the tail region ( KR332 , 333AA ) that disrupts binding to CP . ( B ) Fluorescence anisotropy measurement of 100 nM HiLyte488 - labeled mTwf1 tail peptide mixed with increasing concentrations of the indicated CP construct . ( C ) Fluorescence anisotropy measurement of 100 nM HiLyte488 - labeled mTwf1 tail peptide incubated in the presence 1 m M CP and increasing concentrations of either mTwf1 or mTwf1 - 11 . Anisotropy values for each condition averaged from three independent experiments . ( D , E ) Rates of barbed end depolymerization ( subunits s (cid:0) 1 ) induced by 1 m M of the indicated mouse Twinfilin , in the ( D ) absence or ( E ) presence of 10 nM CP , determined from TIRF assays . Rates for each condition averaged from at least five filaments in each of two independent experiments . From left to right : ( D ) n = 19 , 26 , and 15 and mean depolymerization rates 1 . 13 , 2 . 784 and 2 . 81 subunits s (cid:0) 1 ; ( E ) n = 13 , 15 , and 20 and mean depolymerization rates 1 . 13 , 2 . 784 and 2 . 81 subunits s (cid:0) 1 . ( F ) Rates of barbed end depolymerization ( subunits s (cid:0) 1 ) induced by 1 m M mTwf1 , in the absence or presence of 1 m M of the indicated CP construct , determined from TIRF assays . Rates for each condition averaged from at least five filaments from at least one experiment . From left to right n = 21 , 25 , 6 , and 10 ; mean depolymerization rates 1 . 45 , 2 . 991 , 0 . 11 , and 3 . 58 subunits s (cid:0) 1 . ( G ) Summary of barbed end depolymerization activity of mTwf1 constructs in combination with different CP constructs determined from TIRF assays ( as in D , E , F ) . Error bars , s . e . m . * * * * p (cid:20) 0 . 0001 , n . s . p > 0 . 05 by one - way ANOVA with Tukey post hoc test . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 003 The following video is available for figure 1 : Figure 1\u2014video 1 . Supporting data for Figure 1F . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 015 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 4 of 28 Research article Cell Biology address this possibility , we utilized a CP mutant , CP aD 28 , which truncates the C - terminal tentacle of the alpha subunit , severely inhibiting capping activity ( Kim et al . , 2010 ) . Importantly , in binding assays the mTwf1 tail interacted equally well with wild - type CP and CP aD 28 , demonstrating that this mutant binds normally to mTwf1 ( Figure 1B ) . In TIRF assays , equimolar amounts of CP aD 28 did not significantly alter mTwf1 depolymerization activity ( Figure 1F ; Figure 1\u2014video 1 ; also summarized in Figure 1G ) , suggesting that while CP blocks Twinfilin access to barbed ends , Twinfilin - CP direct interaction does not alter Twinfilin depolymerization activity . The twinfilin tail competes with CARMIL CPI motif for binding to CP Given that CP binding does not affect Twinfilin\u2019s depolymerization activity , or other known activities of Twinfilin ( Falck et al . , 2004 ; Johnston et al . , 2015 ; Palmgren et al . , 2001 ) , we next considered whether Twinfilin binding might influence CP functions in the presence of known regulators of CP . We were particularly interested in how Twinfilin might impact the regulation of CP by CPI - motif pro - teins such as CARMIL , since we noticed that the C - terminal tail regions of evolutionarily diverse Twinfilins share sequence homology with the CPI motifs of several CPI family proteins ( Figure 2A ) . The consensus CPI motif is 17 - amino acids long , with some additional contacts contributed from out - side this motif , and tolerates significant divergence across the CPI - motif family ( Edwards et al . , 2014 ; Hernandez - Valladares et al . , 2010 ) . As an initial test , we used a mutant of CP , CP ( RY ) , which alters two surface residues on the beta subunit ( R15A , Y79A ) that make essential contacts with CPI - motif proteins ( Edwards et al . , 2014 ; Hernandez - Valladares et al . , 2010 ) . The CP ( RY ) mutant is insensitive to inhibition and uncapping by CARMIL and disrupts binding with at least two other CPI - motif proteins , CD2AP and WASHCAP ( FAM21 ) ( Edwards et al . , 2015 ) . In fluorescence anisotropy binding assays , we observed that the CP ( RY ) mutant has approximately 20 - fold reduced affinity for mTwf1 tail compared to wild type CP ( Figure 2B ) . These data are consistent with mTwf1 and CPI - motif proteins sharing at least partially overlapping binding sites on CP . In addition , we asked whether introducing a mutation in the mTwf1 tail peptide at a conserved residue in CPI consensus sequences would alter binding to CP ( Lys 325 in mTwf1 ; see red asterisk , sequence alignment in Figure 2A ) . In fluorescence anisotropy binding assays , we compared the abilities of wild - type and mutant ( K325A ) mTwf1 tail peptides to compete with labeled mTwf1 tail peptide for CP binding . This analysis revealed an ~ 30 fold reduction in binding affinity for the mutant ( K325A ) mTwf1 tail peptide compared to wild type peptide ( Figure 2C ) . We next asked whether the CP - binding region ( CBR ) of CARMIL1 ( residues 964 \u2013 1078 ) competes with mTwf1 tail for binding to CP . We observed that unlabeled CBR peptide competed with the fluorescent mTwf1 tail probe for CP binding ( Figure 2D ) . These results indicate that CARMIL and mTwf1 directly compete for binding CP . Next , we more narrowly defined the region of CARMIL that competes with mTwf1 by using peptides that divide the CBR into its two conserved components , the CPI motif ( 969 \u2013 1005 ) and the CARMIL - specific interaction ( CSI ) motif ( 1019 \u2013 1037 ) . The CSI makes additional contacts with CP , but is found only in CARMIL family members , and not in other CPI - motif proteins ( Edwards et al . , 2014 ) . As expected based on Twinfilin\u2019s sequence similarity to CPI motifs , only the CPI - motif peptide and not the CSI peptide competed with mTwf1 tail for CP binding ( Figure 2D ) . Together , these results suggest that Twinfilin is a divergent CPI - motif protein and has important implications for CP regulation in cells ( see Discussion ) . Twinfilin attenuates CARMIL - mediated displacement of CP from barbed ends Given that CARMIL and Twinfilin compete for binding to CP , we asked whether mTwf1 affects CAR - MIL\u2019s ability to displace CP from barbed ends . We addressed this question in pyrene actin assembly assays , where actin polymerization was initiated at time zero in the presence of CP and increasing concentrations of mTwf1 , and after 400 s CARMIL1 CBR was spiked into the reaction . CARMIL1 alone ( no mTwf1 ) strongly induced uncapping , leading to the rapid polymerization of previously - capped filament seeds ( Figure 3A ) . However , increasing concentrations of mTwf1 attenuated CAR - MIL\u2019s uncapping effects ( Figure 3A ) . These results are consistent with mTwf1 competing with CAR - MIL for binding CP , and thereby blocking uncapping . To more directly observe mTwf1 effects on CARMIL - induced uncapping of barbed ends , we used TIRF microscopy . In these experiments , we used fluorescently labeled SNAP - tagged CP ( SNAP - 649 - Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 5 of 28 Research article Cell Biology Figure 2 . Twinfilin is a Capping Protein Interaction ( CPI ) - motif protein that competes with CARMIL for binding Capping Protein . ( A ) CARMIL domain organization : PH , pleckstrin - homology domain ; L , linker ; N - cap ( N ) , LRR , leucine - rich repeat domain ; C , C - cap ; HD , helical dimerization domain ; CBR , Capping Protein binding domain , consisting of CPI , Capping Protein interaction domain , and CSI , CARMIL - specific interaction sequence ; MBD , membrane binding domain ; PRD , proline - rich domain . Alignment between the Capping Protein Interaction ( CPI ) motif consensus sequence , and the CPI regions of H . sapiens ( H . s . ) CARMIL1 ( UniProtKB Q5VZK9 . 1 ) , CARMIL2 ( UniProtKB Q6F5E8 . 2 ) , CARMIL3 ( UniProtKB Q8ND23 . 2 ) , CKIP1 UniProtKB Q53GL0 . 2 ) , CD2AP ( CBI NP _ 036252 . 1 ) , WASHCAP ( Fam21 ) ( UniProtKB Q9Y4E1 . 3 ) , CapZIP ( CBI NP _ 443094 . 3 ) , CIN85 ( UniProtKB Q96B97 . 2 ) , and the tail sequences of Twinfilin homologs from D . melanogaster ( D . m ) , C . lectularius ( C . l . ) , S . cerevisiae ( S . c . ) , O . Taurus ( O . t . ) , S . litura ( S . l . ) , D . rerio ( D . r . ) , H . sapiens ( H . s . ) , and M . musculus ( M . m . ) . Twinfilin isoforms ( D . m . Twf1 UniProtKB NP _ 650338 , C . l . Twf1 UniProtKB XP _ 014258437 . 1 , S . c . Twf1 GenBank GAX68393 . 1 , O . t . Twf1 XP _ 022917989 . 1 , S . l . Twf1 XP _ 022816377 . 1 , D . r . Twf1 AAH67638 . 1 , H . s . Twf1 UniProtKB NO _ 001229326 . 1 , and M . m . Twf1 GenBank AAH15081 . 1 ) . Amino acid color coding illustrates side chain chemistry similarities . The asterisk marks the residue we mutated in mTwf1 in panel . ( C ) The alignments were generated using the MAFFT algorithm in the DNASTAR Lasergene Suite / MegAlign Pro application ( MegAlign Pro . Version 15 . 0 . DNASTAR . Madison , WI . ) . ( B ) Fluorescence anisotropy measurement of 60 nM HiLyte488 - labeled mTwf1 tail peptide mixed with increasing concentrations of the indicated CP construct . ( C ) Fluorescence anisotropy measurement of 40 nM TAMRA - labeled mTwf1 tail peptide incubated with 1 m M CP and different concentrations of wild type and mutant mTwf1 tail peptides . ( D ) Fluorescence anisotropy measurement of 60 nM HiLyte488 - labeled mTwf1 tail peptide incubated in the presence of 240 nM CP and increasing concentrations of the indicated CARMIL fragment ( CBR , CSI , or CPI ) . CSI failed to compete with HiLyte 488 - mTwf1 tail peptide at the concentrations tested . Anisotropy values for each condition were averaged from three independent experiments . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 004 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 6 of 28 Research article Cell Biology Figure 3 . Direct interactions of Twinfilin with Capping Protein attenuate CARMIL - mediated uncapping . ( A ) Bulk fluorescence assays comparing the rates of actin assembly in the presence of 25 nM muscle Capping Protein ( CP a 1 b 1 ) and increasing concentrations of mTwf1 . To initiate uncapping , 250 nM CBR fragment of CARMIL ( see schematic , Figure 2A ) was spiked into the reaction at 400 s . Data shown are representative curves from experiments repeated three independent times . ( B ) Representative time - lapse images from TIRF microscopy assays monitoring the displacement of labeled CP from barbed ends . Filaments were first polymerized and tethered using 1 m M actin ( 10 % OG - labeled , 0 . 5 % biotin \u2013 actin ) , then capped at their barbed ends by flowing in SNAP - 649 - CP ( 100 % labeled ) . Next , 50 nM CBR fragment of CARMIL and different concentrations of mTwf1 were flowed in , and CP dissociation was monitored over time . Scale bar , 5 m m . ( C ) Quantification of the percentage of filaments retaining CP at the barbed ends in the presence of 50 nM CBR fragment of CARMIL and variable concentrations of mTwf1 , determined from TIRF reactions as in ( B ) . Control curve , buffer alone ( no CBR or mTwf1 ) . n > 45 events measured from at least two independent experiments . ( D ) Representative time - lapse images from TIRF microscopy assays monitoring CP displacement from barbed ends , analyzed as in ( B ) , except using 1 m M mTwf1 - 11 instead of mTwf1 . n > 45 events measured from at least two independent experiments . ( E ) Quantification of the percentage of filaments retaining CP at the barbed end in the presence of 50 nM CBR fragment of CARMIL and different concentrations of mTwf1 - 11 , determined from TIRF assays as in ( D ) . n > 45 events measured from at least two independent experiments . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 005 The following figure supplement is available for figure 3 : Figure supplement 1 . Supporting data for Figure 3 showing that multiple Twinfilin isoforms antagonize CARMIL uncapping of barbed ends . Figure 3 continued on next page Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 7 of 28 Research article Cell Biology CP ; 100 % labeled ) to monitor lifetimes of CP molecules on filament barbed ends ( Bombardier et al . , 2015 ) . Filaments were first polymerized to a desired length ( ~ 10 m m ) and then capped by flowing in SNAP - 649 - CP . Free CP was washed out , and then proteins of interest ( or con - trol buffer ) were flowed in . Capped filaments were identified in the field of view prior to flow - in , and then monitored after flow - in to measure the dwell time of SNAP - 649 - CP . As expected , in the absence of other factors , SNAP - 649 - CP had a long dwell time , remaining on barbed ends for tens of minutes ( Figure 3B and C ) . However , when CARMIL1 CBR was introduced , this led to the rapid dis - placement of SNAP - 649 - CP , with complete loss of CP from barbed ends by 100 s ( Figure 3B and C ) . The addition of mTwf1 with CARMIL1 CBR attenuated the uncapping effects in a concentration - dependent manner ( Figure 3B and C ) . Further , this attenuation required direct interactions between Twinfilin and CP , as mTwf1 - 11 failed to protect CP from CARMIL uncapping ( Figure 3D and E and Figure 3\u2014figure supplement 1 ) . Similar effects were observed for the other major isoform of mouse Twinfilin that is expressed in non - muscle cells , mTwf2a ( Figure 3\u2014figure supplement 1 ) ( Nevalainen et al . , 2011 ; Vartiainen et al . , 2003 ) . Twinfilin accelerates the dissociation of V - 1 from CP We next considered whether Twinfilin binding to CP might affect the activities of CP inhibitor V - 1 / myotrophin , which is distinct from CPI - motif proteins in its mode of CP interaction . Unlike CARMIL , V - 1 does not displace CP from barbed ends ; instead , it sequesters CP and blocks it from binding fila - ment ends ( Bhattacharya et al . , 2006 ; Jung et al . , 2016 ; Taoka et al . , 2003 ) . In contrast to the CARMIL binding site on CP , which partially encircles the \u2018stalk\u2019 of the CP heterodimer ( Hernandez - Valladares et al . , 2010 ; Johnson et al . , 2018 ; Zwolak et al . , 2010 ) , V - 1 interacts with CP on the opposite face , sterically blocking binding to the filament end ( Johnson et al . , 2018 ; Takeda et al . , 2010 ; Zwolak et al . , 2010 ) . To test how Twinfilin might affect the interaction of CP with V - 1 , we used pyrene - actin seeded elongation assays ( Figure 4A ) . As expected , filament seeds pre - incubated with CP and then mixed with pyrene - actin monomers displayed minimal growth , whereas the addi - tion of V - 1 restored actin assembly to uncapped levels . Somewhat to our surprise , the further addi - tion of mTwf1 suppressed V - 1\u2019s effects , restoring capping activity , while mTwf1 - 11 had no effect ( Figure 4A and B ) . These effects were unexpected given the above - mentioned differences in Twinfi - lin\u2019s predicted and V - 1\u2019s known binding sites on CP , and our observation that even high concentra - tions of V - 1 ( 1000 - fold excess to mTwf1 tail probe ) fail to compete with mTwf1 for CP binding in anisotropy assays ( Figure 4C ) . These results suggest that mTwf1 attenuates V - 1 effects on CP via an allosteric mechanism , distinct from a simple steric binding competition . In probing the mechanism further , we drew inspiration from a study by Fujiwara and colleagues , showing that CARMIL forms a transient ternary complex with V - 1 and CP , leading to accelerated dis - sociation of V - 1 from CP ( Fujiwara et al . , 2014 ) . We asked whether mTwf1 might similarly catalyze the dissociation of V - 1 from CP . In stopped - flow fluorescence assays , fluorescently labeled V - 1 ( TAMRA - V - 1 ) was first allowed to bind CP , and then mixed at time zero with an excess of unlabeled V - 1 . The resulting decrease in fluorescence reflects the spontaneous dissociation of TAMRA - V - 1 from CP ( Figure 4D ) . The rate of V - 1 dissociation from CP increased in the presence of increasing concentrations of mTwf1 , pointing to the possible formation of a transient ternary complex that destabilizes V - 1 interactions with CP ( Figure 4D and E ) . Importantly , mTwf1 - 11 failed to enhance V - 1 dissociation ( Figure 4E ) , showing that this effect depends on direct interactions between mTwf1 tail and CP . These results demonstrate that CARMIL and Twinfilin share a common function in cata - lyzing the dissociation of V - 1 from CP using their CPI motifs to bind CP , despite having different effects on the displacement of CP from barbed ends . Structural evidence for the twinfilin tail interacting with the CPI - binding site on CP Given the observed competition between mTwf1 tail peptide and the CPI motif of CARMIL for bind - ing to CP , and the similarity between mTwf1 and CARMIL in catalyzing V - 1 dissociation from CP , we Figure 3 continued DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 006 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 8 of 28 Research article Cell Biology sought structural evidence for the nature of the interaction between mTwf1 and CP . We hypothe - sized that the binding sites for mTwf1 and the CPI motif were likely overlapping . To test this hypoth - esis , we used hydrogen - deuterium exchange with mass spectrometry ( HDX - MS ) to interrogate the conformational dynamics and solvent accessibility of the backbone and sidechains of CP , free and in complex with Twf1 . Further , we compared our results to those in our recent study on the interactions of CARMIL with CP using the same approach ( Johnson et al . , 2018 ) . We tested three different forms of mTwf1 : a short tail peptide ( residues 317 \u2013 350 ) , a longer tail peptide ( residues 305 \u2013 350 ) , and full - length mTwf1 . These constructs were added to CP , either full - length alpha / beta hetero - dimer , or full - length alpha subunit with a beta subunit truncated at its C - terminus , removing the actin - binding beta tentacle . The results were essentially the same in each case . The presence of mTwf1 resulted in protection from H - D exchange at the N - terminal stalk of CP ( Figure 5A , Fig - ure 5\u2014figure supplements 1 and 2 ) . Similar effects to H - D exchange were observed upon CARMIL binding to CP ( Johnson et al . , 2018 ) ; also shown here in Figure 5B ) , which correspond well with the CPI - motif binding site defined by X - ray crystallography and solution NMR studies ( Hernandez - Figure 4 . Twinfilin\u2019s direct binding to Capping Protein accelerates the disassociation of V - 1 to promote capping of filaments . ( A , B ) Seeded elongation assays comparing the rates of actin assembly from spectrin - F - actin seeds ( grey ) in the presence of 0 . 5 m M actin ( 10 % pyrene - labeled ) , 25 nM muscle Capping Protein ( CapZ ) , 500 nM V - 1 , and variable concentrations of mTwf1 ( A ) or mTwf1 - 11 ( B ) as indicated . Data shown are representative curves from experiments performed three independent times . ( C ) Fluorescence anisotropy measurement of 100 nM HiLyte488 - labeled mTwf1 tail peptide mixed with 1 m M mouse Capping Protein ( CP ) and variable concentrations of CBR fragment of CARMIL or V - 1 . Rates for each condition averaged from three independent experiments . ( D ) Stopped - flow fluorescence assays measuring the kinetics of dissociation of 50 nM TAMRA - V - 1 from 1 m M CP upon addition at time zero of 2 . 5 m M unlabeled V - 1 and variable concentrations of mTwf1 as indicated . Apparent dissociation rates are listed for each condition . ( E ) Apparent dissociation rates of TAMRA - V - 1 for different concentrations of mTwf1 are from ( D ) ; and for 12 m M mTwf1 (cid:0) 11 = 1 . 0 \u00b1 0 . 003 s (cid:0) 1 . Anisotropy values for each condition were averaged from five independent experiments . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 007 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 9 of 28 Research article Cell Biology Valladares et al . , 2010 ; Takeda et al . , 2010 ; Zwolak et al . , 2010 ) . For mTwf1 , we also observed H - D exchange protection of a small region on CP corresponding to the V - 1 binding site ( Figure 5A and B , Figure 5\u2014figure supplements 1 and 2 ) , consistent with our results described above for the effects of mTwf1 in promoting V - 1 dissociation from CP . These structural effects are also consistent with our previous results for CARMIL , which alters the V - 1 binding site ( Johnson et al . , 2018 ) . How - ever , it is worth noting that mTwf1 - induced changes in CP conformation at the actin - binding Figure 5 . HDX - MS analysis of Twinfilin reveals effects on Capping Protein structure near the CPI motif - binding site . ( A ) A cartoon representation of a crystal structure of CP , based on PDB 3AAA ( Takeda et al . , 2010 ) . Differences in deuterium uptake induced by mTwf1 binding to CP are displayed as a color gradient ( see scale at bottom of panel ( B ) CPI domain of CARMIL overlaid on to its binding site on CP ( around the stalk ) . Representative comparisons of deuterium uptake curves for free CP ( black ) with mTwf1 bound CP ( red ) for CP alpha subunit ( upper panels ) and CP beta subunit ( lower panels ) . Error bars representing the results of t - tests between samples are shown above each time point to illustrate statistical significance . When error bars are not shown explicitly , the error is within the radius of the symbol . Data shown are representative curves from experiments repeated two independent times . ( B ) A cartoon representation of a crystal structure of CP , showing the differences in deuterium uptake induced by CBR domain of CARMIL binding to CP are displayed as a color gradient ( see scale at the bottom ) . CPI domain of CARMIL overlaid on to its binding site on CP ( around the stalk ) , V - 1 is overlaid on its binding site on CP ( barbed end binding surface ) for comparison . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 008 The following figure supplements are available for figure 5 : Figure supplement 1 . Supporting data for Figure 5 showing differential HDX results . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 009 Figure supplement 2 . Supporting data for Figure 5 showing differential HDX results . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 010 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 10 of 28 Research article Cell Biology interface were not as extensive as those induced by CARMIL , which is consistent with CARMIL , but not mTwf1 , weakening CP binding to actin at the barbed ends . Twinfilin and CP colocalize in cells and have similar knockdown phenotypes To investigate the functional relationship between Twinfilin and CP in cells , we started by asking whether mTwf1 and CP colocalize . While Twinfilin and CP have been localized individually , and are each reported to be enriched at the tips of filopodia and stereocilia , endocytic actin patches , lamelli - podia , and Drosophila bristles ( Avenarius et al . , 2017 ; Falck et al . , 2004 ; Goode et al . , 1998 ; Nevalainen et al . , 2011 ; Peng et al . , 2009 ; Rzadzinska et al . , 2009 ; Sinnar et al . , 2014 ; Vartiainen et al . , 2000 ) , to our knowledge they have never been co - imaged in vertebrate cells . To address this , we performed immunofluorescence on CP and Twinfilin in mouse B16F10 melanoma cells , co - staining the cells with Alexa 568 - phalloidin to visualize F - actin . We observed strong colocali - zation of Twinfilin and CP throughout the cell and a co - enrichment at the actin - rich leading and trail - ing edges ( Figure 6A and B ) . Further , quantitative western blotting showed that Twinfilin and CP are present at ~ 1 : 2 molar ratio in B16F10 cells ( Figure 6C , Figure 6\u2014figure supplement 1 ) . Previ - ous studies reported the concentration of CP in B16F10 cells to be ~ 1 m M ( Fujiwara et al . , 2014 ; Pollard and Borisy , 2003 ) , suggesting that mTwf1 is present at ~ 0 . 5 m M . Given the high affinity of the Twinfilin - CP interaction ( K d = 50 nM ) , these observations are consistent with mTwf1 being associ - ated with a substantial fraction of the CP in cells . The ability of Twinfilin to function as a \u2018pro - capping\u2019 factor in vitro , by antagonizing the inhibitory effects of V - 1 on CP , predicted that genetic loss of mTwf1 might at least partially phenocopy loss of CP . While a number of studies have examined how Twinfilin mutations affect whole animal develop - ment and physiology ( Iwasa and Mullins , 2007 ; Meacham et al . , 2009 ; Nevalainen et al . , 2011 ; Wahlstro\u00a8m et al . , 2001 ; Wang et al . , 2010 ; Yamada et al . , 2007 ) , we are unaware of any studies that have investigated how loss of Twf1 affects the morphology and actin organization of cultured mammalian cells . Using RNAi silencing in B16F10 cells , we separately depleted endogenous mTwf1 and CP , which was verified by both western blotting ( Figure 6E and F ) and immunostaining ( Fig - ure 6\u2014figure supplement 1 ) . Knockdown of either mTwf1 or CP led to a similar , marked increase in the density of peripheral protrusions or microspikes with a concomitant loss of lamellipodial surfaces ( Figure 6F and G ) . Similar phenotypes have been reported for CP depletion in multiple cell lines ( Edwards et al . , 2013 ; Edwards et al . , 2015 ; Mejillano et al . , 2004 ; Sinnar et al . , 2014 ) . Expres - sion of an RNAi - refractive mTwf1 construct , but not mTwf1 - 11 , rescued the defects caused by deple - tion of endogenous mTwf1 ( Figure 6F and G ; Figure 6\u2014figure supplement 1 ) , demonstrating that these cellular functions of mTwf1 critically depend on its interaction with CP . We also made the unexpected observation that knockdown of CP was accompanied by a dra - matic reduction in Twinfilin levels in cells , as seen by both western blotting ( Figure 6D ) and immuno - fluorescence ( Figure 6\u2014figure supplement 1 ) . This effect was confirmed using a second RNAi oligonucleotide that targets a different region of CP ( siCP2 , Figure 6D ) . Further , it was observed in additional cell lines besides B16F10 , including Neuro - 2A and NIH - 3T3 cells ( Figure 6\u2014figure sup - plement 1 ) . These observations support the closely intertwined relationship of CP and Twinfilin in vivo . Our results above also call into question whether the full extent of the phenotype caused by knockdown of CP ( Figure 6G ) is due to loss of CP , or instead is partly due to the accompanying loss of Twinfilin . To address this , we restored mTwf1 levels in cells depleted of CP by driving mTwf1 expression from a rescue plasmid , which was confirmed by western blotting and immunofluores - cence ( Figure 6\u2014figure supplement 1 ) . Forced expression of mTwf1 partially rescued the defects associated with CP depletion , indicating that a portion of the original defects observed after CP knockdown were likely due to the accompanying loss of mTwf1 . These observations also suggest that many previously reported phenotypes arising from CP knockouts and knockdowns should be revisited or reinterpreted with the potential loss of Twinfilin in mind . Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 11 of 28 Research article Cell Biology Figure 6 . mTwf1 and Capping Protein colocalize and have similar phenotypes in B16F10 melanoma cells . ( A ) Representative images from immunofluorescence staining showing colocalization of endogenous mTwf1 ( yellow ) and Capping Protein ( magenta ) . Scale bar , 20 m m . Close ups of boxed regions shown in Zooms ; scale bar , 4 m m . ( B ) Mander\u2019s correlation coefficient ( M1 and M2 ) values of overlap between mTwf1 and Capping Protein ( CP ) measured from cells ( n = 67 cells ) as in ( A ) . Error bars , s . e . m . ( C ) Comparison of the relative abundance of mTwf1 and Capping Protein ( CP ) in B16F10 cells measured from western blot analysis . Data averaged from four separate experiments . Error bars , s . d . n . s . p > 0 . 05 by t - test . ( D , E ) Representative western blots and quantification of cellular levels of mTwf1 ( D ) and CP ( E ) in B16F10 cells treated with siRNA against mTwf1 ( si - Twf1 ) or Figure 6 continued on next page Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 12 of 28 Research article Cell Biology Defects caused by CARMIL1 hyperactivity can be rescued by elevated twinfilin expression Finally , we tested the prediction of our biochemical observations that loss of capping activity in cells caused by overexpressed CARMIL1 should be restored by co - overexpression of Twf1 . B16F10 cells ectopically expressing CARMIL1 showed morphological defects similar to loss of CP , and ectopic mTwf1 expression rescued the defects ( Figure 7A and B ) . Importantly , ectopic expression of mTwf1 alone caused no significant change in cell morphology . These results support our biochemical obser - vations , and suggest that Twf1 promotes capping in vivo , at least in part by competing with CARMIL for CP binding and antagonizing the uncapping effects of CARMIL . Discussion Twinfilin and CP have been inextricably linked as interacting partners in yeast and animal cells for over 15 years ( Palmgren et al . , 2001 ) , yet until now it has remained a mystery what function their interaction serves . Here we discovered that Twinfilin binds to CP using an orphan CPI - like sequence in its C - terminal tail region , and through this interaction protects CP from inhibition and / or barbed end displacement by CARMIL and V - 1 . We found that Twinfilin binds to CP in a competitive manner with the CPI motif of CARMIL , interacts with a site on CP similar to that of CARMIL , and attenuates CARMIL - mediated uncapping of actin filaments . Separately , Twinfilin binding to CP also accelerates V - 1 dissociation from CP , despite Twinfilin and V - 1 having non - overlapping binding sites on CP . This might be achieved by an allosteric mechanism , given that CARMIL uses its CPI motif to induce V - 1 dissociation from CP through allosteric changes ( Fujiwara et al . , 2014 ; Johnson et al . , 2018 ) . Thus , we have demonstrated that Twinfilin promotes capping by protecting CP from interactions with V - 1 and CARMIL . This functional role for Twinfilin is further supported in vivo by our observations of : ( i ) strong colocalization of Twinfilin and CP , ( ii ) knockdowns of mTwf1 and CP that each give rise to sim - ilar defects in cell morphology , and ( iii ) over - expression of mTwf1 suppressing defects caused by CARMIL hyperactivity . Taken together , these results reveal that Twinfilin is a new member of the CPI - motif family of proteins , and the first within this group to show the ability to bind CP without reducing CP affinity for barbed ends , and antagonize the negative regulatory effects of another CPI protein . These functions of Twinfilin provide important new insights into the CP regulatory cycle . The best working model to date has been the Fujiwara model ( Fujiwara et al . , 2014 ) ( depicted here as \u2018Ear - lier Model\u2019 ; Figure 7C ) . It posits that the majority of CP in the cytosol is bound to V - 1 , in an inactive state , which then can be locally \u2018activated\u2019 by CARMIL at the leading edge . However , a caveat to this model is that it suggests CP - CARMIL complexes are the dominant capping species in the cell , despite this complex having ~ 100 fold reduced affinity for barbed ends compared to free CP . While this could potentially explain dynamic capping and uncapping near the plasma membrane , consis - tent with GFP - CP single molecule speckle analysis ( Miyoshi et al . , 2006 ) , it does not explain how cells maintain a pool of \u2018capping competent\u2019 CP further back from the leading edge , where CP is needed to cap barbed ends in stress fibers and other actin networks , and may cap barbed ends gen - erated by severing to promote filament disassembly . This model goes on to suggest that an Figure 6 continued CP ( si - CP ) or negative control ( Control ) . Band intensity for control cells was set to 1 . 0 . Data averaged from at least three separate experiments . , error bars , s . d . ( F ) Representative images showing F - actin immunofluorescence in B16F10 cells treated with siRNA against mTwf1 ( si - Twf1 ) or CP ( si - CP ) or negative control ( Control ) ; siRNA treated cells ( si - Twf1 or si - CP ) were also rescued using plasmids expressing si - resistant FL - myc - mTwf1 ( WT or mTwf1 - 11 ) . Scale bar , 20 m m . Close ups of boxed regions shown in Zooms ; scale bar , 4 m m . ( G ) Microspike density in cells treated as in ( D ) . Box and whisker plots show mean , first and third quartile , and the maximum and minimum values . Data averaged from two experiments . From Left to right : n = 45 , 53 , 51 , 24 , 24 , and 20 and mean microspike density 0 . 69 , 1 . 34 , 1 . 77 , 0 . 59 , 1 . 24 , and 1 . 01 filopodia per 10 m m of cell cortex . Error bars , s . e . m . * * * * p (cid:20) 0 . 0001 , * p (cid:20) 0 . 05 , n . s . p > 0 . 05 by one - way ANOVA with Tukey post hoc test . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 011 The following figure supplement is available for figure 6 : Figure supplement 1 . Supporting data for Figure 6 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 012 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 13 of 28 Research article Cell Biology Figure 7 . Overexpression of Twinfilin suppresses morphological defects caused by CARMIL hyperactivity . ( A ) Representative images of F - actin staining in untreated B16F10 cells ( control ) , and cells transfected with Flag - CARMIL1 , full - length ( FL ) - myc - mTwf1 , or both . Scale bar , 20 m m . ( B ) Average Microspike density in cells treated as in ( A ) . Box and whisker plots show mean , first and third quartile , and the maximum and minimum values . Data averaged from two experiments ( n = 19 \u2013 25 cells per condition ) . Data averaged from two experiments . From Left to right : n = 19 , 25 , 20 , and 25 ; mean microspike density 0 . 75 , 1 . 13 , 0 . 62 , 0 . 58 filopodia per 10 m m of cell cortex . Error bars , s . e . m . * * * p (cid:20) 0 . 001 , n . s . p > 0 . 05 by one - way ANOVA with Tukey post hoc test . ( C ) \u2018Earlier\u2019 model for CP regulatory cycle , adapted from Fujiwara and colleagues ( Fujiwara et al . , 2014 ) . Proposed steps in model : ( 1 ) V - 1 globally inhibits Capping Protein ( CP ) in the cytoplasm , ( 2 ) membrane - associated CARMIL ( at the protruding cell edge ) catalyzes dissociation of V - 1 from CP , ( 3 ) the resulting CARMIL - CP complex is partially active , binding weakly to free barbed ends to provide capping function , ( 4 ) an unknown factor or mechanism promotes dissociation of CARMIL from CP , allowing V - 1 to rebind CP and complete the cycle . ( D ) Our revised working model for the CP regulatory cycle . We propose that V - 1 functions to maintain a cytosolic reservoir of inactive CP , from which Twinfilin and CARMIL activate CP , generating two distinct forms of active CP in cells : Twinfilin - CP complexes and CARMIL - CP complexes . Twinfilin - CP complexes are fully active and support stable capping of barbed ends . In contrast , CARMIL - CP complexes have ~ 100 fold reduced affinity for barbed ends , and may therefore more transiently cap barbed ends , permitting restricted network growth at the cell membrane where CARMIL localizes . CARMIL and Twinfilin directly compete with each other for binding CP ( shown in close up of Transition state ) , which may result in the displacement of CP from Twinfilin . This would leave Twinfilin at the barbed end to catalyze depolymerization , or alternatively return filaments back to the original state of assembly . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 013 The following figure supplement is available for figure 7 : Figure supplement 1 . Structural model for a ternary complex formed by Twinfilin , Capping Protein and the barbed end of an actin filament . Figure 7 continued on next page Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 14 of 28 Research article Cell Biology unknown factor or mechanism dissociates the CP - CARMIL complex , allowing V - 1 to rebind CP , restoring it to an inactive state . In light of our results , we propose several additions and revisions to the Fujiwara model ( see \u2018Revised Model\u2019 ; Figure 7D ) . First , we suggest that Twinfilin\u2019s protective effects on CP , in particular against V - 1 , allow cells to maintain a larger pool of fully active CP ( Twinfilin - CP complexes ) in the cytosol than was previously thought . This view is supported by the relatively high abundance of Twinfilin in cells ( ~ 0 . 5 m M , compared to ~ 1 m M CP ) its high affinity for CP ( K d = 50 nM ) , and its ability to increase the rate of dissociation of V - 1 from CP . Given these observations , we propose that a sub - stantial fraction of CP is available in a fully active state , as Twinfilin - CP complexes , even in the pres - ence of a high concentration of V - 1 in the cytosol ( ~ 3 m M ) ( Fujiwara et al . , 2014 ; Pollard and Borisy , 2003 ) . Second , we propose that V - 1 functions to maintain a cytosolic reservoir of inactive CP , mobilized by Twinfilin and / or CARMIL dissociating V - 1 to generate \u2018stable capping\u2019 ( Twinfilin - CP ) in the cytosol and possibly \u2018transient capping\u2019 ( CARMIL - CP ) complexes at the plasma mem - brane , respectively . CARMIL - CP complexes at the plasma membrane could facilitate actin network growth to drive leading edge protrusion . In contrast , Twinfilin - CP complexes in the cytosol may facil - itate stable capping of barbed ends to limit network growth and promote filament disassembly and turnover . Third , we propose that the association of Twinfilin - CP complexes with barbed ends primes filaments for disassembly . Our data show that CARMIL , and / or other CPI proteins , compete with Twinfilin for binding CP . These interactions may competitively remove CP from barbed ends , leaving Twinfilin at the barbed end to processively depolymerize filaments , either alone or in combination with Srv2 / CAP ( as depicted in Figure 6D ) ( Hilton et al . , 2018 ; Johnston et al . , 2015 ) . In this man - ner , the interaction of Twinfilin with CP could serve not only to initially promote capping , and thus limit network growth , but also to position Twinfilin at barbed ends for subsequently catalyzing the disassembly of filaments . In summary , our results show that functions of mammalian Twinfilin and CP are closely inter - twined . This functional relationship is likely to extend to other species given CPI motif sequence con - servation in the Twinfilin tail region ( Figure 2A ) and the conserved nature of the Twinfilin - CP interaction . Indeed , S . cerevisiae Aim21 was recently identified as the first yeast CPI motif - containing protein , and was shown to regulate CP function at cortical actin patches ( Farrell et al . , 2017 ; Shin et al . , 2018 ) . We generated a structural model to explore the possible ternary complex formed by Twinfilin , CP , and the barbed end of an actin filament ( Figure 7\u2014figure supplement 1 ) . In this model , the Twinfilin tail is long enough to allow for simultaneous binding of Twinfilin\u2019s CPI motif to CP and Twinfilin\u2019s C - terminal ADFH domain to an actin subunit at the barbed end . Further , there are no clashes in binding between CP and Twinfilin on actin . It is worth noting that CP and Twinfilin appear to be able to associate with barbed ends individually or as a CP - Twinfilin complex , but with distinct consequences for the function and dynamics of actin networks . CP alone stably caps barbed ends , blocking subunit addition or loss , and our results suggest that CP - Twinfilin complexes may do the same . However , when Twinfilin alone associates with barbed ends , it drives processive depo - lymerization , while blocking new assembly ( Hilton et al . , 2018 ; Johnston et al . , 2015 ) . Thus , despite key differences in the nature of their associations with barbed ends , CP and Twinfilin each inhibit fila - ment growth , likely explaining why Twinfilin can replace CP in reconstituted actin motility assays in vitro ( Helfer et al . , 2006 ) . Finally , our data add to a broader emerging view that actin dynamics in vivo are controlled by a complex set of barbed end - associated factors , many of which interact with each other and / or stimu - late each other\u2019s dissociation from barbed ends . These multi - component mechanisms may allow cells to control rapid transitions at filament ends through different functional states , including ( i ) formin - bound elongation , ( ii ) paused growth by formin - CP \u2018decision complexes\u2019 , ( iii ) stable or transiently capped states by CP alone or CP - Twinfilin complexes ( Bombardier et al . , 2015 ; Shekhar et al . , 2015 ) , and ( iv ) depolymerization by Twinfilin , Cofilin , and / or Srv2 / CAP . These molecular mechanisms for regulating barbed end growth are vastly more elaborate and dynamic than once thought , and help explain the exquisite spatiotemporal control that cells have in tuning actin network dynamics . Figure 7 continued DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 . 014 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 15 of 28 Research article Cell Biology Materials and methods Key resources table Reagent type ( species ) or resource Designation Source or reference Identifiers Additionalinformation Antibody Rabbit anti - Twinfilin PekkaLappalainen ( Univ . Helsinki ) WB ( 1 : 1000 ) IF ( 1 : 100 ) Antibody mouse anti - Capping Protein DevelopmentStudiesHybridomaBank Cat : 3F2 WB ( 1 : 2000 ) IF ( 1 : 50 ) Antibody Mouse anti - Flag SigmaAldrich Cat : F3165 WB ( 1 : 5000 ) IF ( 1 : 500 ) Antibody Rabbit anti - Myc GeneTex Cat : GTX29106 WB ( 1 : 5000 ) IF ( 1 : 500 ) Antibody Goat anti - mouse - HRP GEHealthcare Cat : 31430 WB ( 1 : 10000 ) Antibody Goat anti - rabbit - HRP GE Health care Cat : 31460 WB ( 1 : 10000 ) Antibody Donkey anti - rabbit Alexa Flour 488 Thermo Fisher Scientific Cat : A21206 IF ( 1 : 1000 ) Antibody Donkey anti - mouse Alexa Flour 488 ThermoFisherScientific Cat : A21202 IF ( 1 : 1000 ) Antibody Goat anti - rabbit Alexa Flour 633 ThermoFisherScientific Cat : A21071 IF ( 1 : 1000 ) Antibody Donkey anti - mouse Alexa Flour 647 ThermoFisherScientific Cat : A31571 IF ( 1 : 1000 ) Cell line ( M . musculus ) B16F10 ATCC CRL - 6475 Cell line ( M . musculus ) Neuro - 2A neuroblast ATCC CCL - 131 Cell line ( M . musculus ) NIH3T3filbroblast ATCC CRL - 1658 Chemicalcompound , drug NHS - XX - Biotin Merck KGaA Cat : 203188 Chemicalcompound , drug Oregon - Green - 488 iodoacetamide Invitrogen Cat : O6010 Chemicalcompound , drug Ni2 + - NTA - agarose beads Qiagen Cat : 30230 Chemicalcompound , drug tetramethylrhodamine ( TAMRA ) (cid:0) 5 - maleimide Invitrogen Cat : T6027 Chemicalcompound , drug methoxy - poly ( ethylene glycol ) - silane LaysanBioInc Chemical compound , drug biotin - poly ( ethylene glycol ) - sil Laysan Bio Inc Continued on next page Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 16 of 28 Research article Cell Biology Continued Reagent type ( species ) or resource Designation Source or reference Identifiers Additionalinformation Chemicalcompound , drug AquaMount Thermo Fisher Scientific Cat : 14 - 390 - 5 Chemicalcompound , drug Alexa Flour 568 - phalloidin Thermo Fisher Scientific Cat : A12380 IF ( 1 : 1000 ) Chemicalcompound , drug Formaldehyde37 % SigmaAldrich Cat : 252549 Commercialassayorkit LipofectamineRNAiMAX Thermo Fisher Scientific Cat : 137780 \u2013 0775 Commercialassayorkit Lipofectamine3000 Thermo Fisher Scientific Cat : L2000 - 015 Commercial assay or kit Pierce ECL Western Blotting Substratedetectionkit Thermo FisherScientific Cat : 34580 Other DMEM - Dulbecco\u2019s Modified Eagle Medium Gibco BRL Life Technologies Cat : 11995 \u2013 073 Other FBS - Fetal Bovine Serum SigmaAldrich Cat : F9423 Other 200 mM L - glutamine ThermoFisherScientific Cat : 25030 \u2013 081 Peptide , recombinantprotein N - terminal HiLyte488mTwf1Tail Anaspec Peptide , recombinantprotein CARMILCPI WatsonBioSciences Peptide , recombinantprotein CARMILCSI WatsonBioSciences Peptide , recombinantprotein mTwf1 A305 - D350 WatsonBioSciences Peptide , recombinantprotein mTwf1A305 - D350 , K325A WatsonBioSciences Peptide , recombinantprotein PreScissionprotease GEHealthcare Cat : GE27 - 0843 - 01 RecombinantDNAreagent chickenCP a 1 b 1 Soeno et al . , 1998 Soeno et al . , 1998 Plasmid RecombinantDNAreagent chicken SNAP - CP a 1 b 1 Bombardier et al . , 2015 Bombardier et al . , 2015 Plasmid RecombinantDNAreagent mouseCP a 1 b 2 Kim et al . , 2012 Plasmid RecombinantDNAreagent mouseCP a 1 D 28 Kim et al . , 2012 Plasmid Continued on next page Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 17 of 28 Research article Cell Biology Continued Reagent type ( species ) or resource Designation Source or reference Identifiers Additionalinformation RecombinantDNAreagent mouseCP a 1 b 2 R15A / Y79A Edwards et al . , 2015 Edwards et al . , 2015 Plasmid RecombinantDNAreagent humanCARMIL1CBR115 ( 964 \u2013 1078 ) Kim et al . , 2012 Plasmid RecombinantDNAreagent humanV - 1 Edwards et al . , 2015 Plasmid RecombinantDNAreagent CARMIL1 Edwards et al . , 2013 Edwards et al . , 2013 Plasmid RecombinantDNAreagent pGEX - 6p - 1 - mTwf1 Hilton et al . , 2018 Hilton et al . , 2018 Plasmid RecombinantDNAreagent pGEX - 6p - 1 - mTwf2a Hilton et al . , 2018 Hilton et al . , 2018 Plasmid Recombinant DNA reagent pGEX - 6p - 1 - mTwf1 - 11 This paper Plasmid RecombinantDNAreagent pEGFP - C1 Clontech Plasmid RecombinantDNAreagent pCMV - M1 Addgene Cat : 23007 Plasmid RecombinantDNAreagent pCMV - myc - mTwf1 This paper Plasmid RecombinantDNAreagent pCMV - myc - mTwf1 - 11 This paper Plasmid Sequence - basedreagent siTwf1 This paper siRNA ; CGUUACCA UUUCUUUCUGUUU Sequence - basedreagent siCP1 This paper siRNA ; CCUCAGCGA UCUGAUCGACUU Sequence - based reagent siCP2 This paper siRNA ; GCACGC UGAAUGAGAUCUA Sequence - based reagent control RNAi oligos ( Stealth RNAi ) Invitrogen Cat : 12935 \u2013 200 Software , algorithm Fiji / Image J Schindelin et al . , 2012 Software , algorithm NIS Elements software - Version 4 . 30 . 02 NikonInstruments Software , algorithm GraphPadPrism6 . 0 GraphPadSoftware Software , algorithm Adobe Creative Cloud Illustrator Adobe Systems Strain , strain background ( E . coli ) BL21 ( DE3 ) pLysS This paper Strain , strain background ( E . coli ) BL21 ( DE3 ) pRIL This paper Strain , strain background ( E . coli ) BL21 ( DE3 ) pRARE This paper Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 18 of 28 Research article Cell Biology Plasmids Plasmids used for expressing the following proteins were previously described : chicken CP a 1 b 1 ( Soeno et al . , 1998 ) , chicken SNAP - CP a 1 b 1 ( Bombardier et al . , 2015 ) , mouse CP a 1 b 2 ( Kim et al . , 2012 ) , mouse CP a 1 D 28 ( Kim et al . , 2012 ) , mouse CP a 1 b 2 R15A / Y79A ( Edwards et al . , 2015 ) , human CARMIL1 CBR115 ( 964 \u2013 1078 ) ( Kim et al . , 2012 ) , human V - 1 ( Edwards et al . , 2015 ) . The plasmid for over - expressing CARMIL1 in mammalian cells has been described ( Edwards et al . , 2013 ) . To generate plasmids for expressing mouse Twinfilin isoforms as glutathione - S - transferase ( GST ) - fusions in E . coli , ORFs were PCR amplified from pHAT2 - mTwf1 and pHAT2 - mTwf2a kindly provided by Pekka Lappalainen ( Univ . of Helsinki ) ( Nevalainen et al . , 2009 ) , and subcloned into the EcoRI and NotI sites of pGEX - 6p - 1 , yielding pGEX - 6p - 1 - mTwf1 and pGEX - 6p - 1 - mTwf2a . pGEX - 6p - 1 - mTwf1 - 11 ( K332A , R333A ) was generated by site - directed mutagenesis of pGEX - 6p - 1 - mTwf1 . For V - 1 fluorescence experiments , we used a previously demonstrated strategy of removing two surface cysteine residues to allow direct labeling on the single remaining cysteine ( Fujiwara et al . , 2014 ) ; this was achieved by performing site - directed mutagenesis on wild type pGEX - GST - V - 1 plasmid to introduce two mutations ( C45S , C83S ) . To generate an RNAi - refractive construct of mTwf1 for expression in cultured cells , the ORF of mTwf1 was PCR amplified from pGEX - 6p - 1 and subcloned into the HindIII and SacI sites of pEGFP - C1 ( Clontech , Mountain View , CA ) . Then , site - directed muta - genesis was used to introduce silent mutations at specific nucleotides of the ORF ( 703 , 709 , 711 , 715 ) , and the RNAi - refractive ORF was subcloned into the EcoRI and NotI sites of pCMV - M1 , a gift from Linda Wordeman ( Stumpff et al . , 2008 ) ( Addgene plasmid # 23007 ) , yielding pCMV - myc - mTwf1 . Site - directed mutagenesis was performed on pCMV - myc - mTwf1 to generate mutant pCMV - myc - mTwf1 - 11 ( K332A , R333A ) . All constructs were verified by DNA sequencing . Protein expression and purification Rabbit skeletal muscle actin ( RMA ) ( Spudich and Watt , 1971 ) , was purified from acetone powder generated from frozen ground hind leg muscle tissue of young rabbits ( PelFreez , Rogers , AR ) . Lyophilized acetone powder stored at (cid:0) 80 \u02da C was mechanically sheared in a coffee grinder , resus - pended in G - buffer ( 5 mM Tris - HCl pH 7 . 5 , 0 . 5 mM DTT , 0 . 2 mM ATP , 0 . 1 mM CaCl 2 ) , and then cleared by centrifugation for 20 min at 50 , 000 (cid:2) g . Actin was polymerized by the addition of 2 mM MgCl 2 and 50 mM NaCl and incubated overnight at 4 \u02da C . F - actin was pelleted by centrifugation for 150 min at 361 , 000 (cid:2) g , and the pellet solubilized by Dounce homogenization and dialyzed against G - buffer for 48 hr at 4 \u02da C . Monomeric actin was then precleared at 435 , 000 (cid:2) g , and loaded onto a S200 ( 16 / 60 ) gel filtration column ( GE healthcare , Marlborough , MA ) equilibrated in G - Buffer . Peak fractions containing actin were stored at 4 \u02da C . For labeling actin with biotin ( Breitsprecher et al . , 2012 ) or Oregon Green ( OG ) ( Kuhn and Pollard , 2005 ) , the F - actin pellet described above was Dounced and dialyzed against G - buffer lacking DTT . Monomeric actin was then polymerized by add - ing an equal volume of 2X labeling buffer ( 50 mM Imidazole pH 7 . 5 , 200 mM KCl , 0 . 3 mM ATP , 4 mM MgCl 2 ) . After 5 min , the actin was mixed with a 5 - fold molar excess of NHS - XX - Biotin ( Merck KGaA , Darmstadt , Germany ) or Oregon - Green - 488 iodoacetamide ( Invitrogen , Carlsbad , CA ) resus - pended in anhydrous DMF , and incubated in the dark for 15 hr at 4 \u02da C . Labeled F - actin was pelleted as above , and the pellet was rinsed briefly with G - buffer , then depolymerized by Dounce homogeni - zation , and dialyzed against G - buffer for 48 hr at 4 \u02da C . Labeled , monomeric actin was purified further on an S200 ( 16 / 60 ) gel filtration column as above . Aliquots of biotin - conjugated actin were snap fro - zen in liquid nitrogen and stored at (cid:0) 80 \u02da C . OG - 488 - actin was dialyzed for 15 hr against G - buffer with 50 % glycerol and stored at (cid:0) 20 \u02da C . For bulk actin assembly assays , RMA was fluorescently labeled with pyrenyl - iodoacetamide on cysteine 374 ( Pollard and Cooper , 1984 ; Graziano et al . , 2013 ) . An RMA pellet stored at 4 \u02da C ( pre - pared as described above ) was dialyzed against pyrene buffer ( 25 mM Tris - HCl , pH 7 . 5 , 100 mM KCl , 0 . 02 % NaN3 , 0 . 3 mM ATP , and 2 mM MgSO4 ) for 3 \u2013 4 hr and then diluted with pyrene buffer to 1 mg / ml ( 23 . 8 m M ) . A sevenfold molar excess of pyrenyl - iodoacetamide was added , the actin solu - tion was incubated overnight at 4 \u02da C , and aggregates were cleared by low - speed centrifugation . The supernatant ( containing F - actin ) was centrifuged for 3 hr at 4 \u02da C at 45 , 000 rpm in a Ti70 rotor ( Beck - man Coulter , Indianapolis , IN ) to pellet F - actin . The actin pellets were disrupted by Douncing , dia - lyzed against G - buffer for 1 \u2013 2 d , and gel filtered on a 16 / 60 S200 column . Peak fractions were pooled , aliquoted , snap frozen , and stored at (cid:0) 80 \u02da C . Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 19 of 28 Research article Cell Biology Mouse non - muscle CP a 1 b 2 was purified as described ( Graziano et al . , 2014 ) . Briefly , the expres - sion vector ( Soeno et al . , 1998 ) was transformed into E . coli strain BL21 pLysS . Cells were grown in LB to log phase , then expression was induced for 3 hr at 37 \u02da C by addition of 0 . 4 mM isopropyl - b - D - thiogalactopyranoside ( IPTG ) . Cells were collected by centrifugation , washed with 25 ml water , and resuspended in lysis buffer ( 20 mM Tris pH 8 . 0 , 1 mM EDTA , 0 . 1 % Triton X - 100 , protease inhibitors ) and lysed by lysozyme treatment and sonication . The cell lysate was clarified by centrifugation at 12 , 500 x g for 30 min at 4 \u02da C . Supernatants were loaded onto a 1 ml Q - HiTrap column ( GE Health - care ) and eluted with a 45 ml salt gradient ( 0 \u2013 500 mM KCl ) in 20 mM Tris , pH 8 . 0 . Peak fractions were pooled , concentrated using a centrifugal filter ( Centiprep , MWCO 10 kDa ; Millipore ) to 3 ml , and loaded onto a 26 / 60 Superdex 75 gel filtration column ( GE Healthcare ) equilibrated in 50 mM KCl , 20 mM Tris , pH 8 . 0 . Peak fractions were pooled and loaded onto a 5 ml Mono Q column ( GE Healthcare ) and eluted with a 30 ml salt gradient ( 0 \u2013 500 mM KCl ) in 20 mM Tris , pH 8 . 0 . Peak frac - tions were pooled , dialyzed overnight at 4 \u02da C into HEK buffer ( 20 mM HEPES , pH 7 . 4 , 1 mM EDTA , 50 mM KCl ) , aliquoted , snap - frozen in liquid N2 , and stored at (cid:0) 80 \u02da C . SNAP - 649 - CP ( CP a 1 b 1 ) was purified and labeled as described ( Bombardier et al . , 2015 ) . SNAP - CP was expressed E . coli strain BL21 pLysS . Cells were grown to log phase at 37 \u02da C , and then expres - sion was induced for 8 hr at 37 \u02da C by addition of 0 . 4 mM isopropyl - b - D - thiogalactopyranoside ( IPTG ) . Cells were collected by centrifugation , and resuspended in 20 mM Tris pH 8 . 0 , 1 mM EDTA , 0 . 1 % Triton X - 100 , protease inhibitors and lysed by lysozyme treatment and sonication . The cell lysate was centrifuged for 80 min at 60 , 000 rpm , 4 \u02da C in a Ti70 rotor ( Beckman / Coulter , Fullerton , CA ) . The supernatant was rotated with 0 . 75 ml of Ni2 + - NTA - agarose beads ( Qiagen , Valencia , CA ) . SNAP - CP was fluorescently labelled using 9 m M ( ~ 4 - fold excess ) dye adduct for 2 hr at room temperature , yielding SNAP - 649 - CP . To remove free dye , beads were washed three times with 20 mM imidazole ( pH 8 . 0 ) , 1X PBS , 1 mM DTT , 200 mM NaCl . Labeled SNAP - 649 - CP was eluted with 0 . 5 ml of 300 mM imidazole pH 8 . 0 , 50 mM Tris pH 8 . 0 , 100 mM NaCl , 1 mM DTT , 5 % glycerol , then purified by gel filtration on a Superose six column ( GE Healthcare ) equilibrated in 20 mM Hepes ( pH 7 . 5 ) , 1 mM EDTA , 150 mM KCl , 5 % glycerol . Peak fractions were pooled , concentrated , aliquoted , snap - frozen in liquid N2 , and stored at (cid:0) 80 \u02da C . For stopped - flow kinetics , fluorescence anisotropy binding and HDX - MS experiments , His - tagged - a 1 and b 2 subunits of mouse CP ( pRSFDuet - 1 , pBJ 2041 ) were co - expressed in E . coli BL21 ( DE3 ) pRIL and purified as described ( Johnson et al . , 2018 ) . For CP lacking the b tentacle , a prema - ture stop codon was introduced , so that the C - terminal residue of the mouse b 2 subunit was L243 instead of C272 ( pBJ 1891 ) . Twinfilin polypeptides were expressed as GST - fusions in E . coli strain BL21 pRARE . Cells were grown to log phase at 37 \u02da C , and then expression was induced for 16 hr at 18 \u02da C by addition of 0 . 4 mM isopropyl - b - D - thiogalactopyranoside ( IPTG ) . Cells were collected by centrifugation , washed with 25 ml water , and resuspended in 10 ml of PBS supplemented freshly with 0 . 5 mM dithiothreitol ( DTT ) , 1 mM phenylmethylsulphonyl fluoride ( PMSF ) , and a standard mixture of protease inhibitors . Cells were incubated with lysozyme ( 0 . 5 mg ml (cid:0) 1 ) on ice for 15 min and then sonicated . The cell lysate was clarified by centrifugation at 12 , 500 g for 20 min and incubated at 4 \u02da C ( rotating ) for at least 2 hr with 0 . 5 ml glutathione \u2013 agarose beads ( Sigma - Aldrich ; St . Louis , MO ) . Beads were washed three times in PBS supplemented with 1M NaCl and then washed two times in PBS . Twinfilin was cleaved from GST by incubation with PreScission Protease ( GE Healthcare ; Marlborough , MA ) over - night at 4 \u02da C ( rotating ) . Beads were pelleted , and the supernatant was concentrated to 0 . 3 ml , and then purified further by size - exclusion chromatography on a Superose12 column ( GE Healthcare ) equilibrated in HEK buffer ( 20 mM Hepes pH 7 . 5 , 1 mM EDTA , 50 mM KCl , 0 . 5 mM DTT ) . Peak frac - tions were pooled , concentrated , aliquoted , snap - frozen in liquid N2 , and stored at (cid:0) 80 \u02da C . CARMIL CBR115 and V - 1 were purified from E . coli as above for mTwf1 proteins , except the GST tag was removed from V - 1 by digestion with thrombin instead of PreScission protease . To purify and label V - 1 ( generating TAMRA - V - 1 ) for fluorescence experiments , BL21 E . coli expressing pGEX - GST - V - 1 ( C45S , C83S ) was lysed in a Microfluidizer ( Microfluidics Corp . ; Westwood , MA ) . Fusion protein was isolated on Glutathione Superflow Agarose ( Thermo Fisher Scientific ; Waltham , MA ) . The GST tag was cleaved by digestion with bovine thrombin ( MP Biomedicals ; Santa Ana , CA ) overnight at 4 \u02da C , then separated from V - 1 on a Sephacryl S - 200 HR 16 / 60 column ( GE Healthcare ) equilibrated in 25 mM HEPES pH 7 . 0 , 1 mM TCEP , 100 mM KCl , 1 mM NaN3 . Residual GST was removed by re - incubating peak fractions with Glutathione Superflow Agarose . Purified V - 1 ( C45S , C83S ) was then Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 20 of 28 Research article Cell Biology labeled with tetramethylrhodamine ( TAMRA ) (cid:0) 5 - maleimide ( Invitrogen ) overnight at 4 \u02da C . Excess TAMRA was removed by dialysis against 20 mM 3 - ( N - morpholino ) propanesulfonic acid ( MOPS ) pH 7 . 2 , 1 . 0 mM TCEP , 100 mM KCl , 1 mM NaN3 . TAMRA - V - 1 was stored at (cid:0) 70 \u02da C The mTwf1 tail peptides used for anisotropy were sourced as follows : N - terminal HiLyte488 labeled mTwf1 ( H317 - D350 ) was purchased from Anaspec ( Fremont , CA ) ; unlabeled CARMIL1 CPI ( G969 - A1005 ) , CARMIL1 CSI ( M1019 - M1037 ) , mTwf1 ( A305 - D350 ) and mTwf1 ( A305 - D350 , K325A ) , as well as N - terminal TAMRA labeled mTwf1 ( A305 - D350 ) , were purchased from WatsonBio Sciences ( Houston , TX ) . Bulk pyrene F - actin assembly assays Pyrene actin assembly assays were performed as previously described ( Chesarone - Cataldo et al . , 2011 ) , with slight modifications for monitoring uncapping . Reactions containing 2 m M G - actin ( 5 % pyrene labeled ) , 25 nM CapZ , and variable concentrations of mTwf1 were mixed to a volume of 52 m l followed by addition of 3 m l of initiation mix ( 40 mM MgCl 2 , 10 mM ATP , 1 M KCl ) . Fluorescence was monitored at excitation and emission wavelengths of 365 and 407 nm , respectively , in a fluores - cence spectrophotometer ( Photon Technology International ; Lawrenceville , NJ ) . Acquisition was paused at 400 s , and 5 m l of CARMIL CBR ( final concentration 250 nM ) was spiked into the reaction , mixed rapidly by pipetting , and measurement was resumed . For pyrene actin elongation assays ( as in Figure 4A and B ) , 5 m l of freshly mechanically sheared F - actin ( 10 m M ) was added to a mixture of the indicated proteins or control buffers , and then imme - diately mixed with 0 . 5 m M monomeric actin ( 10 % pyrene labeled ) in 60 m l reactions and monitored in a plate reader ( Infinite M200 ; Tecan , Ma\u00a8nnedorf , Switzerland ) at excitation and emission wave - lengths of 365 and 407 nm , respectively . Fluorescence anisotropy The following anisotropy experiments were performed in HEK buffer ( 20 mM HEPES pH 7 . 5 , 1 mM EDTA , 50 mM KCl , 0 . 5 mM DTT ) . Reactions were incubated at room temperature for 15 min , and anisotropy was determined by measuring polarized emission intensities at 525 nm when excited at 497 nm using a fluorescence spectrophotometer ( Photon Technology International ) . To compare mTwf1 - tail binding to wild type and mutant CP ( Figure 1B ) , HiLyte - 488 - mTwf1 tail peptide ( 100 nM ) was mixed with different concentrations of wild - type or mutant CP . To compare the abilities of full - length wild type mTwf1 and mutant mTwf1 - 11 polypeptides to compete with labeled mTwf1 - tail for binding CP ( Figure 1C ) , HiLyte - 488 - mTwf1 tail peptide ( 100 nM ) was mixed with 1 m M CP and vari - able concentrations of full - length mTwf1 polypeptides . The following anisotropy experiments were performed in the indicated buffer , incubated at room temperature for 2 min , and anisotropy was determined by measuring polarized emission intensities at 525 nm when excited at 497 nm for HiLyte - 488 , or at 582 nm when excited at 552 nm for TAMRA . To compare mTwf1 tail peptide binding to wild type CP and mutant CP ( RY ) ( Figure 2B ) , HiLyte - 488 - mTwf1 tail peptide ( 60 nM ) was mixed with different concentrations of CP or CP ( RY ) in HEK buffer containing 0 . 005 % TWEEN 20 . To compare the abilities of unlabeled wild type and mutant mTwf1 tail peptides to compete with labeled mTwf1 tail peptide for binding to CP ( Figure 2C ) , TAMRA - mTwf1 tail peptide ( A305 - D350 , 40 nM ) was mixed with 1 m M CP and varying concentrations of the unlabeled tail peptides ( mTwf1 A305 - D350 or mTwf1 A305 - D350 , K325A ) in 20 mM MOPS ( pH 7 . 2 ) , 1 mM TCEP , 100 mM KCl , 1 mM NaN 3 , 0 . 005 % TWEEN 20 . To test the abilities of different frag - ments of CARMIL to compete with mTwf1 tail peptide for binding CP , HiLyte - 488 - mTwf1 tail pep - tide ( 60 nM ) was mixed with 240 nM CP and different concentrations of mouse CARMIL1 CBR ( 964 \u2013 1078 ) , CPI ( 969 \u2013 1005 ) , or CSI ( 1019 \u2013 1037 ) in HEK buffer containing 0 . 005 % TWEEN20 . Stopped - flow fluorescence For kinetic dissociation experiments ( as in Figure 4D and E ) , an SX . 18MV stopped flow instrument with Pro - Data SX software V2 . 2 . 27 ( Applied Photophysics Ltd . , Leatherhead , UK ) was used . 100 nM TAMRA - V - 1 was preincubated with 2 m M CP a 1 b 2 . At time zero , TAMRA - V - 1 : CP complex was rapidly mixed via stopped - flow with an equal volume of a solution containing 5 m M unlabeled V - 1 , along with varied concentrations of mTwf1 or mTwf1 - 11 . Experiments were performed at 25 \u02da C in HEK buffer containing 0 . 005 % TWEEN20 . Excitation occurred at 505 nm , with emission detected using a Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 21 of 28 Research article Cell Biology 570 + nm band - pass filter . All concentrations of mTwf were performed in replicates of 5 \u2013 10 , and traces were averaged . Apparent dissociation rates were determined by fitting the averaged data ( 5 ms . - 120 s . ) to a single exponential model using Pro - Data Viewer software V4 . 2 . 27 ( Applied Photo - physics Ltd . ) . Total internal reflection fluorescence ( TIRF ) microscopy For all experiments , 24 (cid:2) 60 mm coverslips ( Fisher Scientific ; Pittsburg , PA ) were cleaned by succes - sive sonications as follows : 60 min in detergent , 20 min in 1 M KOH , 20 min in 1 M HCl min , and 60 min in ethanol . Coverslips were then washed extensively with ddH 2 O and dried in an N 2 - stream . A solution of 80 % ethanol pH 2 . 0 , 2 mg / ml methoxy - poly ( ethylene glycol ) - silane and 2 m g / ml biotin - poly ( ethylene glycol ) - silane ( Laysan Bio Inc . ; Arab , AL ) was prepared and layered on the cleaned coverslips ( 200 m l per coverslip ) . The coverslips were incubated for 16 hr at 70 \u02da C . To assemble flow cells , PEG - coated coverslips were rinsed extensively with ddH 2 O and dried in an N 2 - stream , then attached to a prepared flow chamber ( Ibidi ; Martinsried , German ) with double sided tape ( 2 . 5 cm x 2 mm x 120 m m ) and five min epoxy resin . Flow cells were prepared immediately before use by sequential incubations as follows : 3 min in HEK - BSA ( 20 mM Hepes pH 7 . 5 , 1 mM EDTA , 50 mM KCl , 1 % BSA ) , 30 s in Streptavidin ( 0 . 1 mg / ml in PBS ) , a fast rinse in HEK - BSA , and then equilibration in 1X TIRF buffer , pH 7 . 5 ( 10 mM imidazole , 50 mM KCl , 1 mM MgCl 2 , 1 mM EGTA , 0 . 2 mM ATP , 10 mM DTT , 15 mM glucose , 20 m g / ml catalase , 100 m g / ml glucose oxidase , and 0 . 5 % methylcellulose ( 4000 cP ) ) . To initiate reactions , actin monomers ( 10 % OG - labeled , 0 . 5 % biotinylated ) were diluted to 1 m M in TIRF buffer , and immediately transferred to a flow chamber . After several minutes , once the actin filaments reached an appropriate length ( approximately 10 m m ) , the reaction mixture was replaced by flow - in . For depolymerization experiments , the solution was replaced with TIRF buffer lacking actin monomers , with or without Twinfilin and / or CP polypeptides . For uncapping experi - ments , the solution was replaced with TIRF buffer lacking actin monomers , with 3 nM SNAP - 649 - CP ( 100 % labeled ) , and filaments were allowed to be capped for 3 min . Subsequently , the solution was again replaced with TIRF buffer lacking actin monomers , with or without 50 nM CARMIL CBR and / or variable concentration of Twinfilin polypeptides . Time - lapse TIRF microscopy was performed using a Nikon - Ti200 inverted microscope equipped with a 150 mW Ar - Laser ( Mellot Griot ; Carlsbad , CA ) , a 60X TIRF - objective with a N . A . of 1 . 49 ( Nikon Instruments Inc . ; New York , NY ) , and an EMCCD cam - era ( Andor Ixon ; Belfast , Northern Ireland ) . During recordings , optimal focus was maintained using the perfect focus system ( Nikon Instruments Inc ) . Images were captured every 5 s . The pixel size cor - responded to 0 . 27 m m . Filament depolymerization rates were determined by tracing filaments in ImageJ ( http : / / rsbweb . nih . gov / ij ) and measuring the change in length of individual filaments for 15 \u2013 20 min after flow - in , or until filaments disappeared . Differences in fluorescence intensity along the length of the filament provided fiduciary marks that allowed us to distinguish barbed - and pointed - ends . Filament uncap - ping was measured by monitoring the as the amount of time that SNAP - 649 - CP puncta remained associated with the barbed end of a filament after the addition of CARMIL to the reaction ( with or without Twinfilin ) and expressing it as a fraction of filaments that remained capped at a given time point . All results shown are data from at least two independent TIRF experiments . Cell culture , transfection , and RNAi silencing Mouse B16 - F10 ( CRL - 6475 ) , Neuro - 2a ( CCL - 131 ) , and NIH / 3T3 ( CRL - 1658 ) cells obtained directly from ATCC ( American Type Culture Collection ; Manassas , VA ) , where their identities were authenti - cated by short tandem repeat DNA profiling and where they were tested for mycoplasma contami - nation . Cells were used for experiments within one year . All cells were grown in DMEM ( Gibco BRL Life Technologies ; Carlsbad , CA ) supplemented with 10 % fetal bovine serum ( FBS ; Sigma ) and 200 mM L - glutamine ( Thermo Fisher Scientific ) at 37 \u02da C and 5 % CO 2 . All cell culture experiments were carried out in 6 - well dishes that were initially seeded with 100 , 000 cells . To knockdown Twinfilin - 1 or Capping Protein cells were transfected 24 hr after seed - ing with 30 pmol siRNA oligo using Lipofectamine RNAiMAX ( Thermo Fisher Scientific ) according to the manufacturer\u2019s instructions . RNAi oligos directed against the mouse Twinfilin - 1 coding region targeting ( siTwf1 ) 5\u2019 - CGUUACCAUUUCUUUCUGUUU (cid:0) 3\u2019 ; and against the Capping Protein b sub - unit coding region targeting ( siCP1 ) 5\u2019 - CCUCAGCGAUCUGAUCGACUU - 3\u2019 , or ( siCP2 ) 5\u2019 - GCACGC Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 22 of 28 Research article Cell Biology UGAAUGAGAUCUA - 3\u2019 . Cells were transfected in parallel with control RNAi oligos ( Invitrogen ) . For over expression experiments cultured cells were transfected using Lipofectamine 3000 ( Thermo Fisher Scientific ) according to the manufacturer\u2019s instructions 24 hr after seeding . For CARMIL over expression experiments , 5 m G of DNA was transfected , and for Twinfilin over expression experi - ments 1 m G of DNA was transfected . Antibodies The rabbit anti - Twinfilin was a generous gift from Pekka Lappalainen ( Univ . Helsinki ) and used a dilu - tion of 1 : 1000 for western blot detection and 1 : 100 in cultured cells . A mouse anti - Capping Protein ( Development Studies Hybridoma Bank ; Iowa City , IA ) was used at a dilution of 1 : 2000 for western blot detection and 1 : 50 in cultured cells . Mouse anti - Flag ( F3165 , Sigma ) and rabbit anti - Myc ( GTX29106 , GeneTex ; Irvine , CA ) was used at 1 : 5000 for western blot detection and 1 : 500 in cul - tured cells . Mouse and Rabbit horseradish peroxidase conjugated secondary antibodies ( GE Health - care ) were used at a dilution of 1 : 10 , 000 for western blot detection . Secondary antibodies for immunofluorescence ( Alexa Fluor 488 or 647 ) and Alexa Fluor 568 - phalloidin ( ThermoFisher ) were used at a dilution of 1 : 1000 . Immunostaining cells For cell - staining experiments , 48 hr post transfection , the cells were re - plated on 3 (cid:2) 1 (cid:2) 1 mm glass coverslip ( VWR International ) that had been acid washed and coated with Laminin ( Invitrogen ) and allowed to adhere for 3 \u2013 6 hr . Cells were fixed for 15 min with 4 % paraformaldehyde in PBS at room temperature and then permeabilized for 15 min in permeabilization solution ( 0 . 5 % Triton X - 100 and 0 . 3 M glycine in PBS ) at room temperature . Slips were then blocked in 3 % BSA dissolved in PBST ( 1X PBS and 0 . 1 % TWEEN 20 ) for 1 hr at room temperature , then incubated in primary antibody ( in PBST ) for 12 hr at 4 \u02da C . Coverslips were then washed three times with 1X PBST and incubated with secondary antibodies ( in PBST ) for 1 hr at room temperature . Slips were washed three times with PBST and two times with PBS , and subsequently mounted on to slides with AquaMount ( Thermo Fisher Scientific ) . Cells were imaged on a Nikon i - E upright confocal microscope equipped with a CSU - W1 spinning disk head ( Yokogawa , Tokyo , Japan ) , 60x oil objective ( NA 1 . 4 ; Nikon Instru - ments ) , and an Ixon 897 Ultra - CCD camera ( Andor Technology ) controlled by NIS - Elements soft - ware . Maximum intensity projections and raw fluorescence values were measured using Fiji . Western blotting To measure protein levels in cells after silencing and rescue , cells were harvest 48 hr after initial oligo transfection and incubated for 10 min at 4 \u02da C in RIPA buffer ( 50 mM Tris , pH 7 . 5 , 150 mM NaCl , 1 % NP - 40 , 0 . 5 % Na - deoxycholate , 0 . 1 % SDS , 2 mM EDTA , 50 mM NaF ) . Samples were incubated on ice for 30 min , vortexed every 10 min , then precleared by centrifugation at 20 , 800 x g for 15 min at 4 \u02da C , quantified by Bradford assay , and immunoblotted . Proteins were detected using a Pierce ECL West - ern Blotting Substrate detection kit ( Thermo Fisher Scientific ) . Bands were quantified using Image - Lab ( Biorad ) . Hydrogen deuterium exchange mass spectrometry ( HDX - MS ) HDX - MS was performed as described ( Johnson et al . , 2018 ) . CP and Twf1 samples were buffer - exchanged with 1X phosphate saline buffer ( PBS ) , pH 7 . 4 . HDX was initiated by diluting samples ( 25 m M , 2 m L ) 10 - fold with 1XPBS prepared in D 2 O buffer , or 1XPBS H 2 O buffer for samples measured for no - deuterium control . At different time intervals ( 10 , 30 , 60 , 120 , 360 , 900 , 3600 , and 14400 s ) , the labeling reaction was quenched by rapidly decreasing the pH to 2 . 5 with 30 m L of quench buffer ( 3 M urea , 1 % trifluoroacetic acid , H 2 O ) at 4 \u02da C . The protein mixture was immediately injected into a custom - built HDX sample - handling device that enabled digestion with a column containing immobi - lized pepsin ( 2 mm (cid:2) 20 mm ) at a flow rate of 100 m L / min in 0 . 1 % formic acid . The resulting peptic peptides were captured on a ZORBAX Eclipse XDB C8 column ( 2 . 1 mm (cid:2) 15 mm , Agilent ) for desalt - ing ( 3 min ) . The C8 column was then switched in - line with a Hypersil Gold C18 column ( 2 . 1 mm (cid:2) 50 mm , Thermo Fisher ) , and a linear gradient ( 4 \u2013 40 % acetonitrile , 0 . 1 % formic acid , 50 m L / min flow rate , over 5 min ) was used to separate the peptides and direct them to an LTQ - FTICR mass spec - trometer ( Thermo Fisher ) equipped with an electrospray ionization source . Valves , columns , and Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 23 of 28 Research article Cell Biology tubing for protein digestion and peptide separation were immersed in an ice - water bath to minimize back - exchange . To map the peptic peptides , the digest , in the absence of HDX , was submitted to accurate mass analysis by LC \u2013 MS / MS with the LTQ - FTICR , and the peptic peptides identified using Mascot ( Matrix Science ) . For samples that underwent HDX , raw mass spectra and peptide sets were submitted to HDX Workbench ( Pascal et al . , 2012 ) for calculation and data visualization in a fully automated fash - ion . Peptides for each run were assessed based on relative representation and statistical validation as implemented within HDX Workbench . Appropriate approach to determine statistical significance between these data is by using Tukey\u2019s multiple comparison test . A representative time point was manually selected , replicate data points from multiple samples at this time point used to conduct a one - way analysis of variance ( ANOVA ) the divergence between the means of the experiments were assessed . In instances with large differences , Tukey method was used to determine statistical signifi - cance if the resulting P value is less than 0 . 05 . In the case where there was a comparison between two experiments , a t - test was used . Only the top six peptides from each MS scan were used in the final analysis . The extent of HDX at each time point was calculated by subtracting the centroid of the isotopic distribution of the nondeuterated peptide from that of the deuterated peptide . The relative deuterium uptake was plotted versus the labeling time to yield kinetic curves ( % D vs time ) . Error bars represent the results of t - tests between samples are shown above each time point to illustrate statistical significance . For comparison between apo states and the complexes , differences in HDX for all time points were calculated . Absolute differences in perturbation values larger than 5 % D were considered significant . HDX values at 15 min time point were mapped onto the protein three - dimensional ( 3D ) structure for data visualization . Peptide digestions were optimized under HDX assay conditions , and the mass calculations included accommodation for back exchange with solvent . Acknowledgements We are grateful to Julian Eskin , Sean Guo , Silvia Jansen , M Angeles Juanes , and Daisy Leung for helpful discussions and / or comments on the manuscript . We thank Ms . S Smith and Ms . M Torres for general support and coordination . In addition , we thank Steve DelSignore for help with analysis of cultured cells , Alex Kozlov and Timothy Lohman for instrument access , assistance , and advice on stopped - flow experiments , and Marlene Mekel for assistance and advice with biochemistry experi - ments . This work was supported by a grant from the NIH ( R01 GM063691 ) to BLG , a grant from the NIH ( R35 GM118171 ) to JAC . , by Brandeis NSF MRSEC DMR - 1420382 , by grants from NIH ( P01AI120943 and R01AI123926 ) and a grant from the Department of the Defense ( Defense Threat Reduction Agency grant HDTRA1 - 16 - 1 - 0033 ; C Basler PI ) to GKA . We thank Dr . Michael L Gross for the use of the mass spectrometry facilities supported by P41GM103422 and for facilitating the mass spectrometry studies . The content of the information does not necessarily reflect the position or the policy of the federal government , and no official endorsement should be inferred . Additional information Funding Funder Grant reference number Author National Institutes of Health R01 GM063691 Bruce L Goode Defense Threat Reduction Agency HDTRA1 - 16 - 1 - 0033 Gaya K Amarasinghe National Institutes of Health R35 GM118171 John A Cooper National Science Foundation MRSEC DMR - 1420382 Bruce L Goode National Institutes of Health P01 AI120943 Gaya K Amarasinghe National Institutes of Health R01 AI123926 Gaya K Amarasinghe The funders had no role in study design , data collection and interpretation , or the decision to submit the work for publication . Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 24 of 28 Research article Cell Biology Author contributions Adam B Johnston , Conceptualization , Data curation , Formal analysis , Validation , Investigation , Writ - ing\u2014original draft ; Denise M Hilton , Conceptualization , Data curation , Formal analysis , Validation , Investigation , Writing\u2014original draft , Writing\u2014review and editing ; Patrick McConnell , Britney John - son , Data curation , Formal analysis , Validation , Writing\u2014review and editing ; Meghan T Harris , Avital Simone , Data curation , Formal analysis ; Gaya K Amarasinghe , Conceptualization , Formal analysis , Supervision , Funding acquisition , Investigation , Writing\u2014original draft , Project administration , Writ - ing\u2014review and editing ; John A Cooper , Conceptualization , Supervision , Funding acquisition , Vali - dation , Writing\u2014original draft , Project administration , Writing\u2014review and editing ; Bruce L Goode , Conceptualization , Supervision , Funding acquisition , Writing\u2014original draft , Project administration , Writing\u2014review and editing Author ORCIDs Adam B Johnston http : / / orcid . org / 0000 - 0002 - 1210 - 4929 Denise M Hilton http : / / orcid . org / 0000 - 0003 - 1577 - 1855 John A Cooper http : / / orcid . org / 0000 - 0002 - 0933 - 4571 Bruce L Goode https : / / orcid . org / 0000 - 0002 - 6443 - 5893 Decision letter and Author response Decision letter https : / / doi . org / 10 . 7554 / eLife . 41313 . 017 Author response https : / / doi . org / 10 . 7554 / eLife . 41313 . 018 Additional files Data availability All datasets associated with this article are included in the manuscript and supporting files . References Avenarius MR , Krey JF , Dumont RA , Morgan CP , Benson CB , Vijayakumar S , Cunningham CL , Scheffer DI , Corey DP , Mu\u00a8ller U , Jones SM , Barr - Gillespie PG . 2017 . Heterodimeric capping protein is required for stereocilia length and width regulation . The Journal of Cell Biology 216 : 3861 \u2013 3881 . DOI : https : / / doi . org / 10 . 1083 / jcb . 201704171 , PMID : 28899994 Bhattacharya N , Ghosh S , Sept D , Cooper JA . 2006 . Binding of myotrophin / V - 1 to actin - capping protein : implications for how capping protein binds to the filament barbed end . The Journal of Biological Chemistry 281 : 31021 \u2013 31030 . DOI : https : / / doi . org / 10 . 1074 / jbc . M606278200 , PMID : 16895918 Bombardier JP , Eskin JA , Jaiswal R , Corre\u02c6a IR , Xu MQ , Goode BL , Gelles J . 2015 . Single - molecule visualization of a formin - capping protein \u2019decision complex\u2019 at the actin filament barbed end . Nature Communications 6 : 8707 . DOI : https : / / doi . org / 10 . 1038 / ncomms9707 , PMID : 26566078 Breitsprecher D , Jaiswal R , Bombardier JP , Gould CJ , Gelles J , Goode BL . 2012 . Rocket launcher mechanism of collaborative actin assembly defined by single - molecule imaging . Science 336 : 1164 \u2013 1168 . DOI : https : / / doi . org / 10 . 1126 / science . 1218062 , PMID : 22654058 Cooper JA , Sept D . 2008 . New insights into mechanism and regulation of actin capping protein . International Review of Cell and Molecular Biology 267 : 183 \u2013 206 . DOI : https : / / doi . org / 10 . 1016 / S1937 - 6448 ( 08 ) 00604 - 7 , PMID : 18544499 Edwards M , Liang Y , Kim T , Cooper JA . 2013 . Physiological role of the interaction between CARMIL1 and capping protein . Molecular Biology of the Cell 24 : 3047 \u2013 3055 . DOI : https : / / doi . org / 10 . 1091 / mbc . e13 - 05 - 0270 , PMID : 23904264 Edwards M , Zwolak A , Schafer DA , Sept D , Dominguez R , Cooper JA . 2014 . Capping protein regulators fine - tune actin assembly dynamics . Nature Reviews Molecular Cell Biology 15 : 677 \u2013 689 . DOI : https : / / doi . org / 10 . 1038 / nrm3869 , PMID : 25207437 Edwards M , McConnell P , Schafer DA , Cooper JA . 2015 . CPI motif interaction is necessary for capping protein function in cells . Nature Communications 6 : 8415 . DOI : https : / / doi . org / 10 . 1038 / ncomms9415 Falck S , Paavilainen VO , Wear MA , Grossmann JG , Cooper JA , Lappalainen P . 2004 . Biological role and structural mechanism of twinfilin \u2013 capping protein interaction . The EMBO Journal 23 : 3010 \u2013 3019 . DOI : https : / / doi . org / 10 . 1038 / sj . emboj . 7600310 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 25 of 28 Research article Cell Biology Farrell KB , McDonald S , Lamb AK , Worcester C , Peersen OB , Di Pietro SM . 2017 . Novel function of a dynein light chain in actin assembly during clathrin - mediated endocytosis . The Journal of Cell Biology 216 : 2565 \u2013 2580 . DOI : https : / / doi . org / 10 . 1083 / jcb . 201604123 Fujiwara I , Remmert K , Piszczek G , Hammer JA . 2014 . Capping protein regulatory cycle driven by CARMIL and V - 1 may promote actin network assembly at protruding edges . PNAS 111 : E1970 \u2013 E1979 . DOI : https : / / doi . org / 10 . 1073 / pnas . 1313738111 Goode BL , Drubin DG , Lappalainen P . 1998 . Regulation of the cortical actin cytoskeleton in budding yeast by Twinfilin , a ubiquitous actin monomer - sequestering protein . The Journal of Cell Biology 142 : 723 \u2013 733 . DOI : https : / / doi . org / 10 . 1083 / jcb . 142 . 3 . 723 , PMID : 9700161 Graziano BR , Yu HY , Alioto SL , Eskin JA , Ydenberg CA , Waterman DP , Garabedian M , Goode BL . 2014 . The F - BAR protein Hof1 tunes formin activity to sculpt actin cables during polarized growth . Molecular Biology of the Cell 25 : 1730 \u2013 1743 . DOI : https : / / doi . org / 10 . 1091 / mbc . e14 - 03 - 0850 , PMID : 24719456 Hart MC , Cooper JA . 1999 . In Vertebrate isoforms of actin capping protein beta have distinct functions in vivo . The Journal of Cell Biology 147 : 1287 \u2013 1298 . DOI : https : / / doi . org / 10 . 1083 / jcb . 147 . 6 . 1287 , PMID : 10601341 Helfer E , Nevalainen EM , Naumanen P , Romero S , Didry D , Pantaloni D , Lappalainen P , Carlier M - F . 2006 . Mammalian twinfilin sequesters ADP - G - actin and caps filament barbed ends : implications in motility . The EMBO Journal 25 : 1184 \u2013 1195 . DOI : https : / / doi . org / 10 . 1038 / sj . emboj . 7601019 Hernandez - Valladares M , Kim T , Kannan B , Tung A , Aguda AH , Larsson M , Cooper JA , Robinson RC . 2010 . Structural characterization of a capping protein interaction motif defines a family of actin filament regulators . Nature Structural & Molecular Biology 17 : 497 \u2013 503 . DOI : https : / / doi . org / 10 . 1038 / nsmb . 1792 Hilton DM , Aguilar RM , Johnston AB , Goode BL . 2018 . Species - Specific functions of twinfilin in actin filament depolymerization . Journal of Molecular Biology 430 : 3323 \u2013 3336 . DOI : https : / / doi . org / 10 . 1016 / j . jmb . 2018 . 06 . 025 , PMID : 29928893 Iwasa JH , Mullins RD . 2007 . Spatial and temporal relationships between actin - filament nucleation , capping , and disassembly . Current Biology 17 : 395 \u2013 406 . DOI : https : / / doi . org / 10 . 1016 / j . cub . 2007 . 02 . 012 , PMID : 17331727 Johnson B , McConnell P , Kozlov AG , Mekel M , Lohman TM , Gross ML , Amarasinghe GK , Cooper JA . 2018 . Allosteric coupling of CARMIL and V - 1 binding to capping protein revealed by Hydrogen - Deuterium exchange . Cell Reports 23 : 2795 \u2013 2804 . DOI : https : / / doi . org / 10 . 1016 / j . celrep . 2018 . 04 . 096 , PMID : 29847807 Johnston AB , Collins A , Goode BL . 2015 . High - speed depolymerization at actin filament ends jointly catalysed by Twinfilin and Srv2 / CAP . Nature Cell Biology 17 : 1504 \u2013 1511 . DOI : https : / / doi . org / 10 . 1038 / ncb3252 Jung G , Alexander CJ , Wu XS , Piszczek G , Chen B - C , Betzig E , Hammer JA . 2016 . V - 1 regulates capping protein activity in vivo . PNAS 113 : E6610 \u2013 E6619 . DOI : https : / / doi . org / 10 . 1073 / pnas . 1605350113 Kim K , McCully ME , Bhattacharya N , Butler B , Sept D , Cooper JA . 2007 . Structure / function analysis of the interaction of phosphatidylinositol 4 , 5 - bisphosphate with actin - capping protein : implications for how capping protein binds the actin filament . The Journal of Biological Chemistry 282 : 5871 \u2013 5879 . DOI : https : / / doi . org / 10 . 1074 / jbc . M609850200 , PMID : 17182619 Kim T , Cooper JA , Sept D . 2010 . The interaction of capping protein with the barbed end of the actin filament . Journal of Molecular Biology 404 : 794 \u2013 802 . DOI : https : / / doi . org / 10 . 1016 / j . jmb . 2010 . 10 . 017 , PMID : 20969875 Kim T , Ravilious GE , Sept D , Cooper JA . 2012 . Mechanism for CARMIL protein inhibition of heterodimeric actin - capping protein . Journal of Biological Chemistry 287 : 15251 \u2013 15262 . DOI : https : / / doi . org / 10 . 1074 / jbc . M112 . 345447 , PMID : 22411988 Kuhn JR , Pollard TD . 2005 . Real - time measurements of actin filament polymerization by total internal reflection fluorescence microscopy . Biophysical Journal 88 : 1387 \u2013 1402 . DOI : https : / / doi . org / 10 . 1529 / biophysj . 104 . 047399 , PMID : 15556992 Liang Y , Niederstrasser H , Edwards M , Jackson CE , Cooper JA . 2009 . Distinct roles for CARMIL isoforms in cell migration . Molecular Biology of the Cell 20 : 5290 \u2013 5305 . DOI : https : / / doi . org / 10 . 1091 / mbc . e08 - 10 - 1071 , PMID : 19846667 Meacham CE , Ho EE , Dubrovsky E , Gertler FB , Hemann MT . 2009 . In vivo RNAi screening identifies regulators of actin dynamics as key determinants of lymphoma progression . Nature Genetics 41 : 1133 \u2013 1137 . DOI : https : / / doi . org / 10 . 1038 / ng . 451 Mejillano MR , Kojima S , Applewhite DA , Gertler FB , Svitkina TM , Borisy GG . 2004 . Lamellipodial versus filopodial mode of the actin nanomachinery : pivotal role of the filament barbed end . Cell 118 : 363 \u2013 373 . DOI : https : / / doi . org / 10 . 1016 / j . cell . 2004 . 07 . 019 , PMID : 15294161 Miyoshi T , Tsuji T , Higashida C , Hertzog M , Fujita A , Narumiya S , Scita G , Watanabe N . 2006 . Actin turnover \u2013 dependent fast dissociation of capping protein in the dendritic nucleation actin network : evidence of frequent filament severing . The Journal of Cell Biology 175 : 947 \u2013 955 . DOI : https : / / doi . org / 10 . 1083 / jcb . 200604176 Narita A , Takeda S , Yamashita A , Mae\u00b4da Y . 2006 . Structural basis of actin filament capping at the barbed - end : a cryo - electron microscopy study . The EMBO Journal 25 : 5626 \u2013 5633 . DOI : https : / / doi . org / 10 . 1038 / sj . emboj . 7601395 Nevalainen EM , Skwarek - Maruszewska A , Braun A , Moser M , Lappalainen P . 2009 . Two biochemically distinct and tissue - specific twinfilin isoforms are generated from the mouse Twf2 gene by alternative promoter usage . Biochemical Journal 417 : 593 \u2013 600 . DOI : https : / / doi . org / 10 . 1042 / BJ20080608 Nevalainen EM , Braun A , Vartiainen MK , Serlachius M , Andersson LC , Moser M , Lappalainen P . 2011 . Twinfilin - 2a is dispensable for mouse development . PLoS ONE 6 : e22894 . DOI : https : / / doi . org / 10 . 1371 / journal . pone . 0022894 , PMID : 21876732 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 26 of 28 Research article Cell Biology Paavilainen VO , Hellman M , Helfer E , Bovellan M , Annila A , Carlier M - F , Permi P , Lappalainen P . 2007 . Structural basis and evolutionary origin of actin filament capping by twinfilin . PNAS 104 : 3113 \u2013 3118 . DOI : https : / / doi . org / 10 . 1073 / pnas . 0608725104 Paavilainen VO , Oksanen E , Goldman A , Lappalainen P . 2008 . Structure of the actin - depolymerizing factor homology domain in complex with actin . The Journal of Cell Biology 182 : 51 \u2013 59 . DOI : https : / / doi . org / 10 . 1083 / jcb . 200803100 Palmgren S , Ojala PJ , Wear MA , Cooper JA , Lappalainen P . 2001 . Interactions with PIP 2 , ADP - actin monomers , and capping protein regulate the activity and localization of yeast twinfilin . The Journal of Cell Biology 155 : 251 \u2013 260 . DOI : https : / / doi . org / 10 . 1083 / jcb . 200106157 Peng AW , Belyantseva IA , Hsu PD , Friedman TB , Heller S . 2009 . Twinfilin 2 regulates actin filament lengths in cochlear stereocilia . Journal of Neuroscience 29 : 15083 \u2013 15088 . DOI : https : / / doi . org / 10 . 1523 / JNEUROSCI . 2782 - 09 . 2009 , PMID : 19955359 Pollard TD , Borisy GG . 2003 . Cellular motility driven by assembly and disassembly of actin filaments . Cell 112 : 453 \u2013 465 . DOI : https : / / doi . org / 10 . 1016 / S0092 - 8674 ( 03 ) 00120 - X , PMID : 12600310 Poukkula M , Kremneva E , Serlachius M , Lappalainen P . 2011 . Actin - depolymerizing factor homology domain : A conserved fold performing diverse roles in cytoskeletal dynamics . Cytoskeleton 68 : 471 \u2013 490 . DOI : https : / / doi . org / 10 . 1002 / cm . 20530 Rzadzinska AK , Nevalainen EM , Prosser HM , Lappalainen P , Steel KP . 2009 . MyosinVIIa interacts with Twinfilin - 2 at the tips of mechanosensory stereocilia in the inner ear . PLoS ONE 4 : e7097 . DOI : https : / / doi . org / 10 . 1371 / journal . pone . 0007097 , PMID : 19774077 Schafer DA , Korshunova YO , Schroer TA , Cooper JA . 1994 . Differential localization and sequence analysis of capping protein beta - subunit isoforms of vertebrates . The Journal of Cell Biology 127 : 453 \u2013 465 . DOI : https : / / doi . org / 10 . 1083 / jcb . 127 . 2 . 453 , PMID : 7929588 Schafer DA , Hug C , Cooper JA . 1995 . Inhibition of CapZ during myofibrillogenesis alters assembly of actin filaments . The Journal of Cell Biology 128 : 61 \u2013 70 . DOI : https : / / doi . org / 10 . 1083 / jcb . 128 . 1 . 61 , PMID : 7822423 Schafer DA , Jennings PB , Cooper JA . 1996 . Dynamics of capping protein and actin assembly in vitro : uncapping barbed ends by polyphosphoinositides . The Journal of Cell Biology 135 : 169 \u2013 179 . DOI : https : / / doi . org / 10 . 1083 / jcb . 135 . 1 . 169 , PMID : 8858171 Schindelin J , Arganda - Carreras I , Frise E , Kaynig V , Longair M , Pietzsch T , Preibisch S , Rueden C , Saalfeld S , Schmid B , Tinevez JY , White DJ , Hartenstein V , Eliceiri K , Tomancak P , Cardona A . 2012 . Fiji : an open - source platform for biological - image analysis . Nature Methods 9 : 676 \u2013 682 . DOI : https : / / doi . org / 10 . 1038 / nmeth . 2019 , PMID : 22743772 Shekhar S , Kerleau M , Ku\u00a8hn S , Pernier J , Romet - Lemonne G , Je\u00b4gou A , Carlier MF . 2015 . Formin and capping protein together embrace the actin filament in a me\u00b4nage a ` trois . Nature Communications 6 : 8730 . DOI : https : / / doi . org / 10 . 1038 / ncomms9730 , PMID : 26564775 Shin M , van Leeuwen J , Boone C , Bretscher A . 2018 . Yeast Aim21 / Tda2 both regulates free actin by reducing barbed end assembly and forms a complex with Cap1 / Cap2 to balance actin assembly between patches and cables . Molecular Biology of the Cell 29 : 923 \u2013 936 . DOI : https : / / doi . org / 10 . 1091 / mbc . E17 - 10 - 0592 , PMID : 2 9467252 Sinnar SA , Antoku S , Saffin JM , Cooper JA , Halpain S . 2014 . Capping protein is essential for cell migration in vivo and for filopodial morphology and dynamics . Molecular Biology of the Cell 25 : 2152 \u2013 2160 . DOI : https : / / doi . org / 10 . 1091 / mbc . e13 - 12 - 0749 , PMID : 24829386 Soeno Y , Abe H , Kimura S , Maruyama K , Obinata T . 1998 . Generation of functional beta - actinin ( CapZ ) in an E . coli expression system . Journal of Muscle Research and Cell Motility 19 : 639 \u2013 646 . DOI : https : / / doi . org / 10 . 1023 / A : 1005329114263 , PMID : 9742448 Spudich JA , Watt S . 1971 . The regulation of rabbit skeletal muscle contraction . I . biochemical studies of the interaction of the tropomyosin - troponin complex with actin and the proteolytic fragments of myosin . The Journal of Biological Chemistry 246 : 4866 \u2013 4871 . PMID : 4254541 Stark BC , Lanier MH , Cooper JA . 2017 . CARMIL family proteins as multidomain regulators of actin - based motility . Molecular Biology of the Cell 28 : 1713 \u2013 1723 . DOI : https : / / doi . org / 10 . 1091 / mbc . e17 - 01 - 0019 Stumpff J , von Dassow G , Wagenbach M , Asbury C , Wordeman L . 2008 . The kinesin - 8 motor Kif18A suppresses kinetochore movements to control mitotic chromosome alignment . Developmental Cell 14 : 252 \u2013 262 . DOI : https : / / doi . org / 10 . 1016 / j . devcel . 2007 . 11 . 014 , PMID : 18267093 Takeda S , Minakata S , Koike R , Kawahata I , Narita A , Kitazawa M , Ota M , Yamakuni T , Mae\u00b4da Y , Nitanai Y . 2010 . Two distinct mechanisms for actin capping protein regulation \u2013 steric and allosteric inhibition . PLoS Biology 8 : e1000416 . DOI : https : / / doi . org / 10 . 1371 / journal . pbio . 1000416 , PMID : 20625546 Tanaka K , Takeda S , Mitsuoka K , Oda T , Kimura - Sakiyama C , Mae\u00b4da Y , Narita A . 2018 . Structural basis for cofilin binding and actin filament disassembly . Nature Communications 9 . DOI : https : / / doi . org / 10 . 1038 / s41467 - 018 - 04290 - w Taoka M , Ichimura T , Wakamiya - Tsuruta A , Kubota Y , Araki T , Obinata T , Isobe T . 2003 . V - 1 , a protein expressed transiently during murine cerebellar development , regulates actin polymerization via interaction with capping protein . Journal of Biological Chemistry 278 : 5864 \u2013 5870 . DOI : https : / / doi . org / 10 . 1074 / jbc . M211509200 , PMID : 12488317 Uruno T , Remmert K , Hammer JA . 2006 . CARMIL is a potent capping protein antagonist : identification of a conserved CARMIL domain that inhibits the activity of capping protein and uncaps capped actin filaments . The Journal of Biological Chemistry 281 : 10635 \u2013 10650 . DOI : https : / / doi . org / 10 . 1074 / jbc . M513186200 , PMID : 16434392 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 27 of 28 Research article Cell Biology Vartiainen M , Ojala PJ , Auvinen P , Pera\u00a8nen J , Lappalainen P . 2000 . Mouse A6 / twinfilin is an actin monomer - binding protein that localizes to the regions of rapid actin dynamics . Molecular and Cellular Biology 20 : 1772 \u2013 1783 . DOI : https : / / doi . org / 10 . 1128 / MCB . 20 . 5 . 1772 - 1783 . 2000 , PMID : 10669753 Vartiainen MK , Sarkkinen EM , Matilainen T , Salminen M , Lappalainen P . 2003 . Mammals have two twinfilin isoforms whose subcellular localizations and tissue distributions are differentially regulated . Journal of Biological Chemistry 278 : 34347 \u2013 34355 . DOI : https : / / doi . org / 10 . 1074 / jbc . M303642200 , PMID : 12807912 Wahlstro\u00a8m G , Vartiainen M , Yamamoto L , Mattila PK , Lappalainen P , Heino TI . 2001 . Twinfilin is required for actin - dependent developmental processes in Drosophila . The Journal of Cell Biology 155 : 787 \u2013 796 . DOI : https : / / doi . org / 10 . 1083 / jcb . 200108022 Wang D , Zhang L , Zhao G , Wahlstrom G , Heino TI , Chen J , Zhang YQ . 2010 . Drosophila twinfilin is required for cell migration and synaptic endocytosis . Journal of Cell Science 123 : 1546 \u2013 1556 . DOI : https : / / doi . org / 10 . 1242 / jcs . 060251 Yamada S , Uchimura E , Ueda T , Nomura T , Fujita S , Matsumoto K , Funeriu DP , Miyake M , Miyake J . 2007 . Identification of twinfilin - 2 as a factor involved in neurite outgrowth by RNAi - based screen . Biochemical and Biophysical Research Communications 363 : 926 \u2013 930 . DOI : https : / / doi . org / 10 . 1016 / j . bbrc . 2007 . 09 . 069 Yang C , Pring M , Wear MA , Huang M , Cooper JA , Svitkina TM , Zigmond SH . 2005 . Mammalian CARMIL inhibits actin filament capping by capping protein . Developmental Cell 9 : 209 \u2013 221 . DOI : https : / / doi . org / 10 . 1016 / j . devcel . 2005 . 06 . 008 , PMID : 16054028 Zwolak A , Uruno T , Piszczek G , Hammer JA , Tjandra N . 2010 . Molecular basis for barbed end uncapping by CARMIL homology domain 3 of mouse CARMIL - 1 . Journal of Biological Chemistry 285 : 29014 \u2013 29026 . DOI : https : / / doi . org / 10 . 1074 / jbc . M110 . 134221 , PMID : 20630878 Johnston et al . eLife 2018 ; 7 : e41313 . DOI : https : / / doi . org / 10 . 7554 / eLife . 41313 28 of 28 Research article Cell Biology",
    "howes2010clathrinindependent": "JCB : Article The Rockefeller University Press $ 30 . 00 J . Cell Biol . Vol . 190 No . 4 675 \u2013 691 www . jcb . org / cgi / doi / 10 . 1083 / jcb . 201002119 JCB 675 Correspondence to Robert G . Parton : r . parton @ imb . uq . edu . au M . Kirkham\u2019s present address is Department of Cell and Molecular Biology , Karolinska Institute , 17177 Stockholm , Sweden . J . Riches\u2019 present address is European Molecular Biology Laboratory , Meyerhofstrasse 1 , 69117 Heidelberg , Germany . K . Cortese\u2019s present address is Centro di Ricerca MicroSCoBio / IFOM , FIRC Institute of Molecular Oncology , Universit\u00e0 di Genova , 16126 Genova , Italy . F . Simpson\u2019s and M . M . Hill\u2019s present address is The University of Queensland , Diamentina Institute for Cancer , Immunology and Metabolic Medicine , Brisbane , Queeensland 4102 , Australia . R . Lundmark\u2019s present address is Department of Medical Biochemistry and Biophysics , Ume\u00e5 University , 901 87 Ume\u00e5 , Sweden . Abbreviations used in this paper : AA , ascorbic acid ; CIE , clathrin - independent endocytosis ; CLIC , clathrin - independent carrier ; CME , clathrin - mediated endo - cytosis ; CTxB , cholera toxin B subunit ; GPI - AP , glycosylphosphatidylinositol - anchored protein ; MEF , mouse embryonic fibroblast ; PM , plasma membrane ; RE , recycling endosome ; Tf , transferrin ; WT , wild type . Introduction Endocytosis provides a crucial and dynamic interface between the extracellular milieu and the interior of the cell . This interface is paramount for fluid and nutrient uptake , signaling regulation , lipid homeostasis , plasma membrane ( PM ) remodeling , synaptic vesicle recycling , and cellular motility ( Grande - Garc\u00eda et al . , 2007 ; Mayor and Pagano , 2007 ; Idone et al . , 2008b ; Sigismund et al . , 2008 ) . Correspondingly , the endocytic system entails a rich diversity of mechanisms coordinated across several pathways . Mounting evidence suggests that eukaryotic cells maintain up to five distinct , constitutive pinocytic pathways ( excluding A lthough the importance of clathrin - and caveolin - independent endocytic pathways has recently emerged , key aspects of these routes remain un - known . Using quantitative ultrastructural approaches , we show that clathrin - independent carriers ( CLICs ) ac - count for approximately three times the volume internal - ized by the clathrin - mediated endocytic pathway , forming the major pathway involved in uptake of fluid and bulk membrane in fibroblasts . Electron tomographic analysis of the 3D morphology of the earliest carriers shows that they are multidomain organelles that form a complex sorting station as they mature . Proteomic analysis pro - vides direct links between CLICs , cellular adhesion turn - over , and migration . Consistent with this , CLIC - mediated endocytosis of key cargo proteins , CD44 and Thy - 1 , is polarized at the leading edge of migrating fibroblasts , while transient ablation of CLICs impairs their ability to migrate . These studies provide the first quantitative ultra - structural analysis and molecular characterization of the major endocytic pathway in fibroblasts , a pathway that provides rapid membrane turnover at the leading edge of migrating cells . Clathrin - independent carriers form a high capacity endocytic sorting system at the leading edge of migrating cells Mark T . Howes , 1 , 2 Matthew Kirkham , 1 James Riches , 2 Katia Cortese , 1 , 2 Piers J . Walser , 1 Fiona Simpson , 1 Michelle M . Hill , 1 Alun Jones , 1 Richard Lundmark , 3 Margaret R . Lindsay , 1 , 2 Delia J . Hernandez - Deviez , 1 Gordana Hadzic , 4 Adam McCluskey , 4 Rumasia Bashir , 5 Libin Liu , 6 Paul Pilch , 6 Harvey McMahon , 3 Phillip J . Robinson , 7 John F . Hancock , 1 Satyajit Mayor , 8 and Robert G . Parton 1 , 2 1 The Institute for Molecular Bioscience and 2 The Centre for Microscopy and Microanalysis , The University of Queensland , Brisbane , Queensland 4072 , Australia 3 MRC Laboratory of Molecular Biology , Cambridge CB2 0QH , England , UK 4 Chemistry , School of Environmental and Life Sciences , The University of Newcastle , Callaghan , NSW 2308 , Australia 5 School of Biological and Biomedical Sciences , Durham University , Durham DH1 3HP , England , UK 6 Department of Biochemistry , Boston University School of Medicine , Boston , MA 02118 7 Children\u2019s Medical Research Institute , The University of Sydney , Sydney , NSW 2145 , Australia 8 National Centre for Biological Science ( TIFR ) , Bangalore 560 065 , India \u00a9 2010 Howes et al . This article is distributed under the terms of an Attribution \u2013 Noncommercial \u2013 Share Alike \u2013 No Mirror Sites license for the first six months after the pub - lication date ( see http : / / www . rupress . org / terms ) . After six months it is available under a Creative Commons License ( Attribution \u2013 Noncommercial \u2013 Share Alike 3 . 0 Unported license , as described at http : / / creativecommons . org / licenses / by - nc - sa / 3 . 0 / ) . T H E J O U R N A L O F C E L L B I O L O G Y D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 JCB \u2022 VOLUME 190 \u2022 NUMBER 4 \u2022 2010 676 Results CLICs account for the major constitutive uptake of fluid and bulk membrane in fibroblasts Quantitation of the volume and surface area internalizing through the CLIC pathway is vital for understanding the contribution of this route to total endocytosis and PM homeostasis . Generic en - docytic markers that enter via CME , caveolae , and CIE routes , such as HRP or CTxB , have been used in conjunction with ultra - structural morphology and dynamin independence to define CLICs ( Kirkham et al . , 2005 ) . Therefore , we first used HRP - conjugated CTxB ( CTxB - HRP ) as an endocytic membrane marker with a previously characterized diaminobenzidine ( DAB ) incubation protocol for EM visualization ( Kirkham et al . , 2005 ) . This method reveals only internalized carriers and can be used to label such carriers within seconds of scission . Although CTxB - HRP is an excellent membrane marker , a multivalent toxin could theoretically affect endocytosis , particularly when combined with an incubation step at low temperature ( R\u00f6mer et al . , 2007 ) . Therefore , quantitative measurements of the very early uptake of the fluid phase marker , HRP ( at 20 mg . ml 2 1 ) , were compared with CTxB - HRP and assessed in the presence or absence of unlabeled CTxB ( Fig . 1 , A and B ) to determine the relative contribution of CME and CLICs in both wild - type ( WT ) and caveolin1 - null ( Cav1 2 / 2 ) mouse embryonic fibroblasts ( MEFs ) . These experiments allowed us to eliminate a cold step and showed that HRP and CTxB - HRP labeled identical compart - ments , whereas CTxB did not quantitatively affect endocytosis ( Fig . 1 , C \u2013 E ) . A defining characteristic of the CLIC pathway is indepen - dence of dynamin for PM scission ( Damke et al . , 1994 ; Parton et al . , 1994 ; Henley et al . , 1998 ; Lamaze et al . , 2001 ; Pelkmans et al . , 2002 ; Sabharanjak et al . , 2002 ; Kirkham et al . , 2005 ) . Therefore , we used a well - characterized small molecule inhibi - tor of dynamin , dynasore ( unpublished data ) , as well as a more potent , hydroxylated analogue of dynasore , Dyngo4a ( unpub - lished data ) to inhibit CME , caveolar endocytosis , and dynamin - dependent CIE pathways . The activity of these inhibitors in WT and Cav1 2 / 2 MEFs was confirmed by assessing the fluorescence uptake of the CME cargo protein transferrin ( Tf - 647 ) and the generic endocytic marker CTxB - 555 ( Fig . 1 F ) , as well as visu - alization of CTxB - HRP \u2013 labeled carriers by EM ( Fig . 1 G ) . Although Tf - 647 and CTxB - 555 were both efficiently internal - ized in untreated control cells , after treatment with 30 \u00b5M Dyngo4a ( or 80 \u00b5M dynasore ) only CTxB - 555 internaliza - tion occurred . In agreement with this observation , in untreated control cells , CTxB - HRP labeling was found in both CLICs ( Fig . 1 G , arrows ) and clathrin - coated vesicles ( CCVs ; Fig . 1 G , double arrowheads ) , visualized by EM . However , in the pres - ence of Dyngo4a , only CLICs , defined by ultrastructure , were active ( Fig . 1 G , arrows ) . Unlabeled clathrin - coated pits ( CCPs ) were readily observed but were surface connected ( Fig . 1 G , arrowheads ) . Together , these data show that CTxB is a useful marker for CLICs and , when combined with acute dynamin inhibition using Dyngo4a , CLICs remain the only detectably active endocytic pathway . phagocytosis and macropinocytosis ; Doherty and McMahon , 2009 ) . These are clathrin - mediated endocytosis ( CME ) , caveolae , and three noncaveolar clathrin - independent endocytic ( CIE ) routes , which are defined based on specific regulation by RhoA ( Lamaze et al . , 2001 ) , Arf6 ( Radhakrishna and Donaldson , 1997 ) , or Cdc42 ( Sabharanjak et al . , 2002 ) . Many of the cargo and regula - tors of the classical CME and caveolae pathways have been identi - fied and the dynamics of these pathways are the best understood ( for reviews see Conner and Schmid , 2003 ; Parton and Simons , 2007 ) . Comparatively , very little is known about the mechanisms governing CIE , although recent work has emphasized the crucial importance of these pathways ( Mayor and Pagano , 2007 ) . The existence of CIE pathways was first proposed from the study of plant and bacterial toxin internalization , in combi - nation with ultrastructural methods or with tools that inhibited CME ( Montesano et al . , 1982 ; Sandvig et al . , 1987 ; Ricci et al . , 2000 ) . The plant toxin ricin was used to first describe the clathrin - independent , apical endocytic pathway in endothelial cells ( Sandvig and van Deurs , 1994 ) . Bacterial toxins , such as the Helicobacter pylori VacA toxin ( Gauthier et al . , 2007 ) and chol - era toxin ( Torgersen et al . , 2001 ) provided further evidence for clathrin - independent endocytic pathways . However , a lack of specific tools to study these noncanonical routes has made it difficult to identify their guiding parameters . Ultrastructural work on the CIE pathway regulated by Cdc42 has identified the morphologically distinct clathrin - independent carriers ( CLICs ) as the primary carriers involved in uptake within this route ( Kirkham et al . , 2005 ) . Parameters currently identified for the CLIC pathway include dynamin - independent PM scission ( Sabharanjak et al . , 2002 ) , enrichment in GPI - anchored proteins ( Sabharanjak et al . , 2002 ) , reliance on Arf1 activity ( Sabharanjak et al . , 2002 ; Kumari and Mayor , 2008 ) , sensitivity to cholesterol depletion ( Kirkham et al . , 2005 ; Chadda et al . , 2007 ) , and contribution to a significant fraction of cholera toxin B subunit ( CTxB ) and fluid internalization ( Kirkham et al . , 2005 ) . CLICs arise directly from the PM and mature into the GPI - enriched early endosomal compartment ( GEEC ) , as they acquire Rab5 and EEA - 1 and , consequently , merge with early endosomes ( EEs ; Kalia et al . , 2006 ) . Ongoing work is linking cargo and regulators to these carriers , such as GTPase regulator associated with FAK ( GRAF1 ) , dysferlin , and Snx9 ( Yarar et al . , 2007 ; Hern\u00e1ndez - Deviez et al . , 2008 ; Lundmark et al . , 2008 ) . The basic parameters of these carriers , however , remain unknown . Here we characterize key parameters of the CLIC pathway . A quantitative analysis of the volume and surface area of CLICs identifies them as the major constitutive pathway involved in bulk endocytosis within fibroblasts . High resolution electron to - mography shows the complexity of these primary endocytic car - riers and , in conjunction with recruitment of classical trafficking regulators , they are realized to be intricate sorting compartments . Identification of novel CLIC cargo suggests links to cellular ad - hesion turnover and plasma membrane repair . Consistent with this , CLICs become polarized in migrating fibroblasts , turning over adhesion complex components , such as Thy - 1 and CD44 , at the leading edge . Inhibition of the CLIC pathway subsequently impairs the capacity of fibroblasts to migrate . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 677 Characterization of the CLIC pathway \u2022 Howes et al . Figure 1 . CTxB does not affect CLIC endocytosis . ( A ) Cav1 2 / 2 MEFs were incubated with HRP in the presence or absence of CTxB for 15 s . Cells were cooled and DAB reaction perfomed in the presence of ascorbic acid ( AA ) before fixation . Bars , 200 nm . ( B ) Quantitation of the number of HRP - filled carriers per cell across 10 \u2013 12 cells treated as in A . CLICs were defined by their characteristic ring - shaped morphology , CCVs were defined as coated vesicular carriers . Error bars show SEM . ( C ) Cav1 2 / 2 MEFs were grown in normal media ( plus energy ) , media containing 2 - deoxyglucose ( no energy ) , or 2 - deoxyglucose media for 1 h followed by a 1 - h washout in normal media ( recovery ) before CTxB - HRP internalization and DAB reaction . Labeled structures were counted across 10 cells . Error bars show SEM . ( D and E ) Cav1 2 / 2 MEFs were incubated with CTxB - HRP for 5 min before the DAB reaction for either 5 min at 37\u00b0C or 10 min at 4\u00b0C , followed by fixation at 37\u00b0C . All CTxB - HRP \u2013 positive structures were counted in 10 \u2013 12 cells across two independent experiments . Error bars show SEM . Bar , 200 nm . ( F ) NIH3T3s were treated for 20 min with vehicle ( Untreated ) or Dyngo4a before internalization of CTxB - 555 ( left panels ) and Tf - 647 ( right panels ) for 5 min . Cells were placed on ice and acid stripped to remove surface labeling . Cells were then bound with CTxB - 488 ( middle panels ) . Bar , 10 \u00b5m . ( G ) NIH3T3 cells were left untreated or were treated with Dyngo4a before internalization of CTxB - HRP , followed by DAB reaction . Arrows show CTxB - HRP \u2013 laden CLICs , double arrowheads show internalized , labeled CCVs , arrowheads show surface connected unlabeled CCPs . Bars , 200 nm . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 JCB \u2022 VOLUME 190 \u2022 NUMBER 4 \u2022 2010 678 We next estimated the contribution of this pathway to total plasma membrane surface endocytosis . By measuring the surface - to - volume ratio ( S ( v ) ; see Materials and methods ) of a typical CLIC ( Fig . 2 F ) and the absolute volume of an average Cav1 2 / 2 MEF ( 2 , 346 . 7 \u00b1 289 . 3 \u00b5m 3 ; see Materials and meth - ods ) , the compiled membrane surface area of CLICs was calcu - lated after 15 s of internalization to be 97 . 1 \u00b1 10 . 9 \u00b5m 2 . Given that the estimated average PM surface area of a Cav1 2 / 2 MEF was found to be 4 , 375 . 7 \u00b5m 2 ( see Materials and methods ) , the equivalent of the total PM surface area could internalize through CLICs in less than 12 min ( or 17 min after most extreme correc - tion factor ; see Materials and methods ) . Together these data indicate that CLICs provide a substantial capacity of internal - ization , more than any other constituite endocytic pathway in fibroblast cells . CLICs exhibit the properties of a complex sorting compartment Aside from the 2D morphological description of tubular CLICs , there is currently no data describing the 3D attributes of the pri - mary carriers . To comprehensively visualize the ultrastructural properties of CLICs , high resolution electron tomography was performed on CTxB - HRP \u2013 laden structures ( Fig . 3 A ; Video 1 ) . As a membrane marker , CTxB - HRP provided superior delinea - tion of membrane profiles compared with free HRP at the con - centrations and very early time points used here . CTxB - HRP was internalized for only 15 s to capture the earliest morphology of the carriers . The DAB reaction was performed on live cells to stabilize the morphology of the structures before glutaraldehyde fixation , simultaneously providing the necessary contrast for high resolution tomography and maintaining a close - to - native morphology . Morphology of DAB - labeled structures was simi - lar in cells fixed before DAB reaction ( not depicted ) . Images of 3D rendering from electron densities are shown ( Fig . 3 A ) . There is clear evidence of vesicular membrane invaginations ( Fig . 3 A , arrows ) , excluded from CTxB - HRP labeling , indicating that CLICs may have a capacity to bud internal membranes . Addi - tionally , the tubular structure forms a complete ring , a facet pre - viously overlooked ( Fig . 3 A , arrowhead ) . As chemical fixatives can alter the morphology of endosomes ( Murk et al . , 2003 ) , it was also important to confirm these observations without the use of aldehyde fixation . Therefore , CTxB - HRP was internalized for 15 s , DAB reaction performed , and the samples were high pres - sure frozen ( HPF ) as described previously ( Richter et al . , 2008 ) . HPF CTxB - HRP \u2013 laden CLICs showed identical features to chemically fixed samples ( Fig . 3 B ) , with vesicular invaginations from the limiting membrane and tubular ring - shaped extensions . Rather than simple tubules or vesicles , these primary carriers have a surprising degree of complexity . Identification of novel CLIC - associated proteins To identify CLIC associated components , we next sought to en - rich CLICs away from other cellular material by density gradi - ent fractionation ( see Materials and methods ) . To increase the number of CLICs within a population of cells , the PI 3 K inhibi - tor wortmannin was used within a previously developed system Having established the constitutive nature of the path - way and lack of detectable perturbation by the used markers , we used EM - based stereology to calculate the proportion of total cytoplasmic volume ( V ( v ) ) contributed by the CLIC pathway . We used both Cav1 2 / 2 and WT MEFs to compare cells with and without caveolae , as well as a fibroblast cell line , NIH3T3 ( Fig . 2 A ) . In untreated Cav1 2 / 2 MEFs , total CTxB - labeled structures occupied 0 . 10 \u00b1 0 . 01 % of the total cytoplasmic volume after 15 s of internalization , 0 . 19 \u00b1 0 . 01 % after 1 min , and 0 . 50 \u00b1 0 . 04 % after 2 min ( Fig . 2 B , Untreated ) . WT MEFs internalized similar volumes as Cav1 2 / 2 MEFs , in - dicating that caveolae contributed an insignificant proportion of internalization at this time . Intriguingly , when treated with Dyngo4a , under conditions where uptake of Tf is completely blocked and CLICs remain the only detectably active pathway , internalized CTxB - HRP still accounted for 0 . 09 \u00b1 0 . 01 % of cytoplasmic volume after 15 s , 0 . 16 \u00b1 0 . 02 % after 1 min , and 0 . 37 \u00b1 0 . 02 % after 2 min ( Fig . 2 B , Dyngo4a ) . This indicates that 90 . 0 % of the total endocytic volume is contributed by the CLIC pathway after 15 s of uptake , 85 . 9 % after 1 min , and 74 . 0 % after 2 min . Confirmation of such high contribution of CLICs to total internalization was provided using strict ultrastructural criteria to compare the volume of tubular / ring - shaped structures ( CLICs ) and coated / vesicular structures ( CCVs ) in NIH3T3 cells that were not treated with any inhibitor ( Fig . 2 C ) . Note that previ - ous studies have shown that at the earliest times used here , Tf - HRP only labels CCVs as defined morphologically ( Kirkham et al . , 2005 ) but CTxB - HRP labels CCVs , caveolae , and CLICs . NIH3T3 cells internalized similar volumes to primary MEFs , with total CTxB - labeled structures after 2 min accounting for 0 . 47 \u00b1 0 . 014 % of cytoplasmic volume and tubular / ring ele - ments ( CLICs ) contributing 78 . 7 % ( 0 . 37 \u00b1 0 . 014 % ) of that vol - ume ( Fig . 2 C ) . This alternative method avoids the need for inhibition of the dynamin - dependent pathways and suggests that under these conditions CLICs are not a compensatory pathway induced after acute dynamin inhibition ( Damke et al . , 1995 ) . To confirm the striking contribution of the CLIC pathway to cellular internalization , V ( v ) measurements were next collected using free HRP , which cannot bind to mem - branes and induce clustering of microdomains ( Fig . 2 D ) . Again , CLICs , morphologically defined as tubules / rings , con - stituted \ue07a 74 % of all HRP internalization after 2 min ( V ( v ) 0 . 32 \u00b1 0 . 05 % ) . To further confirm these results using an alternative method for assessment of endocytic contribution , CTxB was biotinyl - ated and internalized in Cav1 2 / 2 MEFs , in the presence or ab - sence of Dyngo4a . Surface - accessible biotin was cleaved and the remaining , internalized biotin measured by Western blot using streptavidin - HRP . Using this alternative biochemcial method CTxB - biotin levels in Dyngo4a - treated cells were equivalent to 76 . 0 \u00b1 8 . 4 % of that in untreated control cells after 2 min of uptake ( Fig . 2 E ) . Agreement between morphometric and biochemcial approaches ( and with V ( v ) measurements using a specific marker of the CLIC pathway : see below ) provides important confirma - tion of the validity of the methods used and the significant mag - nitude of the CLIC pathway . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 679 Characterization of the CLIC pathway \u2022 Howes et al . Figure 2 . Quantitation of CLIC endocytosis . ( A ) Cav1 2 / 2 or wild - type ( WT ) MEFs were left untreated or were treated with Dyngo4a . CTxB - HRP was inter - nalized for 15 s , 1 min , or 2 min . Examples of CTxB - labeled structures after 15 s of uptake are shown ( inset , left panel ) . Substratum indicated by large arrowhead , grid sizes are 2 , 000 or 200 nm , examples of intersections shown by arrows . ( B ) 20 \u2013 25 cells treated as in A were used to calculate the volume fraction ( V ( v ) ) . Error bars show SEM . ( C ) NIH3T3 cells not treated with inhibitor were processed as in A and counted as in B . V ( v ) was calculated for both tubular and vesicular structures . Error bars show SEM . ( D ) Cav1 2 / 2 MEFs were incubated with HRP for 15 s , 1 min or 2 min and HRP - laden carriers counted as in B . Error bars show SEM . ( E ) CTxB was conjugated with NHS - SS - biotin and added to untreated Cav1 2 / 2 MEFs or Cav1 2 / 2 MEFs treated with Dyngo4a for 15 s or 2 min or was bound to untreated cells on ice for 10 min ( Surface + MesNa ) . Cells were placed on ice and residual surface biotin cleaved with MesNa . Western blots of cell lysates were probed with streptavidin - HRP . Chart represents the average intensity of streptavidin - HRP across three independent experiments . Residue luminescence in Surface + MesNa samples indicates level of uncleavable biotin . Error bars show SEM . ( F ) Absolute volume of CLICs was estimated from the volume fraction , V ( v ) , multiplied by the average volume of a Cav1 2 / 2 MEF , 2 , 347 \u00b5m 2 . Surface density ( S ( v ) ) was calculated from high resolution images of labeled structures using a cycloid grid as described in Materials and methods and multiplied by the absolute volume to give absolute surface area . Volume of a single carrier was calculated as described in Materials and methods . Number of CLIC budding events per minute per cell was calculated based on the absolute volume internalized by all CLICs divided by the volume of a single carrier . Volume adjustments for overprojection effects are in brackets ( see Materials and methods ) . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 JCB \u2022 VOLUME 190 \u2022 NUMBER 4 \u2022 2010 680 Figure 3 . 3D morphology of CLICs . ( A ) Cav1 2 / 2 MEFs were incubated with CTxB - HRP for 20 min on ice , before internalization for 15 s . The DAB reac - tion was performed and cells were processed for electron tomography ( see Materials and methods ) . A single section of the original tomogram is shown ( left ) . Various rotations of a 3D contoured electron density render were generated ( middle ) . Enlarged sections selected from the tomogram ( right ) show internal vesicles ( arrows ) and a complete connection around the circumference of the structure ( arrowhead ) . ( B ) WT MEFs grown on sapphire discs were incubated with CTxB - HRP for 15 s before DAB reaction and high - pressure freezing . Tubular extensions ( large arrows ) are seen emanating from vesicular bulbs ( arrowheads ) in the characteristic ring - shaped CLIC morphology . CCPs without CTxB label ( double arrowheads ) indicate that they are still surface connected . Bars , 200 nm . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 681 Characterization of the CLIC pathway \u2022 Howes et al . Therefore , colocalization between dynamin and anti - GFP in GFP - GPI \u2013 expressing cells was assessed after treatment with wortmannin ( Fig . S3 ) . Colocalization after 10 min was signifi - cantly increased in wortmannin - treated cells compared with untreated controls ( 13 . 1 \u00b1 1 . 1 % untreated ; 39 . 2 \u00b1 2 . 0 % 100 nM wortmannin ; Fig . S3 , A and B ) , indicating that dynamin is re - cruited to CLICs rather than localizing to them from the PM . By EM , dynamin was also found on a small proportion of CTxB - HRP containing tubular / ring - shaped structures ( Fig . S3 C ) . Although the majority of dynamin coincides with the CHC profile , the small but specific concentration in Fraction 2 . 8 , as well as colocalization with GFP - GPI and CTxB , is consistent with the hypothesis that dynamin is recruited to CLICs , after PM scission . To verify that membranes in Fraction 2 . 8 represent bona fide CLICs , 3T3 - GPI cells were transfected with one of two markers for the CLIC pathway , Flot1 - HA and GRAF1 - myc . By Western blot analysis both endogenous and transfected Flot1 and GRAF1 were concentrated in Fraction 2 . 8 ( Fig . 4 F ) . When fractions from transfected cells were labeled with anti - HA or anti - myc antibod - ies , a significant proportion of the structures within Fraction 2 . 8 were labeled ( Fig . 4 G , arrows ; 31 . 5 \u00b1 2 . 5 % Flot - HA and 41 . 3 \u00b1 6 . 4 % Graf1 - myc positive ) . Comparatively , when Fraction 2 . 8 was labeled for endogenous Cav1 , very few of the structures were pos - itive ( 5 . 5 \u00b1 0 . 8 % ) and , notably , these smaller vesicular structures were distinct from the CLIC - like structures ( Fig . 4 G , arrowheads ) . Endogenous dynamin labeled a significant population of struc - tures within Fraction 2 . 8 ( Fig . 4 G , arrows ; Fig . S3 D ) , supporting the notion that dynamin can associate with CLICs . We next used this fractionation approach to identify novel CLIC components . The protein complement of Fraction 2 . 8 was examined after separation using 5 \u2013 15 % SDS - PAGE and the entire lane was cut into 2 - mm slices , digested with trypsin , and analyzed by liquid chromatography mass spectrometry ( LC - MS / MS ; see Materials and methods ) . This identified 80 proteins with two or more peptide hits of 99 % confidence ( key targets and links to endocytosis in Table I , full list in Table S1 ) . A significant cluster of proteins identified play a role in cellular adhesion and inter - action with the ECM . Thy - 1 , a glycosylphosphatidylinositol - anchored protein ( GPI - AP ) , is a critical component of focal adhesion disassembly through FAK phosphorylation ( Rege et al . , 2006 ) . This interaction is dependent on the GPI anchor of Thy - 1 , implicating the GPI - AP \u2013 enriched CLIC pathway as a mediator of FAK activation . Integrin - \ue062 1 is also a key component of focal adhesions and galectin3 is directly involved in clathrin - independent integrin - \ue062 1 endocytosis ( Furtak et al . , 2001 ) . Although a role for CME during integrin - \ue062 1 endocytosis and cellular motility has been established ( Palamidessi et al . , 2008 ) , it has also been found that integrin - \ue062 1 uptake can be indepen - dent from clathrin and is mediated by lipid rafts ( Vassilieva et al . , 2008 ) . Rap1 transmits outside - in integrin signals , leading to cell polarization , and its activation at the leading edge is necessary for chemokine - induced cellular migration ( Shimonaka et al . , 2003 ) . CD98 also mediates integrin signaling during cell spreading and is localized to cell \u2013 cell contacts ( Feral et al . , 2005 ) , where it acti - vates Rap1 ( Suga et al . , 2001 ) . Activation of Rap1 also leads to recruitment of CD44 to the leading edge ( Shimonaka et al . , 2003 ) , that inhibits the fusion of CLICs with EEs ( Kalia et al . , 2006 ; Fig . S1 ) . Using anti - GFP in GFP - GPI \u2013 expressing NIH3T3 cells ( 3T3 - GPIs ) as a marker for CLICs , we optimized conditions to concentrate GFP - GPI \u2013 positive membranes ( Fraction 1 . 2 ; Fig . 4 A ) . CCVs ( identified by clathrin heavy chain ; CHC ) , EEs and recycling endosomes ( REs ; identified by TfR ) , and Golgi ( identified by GM130 ) were below the detection limits within this fraction . However , this fraction also contained Cav1 ( caveolae ) and Grp78 ( ER ) . The 10 % fraction ( Fraction 1 . 2 ) was , therefore , subjected to a second , continuous Nycodenz gradient , which pro - vided a reproducible enrichment of anti - GFP \u2013 positive mem - branes ( Fraction 2 . 8 ; Fig . 4 A ) . To further characterize the enriched fraction , we examined the membranes in detail by whole - mount EM ( Fig . 4 B ) . Consistent with CTxB - HRP \u2013 laden CLICs visualized by EM in intact cells ( Kirkham et al . , 2005 ) , the predominant structures within Fraction 2 . 8 have the striking mor - phology of CLICs . A low magnification image shows the consis - tent morphology of the structures enriched within this fraction ( Fig . S2 A ) . At high magnification ( Fig . 4 C , arrows ) , the struc - tures within Fraction 2 . 8 were also consistent with tomography and HPF data ( Fig . 3 ) . Other structures displayed a clear connec - tion between tubular extensions and spherical bulbs ( Fig . 4 C , arrowheads ) . There was no evidence of contamination by other cellular compartments , such as mitochondria . Other fractions were enriched in structures distinct from those found in Fraction 2 . 8 ( Fig . S2 C ) . This system also gave optimal , but incomplete , separation of GFP - GPI \u2013 positive carriers from Cav1 and Grp78 . To quantify the membrane enrichment in Fraction 2 . 8 we ana - lyzed internalized CTxB - HRP within the fraction after 5 min of internalization ( Fig . 4 D , arrows ) . CTxB - HRP \u2013 labeled structures constituted 70 . 0 \u00b1 0 . 9 % of the membranes in the CLIC - enriched fraction as calculated from stereology ( Fig . S2 B ) . The validity of this method was further examined by ana - lyzing membranes within Fraction 2 . 8 after inhibition of CLIC endocytosis . Using two independent methods of CLIC perturba - tion , expression of a Cdc42 dominant - negative mutant ( Cdc42 - DN ; Sabharanjak et al . , 2002 ) and cholesterol depletion with methyl - \ue062 - cyclodextrin ( m \ue062 CD ) under conditions where CME is unaffected ( Kirkham et al . , 2005 ; Glebov et al . , 2006 ) , there was a substantial reduction of CTxB and anti - GFP \u2013 positive mem - branes within Fraction 2 . 8 ( Fig . 4 E ; 59 . 7 % reduction in CTxB levels in cells transfected with Cdc42 - DN compared with Cdc42 - WT ; 53 . 6 % reduction in anti - GFP antibody levels after treatment with m \ue062 CD ) . Thus , this approach shows concentra - tion of CLIC marker positive membranes within Fraction 2 . 8 and reduced accumulation of these membranes after treatments known to inhibit CLIC endocytosis . Unlike CHC and TfR , there was a small proportion of dynamin specifically concentrated in Fraction 2 . 8 ( Fig . 4 F ) . Although dynamin is not necessary for the formation of nascent CLICs ( Sabharanjak et al . , 2002 ; Kirkham et al . , 2005 ) , it has been previously noted that expression of the DynK44A mutant re - duces CTxB trafficking to the Golgi ( Le and Nabi , 2003 ; Kirkham et al . , 2005 ) , independently from clathrin ( Torgersen et al . , 2001 ) , suggesting a role in clathrin - independent CTxB uptake . As CLICs rapidly mature , however , it is difficult to assess a possible tran - sient recruitment of dynamin under steady - state conditions . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 JCB \u2022 VOLUME 190 \u2022 NUMBER 4 \u2022 2010 682 Figure 4 . Biochemical enrichment of CLICs . ( A ) 3T3 - GPI cells were subjected to density fractionation as described in Materials and methods . Western blots of membrane markers are shown . Within the first gradient ( left ) anti - GFP \u2013 , Cav1 - , and Grp78 - positive membranes are present in Fraction 1 . 2 , highlighted by outline . CHC - , GM130 - , and TfR - positive membranes are below the detection limit within this fraction . Within the second gradient , anti - GFP \u2013 positive membranes concentrate within Fraction 2 . 8 , highlighted by outline . This represnts a yield of 9 . 6 \u00b1 0 . 2 % of anti - GFP \u2013 positive membranes . Cav1 concen - trates in Fraction 2 . 6 and Grp78 in Fraction 2 . 10 . ( B ) Fraction 2 . 8 was fixed and visualized by EM . Structures within Fraction 2 . 8 share similar profiles to CLICs seen within intact cells . Bar , 200 nm . ( C ) High resolution electron micrographs of Fraction 2 . 8 structures , providing examples of vesiculation ( arrows ) and a spherical bulb connected to tubular extension ( arrowheads ) . Bar , 100 nm . ( D ) NIH3T3 cells were incubated with CTxB - HRP before fractionation and the DAB reaction was performed on Fraction 2 . 8 . Bar , 200 nm . ( E ) NIH3T3 cells transfected with Cdc42 - WT or - DN were incubated with CTxB before fractionation . Western blots were probed with anti - CTxB . 3T3 - GPI cells were treated or not with m \ue062 CD before incubation with anti - GFP antibodies . Western blots of fractions were probed with anti \u2013 Rb - HRP . ( F ) Western blots of fractions from cells transfected with Flot1 - HA , GRAF1 - myc , or left untransfected . Frac - tions were probed with anti - HA , - myc , - dynamin , - Flotillin1 , or - GRAF1 antibodies as appropriate . ( G ) Fixed sampled from F of Fraction 2 . 8 were labeled for anti - HA , anti - myc , endogenous Cav1 , or dynamin . Structures labeled with Flotillin - 1 - HA , GRAF1 - myc , and dynamin ( arrows ) or Cav - 1 ( arrowheads ) are shown . Bars , 200 nm . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 683 Characterization of the CLIC pathway \u2022 Howes et al . Anti \u2013 Thy - 1 , anti - CD44 , and myoferlin colocalized with CTxB - 555 but not Tf - 647 in NIH3T3 cells after 2 min of uptake , indicat - ing they are specific cargo for the CLIC pathway . Interestingly , although only a low level of colocalization was seen between anti - GFP and Rab11 after 10 min ( 17 . 07 \u00b1 1 . 49 % ) in untreated 3T3 - GPI cells ( Fig . 5 , B and C ) , after treatment with wortman - nin there was a significant increase in their colocalization after 10 min ( 41 . 12 \u00b1 2 . 79 % ; Fig . 5 , B and C ) . As wortmannin inhib - its the maturation of CLICs , this localization is not likely to occur within REs . Additionally , there was no detectable TfR in the CLIC - enriched fraction , arguing against significant contam - ination of RE membranes within this fraction . Additional evidence for the sorting capability of CLICs was identified by assessment of the endocytic maturation of CD44 . Endocytosis of CD44 was consistently Tf negative ( Fig . 5 D ) , while it colocalized with CTxB after 2 min , with less overlap after 10 min . Persistent colocalization was also seen be - tween CD44 and dysferlin from 2 to 40 min ( Fig . 5 D ) . As dys - ferlin also remains distinct from the classical endocytic network while CD44 is linked to integrin - \ue062 1 signaling by forming special - ized lipid raft microdomains ( Lee et al . , 2008 ) . CD44 is also an important component of extracellular contact , involved in the degradation of ECM , and is involved in maintenance of polar - ization and directionally persistent migration ( Mrass et al . , 2008 ) . This also implicates CLICs as a mediator of cell adhesion and , given the proposed ability for CLICs to recycle directly to the PM , presents a rapid recycling mechanism during cellular adhe - sion turnover and motility . Supporting this , moesin directs actin - based rearrangements of lipid rafts directly through CD44 to orchestrate the topology of cell surface structures , such as filopo - dia ( Bretscher et al . , 1997 ) , and the CD44 - ERM ( ezrin - radixin - moesin ) association is a critical component of directional cellular motility ( Legg et al . , 2002 ) . We next sought to verify some key targets as novel cargo for the CLIC pathway . Verification of novel CLIC proteins Preliminary verification of some key targets was performed by comparison to internalized CTxB - 555 and Tf - 647 ( Fig . 5 A ) . Table I . Key targets identified from isolation of CLIC - positive fraction Target Location Link to endocytosis 4F2 cell surface antigen heavy chain / CD98 PM Receptor for Galectin3 and Integrin \ue062 - 1 endocytosis 78 - kD glucose - regulated protein PM / ER Associates with lipid rafts and receptor for MHCI Actin , cytoplasmic 1 Cytosol Drives endocytic carrier formation Alpha - 2 - HS - glycoprotein / Fetuin PM / E Endocytosis mediated by AnnexinA2 Annexin A2 E Binds glycoproteins during endocytosis CD109 antigen PM Cell surface GPI - AP CD44 antigen PM Hyaluronan receptor Collagen ECM ECM constituent Fructose - bisphosphate aldolase A Cytosol Mediates Snx9 - dynamin interaction Galectin - 3 PM Mediates Integrin \ue062 - 1 endocytosis Guanine nucleotide - binding protein G \ue061 ( i , o ) PM Involved in STAT3 signaling Integrin \ue062 - 1 PM Receptor for ECM Lysosome - associated membrane glycoprotein 2 E Lysosomal membrane regulator Membrane - associated progesterone receptor component 1 PM / ER Involved in cholesterol homeostasis , rapidly cleared from the PM by endocytosis Moesin PM Links CD44 to cortical actin to drive features in the PM Myoferlin ( Fer - 1 \u2013 like protein 3 ) PM Involved in Ca 2 + - promoted membrane fusion ; binds to AHNAK Neuroblast differentiation - associated protein AHNAK PM Binds Myoferlin and Dysferlin and with Annexin2 mediates membrane fusion Prolow - density lipoprotein receptor - related protein 1 ( LRP1 ) PM Contains both tyrosine and di - leucine sorting motifs ; binds saposin on cell surface Prosaposin PM / E Co - receptor with LRP1 ; binds GM1 Rac1 PM / E Induces membrane ruffling Rab - 11A E Endosome recycling Rab - 14 E Endosome recycling Rab - 7a E Late endosome fusion and motility Rab - 5A E Early endosome fusion and motility Rap - 1b PM Integrin \ue062 - 1 endocytosis and focal adhesion turnover ; mediates STAT3 Receptor expression - enhancing protein 5 PM Involved in Rab - mediated membrane trafficking Reticulon - 4 PM / ER Tubulates membranes ; binds NogoR , a GPI - AP at the PM Thy - 1 membrane glycoprotein PM GPI - AP ; regulates FAK signaling Tmp21 PM Contains KDEL sequence , traffics to the PM independently from p24 Vimentin Cytosol With Annexin A2 regulates FAK signaling E , Endosomes ; ECM , extracellular matrix ; ER , endoplasmic reticulum ; PM , plasma membrane . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 JCB \u2022 VOLUME 190 \u2022 NUMBER 4 \u2022 2010 684 Figure 5 . Verification of novel CLIC cargo . ( A ) NIH3T3 cells were incubated with either anti \u2013 Thy - 1 or anti - CD44 antibodies and CTxB - 555 and Tf - 647 for 2 min . Myoferlin was detected after fixation , showing steady - state localization . Arrows indicate colocalization . Bar , 10 \u00b5m . ( B ) 3T3 - GPI cells were left untreated or were treated with 100 nM wortmannin before internalization of anti - GFP for 10 min . Arrows indicate colocalization . Bar , 10 \u00b5m . ( C ) Quantitation of B from 12 \u2013 15 cells in three independent experiments . ( D ) Quantitation of colocalization between internalized anti - CD44 antibodies and Tf - 647 , CTxB - 555 , or anti - myc antibodies for GFP - dysferlin - myc \u2013 expressing cells after 2 , 10 , and 40 min . Error bars show SEM . ( E ) Anti \u2013 CD44 - HRP was internalized into WT MEFs before DAB reaction . Arrows show anti - CD44 - HRP \u2013 positive carriers with morphology of CLICs . Arrowheads show large , tubular ring - shaped anti - CD44 - HRP \u2013 positive compartment . Bars , 200 nm . ( F ) Anti \u2013 CD44 - HRP or Tf - HRP was pulsed into Cav1 2 / 2 MEFs for 15 s , 1 min , or 2 min . Cells were fixed and processed for vertical sections . Arrows show HRP - labeled carriers . Bar , 200 nm . ( G ) Stereology measurements were captured across 20 \u2013 25 cells in three independent areas as treated in F . Error bars show SEM . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 685 Characterization of the CLIC pathway \u2022 Howes et al . To verify the polarization of CLIC endocytosis in migrat - ing fibroblasts , ultrastructure of CTxB - HRP \u2013 laden carriers was assessed after a 2 - min uptake in migrating MEFs ( Fig . 5 C , large arrow indicates direction of migration ) . Labeled CLICs were predominantly found at the leading edge ( Fig . 6 C ; 1 , 2 ) . Con - versely , an accumulation of unlabeled , surface - connected caveo - lae was seen at the trailing edge ( Fig . 6 C ; 3 , 4 insets ) . Quantitation of the number of endocytic and surface - connected structures at the leading and trailing edges confirmed this polarized distribu - tion ( Fig . 6 D ) . Both surface - connected CCPs and internalized CCVs showed no preference for the leading or trailing edge . To our knowledge , this is the first evidence for a striking spatial distinction between the CLIC pathway , caveolae , and CME , and is suggestive of a role for CLICs during dynamic adhesion turn - over specifically from the leading edge . As the CLIC pathway is polarized during migration and internalizes key adhesion components at the leading edge , we speculated that this pathway could play a critical role during cell migration . Thus , a system was developed in which CLIC - dependent uptake could be inhibited without affecting CME ( see Materials and methods ) . To optimize conditions for tran - sient ablation of the CLIC pathway , we internalized CTxB - HRP for various short times , performed DAB - mediated ablation of the labeled compartments , and then analyzed the uptake of a subsequent pulse of CTxB - 555 , Tf - 647 , and antibodies to CD44 . 2 min of CTxB - HRP internalization followed by the DAB reaction on living cells was found to give the strongest inhibition of CD44 and CTxB - 555 endocytosis without any effect on Tf - 647 uptake ( Fig . 6 E ) . Under these conditions Tf - 647 endocytosis and maturation was unaffected , even after a 40 - min pulse , indicating the endosomal network was still func - tional ( not depicted ) . The inhibition of CD44 and CTxB - 555 internalization following CTxB - HRP \u2013 based ablation persisted for at least 4 h after the ablation step , indicating this assay could be used to assess possible effects of CLIC inhibition on migration . Quantitation of fluorescence intensity shows that CTxB - HRP inactivation of the CLIC pathway resulted in an 81 . 8 \u00b1 7 . 2 % reduction in CTxB - 555 and a 93 . 8 \u00b1 2 . 6 % reduc - tion in CD44 endocytosis , whereas Tf - 647 internalization was not reduced ( increased by 28 . 6 \u00b1 24 . 8 % ) after a 4 - h incuba - tion ( Fig . 6 F ) . To test the requirement for CLIC endocytosis during mi - gration , this inactivation assay was applied to migrating fibro - blasts . The capacity of cell monolayers to migrate into a scratched wound was assessed by measuring the total distance between two cell fronts of a wound after CLIC inactivation . When CTxB - HRP was used to inactivate CLICs , without effect on other endo - cytic routes , cell monolayers had a reduced capacity to migrate into a wound after 4 h ( 15 . 8 \u00b1 1 . 7 % distance of wound closed ; Fig . 6 G ) compared with DAB only \u2013 treated cells ( 33 . 4 \u00b1 3 . 7 % closed ) . A similar effect was observed in Cav1 2 / 2 fibroblasts , ruling out the possibility that the migration defect was due to in - activation of caveolae . Together with previous studies ( Grande - Garc\u00eda et al . , 2007 ; Nishimura and Kaibuchi , 2007 ) , these findings suggest that three major endocytic routes in eukaryotic cells , CME , caveolae , and the CLIC pathway , are all important for normal directional cell migration . ( not depicted ) , this suggests that CLICs are directly sorting cargo rather than representing distinct populations of similar mecha - nisms . Confirming this , the ultrastructure of CD44 - positive car - riers , as visualized using anti \u2013 CD44 - HRP , showed identical morphology to that of CTxB - HRP \u2013 laden CLICs ( Fig . 5 E , arrows ) , whereas no label was found in the classical endosomal network . In some instances , intriguing tubular networks were seen with labeling for anti \u2013 CD44 - HRP after 40 min of internal - ization ( Fig . 5 E , arrowheads ) , possibly representing a mature compartment arisen from the CLIC pathway . The presence of anti - CD44 antibodies had no effect on the number or morphol - ogy of CTxB - HRP \u2013 laden CLICs between 2 and 40 min ( not de - picted ) , indicating that the CD44 antibodies do not influence the kinetics or ultrastructure of the pathway . Using CD44 as a novel and specific marker for CLICs , it was also possible to confirm stereology calculations using CTxB - HRP in the presence of dynamin inhibition . Anti - CD44 - HRP \u2013 conjugated antibodies or Tf - HRP was internalized in Cav1 2 / 2 MEFs and the DAB reaction was performed ( Fig . 5 , F and G ) . This method has the advantage of relying on the specificity of CD44 for the CLIC pathway and Tf - HRP for CME . We found that CD44 - positive CLICs accounted for 0 . 09 \u00b1 0 . 02 % of the total cytoplasmic volume after 15 s , 0 . 14 \u00b1 0 . 03 % after 1 min , and 0 . 34 \u00b1 0 . 01 % after 2 min ( Fig . 5 G ) . These volume estimates are in excellent agreement with those obtained using morphological criteria or using specific inhibitors of dynamin - dependent endo - cytosis ( see above ) . These data identify a range of novel CLIC cargo , confirm the significant magnitude of the pathway , and pro - vide evidence for the direct recruitment of dynamin and Rab11 to CLICs and the ability to sort cargo to distinct destinations . Polarization of CLICs during cellular migration Based on the association of novel CLIC cargo , such as Thy - 1 and CD44 with extracellular attachment , we sought to visualize CLICs within migrating cells . As regulation of adhesion compo - nents is a fundamental process during cellular migration the CLIC pathway may be either up - regulated or spatially restricted within motile cells . A scratch - wound was applied to a confluent monolayer of fibroblasts and cells were allowed to migrate into the space provided by the wound . Early endocytosis of CTxB and Tf was assessed in the presence of either internalizing anti - CD44 or anti \u2013 Thy - 1 antibodies or was followed by labeling for endogenous Cav1 ( Fig . 6 A ) . Colocalization was seen between CTxB - 555 and both CD44 and Thy - 1 at the leading edge of mi - grating cells ( arrows ) . In cells expressing GFP - GPI , anti - GFP antibody endocytosis also preferentially occurred at the leading edge ( Fig . 6 A ) , indicating polarization of CLICs rather than po - larized uptake of specific GPI - APs involved in migration , such as Thy - 1 . There was a clear spatial distinction between CTxB - 555 , Cav1 , and Tf - 647 in these cells , which was confirmed by assessing average pixel intensity across a selected field ( Fig . 6 B ) . It has been well documented that Cav1 localizes to the trailing edge during 2D migration ( Grande - Garc\u00eda et al . , 2007 ) . We also found that CTxB binding after fixation was uniform over the en - tire cell , excluding the possibility that GM1 was polarized within these cells ( not depicted ) . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 JCB \u2022 VOLUME 190 \u2022 NUMBER 4 \u2022 2010 686 Figure 6 . CLICs become polarized during and are necessary for efficient cellular migration . ( A ) Confluent monolayers of WT MEFs or NIH3T3s were scratched and cells were allowed to migrate for 8 \u2013 12 h . CTxB - 555 and Tf - 647 were pulsed into migrating cells for 2 min in the presence of anti - CD44 ( WT MEF ) , anti - GFP ( GFP - GPI expressing WT MEF ) , or anti - Thy - 1 ( NIH3T3 ) antibodies or cells were labeled for Cav - 1 ( WT MEF ) . Dotted lines indicate leading edges . Arrows show colocalization between anti - CD44 or anti \u2013 Thy - 1 and CTxB - 555 but not Tf - 647 at the leading edge . Bar , 20 \u00b5m . ( B ) Quantita - tion of average pixel intensity from the leading to trailing edges for CTxB - 555 , Cav - 1 , and Tf - 647 . Inset shows the concentration of tubular , Cav - 1 , and Tf - negative CTxB - labeled CLICs at the leading edge . A rectangular area , outlined , was used to calculate the average pixel intensity ( along the y axis ) across the leading to trailing edge ( along the x axis ) for each endocytic marker . Plot profiles identify a concentration of CTxB in the leading edge and Cav - 1 in the trailing edge whereas Tf shows uniform intensity across the cell . Bar , 10 \u00b5m . ( C ) Electron micrographs of a migrating WT MEF . Large arrow indicates direction of migration . Magnifications from the leading edge and the trailing edge show representative images of CTxB - HRP \u2013 labeled CLICs D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 687 Characterization of the CLIC pathway \u2022 Howes et al . from the PM ( Kirkham et al . , 2005 ; Sabharanjak et al . , 2002 ) , such recruitment of dynamin to maturing CLICs , after the initial scission , provides important machinery for CLICs to compart - mentalize . Clearly , however , dynamin is not required for the generation of CLIC architecture , as the morphology of CLICs is identical in the presence or absence of dynamin inhibition . Recruitment of dynamin to CLICs as they mature would reconcile the direct interaction between GRAF1 and dynamin within this dynamin - independent pathway ( Hansen and Nichols , 2009 ) . Myoferlin , a novel CLIC - associated target identified here , is a member of the ferlin family of proteins implicated in mem - brane fusion ( Davis et al . , 2000 ) . We have previously shown that another ferlin family member , dysferlin , associates with the CLIC pathway ( Hern\u00e1ndez - Deviez et al . , 2008 ) . Myoferlin , like dysferlin , is involved in calcium - dependent membrane fusion after mechanical disruption to the PM ( Bernatchez et al . , 2007 ) . Recently , a clathrin - independent , GPI - enriched endocytic com - partment has been implicated in membrane repair after PM shearing ( Idone et al . , 2008a , b ) . This raises the exciting possibil - ity of links between the role of dysferlin / myoferlin in this pro - cess and the ability of CLICs to provide the necessary membrane for repair of PM lesions . Supporting this , two targets identified here , AHNAK and Annexin A2 , have been shown to bind dysfer - lin and myoferlin during membrane repair ( Huang et al . , 2007 ) . Localization of dysferlin / myoferlin to a subdomain of CLICs that does not merge with Tf - positive EEs highlights that this pathway has the capability to rapidly recycle membrane back to the PM under specific conditions , such as PM repair ( Fig . 7 B ) . Evidence for an association between CLICs and cellular migration is provided by the distinct polarization of CLICs in motile cells , as well as the impairment of migration after tran - sient ablation of the pathway . Specific internalization of Thy - 1 and CD44 through the CLIC pathway at the leading edge pro - vides an explicit role for CLICs in rapid turnover of adhesion receptors within a spatially restricted mechanism . How CLICs become polarized during migration remains an open question ; however , one possibility arises from activity of the initial regula - tor of the CLIC / GEEC pathway , Cdc42 . Cdc42 was originally described as directing cell polarity during yeast budding ( Johnson and Pringle , 1990 ) and plays an assortment of fundamental roles during eukaryotic cell polarization ( Etienne - Manneville , 2004 ) . After wounding of cell monolayers Cdc42 activity is specifically directed to the front of the cell , resultant from signaling between contacts and the ECM ( Etienne - Manneville and Hall , 2001 ) . As Cdc42 activity is necessary for formation of CLICs , this would lead to specific localization of CLICs at the leading edge of wounded monolayers . The clathrin adaptor , Numb , is also Discussion Here we identify the CLIC pathway as a high capacity route that accounts for the vast bulk of constitutive fluid uptake in fibro - blasts , implicating CLICs as key mediators of PM remodeling . Previous estimations of endosomal tubule volume using stereol - ogy have noted that tubules close to section thickness ( 55 nm ) can lead to overprojection effects ( Griffiths et al . , 1989 ) . When correction factors are applied to volume estimates CLICs would still account for over 70 % of internalization . Bulk phase early endocytosis in rat fetal fibroblasts has also been shown to be in - dependent from clathrin and occurs via a constitutive , noncave - olar pathway , which rapidly regurgitates substantial amounts of endocytic volume ( Cupers et al . , 1994 ) . We have previously shown that up to 40 % of endocytosed fluid , mainly internalized through the CLIC / GEEC pathway , is regurgitated in 5 min ( Chadda et al . , 2007 ) . Together , this suggests that the CLIC pathway in - ternalizes and returns significant portions of the PM quickly , rapidly turning over membrane during key cellular processes , such as PM repair and homeostasis . It also provides the endo - cytic system with a mechanism to remodel features of the PM quickly , critical for extracellular interactions during various processes ranging from adhesion to signaling . Such vast mem - brane flow and recycling has recently been observed from a GPI - AP \u2013 positive CIE compartment , which provides a signfi - cant fraction of the required membrane during cellular spread - ing ( Gauthier et al . , 2009 ) . Calculation of fluid uptake via clathrin - independent endo - cytosis has generally estimated the contribution of these routes to be 40 \u2013 60 % of total fluid internalization ( Sandvig et al . , 1987 ; Sabharanjak et al . , 2002 ; Cheng et al . , 2006 ) . The higher contri - bution of 74 % , calculated here , may be explained through differ - ences in the resolution , both spatial and temporal , of experimental procedures used . By calculating volume internalized based on ultrastructure rather than fluorescence this work provides a higher resolution analysis than previous studies , which were limited by the optical resolution of fluorescence microscopy . The 3D structure of CLICs identifies them as pleiomorphic , complex , multicomponent carriers . Although it is recognized that the bulk of CLIC - internalized fluid and membrane fuses with EEs , discrepancies between the ability of CLIC cargos , such as dysferlin / myoferlin and CD44 compared with GFP - GPI and CTxB , to fuse with Tf - positive endosomes strengthens the argument that some domain of CLICs traffics distinctly from the bulk ( Fig . 7 A ) . Recruitment of dynamin to CLICs , after budding from the PM , further demonstrates the sorting complexity of the route . Although dynamin is not necessary for scission of CLICs ( 1 , 2 ) and surface - connected caveolae ( 3 , 4 ) . Bar , 500 nm . ( D ) Quantitation of the number of endocytic structures at the leading and trailing edge of cells treated as in C . Budded caveolae and CCVs are positive for CTxB - HRP label , whereas caveolae and CCPs are not . Error bars show SEM . ( E ) WT MEFs were incubated with or without CTxB - HRP for 2 min . The DAB reaction was performed on live cells for 5 min . Cells were washed and allowed to grow for a further 4 h . CTxB - 555 , anti - CD44 antibodies , and Tf - 647 were added directly to cells for 2 min of uptake before acid stripping and fixation . Bar , 10 \u00b5m . ( F ) 12 \u2013 15 cells treated as in E were quantitated for average fluorescence intensity of CTxB - 555 , Tf - 647 , or goat anti \u2013 mouse - 488 for mouse \u2013 anti - CD44 . ( G ) Confluent monolayers of WT MEFs were scratched and cells were allowed to migrate into the space for 1 h . Cells were incubated with serum alone or serum with 10 \u00b5g ml 2 1 CTxB - HRP for 2 min and the DAB reaction was performed as in E . After 4 h of migration , cells were fixed and distance migrated was determined by measuring the distance of the gap between both sides of the wound at time zero and after 4 h of incubation . Error bars show SEM from three independent experiments . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 JCB \u2022 VOLUME 190 \u2022 NUMBER 4 \u2022 2010 688 and rapidly mature into a high capacity sorting station . We have shown that CLICs are a high capacity , dynamic , and complex sorting system and propose that they have the ability to sort subcompartments to distinct destinations for critical cellular processes such as membrane repair , migration , and maintenance of the PM . Materials and methods Internalization assay Cells on 12 - mm coverslips or 35 - mm dishes were serum starved for 4 h in serum - free DME ( Invitrogen ) before two times 2 - min wash in precooled CO 2 - independent media ( Invitrogen ) on ice and placed on 50 - \u00b5l drops of 10 \u00b5g . ml 2 1 CTxB - 555 ( Invitrogen ) , 5 \u00b5g . ml 2 1 Tf - 488 ( Invitrogen ) , and / or anti - GFP as desired for 20 min . For stereology , 35 - mm dishes were not in - cubated on ice but were subjected to constitutive uptake at 37\u00b0C , 5 % CO 2 for desired times . Coverslips were moved to prewarmed serum - free DME and incubated at 37\u00b0C , 5 % CO 2 for desired times . Coverslips were placed back on ice in precooled CO 2 - independent media . Two times 30 - s 0 . 5 M glycine ( pH 2 . 2 ) acid stripping was routinely performed before fixation in 2 % paraformaldehyde . For immunocytochemistry , cells were permeabi - lized in freshly made 0 . 1 % saponin for 10 min and blocked in 0 . 2 % fish skin gelatin / 0 . 2 % BSA in PBS . Coverslips were incubated with primary antibodies for 1 h at room temperature , followed by three times 10 - min wash in PBS before incubation with secondary Alexa Fluor \u2013 conjugated antibodies at 1 : 400 for 1 h . Fluorescence micrographs were captured on a confocal laser - scanning microscope ( Axiovert 200 m SP LSM 510 META ; Carl Zeiss , Inc . ) . Images were captured under oil with a 63x Plan - Apochromat objective , at excitation wavelengths of 488 , 543 , and 633 nm for GFP - tagged con - structs or Alexa Fluor 488 \u2013 , Alexa Fluor 555 \u2013 , or Alexa Fluor 647 / 660 \u2013 conjugated antibodies , respectively . Band - pass and meta filters were applied as appropriate for respective fluorophores and images were cap - tured from photomultiplier tube detectors ( Carl Zeiss , Inc . ) . Colocalization was quantitated using Volocity v3 . 7 . Images of individual cells were split into RGB channels , automatic thresholding applied , and colocalization co - efficients calculated on percentage of overlapping voxels . Antibodies used include Thy - 1 and CD44 ( Abcam ) ; CHC , GM130 , Grp78 , Cav1 , and HA ( BD ) ; myc ( Sigma - Aldrich ) ; and TfR ( Invitrogen ) . HRP - conjugated anti - CD44 was generated with Lightning - Link HRP conjugation kit ( Innova Bioscience ) as per the manufacturer\u2019s instructions . Dynamin and PI3K inhibition Cells were serum starved for 4 h in serum - free DME before incubation for 20 min with 30 \u00b5M Dyngo4a , 80 \u00b5M dynasore ( Sigma - Aldrich ) , or 100 nM wortmannin ( Roche ) . Cells were then incubated with an endocytic marker of interest in the presence of Dyngo4a and wortmannin for indi - cated periods at 37\u00b0C . Electron microscopy and DAB reaction After constitutive internalization of 10 \u00b5g . ml 2 1 CTxB - HRP ( Invitrogen ) at 37\u00b0C , cells were placed on ice and washed two times for 1 min in ice cold PBS , followed by 10 min in 1 mg . ml 2 1 DAB ( Sigma - Aldrich ) \u00b1 50 \u00b5M ascor - bic acid ( AA ) in PBS . Cells were then incubated with DAB \u00b1 50 \u00b5M AA with 0 . 012 % H 2 O 2 for 20 min , followed by fixation in 2 . 5 % glutaraldehyde ( ProSciTech ) in PBS for 1 h at room temperature . For plastic sections , cells were contrasted with 1 % osmium tetroxide and 4 % uranyl acetate . Cells were dehydrated in successive washes of 70 % , 90 % and 100 % ethanol before embedding in LX - 112 resin . For frozen sections , cells were scraped and embedded in 10 % gelatin and gelatin blocks were infused with 2 . 3 M sucrose . Blocks were frozen onto mounting pins and 55 - nm sections were collected using a UC6 ultramicrotome ( Leica ) . Tomograms of 300 - nm plastic sections were captured in 1\u00b0 increments , tilted \u00b160\u00b0 on a transmission elec - tron microscope ( model F30 ; Tecnai ) using UltraScan 4000 semi - automated software . High pressure frozen samples were essentially processed as de - scribed previously ( Richter et al . , 2008 ) . In brief , CTxB - HRP was added di - rectly to cells grown on sapphire discs and DAB reaction was performed on ice before discs were frozen in a Balzers HPM010 high pressure freezer ( BAL - TEC AG ) and stored under liquid nitrogen . Freeze substitution was performed in acetone with 0 . 7 % uranyl acetate ; 1 % osmium tetroxide at 2 90\u00b0C before warming to 0\u00b0C over 3 d and infiltrated with Embed 812 - Araldite resin . specifically localized to the leading edge where it mediates integrin - \ue062 1 endocytosis ( Nishimura and Kaibuchi , 2007 ) . It appears surprising that Numb - positive CCVs and CLICs both polarize to the leading edge during migration . One possible explanation for this is the types of cargo internalizing through each pathway . Adhesion components without a CCP sorting motif , such as CD44 and Thy - 1 , still need to be internalized for efficient migra - tion and this could occur via CLICs at the leading edge . Con - sistent with such a hypothesis , a screen for genes involved in epithelial cell migration identified that knock - down of GRAF1 , a regulator of the CLIC pathway ( Lundmark et al . , 2008 ) , reduced the capacity of these cells to migrate ( Simpson et al . , 2008 ) . This study has shown the significant magnitude of traf - fic through this largely uncharacterized sorting compartment . Rather than acting as passive carriers that shuttle from the PM to EEs , we propose that CLICs arise from the plasma membrane Figure 7 . Model of CLIC endocytosis . ( A ) ( 1 ) CLIC cargo , such as Thy - 1 , CD44 , and myoferlin are concentrated within Flotillin - 1 and cholesterol - enriched microdomains . ( 2 ) Actin and GRAF - 1 drive the initial formation of the carriers , within 15 s . ( 3 ) Recruitment of dynamin , Rab11 , and Rab5 / EEA - 1 complexes within 2 min provides the ability for these carriers to facilitate bulk membrane flow to early endosomes ( 4a ) and fast plasma membrane recycling ( 4b ) . ( B ) After abrasion to the PM an influx of Ca 2 + activates the fusogenic C2 domains of dysferlin / myoferlin , resulting in the preferential recycling of the CLIC pathway . D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 689 Characterization of the CLIC pathway \u2022 Howes et al . S ( v ) of plasma membrane across 25 cells , from low magnification electron micrographs , multiplied by the total average volume . To capture the average volume of single CTxB - HRP \u2013 positive CLIC , high resolution electron micrographs from whole - mount profiles were used . As CTxB - HRP label ( as well as HRP and anti - CD44 - HRP label ) was only found within the tubular ring of CLICs , we assumed that the entire volume of CLICs is comprised solely within the tubule space . Therefore , the volume of a sphere based on the diameter of the inner membrane was subtracted from the volume of a sphere based on the diameter of the outer membrane of the rings . This calculated that an average CLIC had an estimated volume of 0 . 00263 \u00b1 0 . 00048 \u00b5m 3 . In comparison , the volume of a single CCV was estimated to be five times smaller ( 0 . 0005 \u00b5m 3 assuming sphere based on diameter of 100 nm ) . Mass spectrometry An Agilent 1100 Binary HPLC system ( Agilent Technologies ) was used to per - form reversed phase separation of the in - gel digests using a Vydac MS C18 300A column ( 150 mm \u00d7 2 mm ) with a particle size of 5 \u00b5m ( Vydac ) . Mass spectrometry experiments were performed on a hybrid quadrupole / linear ion trap 4000 QTRAP MS / MS system ( Applied Biosystems ) . All analyses were performed using information - dependent acquisition and the linear ion trap ( LIT ) acquisition modes . Analyst 1 . 4 . 1 software was used for data analysis . Cell migration Confluent WT MEFs grown on coverslips were scratched with a 200 - \u00b5l pipette tip . Cells were washed and allowed to recover for 4 h before inter - nalization assays . Average pixel intensity from leading to trailing edges within a selected area was calculated using the Plot Profile option in ImageJ . Plot profiles show a representative image of 5 \u2013 10 cells across three independent experiments . DAB ablation 10 \u00b5g . ml 2 1 CTxB - HRP was added to cell monolayers for 2 min at 37\u00b0C . Cells were then incubated with 10 mg . ml 2 1 DAB , 50 \u00b5M AA in DME for 2 min before incubation with 10 mg . ml 2 1 DAB , 50 \u00b5M AA , 0 . 012 % H 2 O 2 in DME for a further 2 min . Cells were washed five times with DME before incubation in normal growth media for 4 h . CTxB - 555 , Tf - 647 , and anti - CD44 antibodies were then directly added to cells for 2 min before acid stripping , fixation , permeabilization , and labeling for anti - CD44 antibodies . For quantitation of distance migrated , monolayers were imaged after DAB inactivation ( time zero ) and again after 4 h . Distance between front edges of monolayers on both sides of the wound was measured using Adobe Photoshop CS2 . Anti - CD44 - HRP and Tf - HRP failed to inhibit either clathrin - independent or - dependent endocytosis at concentrations of 10 \u2013 100 \u00b5g . ml 2 1 as used at early times in this assay . Online supplemental material Fig . S1 shows that wortmannin treatment of MEFs inhibits the fusion of CLICs with early endosomes , leading to a threefold increase in the number of CLICs per cell . Wortmannin treatment does not noticeably affect the pro - tein profiles of the CLIC - enriched fraction . Fig . S2 shows the consistent na - ture of membrane profiles in the CLIC - enriched fraction and that CTxB - HRP labels a majority of these profiles when added to cells before the fraction - ation . Fig . S3 shows the localization of dynamin to CLICs after wortmannin treatment by immunofluorescence and by ultrastructure . Dynamin also labels the neck of profiles in the CLIC - enriched fraction . Video 1 shows the tilt series of CTxB - HRP \u2013 laden CLICs after 15 s of internalization . Table S1 lists all the peptides identified from fraction 2 . 8 with two or more 99 % confi - dence hits from the LC - MS / MS . Online supplemental material is available at http : / / www . jcb . org / cgi / content / full / jcb . 201002119 / DC1 . The Institute for Molecular Bioscience is a Special Research Centre of the Aus - tralian Research Council . Confocal microscopy was performed at the Austra - lian Cancer Research Foundation ( ACRF ) / Institute for Molecular Bioscience Dynamic Imaging Facility for Cancer Biology , which was established with the support of the ACRF . The authors acknowledge the facilities as well as scientific and technical assistance from staff in the Australian Microscopy and Micro - analysis Facility ( AMMRF ) at the Centre for Microscopy and Microanalysis at The University of Queensland . We thank Jean Gruenberg for critical appraisal of the manuscript . This work was supported by grants from the National Health and Medi - cal Research Council of Australia ( 511005 ; R . G . Parton and J . F . Hancock ) and the Human Frontier Science Program . Submitted : 22 February 2010 Accepted : 26 July 2010 Nycodenz density fractionation NIH3T3 \u2013 GFP - GPI cells were serum starved and treated for 20 min with 100 nM wortmannin and 30 \u00b5M Dyngo4a . Anti - GFP antibodies were di - rectly added to treated cells in the presence of wortmannin and Dyngo4a for 5 min at 37\u00b0C . Cells were placed on ice and harvested by scraping in TES buffer ( 10 mM tricine , 1 mM EDTA , and 250 mM sucrose , pH 7 . 4 ) with dissolved complete protease inhibitor cocktail ( Roche ) . Cells were ho - mogenized by passaging 10 times each through 21 - and 27 - gauge nee - dles . A 17 , 000 - g spin at 4\u00b0C for 10 min was applied to sediment nonlysed cells and nuclei . Post - nuclear supernatants were mixed with an equal vol - ume 70 % Nycodenz in TES and placed on the bottom of a four - step gradi - ent , consisting of 25 % , 10 % , and 0 % Nycodenz steps . Gradients were spun at 160 , 000 g for 2 h . The 10 % fraction ( excluding the interface ) was collected and diluted 1 / 3 in TES with protease inhibitors and placed on top of a 5 \u2013 20 % continuous Nycodenz gradient . This was spun at 206 , 000 g for 2 h . Fractions were collected from the top of the gradient and TCA - precipitable samples analyzed by SDS - PAGE and LC / MS - MS . For visual - ization of Nycodenz fraction morphology by EM , 10 \u00b5l of fraction of interest was mixed with 10 \u00b5l 5 % glutaraldehyde and a 100 mesh copper EM grid was placed on top for 30 min at room temperature . Grids were washed three times 5 min with PBS , followed by five times 2 min H 2 O before staining with 1 : 9 ( by volume ) of 4 % uranyl acetate to 2 % methyl cellulose for 10 min on ice . CTxB biotinylation CTxB - 555 was biotinylated with 2 mg . ml 2 1 EZ - link Sulfo - NHS - SS - biotin ( Thermo Fisher Scientific ) for 10 min at 37\u00b0C and biotin quenched with 1 : 1 volume of 200 mM glycine for 5 min . CTxB - biotin was added to Cav1 2 / 2 MEFs either treated with 30 \u00b5M Dyngo4a or left untreated for the times indi - cated . After CTxB - biotin internalization , cells were placed on ice and stripped with three times 5 - min washes with 100 mM MesNa ( Sigma - Aldrich ) . MesNa was quenched with three times 5 - min washes with 54 mM iodo - acetamide ( Sigma - Aldrich ) . Cells were harvested in TNE ( 20 mM Tris pH 7 . 5 , 150 mM NaCl , and 5 mM EDTA ) with complete protease inhibitors ( Roche ) and solubilized with 1 % Triton X - 100 . BCA protein assays ( Thermo Fisher Scientific ) were routinely performed to accommodate equivalent loading . Western blots were probed with streptavidin - HRP ( Invitrogen ) in 5 % skim milk , 0 . 1 % Tween 20 . Intensity of developed HRP chemiluminescent bands was determined for multiple exposures per sample in ImageJ using Gel Analyzer . Stereology Serum - starved cells were incubated directly with 10 \u00b5g . ml 2 1 CTxB - HRP , 20 mg . ml 2 1 HRP ( Sigma - Aldrich ) , 10 \u00b5g . ml 2 1 anti - CD44 - HRP antibodies or 20 \u00b5g . ml 2 1 Tf - HRP ( Invitrogen ) for desired times before being placed on ice , washed with precooled CO 2 - independent media , and processed for the DAB reaction and EPON embedding . Vertical sections , visualized at 20 , 000x on a transmission electron microscope ( model 1011 ; JEOL ) , were overlayed with a 2 , 000 - nm square lattice grid using iTEM software . The number of intersections that fell on top of the cytoplasm , excluding the nu - cleus , were recorded across 20 \u2013 25 cells and three independent areas . A 200 - nm square lattice grid was also applied and intersections that fell on top of DAB - labeled structures were recorded as above . Number of points lying over DAB - labeled structures multiplied by the grid size was com - pared , as a percentage , to points over cytoplasm by grid size to give the percentage of labeled structures to cytoplasmic volume ( V ( v ) ) . Overprojec - tion effects on endosomal tubules can lead to volume overestimation as de - scribed previously ( Griffiths et al . , 1989 ) . Based on the section thickness , average diameter , and length of CLICs , a correction factor of 0 . 65 can be applied to the V ( v ) measurements , based on graphs previously developed ( Weibel , 1979 ) . Correction factors are approximate only and give an esti - mate of possible error . Based on tomography data ( Fig . 3 ) , however , CLICs display a depth greater than section thickness ( entire volume is still within 300 - nm section , far thicker than 55 - nm sections used for volume calcula - tion ) , in which case overprojection effects become negligible . Surface density , or surface volume fraction ( S ( v ) ) , was calculated by placing a cycloid grid over high resolution images of CTxB - HRP \u2013 labeled structures and using the formula : S ( v ) = 2I / P\u2022l ( p ) , where I = total intersec - tions of cycloid grid with structure of interest , P = total points of cycloid grid lying over structure of interest , and l ( p ) = length of the cyloid in \u00b5m , which was an absolute distance of 75 nm in this study . Total cellular volume of a Cav1 2 / 2 MEF was calculated by binding CTxB - 555 to fixed cells . Confocal Z - stacks were obtained from five repre - sentative cells and total average cellular volume was calculated using Volocity v3 . 7 . Surface area of Cav1 2 / 2 MEF was calculated by finding the D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 JCB \u2022 VOLUME 190 \u2022 NUMBER 4 \u2022 2010 690 cytoplasm with dysferlin during skeletal muscle regeneration . FASEB J . 21 : 732 \u2013 742 . doi : 10 . 1096 / fj . 06 - 6628com Idone , V . , C . Tam , and N . W . Andrews . 2008a . Two - way traffic on the road to plasma membrane repair . Trends Cell Biol . 18 : 552 \u2013 559 . doi : 10 . 1016 / j . tcb . 2008 . 09 . 001 Idone , V . , C . Tam , J . W . Goss , D . Toomre , M . Pypaert , and N . W . Andrews . 2008b . Repair of injured plasma membrane by rapid Ca2 + - dependent endo - cytosis . J . Cell Biol . 180 : 905 \u2013 914 . doi : 10 . 1083 / jcb . 200708010 Johnson , D . I . , and J . R . Pringle . 1990 . Molecular characterization of CDC42 , a Saccharomyces cerevisiae gene involved in the development of cell polar - ity . J . Cell Biol . 111 : 143 \u2013 152 . doi : 10 . 1083 / jcb . 111 . 1 . 143 Kalia , M . , S . Kumari , R . Chadda , M . M . Hill , R . G . Parton , and S . Mayor . 2006 . Arf6 - independent GPI - anchored protein - enriched early endosomal compartments fuse with sorting endosomes via a Rab5 / phosphatidyl - inositol - 3 \ue039 - kinase - dependent machinery . Mol . Biol . Cell . 17 : 3689 \u2013 3704 . doi : 10 . 1091 / mbc . E05 - 10 - 0980 Kirkham , M . , A . Fujita , R . Chadda , S . J . Nixon , T . V . Kurzchalia , D . K . Sharma , R . E . Pagano , J . F . Hancock , S . Mayor , and R . G . Parton . 2005 . Ultrastructural identification of uncoated caveolin - independent early endocytic vehicles . J . Cell Biol . 168 : 465 \u2013 476 . doi : 10 . 1083 / jcb . 200407078 Kumari , S . , and S . Mayor . 2008 . ARF1 is directly involved in dynamin - independent endocytosis . Nat . Cell Biol . 10 : 30 \u2013 41 . doi : 10 . 1038 / ncb1666 Lamaze , C . , A . Dujeancourt , T . Baba , C . G . Lo , A . Benmerah , and A . Dautry - Varsat . 2001 . Interleukin 2 receptors and detergent - resistant membrane domains define a clathrin - independent endocytic pathway . Mol . Cell . 7 : 661 \u2013 671 . doi : 10 . 1016 / S1097 - 2765 ( 01 ) 00212 - X Le , P . U . , and I . R . Nabi . 2003 . Distinct caveolae - mediated endocytic pathways target the Golgi apparatus and the endoplasmic reticulum . J . Cell Sci . 116 : 1059 \u2013 1071 . doi : 10 . 1242 / jcs . 00327 Lee , J . L . , M . J . Wang , P . R . Sudhir , and J . Y . Chen . 2008 . CD44 engagement promotes matrix - derived survival through the CD44 - SRC - integrin axis in lipid rafts . Mol . Cell . Biol . 28 : 5710 \u2013 5723 . doi : 10 . 1128 / MCB . 00186 - 08 Legg , J . W . , C . A . Lewis , M . Parsons , T . Ng , and C . M . Isacke . 2002 . A novel PKC - regulated mechanism controls CD44 ezrin association and direc - tional cell motility . Nat . Cell Biol . 4 : 399 \u2013 407 . doi : 10 . 1038 / ncb797 Lundmark , R . , G . J . Doherty , M . T . Howes , K . Cortese , Y . Vallis , R . G . Parton , and H . T . McMahon . 2008 . The GTPase - activating protein GRAF1 regu - lates the CLIC / GEEC endocytic pathway . Curr . Biol . 18 : 1802 \u2013 1808 . doi : 10 . 1016 / j . cub . 2008 . 10 . 044 Mayor , S . , and R . E . Pagano . 2007 . Pathways of clathrin - independent endo - cytosis . Nat . Rev . Mol . Cell Biol . 8 : 603 \u2013 612 . doi : 10 . 1038 / nrm2216 Montesano , R . , J . Roth , A . Robert , and L . Orci . 1982 . Non - coated membrane invaginations are involved in binding and internalization of cholera and tetanus toxins . Nature . 296 : 651 \u2013 653 . doi : 10 . 1038 / 296651a0 Mrass , P . , I . Kinjyo , L . G . Ng , S . L . Reiner , E . Pur\u00e9 , and W . Weninger . 2008 . CD44 mediates successful interstitial navigation by killer T cells and enables efficient antitumor immunity . Immunity . 29 : 971 \u2013 985 . doi : 10 . 1016 / j . immuni . 2008 . 10 . 015 Murk , J . L . , G . Posthuma , A . J . Koster , H . J . Geuze , A . J . Verkleij , M . J . Kleijmeer , and B . M . Humbel . 2003 . Influence of aldehyde fixation on the morphology of endosomes and lysosomes : quantitative analysis and electron tomography . J . Microsc . 212 : 81 \u2013 90 . doi : 10 . 1046 / j . 1365 - 2818 . 2003 . 01238 . x Nishimura , T . , and K . Kaibuchi . 2007 . Numb controls integrin endocytosis for directional cell migration with aPKC and PAR - 3 . Dev . Cell . 13 : 15 \u2013 28 . doi : 10 . 1016 / j . devcel . 2007 . 05 . 003 Palamidessi , A . , E . Frittoli , M . Garr\u00e9 , M . Faretta , M . Mione , I . Testa , A . Diaspro , L . Lanzetti , G . Scita , and P . P . Di Fiore . 2008 . Endocytic trafficking of Rac is required for the spatial restriction of signaling in cell migration . Cell . 134 : 135 \u2013 147 . doi : 10 . 1016 / j . cell . 2008 . 05 . 034 Parton , R . G . , and K . Simons . 2007 . The multiple faces of caveolae . Nat . Rev . Mol . Cell Biol . 8 : 185 \u2013 194 . doi : 10 . 1038 / nrm2122 Parton , R . G . , B . Joggerst , and K . Simons . 1994 . Regulated internalization of caveolae . J . Cell Biol . 127 : 1199 \u2013 1215 . doi : 10 . 1083 / jcb . 127 . 5 . 1199 Pelkmans , L . , D . P\u00fcntener , and A . Helenius . 2002 . Local actin polymerization and dynamin recruitment in SV40 - induced internalization of caveolae . Science . 296 : 535 \u2013 539 . doi : 10 . 1126 / science . 1069784 Radhakrishna , H . , and J . G . Donaldson . 1997 . ADP - ribosylation factor 6 regu - lates a novel plasma membrane recycling pathway . J . Cell Biol . 139 : 49 \u2013 61 . doi : 10 . 1083 / jcb . 139 . 1 . 49 Rege , T . A . , M . A . Pallero , C . Gomez , H . E . Grenett , J . E . Murphy - Ullrich , and J . S . Hagood . 2006 . Thy - 1 , via its GPI anchor , modulates Src family kinase and focal adhesion kinase phosphorylation and subcellular localization , and fibroblast migration , in response to thrombospondin - 1 / hep I . Exp . Cell Res . 312 : 3752 \u2013 3767 . doi : 10 . 1016 / j . yexcr . 2006 . 07 . 029 References Bernatchez , P . N . , L . Acevedo , C . Fernandez - Hernando , T . Murata , C . Chalouni , J . Kim , H . Erdjument - Bromage , V . Shah , J . P . Gratton , E . M . McNally , et al . 2007 . Myoferlin regulates vascular endothelial growth factor receptor - 2 stability and function . J . Biol . Chem . 282 : 30745 \u2013 30753 . doi : 10 . 1074 / jbc . M704798200 Bretscher , A . , D . Reczek , and M . Berryman . 1997 . Ezrin : a protein requiring conformational activation to link microfilaments to the plasma membrane in the assembly of cell surface structures . J . Cell Sci . 110 : 3011 \u2013 3018 . Chadda , R . , M . T . Howes , S . J . Plowman , J . F . Hancock , R . G . Parton , and S . Mayor . 2007 . Cholesterol - sensitive Cdc42 activation regulates actin polymerization for endocytosis via the GEEC pathway . Traffic . 8 : 702 \u2013 717 . doi : 10 . 1111 / j . 1600 - 0854 . 2007 . 00565 . x Cheng , Z . J . , R . D . Singh , D . K . Sharma , E . L . Holicky , K . Hanada , D . L . Marks , and R . E . Pagano . 2006 . Distinct mechanisms of clathrin - independent endocytosis have unique sphingolipid requirements . Mol . Biol . Cell . 17 : 3197 \u2013 3210 . doi : 10 . 1091 / mbc . E05 - 12 - 1101 Conner , S . D . , and S . L . Schmid . 2003 . Regulated portals of entry into the cell . Nature . 422 : 37 \u2013 44 . doi : 10 . 1038 / nature01451 Cupers , P . , A . Veithen , A . Kiss , P . Baudhuin , and P . J . Courtoy . 1994 . Clathrin polymerization is not required for bulk - phase endocytosis in rat fetal fibroblasts . J . Cell Biol . 127 : 725 \u2013 735 . doi : 10 . 1083 / jcb . 127 . 3 . 725 Damke , H . , T . Baba , D . E . Warnock , and S . L . Schmid . 1994 . Induction of mutant dynamin specifically blocks endocytic coated vesicle formation . J . Cell Biol . 127 : 915 \u2013 934 . doi : 10 . 1083 / jcb . 127 . 4 . 915 Damke , H . , T . Baba , A . M . van der Bliek , and S . L . Schmid . 1995 . Clathrin - independent pinocytosis is induced in cells overexpressing a temperature - sensitive mutant of dynamin . J . Cell Biol . 131 : 69 \u2013 80 . doi : 10 . 1083 / jcb . 131 . 1 . 69 Davis , D . B . , A . J . Delmonte , C . T . Ly , and E . M . McNally . 2000 . Myoferlin , a candidate gene and potential modifier of muscular dystrophy . Hum . Mol . Genet . 9 : 217 \u2013 226 . doi : 10 . 1093 / hmg / 9 . 2 . 217 Doherty , G . J . , and H . T . McMahon . 2009 . Mechanisms of endocytosis . Annu . Rev . Biochem . 78 : 857 \u2013 902 . doi : 10 . 1146 / annurev . biochem . 78 . 081307 . 110540 Etienne - Manneville , S . 2004 . Cdc42\u2014the centre of polarity . J . Cell Sci . 117 : 1291 \u2013 1300 . doi : 10 . 1242 / jcs . 01115 Etienne - Manneville , S . , and A . Hall . 2001 . Integrin - mediated activation of Cdc42 controls cell polarity in migrating astrocytes through PKCzeta . Cell . 106 : 489 \u2013 498 . doi : 10 . 1016 / S0092 - 8674 ( 01 ) 00471 - 8 Feral , C . C . , N . Nishiya , C . A . Fenczik , H . Stuhlmann , M . Slepak , and M . H . Ginsberg . 2005 . CD98hc ( SLC3A2 ) mediates integrin signaling . Proc . Natl . Acad . Sci . USA . 102 : 355 \u2013 360 . doi : 10 . 1073 / pnas . 0404852102 Furtak , V . , F . Hatcher , and J . Ochieng . 2001 . Galectin - 3 mediates the endocyto - sis of beta - 1 integrins by breast carcinoma cells . Biochem . Biophys . Res . Commun . 289 : 845 \u2013 850 . doi : 10 . 1006 / bbrc . 2001 . 6064 Gauthier , N . C . , P . Monzo , T . Gonzalez , A . Doye , A . Oldani , P . Gounon , V . Ricci , M . Cormont , and P . Boquet . 2007 . Early endosomes associated with dy - namic F - actin structures are required for late trafficking of H . pylori VacA toxin . J . Cell Biol . 177 : 343 \u2013 354 . doi : 10 . 1083 / jcb . 200609061 Gauthier , N . C . , O . M . Rossier , A . Mathur , J . C . Hone , and M . P . Sheetz . 2009 . Plasma membrane area increases with spread area by exocytosis of a GPI - anchored protein compartment . Mol . Biol . Cell . 20 : 3261 \u2013 3272 . doi : 10 . 1091 / mbc . E09 - 01 - 0071 Glebov , O . O . , N . A . Bright , and B . J . Nichols . 2006 . Flotillin - 1 defines a clathrin - independent endocytic pathway in mammalian cells . Nat . Cell Biol . 8 : 46 \u2013 54 . doi : 10 . 1038 / ncb1342 Grande - Garc\u00eda , A . , A . Echarri , J . de Rooij , N . B . Alderson , C . M . Waterman - Storer , J . M . Valdivielso , and M . A . del Pozo . 2007 . Caveolin - 1 regulates cell polarization and directional migration through Src kinase and Rho GTPases . J . Cell Biol . 177 : 683 \u2013 694 . doi : 10 . 1083 / jcb . 200701006 Griffiths , G . , R . Back , and M . Marsh . 1989 . A quantitative analysis of the endo - cytic pathway in baby hamster kidney cells . J . Cell Biol . 109 : 2703 \u2013 2720 . doi : 10 . 1083 / jcb . 109 . 6 . 2703 Hansen , C . G . , and B . J . Nichols . 2009 . Molecular mechanisms of clathrin - independent endocytosis . J . Cell Sci . 122 : 1713 \u2013 1721 . doi : 10 . 1242 / jcs . 033951 Henley , J . R . , E . W . Krueger , B . J . Oswald , and M . A . McNiven . 1998 . Dynamin - mediated internalization of caveolae . J . Cell Biol . 141 : 85 \u2013 99 . doi : 10 . 1083 / jcb . 141 . 1 . 85 Hern\u00e1ndez - Deviez , D . J . , M . T . Howes , S . H . Laval , K . Bushby , J . F . Hancock , and R . G . Parton . 2008 . Caveolin regulates endocytosis of the muscle re - pair protein , dysferlin . J . Biol . Chem . 283 : 6476 \u2013 6488 . doi : 10 . 1074 / jbc . M708776200 Huang , Y . , S . H . Laval , A . van Remoortere , J . Baudier , C . Benaud , L . V . Anderson , V . Straub , A . Deelder , R . R . Frants , J . T . den Dunnen , et al . 2007 . AHNAK , a novel component of the dysferlin protein complex , redistributes to the D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023 691 Characterization of the CLIC pathway \u2022 Howes et al . Ricci , V . , A . Galmiche , A . Doye , V . Necchi , E . Solcia , and P . Boquet . 2000 . High cell sensitivity to Helicobacter pylori VacA toxin depends on a GPI - anchored protein and is not blocked by inhibition of the clathrin - mediated pathway of endocytosis . Mol . Biol . Cell . 11 : 3897 \u2013 3909 . Richter , T . , M . Floetenmeyer , C . Ferguson , J . Galea , J . Goh , M . R . Lindsay , G . P . Morgan , B . J . Marsh , and R . G . Parton . 2008 . High - resolution 3D quanti - tative analysis of caveolar ultrastructure and caveola - cytoskeleton inter - actions . Traffic . 9 : 893 \u2013 909 . doi : 10 . 1111 / j . 1600 - 0854 . 2008 . 00733 . x R\u00f6mer , W . , L . Berland , V . Chambon , K . Gaus , B . Windschiegl , D . Tenza , M . R . Aly , V . Fraisier , J . C . Florent , D . Perrais , et al . 2007 . Shiga toxin in - duces tubular membrane invaginations for its uptake into cells . Nature . 450 : 670 \u2013 675 . doi : 10 . 1038 / nature05996 Sabharanjak , S . , P . Sharma , R . G . Parton , and S . Mayor . 2002 . GPI - anchored proteins are delivered to recycling endosomes via a distinct cdc42 - regulated , clathrin - independent pinocytic pathway . Dev . Cell . 2 : 411 \u2013 423 . doi : 10 . 1016 / S1534 - 5807 ( 02 ) 00145 - 4 Sandvig , K . , and B . van Deurs . 1994 . Endocytosis and intracellular sorting of ricin and Shiga toxin . FEBS Lett . 346 : 99 \u2013 102 . doi : 10 . 1016 / 0014 - 5793 ( 94 ) 00281 - 9 Sandvig , K . , S . Olsnes , O . W . Petersen , and B . van Deurs . 1987 . Acidification of the cytosol inhibits endocytosis from coated pits . J . Cell Biol . 105 : 679 \u2013 689 . doi : 10 . 1083 / jcb . 105 . 2 . 679 Shimonaka , M . , K . Katagiri , T . Nakayama , N . Fujita , T . Tsuruo , O . Yoshie , and T . Kinashi . 2003 . Rap1 translates chemokine signals to integrin activa - tion , cell polarization , and motility across vascular endothelium under flow . J . Cell Biol . 161 : 417 \u2013 427 . doi : 10 . 1083 / jcb . 200301133 Sigismund , S . , E . Argenzio , D . Tosoni , E . Cavallaro , S . Polo , and P . P . Di Fiore . 2008 . Clathrin - mediated internalization is essential for sustained EGFR signaling but dispensable for degradation . Dev . Cell . 15 : 209 \u2013 219 . doi : 10 . 1016 / j . devcel . 2008 . 06 . 012 Simpson , K . J . , L . M . Selfors , J . Bui , A . Reynolds , D . Leake , A . Khvorova , and J . S . Brugge . 2008 . Identification of genes that regulate epithelial cell migration using an siRNA screening approach . Nat . Cell Biol . 10 : 1027 \u2013 1038 . doi : 10 . 1038 / ncb1762 Suga , K . , K . Katagiri , T . Kinashi , M . Harazaki , T . Iizuka , M . Hattori , and N . Minato . 2001 . CD98 induces LFA - 1 - mediated cell adhesion in lymphoid cells via activation of Rap1 . FEBS Lett . 489 : 249 \u2013 253 . doi : 10 . 1016 / S0014 - 5793 ( 00 ) 02222 - 5 Torgersen , M . L . , G . Skretting , B . van Deurs , and K . Sandvig . 2001 . Internalization of cholera toxin by different endocytic mechanisms . J . Cell Sci . 114 : 3737 \u2013 3747 . Vassilieva , E . V . , K . Gerner - Smidt , A . I . Ivanov , and A . Nusrat . 2008 . Lipid rafts mediate internalization of beta1 - integrin in migrating intestinal epithe - lial cells . Am . J . Physiol . Gastrointest . Liver Physiol . 295 : G965 \u2013 G976 . doi : 10 . 1152 / ajpgi . 00082 . 2008 Weibel , E . R . 1979 . Stereological Methods . 1 . Practical methods for biological morphometry . Academic Press Inc . , New York . Yarar , D . , C . M . Waterman - Storer , and S . L . Schmid . 2007 . SNX9 couples actin assembly to phosphoinositide signals and is required for mem - brane remodeling during endocytosis . Dev . Cell . 13 : 43 \u2013 56 . doi : 10 . 1016 / j . devcel . 2007 . 04 . 014 D o w n l oaded f r o m h tt p : / / r up r e ss . o r g / j c b / a r t i c l e - pd f / 190 / 4 / 675 / 1348109 / j c b _ 201002119 . pd f b y U n i v e r s i t y O f W a s h i ng t on u s e r on 30 J une 2023",
    "ayyar2023clic": "Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z CLIC and membrane wound repair pathways enable pandemic norovirus entry and infection B . Vijayalakshmi Ayyar 1 , Khalil Ettayebi 1 , Wilhelm Salmen 2 , Umesh C . Karandikar 1 , Frederick H . Neill 1 , Victoria R . Tenge 1 , Sue E . Crawford 1 , Erhard Bieberich 3 , B . V . Venkataram Prasad 2 , Robert L . Atmar 1 , 4 & Mary K . Estes 1 , 4 Globally , most cases of gastroenteritis are caused by pandemic GII . 4 human norovirus ( HuNoV ) strains with no approved therapies or vaccines available . The cellular pathways that these strains exploit for cell entry and internaliza - tion are unknown . Here , using nontransformed human jejunal enteroids ( HIEs ) that recapitulate the physiology of the gastrointestinal tract , we show that infectious GII . 4 virions and virus - like particles are endocytosed using a unique combination of endosomal acidi \ufb01 cation - dependent clathrin - independent carriers ( CLIC ) , acid sphingomyelinase ( ASM ) - mediated lysosomal exocytosis , and membrane wound repair pathways . We found that besides the known interaction of the viral capsid Protruding ( P ) domain with host glycans , the Shell ( S ) domain interacts with both galectin - 3 ( gal - 3 ) and apoptosis - linked gene 2 - interacting protein X ( ALIX ) , to orchestrate GII . 4 cell entry . Recognition of the viral and cellular determinants regulating HuNoV entry provides insight into the infection process of a non - enveloped virus highlighting unique pathways and targets for developing effective therapeutics . Human noroviruses ( HuNoVs ) , non - enveloped , single - stranded , positive - sense RNA viruses , are classi \ufb01 ed into at least ten genogroups ( GI - GX ) and 48 genotypes and belong to the family Caliciviridae . HuNoVs are the leading cause of acute gastroenteritis worldwide 1 , posing a signi \ufb01 cant risk to global health . Speci \ufb01 c treatments and vac - cines are lacking , partly due to the extended time before a reliable and tractable culture system became available to cultivate HuNoV , which limited understanding of norovirus pathophysiology , drivers of host - virus interactions , and systemsto testapproaches to controlinfection . A breakthrough came when nontransformed human intestinal enteroids ( HIEs ) were proven to be a reproducible and biologically relevant system that supports the replication of multiple HuNoV strains 2 , 3 . HIEs have been successfully applied to ( 1 ) identify HuNoV strain - speci \ufb01 c growth requirements 2 , 4 \u2013 6 , ( 2 ) identify neutralizing monoclonal antibodies and improve understanding of host responses to HuNoV for vaccine development 7 \u2013 10 , and ( 3 ) monitor the environ - mental HuNoV load and the ef \ufb01 cacy of disinfection strategies 11 , 12 . HIEs recapitulate the cellular complexity , diversity , and physiology of the human gastrointestinal tract along with host restriction and genetic factors and mimic strain - speci \ufb01 c epidemiological host - virus infection patterns , making them an ideal system to dissect HuNoV infection and pathophysiology 13 . Virusentryintothecellisthe \ufb01 rststepininfection . Itisinitiatedby interactions between speci \ufb01 c motifs on both viral and host surface proteins and activates signaling cascades that destabilize the mem - branebarrier . Although virus - cellsurfaceinteractionsare complex and highly variable , the number of pathways utilized by viruses for cell entry is relatively limited based on certain key surface components . Received : 7 March 2022 Accepted : 30 January 2023 Check for updates 1 Department of Molecular Virology and Microbiology , Baylor College of Medicine , Houston , TX , USA . 2 Verna and Marrs McLean Department of Biochemistry and Molecular Biology , Baylor College of Medicine , Houston , TX , USA . 3 Department of Physiology , University of Kentucky , Lexington , KY 40506 and VAMC , Lexington , KY 40502 , USA . 4 Department of Medicine , Baylor College of Medicine , Houston , TX , USA . e - mail : mestes @ bcm . edu Nature Communications | ( 2023 ) 14 : 1148 1 1 2 3 4 5 6 7 8 9 0 ( ) : , ; 1 2 3 4 5 6 7 8 9 0 ( ) : , ; Herein , we sought to uncover entry determinants of a globally domi - nant GII . 4 HuNoV straininto physiologicallyrelevant HIEs that support virus replication . Our studies show that GII . 4 HuNoV binding to epi - thelial cells induces plasma membrane wounding that triggers cellular components to the injury site for membrane repair . Surprisingly , we foundtheviruscapsidinteractswiththeendosomalsortingcomplexes requiredforthetransport ( ESCRT ) proteinALIXandinduceslysosomal exocytosis along with other membrane repair mechanisms . GII . 4 HuNoV enters HIEs using a speci \ufb01 c mechanism driven by endosomal acidi \ufb01 cation and requires effectors of the CLIC pathway [ cholesterol , Cdc42 , ADP - ribosylation factor 1 ( Arf1 ) , and galectin - 3 ( gal - 3 ) ] . These studies demonstrate an active CLIC pathway in HIEs , with a crosstalk between CLIC - mediated endocytosis and host repair mechanisms highlighting these pathways as targets for interfering with pathogen entry in human intestinal cells . Results GII . 4 entry in jejunal HIEs requires endosomal acidi \ufb01 cation to initiate endocytosis To probe HuNoV entry into a permissive , secretor positive jejunal ( J2 ) HIE line , we used virus - like particles ( VLPs ) from pandemic GII . 4 Syd - ney / 2012 ( GII . 4 Syd VLPs ) and GII . 4 Sydney / 2012 virus . We \ufb01 rst validatedVLPsasasurrogateforvirusinthesestudiesbyshowingGII . 4 Syd VLPs inhibited virus replication of a GII . 4 stool isolate in a dose - dependent manner ( Fig . 1a ) . Several HuNoV strains are dependent on intestinal components such as bile acid ( BA ) and ceramide for cell entry and replication 6 . GII . 4 infection is enhanced by , but does not require , bile or BA for replication 2 , while initiation of endosomal acidi \ufb01 cation with BAglycochenodeoxycholic acid ( GCDCA ) is required for GII . 3 infection 6 . To understand if endosomal acidi \ufb01 cation is a requirement for GII . 4 entry , HIEs were inoculated with GII . 4 virus , and acidic compartments were labeled with LysoTracker . In GII . 4 - inocu - latedcells , asigni \ufb01 cantincreaseinLysoTracker - positiveorganelleswas seen , which was signi \ufb01 cantly reduced in the presence of neutralizing GII . 4 - speci \ufb01 c polyclonal antibody ( pAb ) , indicating that acidic orga - nelles are induced in response to virus - cell interactions ( Fig . 1b ) . Endosomal acidi \ufb01 cation studies showed that GII . 4 Syd VLPs , like virus , induced endosomal acidi \ufb01 cation that was not observed with GII . 3 VLPs , indicating strain - speci \ufb01 c differences in the ability to trigger endocytosis ( Fig . 1c ) . Moreover , GII . 4 infection in the presence of ba \ufb01 lomycin A1 , a vacuolar ATPase ( V - ATPase ) inhibitor , reduced the number of GII . 4 - induced acidic compartments ( Fig . S1a ) , completely inhibiting GII . 4 replication in a dose - dependent manner ( Fig . 1d ) . Time course experiments showed a partial inhibition of GII . 4 replication Untreated BafilomycinA1 ( 1 . 5nM ) BafilomycinA1 ( 3 . 1nM ) BafilomycinA1 ( 6 . 2nM ) BafilomycinA1 ( 12 . 5nM ) 0 2 4 6 8 G e no m e e qu i va l e n t s / w e ll 1h24h P = < 0 . 0001 * * * * d * * * * P = < 0 . 0001 P = < 0 . 0001 P = < 0 . 0001 * * * * * * * * Untreated GII . 4 VLP ( 10 4 ) GII . 4 VLP ( 10 5 ) GII . 4 VLP ( 10 6 ) 0 2 4 6 8 G e no m e e qu i va l e n t s / w e ll 1 h 24 h P = < 0 . 0001 P = 0 . 0005 P = 0 . 0085 * * * * * * * * * H u m a n p A b G II . 4 s t oo l G II . 4 s t oo l + hu m a n p A b 0 20 40 60 80 M ea n f l uo r esce n se i n t e n s i t y a b c e P = < 0 . 0001 * * * * F M 1 - 43 F X Labe l ed S y d V L P s f 10 m i n Media GII . 4 VLP GII . 3 VLP 50 \u00b5M FM1 - 43FX 10 m i n Media GII . 4 VLP GII . 3 VLP 0 50 100 150 200 M ea n f l uo r esce n se i n t e n s i t y P = < 0 . 0001 * * * * M e d i a G II . 3 V L P G C D C A G II . 4s t oo l G II . 4 V L P 0 1000 2000 3000 4000 M ea n f l uo r esce n se i n t e n s i t y P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * * * * * P = < 0 . 0001 20 \u00b5m 50 \u00b5m Human pAb GII . 4 stool GII . 4 stool + pAb Fig . 1 | GII . 4 capsid protein elicits acidi \ufb01 cation and endocytosis in HIEs . a Viral replication at 24h ( black dots ) compared to bound virus at 1h ( blue dots ) and inhibition of replication in the presence of VLPs compared to untreated at 24h . Replication was quanti \ufb01 ed using n = 2 independent HIE replicates for the 1h and n = 3 independent HIE replicates for 24h with 2 technical replicates / sample . b LysoTracker staining of acidic compartments in the presence of an anti - GII . 4 polyclonalantibody ( pAb ) , GII . 4virus , andpAbmixedwithGII . 4virusat37 o C . Right panel : Mean \ufb02 uorescenceintensitywasquanti \ufb01 edfromdifferentregionsofinterest ( ROIs ) forpAb ( bluebar , ROIs = 32 ) , GII . 4stool ( redbar , ROIs = 31 ) , GII . 4stool + pAb ( green bar , ROIs = 32 ) . c LysoTracker staining of acidic compartments induced by GII . 3 VLP ( green , ROIs = 100 ) , GCDCA ( purple , ROIs = 100 ) , GII . 4 virus ( cyan , ROIs = 101 ) , and GII . 4 VLP ( red , ROIs = 97 ) compared to media ( black , ROIs = 140 ) . d GII . 4 replication in the presence / absence of V - ATPase inhibitor ba \ufb01 lomycin A1 at 1h ( boundvirus , graydots ) andat24h ( blackdots ) . ViralGEswerequanti \ufb01 edusing n = 2independentHIEreplicatesforthe1hand n = 3independentHIEreplicatesfor 24h with 2 technical replicates / sample . e FM1 - 43FX ( green ) uptake showing GII . 4 VLP - induced endocytosis . VLP - induced endocytosis compared to media ( n = 4 HIE replicates ) . Right panel : Mean \ufb02 uorescence intensity quanti \ufb01 ed from ROI = 10 . f Time lapse microscopy showing GII . 4 VLP ( green ) endocytosis and FM1 - 43x uptake ( red ) . All the experiments were repeated independently three times with similar results . In a \u2013 e , error bars represent mean\u00b1SD with signi \ufb01 cance ( P values ) calculated using one - way ANOVA , Dunnett \u2019 s multiple comparisons test . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 2 when ba \ufb01 lomycin A1 was present only in the early stages of infection , while complete inhibition was observed when the inhibitor was pre - sent throughout the entire infection period [ 24h post infection ( h ) ] ( Fig . S1b , upper right panel ) . This was due to the presence of membrane - bound VLPs , which after ba \ufb01 lomycin A1 removal , activated endocytosis reversing inhibition of V - ATPase ( Fig . S1b , lower right panel ) . GII . 4 - induced endocytosis was con \ufb01 rmed using FM1 - 43FX epi - \ufb02 uorescence and time lapse microscopy ( Fig . 1e , f ) . FM1 - 43FX labels extracellular membranes and \ufb02 uorescent puncta are observed fol - lowing endocytosis . An upregulation of FM1 - 43FX - labeled compart - ments was observed in GII . 4 VLP - treated cells further con \ufb01 rming that unlikeGII . 3 ( whichrequiresBA ) , theGII . 4Sydcapsidaloneiscapableof triggering endocytosis upon interaction with the host cells ( Fig . 1e ) . Endocytosis experiments in the presence of ba \ufb01 lomycin A1 indicated that GII . 4 endocytosis is V - ATPase dependent ( Fig . S1c ) and required the intact VLP as the glycan - binding P domain alone did not trigger endocytosis ( Fig . S1d ) . FM1 - 43FX labeling experiments in the presence ofadditionalVLPsshowedthatfourGII . 4variants [ HoustonVirus / 2002 ( HOV ) , New Orleans / 2009 ( NO ) , Grimsby / 1995 ( GRV ) , Farmington Hills / 2002 ( FH ) ] induced endocytosis while HuNoV strains thatrequire BA ( GII . 3 , GI . 1 and GII . 17 ) did not induce endocytosis , further sup - porting the use of different cell entry mechanisms between BA - dependent and independent HuNoV strains ( Fig . S1d ) . GII . 4 uses a dynamin - independent pathway that requires mem - brane cholesterol and is in \ufb02 uenced by actin GII . 4 entry was explored further by using speci \ufb01 c inhibitors of membrane components known to be involved in endocytic path - ways ( Fig . 2a ) . Dynamin , important for both clathrin - and caveolin - mediated endocytosis , was \ufb01 rst tested considering the entry of murine norovirus ( MNV ) in RAW264 . 7 macrophages is dynamin - dependent 14 . Inhibition of dynamin with speci \ufb01 c inhibitors , dynasore and mitmab , did not alter GII . 4 infection , showing that clathrin - and caveolin - mediated endocytosis pathways are not involved in GII . 4 cell entry ( Fig . 2b ) . The ef \ufb01 cacy of dynasore and mitmab in HIEs was con \ufb01 rmed using a \ufb02 uorescently labeled , low - density lipoprotein ( LDL ) - uptake assay , where both inhibitors reduced LDL uptake ( Fig . 2c ) at the tested concentrations ( 40 and 100 \u00b5 M ) . We next targeted cholesterol , another abundant plasma mem - brane component implicated in the entry and replication of many viruses , including GII . 3 HuNoV 6 , 14 \u2013 18 . Cholesterol depletion in HIEs , by inhibitors methyl - \u03b2 - cyclodextrin ( M \u03b2 CD ) and \ufb01 lipin , reduced GII . 4 infection signi \ufb01 cantly ( Fig . 2d , Table S1 ) . In addition , a lower level of inhibition in GII . 4 replication with U18666A , an intracellular choles - terol transport inhibitor , was observed ( Fig . S2a ) compared to M \u03b2 CD and \ufb01 lipin . The activity of U18666A in HIEs was con \ufb01 rmed using \ufb01 lipin and CD63 staining , indicating defective intracellular cholesterol f Untreated M\u00dfCD ( 1mM ) Filipin ( 4\u03bcg / ml ) 10 3 10 4 10 5 10 6 G e no m e e qu i va l e n t s / w e ll 1 h 24h d * * * * P = < 0 . 0001 * * * * P = < 0 . 0001 Untreated CytoD ( 2 . 5\u03bcM ) CytoD ( 5\u03bcM ) 10 3 10 4 10 5 10 6 G e no m i ce qu i va l e n t s / w e ll 1h24h 20 \u00b5m Untreated Mitmab ( 100 \u03bcM ) Dynasore ( 40 \u03bcM ) Dil - LDL a b h c e g T ubu li n Untreated Dynasore ( 40 \u03bcM ) Mitmab ( 100 \u03bcM ) 10 3 10 4 10 5 10 6 10 7 G e no m e e qu i va l e n t s / w e ll 1h 24h Untreated Noco ( 5\u03bcM ) Noco ( 20\u03bcM ) 10 2 10 3 10 4 10 5 10 6 10 7 G e no m e e qu i va l e n t s / w e ll 1 h 24h Untreated Genistein ( 40\u03bcM ) Genistein ( 80\u03bcM ) 10 3 10 4 10 5 10 6 G e no m e e qu i va l e n t s / w e ll 1h24h * * * * P = < 0 . 0001 * * * * P = < 0 . 0001 U n t r ea t e d D y n as o r e ( 40 \u03bc M ) M i t m a b ( 100 \u03bc M ) 0 50 100 150 M ea nF l uo r esce n se i n t e n s i t y P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * Nocodazole ( 20 \u00b5M ) Media Genistein ( 80 \u00b5M ) 10 \u00b5m Fig . 2 | GII . 4 entry is dynamin - independent but depends on cholesterol and actinforinfection . a Schematicofmajorendocytosispathways , keyregulatorsand their inhibitors . b GII . 4 replication ( at 37 o C ) was assessed in the presence of dynamininhibitors , dynasoreandmitmab . ViralRNAreplicationwasquanti \ufb01 edat1 ( black ) and 24h ( green ) by RT - qPCR . c Validation of dynamin inhibitor activity by Dil - complexed low density lipoprotein ( Dil - LDL ) uptake ( red ) . Right panel : Quan - titation of Dil - LDL uptake in untreated ( 80 ROI ) , dynasore - treated ( 38 ROI ) and mitmab - treated ( 55 ROI ) HIEs . d GII . 4 replication in the presence of cholesterol sequestrants , M\u00dfCD and \ufb01 lipin , at 1h ( black ) and 24h ( red ) at 37 o C . e GII . 4 repli - cationinthe presenceof actindepolymerizingagent cytochalasinD ( Cyto D ) at1h ( black ) and 24h ( green ) . f GII . 4 replication in the presence of receptor tyrosine kinase ( RTK ) inhibitor genistein at 1h ( black ) and 24h ( blue ) at 37 o C . g GII . 4 replication in the presence of nocodazole , a microtubule disrupting agent , at 1h ( black ) and24h ( pink ) at37 o C . h Tubulinstaining ( red ) inthepresenceofgenistein andnocodazole . Alltheexperimentswererepeatedindependentlythreetimeswith similar results . In b , d \u2013 g viral GEs were quanti \ufb01 ed using n = 2 independent HIE replicates for the 1h and n = 3 independent HIE replicates for 24h with two tech - nical replicates / sample . In b , viralGEs werequanti \ufb01 ed using n = 3 HIE replicates . In b \u2013 g , errorbars represent mean\u00b1SDwith signi \ufb01 cancerelative to untreated control calculated using one - way ANOVA , Dunnett \u2019 s multiple comparisons test . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 3 transport ( Fig . S2b ) . Altogether , these results show that cholesterol is critical for GII . 4 replication . We next evaluated GII . 4 entry dependence on cytoskeletal com - ponentsusinginhibitorsagainstactin , receptortyrosinekinases ( RTK ) , and microtubules . Cytochalasin D , an actin - depolymerizing agent , reduced GII . 4 replication signi \ufb01 cantly ( 53 % ) ( Fig . 2e ) , whereas genis - tein ( RTK and caveolin inhibitor ) and nocodazole ( microtubule inhi - bitor ) had no effect on GII . 4 replication ( Fig . 2f , g ) . Tubulin staining con \ufb01 rmed the activities of genistein and nocodazole in HIEs ( Fig . 2h ) , suggesting that actin , but not microtubule or RTKs , is essential for GII . 4 entry . GII . 4 depends on the CLIC pathway for cell entry Inhibition of GII . 4 entry by cytochalasin D showed that viral entry was in \ufb02 uenced by actin manipulation , but this drug was cytotoxic above the concentrations tested . This led us to examine other regulators of actinmechanicsthatin \ufb02 uenceclathrin - independentendocytosis ( CIE ) . Aftercon \ufb01 rmingthat theconcentrationofinhibitorsusedinassaysdid not induce cytotoxicity ( Table S1 ) , we tested macropinocytosis and CLICendocytosis ( Fig . 3a ) , twomajorCIEpathways , focusingonCdc42 [ EIPA , ML141 , wiskostatin ( Wisko ) ] , Rac1 [ NSC23766 ( NSC ) ] , and RhoA ( CT04 ) . Inhibition of Cdc42 ( EIPA , ML141 , and Wisko ) dramatically reduced GII . 4 replication ( 86 \u2013 90 % ) ( Figs . 3b and S3a ) , whereas inhi - biting Rac1 or RhoA had no effect on GII . 4 replication . Other inhibitors of classical macropinocytosis such as smooth muscle myosin II inhi - bitor [ blebbistatin ( Bleb ) ] and PI3K inhibitor [ LY29402 ( LY ) ] , known to ef \ufb01 ciently block the entry of vaccinia virus 19 , human adenovirus type 3 20 , and respiratory syncytial virus 21 , were also tested along with inhi - bitors of epidermal growth factor receptor ( EGFR ) ( CAS879127 - 07 - 8 ) andproteinkinaseC ( PKC ) ( calphostin ) . NoneoftheseinhibitorsofCIE had any effect on GII . 4 replication ( Figs . 3b and S3b ) , emphasizing that GII . 4 entry is via a Cdc42 - speci \ufb01 c mechanism distinct from a classical macropinocytosis pathway . The ef \ufb01 cacy of inhibitors in HIEs was con \ufb01 rmed by actin staining , which detected morphological changes in treated cells ( Fig . S3c ) . Next , we evaluated GII . 4 infection in the pre - sence of an Arf1 inhibitor , Golgicide A ( GCA ) , and found that GCA signi \ufb01 cantly inhibited GII . 4 replication ( 65 \u2013 75 % ) at 24 h ( Fig . 3c ) . Taken together , the inhibitor data suggest that GII . 4 entry relies on cholesterol , Cdc42 and Arf1 , indicating a CLIC - mediated virus entry pathway . However , inhibitors can have many off - target effects , so we used a multipronged approach by evaluating the morphology of CLIC carriers , assessing GII . 4 association with gal - 3 ( a regulator of the CLIC pathway ) 22 and testing GII . 4 colocalization with known markers of the CLIC pathway . Fluorescence microscopy showed that GII . 4 induces tubularstructureformationreminiscentofCLICcarriersinVLP - treated cellswith variable lengths ( 200nm \u2013 2 \u00b5 m ) ( Fig . 3d ) , which were further con \ufb01 rmedbyelectron microscopy ( EM ) ( Fig . 3e ) . EMalsoshowedVLPs in vacuolar structures similar to those observed in mouse jejunal * * * * * P = < 0 . 0001 c e d a h b i VLP - treated ( 1 h ) Mock - treated ( 1 h ) 500 nm 500 nm 2 \u00b5m 2 \u00b5m G e no m e e qu i va l e n t s / w e ll * P = 0 . 0133 * * * * P = < 0 . 0001 * * * * P = < 0 . 0001 P = 0 . 0322 * * * * P = < 0 . 0001 g G e no m e e qu i va l e n t s / w e ll P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * 1 h 10 min Media G a l - 3 VP 1 M e r ge 10 \u00b5m VP1 Actin 5 \u00b5m f Fig . 3 | GII . 4 infection in HIEs is sensitive to effectors of CLIC pathway . a Schematic of macropinocytosis and the Clathrin Independent Carrier ( CLIC ) pathways , their effectors and inhibitors . b GII . 4 replication inthe presenceof EIPA ( Na + / H + exchanger inhibitor ) , ML141 ( Cdc42 inhibitor ) , Wiskostatin ( N - WASP inhi - bitor ) , NSC23766 ( RAC1 inhibitor ) , CT04 ( RhoI inhibitor ) , Blebbistatin ( myosin inhibitor ) , and LY29402 ( PI3K inhibitor ) at 1h ( gray ) and 24h ( orange ) . c GII . 4 replicationinthepresenceofArf1inhibitorGolgicideA ( GCAat1h ( black ) and24h ( blue ) . d GII . 4 - inducedtubularcarriersat1h ( 37 o C ) detectedusingguineapiganti - Sydney VLP polyclonal antibody ( Gp Syd - pAb ) for viral capsid ( green ) and phal - loidin for actin ( red ) . Images were taken using a ZEISS Laser Scanning Micro - scope LSM 980 with Airyscan 2 . e Electron microscopy to identify CLIC structures inGII . 4VLP - treatedHIEs . ( 1 \u2013 7structures / cell ) comparedto , mock - treatedcells ( no structures in n = 25cells ) . f Confocalmicroscopytodetect GII . 4VP1capsid ( green ) colocalizationwithgal - 3 ( red ) onthecellsurfaceat10minand1h ( 37 o C ) afterVLP treatment using anti - gal - 3 and Gp Syd - pAb ( n = 3 HIE replicates ) . g Effect of blocking GII . 4 virus - galectin - 3 ( gal - 3 ) surface interaction using anti - gal - 3 antibody on GII . 4 replication at 1h ( black ) and 24h ( red ) . h Dot blot analysis investigating GII . 4 VLP interaction withpuri \ufb01 ed gal - 3 . i Probing CLIC carriersutilized in HIEs for endocytosis with Alexa Fluor \u2122 594 conjugated cholera toxin B ( CTxB ) ( red ) and GII . 4 VLPs ( green ) with similar cargoes marked by white arrows ( n = 2 HIE repli - cates ) . Inset : Co - occurrence of CTxB and GII . 4 VLPs in similar cargos , scale = 5 \u00b5 m . All the experiments were repeated independently three times with similar results . In b , c , and g viral GEs were quanti \ufb01 ed using n = 2 independent HIE replicates for the 1h and n = 3 independent HIE replicates for 24h with two technical replicates / sample . Theerrorbarsrepresentmean\u00b1SDwith P valuescalculatedusingone - way ANOVA , Dunnett \u2019 s multiple comparisons test with comparisons at 24h relative to untreated control . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 4 enterocytes ( Fig . S3d ) 23 . Next , GII . 4association with gal - 3wasassessed using \ufb02 uorescence microscopy , gal - 3 inhibition assays , and direct binding respectively ( Figs . 3f , S3e , S3f , 3g , 3h , and 3i ) . We found that GII . 4 colocalizes with gal - 3 ( Fig . 3f ) in a time - dependent manner . In the presence of TD - 139 , a speci \ufb01 c gal - 3 inhibitor , both GII . 4 colocalization and endocytosis is rapidly ( 10 min ) inhibited suggesting the impor - tance of gal - 3 in GII . 4 entry into cells ( Figs . S3e and S3f ) . Gal - 3 is a multifunctional protein involved in many cellular processes and is foundinboththebiotinylatedsurfaceprotein ( BSP ) plasmamembrane and cytosolic fractions of cells ( Fig . S3g ) . As inhibitors are unable to differentiatebetween the differentpopulationsof gal - 3 , we speci \ufb01 cally used anti - gal - 3 antibody to target gal - 3 interactions on the plasma membrane . Our results showed that anti - gal - 3 antibody signi \ufb01 cantly inhibited GII . 4 replication ( 98 % reduction ) , further asserting the importance of gal - 3 - GII . 4 interaction in GII . 4 entry and infection ( Fig . 3g ) . Direct binding assays showed that gal - 3 bound to VLPs ( Fig . 3h ) , and to the GII . 4 capsid S domain , but not to the P domain , of VP1 of VLPs , with micromolar af \ufb01 nity ( Figs . S3h and S3i ) . Finally , we investigated GII . 4 colocalization using speci \ufb01 c markers of the CLIC pathway such as cholera toxin B ( CTxB ) , CD44 and glycosylphosphatidylinositol - anchored protein ( GPI - AP ) . All these proteins are well studied and characterized for their usage in CLIC - mediated endocytosis 24 \u2013 26 . Fluorescence microscopy showed that GII . 4 shared cargoes with CTxB ( Fig . 3i ) despite not having any effect on replication when recombinant CTxB was used as a competitor ( Fig . S3j ) . ThissuggeststhatalthoughGII . 4usesasimilarentrypathway as CTxB , the receptor / s used for each is distinct . GII . 4 colocalization with CD44 and GPI - AP further veri \ufb01 ed that GII . 4 uses CLIC - mediated endocytosis for entry ( Figs . S3k and S3l ) . ESCRT protein ALIX is critical for GII . 4 entry ESCRTs are known regulators of endocytosis . The unexpected pre - sence of conserved ESCRT binding motifs for ALIX and TSG101 in the GII . 4 capsid sequence ( Fig . S4a ) led us to further investigate if these proteinsareinvolvedinGII . 4entry . Bindingassays ( ELISAanddot - blot ) showedthatGII . 4SydVLPsinteractwithrecombinantALIXandTSG101 ( Figs . 4a , b , S4b ) . Confocal microscopy of virus - infected HIEs showed colocalization between replicating virus , and ALIX and TSG101 at 24 h , suggesting their role in GII . 4 replication ( Fig . S4c ) . Putative ESCRT bindingsequencesarepresentinboththeGII . 4capsidSandPdomains ( Fig . S4a ) ; however , ELISAdatashowedthatwhileTSG101bindstoboth the S and P domains ( Fig . 4b ) , ALIX binding is limited to the S domain only ( Fig . 4a ) . Bio - Layer Interferometry ( BLI ) analysis con \ufb01 rmed the ELISA results , demonstrating that binding af \ufb01 nity of ALIX to the S domain ( K D = 9 . 92 \u00b5 M ) and TSG101 to the S domain ( K D = 21 . 5 \u00b5 M ) and the P domain ( K D = 15 . 5 \u00b5 M ) were in the micromolar range ( Fig . 4c ) . We further probed the contribution of ALIX and TSG101 in GII . 4 entry using protein - speci \ufb01 c antibodies as blocking agents . While anti - ALIX pAb blocked GII . 4 replication completely in a dose - dependent manner , anti - TSG101 pAb had no effect on GII . 4 replication ( Figs . 4d and S4d ) . These data suggest that ALIX is accessible to GII . 4 VLPs on the membrane surface , which was further con \ufb01 rmed by Western blot detection of ALIX in the biotinylated surface protein ( BSP ) fraction of HIEswithananti - ALIXpAb ( Fig . S4e ) . Confocalmicroscopyestablished A b s o r b a n ce ( 450 n m ) A b s o r b a n ce ( 450 n m ) Untreated Anti - ALIX ab ( 80 \u03bcg ) Anti - TSG ab ( 80 \u03bcg ) 10 3 10 4 10 5 10 6 10 7 1 h24 h Media GII . 4 VLP \u0394VLP 50 \u00b5M i FM1 - 43x A b s o r b a n ce ( 450 n m ) c b a d e g h f bALIXvs GII . 4 S - Domain bTSG101 vs GII . 4 S - Domain bTSG101 vs GII . 4 P - Domain Dimer 10 m i n % Fo l d c h a ng e i n v i r a l RNA ( 24 h - 1 h ) no r m a li z e d t o G A P DH * * * P = 0 . 0003 P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * 25 \u00b5m Fig . 4 | GII . 4 interacts with ALIX and TSG101 , withALIX being criticalfor virus entry . a GII . 4VLP , SandPdomainbindingtoimmobilizedALIXbyELISA . Errorbars indicate mean\u00b1SD with three replicates . b GII . 4 VLP , S and P domain binding to immobilized TSG101byELISA . Error bars indicate mean\u00b1SDwith three replicates . c Bio - Layer Interferometry ( BLI ) to determine the binding af \ufb01 nity of GII . 4 S and P domains to biotinylated ALIX and TSG101 ( bALIX and bTSG101 ) . d Effect of blocking virus interaction with ALIX and TSG101 on viral replication using speci \ufb01 c pAbsat1h ( gray ) and 24h ( pink ) . Error barsindicate mean\u00b1SDcalculated using2 HIE replicates for 1h and 4 HIE replicates for 24h ( with two technical replicates / sample ) . e Confocal microscopy to probe GII . 4 - ALIX colocalization on the cell surface by detecting VP1 ( green ) and ALIX ( red ) using Gp Syd - pAb and rabbitanti - ALIX pAb . f ELISA to evaluate binding of GII . 4 S - and mutant \u0394 S - domains to immobilized ALIX . Error bars indicate mean\u00b1SD with 3 replicates . g Endocytosis induced by wild type GII . 4 VLP and GII . 4 VLP lacking the ALIX \u2013 binding motif ( \u0394 VLP ) usingFM1 - 43FXinHIEsat37 o C ( n = 3HIEreplicates ) . h Viralreplicationinthe presence of GII . 4 WT VLPs ( pink bars ) and \u0394 VLPs ( blue bars ) was compared to untreated ( black bar ) at 24h ( blue dots ) . 1h ( gray dots ) represents bound virus . Error bars indicate mean\u00b1SD calculated using 2 HIE replicates for 1h and 3 HIE replicatesfor24h ( eachconditionwithtwotechnicalreplicates ) . i GII . 4replication inWT J2and J2 ALIX - KD HIEmonolayersindicated bypercentfoldchange inviralRNA using the 2 \u2212 \u0394\u0394 CT method . Error bars indicate mean\u00b1SD calculated using 3 HIE replicates for each condition with two technical replicates . All the experiments wererepeatedindependentlythreetimeswithsimilarresults . P valuesfor d , h , and i , relative to untreated control were calculated using one - way ANOVA , Dunnett \u2019 s multiple comparisons test . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 5 a direct link between GII . 4 and ALIX as HIEs incubated with VLPs ( 10 min or 1h ) showed GII . 4 colocalization with ALIX at both time points , withmoreVLP - ALIXcolocalizationobservedat1hcomparedto 10min ( Fig . 4e ) . Notably , media - treated control cells exhibited no signi \ufb01 cant cell surface ALIX staining compared to VLP - treated cells , implying that GII . 4 binding induced ALIX recruitment to the cell sur - face ( Fig . 4e ) . Both ELISA and BLI experiments showed that ALIX binds the GII . 4 capsid S - domain , with a conserved \u201c LYTPL \u201d sequence ( Fig . S4a ) . This sequence was mutated ( \u0394 ) to \u201c AATAL \u201d and this \u0394 S - domain was expressed and puri \ufb01 ed from E . coli . ELISA showed that the \u0394 S - domain does not bind ALIX , thus , con \ufb01 rming the ALIX binding site ( Fig . 4f ) . To further understand the role of ALIX in entry , an ALIX mutant VLP ( \u0394 VLP ) was generated by substituting the \u201c LYTPL \u201d sequence with \u201c AATAL \u201d in GII . 4 Syd VLP . After con \ufb01 rming that the ALIX mutation did not interfere with VLP formation ( Fig . S4f ) , GII . 4 Syd VLP and \u0394 VLP were analyzed using both endocytosis and viral inhibition assays . Both assays showed that the \u0394 VLP was identical to GII . 4 Syd VLP in terms of inducing endocytosis and the ability to block GII . 4 viral replication ( Figs . 4g and 4h ) , which led us to reassess \u0394 VLP binding to ALIX . Binding assays showed that although \u0394 VLP ( unlike the \u0394 S - domain ) boundtoALIX , thebindingwasreducedcomparedtotheGII . 4SydVLP ( Figs . S4g and S4h ) . Thissuggests that , like TSG101 , there maybe more thanoneALIXbindingsitepresentontheviralcapsidinadditiontothe site present in the S - domain . To study the role of ALIX in GII . 4 infection , attempts were made to generate a J2ALIXknock - out HIE but were unsuccessful because the KO cultures did not survive . This led us to generate J2 ALIX knock - down ( J2 ALIX - KD ) HIEs expressing an ALIX - shRNA and a TurboGFP sensor ( Fig . S4i ) . The J2 ALIX - KD cells demon - strated a signi \ufb01 cant reduction in ALIX expression ( 75 % reduction ) compared to the parental J2 cultures ( Fig . S4j ) and in GII . 4 replication ( 75 % reduction ) ( Fig . 4i ) indicating an active role of ALIX in GII . 4 infection . Endolysosomal dynamics and membrane repair mechanisms regulate GII . 4 entry Endolysosomal dynamics play an important role in GII . 3 replication 6 , which led us to investigate the same for GII . 4 . This was achieved by disrupting endosomal maturation and traf \ufb01 cking using YM201636 [ FYVE - type zinc \ufb01 nger - containing phosphoinositide kinase ( PIKfyve ) inhibitor ] and vacuolin 1 ( PIKfyve and lysosomal exocytosis inhibitor ) , which resulted in up to 70 \u2013 80 % reduction in GII . 4 replication com - pared to untreated cells ( Fig . 5a , b ) . Time course experiments showed the presence of inhibitors up to 1 h reduced virus yield by 40 % , whereasupto80 % inhibitioninyieldwasobservedwhentheinhibitors were present up to 2 h or 24 h , thus , suggesting the inhibitor was effective in the early stages of GII . 4 entry ( Figs . S5a and S5b ) . Inhibition of lysosomal exocytosisby vacuolin - 1 led toa reduction in GII . 4 replication , indicating that GII . 4 might require acid sphingo - myelinase ( ASM ) translocation and surface ceramides for infection like P = < 0 . 0001 * * * * G e no m e e qu i va l e n t s / w e ll P = 0 . 0090 P = 0 . 0007 * * * P = < 0 . 0001 * * * * * * a c b d e f P = 0 . 0012 * * P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * C e ll s w i t h P I a nd DA P I c o - s t a i n i ng P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * P ea r s on C o rr e l a t i on C o e ff i c i e n t ( P CC ) P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * P = < 0 . 0001 * * * * VP1 Gal - 3 ALIX / Gal - 3 VP1 / ALIX / Gal - 3 M ed i a V L P ( 1 h ) ALIX 10 \u00b5m 10 \u00b5m VLP - treated ( 1 h ) Media VP 1 L A M P - 1 M e r ge Fig . 5 | GII . 4 entry is sensitive to factors controlling endo - lysosomal home - ostasis and induces membrane wounding and subsequent wound repair mechanisms . a GII . 4 replication in the presence of PIKfyve inhibitor YM201636 at 1h ( black ) and24h ( red ) at37 o C . b GII . 4replicationinthepresenceofvacuolin - 1 ( a lysosomalexocytosisinhibitor ) at1h ( gray ) and24h ( purple ) . c GII . 4replicationin the presence of acid sphingomyelinase ( ASM , amitriptyline ) and neutral sphingo - myelinase ( NSM , GW4869 ) inhibitors at 1h ( black ) and 24h ( blue ) . d Cell injury determination using propidium iodide ( PI ) uptake assay . Right panel : Graph quantitating membrane injury calculated by counting number of PI spots ( red ) counterstained with DAPI ( blue ) when HIEs were incubated with VLPs ( ROI = 10 ) , VLP + Ca 2 + ( ROI = 12 ) and media ( ROI = 10 ) . Signi \ufb01 cance was calculated using two - way ANOVA , Tukey \u2019 s multiple comparisons test . e Immuno \ufb02 uorescence staining showingGII . 4 - inducedlysosomalexocytosisrepresentedbythepresenceofLAMP - 1 on the apical cell surface in media - treated cells compared to VLP - treated cells . LAMP - 1 ( red ) andVP1 ( green ) colocalizationwasdetectedusingmouseanti - LAMP - 1 mAband Gp Syd - pAb . ( n = 3 HIE replicates ) . f VP1 , ALIX and Gal - 3 colocalization in VLP - treated and media - treated cells ( 1h at 37 o C ) using confocal microscopy . VP1 ( green ) , ALIX ( red ) and gal - 3 ( white ) were detected using Gp Syd - pAb , rabbit anti - ALIX pAb and rat - anti - gal3 . Right panel : Graph showing colocalization between VP1 , gal - 3 and ALIX as estimated by Pearson Correlation Coef \ufb01 cient using EzCo - localization ( ROI = 16 , black for VLP - and ROI = 17 , blue for media - treated HIEs ) . Errorbarsindicate mean\u00b1SDand P valueswerecalculated using two - wayANOVA , \u0160 \u00edd\u00e1k \u2019 s multiple comparisons test . All the experiments were repeated indepen - dently three times with similar results . Error bars for a \u2013 c indicate mean\u00b1SD cal - culatedusing2HIEreplicatesfor1hand3HIEreplicatesfor24hwithtwotechnical replicates / sample . P values were calculated using one - way ANOVA , Dunnett \u2019 s multiple comparisons test by comparing replication at 24h to untreated control . Source data are provided as a Source Data \ufb01 le . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 6 other caliciviruses 6 , 18 . Therefore , we assayed different enzymes involved in ceramide synthesis for their effect on GII . 4 replication using inhibitorsagainst ASM ( amitriptyline ) , neutral sphingomyelinase ( NSM ; GW4869 ) , and ceramide synthase [ fumonisin 1 ( FB1 ) ] . In addi - tion , we tested if addition of exogenous sphingomyelin ( SM ) , a lipid product upstream of ceramide , has any effect on GII . 4 replication . Among the inhibitors tested , amitriptyline signi \ufb01 cantly inhibited GII . 4 replication ( up to 85 % ) whereas GW4869 , SM addition and FB1 , had no effect con \ufb01 rming that like GII . 3 , GII . 4 requires ASM activity ( Figs . 5c , S5c and S5d ) . FB1 - treated cells had reduced CD95 expression and enhanced intracellular lysosomal associated membrane protein 1 ( LAMP - 1 ) expression compared to untreated cells , which con \ufb01 rmed the activity of FB1 in HIEs ( Fig . S5e ) . Since GII . 4 does not require BA for infection , we postulated that GII . 4 - induced lysosomal exocytosis may catalyze ceramide generation as reported with feline calicivirus ( FCV ) or MNV in porcine kidney LLC - PK cells 18 . Using a ceramide - speci \ufb01 c rabbit antibody 27 , we established that GII . 4 VLPs induce surface cer - amides in HIEs by confocal microscopy ( Fig . S5f ) , con \ufb01 rming that ASM translocation and ceramide generation are associated with GII . 4 infection . Lastly , we examined cellular mechanisms associated with lysoso - mal exocytosis , ASM translocation and surface ceramide generation , which are known to be associated with endocytosis and wound repair mechanisms 28 . GII . 4 association with membrane wounding was inves - tigated using a propidium iodide ( PI ) uptake assay . PI uptake was sig - ni \ufb01 cantly increased in the presence of GII . 4 Syd VLPs ( 10min incubation ) compared to untreated cells and GII . 4 Syd VLP - treated cells with calcium , which facilitates membrane repair by early signaling 29 ( Fig . 5d ) . GII . 4 induction of membrane wounding was associated with VLP binding to cells at 4 o C ( Fig . S5g ) and requires initial VLP interaction with cell surface histo - blood group antigens ( HBGA ) , glycans known to be essential for virus infection . Thus VLP - induced membrane wounding did not occur in isogenic HIEs in which the fucosyltransferase 2 ( FUT2 ) gene was knocked out and thus do not express secretor - dependent HBGAs on their plasma membranes 4 ( Fig . S5h ) . GII . 4 - induced membrane wounding was further con \ufb01 rmed by staining the VLP - treated and untreated cells for ( 1 ) LAMP - 1 and ( 2 ) regulators of endo - lysosomal processes and wound healing mechan - isms , Rab11 and Rab14 30 , 31 . GII . 4 VP1 localized with LAMP - 1 ( Fig . 5e ) , Rab11 ( Fig . S5i ) , and Rab14 ( Fig . S5j ) , on the plasma membrane of HIEs treated with GII . 4 Syd VLPs , validating GII . 4 entry is associated with membrane wounding and healing mechanisms . Subsequently , we revisited gal - 3 and ALIX , as both are known to be coordinating factors in detecting membrane damage and repair 32 , 33 . As expected , we saw that traf \ufb01 cking of both gal - 3 and ALIX to the cell surface was induced by GII . 4 Syd VLP - treatment compared to untreated cells and a high degree of colocalization was observed between gal - 3 , ALIX , and GII . 4 VP1 ( Fig . 5f ) , consistent with the fact that GII . 4 enters the cell by compromising membrane integrity . Discussion Virusesexploitmanycellularpathwaysforcellentry . Sometimessimilar effectors in \ufb02 uence multiple endocytosis routes making it challenging to pinpoint a speci \ufb01 c entry mechanism 34 . For example , both porcine sapovirus ( PoSaV ) and feline calicivirus ( FCV ) are endocytosed via dynamin - and cholesterol - dependent clathrin - mediated endocytosis into kidney cells 35 , 36 , whereas murine norovirus ( MNV ) is endocytosed via dynamin - and cholesterol - dependent , clathrin - and caveoli - independent endocytosis into RAW264 . 7 macrophages 14 . Since most studies of virus entry are carried out in immortalized cell lines that do not fully recapitulate the properties of the original host , characteriza - tion of virus entry into HIEs offers previously unrealized insights into intestinal biology 37 . HIEs are relevant in that they are multicellular , nontransformed , physiologically active human cultures that contain the necessary susceptibility factors for human tropic HuNoV infection . Here , we report the pandemic GII . 4 virus and VLPs initiate a pre - viously uncharacterized and complex entry process into HIEs that involves carriers of a dynamin - independent , clathrin - independent endocytosis pathway , gal - 3 , the ESCRT protein ALIX , and wound repair mechanisms ( Fig . 6 ) . To our knowledge , this viral entry process into nontransformed HIE cells was not previously described , although individual components have been associated with entry of several other viruses into a range of different cell types . VLPs were used for many of our entry studies because infectious virus cannot yet be passaged inde \ufb01 nitely and current yields of virus from infected HIEs are not suf \ufb01 cient to produce puri \ufb01 ed particles for extensivebiochemicalstudies . Since1992 , whenHuNoVVLPswere \ufb01 rst produced and characterized 38 , they have been extensively docu - mented as useful surrogates for studies of virus structure 39 \u2013 42 , anti - genicity , immunogenicity , and glycan binding 43 \u2013 45 . Here , we validated that VLPs are useful for studying virus entry by documenting their effectiveness as competitors of homologous virus replication . VLPs also recapitulate virus differences as VLPs of different strains exhibit differencesinvirusentry ( Fig . S1d ) . GII . 4isrecognizedtobindtoHBGA glycans through the P domain of the capsid 46 , 47 , yet surprisingly , endocytosis into HIEs was not observed with the P domain ( dimer ) alone but wasobserved only with complete VLPs ( having 180 copies of the capsid VP1 protein ) . These results indicate that the capsid shell domain participates in virus entry into cells and add evidence to the hypothesis that cells contain an entry co - receptor in addition to HBGA glycans that serve as initial binding factors for HuNoV infection . Alternatively , it could be that avidity effects , such as multiple inter - actions linked together , are important for endocytosis . We show that GII . 4 uses the CLIC pathway for entry in a HBGA glycan - dependent manner ( Fig . 6 ) . The CLIC pathway is reported to be involved in uptake of abundant surface proteins such as CD44 and glycosylphosphatidylinositol - anchored proteins 25 , 26 , bacterial toxins ( CTxB and Shiga ) 24 , 48 , adeno - associated virus 49 and SARS - CoV - 2 50 . Use of the CLIC pathway by GII . 4 HuNoV in nontransformed HIEs is con - sistent with the CLIC pathway being operational in vivo in mouse enterocytes 23 and tubular invaginations observed in giant unilamellar vesicles treated with GII . 4 VLPs 51 . Our study shows that gal - 3 is recruited by GII . 4 at the site of the GII . 4 - membrane interface and GII . 4 - gal - 3 interactions are necessary for infection , suggesting a putative role of gal - 3 in endocytosis . Although intracellular gal - 3 and extra - cellular gal - 3 have a myriad of functions , extracellular gal - 3 is a key player in the biogenesis of CLIC structures 22 , 23 . Gal - 3 is also known for its crosslinking abilities , organizing membrane proteins / receptors and in \ufb02 uencing their function and signaling 52 . We hypothesize that , as reportedwithCD44 , monomericgal - 3interactswithGII . 4 ( boundtoits still uncharacterized receptor ) on the membrane surface leading to oligomerization ofthereceptor - GII . 4 - gal - 3 complex , and interactionof this complex with multipleglycosphingolipid head groups in the outer plasma membrane lea \ufb02 et causes membrane bending due to an avidity effect 22 andsubsequentlyleadstoendocytosis . AlthoughGII . 4VLPsare shownto beintrinsicallycapable of forming tubular structures in giant unilamellar vesicles , it is unknown if the formation of such structures alone is suf \ufb01 cient for endocytosis . Moreover , the GII . 4 - cell interaction causes recruitment of ALIX , an essential component of CLIC - mediated endocytosis 53 and an adaptor for several ESCRT - III subunits and phospholipids capable of regulating membrane curvature and tubular structure formation 54 \u2013 56 , consistent with the notion that GII . 4 triggers signaling pathways that accrue an ensemble of cellular factors with membrane modulating functions for cell entry . This leads to many interesting questions such as how does GII . 4 engage gal - 3 and ALIX , what prompts their recruitment to the mem - brane surface to induce endocytosis and is there an association between gal - 3 and ALIX ? Interaction of gal - 3 and ALIX with the nor - mally inaccessible S domain of the GII . 4 capsid suggests that the GII . 4 capsid undergoes conformational changes , which can be attributed to Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 7 the newly recognized plasticity of VLPs under certain conditions 42 , 57 . These conformational changes in the GII . 4 capsid may be due to its binding to the initial glycan receptor through the P domain , thus , exposing the S domain to participate in virus - host interaction . This suggests that the GII . 4 P domain interaction with the glycan receptor may be instrumental in triggering gal - 3 and ALIX recruitment to the cell surface , which is further supported by the fact that we see increased gal - 3 and ALIX recruitment with increased VLP binding . However , the GII . 4S domain interaction is still required with gal - 3 and ALIX to drive receptor oligomerization and subsequent endocytosis , which is not possible with P domain alone . To understand GII . 4 - triggered gal - 3 and ALIX translocation to the membrane , we investigated both endolysosomal traf \ufb01 cking and membrane wounding mechanismsdue to the well - documented role of gal - 3 and ALIX in both processes 32 , 33 . Wounding experiments demon - strated thatGII . 4 particles directlyinducemembranewounding , which requires plasma membrane glycan expression ( Figure S5h ) and leads to lysosomal exocytosis and generation of surface ceramides ( Figs . 5e and S5f ) . These events indicate that GII . 4 exploits membrane wound - ing and subsequent repair mechanisms , similar to non - enveloped adenovirus 58 and the protozoan parasite Trypanosoma cruzi in HeLa cells 59 , for infection . Membrane wounds are repaired by reducing membrane tension , lesion removal by endocytosis , or shedding of wounded areas 28 , 29 , 60 , and compelling evidence implicate the involvement of the gal - 3 , ALIX and ESCRT proteins in membrane repair mechanisms 32 , 33 , 60 . Gal - 3 is reported to recruit ALIX in response to lysosomal damage , further promoting interaction between ALIX and the ESCRT III effector CHMP4 33 , which is known for its membrane - modulating functions . Moreover , ALIX is known to regulate epithelial cell polarity and maintain the integrity of the epithelial barrier by physically interacting with F - actin by binding the Rho family of small GTPases Cdc42 and Rac1 61 , responsible for actin polymerization and endocytosis . Cdc42 - induced cell polarization is regulated by membrane recycling pathways 62 , 63 ( involving Rab11 and Rab14 ) , which are critical for both wound healing and CLIC - mediated endocytosis . This indicates that GII . 4 - cell interactions trigger a series of signaling events and cross talk between cellular components activating membrane repair and recy - cling events that culminates in endocytosis . It is known that inhibiting regulators of the CLIC pathway deters membrane repair processes 64 , 65 , further validating our results on GII . 4 entry , where inhibiting ALIX , lysosomal exocytosis , and ASM translocation inhibited GII . 4 replica - tion . Although , the role of ASM or ceramide in the CLIC pathway is unclear , considering the fact that ASM mediates traf \ufb01 cking of palmi - toylated proteins to the plasma membrane ( in \ufb02 uencing lipid raft formation ) 66 and ceramide production leads to plasma membrane repairvia endocytosis 28 , both ASM and ASM - generated ceramide seem to be essential for GII . 4 entry and infection . Our observation that endocytosis is not observed by adding ceramide exogenously 6 , con - \ufb01 rms that endocytosis requires both ASM translocation and sub - sequent conversion of membrane sphingomyelin to ceramide . Ceramides andsphingolipidsplaymultifacetedrolesasmediators of membrane dynamics by in \ufb02 uencing cellular signaling , modulating receptor conformation , and clustering , which can critically affect endocytosis 67 , 68 . GII . 4 binds HBGAs in glycosphingolipid - containing membranes similar to gal - 3 , which is a glycan sensor and has a high Gal - 3 Rab11 Rab14 ASM Cdc42 Cholesterol Membrane repair EndocytosisAdaptor for membrane modulating proteins Maintain cell polarity and epithelial barrier by interacting with Cdc42 Membrane organizer Membrane repair EndocytosisAdaptor for ALIX Membrane recycling EndocytosisCdc42 regulator Membrane recycling EndocytosisCdc42 regulator Arf6 - mediated regulation of lipid domains Membrane recycling Endocytosis Cdc42 regulator Arf6 - mediated regulation of lipid domains Regulates membrane fluidity EndocytosisRegulates receptor trafficking Regulates sphingomyelin homeostasis on the plasma membrane Lysosomal trafficking Membrane repair Endocytosis Lysosome ASM C e ll i n j u r y triggers SM Cer Membrane repair machinery Membrane reorganization and receptor clustering CLICs V - ATPase Endosomal acidification Virus / genome release Cdc42Cholesterol 4 1 3 5 6 7 8 9 2 Rab11 / Rab14 HBGA ASM HBGA HuNoVbinds to these complex glycans expressed by secretor positive individuals Regulators of CLIC pathway ESCRT Membrane repair machinery Cofactors ? Fig . 6 | GII . 4 uses a complex entry mechanism involving cellular wound repair mechanisms and CLIC - mediated endocytosis . ( 1 ) Binding of GII . 4 with its glycan receptor ( HBGAs and possibly with a still unidenti \ufb01 ed co - receptor ) on the cell surface ( 2 ) inducesplasmamembrane wounding ( 3 ) triggeringsignaling responses that direct multiple membrane repair cellular components to the injury site . ( 4 ) ASM translocation to the plasma membrane surface results in conversion of sphingomyelin ( SM ) to ceramide . ( 5 ) Ceramide formation along with other mem - brane repair processes involving gal - 3 ( glycan damage sensor ) , ALIX ( Ca 2 + sensor ) and membrane recycling processes regulated by Rab11 and Rab14 result in mem - brane reorganization and receptor clustering leading to ( 6 ) tubular carrier forma - tion due to multiple interactions of virus with host factors causing membrane bending and ( 7 ) endocytosis regulated by Cdc42 and cholesterol . ( 8 ) GII . 4 entry into the cell results in V - ATPase regulated endosomal acidi \ufb01 cation causing ( 9 ) conformational changes in the virus capsid and release of the viral genome from the endosomal compartment . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 8 af \ufb01 nity for HBGAs 69 , 70 . Our data in isogenic HIEs lacking FUT2 - dependent HBGAs on the cell surface show that GII . 4 binding to suchHBGAs isnecessarytoinducemembrane wounding . We posit this triggers ceramide - mediated signaling and changes in membrane / receptor dynamics leading to gal - 3 recruitment to the interaction site and promoting cross - talk between GII . 4 and effectors of membrane repair and recycling . We postulate that recruited gal - 3 acts both as crosslinking membrane organizers and as a switch in immobilizing the ESCRT machinery followed by subsequent endocytosis and scission of CLIC structures . Although gal - 3 is a multifunctional protein strictly regulated by its ability to recognize exposed glycans , the exact mechanism of how it \ufb01 ts into the cell signaling process and recruits ESCRT proteins remains unclear and requires further investigation . In this manuscript , we have shown that GII . 4 HuNoV exploits unexpected entry mechanisms to enter the cell by causing membrane wounding , triggering membrane repair mechanisms and inducing endocytosis and endosomal acidi \ufb01 cation for viral infection ( Fig . 6 ) . Our data clearly highlight the strain - speci \ufb01 c similarities and differences between non - pandemic , BA - dependent HuNoV strains and the pandemic , BA - independent GII . 4 HuNoV . Based on our results , we posit that the dominance of GII . 4 over other HuNoV strains may be attributed to its superior ability to exploit constitutive endocytic pathways reserved for recycling abundant surface proteins and membrane repair , in addition to its ability to overcome the host innate response 5 . Other highly pathogenic microbes and toxins ( SARS - CoV - 2 , Shiga toxin and CTxB ) share many common features of entry into cells with GII . 4 HuNoV entry mechanisms described herein 24 , 48 , 50 . Fur - ther identi \ufb01 cation of a speci \ufb01 c receptor coupled with a detailed understanding of the role of each of the cellular factors identi \ufb01 ed here for infection will be helpful in deciphering the speci \ufb01 c cel - lular tropism of GII . 4 HuNoVs . This raises interesting questions regarding whether utilization of a particular endocytic pathway by these highly infectious pathogens is associated with their pathogenesis and whether features of the CLIC route of entry and associated membrane repair mechanisms represent new ther - apeutic targets . Methods Virus and VLPs The virus used in these studies was TCH12 - 580 , a GII / Hu / US / 2012 / GII . 4 Sydney [ P31 ] / TCH12 \u2013 580strain . Thestoolsamplewaspreparedas10 % stool \ufb01 ltrates in phosphate - buffered saline ( PBS ) with viral titer determinedbyreal - timeRT - PCR 2 . VLPswereproducedinabaculovirus expression system using open reading frame 2 ( ORF2 ) + ORF3 + untranslated region ( UTR ) sequences representing HuNoV strains ( Table 1 ) . Pharmacological inhibitors , additives , recombinant proteins , and antibodies A list of the various inhibitors , additives and recombinant proteins used in this study is provided in Table 2 . Preparation of J2 HIE monolayers Human jejunal HIE was established from tissue from a patient under - going bariatric surgery . The study protocol was approved by the Baylor College of Medicine Institutional Review Board . Preparation of HIE cultures used in the current study were described previously 2 , 3 . Brie \ufb02 y , a permissive , jejunal secretor positive ( J2 ) HIE line was used for almost all the experiments described in this work . An isogenic , genetically modi \ufb01 ed J2Fut2 knockout HIE that lacks the FUT2 enzyme , as previously described 4 was used in select membrane wounding studies . To prepare monolayers , 3D cultures of HIE were dissociated by trypsinization followed by gentle pipetting and passing the cells through a 40 \u03bc m cell strainer . Cells were pelleted and resuspended in a proliferation medium [ CMGF ( + ) containing the ROCK inhibitor Y - 27632 ( 10 \u03bc M , Sigma ) ] , and seeded in a 96 - well plate ( 100 , 000 cells per well ) . After 24 h of cell growth , differentiation medium was added to the cells and the cells were allowed to differ - entiate for 5 days with intermittent medium change . Table 1 | Norovirus VLP speci \ufb01 cations Genotype Variant GenBank Accession No . GII . 4 Sydney ( Syd ) JX459908 GII . 4 Grimsby ( GRV ) AJ004864 GII . 4 Lanzou ( HoV ) EU310927 GII . 4 Farmington Hills ( FH ) AY502023 GII . 4 New Orleans ( NO ) GU445325 GII . 4 Sydney Alix mutant NA GII . 3 TCH04 - 577 AB365435 GI . 1 Norwalk M87661https : / / www . ncbi . nlm . nih . gov / nuccore / 106043086 / GII . 17 Katrina ABD95934 Table 2 | List of inhibitors , additives and recombinant proteins Reagent Source Identi \ufb01 er Inhibitors Dynasore SCBT sc - 202592 Mitmab Tocris 4224 M \u03b2 CD Sigma C4555 Filipin Sigma SAE0087 Nystatin Sigma N6803 U18666A SCBT sc - 203306 Amitriptyline Sigma A8404 GW4869 SCBT sc - 218578 Cytochalasin D SCBT sc - 201442 Genistein SCBT sc - 3515 Nocodazole Sigma 487929 Bre \ufb01 ldin A SCBT sc - 200861 Golgicide A SCBT sc - 215103 EIPA SCBT sc - 202458 ML141 SCBT sc - 362768 NSC 23766 SCBT sc - 204823 CT04 Cytoskeleton Inc . CT04 CAS 879127 - 07 - 8 SCBT sc - 203934 Calphostin C Sigma C6303 Blebbistatin SCBT sc - 203532B Wiskostatin SCBT sc - 204399 Ba \ufb01 lomycin A1 SCBT sc - 201550A YM201636 SCBT sc - 204193 LY 294002 Sigma 440202 Vacuolin - 1 Sigma 673000 TD139 Cayman Chemical 28400 Additives / and recombinant proteins GCDCA Sigma G0759 Cholesterol Sigma C4951 His - tagged recombinant ALIX Fitzgerald 80R - 1259 His - tagged recombinant TSG101 Fitzgerald 80R - 1296 Recombinant gal - 3 Biolegend 599706 Cholera Toxin B subunit Sigma C9903 Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 9 HuNoV infection of J2 HIE monolayers HuNoV infection was performed with the J2 HIEs without the addition of bile / bile acids 2 , 3 . Brie \ufb02 y , J2 HIEs were incubated with 9\u00d710 5 GEs of GII . 4 stool \ufb01 ltrate in CMGF ( - ) for 1 h at 37 o C . The monolayers were washedtwicewithCMGF ( - ) toremoveunattachedviruses . Oneplateof infected J2 HIEs was frozen at this time point ( as binding reference for 1 h ) and the other plate of infected J2 HIEs were incubated for 24 h in the differentiation medium followed by Reverse Transcriptase Quan - titative Polymerase Chain Reaction ( RT - qPCR ) to quantitate HuNoV replication . Toxicity assessment in HIEs Pharmacological inhibitors were tested for toxicity in HIEs before testing their effect on HuNoV replication . The HIEs were incubated with the inhibitors for 24 h at 37 o C and the supernatant was subjected to LDH - Glo \u2122 cytotoxicity assay ( Promega ) according to the manu - facturer \u2019 s instructions . HIE lysate was used as positive control and untreated supernatant was used as negative control and Table S1 reports toxicity results for all inhibitors . Infection assays in the presence of VLPs or inhibitors The effect of VLPs or pharmacological inhibitors on GII . 4 entry and infection were studied by pretreating the J2 HIE monolayers with them 1 h prior to infection . Infection was carried out in the presence of inhibitor for 1 h at 37 o C . After washing and unbound virus removal , fresh media containing VLPs or inhibitors was added to the infected monolayers and incubated for 24 h until the cells were harvested for RNA isolation . RNA extraction and RT - PCR Total RNA was extracted from each well using the KingFisher Flex Puri \ufb01 cation System and MagMAX \u2122 Pathogen RNA / DNA Kit . For RT - qPCR , the primer pair COG2R / QNIF2d and probe QNIFS were used for GII . 4 with qScript XLT One - Step RT - qPCR ToughMix reagent with ROX reference dye . Reactions were performed on an Applied Biosystems StepOnePlus thermocycler using the following conditions : 50\u00b0C ( 15 min ) , 95\u00b0C ( 5min ) , followed by 40 cycles of 95\u00b0C ( 15s ) and 60\u00b0C ( 35 s ) . A standard curve based on a recombinant HuNoV HoV RNA transcript was used to quantitate viral GEs in RNA samples 2 . Western blot analysis Samples were boiled with 4X sample buffer supplemented with \u03b2 - mercaptoethanol at 95 \u00b0C for 10 min and subjected to SDS - PAGE for resolving the proteins . The proteins were transferred to nitrocellulose membrane and Western blot analysis was carried out using speci \ufb01 c antibodies against target proteins . Villin and GAPDH were used as cell control which were detected using 1 : 1000 dilution of mouse anti - villin ( # sc - 373997 , Santa Cruz Biotechnology ) and rabbit anti - GAPDH ( Cat # 10494 - 1 - AP , Proteintech ) antibodies . Dot - blot analysis Dot - blotanalysiswascarriedoutbyadding10 \u00b5 lofpuri \ufb01 edproteinson nitrocellulose membrane as dots . After air drying , the membrane was blocked overnight at 4 o C . Blocked membranes were probed with tar - get proteins and their corresponding antibodies to detect binding . Immuno \ufb02 uorescence staining and imaging of J2 HIE monolayers J2 HIE monolayers were either grown in 10 - well glass - bottom culture slides ( # 543979 , Greiner Bio - One ) or 8 - well 15 \u00b5 - slides ( # 80826 , Ibidi ) for imaging . J2 HIEs were differentiated for 5 days and infected with 1\u00d710 6 GEs GII . 4 . For studies using VLPs , the cells were treated with 1\u00d710 12 particles of VLPs for either 10 min or 1h . The cells were \ufb01 xed with 4 % paraformaldehyde ( PFA ) for 20min at room temperature , blockingwith5 % BSAin0 . 1 % TritonX - 100 ( forpermeabilization ) inPBS for 30 min at room temperature . The cellswere incubated overnight at 4 \u00b0Cwithprimaryantibodies . Allthesubsequentstepswereperformed in PBS + 0 . 1 % Triton X - 100 . HuNoV capsid protein ( VP1 ) was detected using guinea pig anti - HuNoV polyclonal Ab ( pAb ) , 1 : 1000 dilution . Gal - 3 , ALIX , TSG101 , LAMP - 1 , Rab11 , Rab14 , Flotilin - 1 , CD44 and glycosylphosphatidylinositol - anchored protein ( GPI - AP ) were detec - ted using 1 : 200 dilution of rat anti - gal - 3 ( # 125402 , Biolegend ) , rabbit anti - ALIX ( # 12422 - 1 - AP , Proteintech ) , anti - TSG101 ( # 14497 - 1 - AP , Pro - teintech ) , rabbit anti - LAMP - 1 ( # 9091 , Cell Signaling Technologies ) , rabbit Rab11 ( # 5589 , Cell Signaling Technologies ) and rabbit Rab14 ( # A12752 , ABclonal ) , mouse anti - Flotilin - 1 ( # 610821 , BD Biosciences ) , rabbit anti - CD44 ( # 15675 - 1 - AP , Proteintech ) , and rabbit anti - GPI - AP ( # 10104 - 1 - AP , Proteintech ) antibodies , respectively . Ceramide detec - tion was carried out as described previously 6 using a speci \ufb01 c rabbit anti - ceramide pAb 28 , 71 , 72 . After washing ( 3 times , 10min each ) , the cells were incubated with 1 : 500 dilution of goat anti - rat 549 ( # 612 - 142 - 120 , Rockland ) , donkey anti - mouse 549 ( # 610 - 742 - 124 , Rockland ) , donkey anti - rabbit 649 ( # 611 - 743 - 127 , Rockland ) , and anti - guinea pig 488 sec - ondary antibodies ( # 606 - 141 - 129 , Rockland ) , to visualize the viral and cellular proteins . Actin staining wasdone using 1 : 500 dilution of Alexa - 647 Phalloidin ( # A22287 , ThermoFisher Scienti \ufb01 c ) . The cells were washed three times , and nuclei were stained with 4 , 6 - diamidino - 2 - phenylindole ( DAPI ) ( 300 nM ) for 5minatroom temperature followed by subsequent Z - stack images captured using a Nikon A1 confocal microscope . TheeffectofinhibitorsonJ2HIEswastestedbyincubatingthecells withpharmacologicalinhibitorsfor24hand \ufb01 xingthecellswith4 % PFA for 20min ( for actin ) or methanol for 5min ( for tubulin ) at room temperature . Actin was stained using Alexa - 647 Phalloidin and tubulin was stained using mouse anti - tubulin antibody ( # T8203 , Sigma ) fol - lowed by using goat anti - mouse 594 antibody ( # 610 - 742 - 124 , Rockland ) as a secondary antibody . The images were captured either by Nikon A1 confocal microscope using NIS - Elements Viewer 4 . 20 , GE Healthcare DeltaVision Deconvolution Microscope using softWoRx - software Acquire Version : 7 . 2 . 1 or by using a ZEISS Laser Scanning Micro - scope LSM 980 confocal microscope using ZEISS ZEN 3 . 5 ( blue edition ) software . The images were further processed using ImageJ2 / FIJI . Speci \ufb01 city testing of commercial antibodies against GII . 4 VLP Antibodies ( gal - 3 , ALIX , LAMP - 1 , Rab11 , and Rab14 ) were incubated with 5 \u00b5 g of GII . 4 VLPs for 1h at room temperature . An antibody pull - down was performed by adding 25 \u00b5 l of PBS - washed protein A / G magnetic beads ( # 88802 , ThermoFisher Scienti \ufb01 c ) to the mixture and incubating the mixture for 30 min at room temperature . The beads were collected by placing tubes on a magnetic stand and the unbound proteins present in the supernatant were collected for analysis . The pelleted beads were washed with PBS ( 5 times ) and the bound proteins were analyzed after adding 100 \u00b5 l SDS - PAGE buffer and heating the beadsat95 o Cfor5min ( pull - downfraction ) . Thesupernatantandpull - down fractions were assessed for the presence of VLPs by SDS - PAGE , followed by Western blot analysis for the capsid VP1 protein ( Fig . S6a ) . Blocking assay to block GII . 4 - protein interaction using antibodies To test the blocking activity of anti - gal - 3 and anti - ALIX antibodies in blocking GII . 4 Syd VLP interactions with gal - 3 and ALIX , antibody - blocking assays were carried out using anti - gal - 3 and anti - ALIX anti - bodies . GII . 4 Syd VLPs ( 2 \u00b5 g ) were incubated with recombinant His - tagged gal - 3 and ALIX with or without anti - gal - 3 ( left ) and anti - ALIX ( right ) antibodies ( 2 and 10 \u00b5 g ) . A pull - down assay was carried out using Ni - NTA beads and GII . 4 VP1 was detected using guinea pig ( Gp ) Syd - pAb ( Fig . S6b ) . Epi \ufb02 uorescence microscopy for measurement of endocytosis Differentiated J2 HIE monolayers ( incubated with virus / VLPs for 10min ; ba \ufb01 lomycin for 1 h ) were incubated with 50nM LysoTracker Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 10 ( ThermoFisher Scienti \ufb01 c ) for 10min . The cells were washed twice with CMGF ( - ) and LysoTracker staining of acidic compartments in treated and untreated J2 HIEs ( mock ) was observed by epi \ufb02 uorescence microscopy using Olympus cellSens Standard Version 2 . 3 software . Endocytosis measurements were carried out using FM1 - 43FX ( Ther - moFisher Scienti \ufb01 c ) . HIE monolayers were treated media containing 10 \u03bc g / mL of FM1 - 43FX with or without VLPs for 10min at 37\u00b0C . Endocytosis was stopped by washing with prechilled PBS and the cells were \ufb01 xed in 4 % PFA for 20 min and nuclei were stained with 300nM DAPI for 5min at room temperature . Quantitation of the \ufb02 uorescence wasdoneusingJ - image . Brie \ufb02 y , everyexperimentwasrepeatedat least three times with 3 \u2013 4 images analyzed per condition in each experi - ment . Identical elliptical regions - of - interests ( ROIs ) were drawn per \ufb01 eld and mean \ufb02 uorescence intensity from these ROIs was measured . Dil - LDL assay Differentiated J2 HIE monolayers were treated with dynamin inhibitors ( 40 \u00b5 M of dynasore and 100 \u00b5 M mitmab ) or dimethyl sulfoxide ( DMSO ) for 1h at 37 o C . Dil - LDL was added to the cells ( 5 \u00b5 l / well ) in the presence of inhibitor and the cells were further incubated for 3 h at 37 o C . The cells were washed three times , \ufb01 xed with 4 % PFA for 20 min . Dil - LDL uptake was measured using epi \ufb02 uorescence microscopy . Time lapse microscopy Timelapsemicroscopywascarriedoutbyadding \ufb02 uorescentlylabeled VLPs with FM1 - 43FX to J2 HIE monolayers and subsequent Z - stack imageswerecaptured ( every 2 . 5 minforaperiod of1 hat 37 o C ) usinga GE Healthcare GE Healthcare DeltaVision Deconvolution Microscope . Image acquisition was done using softWoRx - Acquire Version : 7 . 2 . 1 software . The image stack was then converted into a movie using ImageJ AVI pluggin . Thin section electron microscopy J2 HIE monolayers were incubated with VLPs for 1h at 37 o C . The cells were washed twice with PBS and VLP - treated and untreated control cells were trypsinized and \ufb01 xed in 2 % paraformaldehyde \u2013 3 % phosphate - buffered glutaraldehyde . Thin sections were prepared and electron microscopy was carried out using a FEI Tecnai Spirit trans - mission electron microscope equipped with a 4K Eagle digital camera 73 . Af \ufb01 nity measurements using BLI BLI studies were carried out using an Octet RED96 instrument ( For - teBio ) . Biotinylation of ALIX , TSG101 and gal - 3 was carried out using EZ - Link NHC - LC - LC - biotin following the manufacturer \u2019 s instructions . Thebiotinylatedproteins , bTSG101 ( 0 . 5 \u03bc g / ml ) , bALIX ( 1 \u03bc g / ml ) , bgal - 3 ( 1 \u03bc g / ml ) and were loaded onto streptavidin biosensors ( 18 - 5019 , ForteBio ) in the running buffer ( 20mM HEPES , 150mM NaCl , 0 . 005 % surfactant Tween 20 , and 2 mg / ml BSA , pH 7 . 8 ) for 600 s . Af \ufb01 nity measurements were carried out by passing two - fold serial dilutions of recombinant GII . 4S ( 280 - 70 \u03bc M ) and P domains ( 144 - 36 \u03bc M ) over the captured bALIX , bTSG and bgal - 3 and allowing both association and dissociationfor900susingForteBiodataacquisition software Version 7 . 1 . 0 . 100 . The binding data were \ufb01 tted using the ForteBio data analysis software Version 7 . 1 . 0 . 38 ( 2 : 1 model ) by subtracting buffer blanks to calculate thebinding af \ufb01 nity ofGII . 4 SandP domainswith biotinylated cellular proteins ( ALIX , TSG , and gal - 3 ) . Elisa Direct binding of ALIX with GII . 4 domains ( S and P ) and VLP was determined by ELISA . ALIX ( 5 \u03bc g / ml ) - coated and blocked plates were incubated with equimolar concentrations of GII . 4 VLP , S and P domain for 2 h at room temperature . The plate was washed three times with PBST ( PBS with 0 . 05 % Tween 20 ) , and PBS followed by probing the interaction with guinea pig anti - HuNoV polyclonal Ab , 1 : 5000 dilution in 1 % blocking buffer . Horseradish peroxidase - labeled goatanti - guinea pig secondary antibody ( 1 : 5000 ) was used to detect the binding using tetramethylbenzidine ( TMB ) substrate . Color development was quenched using phosphoric acid and absorbance was measured at 450nm . Generation of GII . 4 \u0394 S - domain TheGII . 4 \u0394 S - domainconstructcontainingthe \u201c AATAL \u201d mutationatthe putative ALIX binding site was generated using the NEBaseChanger software and Q5 site - Directed Mutagenesis Kit . The sequence of the primers used was TGCGCTGCGTGCGAACAATGCT and GTCGCCGC CATTGCGATCAGTTTGATGGTC ( Integrated DNA Technologies ) . The mutant construct was thentransformed andculturedin Novabluecells made competent using a Mix & Go transformation kit ( Zymo Research ) . Cells were then cultured in 50ml Luria Broth media , con - taining 50 \u00b5 g / ml ampicillin , and puri \ufb01 ed using Qiagen Miniprep kit ( Qiagen ) . Puri \ufb01 ed DNA was then submitted to Genewiz for Sanger sequencing to con \ufb01 rm the successful incorporation of the AATAL mutation . His - tagged shell mutant was then expressed , puri \ufb01 ed and assessed using EM and SDS - PAGE . Generation of GII . 4 Syd \u0394 VLP To obtain mutant VLPs , the ORF1 + ORF2 + 3 \u2019 UTR nucleotide sequence of norovirus GII . 4 Sydney ( JX459908 ) was submitted to a commercial laboratory ( Epoch Life Science ) for modi \ufb01 cation . Nucleotides 5628 \u2013 5637 of the cDNA version of the genome were changed from TTGTATACAC to GCAGCTACAG . The coded amino acids 182 \u2013 185 of VP1 were thus mutated from \u201c LYTPL \u201d to \u201c AATAL \u201d . The laboratory synthesized the gene and after veri \ufb01 cation of the sequence , the insert was cloned into the bacmid pFastbac1 vector . The Protein and Monoclonal Antibody Production Core at Baylor College of Medicine recombinantly expressed VLPs by infecting High Five \u2122 ( # B85502 , ThermoFisher Scienti \ufb01 c ) insect cells with the bacmid construct . The VLPs were puri \ufb01 ed by passing through a 30 % sucrose cushion followed by a cesium chloride gradient and dialysis in sodium chloride / sodium phosphate buffer as previously described 74 . shRNA - mediated knockdown of the PDCD6IP gene A microRNA - adapted shRNA technology based on miR - 30 was used to knockdown gene expression in the J2 HIE line . Commercially available pGIPZ lentiviral constructs , comprising of shRNAs complementary to the sequence of the targeted PDCD6IP gene ( # RHS4430 - 200156283 , # RHS4430 - 200158656 , # RHS4430 - 200283070 ; Dharmacon ) , were providedasbacterialglycerolstocks . shRNAconstructswereextracted using Qiagen prep kit ( Qiagen ) . Lentiviral particles with the shRNAs were produced , using the third - generation lentivirus technology , by co - transfecting HEK293FT cells ( # R70007 , ThermoFisher Scienti \ufb01 c ) with a combina - tion of each GIPZ - shRNA lentiviral plasmid and three packaging plas - mids pMDLg / pRRE [ plasmid # 12251 ; Addgene ] , envelope plasmid pMD2 . G [ plasmid # 12259 ; Addgene ] , and pRSV - Rev [ plasmid # 12253 ; Addgene ] ) at a ratio of 3 . 5 : 2 : 1 : 1 , respectively , using polyethyleneimine HCl Max molecular weight ( MW ) 40 , 000 ( Polysciences ) . The culture supernatant was harvested 72 h , passed through a 0 . 45 - \u03bc m \ufb01 lter , concentrated by using LentiX - concentrator ( TaKaRa - Clontech ) according to the manufacturer \u2019 s protocol , and suspended in a high Wnt - 3A proliferation medium supplemented with 10 \u03bc M CHIR99021 and 10 \u03bc M Rho - associated protein kinase ( ROCK ) inhibitor Y - 27632 , and stored at \u2212 80 \u00b0C in 50 \u03bc l aliquots until further use . A cell suspension was prepared from three - dimensional ( 3D ) undifferentiated J2 HIEs cultivated as previously described . After trypsinization and pelleting of the cells at 300 \u00d7 g , the resulting cell pellet was suspended at a concentration of 3\u00d7 10 5 cells per ml in high Wnt - 3A proliferation medium supplemented with 10 \u03bc M CHIR99021 and 10 \u03bc M Y - 27632 . 8 \u03bc g / ml Polybrene and 50 \u03bc l of concentrated Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 11 shRNA lentivirus were added to 1\u00d7 10 5 cells , the mixture was plated in one well of a 48 - well plate and centrifuged for 1 h at 600\u00d7 g at room temperature . After spinoculation , the plate was incubated for 4 h at 37\u00b0C under 5 % CO 2 atmosphere to enhance transduction . The lenti - virus solution was removed by centrifugation , the cells were washed once with CMGF ( - ) medium , centrifuged again , embedded in 30 \u03bc l Matrigel plug , and incubated at 37\u00b0C and 5 % CO 2 in the presence of high Wnt - 3A medium with CHIR99021 and Y - 27632 for recovery . Five days post - transduction , the cells were treated with puromycin ( 2 \u03bc g / ml ) until mock - treated cells were completely dead . Single - cell clones were isolated by serial dilution of cells in 96 - well plates for shRNA - transduced HIEs , and deletion of the gene was con \ufb01 rmed by sequencing of genomic DNA from each single cell clone using primers ALIX - F and ALIX - R that amplify the portion of the shRNA targeted PDCD6IP gene . Wounding assays Differentiated J2 or J2FUT2KO HIEs were incubated with a mixture of VLP and PI ( 5 \u00b5 g / ml ) for 10 min at 37 o C . The cells were washed twice with CMGF ( - ) and \ufb01 xed with 4 % PFA for 20 min . After incubation , the cellswere washed with PBS and counterstained with DAPI for 5 min . PI / DAPI colocalization was measured to determine VLP - induced wounding . Statistical analysis Each experiment was performed three or more times independently with two or more HIE replicates ( represented by n ) . For infection assays , two HIE replicates for 1h time point and three independent HIE replicates for the 24h time point were used . RT - qPCR assays were carried out using two or more technical replicates for each HIE repli - cate depending on the treatment of cells . Samples with RNA levels below the limit of detection of the RT - qPCR assay were assigned a value that was one - half the limit ofdetection of the assay . All statistical analyses were performed on GraphPad Prism ( Version 8 . 4 . 3 ) ( Graph - Pad Software ) using geometric mean values . Comparison between untreated and treated groups at 24 h was performed using one - way ANOVA using data from independent HIE replicates of one indepen - dent experiment , with statistical signi \ufb01 cance ( P value ) determined using Dunnett \u2019 s multiple comparisons test . For imaging experiments , statistical signi \ufb01 cance was calculated either by one - way ANOVA , Dun - nett \u2019 s multiple comparisons test ( when comparing with control ) or using two - way ANOVA with Tukey \u2019 s or \u0160 \u00edd\u00e1k \u2019 s multiple comparisons test ( when comparing within groups ) . A P value above 0 . 05 was con - sidered non - signi \ufb01 cant . Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article . Data availability The authors declare that all data supporting the \ufb01 ndings of this study are available within the paper and in the Supplementary Information \ufb01 les . Rawdata isavailable as Source Data \ufb01 le . Source data are provided with this paper . References 1 . Pires , S . M . et al . Aetiology - speci \ufb01 c estimates of the global and regional incidence and mortality of diarrhoeal diseases commonly transmitted through food . PLoS ONE 10 , e0142927 ( 2015 ) . 2 . Ettayebi , K . et al . Replication of human noroviruses in stem cell \u2013 derived human enteroids . Science 353 , 1387 \u2013 1393 ( 2016 ) . 3 . Ettayebi , K . et al . New insights and enhanced human norovirus cultivation in human intestinal enteroids . mSphere https : / / doi . org / 10 . 1128 / mSphere . 01136 - 20 ( 2021 ) . 4 . Haga , K . et al . Genetic manipulation of human intestinal enteroids demonstrates the necessity of a functional fucosyltransferase 2 gene for secretor - dependent human norovirus infection . mBio 11 , e00251 \u2013 00220 ( 2020 ) . 5 . Lin , S . C . etal . Humannorovirusexhibitsstrain - speci \ufb01 csensitivityto host interferon pathways in human intestinal enteroids . Proc . Natl Acad . Sci . USA 117 , 23782 \u2013 23793 ( 2020 ) . 6 . Murakami , K . et al . Bile acids and ceramide overcome the entry restriction for GII . 3 human norovirus replication in human intestinal enteroids . Proc . Natl Acad . Sci . USA 117 , 1700 \u2013 1710 ( 2020 ) . 7 . Alvarado , G . et al . Broadly cross - reactive human antibodies that inhibit genogroup I and II noroviruses . Nat . Commun . 12 , 4320 ( 2021 ) . 8 . Atmar , R . L . et al . Comparison of microneutralization and histo - blood group antigen \u2013 blocking assays for functional norovirus antibody detection . J . Infect . Dis . 221 , 739 \u2013 743 ( 2020 ) . 9 . Ford - Siltz , L . A . , Wales , S . , Tohma , K . , Gao , Y . & Parra , G . I . Genotype - speci \ufb01 c neutralization of norovirus is mediated by antibodies against the protruding domain of the major capsid protein . J . Infect . Dis . 7 , 1205 \u2013 1214 ( 2020 ) . 10 . Lindesmith , L . C . et al . Virus \u2013 host interactions between non - secretors and human norovirus . Cell Mol . Gastroenterol . Hepatol . 10 , 245 \u2013 267 ( 2020 ) . 11 . Costantini , V . et al . Human norovirus replication in human intestinal enteroidsas model to evaluate virus inactivation . Emerg . Infect . Dis . 24 , 1453 \u2013 1464 ( 2018 ) . 12 . Ettayebi , K . et al . Antiviral activity of olanexidine - containing hand rub against human noroviruses . mBio 13 , e02848 \u2013 02821 ( 2022 ) . 13 . Estes , M . K . et al . Human norovirus cultivation in nontransformed stem cell - derived human intestinal enteroid cultures : Success and challenges . Viruses 11 , 638 ( 2019 ) . 14 . Gerondopoulos , A . , Jackson , T . , Monaghan , P . , Doyle , N . & Roberts , L . O . Murine norovirus - 1 cell entry is mediated through a non - cla - thrin - , non - caveolae - , dynamin - and cholesterol - dependent path - way . J . Gen . Virol . 91 , 1428 \u2013 1438 ( 2010 ) . 15 . Audi , A . , Soudani , N . , Dbaibo , G . & Zaraket , H . Depletion of host and viral sphingomyelin impairs in \ufb02 uenza virus infection . Front . Micro - biol . 11 , 612 ( 2020 ) . 16 . Mart\u00edn , J . J . , Holguera , J . , S\u00e1nchez - Felipe , L . , Villar , E . & Mu\u00f1oz - Barroso , I . Cholesterol dependence of newcastle disease virus entry . Biochim . Biophys . Acta 1818 , 753 \u2013 761 ( 2012 ) . 17 . Miller , M . E . , Adhikary , S . , Kolokoltsov , A . A . & Davey , R . A . Ebolavirus requires acid sphingomyelinase activity and plasma membrane sphingomyelin for infection . J . Virol . 86 , 7473 \u2013 7483 ( 2012 ) . 18 . Shivanna , V . , Kim , Y . & Chang , K . - O . Ceramide formation mediated by acid sphingomyelinase facilitates endosomal escape of calici - viruses . Virology 483 , 218 \u2013 228 ( 2015 ) . 19 . Mercer , J . & Helenius , A . Vaccinia virus uses macropinocytosis and apoptotic mimicry to enter host cells . Science 320 , 531 \u2013 535 ( 2008 ) . 20 . Amstutz , B . et al . Subversion of CtBP1 - controlled macropinocytosis by human adenovirus serotype 3 . EMBO J . 27 , 956 \u2013 969 ( 2008 ) . 21 . Krzyzaniak , M . A . , Zumstein , M . T . , Gerez , J . A . , Picotti , P . & Helenius , A . Host cell entry of respiratory syncytial Virus involves macro - pinocytosis followedbyproteolytic activation ofthe Fprotein . PLoS Pathog . 9 , e1003309 ( 2013 ) . 22 . Lakshminarayan , R . et al . Galectin - 3 drives glycosphingolipid - dependent biogenesis of clathrin - independent carriers . Nat . Cell Biol . 16 , 592 \u2013 603 ( 2014 ) . 23 . Ivashenka , A . et al . Glycolipid - dependent and lectin - driven trans - cytosis in mouse enterocytes . Commun . Biol . 4 , 173 ( 2021 ) . 24 . Kirkham , M . et al . Ultrastructural identi \ufb01 cation of uncoated caveolin - independent early endocytic vehicles . J . Cell Biol . 168 , 465 \u2013 476 ( 2005 ) . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 12 25 . Howes , M . T . et al . Clathrin - independent carriers form a high capacity endocytic sorting system at the leading edge of migrating cells . J . Cell Biol . 190 , 675 \u2013 691 ( 2010 ) . 26 . Sabharanjak , S . , Sharma , P . , Parton , R . G . & Mayor , S . GPI - anchored proteins are delivered to recycling endosomes via a distinct cdc42 - regulated , clathrin - independent pinocytic pathway . Dev . Cell 2 , 411 \u2013 423 ( 2002 ) . 27 . Krishnamurthy , K . , Dasgupta , S . & Bieberich , E . Development and characterization of a novel anti - ceramide antibody . J . Lipid Res . 48 , 968 \u2013 975 ( 2007 ) . 28 . Tam , C . et al . Exocytosis of acid sphingomyelinase by wounded cells promotes endocytosis and plasma membrane repair . J . Cell Biol . 189 , 1027 \u2013 1038 ( 2010 ) . 29 . Steinhardt , R . A . , Bi , G . & Alderton , J . M . Cellmembraneresealingby a vesicular mechanism similar to neurotransmitter release . Science 263 , 390 \u2013 393 ( 1994 ) . 30 . Escrevente , C . , Bento - Lopes , L . , Ramalho , J . S . & Barral , D . C . Rab11 is required for lysosome exocytosis through the interaction with Rab3a , Sec15 and GRAB . J . Cell Sci . https : / / doi . org / 10 . 1242 / jcs . 246694 ( 2021 ) . 31 . Tro \ufb01 menko , E . , Homma , Y . , Fukuda , M . & Widmann , C . The endo - cytic pathway taken by cationic substances requires Rab14 but not Rab5 and Rab7 . Cell Rep . 37 . 5 , 109945 ( 2021 ) . 32 . Jimenez , A . J . et al . ESCRT machinery is required for plasma mem - brane repair . Science 343 , 1247136 ( 2014 ) . 33 . Jia , J . et al . Galectin - 3 coordinates a cellular system for lysosomal repair and removal . Dev . Cell 52 , 69 \u2013 87 . e68 ( 2020 ) . 34 . Sandvig , K . , Kavaliauskiene , S . & Skotland , T . Clathrin - independent endocytosis : an increasing degree of complexity . Histochem . Cell Biol . 150 , 107 \u2013 118 ( 2018 ) . 35 . Soliman , M . et al . Porcine sapovirus Cowden strain enters LLC - PK cells via clathrin - and cholesterol - dependent endocytosis with the requirement of dynamin II . Vet . Res . 49 , 92 ( 2018 ) . 36 . Stuart , A . D . & Brown , T . D . Entryoffelinecalicivirusisdependenton clathrin - mediated endocytosis and acidi \ufb01 cation in endosomes . J . Virol . 80 , 7500 \u2013 7509 ( 2006 ) . 37 . Schutgens , F . & Clevers , H . Human organoids : tools for under - standing biology and treating diseases . Annu . Rev . Pathol . 15 , 211 \u2013 234 ( 2020 ) . 38 . Jiang , X . , Wang , M . , Graham , D . Y . & Estes , M . K . Expression , self - assembly , and antigenicity of the Norwalk virus capsid protein . J . Virol . 66 , 6527 \u2013 6532 ( 1992 ) . 39 . Devant , J . M . & Hansman , G . S . Structural heterogeneityof a human norovirus vaccine candidate . Virology 553 , 23 \u2013 34 ( 2021 ) . 40 . Jung , J . et al . High - resolution cryo - EM structures of outbreak strain human norovirus shells reveal size variations . Proc . Natl Acad . Sci . USA 116 , 12828 \u2013 12832 ( 2019 ) . 41 . Prasad , B . V . V . , Hardy , M . E . , Jiang , X . & Estes , M . K . Structure of Norwalk virus . Arch . Virol . 12 , 237 \u2013 242 ( 1996 ) . 42 . Song , C . etal . Dynamicrotationoftheprotrudingdomainenhances the infectivity of norovirus . PLoS Pathog . 16 , e1008619 ( 2020 ) . 43 . Atmar , R . L . etal . Serologicalcorrelates ofprotection againstaGII . 4 Norovirus . Clin . Vaccin . Immunol . 22 , 923 \u2013 929 ( 2015 ) . 44 . Green , K . Y . , Lew , J . F . , Jiang , X . , Kapikian , A . Z . & Estes , M . K . Comparison of the reactivities of baculovirus - expressed recombi - nant Norwalk virus capsid antigen with those of the native Norwalk virus antigen in serologic assays and some epidemiologic obser - vations . J . Clin . Microbiol . 31 , 2185 \u2013 2191 ( 1993 ) . 45 . Shanker , S . et al . Structural analysis of determinants of histo - blood group antigen binding speci \ufb01 city in genogroup I noroviruses . J . Virol . 88 , 6168 \u2013 6180 ( 2014 ) . 46 . Shanker , S . et al . Structural analysis of histo - blood group antigen binding speci \ufb01 city in a norovirus GII . 4 epidemic variant : implica - tions for epochal evolution . J . Virol . 85 , 8635 \u2013 8645 ( 2011 ) . 47 . Cao , S . et al . Structural basis for the recognition of blood group trisaccharides by norovirus . J . Virol . 81 , 5949 \u2013 5957 ( 2007 ) . 48 . R\u00f6mer , W . et al . Shiga toxin induces tubular membrane invagina - tions for its uptake into cells . Nature 450 , 670 \u2013 675 ( 2007 ) . 49 . Nonnenmacher , M . & Weber , T . Adeno - associated virus 2 infection requires endocytosis through the CLIC / GEEC pathway . Cell Host Microbe 10 , 563 \u2013 576 ( 2011 ) . 50 . Prabhakara , C . et al . Strategies to target SARS - CoV - 2 entry and infection using dual mechanisms of inhibition by acidi \ufb01 cation inhi - bitors . PLoS Pathog . 17 , e1009706 ( 2021 ) . 51 . Rydell , G . E . , Svensson , L . , Larson , G . , Johannes , L . & R\u00f6mer , W . Human GII . 4 norovirus VLP induces membrane invaginations on giant unilamellar vesicles containing secretor gene dependent \u03b1 1 , 2 - fucosylated glycosphingolipids . Biochim . Biophys . Acta 1828 , 1840 \u2013 1845 ( 2013 ) . 52 . Van Deventer , S . , Arp , A . B . & van Spriel , A . B . Dynamic plasma membrane organization : a complex symphony . Trends Cell Biol . 31 , 119 \u2013 129 ( 2021 ) . 53 . Mercier , V . et al . ALG - 2 interacting protein - X ( Alix ) is essential for clathrin - independent endocytosis and signaling . Sci . Rep . 6 , 26986 ( 2016 ) . 54 . McCullough , J . , Fisher , R . D . , Whitby , F . G . , Sundquist , W . I . & Hill , C . P . ALIX - CHMP4 interactions in the human ESCRT pathway . Proc . Natl Acad . Sci . USA 105 , 7687 \u2013 7691 ( 2008 ) . 55 . Bissig , C . et al . Viral infection controlled by a calcium - dependent lipid - binding module in ALIX . Dev . Cell 25 , 364 \u2013 373 ( 2013 ) . 56 . Bertin , A . et al . Human ESCRT - III polymers assemble on positively curved membranes and induce helical membrane tube formation . Nat . Commun . 11 , 2663 ( 2020 ) . 57 . Hu , L . et al . Atomic structure of the predominant GII . 4 human nor - ovirus capsid reveals novel stability and plasticity . Nat . Commun . 13 , 1241 ( 2022 ) . 58 . Luisoni , S . et al . Co - option of membrane wounding enables virus penetration into cells . Cell Host Microbe 18 , 75 \u2013 85 ( 2015 ) . 59 . Fernandes , M . C . et al . Trypanosoma cruzi subverts the sphingomyelinase - mediated plasma membrane repair pathway for cell invasion . J . Exp . Med . 208 , 909 \u2013 921 ( 2011 ) . 60 . Scheffer , L . L . et al . Mechanism of Ca2 + - triggered ESCRT assembly and regulation of cell membrane repair . Nat . Commun . 5 , 5646 ( 2014 ) . 61 . Campos , Y . et al . Alix - mediated assembly of the actomyosin \u2013 tight junction polarity complex preserves epithelial polarity and epithe - lial barrier . Nat . Commun . 7 , 11876 ( 2016 ) . 62 . Rodal , A . A . , Motola - Barnes , R . N . & Littleton , J . T . Nervous wreck and Cdc42 cooperate to regulate endocytic actin assembly during synaptic growth . J . Neurosci . 28 , 8316 \u2013 8325 ( 2008 ) . 63 . Lu , R . & Wilson , J . M . Rab14 speci \ufb01 es the apical membrane through Arf6 - mediated regulation of lipid domains and Cdc42 . Sci . Rep . 6 , 38249 ( 2016 ) . 64 . Nyg\u00e5rd Skalman , L . , Holst , M . R . , Larsson , E . & Lundmark , R . Plasma membrane damage caused by listeriolysin O is not repaired through endocytosis of the membrane pore . Biol . Open 7 , bio035287 ( 2018 ) . 65 . Corrotte , M . , Cerasoli , M . , Maeda , F . Y . & Andrews , N . W . Endophilin - A2 - dependent tubular endocytosis promotes plasma membrane repair and parasite invasion . J . Cell Sci . 134 , jcs249524 ( 2020 ) . 66 . Xiong , X . et al . Acid Sphingomyelinase regulates the localization and traf \ufb01 cking of palmitoylated proteins . Biol . Open 8 , bio040311 ( 2019 ) . 67 . Bollinger , C . R . , Teichgr\u00e4ber , V . & Gulbins , E . Ceramide - enriched membrane domains . Biochim . Biophys . Acta 1746 , 284 \u2013 294 ( 2005 ) . 68 . Orchard , R . C . , Wilen , C . B . & Virgin , H . W . Sphingolipidbiosynthesis induces a conformational change in the murine norovirus receptor and facilitates viral infection . Nat . Microbiol . 3 , 1109 \u2013 1114 ( 2018 ) . Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 13 69 . Barondes , S . H . et al . Galectins : a family of animal beta - galactoside - binding lectins . Cell 76 , 597 \u2013 598 ( 1994 ) . 70 . Stowell , S . R . et al . Galectin - 1 , \u2212 2 , and \u2212 3 exhibit differential recognition of sialylated glycans and blood group antigens . J . Biol . Chem . 283 , 10109 \u2013 10123 ( 2008 ) . 71 . Burgert , A . et al . Characterization of plasma membrane ceramides by super - resolution microscopy . Angew . Chem . Int . Ed . Engl . 56 , 6131 \u2013 6135 ( 2017 ) . 72 . Nganga , R . et al . Receptor - interacting Ser / Thr kinase 1 ( RIPK1 ) and myosin - dependent ceramidosomes form membrane pores that mediate blebbing and necroptosis . J . Biol . Chem . 294 , 502 \u2013 519 ( 2019 ) . 73 . Criglar , J . M . et al . A genetically engineered rotavirus NSP2 phos - phorylation mutantimpaired in viroplasm formation andreplication shows an early interaction between vNSP2 and cellular lipid dro - plets . J . Virol . 94 , e00972 \u2013 00920 ( 2020 ) . 74 . Hutson , A . M . , Atmar , R . L . , Marcus , D . M . & Estes , M . K . Norwalk virus - like particle hemagglutination by binding to h histo - blood group antigens . J . Virol . 77 , 405 \u2013 415 ( 2003 ) . Acknowledgements This research was supported by National Institutes of Health Grant P01 AI57788 , P30DK56338thatsupportstheTexasMedicalCenterDigestive Diseases Center and the Texas Children \u2019 s Hospital EM core , T32 AI055413 ( to V . R . T . ) , S10 OD028480 that supported purchasing the Zeiss Laser Scanning Microscope LSM 980 with Airyscan 2 and the RobertWelchFoundationQ1279 . Theauthorswouldliketoacknowledge the Advanced Technology Core Laboratories ( Baylor College of Medi - cine ) , speci \ufb01 cally the Integrated Microscopy Corewith funding from the NIH ( DK56338 , CA125123 , ES030285 ) , and CPRIT ( RP150578 , RP170719 ) . We thank Xiaomin Yu and Hannah Johnson for technical assistance with human intestinal enteroids and imaging and Drs . Sasirekha Ramani , Joseph Hyser , and Jeanette Criglar for their helpful suggestions . Author contributions B . V . A . and M . K . E . conceived the idea and designed the research ; B . V . A . , K . E . , W . S . , U . C . K . , F . H . N . , and V . R . T . performed experiments ; B . V . A . , K . E . , W . S . , U . C . K . , V . R . T , S . E . C . , B . V . V . P . , R . L . A . and M . K . E . analyzed data ; R . L . A . helped with statistical analysis ; E . B . contributed new reagent ; B . V . A . andM . K . E . wrote thepaper . All authors contributedtomanuscript revision and editing . Competing interests M . K . E . is named as an inventor on patents related to cloning of the Norwalkvirus genomeandHuNoVcultivationandhasreceivedresearch funding from Takeda Vaccines Business Unit ( Cambridge , MA , USA ) . R . L . A . is named as an inventor on patents related to HuNoV cultivation and has received research support from Takeda Vaccines Business Unit ( Cambridge , MA , USA ) . Thefunders hadnorolein thestudy design ; data collection , analyses , or interpretation ; manuscriptwriting , or decision to publish the results so there are no competing interests . The remaining authors declare no competing interests . Additional information Supplementary information The online version contains supplementary material available at https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z . Correspondence and requests for materials should be addressed to Mary K . Estes . Peer review information Nature Communications thanks the anon - ymous reviewer ( s ) for their contribution to the peer review of this work . Reprints and permissions information is available at http : / / www . nature . com / reprints Publisher \u2019 s note Springer Nature remains neutral with regard to jur - isdictional claims in published maps and institutional af \ufb01 liations . Open Access This article is licensed under a Creative Commons Attribution 4 . 0 International License , which permits use , sharing , adaptation , distribution and reproduction in any medium or format , as long as you give appropriate credit to the original author ( s ) and the source , provide a link to the Creative Commons license , and indicate if changes were made . The images or other third party material in this article are included in the article \u2019 s Creative Commons license , unless indicated otherwise in a credit line to the material . If material is not included in the article \u2019 s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use , you will need to obtain permission directly from the copyright holder . To view a copy of this license , visit http : / / creativecommons . org / licenses / by / 4 . 0 / . \u00a9 The Author ( s ) 2023 Article https : / / doi . org / 10 . 1038 / s41467 - 023 - 36398 - z Nature Communications | ( 2023 ) 14 : 1148 14",
    "vasconcelosEffectExplicitInstructions2018": "Artificial Intelligence for Engineering Design , Analysis and Manufacturing cambridge . org / aie Research Article Cite this article : Vasconcelos LA , Neroni MA , Crilly N ( 2018 ) . The effect of explicit instructions in idea generation studies . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 32 , 308 \u2013 320 . https : / / doi . org / 10 . 1017 / S0890060417000658 Received : 28 February 2017 Revised : 28 October 2017 Accepted : 29 October 2017 Key words : Design fixation ; experimental instructions ; feature repetition ; inspiration ; stimuli introduction Author for correspondence : Luis A . Vasconcelos , E - mail : lalv4401 @ gmail . com \u00a9 Cambridge University Press 2018 The effect of explicit instructions in idea generation studies Luis A . Vasconcelos , Maria A . Neroni and Nathan Crilly Engineering Design Centre , Department of Engineering , University of Cambridge , Trumpington Street , Cambridge CB2 1PZ , UK Abstract In inspiration and fixation experiments , example designs are often provided along with the instructions for how participants should treat them . However , research has not reached a con - sensus about the influence of such instructions , leading to difficulties in understanding how the examples and the instructions each affect idea generation . We conducted an experiment in which 303 participants designed for the same design problem , while given different examples and instructions , which ranged from strongly encouraging copying the examples to strongly discouraging copying . Exposure to the examples affected the number and type of ideas gen - erated , whereas exposure to the instructions did not . However , instructions did affect how par - ticipants incorporated features of the examples in their ideas . Encouraged groups incorporated many features of the examples , while also incorporating structural features more than concep - tual ones . Surprisingly , the incorporation of features in discouraged groups was not different from that of groups given no instructions or even no stimulus . This indicates that concrete features may be easier to recognize and reproduce than abstract ones , and that encouraging instructions are more effective than discouraging ones , despite how strict or lenient those instructions are . The manipulation of different features also allowed us to observe how similar approaches to solving a design problem can compete for attention and how the calculation of feature repetition can be misleading depending on how common or obvious the features might be . These findings have implications for the interpretation of results from fixation stud - ies , and for the development of design tools that present stimuli to assist idea generation . Introduction Inspiration is vital to creative design and innovation ; yet , we know very little about it . Understanding more about inspiration and other cognitive aspects of design allows us to improve design processes and outcomes . It also allows us to develop intelligent tools to assist innovative design and computational models of creativity . The potential benefits of under - standing inspiration have driven the design research community to conduct many studies that seek to find out , for instance , what materials inspire designers ( Gon\u00e7alves et al . , 2014 ) , how designers achieve inspiration ( Zhao , 2013 ) , how inspiration can improve designers \u2019 per - formance ( Eckert , 1997 ) , and how to better develop computational tools to inspire designers ( T\u00f6re Yargin & Crilly , 2015 ) . Many of these studies have observed the use of external stimuli during idea generation and have reported that while stimuli can assist idea generation , they can also hinder it . One of these hindrances to creativity is described in the design literature as \u201c design fixation \u201d ( Jansson & Smith , 1991 ) , when prior knowledge of a solution causes designers to inadvertently limit their search for other solutions . Design fixation was originally studied in situations where designers were trying to address a problem and were provided with stimuli which represented possible solutions to the problem that was being addressed ( Jansson & Smith , 1991 ) . Fixation was measured by counting the solutions that were developed in response to the problem and counting the repetition of fea - tures from the stimuli . This general approach to measuring fixation has subsequently been employed in several other related studies ( e . g . , Purcell & Gero , 1996 ; Dahl & Moreau , 2002 ; Linsey et al . , 2010 ; Cardoso & Badke - Schaub , 2011 ; Youmans , 2011 a , b ; Vasconcelos et al . , 2017 ) . Although reproducing features of good designs might be beneficial or efficient , even flawed stimuli ( i . e . , stimuli with negative features ) are reproduced , when it might instead be expected that those stimuli would be identified and avoided ( Jansson & Smith , 1991 ; Linsey et al . , 2010 ; Youmans , 2011 b ) . In general , fixation research suggests that blindly copying features from stimuli is harmful to idea generation , and therefore to the design process itself . As such , some studies have incor - porated constraining textual instructions into the stimuli given to participants , to prevent them from copying their features . However , the efficacy of these instructions has varied across stud - ies : they were effective in some cases ( Chrysikou & Weisberg , 2005 ; Yilmaz et al . , 2010 ) and ineffective in others ( Jansson & Smith , 1991 ; Perttula & Sipil\u00e4 , 2007 ) . Thus , it is still uncertain https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press how textual instructions can influence the copying behavior , and this uncertainty might be attributed to different factors . For instance , individuals may tend to overlook the instructions that accompany the stimuli ( LeFevre & Dixon , 1986 ) , they may inter - pret the instructions in different ways if instructions are not explicit , or the instructions may lead them to question why they are being exposed to the stimuli and change their idea generation process accordingly . Whatever the reason might be , it is impor - tant to understand the relevance of the instructions provided to designers as part of the inspiration material . Methodologically , this would help in determining how fixation studies should be conducted and the results interpreted . More practically , it would also help in understanding how stimuli should be presented when designers are stimulated with examples , such as in those software tools that support idea generation by providing external stimuli [ for examples of such tools , see ( Chakrabarti et al . , 2005 ; Linsey & Wood , 2007 ; Vattam et al . , 2011 ) ] . To better understand the influence of instructions on idea gen - eration , we conducted an experiment in which participants responded to the same design problem , while being exposed to different example solutions , and following different instructions with respect to those examples . Inspiration , fixation , and the introduction of external stimuli Design researchers have studied many different aspects of the inspiration process , including characteristics of external stimuli or inspiration sources ( e . g . , the modality of representation used for the stimuli ) ( Linsey et al . , 2008 ; Sarkar & Chakrabarti , 2008 ; Atilola & Linsey , 2015 ; Viswanathan et al . , 2016 ) and aspects of the design process ( e . g . , time available for ideation ) ( Youmans , 2011 a ; Tsenn et al . , 2014 ; Siangliulue et al . , 2015 ) . One of the stimuli characteristics that has been studied is whether the quality of the examples ( i . e . , good or bad ) would influence feature repeti - tion , or whether those examples would be copied indiscrimi - nately . A series of studies have found that participants still repeat features from previous examples even when they are flawed ( Jansson & Smith , 1991 ; Chrysikou & Weisberg , 2005 ; Perttula & Liikkanen , 2006 ; Fu et al . , 2010 ; Linsey et al . , 2010 ; Youmans , 2011 b ; Viswanathan et al . , 2014 ) . It should be noted that in Jansson and Smith \u2019 s ( 1991 ) original paper , feature repetition was a way in which the researchers operationalized the degree to which the participants \u2019 exploration of the solution space was limited . Furthermore , the repetition of negative features ( or fea - tures that contradicted the problem brief ) was taken to be an indi - cation that this limited search was an unconscious behavior ( Youmans & Arciszewski , 2014 ) . To counteract the indiscriminate repetition of features , some studies have tried warning participants about the flaws in the examples ( Jansson & Smith , 1991 ; Chrysikou & Weisberg , 2005 ; Viswanathan et al . , 2014 ) , while others have tried instructing participants not to copy the examples ( Smith et al . , 1993 ; Perttula & Sipil\u00e4 , 2007 ; Yilmaz et al . , 2010 ) . Considering these two approaches , Chrysikou and Weisberg ( 2005 ) found that warning participants of flaws in the examples was not enough ; they had to be told to avoid repeating those flaws . Yilmaz et al . ( 2010 ) also told participants not to reproduce the examples and this led to a reduction in feature repetition . Conversely , Jansson and Smith ( 1991 ) and Perttula and Sipil\u00e4 ( 2007 ) found repetition even when participants were instructed to avoid using features of the examples provided . In summary , the published literature reports conflicting results about the influence of instructions when providing external stimuli to designers . While many variables have been manipulated in fixation experiments , and some studies have already tested the effective - ness of using textual instructions to some extent , the way the stim - uli are introduced in such experiments has not yet been studied systematically . Such stimuli introductions can typically be divided into two components : a descriptive statement on what the stimu - lus is ( e . g . , \u201c here is an example solution \u201d ) and a prescriptive instruction for how the stimulus should be used ( e . g . , \u201c don \u2019 t copy its features \u201d ) . Currently , the stimuli introductions ( i . e . , descriptions and instructions ) given to participants vary from study to study , with some reporting that the example is there to help , to illustrate previous solutions , or to show how ideas should be represented , while other studies only mention that the stimulus was provided to participants without describing how ( Vasconcelos & Crilly , 2016 ) . See Table 1 for a more detailed description of stimuli introductions as described in related studies . Variation in the way the stimuli are introduced might be attributed to a lack of agreement across studies about which \u201c real - world \u201d situation is being simulated ( e . g . , contexts in which exam - ples are accidentally seen , intentionally searched for , or already known ) . Regardless of the reason , the variation in the way copying is encouraged or discouraged makes it difficult to compare results across studies and to know the extent to which the design work is being affected by instructions . Although previous fixation research makes it hard to formulate a theory for the influence of instructions , there is one study that looked more carefully at how different sets of instructions affected feature repetition during idea generation ( Smith et al . , 1993 ) . In the experiment , participants were asked to either conform to or diverge from the example solutions provided in a creative task . Some participants ( conform group ) were told that the examples were great ideas previously created for that task and that partici - pants should create ideas like those while not copying them exactly . Other participants ( diverge group ) were told that the examples restricted people \u2019 s creativity and that participants should create ideas that were different from the examples . Finally , other participants ( standard group ) were given the same examples but without any specific instructions , and a baseline group ( BG ) was given no examples at all . When compared with the participants in the BG , the researchers found that all other groups generated more ideas containing features of the examples . Additionally , feature repetition in the diverge group did not decrease when compared with the standard group , while feature repetition in the conform group increased when compared with the standard group . Based on their results , the researchers proposed that the partic - ipants did not conform to examples because they had assumed that they should ; they conformed because they could not forget the examples that they had seen . However , the instructions used in the experiment were not strict ( i . e . , they suggested how ideas should be created , but neither forbade nor required the use of fea - tures from the examples , thus allowing participants to interpret the instructions in different ways ) . The description of the stimuli and the instruction for their use also varied between groups , mak - ing it difficult to infer the influence of each piece of information in isolation . Additionally , the experiment reports an analysis on the repetition of features that were always concrete or structural ( e . g . , parts of a physical object ) , thus inviting further investigation to be done on different types of features , such as more abstract or conceptual features ( e . g . , properties of a physical object ) . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 309 https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press We believe that a different experimental setup with additional manipulations can provide further evidence to clarify the influ - ence of the textual instructions used in experimental research , thus having implications for design research methodology and for professional design practice . To investigate this , we designed and conducted a study to explore the effects of instructions in design inspiration and fixation experiments , with instructions that are ( slightly or very ) discouraging or encouraging , or that are neutral . Research methodology We conducted a single experiment ( N = 303 ) designed to examine the role of instructions in creativity studies . However , while the data were captured in a single session ( thus minimizing experi - mental variation and allowing for efficient data capture ) , the experimental manipulation and metrics were divided into two main blocks , and the corresponding analyses were intended to be performed over two different time frames . Thus , we here treat all the data as relating to two separate studies ; the details and results of which are discussed as follows . The experimental setup of study 1 is described in full , whereas study 2 follows up on that with shorter descriptions , in which only the main differ - ences from the previous study are highlighted . Both studies drew participants from the same population ( e . g . , educational back - ground ) , and used identical procedures ( e . g . , time available ) , materials ( e . g . , design problem ) , analysis ( e . g . , fixation metrics ) , and environment ( e . g . , room conditions ) . However , the stimulus manipulated in study 2 incorporated a distinct conceptual feature , thus being significantly different from that used in study 1 . Objective and hypothesis This experiment investigates how textual instructions accompany - ing external stimuli can shape the design work of the participants . We hypothesize that instructions for the incorporation of features provided along with an example design will determine the repeti - tion of the ideas that participants generate . In particular , we expect that positive or encouraging instructions will increase fea - ture repetition , whereas negative or discouraging instructions will reduce feature repetition . Additionally , we also expect to find some level of fixation in all groups exposed to the examples when compared with a BG . From an experimental perspective , we understand design fixation to be the process of generating ideas in response to a design problem under the influence of one or more external stimuli , in which the generation of ideas is constrained due to the repetition of the stimuli into the ideas generated . This would imply decreased productivity of the partic - ipants as well as increased idea repetition . Study 1 Experimental method Participants were randomly allocated to different experimental conditions . They were verbally asked to be creative and to gener - ate , individually , as many ideas as possible to a given problem . They were also instructed , via voice and text , to sketch and describe in writing their ideas on sheets of paper . Except for a BG that designed without any stimuli , all other experimental groups received a sketch of one example solution and a descrip - tion of what the sketch represented . These stimulated groups ( SGs ) received instructions for the use of features from the exam - ple , and these instructions varied with respect to how constraining or encouraging they were ( see the \u201c Materials \u201d section for the com - plete instructions ) . Finally , the participants \u2019 ideas were assessed to evaluate the influence of the instructions on the level of fixation observed . Participants One hundred and sixty - eight engineering students in their first year at the University of Cambridge ( UK ) were assigned to six experimental groups ( n = 28 ) . Participation in the experiment was part of the students \u2019 education ( but was not for credit ) , and was aimed at collecting data that could later be used to introduce Table 1 . Different means of introducing external stimuli to participants ( each instance was extracted from a fixation - related study ; changes to the original text were made to permit better comparison across rows ) Study Stimuli introduction ( or descriptive statement about the stimuli ) ( Cheng et al . , 2014 ) The examples should be analyzed using a set of instructions ( Dahl & Moreau , 2002 ) The example is provided to help participants get started and it might be useful in solving the design problem ( Gon\u00e7alves et al . , 2012 ) The example could be considered ( or not ) when participants were generating ideas ( Liikkanen & Perttula , 2008 ) The example should be used to raise thoughts ( Linsey et al . , 2010 ) The example should be considered as a solution that might be created for the design problem ( Lujun , 2011 ) The example could be referred to during ideation ( Perttula & Sipil\u00e4 , 2007 ) The examples should be used to awaken thoughts , but should not be reproduced as such ( Purcell & Gero , 1996 ) The example design illustrates what is meant by a sketch design ( Vasconcelos et al . , 2016 ) The example shows how participants should present their ideas ( Vasconcelos et al . , 2017 ) The example is a concept that illustrates one way to solve this problem ( Yilmaz et al . , 2010 ) The examples could be used to understand the method , but they should not be repeated in participants \u2019 own designs ( Youmans , 2011 a ) The example illustrates a previous design for the same task ( Youmans , 2011 b ) The example should be rebuilt as to allow familiarization with the materials in a construction set 310 Luis A . Vasconcelos et al . https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press them to the concept of design fixation . No demographic data were collected from the participants , but as first - year undergraduate students , they were broadly similar in age and design experience , drawn from a cohort with a male \u2013 female ratio of approximately 3 : 1 . Task and problem The participants were told to solve the following problem . \u201c Bicycles are a popular mode of transportation and recreation for many people . While growing up , a person might go through a series of ever - larger bikes , sometimes having several models , one after the other . However , having several bikes can be a problem for many reasons . Your task is to generate as many ideas as possi - ble to eliminate the need to have multiple bikes as people grow up . \u201d This problem was selected because it was expected to satisfy four criteria . First , it was unlikely that the participants had designed solutions for this problem before . Second , they were likely to have experienced the situation described in the problem previously ( i . e . , while growing up , they probably had multiple bikes ) , thus helping their understanding of it . Third , the problem could be solved in different ways , with several different underlying principles being applied , thus leaving enough room for creativity . Finally , the design brief held a low level of complexity , which was also expected for the ideas generated , thus being suitable for a quick experiment fitting with the constraints of the session . Procedure ( overview ) The experiment took place in a large lecture theatre with all the participants present . During the first 5 min , the participants lis - tened to an oral explanation about the activities to follow and received all the material they needed . Participants in the five SGs received the design brief , the external stimulus and instruc - tions , and blank sheets of paper , while participants in the BG received the brief without any stimulus . Then , the participants were asked to think of ideas for 3 min without committing any designs to paper ( because different participants had different materials and content , this ensured they all had enough time to read all the materials and start developing some ideas ) . Finally , for the remaining 10 min , all participants individually generated as many ideas as possible in silence , ideally including both a sketch and a written description of each idea . Materials All participants received the same design problem written on an A4 sheet , as well as blank A4 sheets to sketch and annotate their own ideas . Except for the participants in the BG , all partic - ipants received one additional sheet with an example solution , that is , an annotated sketch of a bike ( Fig . 1 ) . The example solution was preceded with the written descrip - tion : \u201c Below is an example of how you should present your ideas ( i . e . , annotated sketches ) \u201d . This description was either immedi - ately followed by an instruction regarding the use of features from the examples ( discouraging or encouraging ) or by no instruction whatsoever . The instructions for the different experi - mental groups are listed below against a code for each experi - mental group . (cid:129) SG \u2212 2 ( strictly forbidding ) : \u201c make sure you do not use features from this example in your own work \u201d ; (cid:129) SG \u2212 1 ( discouraging ) : \u201c avoid using features from this example in your own work \u201d ; (cid:129) SG0 ( neutral ) : no instruction was given ; (cid:129) SG + 1 ( encouraging ) : \u201c consider using features from this example in your own work \u201d ; (cid:129) SG + 2 ( strictly requiring ) : \u201c make sure you use features from this example in your own work \u201d . Analysis The assessment of the participants \u2019 ideas was conducted by three research collaborators , with backgrounds in design research , com - plexity science , and mechanical engineering . Initially , the three evaluators agreed on the metrics to be included in the analysis . Such metrics should be consistent with most of the existing design fixation literature and should be appropriate for testing our hypothesis . After that , the three evaluators analyzed the design work of a random experimental group together to agree on the assessment method , ultimately reaching a consensus with respect to how to interpret and assess the ideas . Finally , each of the eva - luators , who were blind to the experimental groups that they were rating , individually judged a random subset of the remaining ideas . To reduce the chances of mistakes , if any evaluator had trouble judging an idea , this idea was then discussed collectively . We considered \u201c one idea \u201d either to be a sketch or a written description ( usually both ) that presented an understandable way to solve the problem . Participants often generated more than one idea , but in some cases , all ideas could be considered as one , particularly when the idea was a bike . For instance , if the ideas could all be incorporated onto the same bike without inter - ference , then they were considered as a single idea . Conversely , if there were two or more ideas for the same bike component ( e . g . , frame , wheel , handlebar ) , then they were considered to be distinct ideas . The metrics used in the assessment were idea \u201c fluency \u201d and idea \u201c repetition \u201d . These metrics are central to many fixation stud - ies ( e . g . , Jansson & Smith , 1991 ; Purcell & Gero , 1996 ; Dahl & Moreau , 2002 ; Linsey et al . , 2010 ; Cardoso & Badke - Schaub , 2011 ; Youmans , 2011 a , b ; Vasconcelos et al . , 2017 ) and they can provide a fairly objective measurement ( i . e . , a direct count Fig . 1 . Example solution provided to the participants along with the following description : \u201c A modular bike with parts of various sizes that can be connected and swapped to fit people with very different heights . Apart from the socketing parts and expansible / contractible wheels , the angles between tubes can also be modified in specific joints . \u201d The sketch used is a modification of the ECO 07 Compactable Urban Bicycle ( Aleman , 2009 ) . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 311 https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press of ideas and features ) that is consistent with testing our feature repetition hypothesis . Therefore , other design fixation metrics found in the literature , such as diversity and conceptual distance , were not included into the analysis because they are more subjec - tive ( i . e . , increased judgment is required on the part of assessors ) and are not suited to testing our research hypothesis . Because test - ing our hypothesis implied identifying and counting ideas and features without any further judgment , we did not require dupli - cate assessment from our evaluators . Idea fluency is the total number of ideas generated , also called \u201c quantity \u201d ( Shah et al . , 2003 ) or \u201c productivity \u201d ( Nijstad et al . , 2002 ) elsewhere . Idea repetition might occur at different levels ; for instance , the repetition of idea types , conceptual features , or structural features ( although other categorizations and granularity levels might also be identified ) . With respect to the idea type , we divided the ideas into two broad categories : bike ideas and non - bike ideas ; thus , by designing a bike , the participant would be repeating the idea type . With respect to their conceptual features , we also divided the ideas into two categories : modular ideas or non - modular ideas ; thus , by generating a modular idea , the par - ticipant would be repeating the conceptual feature . Finally , we examined the incorporation of structural features in the partici - pants \u2019 ideas , in terms of both the number of ideas containing any feature of the example ( i . e . , Is repetition there ? ) and the num - ber of example features present in each idea ( i . e . , How much repe - tition is there ? ) . These features were intentionally included in the example design to permit a measure of fixation inferred because of repetition . There were five structural features : swappable com - ponents to change bike size ; frame joints ( lugs ) that act as sockets for the tubes ; wheels with bendable spokes ; an hourglass - shaped frame ; and a saddle that cannot be adjusted in height directly . Eight participants ( 4 . 8 % ) either did not generate any idea or generated ideas that could not be interpreted by the evaluators ; the results from such participants were not included in our anal - ysis . The adjusted number of participants per experimental group is used in the following section . Results and discussion To deal with non - normality in the data , we used analysis of var - iance ( ANOVA ) with significance values estimated using 1000 bootstrap resamples and planned contrasts ( a non - parametric ver - sion of the regular ANOVA test ) . To identify differences between groups , we used five planned contrasts in the analysis . The first contrast compares the BG to all SGs , and aims to identify any effect from exposure to the example design ( the fixation test ) . The remaining contrasts incorporate only the SGs , and aim to identify any effect from instructions received as follows . The sec - ond contrast compares the discouraged and neutral groups ( the discouraging instructions test ) ; the third compares the encouraged and neutral groups ( the encouraging instructions test ) ; the fourth compares the two discouraged groups ( the discouraging subtlety test ) ; and the fifth contrast compares the two encouraged groups ( the encouraging subtlety test ) . Idea fluency Instructions had no effect on the number of ideas generated . A one - way ANOVA with the total number of ideas ( per partici - pant ) as the dependent variable showed a significant difference across the groups , F ( 5 , 157 ) = 4 . 37 , p = 0 . 001 , \u03b7 2 = 0 . 122 . However , planned contrasts revealed that this difference is explained by the BG generating on average more ideas than the SGs , t ( 32 . 7 ) = 3 . 64 , p = 0 . 002 , d = 0 . 840 , since no difference was found between the SGs . Table 2 shows summary statistics for these results . These results reveal that idea fluency was not influenced by how constraining or encouraging the instructions were . However , the presence of an example design affected idea genera - tion : designing without exposure to stimuli resulted in more ideas being generated , which we interpret as a beneficial isolation from examples ( Vasconcelos et al . , 2017 ) . This effect is consistent with other studies in which seeing an example caused reduction in the idea fluency ( Linsey et al . , 2010 ) , although studies have also reported an increase in the idea fluency as a result of external stimulation ( Purcell & Gero , 1992 ) or even no effect at all ( Jansson & Smith , 1991 ) . More importantly , our results are differ - ent from those found in ( Smith et al . , 1993 ) , where the number of ideas generated by a BG did not exceed that of SGs . However , we should mention that while all groups in our study were asked to present their ideas with both sketches and textual descriptions , the BG produced many ideas that were presented only in text . As a result , baseline participants spent less time sketching every idea , which allowed them more time to produce a greater number of less - elaborated ideas , as explained by the \u201c normative represen - tation effect \u201d ( Vasconcelos et al . , 2016 ) . Repetition of the idea type Instructions had no effect on the number of bike ideas generated . A one - way ANOVA with the average number of bike ideas ( per participant ) as the dependent variable showed a significant differ - ence across the groups , F ( 5 , 157 ) = 2 . 87 , p = 0 . 017 , \u03b7 2 = 0 . 084 . However , planned contrasts revealed that this difference is explained by participants in the BG generating , on average , fewer bike ideas than the SGs , t ( 31 . 4 ) = \u2212 2 . 59 , p = 0 . 025 , d = 0 . 613 , since no difference was found between the SGs . Table 3 shows summary statistics for these results . If we adopt a \u201c bike \u201d and \u201c non - bike \u201d categorization of the ideas generated , these results reveal that the type of idea generated was not influenced by the instructions . This can be seen as a surpris - ing result : by manipulating how participants were instructed to incorporate features of the example , we also anticipated an indi - rect effect on the type of ideas that they might have generated ( because all features were incorporated into a bike idea ) , but Table 2 . Summary of ideas generated per participant across groups Generated ideas BG SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 Mean ( and SD ) for the number of ideas per participant 2 . 39 ( 1 . 31 ) 1 . 50 ( 1 . 27 ) 1 . 30 ( 0 . 61 ) 1 . 48 ( 0 . 49 ) 1 . 41 ( 0 . 56 ) 1 . 54 ( 1 . 10 ) Range of ideas generated per participant 1 \u2013 5 1 \u2013 6 1 \u2013 3 1 \u2013 4 1 \u2013 4 1 \u2013 6 Total number of ideas 67 39 35 40 38 43 Total number of participants 28 26 27 27 28 28 312 Luis A . Vasconcelos et al . https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press this was not the case . Conversely , while the total number of bike ideas was roughly the same for all groups , the proportion of bike ideas generated per participant across all groups was significantly different . In fact , only 60 % of all ideas generated by the BG were bikes , whereas the SGs had a much greater proportion of bike ideas ( 89 % on average ) . In summary , bike ideas were likely to be generated irrespective of the experimental condition , but not seeing the example allowed participants from the BG to explore different areas of the solution space , again confirming the beneficial isolation effect . Accordingly , design fixation ( inferred because of repetition of the idea type ) occurred across the stimulated conditions , a result that is broadly consistent with other studies in which seeing an example caused participants to conform to certain types of solu - tions , thus reducing the breadth of ideas ( Jansson & Smith , 1991 ; Linsey et al . , 2010 ; Cardoso & Badke - Schaub , 2011 ) . Repetition of the conceptual feature Instructions had no effect on the number of modular ideas gen - erated . A one - way ANOVA with the average number of modular ideas ( per participant ) as the dependent variable showed no sig - nificant difference across the groups , F ( 5 , 157 ) = 1 . 86 , p = 0 . 104 , \u03b7 2 = 0 . 056 . Table 4 shows summary statistics for these results . If we adopt a \u201c modular \u201d and \u201c non - modular \u201d categorization of the ideas , these results reveal that the type of idea generated was not influenced by the instructions . In fact , modular ideas were extremely rare across all groups . Still , the fixation literature tells us that the SGs would generate more modular ideas , an effect sim - ilar to the repetition of the idea type ( i . e . , bikes ) . In particular , participants who were encouraged to use features from the exam - ple in their own work ( and modularity was a visible feature on the example provided ) did not act accordingly , and participants who were discouraged from copying features generated as many mod - ular ideas as the other groups . The fact that encouraged groups did not produce more modular ideas contradicts previous research ( Smith et al . , 1993 ) , where results pointed to an increase in feature repetition in encouraged groups . Additionally , previous research has shown that abstract , conceptual features can fixate designers in a similar way to concrete , structural features ( Vasconcelos et al . , 2017 ) . However , as there was no difference in the proportion of modular ideas between baseline and SGs , we speculate that the idea of modularity ( included in the example as a conceptual feature ) was not obvious enough to induce fixa - tion effects ( inferred because of repetition of the conceptual feature ) . Repetition of structural features Instructions had a significant effect on the number of ideas with features of the example provided . A one - way ANOVA with the average number of ideas that contained structural features of the example ( per participant ) as the dependent variable showed a significant difference between the groups , F ( 5 , 157 ) = 9 . 50 , p = 0 . 000 , \u03b7 2 = 0 . 232 ; a result that was also found for the average number of repeated features per participant , F ( 5 , 157 ) = 14 . 115 , p = 0 . 000 , \u03b7 2 = 0 . 310 . For the number of ideas containing features of the example solution , planned contrasts revealed a significant difference between the two encouraged and the neutral groups , t ( 76 . 0 ) = \u2212 5 . 33 , p = 0 . 001 , d = 1 . 14 , with the encouraged groups generating more ideas with features from the example . For the average number of features included in the participants \u2019 ideas , planned contrasts showed a marginally significant difference between the BG and all SGs , t ( 44 . 6 ) = \u2212 1 . 79 , p = 0 . 086 , d = 0 . 338 , and again , between the two encouraged and the neutral groups , t ( 71 . 0 ) = \u2212 6 . 47 , p = 0 . 001 , d = 1 . 29 , with the SGs repeating on aver - age more features than the baseline condition due to an increased feature repetition of the encouraged groups . Table 5 shows sum - mary statistics for these results . These results reveal that encouraging instructions influenced the incorporation of structural features from the example . On average , participants in encouraged groups incorporated more example features per idea and generated more ideas that incorpo - rated such features . Additionally , participants in discouraged groups produced results similar to the neutral group and the BG . This result is inconsistent with previous research from Chrysikou and Weisberg ( 2005 ) , as constraining instructions did not decrease repetition , but consistent with Smith et al . ( 1993 ) , as encouraging instructions did increase repetition . However , contrary to the results from Smith et al . ( 1993 ) , our dis - couraged ( but still stimulated ) groups did not significantly repeat more features than a BG , contradicting the idea that participants could not forget the example they had seen . It seems that a few structural features were naturally likely to be incorporated into the ideas generated ( supported by the similar results from the non - encouraged groups ) , with encouraged groups incorporating more features because they were instructed to do so . When com - paring the results from the repetition of structural features to the repetition of the conceptual feature used in this study , we can infer that concrete structural features are more easily copied than abstract conceptual features from examples [ similar to what has been suggested in previous studies ( Zahner et al . , 2010 ; Cheong et al . , 2014 ; Feng et al . , 2014 ) ] . This is despite Table 3 . Summary of bike and non - bike ideas generated across groups Bike ideas BG SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 Mean ( and SD ) for the number of bike ideas per participant 1 . 43 ( 0 . 84 ) 1 . 27 ( 0 . 92 ) 1 . 15 ( 0 . 36 ) 1 . 33 ( 0 . 83 ) 1 . 33 ( 0 . 55 ) 1 . 29 ( 0 . 53 ) Number of bike ideas ( and % ) 40 ( 60 % ) 33 ( 85 % ) 31 ( 89 % ) 36 ( 90 % ) 36 ( 95 % ) 36 ( 84 % ) Number of non - bike ideas ( and % ) 27 ( 40 % ) 6 ( 15 % ) 4 ( 11 % ) 4 ( 10 % ) 2 ( 5 % ) 7 ( 16 % ) Table 4 . Summary of modular and non - modular ideas generated across groups Modular ideas BG SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 Mean ( and SD ) for the number of modular ideas per participant 0 . 43 ( 0 . 50 ) 0 . 08 ( 0 . 27 ) 0 . 3 ( 0 . 47 ) 0 . 19 ( 0 . 40 ) 0 . 19 ( 0 . 40 ) 0 . 18 ( 0 . 39 ) Number of modular ideas ( and % ) 12 ( 18 % ) 2 ( 5 % ) 8 ( 23 % ) 5 ( 13 % ) 5 ( 13 % ) 5 ( 12 % ) Number of non - modular ideas ( and % ) 55 ( 82 % ) 37 ( 95 % ) 27 ( 77 % ) 35 ( 87 % ) 33 ( 87 % ) 38 ( 88 % ) Artificial Intelligence for Engineering Design , Analysis and Manufacturing 313 https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press the opportunity that conceptual features offer for inspiring solu - tions across multiple categories of solution ( e . g . , products , ser - vices ) . Alternatively , it is possible that either the conceptual feature was chosen or its incorporation into the example was not effective and thus not fully comprehended by the participants . Again , we should highlight that the BG produced more ideas but proportionally less sketches than the SGs . This puts the SGs in an unfavorable position with respect to the count of structural fea - tures incorporated , since these groups had to represent a shape of the bike in which repetition could be more easily recognized \u2013 it is difficult to identify structural repetition when the idea is represented only by text \u2013 thus possibly biasing the results . General discussion and limitations In summary , except for the repetition of structural features , the instructions manipulation had no effect on participants \u2019 idea gen - eration according to the creativity metrics used in this study . Idea fluency and the repetition of the idea type were only affected by the exposure to the example design , and the repetition of the con - ceptual feature did not seem to be affected by either instructions or exposure to the example . This last set of results conflicts with a previous study that found that a pictorial representation of mod - ularity ( as a conceptual and abstract feature ) induced fixation effects in a similar way to a more concrete bike example ( Vasconcelos et al . , 2017 ) . Additionally , looking more carefully into our data reveals that the generation of modular ideas hap - pened quite rarely and randomly across groups , and it is possible that the feature \u201c modularity \u201d , in particular , failed to inspire par - ticipants . These findings call for further investigation and exten - sion of this study , potentially by incorporating other stimuli . As such , to test whether our previous findings are robust to a differ - ent stimulus with different structural and conceptual features ( while keeping all other experimental variables the same ) , we now report an additional analysis done on a second dataset . Additionally , in study 1 , we did not check the inter - rater agree - ment , a test which is often performed for similar studies to dem - onstrate the reliability of the evaluators \u2019 subjective assessment when working individually . The assessment performed was rela - tively objective and yet involved many interactions between eva - luators ( thus , we expected a high agreement between them and we believe that such interaction allowed a reliable analysis of the idea ) . However , because there was no redundancy in the assessment , we cannot quantify how good this agreement might have been . As such , in study 2 , we now compute and report agree - ment coefficients . Study 2 Experimental method , participants , task and problem , and procedure Studies 1 and 2 share the same experimental method and drew participants from the same population . One hundred and thirty - five participants were assigned to five experimental groups ( n = 26 ) . The task , design problem , and procedure were the same as in study 1 , except for one aspect : we did not have a new base - line condition in study 2 since we already had data from a non - stimulated population to which we could compare our new SGs . Materials To allow a neat comparison of studies 1 and 2 , we chose a similar example solution to be given to the participants . While the prob - lem could be solved in many ways and with several underlying properties being applied , it was expected that developing an \u201c adaptable \u201d solution would be a common approach in this con - text . Such an approach could equally be realized by invoking a modular concept ( e . g . , by interchangeable parts ) or a telescopic concept ( e . g . , by extendable parts ) . As a result , we opted for using a telescopic bike ( Fig . 2 ) with both sketch and a textual description of a similar quality to that used in study 1 to avoid influencing the participants \u2019 perception of the ideas due to differ - ent sketch quality ( Kudrowitz et al . , 2012 ) . The description of the example solution was the same used in study 1 [ \u201c Below is an example of how you should present your ideas ( i . e . , annotated sketches ) \u201d ] , and it was followed by the same set of discouraging , encouraging , or neutral instructions regarding the use of features from the example . Analysis The assessment of the participants \u2019 ideas was conducted by the first two authors of this study , with backgrounds in design research and experimental psychology , respectively . First , the evaluators discussed the metrics of the analysis and assessed a random sample of ideas to reach a satisfactory agreement . As in study 1 , the coding scheme used here was not very subjective \u2013 the Table 5 . Summary of modular features incorporated into the participants \u2019 ideas and ideas with modular features across groups Feature repetition BG SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 Mean ( and SD ) for the number of features incorporated per idea 0 . 36 ( 0 . 62 ) 0 . 12 ( 0 . 33 ) 0 . 30 ( 0 . 47 ) 0 . 15 ( 0 . 36 ) 1 . 00 ( 1 . 14 ) 1 . 43 ( 1 . 07 ) Range of features incorporated per participant 0 \u2013 2 0 \u2013 1 0 \u2013 1 0 \u2013 1 0 \u2013 3 0 \u2013 3 Total number of ideas with features incorporated ( and % ) 8 ( 12 % ) 3 ( 8 % ) 8 ( 21 % ) 4 ( 11 % ) 16 ( 44 % ) 24 ( 57 % ) Fig . 2 . Example solution provided to the participants along with the following description : \u201c A telescoping bike with parts that can be extended or shortened to fit people with very different heights . Apart from the adjustable tubes and wheels , the angles between tubes can also be modified in specific joints . \u201d The sketch used is a modification of the Zee - K Ergonomic Bike ( Floss , 2010 ) . 314 Luis A . Vasconcelos et al . https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press evaluators spotted the corresponding features in the drawings and reported them . However , the drawings were not always clear and the handwriting was often poor . Consequently , the assessment was prone to human error when interpreting what was contained in each idea , and differences between evaluators were more likely to be attributed to that , instead of divergence in individual opi - nions . As such , the two evaluators assessed the complete set of ideas individually to reveal errors . Such errors were identified by reviewing all ratings and looking for ideas with discrepancies in many variables , which often indicated ideas with a poor repre - sentation . After clarifying such cases , we computed Cohen \u2019 s \u03ba coefficients for each metric , as shown in Table 6 . When consider - ing the high agreement coefficients obtained here and the very brief training provided to the evaluators , it is clear that there was little subjectivity when assessing the ideas and similar coeffi - cients should be expected for study 1 . While most particularities of the assessment were the same as in study 1 , including the metrics used ( i . e . , idea fluency and idea repetition ) , the change in the stimulus required a new set of struc - tural features to be present in the example design . Again , there were five features intentionally incorporated into the example : extendable components to change bike size ; frame joints ( lugs ) that rotate to adjust angles between parts ; wheels with no spokes ; an open - shaped frame ; and a cantilevered saddle that can be adjusted directly . Five participants ( 3 . 7 % ) either did not generate any idea or generated ideas that could not be interpreted by the evaluators ; the results from such participants were not included in our anal - ysis . The adjusted number of participants per experimental group is used in the following section . Results and discussion Idea fluency Instructions had no effect on the number of ideas generated . A one - way ANOVA with the total number of ideas ( per partici - pant ) as the dependent variable showed a significant difference across the groups , F ( 5 , 152 ) = 4 . 79 , p = 0 . 000 , \u03b7 2 = 0 . 136 . However , planned contrasts revealed that this difference is explained by the BG generating on average more ideas than the SGs , t ( 32 . 4 ) = 3 . 59 , p = 0 . 002 , d = 0 . 844 , since no difference was found between the SGs ( Table 7 shows summary statistics for these results ) . This set of results matches very closely with those obtained in study 1 , confirming that the manipulation of instructions had no effect on the number of ideas generated , but exposure to an example design did . Repetition of the idea type Instructions had no effect on the number of bike ideas generated . A one - way ANOVA with the average number of bike ideas ( per participant ) as the dependent variable showed a significant difference across the groups , F ( 5 , 152 ) = 2 . 44 , p = 0 . 037 , \u03b7 2 = 0 . 074 . However , planned contrasts revealed that this difference is explained by the BG generating , on average , fewer bike ideas than the SGs , t ( 31 . 4 ) = \u2212 2 . 45 , p = 0 . 029 , d = 0 . 576 , since no differ - ence was found between the SGs ( Table 8 shows summary statis - tics for these results ) . As in study 1 , the manipulation of instructions had no effect on the number of bike ideas generated , even though the instructions had explicit directions toward the use of features from the bike example and whether participants should comply with the instructions , we expected to observe an indirect effect on the number of bike ideas being generated . Repetition of the conceptual feature Instructions had an effect on the number of telescopic ideas gen - erated . A one - way ANOVA with the average number of telescopic ideas ( per participant ) as the dependent variable showed no sig - nificant difference across the groups , F ( 5 , 152 ) = 1 . 75 , p = 0 . 126 , \u03b7 2 = 0 . 055 . However , planned contrasts revealed a significant dif - ference between the two encouraged groups and the neutral group , t ( 152 ) = \u2212 2 . 27 , p = 0 . 021 , d = 0 . 560 , with the encouraged groups generating more telescopic ideas ( Table 9 shows summary statistics for these results ) . If we adopt a \u201c telescopic \u201d and \u201c non - telescopic \u201d categorization of the ideas , these results reveal that the type of idea generated was influenced by the instructions , a finding that differs from that reported in study 1 . As discussed before , it is possible that the notion of extendibility ( included in the example as a conceptual feature ) was more recognizable than modularity , thus being more easily copied . Interestingly , if we look at the number of mod - ular ideas in study 2 , a similar contrast between the two encour - aged and the neutral groups showed a marginally significant result , t ( 41 . 0 ) = 2 . 02 , p = 0 . 053 , d = 0 . 507 , but now with the encour - aged groups generating less modular ideas ( encouraged , M = 0 . 162 , SD = 0 . 334 ; neutral , M = 0 . 347 , SD = 0 . 394 ) . A possible explanation for this is that , whilst those participants who were encouraged to reproduce features from the telescopic example increased the generation of telescopic ideas , they were also diverted or pushed away from the generation of modular ideas ( these two approaches were so commonly adopted that together they represented 46 % of the ideas from the BG ) . Though research has already reported the occurrence of such phenomena ( Vasconcelos et al . , 2017 ) , this last set of results should be inter - preted with caution since they are derived after the formal analysis had been made , rather than being hypothesized and planned prior to the analysis . Repetition of structural features Instructions had a marginally significant effect on the number of ideas with features of the example provided . A one - way ANOVA with the average number of ideas that contained structural fea - tures of the example ( per participant ) as the dependent variable showed a marginally significant difference across the groups , F ( 5 , 152 ) = 2 . 01 , p = 0 . 081 , \u03b7 2 = 0 . 062 , and a significant difference for the average number of repeated features per participant , F ( 5 , 152 ) = 2 . 73 , p = 0 . 021 , \u03b7 2 = 0 . 082 . Planned contrasts revealed Table 6 . Ideation metrics used in this study and their computed agreement coefficients and interpretation Metric Cohen \u2019 s \u03ba Interpretation ( Landis & Koch , 1977 ) A bike idea 0 . 927 Almost perfect A telescopic idea 0 . 893 Almost perfect A modular idea 0 . 877 Almost perfect Extendable components 0 . 823 Almost perfect Rotating joints 0 . 768 Substantial No spokes 0 . 769 Substantial Open frame 0 . 915 Almost perfect Cantilevered saddle 0 . 852 Almost perfect Artificial Intelligence for Engineering Design , Analysis and Manufacturing 315 https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press significant differences between the encouraged groups and the neutral group for both the number of ideas containing features of the example solution , t ( 152 ) = \u2212 2 . 67 , p = 0 . 004 , d = 0 . 710 , and the number of features included in the participants \u2019 ideas , t ( 152 ) = \u2212 3 . 04 , p = 0 . 004 , d = 0 . 736 , with the encouraged groups generating more ideas with these features and repeating more of these features per idea ( Table 10 shows summary statistics for these results ) . As in study 1 , these results show that only the encouraging instructions influenced the incorporation of structural features from the example . On average , participants in encouraged groups incorporated more example features per idea and generated more ideas that incorporated such features , whereas discouraged groups produced results similar to the neutral group and the BG . While these findings are consistent with those of study 1 , here we observed considerably smaller F - statistics and effect sizes . This can be attributed to the BG generating more ideas with features of the telescopic example than of the modular one , t ( 54 ) = \u2212 3 . 59 , p = 0 . 001 , d = 0 . 960 , and incorporating more of such features into the ideas when compared with the incorporation of modular features , t ( 44 . 3 ) = \u2212 3 . 77 , p = 0 . 000 , d = 1 . 01 . Indeed , there was a general tendency of increased feature repetition in the SGs of study 2 , so we can assume that the features of the telescopic exam - ple were more likely to appear in the participants \u2019 ideas in study 2 than the modular features in study 1 . General discussion and limitations Study 2 results replicate most of what we have found in study 1 . In fact , the statistics regarding fluency and the repetition of the idea type were almost identical , indicating that these findings are robust to different stimuli . With respect to the conceptual feature , even though we have found some repetition in the encouraged groups , participants in study 2 generated on average less tele - scopic ideas [ M = 0 . 407 , SD = 0 . 460 , t ( 263 ) = 3 . 02 , p = 0 . 003 , d = 0 . 370 ] and more modular ideas [ M = 0 . 262 , SD = 0 . 396 , t ( 247 . 4 ) = \u2212 2 . 79 , p = 0 . 006 , d = 0 . 344 ] than study 1 participants ( telescopic ideas , M = 0 . 575 , SD = 0 . 448 ; modular ideas , M = 0 . 138 , SD = 0 . 318 ) . This is a surprising result , as it shows an apparent resistance toward incorporating the underlying mechanisms behind the adaptable bike to which participants were exposed . Moreover , it might be the case that this resistance was a conscious choice , as though participants deliberately tried not to conform to whatever they were shown , either because the example already pre - exhausted the solution space they could have explored ( Perttula & Liikkanen , 2006 ) , or because the participants were suspicious or exhibited demand awareness ( Page , 1981 ) . Figure 3 illustrates a selection of participants \u2019 ideas that were bikes , not bikes , modular , and telescopic . The results for structural feature repetition provide further support for the effectiveness of encouraging instructions in inspiration and fixation experiments , as shown in study 1 and reported in previous research ( Smith et al . , 1993 ) . However , the data also revealed a critical aspect of fixation studies : idea repeti - tion results will depend on the kind of feature that experimenters incorporate into the analysis . Presumably , differences in feature repetition results between stimulated and non - stimulated partici - pants are more pronounced when those features represent rare , unusual solutions to the design problem , thus being very unlikely to emerge during idea generation in non - SGs . Conversely , repeti - tion results might be similar if the features analyzed represent obvious solutions , thus being highly likely to appear in ideas with or without external stimulation [ although there is recent evi - dence to the contrary ( Viswanathan et al . , 2016 ) ] . Though the fea - tures used in both our studies are easily comparable ( i . e . , following a similar structure and referring to the same bike parts ) , we believe that the features used in study 2 were more common than those in study 1 , especially the \u201c extendable compo - nents to change bike size \u201d , which frequently appeared in the designs of all groups . This might have biased the idea repetition analysis , and therefore compromised the comparability of the two studies in this respect ( see Fig . 4 for a comparison of feature repetition results between the two studies ) . The main limitations of the studies reported here are the dura - tion of the generation session , the pool of participants chosen , and the design problem used . The idea generation session in this study was 10 min long , which can be considered short when compared with other fixation studies in which generation sessions typically lasted for 30 or 60 min ( Vasconcelos & Crilly , 2016 ) . In addition , the research suggests that novel ideas tend to occur later in the idea generation session ( Kudrowitz & Dippo , 2013 ) ; thus , the short session adopted for this study might have contributed to inflated fixation scores ( when considering fluency and the repetition of the respective idea types ) . However , it is also Table 7 . Summary of ideas generated per participant across groups Generated ideas BG SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 Mean ( and SD ) for the number of ideas per participant 2 . 39 ( 1 . 31 ) 1 . 29 ( 0 . 46 ) 1 . 56 ( 1 . 45 ) 1 . 44 ( 0 . 65 ) 1 . 38 ( 0 . 64 ) 1 . 62 ( 0 . 85 ) Range of ideas generated per participant 1 \u2013 5 1 \u2013 2 1 \u2013 8 1 \u2013 3 1 \u2013 3 1 \u2013 4 Total number of ideas 67 36 39 36 36 42 Total number of participants 28 28 25 25 26 26 Table 8 . Summary of bike and non - bike ideas generated across groups Bike ideas BG SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 Mean ( and SD ) for the number of bike ideas per participant 1 . 43 ( 0 . 84 ) 0 . 929 ( 0 . 22 ) 0 . 962 ( 0 . 14 ) 0 . 867 ( 0 . 33 ) 0 . 904 ( 0 . 25 ) 0 . 936 ( 0 . 18 ) Number of bike ideas ( and % ) 40 ( 60 % ) 33 ( 92 % ) 33 ( 85 % ) 31 ( 86 % ) 30 ( 84 % ) 37 ( 88 % ) Number of non - bike ideas ( and % ) 27 ( 40 % ) 3 ( 8 % ) 6 ( 15 % ) 5 ( 14 % ) 6 ( 16 % ) 5 ( 12 % ) 316 Luis A . Vasconcelos et al . https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press possible that 10 min of ideation might have been too short for fixation to take place ( or to be measured at least ) , although we believe that this is unlikely , as research has already shown that having extra time helps to diminish fixation effects ( Tseng et al . , 2008 ) . The participants in our studies were undergraduate students and the generation session was part of an ongoing engi - neering course . This might have resulted in a more diligent par - ticipant behavior when compared with other studies in which participants and experimenters did not have a student \u2013 lecturer relationship . As a result , the setup adopted for this study might have contributed to the extent to which participants adhered to the instructions . Additionally , the behavior of design students ( or novices ) and practitioners ( or experts ) differs , both with respect to the design process in general ( for a review , see Dinar et al . , 2015 ) , and design fixation in particular ( for a review , see Vasconcelos & Crilly , 2016 ) . Thus , our results might have been different had our participants been more experienced . The design problem used in this study was chosen partly because it was unli - kely that the participants had designed solutions to it before . As such , it is possible that using familiar problems will produce dif - ferent results , although research has demonstrated that fixation effects can be observed with both familiar and unfamiliar prob - lems ( Jansson & Smith , 1991 ) . Finally , although it is not necessarily a limitation of this type of work , but a characteristic of it , the evaluation method used in these studies can be considered subjective to some extent [ yet , consensual assessment is still the standard method in idea genera - tion and design fixation studies ( see Vasconcelos & Crilly , 2016 ) ] . While new methods are available today , such as eye - tracking and neuroimaging , they either require additional techniques for trian - gulation or are still very recent and their potential for providing more objective , meaningful data is yet to be developed ( Editorial board of IJDCI , 2013 ) . With the aim to achieve higher objectivity levels when analyzing data from idea generation experiments , two recent studies had their participants designing in a digital environment , such as computer games ( Neroni et al . , 2017 ) and computer - aided design tools ( Zhang et al . , 2017 ) . Apart from allowing more objective and comparable anal - yses of the design activity , these approaches may enable a series of benefits , such as using design problems and materials that are not easily manipulated in classroom settings , unobtrusive data capture throughout the design task , compatibility with other digital tools for data capture , and better control over the format and quality of the output generated . These are exciting opportunities not just for future work in design fixation ; they show themselves as new methodological paradigms to study human cognition more broadly . Conclusion and future work We have tested the influence of instructions on idea generation . More specifically , we analyzed how discouraging , encouraging , and neutral instructions may affect the number of ideas generated and the incorporation of the example or its parts into the partic - ipants \u2019 ideas . The instructions used were provided along with an external stimulus and its description . It is important to differenti - ate the descriptions from instructions because in this study we have controlled the former but manipulated the latter . We found that instructions shaped the idea generation of our partic - ipants to some extent . When encouraged or required to use features from the example , participants consistently copied struc - tural features but some failed to copy a more abstract conceptual feature . When discouraged or forbidden from using features from the example , however , most participants failed to reduce the num - ber of features copied . This result allows us to infer that more con - crete features are easier to recognize \u2013 and thus reproduce \u2013 than more abstract features such as modularity . However , this may reflect a peculiarity of the modularity feature , since an equivalent approach , such as extendibility , produced slightly different results . Additionally , the results might indicate that positive instructions are more effective than negative ones , irrespective of how flexible or strict the instructions are . This can tell us how to frame future instructions , whether that is with respect to experimental stimuli in research , illustrative cases in design education , and inspira - tional examples in design practice . Regardless of the instructions given to the participants , using different conceptual and structural features between studies allowed us to explore how two alternative but equivalent approaches to solving a design problem can compete for atten - tion . When participants complied with the encouraging instruc - tions and generated more telescopic ideas , they also decreased the number of modular ideas they generated \u2013 we call this design diversion . Two other observations were made when analyzing the datasets . The first is that the telescopic bike conditions generated more modular ideas and less telescopic ideas than the modular bike conditions , and we interpret this as the participants deliber - ately exploring an alternative solution space . The second is that feature repetition calculations can be misleading depending on Table 9 . Summary of telescopic and non - telescopic ideas generated across groups Telescopic ideas BG SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 Mean ( and SD ) for the number of telescopic ideas per participant 0 . 43 ( 0 . 50 ) 0 . 08 ( 0 . 27 ) 0 . 3 ( 0 . 47 ) 0 . 19 ( 0 . 40 ) 0 . 19 ( 0 . 40 ) 0 . 18 ( 0 . 39 ) Number of telescopic ideas ( and % ) 19 ( 28 % ) 16 ( 44 % ) 7 ( 18 % ) 8 ( 22 % ) 15 ( 42 % ) 19 ( 45 % ) Number of non - telescopic ideas ( and % ) 48 ( 72 % ) 20 ( 56 % ) 32 ( 82 % ) 28 ( 78 % ) 21 ( 58 % ) 23 ( 55 % ) Table 10 . Summary of telescopic features incorporated into the participants \u2019 ideas and ideas with telescopic features across groups Feature repetition BG SG \u2212 2 SG \u2212 1 SG0 SG + 1 SG + 2 Mean ( and SD ) for the number features incorporated per idea 1 . 21 ( 1 . 03 ) 1 . 43 ( 1 . 55 ) 1 . 00 ( 1 . 19 ) 0 . 84 ( 1 . 07 ) 1 . 65 ( 1 . 60 ) 2 . 08 ( 1 . 72 ) Range of features incorporated per participant 0 \u2013 4 0 \u2013 6 0 \u2013 4 0 \u2013 4 0 \u2013 8 0 \u2013 5 Total number of ideas with features incorporated ( and % ) 24 ( 36 % ) 23 ( 64 % ) 15 ( 38 % ) 13 ( 36 % ) 23 ( 64 % ) 26 ( 62 % ) Artificial Intelligence for Engineering Design , Analysis and Manufacturing 317 https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press how common or obvious the features might be ( e . g . , the bike has a saddle vs . the bike has wings ) , but also on their level of granularity ( e . g . , the idea is a transportation mode , it is a bike , it is a modular bike , the pedals are made of modular - blocks , and so on ) , with repetition results likely varying according to each level . As such , future studies should reflect on both the commonness and the granularity level of the features under consideration and could report results according to these categories of features ( e . g . , a quadrant chart of features , having commonness and granularity as two orthogonal continuums ) . Nevertheless , all these potential findings need support from more data and are thus open for future investigation . Again , irrespective of how discouraging the instructions were , stimulated participants still exhibited fixation effects due to their exposure to the example design ( i . e . , they generated fewer ideas and proportionally more bike ideas than the BG ) . This result is in line with many other design fixation experiments in which par - ticipants become stuck on a specific idea type . However , it is important to emphasize that in this study the description of the stimulus itself could also be causing fixation as the stimulus was presented to the participants as \u201c an example of how they should present their ideas \u201d . Thus , perhaps there is an implicit suggestion for the participants to produce ideas that are similar to the example , i . e . , a bike . Another possible explanation is that in inspiration and fixation studies , the examples overwhelm the par - ticipants \u2019 interpretation of the problem , to the extent that the examples themselves become part of the participants \u2019 problem representation . Future studies could investigate such possibilities Fig . 3 . Some of the participants \u2019 ideas : ( a ) bike idea , ( b ) non - bike idea , ( c ) modular idea , and ( d ) telescopic idea . Note that these examples are not representative of the design work of any particular experimental group . Fig . 4 . Bar chart showing the average number of ideas with features of the example and an average number of example features per idea across groups ( averages are given per participant ; error bars indicate standard deviation ) . 318 Luis A . Vasconcelos et al . https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press and complement our understanding about stimuli introduction by manipulating the descriptions provided to designers and by testing how examples interfere with problem representation . It is also important to consider here and in similar future studies , that copying and repeating features from examples do not neces - sarily imply fixation . Fixation might be inferred when participants are discouraged from copying ( or maybe when they are given no guidance at all ) , and yet they still copy , otherwise , they may only be complying with the descriptions and instructions in the experi - ments . Similarly , we should be cautious when referring to \u201c idea repetition \u201d for those participants who are given no stimulus but spontaneously generate ideas that incorporate features that also happen to be in the example provided to others . Finally , when considering design practice , design teams and project managers can benefit from being aware of the design fixa - tion knowledge and from the findings discussed in this paper , applying such knowledge depending on the scope and stage of the design project . For instance , the idea of design diversion might be considered by project managers for strategically remov - ing complementary areas of the solution space that are not of interest . The results about the efficacy of instructions and the abstraction of stimuli reported here are also relevant to how inspirational stimuli should be framed when presented to the designers . This is particularly important for the development and implementation of computer - aided design tools that provide designers with external stimuli . Much has been researched on how such software tools might be structured and interacted with , and what form the inspirational stimuli should take ( Shneiderman , 2000 ; T\u00f6re Yargin & Crilly , 2015 ) . However , it is also important to understand how those stimuli should be intro - duced , whether by description , instruction , or both . Our findings show that using positive instructions ( either requiring or encoura - ging ) in inspiration tools can guarantee that designers will incor - porate features from a given stimulus , and that choosing a concrete stimulus will increase feature incorporation even further . However , should designers be steered toward or away from the repetition of structural features , directed to identify conceptual features , or simply be left alone to interpret and respond as they see fit ? By developing a better understanding of how stimuli instructions influence idea generation , we will move closer to answering such questions and thereby become more capable of supporting creative design . Acknowledgments . This work was supported by the CAPES Foundation , Ministry of Education of Brazil ( grant number BEX 11468 / 13 - 0 ) and the UK \u2019 s Engineering and Physical Sciences Research Council ( grant number EP / K008196 / 1 ) . Research data supporting this publication are available from the University of Cambridge data repository : https : / / doi . org / 10 . 17863 / CAM . 7571 References Aleman V ( 2009 ) ECO 07 \u2013 Compactable Urban Bicycle , accessed February 10 , 2017 . Available at https : / / www . behance . net / gallery / 293563 / ECO - 07 - Compactable - Urban - Bicycle . Atilola O and Linsey J ( 2015 ) Representing analogies to influence fixation and creativity : a study comparing computer - aided design , photographs , and sketches . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 29 ( 2 ) , 161 \u2013 171 . Cardoso C and Badke - Schaub P ( 2011 ) The influence of different pictorial representations during idea generation . The Journal of Creative Behavior 45 ( 2 ) , 130 \u2013 146 . Chakrabarti A , Sarkar P , Leelavathamma B and Nataraju BS ( 2005 ) A func - tional representation for aiding biomimetic and artificial inspiration of new ideas . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 19 ( 2 ) , 113 \u2013 132 . Cheng P , Mugge R and Schoormans JPL ( 2014 ) A new strategy to reduce design fixation : presenting partial photographs to designers . Design Studies 35 ( 4 ) , 374 \u2013 391 . Cheong H , Hallihan G and Shu LH ( 2014 ) Understanding analogical reason - ing in biomimetic design : an inductive approach . In Gero JS ( ed . ) . Design Computing and Cognition \u2018 12 . Dordrecht , The Netherlands : Springer Netherlands , pp . 21 \u2013 39 . Chrysikou EG and Weisberg RW ( 2005 ) Following the wrong footsteps : fixation effects of pictorial examples in a design problem - solving task . Journal of Experimental Psychology : Learning , Memory , and Cognition 31 ( 5 ) , 1134 \u2013 1148 . Dahl DW and Moreau P ( 2002 ) The influence and value of analogical thinking during new product ideation . Journal of Marketing Research 39 ( 1 ) , 47 \u2013 60 . Dinar M , Shah JJ , Cagan J , Leifer L , Linsey J , Smith SM and Hernandez NV ( 2015 ) Empirical studies of designer thinking : past , present , and future . Journal of Mechanical Design 137 ( 2 ) , 021101 . Eckert CM ( 1997 ) Design inspiration and design performance . In Proceedings of the 78th World Conference of the Textile Institute , Thessaloniki , Greece , pp . 369 \u2013 387 . Editorial board of IJDCI ( 2013 ) Perspectives on design creativity and innova - tion research . International Journal of Design Creativity and Innovation 1 ( 1 ) , 1 \u2013 42 . Feng T , Cheong H and Shu LH ( 2014 ) Effects of abstraction on selecting rel - evant biological phenomena for biomimetic design . Journal of Mechanical Design 136 ( 11 ) , 111111 . Floss GH ( 2010 ) Zee - K Ergonomic Bike , accessed February 10 , 2017 . Available at http : / / www . coroflot . com / gabrielfloss / zee - k - ergonomic - bike1 . Fu K , Cagan J and Kotovsky K ( 2010 ) Design team convergence : the influ - ence of example solution quality . Journal of Mechanical Design 132 ( 11 ) , 111005 . Gon\u00e7alves M , Cardoso C and Badke - Schaub P ( 2012 ) . Find your inspiration : exploring different levels of abstraction in textual stimuli . In 2 nd International Conference on Design Creativity ( ICDC 2012 ) , Glasgow , UK . Gon\u00e7alves M , Cardoso C and Badke - Schaub P ( 2014 ) What inspires designers ? Preferences on inspirational approaches during idea generation . Design Studies 35 ( 1 ) , 29 \u2013 53 . Jansson DG and Smith SM ( 1991 ) Design fixation . Design Studies 12 ( 1 ) , 3 \u2013 11 . Kudrowitz B and Dippo C ( 2013 ) Getting to the novel ideas : exploring the alternative uses test of divergent thinking . In ASME 2013 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference . American Society of Mechanical Engineers , V005T06A013 . Kudrowitz B , Te P and Wallace D ( 2012 ) The influence of sketch quality on perception of product - idea creativity . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 26 ( 3 ) , 267 \u2013 279 . Landis JR and Koch GG ( 1977 ) The measurement of observer agreement for categorical data . Biometrics 33 ( 1 ) , 159 . LeFevre J - A and Dixon P ( 1986 ) Do written instructions need examples ? Cognition and Instruction 3 ( 1 ) , 1 \u2013 30 . Liikkanen LA and Perttula M ( 2008 ) Inspiring design idea generation : insights from a memory - search perspective . Journal of Engineering Design 21 ( 5 ) , 545 \u2013 560 . Linsey J , Tseng I , Fu K , Cagan J , Wood K and Schunn C ( 2010 ) A study of design fixation , its mitigation and perception in engineering design faculty . Journal of Mechanical Design 132 ( 4 ) , 041003 . Linsey J and Wood K ( 2007 ) Wordtrees : a method for design - by - analogy . In ASEE Annual Conference , AC 2008 - 1669 . Linsey J , Wood KL and Markman AB ( 2008 ) Modality and representation in analogy . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 22 ( 2 ) , 85 \u2013 100 . Lujun Z ( 2011 ) Design fixation and solution quality under exposure to exam - ple solution . In IEEE 2 nd International Conference on Computing , Control and Industrial Engineering ( CCIE ) , 2011 , vol . 1 . IEEE , pp . 129 \u2013 132 . Neroni MA , Vasconcelos LA and Crilly N ( 2017 ) Computer - based \u201c mental set \u201d tasks : an alternative approach to studying design fixation . Journal of Mechanical Design 139 ( 7 ) , 071102 . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 319 https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press Nijstad BA , Stroebe W and Lodewijkx HF ( 2002 ) Cognitive stimulation and interference in groups : exposure effects in an idea generation task . Journal of Experimental Social Psychology 38 ( 6 ) , 535 \u2013 544 . Page MM ( 1981 ) Demand compliance in laboratory experiments . In Tedeschi JT ( ed . ) . Impression Management Theory and Social Psychological Research . New York , NY : Academic Press , pp . 57 \u2013 82 . Perttula M and Liikkanen LA ( 2006 ) . Exposure effects in design idea genera - tion : unconscious conformity or a product of sampling probability . In Proceedings of NordDesign . Reykjavik , Iceland : The Design Society , pp . 42 \u2013 55 . Perttula M and Sipil\u00e4 P ( 2007 ) The idea exposure paradigm in design idea generation . Journal of Engineering Design 18 ( 1 ) , 93 \u2013 102 . Purcell AT and Gero JS ( 1992 ) Effects of examples on the results of a design activity . Knowledge - Based Systems 5 ( 1 ) , 82 \u2013 91 . Purcell AT and Gero JS ( 1996 ) Design and other types of fixation . Design Studies 17 ( 4 ) , 363 \u2013 383 . Sarkar P and Chakrabarti A ( 2008 ) The effect of representation of triggers on design outcomes . AI Artificial Intelligence for Engineering Design , Analysis and Manufacturing 22 ( 2 ) , 101 \u2013 116 . Shah JJ , Smith SM and Vargas - Hernandez N ( 2003 ) Metrics for measuring ideation effectiveness . Design Studies 24 ( 2 ) , 111 \u2013 134 . Shneiderman B ( 2000 ) Creating creativity : user interfaces for supporting innovation . ACM Transactions on Computer - Human Interaction 7 ( 1 ) , 114 \u2013 138 . Siangliulue P , Chan J , Gajos KZ and Dow SP ( 2015 ) Providing timely examples improves the quantity and quality of generated ideas . In ACM SIGCHI Conference on Creativity and Cognition . Glasgow , UK : ACM Press , pp . 83 \u2013 92 . Smith SM , Ward TB and Schumacher JS ( 1993 ) Constraining effects of exam - ples in a creative generation task . Memory & Cognition 21 ( 6 ) , 837 \u2013 845 . T\u00f6re Yargin G and Crilly N ( 2015 ) Information and interaction requirements for software tools supporting analogical design . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 29 ( 2 ) , 203 \u2013 214 . Tseng I , Moss J , Cagan J and Kotovsky K ( 2008 ) The role of timing and ana - logical similarity in the stimulation of idea generation in design . Design Studies 29 ( 3 ) , 203 \u2013 221 . Tsenn J , Atilola O , McAdams DA and Linsey JS ( 2014 ) The effects of time and incubation on design concept generation . Design Studies 35 ( 5 ) , 500 \u2013 526 . Vasconcelos LA , Cardoso CC , S\u00e4\u00e4ksj\u00e4rvi M , Chen C - C and Crilly N ( 2017 ) Inspiration and fixation : the influences of example designs and system properties in idea generation . Journal of Mechanical Design 139 ( 3 ) , 031101 \u2013 031101 - 13 . Vasconcelos LA and Crilly N ( 2016 ) Inspiration and fixation : questions , methods , findings , and challenges . Design Studies 42 , 1 \u2013 32 . Vasconcelos LA , Neroni MA and Crilly N ( 2016 ) . Fluency results in design fixation experiments : an additional explanation . In The 4th International Conference on Design Creativity . Atlanta , GA : The Design Society . Vattam S , Wiltgen B , Helms M , Goel AK and Yen J ( 2011 ) DANE : fostering creativity in and through biologically inspired design . In Taura T and Nagai Y ( eds ) . Design Creativity 2010 . London , UK : Springer London , pp . 115 \u2013 122 . Viswanathan V , Atilola O , Esposito N and Linsey J ( 2014 ) . A study on the role of physical models in the mitigation of design fixation . Journal of Engineering Design 25 ( 1 \u2013 3 ) , 25 \u2013 43 . Viswanathan V , Tomko M and Linsey J ( 2016 ) A study on the effects of example familiarity and modality on design fixation . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 30 ( 2 ) , 171 \u2013 184 . Yilmaz S , Seifert CM and Gonzalez R ( 2010 ) Cognitive heuristics in design : instructional strategies to increase creativity in idea generation . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 24 ( 3 ) , 335 \u2013 355 . Youmans RJ ( 2011 a ) Design fixation in the wild : design environments and their influence on fixation . The Journal of Creative Behavior 45 ( 2 ) , 101 \u2013 107 . Youmans RJ ( 2011 b ) The effects of physical prototyping and group work on the reduction of design fixation . Design Studies 32 ( 2 ) , 115 \u2013 138 . Youmans RJ and Arciszewski T ( 2014 ) Design fixation : classifications and modern methods of prevention . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 28 ( 2 ) , 129 \u2013 137 . Zahner D , Nickerson JV , Tversky B , Corter JE and Ma J ( 2010 ) A fix for fixation ? Rerepresenting and abstracting as creative processes in the design of information systems . Artificial Intelligence for Engineering Design , Analysis and Manufacturing 24 ( 2 ) , 231 . Zhang HZ , Xie C and Nourian S ( 2017 ) Are their designs iterative or fixated ? Investigating design patterns from student digital footprints in computer - aided design software . International Journal of Technology and Design Education , 1 \u2013 23 . https : / / doi . org / 10 . 1007 / s10798 - 017 - 9408 - 1 Zhao M ( 2013 ) Seek it or let it come : how designers achieve inspirations . In CHI \u2019 13 Extended Abstracts on Human Factors in Computing Systems . ACM , pp . 2779 \u2013 2784 . Dr Luis A . Vasconcelos has recently completed his PhD in Engineering Design at the University of Cambridge , where he was also a member of the Design Practice research group at the Engineering Design Centre . With a back - ground in design research and methodology , his research explores new paths in design inspiration and fixation studies . He is interested in the design process at both macro and micro levels . Dr Maria A . Neroni is a Research Associate at the Design Practice Group in the Cambridge Engineering Design Centre . With a background in experi - mental psychology , cognitive neuroscience and education , her current research focuses on developing novel experimental paradigms to investigate creativity in engineering design . Dr Nathan Crilly is a Senior Lecturer in Engineering Design at the University of Cambridge . He leads the Design Practice research group at the Engineering Design Centre . His research interests lie in design , creativity , and communica - tion . He and his research group employ an interdisciplinary approach to study - ing the development of products , the properties they exhibit , and how people respond to them . 320 Luis A . Vasconcelos et al . https : / / doi . org / 10 . 1017 / S0890060417000658 Published online by Cambridge University Press",
    "soriano-castell2017rock1": "1 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x www . nature . com / scientificreports ROCK1 is a novel Rac1 effector to regulate tubular endocytic membrane formation during clathrin - independent endocytosis David Soriano - Castell 1 , Albert Chavero 1 , Carles Rentero 1 , Marta Bosch 1 , Maite Vidal - Quadras 1 , Albert Pol 1 , 2 , Carlos Enrich 1 & Francesc Tebar 1 Clathrin - dependent and - independent pathways contribute for \u03b2 1 - integrin endocytosis . This study defines a tubular membrane clathrin - independent endocytic network , induced with the calmodulin inhibitor W13 , for \u03b2 1 - integrin internalization . This pathway is dependent on increased phosphatidylinositol 4 , 5 - bisphosphate ( PI ( 4 , 5 ) P 2 ) levels and dynamin activity at the plasma membrane . Exogenous addition of PI ( 4 , 5 ) P 2 or phosphatidylinositol - 4 - phosphate 5 - kinase ( PIP5K ) expression mimicked W13 - generated - tubules which are inhibited by active Rac1 . Therefore , the molecular mechanisms downstream of Rac1 , that controls this plasma membrane tubulation , were analyzed biochemically and by the expression of different Rac1 mutants . The results indicate that phospholipase C and ROCK1 are the main Rac1 effectors that impair plasma membrane invagination and tubule formation , essentially by decreasing PI ( 4 , 5 ) P 2 levels and promoting cortical actomyosin assembly respectively . Interestingly , among the plethora of proteins that participate in membrane remodeling , this study revealed that ROCK1 , the well - known downstream RhoA effector , has an important role in Rac1 regulation of actomyosin at the cell cortex . This study provides new insights into Rac1 functioning on plasma membrane dynamics combining phosphatidylinositides and cytoskeleton regulation . Endocytosis is an essential process for eukaryotic cells to internalize growth factors , hormones , and nutrients from the plasma membrane ( PM ) or extracellular fluid 1 \u2013 4 . The internalization routes can be classified into clathrin - dependent endocytosis ( CDE ) and clathrin - independent endocytosis pathways ( CIE ) 2 , 5 \u2013 9 . CIE pathways include different ways of internalization which show high complexity , though all generally share their association with PM microdomains enriched in cholesterol and glycosphingolipids 10 \u2013 13 . The molecular machinery that reg - ulates these different routes is only now beginning to emerge . In comparison with CDE pathways , the morpho - logical features of membrane carriers generated by CIE pathways range from small vesicles to membrane tubular networks of different size and extension 9 , 11 , 14 \u2013 16 . Several laboratories , including ours , have recently described the existence of PM tubular networks belong - ing to CIE pathways , with tubules of tens of micrometers in length 11 , 15 , 17 . Major histocompatibility complex I ( MHCI ) has been reported to be internalized via Arf6 - dependent , clathrin and caveolae - independent endo - cytosis 18 , and both MHCI and Arf6 were detected in tubules that lack the CDE marker transferrin 15 . Induced non - clathrin - mediated tubular membrane invaginations have also been reported for the uptake of Cholera and Shiga toxins , and the simian SV40 virus 16 , 17 , 19 . Formation of these tubules seems to require an intact microtubule network 15 , 16 . Moreover , we have demonstrated the involvement of Rac1 , calmodulin ( CaM ) , and phosphatidy - linositol 4 , 5 - bisphosphate ( PI ( 4 , 5 ) P 2 ) in this process 15 . While expression of the constitutively active Rac1 mutant Rac1 G12V completely abolishes membrane tubules , the dominant negative mutant Rac1 T17N triggers the formation . The same phenotype is generated by phosphatidylinositol 4 - phosphate - 5 - kinase ( PIP5K ) overexpression or by treatment with the CaM inhibitor N - ( 4 - aminobutyl ) - 5 - chloro - 2 - naphthalenesulfonamide ( W13 ) , which increase PI ( 4 , 5 ) P 2 levels at the PM 15 . Involvement of PI ( 4 , 5 ) P 2 in the initiation of endocytic events is determined by its 1 Departament de Biomedicina , Unitat de Biologia Cel\u00b7lular , Centre de Recerca Biom\u00e8dica CELLEX , Institut d\u2019Investigacions Biom\u00e8diques August Pi i Sunyer ( IDIBAPS ) , Facultat de Medicina , Universitat de Barcelona , Casanova 143 , 08036 , Barcelona , Spain . 2 Instituci\u00f3 Catalana de Recerca i Estudis Avan\u00e7ats ( ICREA ) , 08010 , Barcelona , Spain . Correspondence and requests for materials should be addressed to F . T . ( email : tebar @ ub . edu ) Received : 7 November 2016 Accepted : 23 June 2017 Published online : 31 July 2017 OPEN www . nature . com / scientificreports / 2 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x ability to bind and recruit several membrane - bending proteins such as dynamin or BAR - domain containing pro - teins , but also by its role in actin dynamics at the cell surface 20 \u2013 25 . Afterwards , the decrease of PI ( 4 , 5 ) P 2 by specific phosphatases and / or phospholipases , such as synaptojanin or phospholipase C ( PLC ) , is important to promote pinch - off of the plasma membrane and the consequent endocytic vesicle production 26 \u2013 28 . The small GTPases , Rac1 , RhoA , and Cdc42 , are implicated in the regulation of several CIE pathways . Rac1 and RhoA control interleukin - 2 receptor ( IL - 2R ) uptake 29 , 30 , and Rac1 regulates macropinocytosis with Cdc42 , which is also required during clathrin - independent carrier ( CLIC ) and GPI - enriched endocytic com - partment ( GEEC ) endocytosis 31 \u2013 34 . Several CIE pathways also require Pak1 , Pak2 , or cortactin activity , which are Rac1 actin - related targets 30 , 35 , suggesting that Rac1 - dependent actin polymerization plays a key role during these events . The PI ( 4 , 5 ) P 2 - binding protein dynamin , as well as cortactin , have been reported to be important actin - modulating and membrane - remodeling factors during both CDE and CIE 35 \u2013 37 . Therefore , cortactin and dynamin may be acting downstream of Rac1 to regulate the endocytic tubules formation . Moreover , recent stud - ies have identified myosins regulating endocytosis 38 \u2013 41 , and it has been shown that an increased assembly of actomyosin networks at the PM antagonizes membrane invagination and endocytosis 42 , 43 . Actomyosin is mainly regulated by RhoA through its effector ROCK1 ( rho associated coiled - coil containing protein kinase 1 ) , but also by Rac1 , and these two GTPases usually have opposite effects in several cellular processes 44 . The possible con - tribution of Rac1 - dependent actomyosin regulation to CIE has not been investigated in depth , and nor has its contribution to tubule regulation . Actually , Rac1 could control tubule outcomes by regulating PI ( 4 , 5 ) P 2 levels ( via PLC activity ) and cytoskeleton dynamics ( through actin polymerization and myosin activation ) 45 \u2013 51 . In the present study we demonstrate that increased PI ( 4 , 5 ) P 2 levels trigger dynamin - dependent endocytic tubules formation and enhance \u03b2 1 - integrin internalization , and that this process can be neutralized by Rac1 activation . We show that Rac1 regulates PM endocytic tubule formation by controlling PI ( 4 , 5 ) P 2 levels , actin dynamics and myosin activation through activation of PLC , cortactin and ROCK1 , respectively . Importantly , the results reveal ROCK1 as a new Rac1 effector and here we propose a novel Rac1 - dependent ROCK1 activation pathway to regulate membrane dynamics . Results and Discussion Integrin internalization via a clathrin - independent , Rac1 - regulated endocytic pathway . We have previously shown that Rac1 activity can regulate the formation of membrane tubular structures , with the dominant negative Rac1 increasing and the constitutively active mutant reducing the percentage of cells present - ing tubules 15 . These tubular membrane structures , which are also induced after treatment with the calmodulin inhibitor W13 , transported clathrin independent endocytic cargoes like MHCI 15 . Since Rac1 activity can control integrin trafficking , and vice versa 52 \u2013 56 , we have examined whether integrins were also present in these endocytic tubules and if the presence of such tubules affects integrin transport to early endosomes ( EEs ) . COS1 cells were incubated with an antibody that recognizes the \u03b2 1 - integrin ectodomain and treated then with W13 for 10 min - utes at 37 \u00b0C , before fixing and immunostainning cells with anti - EEA1 antibody . The images in Fig . 1a show the presence of \u03b2 1 - integrin ( red ) in W13 - induced tubules , visualized with the expressed membrane marker GFP - mem ( green ) . Whereas \u03b2 1 - integrin was clearly detected in EEA1 - positive endosomes of control cells ( Fig . 1b ) , those cells that contained extensive tubulation ( W13 treated ) showed low \u03b2 1 - integrin labeling in EEs ( Fig . 1c ) . Considering the presence in tubules as internalized molecules , cells exhibiting tubules showed increase in the total internalized \u03b2 1 - integrin after treatment with W13 at different time points ( 5 , 10 and 20 min ) compared with control cells ( Fig . 1d ) . In these settings , the effect of W13 - induced tubules on transferrin internalization , a well - established cargo of the CDE route , was also analyzed ( Fig . 1e ) . Transferrin was not observed in W13 - tubules and increased transferrin internalization was observed only at 20 minutes in W13 - induced tubules com - pared to control cells , which could be explained by the previously reported effect of W13 inhibiting sorting from early endosomes 57 \u2013 60 and consequently transferrin recycling that at later time points contributes to the uptake measurements . Likewise , this could be the reason for increased \u03b2 1 - integrin internalization at later time points in W13 - treated cells not presenting tubules . To corroborate that clathrin did not participate in W13 - induced tubule formation , clathrin expression was inhibited by siRNA knockdown ( Fig . 1f ) . Figure 1f shows that clathrin down - regulation did not modify the extend of W13 - induced tubules and , in agreement with the inhibition of recycling , W13 treatment accumulated transferrin in EEs vesicles at the cell periphery in contrast to its perinuclear locali - zation observed in control cells . These results indicate that induced tubular endocytic membrane structures are a cellular port of entry important for \u03b2 1 - integrin internalization in a CIE pathway . The effect of W13 on \u03b2 1 - integrin internalization was simultaneously analyzed in cells expressing the consti - tutively active Rac1 mutant ( GFP - Rac1 G12V ) . Image quantification showed that Rac1 G12V expression completely abrogated the W13 - increased \u03b2 1 - integrin internalization at all - time points analyzed ( Fig . 1d ) , indicating that active Rac1 blocks tubule formation instead of promoting tubule fission . Besides , the results imply that Rac1 is a \u03b2 1 - integrin internalization regulator , and suggest that it may regulate integrin turnover through CIE . Induction of dynamin - dependent tubules in CIE pathway by increased PI ( 4 , 5 ) P 2 . Next , clathrin - independent endocytic tubules were further characterized and we focused on the molecular mechanisms activated by Rac1 and its regulation . We and others reported that membrane tubules are induced after increasing PI ( 4 , 5 ) P 2 levels by overexpression of PIP5K 15 , 61 . In fact , W13 - induced PM tubules appear to depend on PIP5K activity 15 . The presence of PI ( 4 , 5 ) P 2 in W13 - tubules was confirmed by immunostaining with an anti - PI ( 4 , 5 ) P 2 antibody ( Fig . 2a ) . Moreover , tubule induction by PI ( 4 , 5 ) P 2 increase was supported by a dose response curve with exogenous diC8 - PI ( 4 , 5 ) P 2 ( previously conjugated with the neomycin carrier for its transmembrane delivery ) 62 , 63 ( Fig . 2b ) . Addition of 50 \u00b5 M of diC8 - PI ( 4 , 5 ) P 2 increased the percentage of cells with tubules up to approximately 35 % compared to 12 % observed by the neomycin carrier in control cells ( Fig . 2c , d ) . Similar ratio elicited by W13 treatment was observed by overexpression of PIP5K or diC8 - PI ( 4 , 5 ) P 2 incubation ( Fig . 2e ) . In addition , these www . nature . com / scientificreports / 3 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x Figure 1 . W13 - induced PM tubules provide an internalization pathway for \u03b2 1 - integrin . ( a ) COS1 cells grown on coverslips expressing the membrane marker GFP - mem were incubated with an anti - \u03b2 1 - integrin rat antibody for 30 minutes at 4 \u00b0C to avoid endocytosis , followed by incubation for 10 minutes at 37 \u00b0C to allow internalization in the presence of W13 ( 20 min , 4 . 5 \u00b5 g / ml ) . After fixation , \u03b2 1 - integrin was detected with an Alexa - 555 labeled anti - rat antibody , and images were acquired with a confocal microscope ( Leica TCS SP5 ) . The higher magnification insets show \u03b2 1 - integrin localization in W13 - induced tubules ( green arrowheads ) . ( b , c ) Following the same procedure explained as in ( a ) , \u03b2 1 - integrin was detected with an Alexa - 647 labeled anti - rat antibody and EEA1 with a specific antibody and the secondary Alexa - 555 anti - mouse in untreated ( b ) or W13 - treated cells ( c ) . Insets show \u03b2 1 - integrin in EEA1 - positive endosomes ( red arrowheads ) ( b ) or in tubules ( green arrowheads ) ( c ) ( bars , 10 \u00b5 m ) . ( d , e ) Quantification of internalized \u03b2 1 - integrin ( d ) and transferrin ( e ) , as explained in the Materials and Methods , in COS1 cells expressing GFP - mem or GFP - Rac1 G12V for the indicated conditions ( W13t , cells presenting tubules ; W13nt , cells without tubules ) . Mean values \u00b1 standard error of the mean ( SEM ) from three independent experiments are shown . Statistical significance between different www . nature . com / scientificreports / 4 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x different experimental conditions similarly increased both the number of tubules per cell ( Fig . 2e ) , and PI ( 4 , 5 ) P 2 levels detected by immunofluorescence compared to control cells ( Fig . 2f ) . These results demonstrated a direct relationship between increased PI ( 4 , 5 ) P 2 levels and tubule development . Therefore , W13 - treatment could be used to increase PI ( 4 , 5 ) P 2 levels and tubulation at the PM . It is known that increased PI ( 4 , 5 ) P 2 are necessary for endocytosis to proceed because they recruit several PI ( 4 , 5 ) P 2 - binding proteins , including adaptor proteins , BAR - domain containing proteins , and dynamin ( among others ) 20 . Although dynamin has an important role in the scission of endocytic vesicles from the PM , it has also been involved in membrane deformation and tubular membrane organization 24 , 64 \u2013 68 . Therefore , we investigated the role of dynamin in these PI ( 4 , 5 ) P 2 - induced tubules . Dynamin action was inhibited by dominant negative mutant expression ( dyn K44A ; Fig . 3a ) , treatment with a specific inhibitor ( dynasore ; Fig . 3a ) , or by siRNA knock - down ( Figs 3b and 2f ) . In each of these experimental settings , W13 - induced tubules were prevented , indicating that this tubular endocytic pathway is dynamin - dependent . Dynamin was necessary to initiate tubule formation , but an additional role of dynamin in the fission of tubules cannot be discarded . Dynamin participates in membrane invagination in combination with BAR - domain containing proteins 43 , 67 . In W13 - induced tubules we have also observed the presence of PACSIN2 , an F - BAR - domain protein that binds dynamin , PI ( 4 , 5 ) P 2 and Rac1 69 , 70 . Interestingly , although PACSIN2 interacts with dynamin , it does not bind or colocalize with clathrin 71 . Moreover , in agreement with the report by Kreuk et al . 69 , we showed that expression of the active Rac1 mutant inhibited the presence of PACSIN2 - positive tubules in COS1 cells after W13 treat - ment ( Supplementary Fig . S1 ) . In addition , it has been reported that PACSIN2 regulates caveolae biogenesis and endocytosis in cholesterol rich and plasma membrane ordered domains 69 , 72 , 73 , where active Rac1 is located 52 , 74 . Actually , we have previously shown that cyclodext rin , a PM cholesterol chelator , inhibited W13 - tubule forma - tion 15 . Therefore , we analyzed whether the specific PI ( 4 , 5 ) P 2 increase elicited by W13 treatment was respon - sible for tubulation in these domains . Tubules were inhibited in cells expressing a PI ( 4 , 5 ) P 2 - phosphatase specifically targeted to PM ordered domains by the 10 N - terminal amino acids of Lck ( L10 - GFP - Inp54p ) 75 . Otherwise , no effect was observed with the phosphatase dead mutant ( L10 - GFP - Inp54p D281A ) ( Fig . 3c ) . Together , these results show that increased PI ( 4 , 5 ) P 2 levels in specific PM domains , where clathrin - independent and dynamin - dependent endocytosis takes place , are probably responsible for tubule formation . In agreement with the localization of Rac1 in ordered domains 74 , 76 , tubular endocytic membranes present in control cells , or elicited by W13 treatment , were inhibited by active Rac1 , as well as tubules induced by PIP5K overexpression or by the addition of exogenous diC8 - PI ( 4 , 5 ) P 2 ( Fig . 3d ) . These results , together with the fact that Rac1 G12V expressing cells showed reduced PI ( 4 , 5 ) P 2 immunostaining in W13 - treated cells ( Supplementary Fig . S2 ) , prompted us to study the role of Rac1 effectors in tubulation . The role of PLC - regulated PI ( 4 , 5 ) P 2 levels on tubule inhibition by active Rac1 . Rac1 can modu - late PI ( 4 , 5 ) P 2 levels at the PM by activating PLC 77 , 78 , which hydrolyzes PI ( 4 , 5 ) P 2 generating diacylglycerol ( DAG ) and inositol trisphosphate ( IP 3 ) . Then , we analyzed PLC involvement by two strategies : ( i ) inhibition of PLC activ - ity with its inhibitor U73122 , and ii ) expression of GFP fusion proteins for two previously described constitutively active ( GTP - bound ) , but PLC - deficient , Rac1 mutants ( Rac1 G12V - F37A and Rac1 G12V - W56A ) 77 . Quantification of tubule formation in COS1 cells demonstrates that U73122 impaired the tubule inhibition produced by GFP - Rac1 G12V expression ( Fig . 3e ) . Moreover , the expression of both Rac1 G12V - F37A and Rac1 G12V - W56A did not inhibit tubules in untreated cells ( Fig . 3d ) . These results strongly suggest that PLC plays an important role in Rac1 - dependent tubule inhibition . To further analyze PLC activity involvement in the Rac1 - dependent inhibition of tubule formation , we assessed the effect of Rac1 G12V - F37A and Rac1 G12V - W56A expression on PI ( 4 , 5 ) P 2 - induced tubulation , either by W13 - treatment or PIP5K - overexpression . As expected , W13 and PIP5K induced a similar percentage of cells presenting tubules in control and Rac1 G12V - F37A expressing cells ( Fig . 3f ) , confirming the involvement of PLC activity . However , the expression of the Rac1 G12V - W56A mutant was still able to block tubule formation ( Fig . 3f ) . Together , these results suggest that , although PLC plays a key role in tubule inhibition by active Rac1 , addi - tional mechanism contributes to the inhibition , as revealed through the use of the PLC - deficient mutants . Since Rac1 G12V - F37A mutant is not able to translocate cortactin to the plasma membrane or interact with ROCK1 ( two important factors for cortical actomyosin regulation ) 79 , 80 , the Rac1 G12V - W56A mutant was therefore considered a valuable tool for studying the role of cytoskeleton in the PLC - independent tubular - inhibitory effect of active Rac1 . conditions and the corresponding control was determined by the two - way ANOVA , * p < 0 . 05 , * * * p < 0 . 001 . Statistical significance in integrin internalization assay ( d ) between W13t and W13nt at 5 , 10 and 20 minutes were * p < 0 . 05 , * * * p < 0 . 001 , * * * p < 0 . 001 , respectively . ( f ) COS1 cells co - transfected with GFP - mem and a specific clathrin heavy chain siRNA or a non - specific siRNA as a control ( 72 h ) were incubated with transferrin - TRITC during 15 minutes at 37 \u00b0C in the presence or absence of W13 ( 4 . 5 \u00b5 g / ml ) . After fixation , clathrin was detected with an anti - mouse antibody ( clone \u00d7 22 ) and the corresponding Alexa - 647 secondary antibody . Confocal images were acquired with a confocal microscope ( Leica TCS SP5 ) through the corresponding channels ( bars , 10 \u00b5 m ) . Downregulation of clathrin expression by its specific siRNA is shown by western blotting using a rabbit polyclonal antibody Graph shows the percentage of cells presenting tubules ( > 5 tubules / cell ) in the indicated conditions . Mean values \u00b1 standard error of the mean ( SEM ) from three independent experiments are shown . Statistical significance between W13 - treatment and the corresponding control was determined by the paired Student\u2019s t - test , * * p < 0 . 01 . www . nature . com / scientificreports / 5 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x Figure 2 . Increased levels of PI ( 4 , 5 ) P 2 induces membrane tubules at the PM . ( a ) COS1 cells expressing Cherry - Mem grown on coverslips were incubated with W13 ( 20 min , 4 . 5 \u00b5 g / ml ) . After fixation ( PFA 4 % , 15 min at 37 \u00b0C ) , endogenous PI ( 4 , 5 ) P 2 was detected with a specific antibody and the corresponding anti - mouse Alexa - 488 labeled secondary antibody . The images were acquired with a confocal microscope ( Leica TCS SP5 ) and magnification insets show the presence of PI ( 4 , 5 ) P 2 in W13 - induced membrane tubules . ( b ) The percentage of cells presenting more than 5 tubules was determined after 20 min of diC8 - PI ( 4 , 5 ) P 2 dose - response at the indicated concentrations . ( c ) The percentage of cells with tubules after 20 min incubation with 50 \u00b5 M diC8 - PI ( 4 , 5 ) P 2 or neomycin control . ( d ) Confocal image of cherry - mem depicted tubules in cells treated with diC8 - PI ( 4 , 5 ) P 2 as in ( c ) ( bars , 10 \u00b5 m ) . ( e ) In Cherry - Mem expressing cells , the percentage of cells presenting tubules and the number of tubules per cell were determined in different experimental conditions : W13 ( 20 min , 4 . 5 \u00b5 g / ml ) , diC8 - PI ( 4 , 5 ) P 2 ( 20 min , 50 \u00b5 M ) or GFP - PIP5K expression . Statistical significance between the different conditions and the control ( NT , non - treatment ) was determined by the unpaired Student\u2019s t - test , * * p < 0 . 01 ( n = 100 cells per condition ) . Representative confocal images acquired through the red channel and insets displaying cells presenting cherry - Mem decorated tubules are shown ( bars , 5 \u00b5 m ) . ( f ) In the same experimental conditions as in ( e ) , the intracellular levels of PI ( 4 , 5 ) P 2 were determined by immunofluorescence performed as in ( a ) . Graph shows the percentage of fluorescence ( a . u . , arbitrary units ) of each condition vs NT , mean values \u00b1 standard error of the mean ( SEM ) from three independent experiments . Statistical significance between the different conditions and the NT was determined by the paired Student\u2019s t - test , * p < 0 . 05 . Representative z - stacks confocal projection images are shown ( bars , 10 \u00b5 m ) . www . nature . com / scientificreports / 6 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x Cortactin - dependent actin polymerization inhibits tubular endocytic membrane structures downstream of active Rac1 . Active Rac1 is important to control actin polymerization ( F - actin ) at the PM and F - actin depolymerizing agents are known to generate membrane tubules in many cell types 81 , 82 . Thus , active Rac1 , by increasing F - actin at the cell cortex , could inhibit PM invagination and consequently tubule formation . It has been reported that Rac1 G12V - F37A mutant is defective in cortical actin generation 79 , 83 , 84 . Therefore , it is plausible that the different tubule inhibition response observed with both active mutants in this study ( F37A and W56A ) may be related to their different abilities to regulate actin polymerization . Figure 3 . PI ( 4 , 5 ) P 2 - induced tubulation is dynamin - dependent and its inhibition by active Rac1 involves PLC activity . ( a , b ) In the presence or absence of W13 ( 20 min , 4 . 5 \u00b5 g / ml ) , the percentage of cells with tubules was determined after dynasore treatment ( 30 min , 150 \u00b5 M ) or dynamin K44A expression ( 24 h ) ( a ) and in cells transfected with a specific dynamin siRNA or a non - targeting siRNA as a control ( 48 h ) ( b ) . Downregulation of dynamin expression by its specific siRNA is shown by western blotting . ( c ) The percentage of cells with tubules was determined in cells expressing Cherry - Mem alone or co - expressed with L10 - GFP Inp54p or L10 - GFP Inp54p D281A , in the presence or absence of W13 ( 20 min , 4 . 5 \u00b5 g / ml ) . ( d ) The percentage of cells with tubules was determined in 1 - hour starved cells expressing Cherry - Mem or Cherry - Rac1 G12V in the presence or absence of PI ( 4 , 5 ) P 2 ( 20 min , 50 \u00b5 M ) , W13 ( 20 min , 4 . 5 \u00b5 g / ml ) , or GFP - PIP5K expression . ( e , f ) The percentage of cells displaying tubules among the starved cells expressing GFP - mem , GFP - Rac1 G12V , or active Rac1 mutants ( F37A and W56A ) after the treatment with the PLC - inhibitor U73122 ( 20 min , 5 \u00b5 M ) ( e ) , W13 ( 20 min , 4 . 5 \u00b5 g / ml ) , or PIP5K overexpression ( f ) . Mean values \u00b1 SEM from three independent experiments are shown in all cases . Statistical significance between different conditions and the corresponding controls were determined by Student\u2019s t - test , * p < 0 . 05 , * * p < 0 . 01 , * * * p < 0 . 001 . www . nature . com / scientificreports / 7 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x To determine the effect of these mutants on actin polymerization , F - actin was detected in Vero cells express - ing the GFP - Rac1 G12V , GFP - Rac1 G12V - F37A , or GFP - Rac1 G12V - W56A using conjugated phalloidin - TRITC . Vero cells , which also showed W13 - induced PM tubulation 15 , were used instead of COS1 to improve the visualization of actin cytoskeleton . Fluorescence confocal images showed that GFP - Rac1 G12V and GFP - Rac1 G12V - W56A modified the F - actin pattern by severely reducing stress fibers and increasing cortical F - actin . The effect of G12V was stronger than the G12V - W56A mutant . By contrast , GFP - Rac1 G12V - F37A mutant did not affect the actin organiza - tion ( Fig . 4a ) . Figure 4 . Actin cytoskeleton is involved in Rac1 - dependent tubule inhibition . ( a ) Vero cells expressing GFP - Rac1 G12V , GFP - Rac1 G12V - F37A , or GFP - Rac1 G12V - W56A were grown on coverslips , and F - actin was visualized by confocal microscopy using phalloidin - TRITC ( 15 min , 0 . 04 U / ml ) ( bars , 10 \u00b5 m ) . ( b ) The percentage of COS1 cells presenting tubules was determined in starved cells expressing GFP - Rac1 G12V or GFP - Rac1 G12V - W56A and treated with the F - actin depolymerizing agent latrunculin A ( LatA ; 20 min , 200 nM ) in the presence or absence of W13 . ( c , d ) Percentage of cells with tubules expressing GFP - mem after pre - incubation with nocodazole ( 10 min , 30 \u00b5 M ) ( c ) or the dynein inhibitor EHNA ( 6 hours , 1 mM ) ( d ) and incubation with W13 for 20 min . Mean values \u00b1 standard error of the mean ( SEM ) from three independent experiments are shown in all cases . Statistical significance between W13 - treatment and the corresponding control were determined by Student\u2019s t - test , * p < 0 . 05 , * * p < 0 . 01 . ( e ) In cells expressing Venus - Rac1wt and treated with W13 , tubulin was detected by immunofluorescence using a mouse anti - \u03b2 - tubulin antibody and a secondary anti - mouse - Alexa594 , and F - actin was detected using SiR actin ( SC006 ) . Representative confocal images and STED images from the selected areas are shown ( bars , 10 \u00b5 m ) . www . nature . com / scientificreports / 8 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x To establish a possible connection between the increased cortical F - actin and the tubule inhibition produced by the active Rac1 mutants ( Rac1 G12V and Rac1 G12V - W56A ) , actin filaments were disrupted using the depolym - erizing agent Latrunculin A ( LatA ) in W13 - treated cells . Actin depolymerization decreased the tubule forma - tion inhibition by Rac1 G12V and completely eliminated the inhibitory effect of Rac1 G12V - W56A ( Fig . 4b ) . These results indicate that inhibition of membrane tubulation by Rac1 depends on actin polymerization , and that actin cytoskeleton is not necessary for membrane invagination and elongation to proceed . Actually , consider - ing the critical role of microtubules ( MTs ) for the stabilization of W13 - induced tubules described previously 15 , and further analyzed here using the MT depolymerizing agent nocodazole ( Fig . 4c ) , and the recently described role of dyneins for the stabilization and elongation of PM tubular structures 16 , the general dynein inhibitor erythro - 9 - [ 3 - ( 2 - hydroxynonyl ) ] adenine ( EHNA ) impaired W13 - induced tubulation in COS1 cells ( Fig . 4d ) . In addition , \u03b2 - tubulin and F - actin staining in W13 - induced tubules cells expressing Venus - Rac1wt , showed some association of these tubules with MTs but not with F - actin ( Fig . 4e ) . Although only occasional coincidence of MTs with tubules was observed , STED and confocal microscopy images revealed highly similar pattern and direction - ality between both networks ( Figs 4e and S4a ) , which is consistent with the dependency of W13 - induced tubules on the integrity of MTs ( Fig . 4c ) . Accordingly , nocodazole also inhibited \u03b2 1 - intergrin internalization elicited by W13 treatment in cells presenting tubules ( Supplementary Fig . S4b ) . These results indicate a key role of dyneins and MTs in PI ( 4 , 5 ) P 2 - induced membrane elongation towards the cell center . In summary , actin cytoskeleton is unnecessary for tubule elongation ( operated by MTs and dyneins ) and Rac1 - driven actin polymerization is critical to inhibit basal and PI ( 4 , 5 ) P 2 - induced membrane invagination . However , the results obtained with LatA cannot rule out a role of actin polymerization in PM invagination scission . Rac1 G12V - W56A and Rac1 G12V - F37A have differential effects on cortical F - actin , which may explain the differences in tubule formation inhibition by each mutant . Indeed , it has been described that active Rac1 F37A is not able to translocate cortactin , an actin polymerizing protein , to the PM 79 . To address if Rac1 G12V - W56A translocates cort - actin to the PM to inhibit PI ( 4 , 5 ) P 2 - induced tubule formation , we analyzed the location of endogenous cortac - tin in Rac1 G12V , Rac1 G12V - F37A and Rac1 G12V - W56A expressing Vero cells by immunofluorescence ( Fig . 5a ) . These images showed that Rac1 G12V and Rac1 G12V - W56A mutants translocate cortactin to the cell periphery ( being again more clear in Rac1 G12V expressing cells ) , and Rac1 G12V - F37A does not . Additionally , the involvement of cortactin in Rac1 G12V - W56A - dependent tubule inhibition was examined by overexpression of a dominant negative mutant ( cort - actin \u0394 PHSH3 ) 85 . While expression of the wild type cortactin showed no effect , cortactin \u0394 PHSH3 expression restored W13 - induced tubules in Rac1 G12V - W56A expressing cells ( Fig . 5b ) . The same result was obtained by slencing cort - actin expression through siRNA transfection in cells expressing Rac1 G12V - W56A and treated with W13 ( Fig . 5c ) , demonstrating that Rac1 needs a functional cortactin to prevent PI ( 4 , 5 ) P 2 - induced tubulation . ROCK1 activity inhibits endocytic tubule formation downstream Rac1 . In vitro yeast two - hybrid experiments demonstrated that active Rac1 interacted with ROCK1 , whereas the active Rac1 F37A mutant was defective in such interaction 80 , though the functionality of this association has not been reported yet . ROCK1 is a Ser / Thr kinase that is activated after RhoA - GTP binding to its Rho - binding domain ( RBD ) due to the release of its autoinhibitory conformation 86 . To analyze the interaction between ROCK1 and the different Rac1 G12V mutants , we performed a pull - down assay incubating lysates from GFP - Rac1 G12V , GFP - Rac1 G12V - F37A , or GFP - Rac1 G12V - W56A expressing cells with GST - ROCK1 725 \u2013 1024 immobilized on Sepharose beads . This ROCK1 fragment contains the RBD and a N - terminal portion of its kinase domain that facilitates a proper conformation for RBD / Rho - GTP binding 87 . Western blot analysis showed that Rac1 G12V and Rac1 G12V - W56A were both pulled down by GST - ROCK1 725 \u2013 1024 , whereas Rac1 G12V - F37A did not ( Fig . 6a ) . Co - immunoprecipitation experiments showing interaction between Rac1 G12V - W56A and endogenous ROCK1 were performed as well , although , these results were not consistently reproduced probably due to weak and transient Rac1 - ROCK1 interaction and have not been included in this report . In addition , expression of Rac1 G12V induced the recruitment and co - localization of ROCK1 at the PM ( Fig . 6b ) . After binding to Rac1 - GTP , ROCK1 may be activated and become functional in PM domains where Rac1 is present . Supporting the hypothesis that ROCK1 is an effector involved in Rac1 - dependent tubule formation inhibi - tion , the specific ROCK1 inhibitor Y27632 impaired the inhibition of Rac1 G12V - W56A ( Fig . 6c ) . To further confirm ROCK1 involvement in tubule inhibition by the active Rac1 mutant , we silenced ROCK1 expression by siRNA in Rac1 G12V - W56A expressing COS1 cells . Downregulation of ROCK1 completely restored the tubules induced by W13 ( Fig . 6d ) . Although ROCK1 could inhibit tubulation by recruiting cortactin to the PM , this possibility was ruled out using Y27632 in Rac1 G12V - and Rac1 G12V - W56A - transfected cells . Cortactin translocation appears to be independent of ROCK1 activity ( Supplementary Fig . S3 ) , in agreement with others authors 79 , 84 , 88 , 89 . These results indicate that , in addition to cortactin translocation and actin polymerization at the PM , active Rac1impairs tubu - lation via ROCK1 activity . However , ROCK1 is a well - known RhoA effector 44 , 90 , and to date no functional relationship has been described with other GTPases . In order to exclude RhoA as the upstream activator of ROCK1 responsible for the inhibition of PI ( 4 , 5 ) P 2 - induced tubule formation , the outcome of RhoA activity on W13 - induced tubules was examined in COS1 cells by expressing the constitutively active ( GFP - RhoA G14V ) and inactive ( GFP - RhoA T19N ) RhoA mutants . The expression of GFP - RhoA T19N did not modify the percentage of cells presenting tubules neither in control nor W13 - treated cells ( Fig . 6e ) . In contrast , when we expressed GFP - RhoA G14V ( expected to activate ROCK1 ) , an important and significant increase in tubule - presenting cells was observed after W13 - treatment , instead of inhibition ( Fig . 6e ) . RhoA and Rac1 are mutual antagonists 44 , 91 , and the observed active RhoA - induced tubulation might be a consequence of endogenous Rac1 inhibition . Moreover , while GFP - Rac1 was present in W13 - induced tubules , GFP - RhoA G14V was almost absent ( Fig . 6f ) . Then , RhoA - induced ROCK1 activation takes www . nature . com / scientificreports / 9 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x place in different sites , precluding inhibition of tubule formation by RhoA activity . Together , these data suggest that Rac1 / ROCK1 , and not RhoA / ROCK1 , plays a key role in the inhibition of the endocytic tubule formation . Rac1 / ROCK1 - dependent actomyosin assembly inhibits tubulation . Although there is no reported or conclusive role for ROCK1 as an effector of active Rac1 , this protein controls actomyosin downstream of active RhoA . It is feasible , therefore , that by controlling myosin activation , Rac1 / ROCK1 interplay could stabilize actin polymerization at the specific sites where tubules should be induced . In turn , this may inhibit tubule formation by mechanical hindrance or by membrane tension increase . It is known that phosphorylation of myosin light chain protein ( MLC ) is critical for the interaction between myosin and F - actin , and hence for actomyosin generation 92 . Accordingly , MLC phosphatase ( MLCP ) dephosphorylates MLC and impairs actomyosin formation 93 . In fact , Figure 5 . Cortactin is involved in Rac1 - dependent tubule inhibition . ( a ) Vero cells grown on coverslips were transfected with GFP - Rac1 G12V , GFP - Rac1 G12V - F37A , or GFP - Rac1 G12V - W56A . After fixation , immunofluorescence was performed to detect endogenous cortactin using an antibody and the corresponding Alexa - 555 anti - mouse secondary antibody . Magnification insets show the presence or absence of cortactin at the PM ( bars , 10 \u00b5 m ) . ( b ) Percentage of COS1 cells presenting tubules , after W13 treatment , in cells transfected with cherry - Rac1 G12V - W56A alone or co - transfected with full length cortactin or with a cortactin mutant lacking the PH and SH3 domains , acting as a dominant negative . ( c ) The percentage of cells presenting more than 5 tubules per cell was quantified in untreated ( NT ) and W13 - treated cells co - transfected with GFP - Mem or GFP - Rac1 G12V - W56A and the non - specific ( NS ) or cortactin siRNAs ( 48 h ) . Downregulation of cortactin expression by its specific siRNA is shown by western blotting . The W13 - induced tubules and cortactin downregulation in GFP - Rac1 G12V - W56A and cortactin siRNA - transfected cells is shown by immunofluorescence as in ( a ) ( bars , 10 \u00b5 m ) . Mean values \u00b1 standard error of the mean ( SEM ) from three independent experiments are shown in all cases . Statistical significance between different conditions was determined by Student\u2019s t - test , * p < 0 . 05 , * * p < 0 . 01 . www . nature . com / scientificreports / 10 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x phosphorylation of the MLCP regulatory subunit MYPT1 by ROCK1 results in its inhibition 94 \u2013 96 . Given that ROCK1 activity may promote actomyosin , we hypothesized that actomyosin induced via Rac1 / ROCK1 97 , could be responsible for tubule inhibition . To investigate this hypothesis , myosin IIA localization was analyzed by immunofluorescence in untreated or Y27632 - treated Vero cells expressing GFP - Rac1 G12V , GFP - Rac1 G12V - F37A , or GFP - Rac1 G12V - W56A . Figure 7 shows that while GFP - Rac1 G12V - F37A did not significantly affect myosin IIA localization ( Fig . 7a ) , expression of GFP - Rac1 G12V or GFP - Rac1 G12V - W56A inhibited myosin IIA stress - fiber localization and enhanced its presence at Figure 6 . Rac1 / ROCK1 pathway prevents tubulation without altering PM localization of cortactin . ( a ) Lysates of COS1 cells expressing GFP - Rac1 G12V , GFP - Rac1 G12V - F37A , GFP - Rac1 G12V - W56A or the empty GFP - C1 vector ( EV ) were incubated with immobilized GST - ROCK1 725 \u2013 1024 in glutathione Sepharose beads , as described in the Materials and Methods . GFP - Fusion proteins present in the input lysates and in the bound fraction were detected by western blotting , using an anti - GFP antibody . A representative western blotting is shown ( n = 3 ) . ( b ) Vero cells were co - transfected with cherry - Rac1 G12V and GFP - ROCK1 . Confocal image insets show regions with high co - localization . ( c , d ) COS1 cells were transfected with GFP - Rac1 G12V - W56A and treated with or without W13 ( 20 min , 4 . 5 \u00b5 g / ml ) . The percentage of cells presenting tubules was determined after the inhibition of ROCK1 activity with Y27632 ( c ) or after the inhibition of ROCK1 expression by transfection with a specific siRNA ( d ) . Downregulation of ROCK1 expression by its specific siRNA is shown by western blotting . ( e ) The percentage of cells presenting tubules was determined in untreated and W13 - treated COS1 cells expressing GFP - RhoA T19N , GFP - RhoA G14V , or GFP - mem as a control . Mean values \u00b1 standard error of the mean ( SEM ) from three independent experiments are shown in all cases . Statistical significance between different conditions was determined by Student\u2019s t - test , * p < 0 . 05 , * * p < 0 . 01 , * * * p < 0 . 001 . ( f ) Vero cells co - transfected with Cherry - mem and GFP - Rac1 ( left panel ) or GFP - RhoA G14V ( right panel ) were treated with W13 . Images and insets show localization of GFP - Rac1 , but not of GFP - RhoA G14V , in the cherry - mem tubules . All images were acquired using the Leica TCS SP5 confocal microscope ( bars , 10 \u00b5 m ) . www . nature . com / scientificreports / 11 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x Figure 7 . ROCK1 participates in Rac1 induction of actomyosin , and myosin activity is required for Rac1 - mediated tubule inhibition . ( a , b , c ) By immunofluorescence , myosin IIA and F - actin were detected in starved Vero cells expressing GFP - Rac1 G12V - F37A ( a ) , GFP - Rac1 G12V ( b ) , or GFP - Rac1 G12V - W56A ( c ) after treatment with Y27632 ( 30 min , 25 \u00b5 M ) . The magnification insets show GFP - Rac1 , phalloidin - A555 , and myosin - IIA - A647 in transfected ( 1 ) and non - transfected cells ( 2 ) . In the cells expressing GFP - Rac1 G12V and GFP - Rac1 G12V - W56A , the images show the loss of stress fibers ( and consequently their staining with myosin II ) plus myosin recruitment to cortical F - actin ( arrow heads ) , which is reduced after treatment with Y27632 ( bars , 5 \u00b5 m ) . ( d ) The percentage of cells presenting tubules was determined in the presence or absence of the myosin inhibitor blebbistatin ( 30 min , 50 \u00b5 M ) . Mean values \u00b1 standard error of the mean ( SEM ) from three independent experiments is shown . Statistical significance between different conditions was determined by Student\u2019s t - test , * p < 0 . 05 . www . nature . com / scientificreports / 12 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x the cell periphery colocalizing with cortical actin ( Fig . 7b , c ) , being this effect more evident in GFP - Rac1 G12V than in GFP - Rac1 G12V - W56A expressing cells . Y27632 treatment inhibited the presence of myosin IIA in both stress fibers and cell periphery in all cells regardless whether they expressed the active Rac1 mutants ( Fig . 7 ) , consistent with the restitution of the W13 - induced tubules after Y27632 treatment in GFP - Rac1 G12V - W56A cells ( Fig . 6c ) . The participation of ROCK1 in myosin IIA localization at the leading edge of wound migrating cells has been previously demonstrated 98 and the results presented here further support the involvement of ROCK1 in Rac1 induction of actomyosin . Finally , to clarify the role of myosin activity in tubule inhibition by Rac1 , we pre - incubated cells expressing GFP - Rac1 G12V - W56A with the general myosin inhibitor blebbistatin before W13 - treatment . In this experiment , blebbistatin effectively restored the W13 - induced tubules in cells expressing GFP - Rac1 G12V - W56A ( Fig . 7d ) . In con - clusion , our data establishes ROCK1 , for the first time , as a novel downstream effector of Rac1 involved in the control of membrane dynamics via myosin regulation . Proposed model of the molecular machinery implicated in the dynamics of PI ( 4 , 5 ) P 2 - induced endocytic tubulation . Taken together , these data support the model summarized in Fig . 8 . The regula - tion of PI ( 4 , 5 ) P 2 levels in cholesterol rich PM ordered domains is crucial for membrane invagination , elonga - tion and fission , enabling the correct progression of CIE ( used for \u03b2 1 - integrin internalization ) . When PI ( 4 , 5 ) Figure 8 . Molecular machinery implicated in the biogenesis and inhibition of PI ( 4 , 5 ) P 2 - induced endocytic tubulation . An increase of PI ( 4 , 5 ) P 2 in PM ordered domains could induce the recruitment of several PI ( 4 , 5 ) P 2 - binding proteins to generate an incipient membrane deformation ( 1 ) . When Rac1 activity is low ( 2 ) , the invagination can be elongated by dyneins toward the center of the cell along microtubules . PI ( 4 , 5 ) P 2 accumulation , as well as the high degree of membrane curvature , could lead to the recruitment of dynamin or BAR - domain proteins , which in turn , could propagate and stabilize the tubule . By contrast , when Rac1 activity is high , tubulation process could be inhibited by either PLC activation ( reducing PIP 2 levels ) or cytoskeleton regulation ( inducing cortical actomyosin at the PM ) ( 3 ) . Rac1 appears to stimulate cortactin PM translocation and ROCK1 activity , thereby triggering cortical actin polymerization and association with myosin ( i . e . , actomyosin ) . The resulting over - activation of local actomyosin networks could potentially impede tubulation in one of two ways : ( i ) by forming a local barrier to increasing PM tension or by causing a steric hindrance that impedes the recruitment of tubulating proteins ( 4 ) ; or ( ii ) by generating mechanical forces needed to pinch off membrane invaginations more efficiently ( 5 ) . www . nature . com / scientificreports / 13 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x P 2 levels increase due to CaM inhibition , PIP5K overexpression , exogenous diC8 - PI ( 4 , 5 ) P 2 administration or Rac1 inhibition , a long tubular plasma membrane network is formed ( Fig . 8 , points 1 and 2 ) . This membrane process requires dynamin , dynein and MTs ( point 2 ) . The results presented above demonstrated that activation of Rac1 ( overexpression of Rac1 G12V ) inhibits the PI ( 4 , 5 ) P 2 - dependent tubular PM network formation by two main molecular mechanisms : [ i ] reducing PI ( 4 , 5 ) P 2 levels through PLC activation ; and [ ii ] inducing cortical F - actin mesh ( via cortactin ) and actomyosin ( via ROCK1 ) formation . Rac1 - mediated cortactin recruitment is insufficient for tubule inhibition , and requires actomyosin formation ( myosin activation ) . Rac1 - induced actomyosin forma - tion prevents PI ( 4 , 5 ) P 2 - dependent tubule establishment either by an actively actin - dependent tubule scission process ( Fig . 8 , point 5 ) or by generating a cortical actomyosin network that produces a local mechanical barrier or increases PM tension to impede membrane invagination ( Fig . 8 , point 4 ) . For the first time , we identify ROCK1 as a novel downstream effector of Rac1 acting in a RhoA - independent manner to regulate membrane dynamics during a tubular CIE . Both Rac1 and RhoA GTPases stimulate acto - myosin formation , but at different times and locations within the cell , and it is possible that both proteins share or compete for ROCK1 . These results suggest that Rac1 activation at the leading edge of migrating cells may be important to stabilize \u03b2 1 - integrin in the newly generated adhesions , preventing its internalization and turnover , and therefore facilitating cell movement as a result . Material and Methods Reagents and Antibodies . W13 hydrochloride , U73122 and latrunculin A were from Calbiochem ( Merck Millipore ) . Y27632 , blebbistatin , neomycin and dynasore were from Sigma - Aldrich . DiC8 - PI ( 4 , 5 ) P 2 was from Echelon Biosciences . Primary antibodies used were as follows : rabbit polyclonal anti - GFP and mouse anti - actin ( Abcam ) , mouse monoclonal anti - cortactin ( Upstate ) , rabbit polyclonal anti - PACSIN2 ( Abgent ) , rat monoclonal anti - \u03b2 1 - Integrin ( AIIB2 ) ( Damsky , C . H . , Developmental Studies Hybridoma Bank ) , mouse monoclonal anti - PI ( 4 , 5 ) P 2 ( Echelon Biosciences ) , mouse monoclonal anti - early endosomal antigen1 ( EEA1 ) ( BD Transduction Laboratories ) , rabbit polyclonal anti - Myosin Heavy Chain IIA ( Biolegend Inc . ) , rab - bit polyclonal anti - Clathrin heavy chain ( CHC ) antibody ( # PA5 - 25804 ) and transferrin - Alexa546 ( # T23364 ) from ThermoFisher Scientific , and mouse monoclonal anti - CHC antibody clone \u00d7 22 ( # MA1 - 065 , Affinity BioReagents ) . Phalloidin conjugated with TRITC or Alexa - 350 , secondary Alexa - labeled antibodies and ProtA - HRP were from Molecular Probes ( Invitrogen - Life Technologies ) . SiR actin ( SC006 ) was from Spirochrome . Secondary HRP - labeled antibodies , SDS \u2013 polyacrylamide gel electrophoresis ( PAGE ) and molecular weight markers were from Bio - Rad . Glutathione - Sepharose beads were purchased from GE Healthcare . Human ROCK1 ( 4390824 ) siRNA was from Ambion , Human dynamin ( s12097 ) siRNA was from Santa Cruz Biotechnologies , Human cortactin ( CTTN , GS2017 ) and Human Clathrin heavy chain ( CLTC , GS1213 ) Flexitube Gene Solution siRNAs were from Qiagen . Cell culture . African green monkey kidney fibroblast COS1 or Vero cells were grown in Dulbecco\u2019s modified Eagle\u2019s medium ( DMEM ) supplemented with 5 % ( v / v ) or 10 % fetal calf serum ( FCS ) respectively , pyruvic acid , antibiotics and glutamine . DMEM and FCS were purchased from Biological Industries . Plasmids and transfection . Plasmid encoding the constitutively active Rac1 mutant ( Rac1 G12V ) was kindly pro - vided by Dr Michiyuki Matsuda ( University of Kyoto ) 99 and subcloned into living color vectors ( Clontech ) . Rac1 F37A point mutation was introduced into pEGFPC1 - Rac1 G12V by polymerase chain reaction ( PCR ) with the following primers : 5 \u2032 - CAGATTCACCGGTTTTCCATCTACCATAACATTGGCAGAATAATTGTCAGCGACAGTAGGG - 3 \u2032 and 5 \u2032 - GGACGAGCTGTACAAGTCCCTCATGCAGGCCATCAAGTG - 3 \u2032 , using BsrGI and BstXI restriction sites ; Rac1 W56A point mutation was introduced into pEGFPC1 - Rac1 G12V by polymerase chain reaction ( PCR ) with the following primers : 5 \u2032 - CAATTATTCTGCCAATGTTATGGTAGATGGAAAGCCAGTGAATCTGGGCTTAGCG GATACAGCTGG - 3 \u2032 and 5 \u2032 - CAGTCACGATGAATTCTTACAACAGCAGGC - 3 \u2032 , using BstXI and EcoRI restric - tion sites . Both mutants were subcloned into mCherry vector ( Clontech ) . GFP - and Cherry - mem are fusion proteins that contain the N - terminal amino acids of GAP - 43 and a GFP and mCherry fluorescent protein , respec - tively . The GAP - 43 fragment contains a signal for post - translational palmitoylation of cysteines 3 and 4 that tar - gets fusion protein to cellular membranes . Plasmid encoding dominant negative dynamin ( dynamin - K44A ) was obtained from ATCC ( MBA - 93 ) . L10 - GFP - Inp54p and L10 - GFP - Inp54p D281A were cloned from plasmids pro - vided by Dr Tobias Meyer through Addgene ( # 20155 and # 20156 respectively ) 100 . Firstly , a peptide containing the N - terminal 10 residues of Lck ( L10 ) was fused to the N - terminal GFP encoding sequence ( L10 - EGFP - C1 ) ; then , vectors from addgene were digested with EcoRI and BamHI and inserted into L10 - EGFP - C1 vector . Cortactin - WT and Cortactin - \u0394 PHSH3 were kindly provided by Dr . Thomas Parsons . Plasmid encoding GST - ROCK1 725 \u2013 1024 was generated by PCR using the primers 5 \u2032 - GACCGGTGGATCCCGGGCTGTATTAGCTTTCTTTCTATC - 3 \u2032 and 5 \u2032 - CACATGGTCCTGCTGGAGTTCGTG - 3 \u2032 with pECFP - ROCK1 , kindly provided by Dr Gareth Jones 101 , as a tem - plate . The resulting PCR product was introduced into CFP - N1 vector using XhoI and XmaI restriction sites , and then it was subcloned into pGEX - 4T - 2 for the expression in bacteria using BamHI and XmaI restriction sites . GFP - RhoA constitutively active ( G14V ) and dominant negative ( T19N ) were kindly provided by Michael Way ( Cancer Research UK , London , UK ) . COS1 and Vero cells were transfected with DNA using Effectene ( QIAGEN ) or GenJet ( Signagen ) , and transfected with combined DNA and siRNA transfection using RNAiMAX ( Invitrogen - Life Technologies ) . Cells were used for experiments 24 h after DNA transient expression or 48 \u2013 72 h after siRNA transfection . Immunofluorescence staining . Cells grown on coverslips were fixed with freshly prepared 4 % paraform - aldehyde ( PFA ) in cytoskeleton buffer ( CB ; 10 mM MES pH6 . 1 , 138 mM KCl , 3 mM MgCl2 , 2 mM EGTA ) at 37 \u00b0C for 15 min and permeabilized with 0 . 5 % Triton X - 100 in CB at room temperature for 3 min . After 5 - min incubation with blocking solution ( TBST , 2 % BSA ) , coverslips were incubated with the primary antibody diluted www . nature . com / scientificreports / 14 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x in blocking solution for 50 min at room temperature , washed intensively and then incubated with the adequate secondary antibodies labeled with Alexa488 , Alexa555 or Alexa647 . After staining , the coverslips were mounted in Mowiol ( Calbiochem , Merck ) . The images were acquired using a Leica TCS SP5 laser scanning confocal microscope ( Leica Microsystems Heidelberg GmbH ) equipped with DMI6000 inverted microscope , Argon ( 458 / 476 / 488 / 514 ) , diode pumped solid state ( 561 nm ) and HeNe ( 633 ) lasers . GFP , TRITC or Alexa - 555 and Alexa - 647 images were acquired sequentially using 488 , 561 and 633 laser lines , acoustic optical beam splitter ( AOBS ) as beam splitter , and emission detection ranges 500 \u2013 555 , 571 \u2013 625 and 640 \u2013 700 nm , respectively . STED confocal images were acquired using a Leica TCS SP8 . Final analysis of all images was performed using IMAGEJ software . \u03b2 1 - integrin internalization analysis . COS1 cells grown on coverslips were tempered to 4 \u00b0C to defuse endocytosis and then were incubated with anti - \u03b2 1 - integrin antibody and transferrin - TRITC for 30 min at 4 \u00b0C . After washing the unbound antibody and transferrin excess with PBS , cells were incubated at 37 \u00b0C for 5 , 10 and 20 min under the corresponding treatment . Cells were washed twice with PBS at 4 \u00b0C and then were subjected to a surface acid wash ( 0 . 5 % glacial acetic acid , 0 . 5 M NaCl , pH 3 . 0 ) at 4 \u00b0C for 2 min . After fixation with freshly prepared 4 % PFA at 37 \u00b0C for 15 min , immunostaining of the antigen - antibody complexes was performed as described above . Images were acquired along the Z - axis , in order to cover the whole cell , using a Leica TCS SP5 laser scanning confocal microscope ( Leica Microsystems Heidelberg GmbH ) equipped with DMI6000 inverted microscope . To determine the amount of internalized \u03b2 1 - integrin and transferrin , fluorescence intensity was normalized per cell area . Pull - down assay . Cleared TGH ( 1 % Triton X - 100 , 10 % glycerol , 50 mM Hepes with proteases inhibitors and 50 mM NaCl ) lysates of COS1 cells , transiently expressing GFP - tagged Rac1 constructs , were split and incubated for 2 h at 4 \u00b0C with GST - ROCK1 - 725 - 1024 bound to gluthation - Sepharose beads . The unbound fraction was col - lected by centrifugation , and the remaining bound fraction was washed twice in lysis buffer supplemented with 150 mM NaCl and then once without NaCl . The total of the bound fraction was resolved by electrophoresis , and the proteins of interest were detected by western blotting . References 1 . Goldstein , J . L . , Anderson , R . G . & Brown , M . S . Coated pits , coated vesicles , and receptor - mediated endocytosis . Nature 279 , 679 \u2013 685 ( 1979 ) . 2 . Kirkham , M . & Parton , R . G . Clathrin - independent endocytosis : new insights into caveolae and non - caveolar lipid raft carriers . Biochimica et biophysica acta 1745 , 273 \u2013 286 , doi : 10 . 1016 / j . bbamcr . 2005 . 06 . 002 ( 2005 ) . 3 . Le Roy , C . & Wrana , J . L . Signaling and endocytosis : a team effort for cell migration . Developmental cell 9 , 167 \u2013 168 , doi : 10 . 1016 / j . devcel . 2005 . 07 . 007 ( 2005 ) . 4 . Sandvig , K . , Pust , S . , Skotland , T . & van Deurs , B . Clathrin - independent endocytosis : mechanisms and function . Current opinion in cell biology 23 , 413 \u2013 420 , doi : 10 . 1016 / j . ceb . 2011 . 03 . 007 ( 2011 ) . 5 . Weinberg , J . & Drubin , D . G . Clathrin - mediated endocytosis in budding yeast . Trends in cell biology 22 , 1 \u2013 13 , doi : 10 . 1016 / j . tcb . 2011 . 09 . 001 ( 2012 ) . 6 . Schmid , S . L . Clathrin - mediated endocytosis : a universe of new questions . Molecular biology of the cell 21 , 3818 \u2013 3819 , doi : 10 . 1091 / mbc . E10 - 05 - 0386 ( 2010 ) . 7 . Montesano , R . , Roth , J . , Robert , A . & Orci , L . Non - coated membrane invaginations are involved in binding and internalization of cholera and tetanus toxins . Nature 296 , 651 \u2013 653 ( 1982 ) . 8 . Sandvig , K . & van Deurs , B . Selective modulation of the endocytic uptake of ricin and fluid phase markers without alteration in transferrin endocytosis . The Journal of biological chemistry 265 , 6382 \u2013 6388 ( 1990 ) . 9 . Lundmark , R . et al . The GTPase - activating protein GRAF1 regulates the CLIC / GEEC endocytic pathway . Current biology : CB 18 , 1802 \u2013 1808 , doi : 10 . 1016 / j . cub . 2008 . 10 . 044 ( 2008 ) . 10 . Brameshuber , M . et al . Imaging of mobile long - lived nanoplatforms in the live cell plasma membrane . J Biol Chem 285 , 41765 \u2013 41771 , doi : 10 . 1074 / jbc . M110 . 182121 ( 2010 ) . 11 . Lakshminarayan , R . et al . Galectin - 3 drives glycosphingolipid - dependent biogenesis of clathrin - independent carriers . Nat Cell Biol 16 , 595 \u2013 606 , doi : 10 . 1038 / ncb2970 ( 2014 ) . 12 . Cheng , Z . J . et al . Distinct mechanisms of clathrin - independent endocytosis have unique sphingolipid requirements . Molecular biology of the cell 17 , 3197 \u2013 3210 , doi : 10 . 1091 / mbc . E05 - 12 - 1101 ( 2006 ) . 13 . Nimmervoll , B . et al . Cell surface localised Hsp70 is a cancer specific regulator of clathrin - independent endocytosis . FEBS Lett 589 , 2747 \u2013 2753 , doi : 10 . 1016 / j . febslet . 2015 . 07 . 037 ( 2015 ) . 14 . Nabi , I . R . & Le , P . U . Caveolae / raft - dependent endocytosis . The Journal of cell biology 161 , 673 \u2013 677 , doi : 10 . 1083 / jcb . 200302028 ( 2003 ) . 15 . Vidal - Quadras , M . et al . Rac1 and calmodulin interactions modulate dynamics of ARF6 - dependent endocytosis . Traffic 12 , 1879 \u2013 1896 , doi : 10 . 1111 / j . 1600 - 0854 . 2011 . 01274 . x ( 2011 ) . 16 . Day , C . A . et al . Microtubule motors power plasma membrane tubulation in clathrin - independent endocytosis . Traffic 16 , 572 \u2013 590 , doi : 10 . 1111 / tra . 12269 ( 2015 ) . 17 . Romer , W . et al . Shiga toxin induces tubular membrane invaginations for its uptake into cells . Nature 450 , 670 \u2013 675 , doi : 10 . 1038 / nature05996 ( 2007 ) . 18 . Naslavsky , N . , Weigert , R . & Donaldson , J . G . Convergence of non - clathrin - and clathrin - derived endosomes involves Arf6 inactivation and changes in phosphoinositides . Molecular biology of the cell 14 , 417 \u2013 431 , doi : 10 . 1091 / mbc . 02 - 04 - 0053 ( 2003 ) . 19 . Ewers , H . et al . GM1 structure determines SV40 - induced membrane invagination and infection . Nature cell biology 12 , 11 \u2013 18 ; sup pp 11 \u2013 12 , doi : 10 . 1038 / ncb1999 ( 2010 ) . 20 . Suetsugu , S . , Kurisu , S . & Takenawa , T . Dynamic shaping of cellular membranes by phospholipids and membrane - deforming proteins . Physiol Rev 94 , 1219 \u2013 1248 , doi : 10 . 1152 / physrev . 00040 . 2013 ( 2014 ) . 21 . Kwiatkowska , K . One lipid , multiple functions : how various pools of PI ( 4 , 5 ) P ( 2 ) are created in the plasma membrane . Cell Mol Life Sci 67 , 3927 \u2013 3946 , doi : 10 . 1007 / s00018 - 010 - 0432 - 5 ( 2010 ) . 22 . Tolias , K . F . et al . Type Ialpha phosphatidylinositol - 4 - phosphate 5 - kinase mediates Rac - dependent actin assembly . Current biology : CB 10 , 153 \u2013 156 ( 2000 ) . 23 . van den Bout , I . & Divecha , N . PIP5K - driven PtdIns ( 4 , 5 ) P2 synthesis : regulation and cellular functions . J Cell Sci 122 , 3837 \u2013 3850 , doi : 10 . 1242 / jcs . 056127 ( 2009 ) . www . nature . com / scientificreports / 15 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x 24 . Ferguson , S . M . & De Camilli , P . Dynamin , a membrane - remodelling GTPase . Nat Rev Mol Cell Biol 13 , 75 \u2013 88 , doi : 10 . 1038 / nrm3266 ( 2012 ) . 25 . Zoncu , R . et al . Loss of endocytic clathrin - coated pits upon acute depletion of phosphatidylinositol 4 , 5 - bisphosphate . Proceedings of the National Academy of Sciences of the United States of America 104 , 3793 \u2013 3798 , doi : 10 . 1073 / pnas . 0611733104 ( 2007 ) . 26 . Cremona , O . et al . Essential role of phosphoinositide metabolism in synaptic vesicle recycling . Cell 99 , 179 \u2013 188 ( 1999 ) . 27 . Sun , Y . , Carroll , S . , Kaksonen , M . , Toshima , J . Y . & Drubin , D . G . PtdIns ( 4 , 5 ) P2 turnover is required for multiple stages during clathrin - and actin - dependent endocytic internalization . The Journal of cell biology 177 , 355 \u2013 367 , doi : 10 . 1083 / jcb . 200611011 ( 2007 ) . 28 . Antonescu , C . N . , Aguet , F . , Danuser , G . & Schmid , S . L . Phosphatidylinositol - ( 4 , 5 ) - bisphosphate regulates clathrin - coated pit initiation , stabilization , and size . Molecular biology of the cell 22 , 2588 \u2013 2600 , doi : 10 . 1091 / mbc . E11 - 04 - 0362 ( 2011 ) . 29 . Lamaze , C . et al . Interleukin 2 receptors and detergent - resistant membrane domains define a clathrin - independent endocytic pathway . Molecular cell 7 , 661 \u2013 671 ( 2001 ) . 30 . Grassart , A . , Dujeancourt , A . , Lazarow , P . B . , Dautry - Varsat , A . & Sauvonnet , N . Clathrin - independent endocytosis used by the IL - 2 receptor is regulated by Rac1 , Pak1 and Pak2 . EMBO reports 9 , 356 \u2013 362 , doi : 10 . 1038 / embor . 2008 . 28 ( 2008 ) . 31 . Nobes , C . & Marsh , M . Dendritic cells : new roles for Cdc42 and Rac in antigen uptake ? Current biology : CB 10 , R739 \u2013 741 ( 2000 ) . 32 . West , M . A . , Prescott , A . R . , Eskelinen , E . L . , Ridley , A . J . & Watts , C . Rac is required for constitutive macropinocytosis by dendritic cells but does not control its downregulation . Current biology : CB 10 , 839 \u2013 848 ( 2000 ) . 33 . Kumari , S . & Mayor , S . ARF1 is directly involved in dynamin - independent endocytosis . Nature cell biology 10 , 30 \u2013 41 , doi : 10 . 1038 / ncb1666 ( 2008 ) . 34 . Sabharanjak , S . , Sharma , P . , Parton , R . G . & Mayor , S . GPI - anchored proteins are delivered to recycling endosomes via a distinct cdc42 - regulated , clathrin - independent pinocytic pathway . Developmental cell 2 , 411 \u2013 423 ( 2002 ) . 35 . Grassart , A . et al . Pak1 phosphorylation enhances cortactin - N - WASP interaction in clathrin - caveolin - independent endocytosis . Traffic 11 , 1079 \u2013 1091 , doi : 10 . 1111 / j . 1600 - 0854 . 2010 . 01075 . x ( 2010 ) . 36 . Cao , H . et al . Cortactin is a component of clathrin - coated pits and participates in receptor - mediated endocytosis . Mol Cell Biol 23 , 2162 \u2013 2170 ( 2003 ) . 37 . Sauvonnet , N . , Dujeancourt , A . & Dautry - Varsat , A . Cortactin and dynamin are required for the clathrin - independent endocytosis of gammac cytokine receptor . The Journal of cell biology 168 , 155 \u2013 163 ( 2005 ) . 38 . Missirlis , D . The effect of substrate elasticity and actomyosin contractility on different forms of endocytosis . PloS one 9 , e96548 , doi : 10 . 1371 / journal . pone . 0096548 ( 2014 ) . 39 . Collaco , A . , Jakab , R . , Hegan , P . , Mooseker , M . & Ameen , N . Alpha - AP - 2 directs myosin VI - dependent endocytosis of cystic fibrosis transmembrane conductance regulator chloride channels in the intestine . The Journal of biological chemistry 285 , 17177 \u2013 17187 , doi : 10 . 1074 / jbc . M110 . 127613 ( 2010 ) . 40 . Chandrasekar , I . et al . Nonmuscle myosin II is a critical regulator of clathrin - mediated endocytosis . Traffic 15 , 418 \u2013 432 , doi : 10 . 1111 / tra . 12152 ( 2014 ) . 41 . Levayer , R . , Pelissier - Monier , A . & Lecuit , T . Spatial regulation of Dia and Myosin - II by RhoGEF2 controls initiation of E - cadherin endocytosis during epithelial morphogenesis . Nature cell biology 13 , 529 \u2013 540 , doi : 10 . 1038 / ncb2224 ( 2011 ) . 42 . Gauthier , N . C . , Masters , T . A . & Sheetz , M . P . Mechanical feedback between membrane tension and dynamics . Trends in cell biology 22 , 527 \u2013 535 , doi : 10 . 1016 / j . tcb . 2012 . 07 . 005 ( 2012 ) . 43 . Lee , D . M . & Harris , T . J . An Arf - GEF regulates antagonism between endocytosis and the cytoskeleton for Drosophila blastoderm development . Current biology : CB 23 , 2110 \u2013 2120 , doi : 10 . 1016 / j . cub . 2013 . 08 . 058 ( 2013 ) . 44 . Guilluy , C . , Garcia - Mata , R . & Burridge , K . Rho protein crosstalk : another social network ? Trends in cell biology 21 , 718 \u2013 726 , doi : 10 . 1016 / j . tcb . 2011 . 08 . 002 ( 2011 ) . 45 . Halstead , J . R . et al . Rac controls PIP5K localisation and PtdIns ( 4 , 5 ) P ( 2 ) synthesis , which modulates vinculin localisation and neurite dynamics . J Cell Sci 123 , 3535 \u2013 3546 , doi : 10 . 1242 / jcs . 062679 ( 2010 ) . 46 . Bosco , E . E . , Mulloy , J . C . & Zheng , Y . Rac1 GTPase : a \u201cRac\u201d of all trades . Cell Mol Life Sci 66 , 370 \u2013 374 , doi : 10 . 1007 / s00018 - 008 - 8552 - x ( 2009 ) . 47 . Bustelo , X . R . , Sauzeau , V . & Berenjeno , I . M . GTP - binding proteins of the Rho / Rac family : regulation , effectors and functions in vivo . Bioessays 29 , 356 \u2013 370 , doi : 10 . 1002 / bies . 20558 ( 2007 ) . 48 . Etienne - Manneville , S . & Hall , A . Rho GTPases in cell biology . Nature 420 , 629 \u2013 635 , doi : 10 . 1038 / nature01148 ( 2002 ) . 49 . Heasman , S . J . & Ridley , A . J . Mammalian Rho GTPases : new insights into their functions from in vivo studies . Nat Rev Mol Cell Biol 9 , 690 \u2013 701 , doi : 10 . 1038 / nrm2476 ( 2008 ) . 50 . Brzeska , H . , Szczepanowska , J . , Matsumura , F . & Korn , E . D . Rac - induced increase of phosphorylation of myosin regulatory light chain in HeLa cells . Cell Motil Cytoskeleton 58 , 186 \u2013 199 , doi : 10 . 1002 / cm . 20009 ( 2004 ) . 51 . Pasapera , A . M . et al . Rac1 - dependent phosphorylation and focal adhesion recruitment of myosin IIA regulates migration and mechanosensing . Current biology : CB 25 , 175 \u2013 186 , doi : 10 . 1016 / j . cub . 2014 . 11 . 043 ( 2015 ) . 52 . del Pozo , M . A . et al . Integrins regulate Rac targeting by internalization of membrane domains . Science 303 , 839 \u2013 842 , doi : 10 . 1126 / science . 1092571 ( 2004 ) . 53 . del Pozo , M . A . , Price , L . S . , Alderson , N . B . , Ren , X . D . & Schwartz , M . A . Adhesion to the extracellular matrix regulates the coupling of the small GTPase Rac to its effector PAK . Embo J 19 , 2008 \u2013 2014 , doi : 10 . 1093 / emboj / 19 . 9 . 2008 ( 2000 ) . 54 . Yron , I . et al . Integrin - dependent tyrosine phosphorylation and growth regulation by Vav . Cell Adhes Commun 7 , 1 \u2013 11 ( 1999 ) . 55 . Miao , H . et al . Differential regulation of Rho GTPases by beta1 and beta3 integrins : the role of an extracellular domain of integrin in intracellular signaling . J Cell Sci 115 , 2199 \u2013 2206 ( 2002 ) . 56 . Jacquemet , G . et al . Rac1 is deactivated at integrin activation sites through an IQGAP1 - filamin - A - RacGAP1 pathway . J Cell Sci 126 , 4121 \u2013 4135 , doi : 10 . 1242 / jcs . 121988 ( 2013 ) . 57 . Tebar , F . et al . Calmodulin regulates intracellular trafficking of epidermal growth factor receptor and the MAPK signaling pathway . Mol Biol Cell 13 , 2057 \u2013 2068 , doi : 01 - 12 - 0571 ( 2002 ) . 58 . Llado , A . et al . Protein kinaseCdelta - calmodulin crosstalk regulates epidermal growth factor receptor exit from early endosomes . Mol Biol Cell 15 , 4877 \u2013 4891 , doi : 10 . 1091 / mbc . E04 - 02 - 0127 ( 2004 ) . 59 . Llado , A . et al . Protein kinase Cdelta and calmodulin regulate epidermal growth factor receptor recycling from early endosomes through Arp2 / 3 complex and cortactin . Mol Biol Cell 19 , 17 \u2013 29 , doi : 10 . 1091 / mbc . E07 - 05 - 0411 ( 2008 ) . 60 . Apodaca , G . , Enrich , C . & Mostov , K . E . The calmodulin antagonist , W - 13 , alters transcytosis , recycling , and the morphology of the endocytic pathway in Madin - Darby canine kidney cells . J Biol Chem 269 , 19005 \u2013 19013 ( 1994 ) . 61 . Shinozaki - Narikawa , N . , Kodama , T . & Shibasaki , Y . Cooperation of phosphoinositides and BAR domain proteins in endosomal tubulation . Traffic 7 , 1539 \u2013 1550 , doi : 10 . 1111 / j . 1600 - 0854 . 2006 . 00480 . x ( 2006 ) . 62 . Ozaki , S . , DeWald , D . B . , Shope , J . C . , Chen , J . & Prestwich , G . D . Intracellular delivery of phosphoinositides and inositol phosphates using polyamine carriers . Proc Natl Acad Sci USA 97 , 11286 \u2013 11291 , doi : 10 . 1073 / pnas . 210197897 ( 2000 ) . 63 . Casas , J . et al . Phosphatidylinositol 4 , 5 - bisphosphate anchors cytosolic group IVA phospholipase A2 to perinuclear membranes and decreases its calcium requirement for translocation in live cells . Mol Biol Cell 17 , 155 \u2013 162 , doi : 10 . 1091 / mbc . E05 - 06 - 0545 ( 2006 ) . 64 . Sweitzer , S . M . & Hinshaw , J . E . Dynamin undergoes a GTP - dependent conformational change causing vesiculation . Cell 93 , 1021 \u2013 1029 ( 1998 ) . www . nature . com / scientificreports / 16 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x 65 . Stowell , M . H . , Marks , B . , Wigge , P . & McMahon , H . T . Nucleotide - dependent conformational changes in dynamin : evidence for a mechanochemical molecular spring . Nature cell biology 1 , 27 \u2013 32 , doi : 10 . 1038 / 8997 ( 1999 ) . 66 . Praefcke , G . J . & McMahon , H . T . The dynamin superfamily : universal membrane tubulation and fission molecules ? Nat Rev Mol Cell Biol 5 , 133 \u2013 147 , doi : 10 . 1038 / nrm1313 ( 2004 ) . 67 . Meinecke , M . et al . Cooperative recruitment of dynamin and BIN / amphiphysin / Rvs ( BAR ) domain - containing proteins leads to GTP - dependent membrane scission . The Journal of biological chemistry 288 , 6651 \u2013 6661 , doi : 10 . 1074 / jbc . M112 . 444869 ( 2013 ) . 68 . Roux , A . , Uyhazi , K . , Frost , A . & De Camilli , P . GTP - dependent twisting of dynamin implicates constriction and tension in membrane fission . Nature 441 , 528 \u2013 531 , doi : 10 . 1038 / nature04718 ( 2006 ) . 69 . de Kreuk , B . J . et al . The F - BAR domain protein PACSIN2 associates with Rac1 and regulates cell spreading and migration . J Cell Sci 124 , 2375 \u2013 2388 , doi : 10 . 1242 / jcs . 080630 ( 2011 ) . 70 . Dharmalingam , E . et al . F - BAR proteins of the syndapin family shape the plasma membrane and are crucial for neuromorphogenesis . J Neurosci 29 , 13315 \u2013 13327 , doi : 10 . 1523 / JNEUROSCI . 3973 - 09 . 2009 ( 2009 ) . 71 . Modregger , J . , Ritter , B . , Witter , B . , Paulsson , M . & Plomann , M . All three PACSIN isoforms bind to endocytic proteins and inhibit endocytosis . J Cell Sci 113 ( Pt 24 ) , 4511 \u2013 4521 ( 2000 ) . 72 . Senju , Y . , Itoh , Y . , Takano , K . , Hamada , S . & Suetsugu , S . Essential role of PACSIN2 / syndapin - II in caveolae membrane sculpting . J Cell Sci 124 , 2032 \u2013 2040 , doi : 10 . 1242 / jcs . 086264 ( 2011 ) . 73 . Hansen , C . G . , Howard , G . & Nichols , B . J . Pacsin 2 is recruited to caveolae and functions in caveolar biogenesis . J Cell Sci 124 , 2777 \u2013 2785 , doi : 10 . 1242 / jcs . 084319 ( 2011 ) . 74 . Navarro - Lerida , I . et al . A palmitoylation switch mechanism regulates Rac1 function and membrane organization . Embo J 31 , 534 \u2013 551 , doi : 10 . 1038 / emboj . 2011 . 446 ( 2012 ) . 75 . Johnson , C . M . , Chichili , G . R . & Rodgers , W . Compartmentalization of phosphatidylinositol 4 , 5 - bisphosphate signaling evidenced using targeted phosphatases . The Journal of biological chemistry 283 , 29920 \u2013 29928 , doi : 10 . 1074 / jbc . M805921200 ( 2008 ) . 76 . Tsai , F . D . & Philips , M . R . Rac1 gets fattier . Embo J 31 , 517 \u2013 518 , doi : 10 . 1038 / emboj . 2011 . 481 ( 2012 ) . 77 . Jezyk , M . R . et al . Crystal structure of Rac1 bound to its effector phospholipase C - beta2 . Nat Struct Mol Biol 13 , 1135 \u2013 1140 , doi : 10 . 1038 / nsmb1175 ( 2006 ) . 78 . Illenberger , D . et al . Stimulation of phospholipase C - beta2 by the Rho GTPases Cdc42Hs and Rac1 . Embo J 17 , 6241 \u2013 6249 , doi : 10 . 1093 / emboj / 17 . 21 . 6241 ( 1998 ) . 79 . Weed , S . A . , Du , Y . & Parsons , J . T . Translocation of cortactin to the cell periphery is mediated by the small GTPase Rac1 . J Cell Sci 111 ( Pt 16 ) , 2433 \u2013 2443 ( 1998 ) . 80 . Lamarche , N . et al . Rac and Cdc42 induce actin polymerization and G1 cell cycle progression independently of p65PAK and the JNK / SAPK MAP kinase cascade . Cell 87 , 519 \u2013 529 ( 1996 ) . 81 . Hattula , K . et al . Characterization of the Rab8 - specific membrane traffic route linked to protrusion formation . J Cell Sci 119 , 4866 \u2013 4877 , doi : 10 . 1242 / jcs . 03275 ( 2006 ) . 82 . Brown , F . D . , Rozelle , A . L . , Yin , H . L . , Balla , T . & Donaldson , J . G . Phosphatidylinositol 4 , 5 - bisphosphate and Arf6 - regulated membrane traffic . The Journal of cell biology 154 , 1007 \u2013 1017 ( 2001 ) . 83 . Zugaza , J . L . , Caloca , M . J . & Bustelo , X . R . Inverted signaling hierarchy between RAS and RAC in T - lymphocytes . Oncogene 23 , 5823 \u2013 5833 , doi : 10 . 1038 / sj . onc . 1207768 ( 2004 ) . 84 . D\u2019Souza - Schorey , C . , Boshans , R . L . , McDonough , M . , Stahl , P . D . & Van Aelst , L . A role for POR1 , a Rac1 - interacting protein , in ARF6 - mediated cytoskeletal rearrangements . Embo J 16 , 5445 \u2013 5454 , doi : 10 . 1093 / emboj / 16 . 17 . 5445 ( 1997 ) . 85 . Weed , S . A . et al . Cortactin localization to sites of actin assembly in lamellipodia requires interactions with F - actin and the Arp2 / 3 complex . The Journal of cell biology 151 , 29 \u2013 40 ( 2000 ) . 86 . Riento , K . & Ridley , A . J . Rocks : multifunctional kinases in cell behaviour . Nat Rev Mol Cell Biol 4 , 446 \u2013 456 , doi : 10 . 1038 / nrm1128 ( 2003 ) . 87 . Dvorsky , R . , Blumenstein , L . , Vetter , I . R . & Ahmadian , M . R . Structural insights into the interaction of ROCKI with the switch regions of RhoA . The Journal of biological chemistry 279 , 7098 \u2013 7104 , doi : 10 . 1074 / jbc . M311911200 ( 2004 ) . 88 . Van Aelst , L . , Joneson , T . & Bar - Sagi , D . Identification of a novel Rac1 - interacting protein involved in membrane ruffling . Embo J 15 , 3778 \u2013 3786 ( 1996 ) . 89 . Bourguignon , L . Y . , Singleton , P . A . & Diedrich , F . Hyaluronan - CD44 interaction with Rac1 - dependent protein kinase N - gamma promotes phospholipase Cgamma1 activation , Ca ( 2 + ) signaling , and cortactin - cytoskeleton function leading to keratinocyte adhesion and differentiation . The Journal of biological chemistry 279 , 29654 \u2013 29669 , doi : 10 . 1074 / jbc . M403608200 ( 2004 ) . 90 . Leung , T . , Chen , X . Q . , Manser , E . & Lim , L . The p160 RhoA - binding kinase ROK alpha is a member of a kinase family and is involved in the reorganization of the cytoskeleton . Molecular and cellular biology 16 , 5313 \u2013 5327 ( 1996 ) . 91 . Sander , E . E . , ten Klooster , J . P . , van Delft , S . & van der Kammen , R . A . & Collard , J . G . Rac downregulates Rho activity : reciprocal balance between both GTPases determines cellular morphology and migratory behavior . The Journal of cell biology 147 , 1009 \u2013 1022 ( 1999 ) . 92 . Vicente - Manzanares , M . , Ma , X . , Adelstein , R . S . & Horwitz , A . R . Non - muscle myosin II takes centre stage in cell adhesion and migration . Nat Rev Mol Cell Biol 10 , 778 \u2013 790 , doi : 10 . 1038 / nrm2786 ( 2009 ) . 93 . Matsumura , F . & Hartshorne , D . J . Myosin phosphatase target subunit : Many roles in cell function . Biochem Biophys Res Commun 369 , 149 \u2013 156 , doi : 10 . 1016 / j . bbrc . 2007 . 12 . 090 ( 2008 ) . 94 . Kimura , K . et al . Regulation of myosin phosphatase by Rho and Rho - associated kinase ( Rho - kinase ) . Science 273 , 245 \u2013 248 ( 1996 ) . 95 . Khromov , A . , Choudhury , N . , Stevenson , A . S . , Somlyo , A . V . & Eto , M . Phosphorylation - dependent autoinhibition of myosin light chain phosphatase accounts for Ca2 + sensitization force of smooth muscle contraction . The Journal of biological chemistry 284 , 21569 \u2013 21579 , doi : 10 . 1074 / jbc . M109 . 019729 ( 2009 ) . 96 . Julian , L . & Olson , M . F . Rho - associated coiled - coil containing kinases ( ROCK ) : structure , regulation , and functions . Small GTPases 5 , e29846 , doi : 10 . 4161 / sgtp . 29846 ( 2014 ) . 97 . Amano , M . et al . Phosphorylation and activation of myosin by Rho - associated kinase ( Rho - kinase ) . The Journal of biological chemistry 271 , 20246 \u2013 20249 ( 1996 ) . 98 . Sandquist , J . C . & Means , A . R . The C - terminal tail region of nonmuscle myosin II directs isoform - specific distribution in migrating cells . Molecular biology of the cell 19 , 5156 \u2013 5167 , doi : 10 . 1091 / mbc . E08 - 05 - 0533 ( 2008 ) . 99 . Nakaya , M . , Kitano , M . , Matsuda , M . & Nagata , S . Spatiotemporal activation of Rac1 for engulfment of apoptotic cells . Proceedings of the National Academy of Sciences of the United States of America 105 , 9198 \u2013 9203 , doi : 10 . 1073 / pnas . 0803677105 ( 2008 ) . 100 . Suh , B . C . , Inoue , T . , Meyer , T . & Hille , B . Rapid chemically induced changes of PtdIns ( 4 , 5 ) P2 gate KCNQ ion channels . Science 314 , 1454 \u2013 1457 , doi : 10 . 1126 / science . 1131163 ( 2006 ) . 101 . Shea , K . F . , Wells , C . M . , Garner , A . P . & Jones , G . E . ROCK1 and LIMK2 interact in spread but not blebbing cancer cells . PloS one 3 , e3398 , doi : 10 . 1371 / journal . pone . 0003398 ( 2008 ) . Acknowledgements This research was supported by grants BFU2012 \u2013 38259 and BFU2015 \u2013 66785 - P from Ministerio de Economia y Competitividad of Spain to F . T . ; Consolider - Ingenio from Ministerio de Innovaci\u00f3n , Ciencia y Tecnolog\u00eda of Spain to C . E . and A . P . ; BFU2011 \u2013 23745 to A . P . ( MICINN ) . D . S - C . and A . C . were recipient of FI fellowship www . nature . com / scientificreports / 17 SCIEntIFIC REPORts | 7 : 6866 | DOI : 10 . 1038 / s41598 - 017 - 07130 - x ( Generalitat de Catalunya ) . We thank Maria Calvo , Anna Bosch and Elisenda Coll for assistance in the confocal imaging ( Unitat Microscopia \u00d2ptica Avan\u00e7ada , Centres Cient\u00edfics i Tecnol\u00f2gics , Universitat de Barcelona ) ; Timo Zimmermann , Raquel Garc\u00eda , Xavier Sanjuan and Arrate Mallabiabarrena for assistance in STED microscopy ( Advanced Light Microscopy Unit at the Centre for Genomic Regulation ( CRG ) in Barcelona ) and Maria Molinos for technical assistance . We are also grateful to Salvador Soriano for their comments and critical reading of the manuscript . Author Contributions D . S . - C . , A . C . and M . V . - Q . conducted most experiments and analyzed data . C . R . and M . B . assisted with experimental design , analyses and interpretation data . A . P . and C . E . helped with reagents , equipment , and discussion of results . D . S . - C . and F . T . wrote the manuscript . All authors contributed to critical revising the manuscript . Additional Information Supplementary information accompanies this paper at doi : 10 . 1038 / s41598 - 017 - 07130 - x Competing Interests : The authors declare that they have no competing interests . Publisher ' s note : Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations . Open Access This article is licensed under a Creative Commons Attribution 4 . 0 International License , which permits use , sharing , adaptation , distribution and reproduction in any medium or format , as long as you give appropriate credit to the original author ( s ) and the source , provide a link to the Cre - ative Commons license , and indicate if changes were made . The images or other third party material in this article are included in the article\u2019s Creative Commons license , unless indicated otherwise in a credit line to the material . If material is not included in the article\u2019s Creative Commons license and your intended use is not per - mitted by statutory regulation or exceeds the permitted use , you will need to obtain permission directly from the copyright holder . To view a copy of this license , visit http : / / creativecommons . org / licenses / by / 4 . 0 / . \u00a9 The Author ( s ) 2017",
    "jin2022branched_supp": "1 Supplementary information Branched actin networks are organized for asymmetric force production during clathrin - mediated endocytosis in mammalian cells Meiyan Jin 1 , Cyna Shirazinejad 1 , 2 , Bowen Wang 3 , Amy Yan 1 , Johannes Sch\u00f6neberg 1 , 4 , Srigokul Upadhyayula 1 , Ke Xu 3 , David G . Drubin 1 * 1 Department of Molecular and Cell Biology , University of California , Berkeley , CA 94720 2 Biophysics Graduate Group , University of California Berkeley ; Berkeley , CA , 94720 3 Department of Chemistry , University of California , Berkeley , CA 94720 4 Current address : Department of Pharmacology , and Department of Chemistry and Biochemistry , University of California , San Diego , CA 92093 These authors contributed equally : Meiyan Jin , Cyna Shirazinejad Correspondence to DGD ( drubin @ berkeley . edu ) 2 Supplementary Fig . 1 : Genome - edited iPSCs show dynamic CME sites . a Schematic model of CME . Mammalian CME proteins can be grouped into several modules , including the coat , WASP and Myosin / actin nucleation promoting factor ( NPF ) , actin and scission modules . Actin networks provide pulling forces to invaginate the membrane against membrane tension . b Immunoblot analysis of cell extracts from the WT ( WTC ) and genome - edited ( AP2M1 - tagRFP - T / DNM2 - tagGFP2 / ARPC3 - HaloTag ; ADA ) human iPSCs . The labeled proteins were detected with tag ( CGY ) FP , HaloTag , and GAPDH ( loading control ) antisera respectively . N = 1 . Uncropped and unprocessed scans of the blots are provided in the Source Data file . c Kymograph of representative CME sites of double - edited ( AP2M1 - tagRFP - 3 T / DNM2 - tagGFP2 ; AD ) and triple - edited ( AP2M1 - tagRFP - T / DNM2 - tagGFP2 / ARPC3 - HaloTag ; ADA ) cells . Scale bar : 5\u00b5m . 4 Supplementary Fig . 2 : Actin assembles at different types of CME sites . a Montage of a representative ARPC3 positive CME plaque from a TIRF movie of triple - edited ( AP2M1 - tagRFP - T / DNM2 - tagGFP2 / ARPC3 - HaloTag ; ADA ) human iPSCs ( Supplementary Movie 2 ) . b Montage of a representative ARPC3 positive splitting CME site from a TIRF movie of triple - edited ( AP2M1 - tagRFP - T / DNM2 - tagGFP2 / ARPC3 - HaloTag ; ADA ) human iPSCs ( Supplementary Movie 2 ) . 5 6 Supplementary Fig . 3 : Filtering methods for selection of CME sites . a 2 - D histogram of the first two principal components ( PCs ) of AP2 and DNM2 dynamic features . Of the 121 , 349 total tracks detected by CMEanalysis , ~ 49 % were valids ( as defined previously 30 ) , ~ 31 % had detection gaps , ~ 15 % were persistent events , and ~ 5 % were split / merges . Only valid tracks ( N = 59 , 239 ) , which appear and disappear during the course of movie acquisition , were used to generate filtering methods and for subsequent analysis . The shaded underlay represents simulated data points in principal component space and their individual probabilities of belonging to the nearest cluster center . Cluster 0 shows data points in the DNM2 - positive cluster , which contains 10 . 02 % of total valid tracks . Clusters 1 - 4 represented 24 . 60 % , 10 . 43 % , 17 . 03 % , and 37 . 91 % of valid tracks , respectively . b Cohort plots of the shortest AP2 events ( < 40 seconds ) from each cluster . Cluster 0 represents DNM2 - positive events where a strong DNM2 signal is detected . Data are presented as mean values + / - 1 / 4 standard deviation . c DNM2 - positive events are sorted by the number of DNM2 peaks using a peak - detection scheme . Representative intensity vs time plots of a single - peaked event ( top ) , non - peak event ( bottom left ) and a multi - peaked event ( bottom right ) . Percentage of the number of the events in each class is shown next to the plot . d Single - peaked DNM2 events ( N = 2 , 538 ) , hereon named CME sites , are grouped into lifetime cohorts and aligned to the peak of the DNM2 channel . Percentage of the number of the CME sites in each cohort is shown next to the plot . Data are presented as mean values + / - 1 / 4 standard deviation . 7 8 Supplementary Fig . 4 : Computational analysis reveals actin network assembly at the late stage of CME . a We generated a randomized data set by pairing ARPC3 images with AP2 and DNM2 images from an unrelated movie . b The fraction of CME sites without actin assembly ( Negative ) , with early actin assembly ( ARPC3 signal disappears before DNM2 signal peak ) , and with late actin assembly ( ARPC3 signal overlaps with DNM2 signal peak ) were calculated for real data movies and randomized control movies ( a ) . In the randomized dataset , we detected early \u201cassembly\u201d of actin in a similar fraction of CME events as in the real data set , indicating that presence of actin detected early in CME is very likely an artificial association caused when nearby actin structures overlap with CME sites by chance . Mean and standard deviation are shown on the graph . Source data are provided in the Source Data file . 9 10 Supplementary Fig . 5 : AP2 - ARPC3 separation is not due to imaging artifacts . a Montage from a TIRF movie of a multi - fluorescence bead ( Supplementary Movie 4 ) . Size of field of view : 2\u00b5m x 2\u00b5m . Intervals : 1sec . b A boxplot of inter - channel distances ( note : y - axis is base - 10 log scale ) between centroids of either CME sites or beads to highlight that the observed separation between marked proteins exceeds the measured chromatic aberration . Beads were compared between the same pairs of channels as used to image the tagged proteins . Box plot elements : center line , median ; box limits , upper and lower quartiles ; whiskers , 1 . 5x interquartile range ; points , outliers . Source data are provided in the Source Data file . c A heat map graph of distance between AP2 and ARPC3 signals before scission , and average AP2 frame to frame displacement within 6 seconds before scission . Over 95 % of the CME events present larger AP2 - ARPC3 separation than AP2 displacement . N = 1 , 385 . 11 Supplementary Fig . 6 : Arp2 / 3 - mediated actin assembly facilitates CME . a Intensity vs time plots of averaged ARPC3 negative ( top ) and positive ( bottom ) CME sites in ADA cells after 1 - 5min treatment with 2 % DMSO ( solid lines ) or 100 \u03bcM CK666 ( dashed lines ) . Events were aligned to the frames showing the maximum DNM2 intensity . Error bar : \u00bc standard deviation . ARPC negative CME sites : DMSO : N = 402 , CK666 : N = 261 , ARPC3 positive CME sites : DMSO : N = 1 , 006 , CK666 : N = 580 . b CK666 - treated cells have longer CME lifetimes than DMSO - treated control cells . p - value from two - sided Kolmogorov - Smirnov test : 1 . 83e - 3 . Source data are provided in the Source Data file . 12 Supplementary Fig . 7 : Dynamics of N - WASP at CME sites . a Immunoblot analysis of cell extracts from the control and genome - edited ( AP2M1 - tagRFP - T / DNM2 - tagGFP2 / HaloTag - WASL ; ADW ) human iPSCs . The labeled proteins were detected with HaloTag and GAPDH ( loading control ) antisera respectively . N = 1 . Uncropped and unprocessed scans of the blots are provided in the Source Data file . b Histogram of N - WASP lifetime at CME sites . The lifetime is measured from the first frame of the N - WASP signal to the presumed scission time ( the peak of DNM2 signal ) . c Intensity vs time plots of cohorts of N - WASP positive CME sites in ADW cells . Events are grouped into cohorts by the lifetimes of AP2 and aligned to the frames showing the maximum DNM2 intensity . Data are presented as mean values + / - 1 / 4 standard deviation . N = 1 , 199 . Source data are provided in the Source Data file .",
    "numasawa2020fluorescent_supp": "Supporting Information A Fluorescent Probe for Rapid , High - Contrast Visualization of Folate - Receptor - Expressing Tumors In Vivo Koji Numasawa , Kenjiro Hanaoka , * Naoko Saito , Yoshifumi Yamaguchi , Takayuki Ikeno , Honami Echizen , Masahiro Yasunaga , Toru Komatsu , Tasuku Ueno , Masayuki Miura , Tetsuo Nagano , and Yasuteru Urano * anie _ 201914826 _ sm _ miscellaneous _ information . pdf S2 Materials and general information Reagents and solvents were of the best grade available , supplied by Tokyo Chemical Industries , Co . , Ltd . , Wako Pure Chemical Industries , Ltd . , Sigma - Aldrich Co . , LLC . , Kanto Chemical Co . , Inc . , Watanabe Chemical Industries Ltd . , Carbosynth Ltd . , Novabiochem or Thermo Fisher Scientific Inc . , and were used without further purification . Silica gel column chromatography was performed on Silica gel 60 ( Spherical ) ( Kanto Chmical Co . ) . Saline was purchased from Otsuka Pharmaceutical Co . Ltd . . Mice ( BALB / cAJcl - nu / nu ) were purchased from CLEA Japan . 1 H NMR and 13 C NMR spectra were recorded on a JEOL INM - LA300 ( 300 MHz ) , JNM - LA400 or ECZ - 400S instrument ( 400 MHz ) ; \u03b4 values are in ppm relative to tetramethylsilane ( TMS ) . Mass spectra ( MS ) were measured with a JEOL JMS - T100LC AccuTOF ( ESI ) . UV - vis spectra were obtained on a UV - 2550 ( Shimadzu ) . Fluorescence spectroscopic studies were performed with a Hitachi F - 7000 ( Tokyo , Japan ) ; the slit widths were 2 . 5 nm for excitation and 2 . 5 nm for emission , and the photomultiplier voltage was 700 V . Fluorescence quantum yields of SiR - based probes were measured with a Quantaurus - QY ( Hamamatsu ) . HPLC analysis was performed on an Inertsil ODS - 3 column ( GL Science Inc . ; 4 . 6 mm \u00d7 250 mm ) using a system composed of a pump ( PU - 980 , JASCO ) and a detector ( MD - 2015 , JASCO ) . HPLC purifications were performed on an Inertsil ODS - 3 column ( GL Science Inc . , 10 . 0 mm\u00d7 250 nm ) using a system composed of a pump ( PU - 2080 , JASCO ) and a detector ( MD - 2015 , JASCO ) . MPLC purifications were performed on a Yamazen Smart Flash EPCLC AI - 5805 ( Tokyo , Japan ) . UV - Vis Absorption and Fluorescence Spectroscopy UV - Visible spectra were obtained on a Shimadzu UV - 2550 . Fluorescence spectroscopic studies were performed with a Hitachi F7000 with a slit width of 2 . 5 nm for excitation and emission . The photomultiplier voltage was 700 V . Absolute quantum yields ( \u03a6 ) were measured with a Hamamatsu Photonics Quantaurus QY . Relative fluorescence quantum yields were obtained by comparing the area under the emission spectra of the test and standard samples according to the following equation . \u03a6 x / \u03a6 st = [ A st / A x ] [ n x2 / n st2 ] [ D x / D st ] where st : standard ; x : sample ; A : absorbance at the excitation wavelength ; n : refractive index ; D : area under the fluorescence spectra on an energy scale . S3 Preparation of Cultured Cells Cells of the human oral epidermoid carcinoma cell line KB ( a subline of HeLa ) were cultured in Eagle ' s minimal essential medium ( EMEM ) ( FUJIFILM Wako Pure Chemical Corp . , Japan ) , supplemented with 10 % ( v / v ) fetal bovine serum ( Invitrogen ) and penicillin ( 100 units / mL ) - streptomycin ( 100 \u03bcg / mL ) mixture ( Invitrogen ) . Epithelial ovarian cancer cell line OVCAR - 3 cells ( ATCC , USA ) were cultured in RPMI 1640 medium ( Invitrogen ) , supplemented with 20 % ( v / v ) fetal bovine serum ( Invitrogen ) and penicillin ( 100 units / mL ) - streptomycin ( 100 \u03bcg / mL ) ( Invitrogen ) . Human fibrosarcoma cell line HT1080 cells ( Health Science Research Resource Bank , Osaka , Japan ) were cultured in Dulbecco\u2019s modified Eagle\u2019s medium ( DMEM ) ( Invitrogen ) , supplemented with 10 % ( v / v ) fetal bovine serum ( Invitrogen ) and penicillin ( 100 units / mL ) - streptomycin ( 100 \u03bcg / mL ) ( Invitrogen ) . All cell lines were cultured at 37\u00b0C in a humidified incubator containing 5 % CO 2 in air . Confocal Fluorescence Imaging of Cultured Cells A confocal imaging system ( TCS - SP5 ; Leica ) equipped with a white - light laser was used . Fluorescence images were captured with excitation and emission wavelengths of 490 and 510 - 540 nm for fluorescein folate , Alexa 488 folate and Alexa 488 - conjugated IgG , 555 and 575 - 600 nm for TAMRA folate , and 650 and 670 - 750 nm for FolateSiR - 1 and FolateSiR - 2 . Immunofluorescence Staining of Cultured Cells KB cells , OVCAR - 3 cells and HT1080 cells were fixed with 4 % formaldehyde in phosphate - buffered saline ( PBS ) solution for 20 min . The medium was removed and the cells were washed with PBS three times , then incubated for 1 hr in PBS containing 0 . 1 % Triton X - 100 and 1 % bovine serum albumin ( BSA ) for blocking . Immunostaining was carried out for 1 h at room temperature with anti - folate binding protein antibody ( Abcam , Inc . ) at 1 : 100 dilution in PBS containing 0 . 1 % Triton X - 100 and 1 % BSA . Immunodetection was performed using Alexa Fluor \u00ae 488 - conjugated AffiniPure goat anti - mouse IgG ( Jackson Immuno Research Laboratories , Inc . ) at 1 : 500 dilution in PBS containing 0 . 1 % Triton X - 100 and 1 % BSA . Fluorescence images were captured with TCS - SP5 confocal S4 laser scanning microscope ( Leica ) . Fluorescence Imaging of Cultured KB Cells Cells were plated onto a 35 - mm poly - L - lysine - coated glass - bottomed dish ( Matsunami Glass Ind . , Ltd . ) in EMEM supplemented with 10 % ( v / v ) fetal bovine serum , 1 % penicillin and 1 % streptomycin . Before loading dyes , the medium was removed and the cells were washed with Hank ' s balanced salt solution ( HBSS ) three times . The cells were incubated with 5 \u03bcM folate - conjugated fluorescent dye in EMEM containing 0 . 3 % DMSO as a cosolvent at 37\u00b0C for 30 min , then the medium was removed and the cells were washed with HBSS three times . The cells were observed in HBSS . When cells were fixed with 4 % formaldehyde in PBS , the dish was kept on ice for 15 min . The medium was warmed to 37\u00b0C before all procedures . Fluorescence Imaging of Cultured OVCAR - 3 Cells Cells were plated onto a 35 - mm poly - L - lysine - coated glass - bottomed dish ( Matsunami Glass Ind . , Ltd . ) in RPMI 1640 medium supplemented with 20 % ( v / v ) fetal bovine serum , 1 % penicillin and 1 % streptomycin . Before dye loading , the medium was removed and the cells were washed with HBSS three times . The cells were incubated with 5 \u03bcM folate - conjugated fluorescent dye in RPMI 1640 containing 0 . 3 % DMSO as a cosolvent at 37\u00b0C for 30 min , then the medium was removed and the cells were washed with HBSS three times . The medium was warmed to 37\u00b0C before all procedures . Ex vivo Fluorescence Imaging of Mouse Embryos All procedures were approved by the Animal Care and Use Committee of the University of Tokyo . Imaging of embryos was done as described previously with slight modifications . SR1 Briefly , before the embryos were dissected , all media and dishes were prewarmed to 37\u00b0C . Then , about 300 \u03bcL of 2 % low - melting agarose was poured into a glass - bottomed dish ( thickness of 100 \u03bcm ; Matsunami Glass Ind . , Ltd . ) , and holes for receiving embryos were made at least 30 min before filling the dish with medium . Pregnant mice were killed at 8 . 5 d postcoitum by cervical dislocation , and the uterus was removed and placed in custom - made DMEM without phenol red or folic acid . S5 Embryos in their yolk sac were removed from the uterus and transferred into folic acid - free DMEM containing 10 % fetal bovine serum , and the yolk sac and amnion were carefully removed while the embryo was kept warm . The dissected embryos were cultured with FolateSiR - 1 ( 10 \u03bcM or 20 \u03bcM ) or FolateSiR - 2 ( 20 \u03bcM ) in DMEM without folic acid or in DMEM containing 1 mM folic acid for the competitive assay for 30 minutes in a humidified incubator at 37\u00b0C under 5 % CO 2 , and then washed twice with DMEM without folic acid . For live - imaging analysis , the embryos were immediately transferred to agarose - containing glass - bottomed dishes filled with folic acid - free DMEM containing 50 % centrifuged rat serum ( Charles River ) . Observation was done with an inverted confocal microscope ( TCS - SP5 ; Leica ) equipped with a galvo stage and a resonant scanner for fast scanning , using an HCX PL FLUOTAR 10\u00d7 0 . 3 NA dry objective ( Leica ) . During imaging , the dishes were kept in a humidified cell culture incubator with a continuous supply of 5 % CO 2 / air at 37\u00b0C ( Tokai Hit Company ) . For static observation , stained embryos were fixed with 4 % paraformaldehyde on ice for 10 min immediately after culture , then transferred to HBSS and observed . During imaging , the embryos were placed in holes in the agarose as described above . FolateSiR - 1 or FolateSiR - 2 was excited with a 633 - nm laser ( 20 \u2013 40 % power ) with the resonant scanner ( 8 , 000 Hz ; Leica ) , and the emission and bright - field signals were detected with photomultiplier tubes . The thickness of the z slices was 4 \u03bcm , and acquired images were processed to create maximum intensity projections . In Vivo Fluorescence Imaging of Mouse Models All procedures were approved by the Animal Care and Use Committee of the University of Tokyo . Tumor model mice were prepared as follows . BALB / cAJcl mice ( 7 - 8 weeks , \u2640 ) were inoculated with 3 - 5\u00d710 6 KB cells or 1\u00d710 7 HT1080 cells into the root of the forefoot at 7 - 10 days before fluorescence imaging . After intravenous injection into the tail of 100 \u03bcM FolateSiR - 1 or FolateSiR - 2 in 100 \u03bcL saline containing 1 % DMSO as a cosolvent , bright - field and fluorescence images were captured at different time points with a Maestro In - vivo Imaging System ( CRi Inc . , Woburn , MA ) equipped with an excitation filter 661 ( 641 - 681 nm ) , emission filter 704 ( 684 - 729 nm ) , using a bright - field setting . The tumor - bearing mice were carried back from the in vivo imaging system to the cage after the fluorescence image was captured at each time point . S6 Tissue Microarray - based Binding Analysis Frozen tissue array slides of ovary tumors ( Catalog No . : T6235183 - 5 , Lot No . : B705061 ) were purchased from BioChain Institute , Inc . ( Newark , CA , USA ) . For immunostaining folate receptors , the procedure was as follows . The slides were fixed with 4 % paraformaldehyde ( Wako ) at room temperature for 10 min , and washed with PBS containing 0 . 05 % Tween 20 ( PBST ) for 3 min three times . Endogenous peroxidases were blocked in 3 % H 2 O 2 for 20 min , and the slides were washed with PBST for 3 min three times . The slides were then blocked with 5 % skim milk ( Becton Dickinson ) in PBST at room temperature for 30 min , and incubated with mouse anti - folate binding protein antibody ( Abcam ) in PBST with 5 % skim milk at room temperature for 1 h . Then , the slides were washed with PBST three times , and treated with EnVision + System - HRP Labeled Polymer Anti - mouse ( Dako ) . DAB staining was performed at room temperature for 5 min with DAB chromogen substrate ( K3468 , Dako ) . The slide was dipped in water for 3 min , then subjected to hematoxylin staining at room temperature for 5 min , incubated again in water for 10 min , and washed with ethanol ( 2 min , 2 min and 3 min ) and xylene ( 3 min , 3 min and 3 min ) . MOUNT - QUICK ( Daido Sangyo Co . , Ltd . ) was used as a mounting agent , and immunostaining was observed with a VS120 virtual slide microscope ( Olympus , Inc . ) . For the fluorescence imaging of folate receptors with FolateSiR - 1 , the procedure was as follows . The slide was dipped in PBS at room temperature for 10 min , washed with PBST for 3 min three times , and blocked with 5 % skim milk in PBST at room temperature for 30 min . Then 5 \u03bcM FolateSiR - 1 and 2 . 9 \u03bcM DAPI ( nuclear stain ) in PBST containing 5 % skim milk and 0 . 1 % DMSO as a cosolvent were added and the slide was incubated at room temperature for 2 h , and washed with PBST three times . Fluoromount - G ( SouthernBiotech ) was used as a mounting agent , and fluorescence imaging was performed with a VS120 virtual slide microscope ( Olympus , Inc . ) equipped with a blue filter for DAPI , and a Cy5 filter for FolateSiR - 1 . The data were analyzed using the OlyVIA program ( Olympus , Inc . ) . S7 Synthesis and characterization of compounds Scheme S1 . Synthesis of 2 , 5 - diCOOH SiR650 Synthesis of N , N , N\u2019 , N\u2019 - tetramethyldiamino - Si - xanthone The compound was synthesized according to reference SR2 . Synthesis of 1 , 1 ' - ( 4 - bromo - 1 , 3 - phenylene ) bis ( 4 - methyl - 2 , 6 , 7 - trioxabicyclo [ 2 . 2 . 2 ] octane ) The compound was synthesized according to reference SR3 . S8 Synthesis of 2 , 5 - diCOOH SiR650 1 , 1 ' - ( 4 - Bromo - 1 , 3 - phenylene ) bis ( 4 - methyl - 2 , 6 , 7 - trioxabicyclo [ 2 . 2 . 2 ] octane ) ( 406 mg , 0 . 983 mmol ) and anhydrous THF ( 10 mL ) were added to a flame - dried flask flushed with argon . The solution was cooled to \u2013 78\u00b0C , and 1 M sec - BuLi solution in cyclohexane / n - hexane mixture ( 0 . 98 mL , 0 . 98 mmol ) was added to it . The mixture was stirred for 1 hr and N , N , N\u2019 , N\u2019 - tetramethyldiamino - Si - xanthone ( 106 mg , 0 . 327 mmol ) in anhydrous THF ( 10 mL ) was slowly added at the same temperature . The whole was warmed to room temperature and stirred for 3 . 5 h . AcOH ( 5 . 0 mL ) was added , and the mixture was evaporated to dryness . The residue was dissolved in 6 N HCl aq . . The resulting solution was refluxed overnight , allowed to cool to room temperature , and evaporated to dryness . The residue was purified by RP - HPLC to obtain 2 , 5 - diCOOH SiR650 ( 54 . 3 mg , 0 . 115 mmol , 35 % yield ) . 1 H NMR ( 300 MHz , CD 3 OD ) : \u03b4 = 0 . 58 ( s , 3H ) , 0 . 66 ( s , 3H ) , 3 . 30 ( s , 12H ) , 6 . 77 ( dd , J = 9 . 5 Hz , 2 . 9 Hz , 2H ) , 6 . 97 ( d , J = 9 . 5 Hz , 2H ) , 7 . 34 ( d , J = 2 . 9 Hz , 2H ) , 7 . 83 ( s , 1H ) , 8 . 30 ( s , 2H ) ; HRMS ( ESI + ) : Calcd for [ M ] + 473 . 1897 , Found , 473 . 1937 ( + 4 . 1 mmu ) ; HPLC analysis : eluent : A / B = 95 / 5 to 0 / 100 , 20 min , linear gradient ; solvent A : H 2 O , 0 . 1 % TFA ; solvent B : acetonitrile / H 2 O = 80 / 20 , 0 . 1 % TFA ; flow rate , 1 . 0 mL / min ; detection wavelength 650 nm . S9 Scheme S2 . Synthesis of 2 - Me - 5 - COOH SiR650 Synthesis of tert - butyl 3 - bromo - 4 - methylbenzoate 3 - Bromo - 4 - methylbenzoic acid ( 4 . 07 g , 18 . 9 mmol ) , Boc 2 O ( 11 . 2 g , 51 . 2 mmol ) and DMAP ( 62 . 5 mg , 5 . 12 mmol ) were dissolved in anhydrous THF ( 68 mL ) . The solution was refluxed overnight , then allowed to cool to room temperature , and evaporated to dryness . The residue was dissolved in ethyl acetate . The resulting solution was washed with sat . NaHCO 3 aq . and brine , dried over anhydrous Na 2 SO 4 , and filtered . The filtrate was evaporated to dryness , and the residue was purified by silica gel column chromatography ( n - hexane ) to afford tert - butyl 3 - bromo - 4 - methylbenzoate ( 3 . 47 g , 12 . 8 mmol , 68 % yield ) . 1 H NMR ( 300 MHz , CDCl 3 ) : \u03b4 = 1 . 58 ( s , 9H ) , 2 . 44 ( s , 3H ) , 7 . 27 ( d , J = 8 . 1 Hz , 1H ) , 7 . 81 ( dd , J = 8 . 1 Hz , 1 . 5 Hz , 1H ) , 8 . 12 ( d , J = 1 . 5 Hz , 1H ) ; 13 C NMR ( 75 MHz , CDCl 3 ) : \u03b4 = 23 . 1 , 28 . 1 , 81 . 4 , 124 . 6 , 128 . 2 , 130 . 5 , 131 . 3 , 133 . 2 , 142 . 6 , 164 . 5 ; HRMS ( ESI \u2012 ) : Calcd for [ M \u2012 t Bu ] \u2012 212 . 9551 , Found , 212 . 9553 ( + 0 . 2 mmu ) . Synthesis of 2 - Me - 5 - COOH SiR650 S10 tert - Butyl 3 - bromo - 4 - methylbenzoate ( 627 mg , 2 . 31 mmol ) and anhydrous THF ( 5 . 0 mL ) were added to a flame - dried flask flushed with argon . The solution was cooled to \u2012 78 \u02daC , and 1 M sec - BuLi in cyclohexane / n - hexane mixture ( 2 . 0 mL , 2 . 0 mmol ) was added to it . The mixture was stirred for 10 min at the same temperature , and N , N , N\u2019 , N\u2019 - tetramethyldiamino - Si - xanthone ( 72 . 3 mg , 0 . 223 mmol ) in anhydrous THF ( 12 . 0 mL ) was slowly added . The reaction mixture was warmed to room temperature and stirred for 2 . 5 h , then 2 N HCl aq . was added . Stirring was continued for 20 min , and the mixture was extracted with CH 2 Cl 2 . The organic layer was washed with brine , dried over anhydrous Na 2 SO 4 and evaporated to dryness . The residue was dissolved in TFA . The solution was stirred at room temperature for 2 h , then evaporated , and the residue was purified to give 2 - Me - 5 - COOH SiR650 ( 53 . 5 mg , 0 . 121 mmol , 54 % yield ) . 1 H NMR ( 300 MHz , CD 3 OD ) : \u03b4 = 0 . 61 ( s , 3H ) , 0 . 63 ( s , 3H ) , 2 . 11 ( s , 3H ) , 3 . 35 ( s , 12H ) , 6 . 78 ( dd , J = 9 . 9 Hz , 2 . 6 Hz , 2H ) , 7 . 03 ( d , J = 9 . 5 Hz , 2H ) , 7 . 37 ( d , J = 2 . 9 Hz , 2H ) , 7 . 54 ( d , J = 8 . 1 Hz , 1H ) , 7 . 73 ( s , 1H ) , 8 . 11 ( d , J = 8 . 1 Hz , 1H ) ; 13 C NMR ( 100 MHz , CD 3 OD ) : \u03b4 = \u20121 . 3 , \u20121 . 1 , 19 . 6 , 40 . 9 , 115 . 4 , 122 . 4 , 128 . 3 , 129 . 8 , 131 . 2 , 131 . 3 , 131 . 7 , 140 . 4 , 142 . 0 , 142 . 6 , 149 . 6 , 155 . 9 , 169 . 0 , 169 . 1 ; HRMS ( ESI + ) : Calcd for [ M ] + 455 . 2155 , Found , 455 . 2131 ( \u20122 . 3 mmu ) ; HPLC analysis : eluent : A / B = 95 / 5 to 0 / 100 , 20 min , linear gradient ; solvent A : H 2 O , 0 . 1 % TFA ; solvent B : acetonitrile / H 2 O = 80 / 20 , 0 . 1 % TFA ; flow rate , 1 . 0 mL / min ; detection wavelength 650 nm . S11 Synthesis of 2 - Me - 4 - COOH dichloro - TokyoMagenta ( DCTM ) The compound was synthesized according to reference SR4 . Synthesis of 2 , 4 - diCOOH DCTM The compound was synthesized according to reference SR5 . S12 Scheme S3 . Synthesis of Folate - peptide . Synthesis of Folate - peptide Folate - peptide was synthesized on an automatic peptide synthesizer using the standard protocol for fluorenylmethoxycarbonyl ( Fmoc ) solid - phase synthesis with 2 - chlorotrityl chloride resin ( 302 mg , 1 . 42 mmol / g reaction site , 0 . 428 mmol ) ( Novabiochem ) . O - ( 7 - Azabenzotriazol - 1 - yl ) - N , N , N ' , N ' - tetramethyluronium hexafluorophosphate ( HATU ) and N , N - diisopropylethylamine ( DIPEA ) were used as coupling reagents and DMF S13 and N - methylpyrrolidone ( NMP ) were used as solvents . Fmoc - Gly - OH , Fmoc - Lys ( Boc ) - OH , Fmoc - Asp ( O t Bu ) - OH and Fmoc - Glu - O t Bu were used as Fmoc amino acids . After the construction of the peptide moiety of Folate - peptide on the solid - phase , the Fmoc group at the N - terminus was deprotected with 40 % piperidine in DMF . Then , N - trifluoroacetyl pteroic acid ( 368 mg , 0 . 902 mmol ) was coupled to the peptide on the solid phase by adding a mixture of HOBt , WSCD\u2027HCl , Et 3 N and DMSO . The trifluoroacetyl group was deprotected with 2 % hydrazine in DMF . Folate - peptide was cleaved with 2 mL of TFA / triisopropylsilane / H 2 O ( 95 / 2 . 5 / 2 . 5 v / v / v ) for 90 min . The solution was evaporated to dryness and the residue was purified by RP - HPLC to afford a brown solid ( 42 . 1 mg , 12 % ) . HRMS ( ESI + ) : Calcd for [ M + H ] + 742 . 2909 Found , 742 . 2862 ( \u2012 4 . 7 mmu ) . HPLC analysis : eluent , A / B = 99 / 1 \uf0e0 70 / 30 , 10 min \uf0e0 0 / 100 , 35 min ; A : H 2 O containing 0 . 1 % TFA ( v / v ) , B : CH 3 CN / H 2 O = 80 / 20 containing 0 . 1 % TFA ( v / v ) ; 1 . 0 mL / min flow rate ; Detection at 286 nm : Retention time was 12 . 0 min . Scheme S4 . Synthesis of FolateSiR - 1 S14 Synthesis of FolateSiR - 1 NHS ( 17 . 9 mg , 150 \u03bcmol ) and WSCD\u2027HCl ( 31 . 0 mg , 162 \u03bcmol ) were added to a solution of 2 , 5 - diCOOH SiR ( 27 . 9 mg , 47 . 0 \u03bcmol ) in THF ( 3 mL ) . The mixture was stirred at 35 \u02da C for 3 h , then evaporated to dryness , and the residue was partially purified by RP - HPLC to obtain crude 2 , 5 - diCOOH SiR650 succinimidyl ester ( 15 . 4 mg ) . DIPEA ( 57 \u03bcL , 0 . 327 mmol ) was added to a solution of the crude 2 , 5 - diCOOHSiR succinimidyl ester ( 7 . 2 mg ) and Folate - peptide ( 10 . 1 mg , 0 . 012 mmol ) in THF ( 1 . 5 mL ) . The mixture was stirred at room temperature overnight , then evaporated to dryness , and the residue was purified by RP - HPLC to obtain FolateSiR - 1 ( 2 . 7 mg , 2 . 0 \u03bcmol , 4 % yield in 2 steps ) . HRMS ( ESI + ) : Calcd for [ M ] + , 1196 . 4621 , Found , 1196 . 4621 ( 0 . 0 mmu ) . HPLC analysis : eluent , A / B = 80 / 20 \uf0e0 0 / 100 , 20 min ; A : H 2 O containing 0 . 1 % TFA ( v / v ) , B : CH 3 CN / H 2 O = 80 / 20 containing 0 . 1 % TFA ( v / v ) ; 1 . 0 mL / min flow rate ; Detection at 650 nm : Retention time was 11 . 3 min . S15 Scheme S5 . Synthesis of FolateSiR - 2 Synthesis of FolateSiR - 2 NHS ( 45 . 7 mg , 0 . 398 mmol ) and WSCD\u2027HCl ( 38 . 0 mg , 0 . 199 mmol ) were added to a solution of 2 - Me - 5 - COOH SiR650 ( 9 . 7 mg , 17 . 5 \u03bcmol ) in THF ( 3 . 0 mL ) . The mixture was stirred at room temperature overnight , then evaporated to dryness , and the residue was partially purified by RP - HPLC to obtain crude 2 - Me - 5 - COOH SiR650 succinimidyl ester ( 5 . 4 mg ) . DIPEA ( 17 \u03bcL , 0 . 098 mmol ) was added to a solution of the crude 2 - Me - 5 - COOH SiR650 succinimidyl ester ( 6 . 5 mg ) and Folate - peptide ( 5 . 7 mg , 6 . 7 \u03bcmol ) in DMF ( 3 . 0 mL ) . The mixture was stirred at room temperature overnight , then evaporated to dryness , and the residue was purified by RP - HPLC to obtain FolateSiR - 2 ( 1 . 0 mg , 0 . 78 \u03bcmol , 4 % yield in 2 steps ) . HRMS ( ESI + ) : Calcd for [ M ] + , S16 1166 . 4880 , Found , 1166 . 4881 ( + 0 . 1 mmu ) . HPLC analysis : eluent , A / B = 80 / 20 \uf0e0 0 / 100 , 20 min ; A : H 2 O containing 0 . 1 % TFA ( v / v ) , B : CH 3 CN / H 2 O = 80 / 20 containing 0 . 1 % TFA ( v / v ) ; 1 . 0 mL / min flow rate ; Detection at 650 nm : Retention time was 13 . 2 min . Scheme S6 . Synthesis of 2 - Me DCTM folate Synthesis of 2 - Me DCTM folate 2 - Me - 4 - COOH DCTM ( 11 mg , 24 \uf06d mol ) , N - hydroxysuccinimide ( NHS ) ( 8 . 4 mg , 73 \uf06d mol ) and WSCD\u2027HCl ( 14 . 0 mg , 73 \uf06d mol ) were dissolved in THF ( 1 . 0 mL ) . The mixture was stirred at 35 \u02daC for 5 h , then evaporated to dryness , and the residue was partially purified by RP - HPLC to obtain crude 2 - Me - 4 - COOH DCTM succinimidyl ester ( 4 . 8 mg ) . The crude product and Folate - peptide ( 18 mg , 24 \uf06d mol ) were dissolved in DMF , then DIPEA ( 10 S17 \uf06d L , 57 \uf06d mol ) was added , and the reaction mixture was stirred at 35 \u02daC for 2 days . The solution was evaporated to dryness and the residue was purified by RP - HPLC to obtain 2 - Me DCTM folate ( 0 . 12 mg , 0 . 10 \uf06d mol , 4 % yield in 2 steps ) . HRMS ( ESI + ) : Calcd for [ M + H ] + , 1180 . 3155 , Found , 1180 . 3183 ( + 2 . 8 mmu ) . HPLC analysis : eluent , A / B = 30 / 70 \uf0e0 0 / 100 , 20 min ; A : H 2 O containing 0 . 1 % TFA ( v / v ) , B : CH 3 CN / H 2 O = 80 / 20 containing 0 . 1 % TFA ( v / v ) ; 1 . 0 mL / min flow rate ; Detection at 470 nm : Retention time was 14 . 8 min . Scheme S7 . Synthesis of 2 - COOH DCTM folate Synthesis of 2 - COOH DCTM Folate 2 , 4 - diCOOH DCTM ( 11 mg , 23 \uf06d mol ) , NHS ( ( 8 . 1 mg , 70 \uf06d mol ) and WSCD\u2027HCl ( 14 . 8 mg , 77 \uf06d mol ) were S18 dissolved in THF ( 1 . 0 mL ) . The mixture was stirred at 35 \u02daC overnight , then evaporated to dryness , and the residue was partially purified by RP - HPLC to obtain crude 2 , 4 - diCOOH DCTM succinimidyl ester ( 4 . 1 mg ) . The crude product and Folate - peptide ( 15 mg , 20 \uf06d mol ) were dissolved in DMF ( 2 . 0 mL ) , then DIPEA ( 10 \uf06d L , 57 \uf06d mol ) was added . The reaction mixture was stirred at 35 \u02daC for 2 days , then evaporated to dryness , and the residue was purified by RP - HPLC to obtain 2 - COOH DCTM folate ( 0 . 63 mg , 0 . 52 \uf06d mol , 23 % yield ) . HRMS ( ESI + ) : Calculated for [ M + H ] + , 1210 . 2896 , Found , 1210 . 2902 ( + 0 . 6 mmu ) . HPLC analysis : eluent , A / B = 70 / 30 \uf0e0 0 / 100 , 20 min ; A : H 2 O containing 0 . 1 % TFA ( v / v ) , B : CH 3 CN / H 2 O = 80 / 20 containing 0 . 1 % TFA ( v / v ) ; 1 . 0 mL / min flow rate ; Detection at 283 nm : Retention time was 13 . 1 min . Scheme S8 . Synthesis of Fluorescein folate Fluorescein folate DIPEA ( 5 \u03bcL , 0 . 028 mmol ) was added to a solution of 5 - carboxyfluorescein succinimidyl ester ( 5 . 4 mg , 0 . 011 mmol ) ( Tokyo Chemical Industry Co . , Ltd . , Japan ) and Folate - peptide ( 14 . 5 mg , 0 . 019 mmol ) in DMF ( 500 \u03bcL ) . S19 The mixture was stirred at room temperature overnight , then evaporated , and the residue was purified by HPLC to give fluorescein folate ( 1 . 5 mg , 1 . 3 \u03bcmol , 12 % yield ) . HRMS ( ESI - ) : Calcd for [ M - H ] - 1098 . 3230 , Found , 1098 . 3190 ( \u20124 . 0 mmu ) . HPLC analysis : eluent : A / B = 80 / 20 , 6 min , 80 / 20 to 20 / 80 , 20 min , linear gradient ; solvent A : H 2 O , 0 . 1 % TFA ; solvent B : acetonitrile / H 2 O = 80 / 20 , 0 . 1 % TFA ; flow rate , 1 . 0 mL / min ; detection wavelength 450 nm : Retention time was 27 . 0 min . Scheme S9 . Synthesis of Alexa488 folate Alexa 488 folate DIPEA ( 5 \u03bcL , 0 . 028 mmol ) was added to a solution of Alexa 488 succinimidyl ester isomer mixture ( Invitrogen , Carlsbad , CA ) ( 4 . 5 mg , 7 . 1 \u03bcmol ) and Folate - peptide ( 11 . 5 mg , 0 . 015 mmol ) in DMF ( 500 \u03bcL ) . The mixture was stirred at room temperature overnight , then evaporated to dryness , and the residue was purified by RP - HPLC to obtain Alexa 488 folate ( 2 . 1 mg , 1 . 7 \u03bcmol , 24 % yield ) . HRMS ( ESI \u2012 ) : Calcd for [ M \u2012 H ] \u2012 1256 . 2686 , Found , 1256 . 2711 ( + 2 . 5 mmu ) . HPLC analysis : eluent , A / B = 80 / 20 , 6 min ; 80 / 20 \uf0e0 20 / 80 , 20 min ; A : H 2 O containing S20 0 . 1 % TFA ( v / v ) , B : CH 3 CN / H 2 O = 80 / 20 containing 0 . 1 % TFA ( v / v ) ; 1 . 0 mL / min flow rate ; Detection at 490 nm : Retention time was 22 . 4 min . Scheme S10 . Synthesis of TAMRA folate TAMRA folate DIPEA ( 5 \u03bcL , 0 . 028 mmol ) was added to a solution of 5 - carboxytetramethylrhodamine succinimidyl ester ( Invitrogen , Carlsbad , CA ) ( 5 . 6 mg , 0 . 011 mmol ) and Folate - peptide ( 12 . 1 mg , 0 . 016 mmol ) in DMF ( 0 . 5 mL ) . The mixture was stirred at room temperature overnight , then evaporated to dryness , and the residue was purified by RP - HPLC to obtain TAMRA folate ( 2 . 5 mg , 2 . 1 \u03bcmol , 20 % yield ) . HRMS ( ESI + ) : Calcd for [ M ] + 1154 . 4332 , Found , 1154 . 4315 ( \u20121 . 7 mmu ) . HPLC analysis : eluent , A / B = 80 / 20 , 6 min ; 80 / 20 \uf0e0 20 / 80 , 20 min ; A : H 2 O containing 0 . 1 % TFA ( v / v ) , B : CH 3 CN / H 2 O = 80 / 20 containing 0 . 1 % TFA ( v / v ) ; 1 . 0 mL / min flow rate ; Detection at 547 nm : Retention time was 27 . 0 min . S21 Scheme S11 . Synthesis of 2 - COOH SiR650 Synthesis of tert - butyl 2 - bromobenzoate 2 - Bromobenzoic acid ( 5 . 65 g , 28 . 3 mmol ) , Boc 2 O ( 8 . 65 g , 39 . 5 mmol ) and N , N - dimethyl - 4 - aminopyridine ( DMAP ) ( 765 mg , 6 . 27 mmol ) were dissolved in dry THF ( 40 mL ) . The solution was refluxed overnight , then allowed to cool to room temperature , and evaporated to dryness . The residue was dissolved in ethyl acetate . This solution was washed with saturated NaHCO 3 aq . and brine , dried over Na 2 SO 4 , filtered and evaporated to dryness . The resulting crude oil was purified by silica gel column chromatography ( EtOAc / n - hexane = 1 / 10 ) to afford tert - butyl 2 - bromobenzoate ( 4 . 89 g , 19 . 0 mmol , y . 67 % ) . 1 H NMR ( 400 MHz , CDCl 3 ) : \u03b4 = 1 . 61 ( s , 9H ) , 7 . 25 - 7 . 29 ( m , 1H ) , 7 . 33 ( td , J = 7 . 4 , 1 . 3 Hz , 1H ) , 7 . 61 ( dd , J = 8 . 0 Hz , 1 . 2 Hz , 1H ) , 7 . 68 ( dd , J = 7 . 8 Hz , 2 . 0 Hz , 2H ) ; 13 C NMR ( 100 MHz , CDCl 3 ) : \u03b4 = 28 . 1 , 82 . 5 , 120 . 9 , 127 . 0 , 130 . 7 , 131 . 8 , 134 . 0 , 134 . 3 , 165 . 7 ; HRMS ( ESI + ) : Calcd for [ M \u2012 t Bu ] + 200 . 9551 , Found , 200 . 9537 ( \u20121 . 4 mmu ) . Synthesis of 2 - COOH SiR650 tert - Butyl 2 - bromobenzoate ( 186 mg , 0 . 723 mmol ) and anhydrous THF ( 5 . 0 mL ) were added to a flame - dried S22 flask that had been flushed with argon . The solution was cooled to \u2013 78 \u00b0C , 1 M sec - BuLi ( 0 . 3 mmol ) was added , and the mixture was stirred for 10 min . At the same temperature , Si - xanthone ( 33 . 5 mg , 0 . 103 mmol ) dissolved in anhydrous THF ( 5 . 0 mL ) was slowly added . The mixture was warmed to room temperature and stirred for 2 . 5 h , then 2 N HCl aq . ( 10 mL ) was added to it . Stirring was continued for 20 min , then the mixture was extracted with CH 2 Cl 2 . The organic layer was washed with brine , dried over Na 2 SO 4 and evaporated to dryness . The residue was dissolved in TFA ( 1 . 5 mL ) . The solution was stirred for 2 h , then evaporated to dryness , and the residue was purified by HPLC to give pure 2 - COOH SiR650 ( 6 . 3 mg , 0 . 012 mol , 11 % yield ) . 1 H NMR ( 300 MHz , CD 3 COCD 3 ) : \u03b4 = 0 . 11 ( s , 3H ) , 0 . 21 ( s , 3H ) , 2 . 61 ( s , 12H ) , 6 . 31 ( dd , J = 9 . 2 Hz , 2 . 6 Hz , 1H ) , 6 . 40 ( d , J = 9 . 5 Hz , 2H ) , 6 . 83 ( d , d , J = 2 . 9 Hz , 2H ) 6 . 87 ( d , J = 7 . 2 Hz , 1H ) , 7 . 21 ( t , J = 7 . 3 Hz , 1H ) , 7 . 32 ( m 1H ) , 7 . 54 ( d , J = 7 . 3 Hz , 1H ) ; HRMS ( ESI + ) : Calcd for [ M ] + 429 . 1998 , Found , 429 . 2006 ( + 0 . 8 mmu ) ; HPLC analysis : eluent : A / B = 80 / 20 to 0 / 100 , 20 min , linear gradient ; solvent A : H 2 O , 0 . 1 % TFA ; solvent B : acetonitrile / H 2 O = 80 / 20 , 0 . 1 % TFA ; flow rate , 1 . 0 mL / min ; detection wavelength 650 nm . S23 Figure S1 . Immunostaining images of KB cells ( a ) , OVCAR - 3 cells ( b ) , and HT1080 cells ( c ) . Cells were fixed with 4 % PFA / PBS for 20 min , labeled with human anti - folate binding protein antibody for 1 hr , washed , and labeled with second antibody conjugated with Alexa Fluor \u24c7 488 for 1 hr . Ex . 488 nm ; Em . 500 - 535 nm . Scale bars : 20 \u03bcm . Figure S2 . Fluorescence images of KB cells ( FR + ) ( a ) and OVCAR - 3 cells ( FR\u2012 ) ( b ) obtained with a commercially available near - infrared fluorescent probe , FolateRSense ( PerkinElmer Inc . ) . White arrows indicate bright dots inside cells . Ex . 650 nm , Em . 670 - 750 nm . Scale bars : 20 \u03bcm . 0 100 KB cells b Fluorescence Bright field a OVCAR - 3 cells Fluorescence Bright field 0 100 c HT1080 cells Fluorescence Bright field a 5 50 0 150 KB cells ( FR + ) OVCAR - 3 cells ( FR \u2013 ) Bright field Fluorescence image Bright field Fluorescence image b S24 Figure S3 . ( a ) Fluorescence images of KB cells ( left ) and OVCAR - 3 cells ( right ) incubated with fluorophore - labeled folates . Emission : 490 nm ( Fluorescein Folate , Alexa488 Folate ) , 555 nm ( TAMRA Folate ) and 595 nm ( 2 - Me DCTM Folate , 2 - COOH DCTM Folate ) . Excitation : 510 - 540 nm ( Fluorescein Folate , Alexa488 Folate ) , 575 - 600 nm ( TAMRA Folate ) and 610 - 650 nm ( 2 - Me DCTM Folate , 2 - COOH DCTM Folate ) . Scale bars : 20 \u03bcm . ( b ) Photophysical properties of fluorophore - labeled folates . a Photophysical properties were measured in 0 . 1 N NaOH aq . with fluorescein in 0 . 1 N NaOH aq . ( \uf046 fl = 0 . 85 ) as a standard . b Photophysical properties were measured in 100 mM sodium phosphate buffer ( pH 7 . 4 ) with fluorescein in 0 . 1 N NaOH aq . ( \uf046 fl = Fluorophore Folate moiety Negatively charged linker Fluorescein Folate Fluorophore = KB cells OVCAR - 3 cells Fluorescence Alexa488 Folate KB cells OVCAR - 3 cells Fluorescence 0 100 5 100 TAMRA Folate KB cells OVCAR - 3 cells Fluorescence 5 100 2 - Me DCTM Folate KB cells OVCAR - 3 cells Fluorescence 5 30 2 - COOH DCTM Folate KB cells OVCAR - 3 cells Fluorescence 5 30 a b Dye Abs max ( nm ) Em max ( nm ) \uf046 fl Fluorescein Folate a Alexa488 Folate b TAMRA Folate c 2 - Me DCTM Folate d 2 - COOH DCTM Folate d 498 523 0 . 25 496 520 0 . 19 558 587 0 . 069 598 611 0 . 12 596 612 0 . 17 S25 0 . 85 ) as a standard . c Photophysical properties were measured in 100 mM sodium phosphate buffer ( pH 7 . 4 ) with Rhodamine B in EtOH ( \uf046 fl = 0 . 65 ) as a standard . d Photophysical properties were measured in 100 mM sodium phosphate buffer ( pH 9 . 0 ) with 2 - Me TokyoMagenta in 100 mM sodium phosphate buffer ( pH 9 . 0 ) ( \uf046 fl = 0 . 42 ) as a standard . Figure S4 . ( a ) KB cells were incubated with FolateSiR - 1 , then fixed with 4 % formaldehyde , and bright - field and fluorescence images were obtained . ( b ) Fluorescence image of mouse embryo incubated with 10 \u03bcM FolateSiR - 1 a 20 \u03bcm 20 \u03bcm 5 50 Fluorescence image Bright field b 50 \u03bcm 50 \u03bcm Closure Caudal Rostral 0 80 Magnified image Fluorescence image c 50 \u03bcm 50 \u03bcm Closure Rostral Caudal 0 80 Magnified image Fluorescence image Strong surface signal in closing neural tube S26 in DMEM containing 10 % rat IC - serum . Then , the mouse embryo was fixed with 4 % formaldehyde . Locations stained with FolateSiR - 1 are indicated by white arrowheads . ( c ) Fluorescence image of mouse embryo incubated with 10 \u03bcM FolateSiR - 1 in the presence of 1 mM folic acid , and then fixed with 4 % formaldehyde . Figure S5 . ( a ) Fluorescence images of mouse embryos incubated with 10 \uf06d M FolateSiR - 1 and fixed with 4 % formaldehyde . The right images are magnifications of the indicated portions of the left images . The regions stained with FolateSiR - 1 are indicated by white arrowheads . The cells indicated by yellow arrows appear to be dead cells . Scale bars : 50 \uf06d m . ( b , c ) Bright - field ( left ) and fluorescence ( right ) images of mouse embryos fixed with 4 % formaldehyde in the absence ( b ) or presence ( c ) of 1 mM folic acid . These data indicate that the autofluorescence of the fixed embryos with or without 1 mM folic acid is sufficiently low compared with those in ( a ) . Scale bars : 50 \uf06d m . S27 Figure S6 . ( a ) White - light and fluorescence images of extracted organs of the mouse in Figure 4a . ( b ) White - light and fluorescence images of extracted organs of the mouse in Figure 4c . a White light Fluorescence Stomach Tumor b White light Fluorescence Kidney Tumor S28 Figure S7 . In vivo fluorescence imaging of HT1080 tumor - bearing mice injected with 100 \uf06d M FolateSiR - 1 in 100 \uf06d L saline ( n = 3 ) . ( a ) Time - lapse white - light ( top ) and fluorescence ( bottom ) images of a mouse . The images were obtained before the probe injection and 0 , 0 . 5 , 1 , 2 , 3 and 6 h after the probe injection . Ex . / Em . = 661 / 700 - 800 nm . T : Tumor ; M : Muscle . Fluorescence intensity scale : gray scale 0 to 255 . ( b ) The time - dependent fluorescence intensity changes of tumor and non - tumor ( muscle ) areas of three mice . Error bar shows S . E . ( c ) White - light ( top ) and fluorescence ( bottom ) images of extracted mouse organs . S29 Figure S8 . In vivo fluorescence imaging of HT1080 tumor - bearing mice injected with 100 \uf06d M FolateSiR - 2 in 100 \uf06d L saline ( n = 3 ) . ( a ) Time - lapse white - light ( top ) and fluorescence ( bottom ) images of a mouse . The images were obtained before the probe injection and 0 , 0 . 5 , 1 , 2 , 3 and 6 h after the probe injection . Ex . / Em . = 661 / 700 - 800 nm . T : Tumor ; M : Muscle . Fluorescence intensity scale : gray scale 0 to 255 . ( b ) The time - dependent fluorescence intensity changes in tumor and non - tumor ( muscle ) areas of three mice . Error bar shows S . E . ( c ) White - light ( top ) and fluorescence ( bottom ) images of extracted mouse organs . S30 Figure S9 . In vivo fluorescence imaging of KB tumor - bearing mice injected with 6 mM folic acid in 100 \uf06d L saline , then with 100 \uf06d M FolateSiR - 1 in 100 \uf06d L saline ( n = 3 ) . ( a ) Time - lapse white - light ( top ) and fluorescence ( bottom ) images of a mouse . The images were obtained before the probe injection and 0 , 0 . 5 , 1 , 2 , 3 and 6 h after the probe injection . Ex . / Em . = 661 / 700 - 800 nm . T : Tumor ; M : Muscle . Fluorescence intensity scale : gray scale 0 to 255 . ( b ) The time - dependent fluorescence intensity changes of tumor and non - tumor ( muscle ) areas of three mice . Error bar shows S . E . ( c ) White - light ( top ) and fluorescence ( bottom ) images of extracted mouse organs . S31 Position Age Sex Organ Pathological Diagnosis Tumor history Tumor size ( cm ) Differentiation A1 50 F Ovary Adenocarcinoma 1 year 4\u00d72\u00d71 . 5 \u2012 A2 A 45 F Ovary Adenocarcinoma 14 days 11\u00d78\u00d76 Poor A3 62 F Ovary Mucous cystadenocarcinoma 20 days 6\u00d74\u00d74 \u2012 A4 69 F Ovary Serous adenocarcinoma 1 month diameter 1 . 5 \u2012 A5 49 F Ovary Serous adenocarcinoma \u2012 \u2012 Poor A6 53 F Ovary Serous adenocarcinoma \u2012 \u2012 Moderate A7 80 F Ovary Serous adenocarcinoma \u2012 \u2012 Poor A8 54 F Ovary Serous cystadenocarcinoma 2 months 6\u00d76\u00d75 Poor B1 53 F Ovary Mucous adenocarcinoma 1 month 5\u00d74\u00d73 \u2012 B2 40 F Ovary Adenocarcinoma , signet ring cell carcinoma 2 months 2 . 5\u00d72 . 5\u00d72 Poor B3 55 F Ovary Clear cell adenocarcinoma 3 months diameter 13 Poor B4 67 F Ovary Serous cystadenocarcinoma 2 months diameter 2 Poor B5 59 F Ovary Serous cystadenocarcinoma 1 month 7\u00d75\u00d72 . 8 Poor B6 74 F Ovary Cystadenocarcinoma , serous & papillary 20 days 5\u00d73 . 5\u00d72 Moderate B7 71 F Ovary Endometrioid carcinoma \u2012 \u2012 Moderate B8 49 F Ovary Endometrioid carcinoma 20 days 10\u00d78\u00d74 . 5 Moderate C1 40 F Ovary Mucinous cystodenocarcinoma \u2012 \u2012 Moderate C2 42 F Ovary Poorly differentiated carcinoma \u2012 \u2012 Poor C3 52 F Ovary Serous carcinoma \u2012 \u2012 Poor C4 54 F Ovary Serous cystoadenocarcinoma \u2012 \u2012 \u2012 C5 60 F Ovary Serous cystoadenocarcinoma , papillary \u2012 \u2012 \u2012 C6 50 F Ovary Serous papillary cystoadenocarcinoma \u2012 \u2012 Moderate C7 42 F Ovary Signet ring cell carcinoma 2 months 6\u00d74\u00d74 \u2012 C8 67 F Ovary Transitional cell carcinoma 2 months 8\u00d76\u00d73 Poor D1 47 F Ovary Transitional epithelial cell Carcinoma 20 days 12 . 5\u00d710\u00d7 3 . 5 Poor D2 20 F Ovary Yolk sac cystadenoma \u2012 \u2012 \u2012 D3 26 F Ovary Leimyoma 2 months 8\u00d76 . 5\u00d74 . 5 \u2012 S32 D4 45 F Ovary Malignancy , Brenner\u2019s tumor 3 years diameter 10 \u2012 D5 21 F Ovary Mature teratoma 6 years 18\u00d716\u00d75 Moderate D6 39 F Ovary Krukenberg\u2019s tumor \u2012 \u2012 \u2012 D7 25 F Ovary Cystic mature teratoma 29 days 8\u00d77\u00d76 \u2012 D8 18 F Ovary Dysgerminoma 1 month 18\u00d712\u00d78 Poor E1 26 F Ovary Dysgerminoma \u2012 \u2012 \u2012 E2 42 F Ovary Endodermal sinus tumor 3 months 15\u00d712\u00d78 Poor E3 60 F Ovary Thecoma 6 months 9\u00d76 . 5\u00d76 \u2012 E4 20 F Ovary Thecoma 1 week 12\u00d78\u00d76 \u2012 E5 32 F Ovary Thecoma 3 months 8\u00d77 \u2012 E6 54 F Ovary Normal E7 40 F Ovary Normal E8 56 F Ovary Normal Figure S10 . Components of the frozen human ovary tumor tissue array ( Catalog No . : T6235183 - 5 ; Lot No . : B705061 ) . Figure S11 . Fluorescence staining of the tissue microarray described in Figure S8 with FolateSiR - 1 ( pink ) and DAPI ( Blue ) . Scale bar represents 2 mm . S33 Figure S12 . Immunostaining of folate receptors in the tissue microarray described in Figure S8 . Scale bar represents 2 mm . Figure S13 . ( a ) The magnified fluorescence images of C4 , B8 , C3 and A5 specimens in Figure S11 are shown . Red color indicates the fluorescence of FolateSiR - 1 . All specimens showed the fluorescence signal , though the A5 A5 A5 A5 B8 B8 C3 C3 C4 C4 a b c Magnified image S34 fluorescence intensities were weak in C3 and A5 specimens . ( b ) The magnified folate receptor - immunostaining images of C4 , B8 , C3 and A5 in Figure S12 are shown . All specimens were immunostained , though the sample condition of B8 , C3 and A5 were not good . ( c ) The magnified fluorescence image ( left ) and the magnified immunostaining image ( right ) of C4 specimen in ( a ) and ( b ) are shown . Arrows indicate the boundary between tumor tissues and non - tumor tissues ( vascular tissues and fibrous tissues ) . The tumor tissues are selectively stained by the fluorescent probe , FolateSiR - 1 . Figure S14 . Absorption spectra of 1 \u03bc M 2 - COOH SiR650 in 100 mM sodium phosphate buffer at pH 7 . 4 with or without 10 % fetal bovine serum ( FBS ) . Figure S15 . Absorbance spectra of 1 \u03bc M FolateSiR - 1 in phosphate buffered saline ( PBS ) containing 0 . 1 % DMSO with or without 10 % FBS ( Biowest , Cat . No . : S1810 - 500 ) , or in mouse serum ( FUJIFILM Wako Pure Chemical Corp . , Cat . No . : 146 - 06551 ) . - 0 . 02 0 0 . 02 0 . 04 0 . 06 0 . 08 0 . 1 0 . 12 500 550 600 650 700 PBS PBS + 10 % FBS mouse serum A b s o r ban c e Wavelength ( nm ) S35 Supporting References SR1 ) Yamaguchi , Y . , Shinotsuka , N . , Nonomura , K . , Takemoto , K . , Kuida , K . , Yosida , H . , Miura , M . Live imaging of apoptosis in a novel transgenic mouse highlights its role in neural tube closure . J . Cell Biol . 195 , 1047 - 1060 ( 2011 ) . SR 2 ) Koide , Y . , Urano , Y . , Hanaoka , K . , Terai , T . & Nagano , T . Evolution of group 14 rhodamines as platforms for near - infrared fluorescence probes utilizing photoinduced electron transfer . ACS Chem . Biol . 6 , 600 - 608 ( 2011 ) . SR3 ) Butkevich , A . N . et al . Fluorescent rhodamines and fluorogenic carbopyronines for super - resolution STED microscopy in living cells . Angew . Chem . Int . Ed . 55 , 3290 - 3294 ( 2016 ) . SR4 ) Egawa , T . et al . Red fluorescent probe for monitoring the dynamics of cytoplasmic calcium ions . Angew . Chem . Int . Ed . 52 , 3874 - 3877 ( 2013 ) . SR5 ) Hirabayashi , K . et al . Synthesis of practical red fluorescent probe for cytoplasmic calcium ions with greatly improved cell - membrane permeability . Data in Brief 12 , 351 - 357 ( 2017 ) ."
}